Antenatal maternal anaemia and infant brain structure: high-field (3 T) and ultra-low-field (64 mT) MRI findings from South Africa.

Code ↔ Paper

9 matches between paragraphs of the paper and lines of its authors' code, computed by the harvester (lexical-v1). Click a colored paragraph or line to see its counterpart.

The 9 matches

[1] § Materials and methods › Measures › Neuroimaging outcomes › Neuroimaging processing ↔ app/main.sh, lines 1–40 · score 0.99 · callosal parcellations, ANTs Atropos, supratentorial tissue, Subcortical grey matter, template space, native space
[2] § Materials and methods › Statistical analysis › Sample characteristics ↔ Khula_AnaemiaImagingAnalysis.qmd, lines 5522–5562 · score 0.81 · duplicated demographic, chi squared, avoid skewing, clinical observations, multiple scans acquired, categorical variables
[3] § Materials and methods › Statistical analysis › Neuroimaging regions of interest ↔ Khula_AnaemiaImagingAnalysis.qmd, lines 2253–2296 · score 0.75 · mid anterior, mid posterior, basal ganglia, corpus callosum, brain volume, caudate nucleus
[4] § Results › Primary analysis: antenatal maternal anaemia | HF and ULF MRI › Modelling: antenatal maternal anaemia status ↔ Khula_AnaemiaImagingAnalysis.qmd, lines 2253–2296 · score 0.72 · linear growth trajectories, logarithmic transformation, linear relationship, absolute age, corpus callosum, LME models
[5] § Materials and methods › Statistical analysis › Secondary analysis: postnatal child anaemia status ↔ Khula_AnaemiaImagingAnalysis.qmd, lines 6863–6884 · score 0.69 · postnatal exposures, regional child brain, postnatal child anaemia, volumes remained, regional brain volumes, predictor
[6] § Materials and methods › Measures › Anaemia status ↔ Khula_AnaemiaImagingAnalysis.qmd, lines 5382–5416 · score 0.68 · Minimum child haemoglobin, corresponding age, 3–24, multiple scans, diagnosis, child anaemia status
[7] § Materials and methods › Statistical analysis › Primary analysis: antenatal maternal anaemia status › Statistical modelling ↔ Khula_AnaemiaImagingAnalysis.qmd, lines 338–363 · score 0.66 · absolute regional brain, linear mixed, individual infant, volume trajectories, ULF subsamples, fitted
[8] § Materials and methods › Measures › Contextual measures ↔ Khula_AnaemiaImagingAnalysis.qmd, lines 76–171 · score 0.63 · maternal education, maternal employment, household income, maternal HIV, variables, sex
[9] § Results › Primary analysis: antenatal maternal anaemia | HF and ULF MRI › Exploratory analyses: antenatal maternal anaemia status ↔ Khula_AnaemiaImagingAnalysis.qmd, lines 4870–4903 · score 0.61 · basal ganglia, rapid growth, growth trajectory, corpus callosum, LOESS curves, brain volumes

Paper

Loaded from Europe PMC by your browser, not stored by OSCR: doi.org · Europe PMC

The paper is loaded when this pane is shown.

The paper is loaded by your browser, with JavaScript: read it at doi.org or on Europe PMC.

The authors' code

Quarto · 6,955 lines · 196 KB · no license · 8 matches

---
title: "Khula_AnaemiaImagingAnalysis"
format: pdf
editor: visual
---
------------------------------------------------------------------------
------------------------------------------------------------------------
\*This is the final script used in the analyses for the following research: "**Antenatal Maternal Anaemia and Infant Brain Structure: 3 T and 64 mT MRI Findings from South Africa**"
*Details of the datasets:*
- Both high-field (HF; 3 T) and ultra-low-field (ULF; 64 mT) datasets are in long format with some infants having multiple scans across study timepoints (3mo, 6mo, 12mo, 18mo, 24mo).
- These datasets represent the final samples after the exclusion of 1) scans that failed quality check (QC), 2) infants outside of the age range (\>25.5 months or \<2.5 months), and 3) statistical outliers.
- HF: All acquired scans (*n* = 264) –\> passed QC (*n* = 247) –\> within the defined age range (*n* = 239) –\>removal of outliers (*n* = 232)
- ULF: All acquired scans (*n* = 553) –\> passed QC (*n* = 456) —\> within the defined age range (*n* = 439) –\> removal out statistical outliers (*n* = 418)
- The script for the HF analyses is presented first, followed by the script for the ULF analyses.
#### --------------------------------------------------------------------------
**Install Packages**
```{r}
# install packages if needed
# install.packages('pacman')
library(pacman)
p_load(readr, readxl, tidyverse, magrittr, writexl, stringr, dplyr)
```
------------------------------------------------------------------------
[***High-Field (HF) Analyses***]{.underline}
------------------------------------------------------------------------
**Import HF Dataset**
```{r}
# import dataset
HF_ida_final <- read_excel("HF_ida_final.xlsx")
```
**Create New Variables for Total Brain Volumes Across Regions of Interest (ROIs)**
```{r}
# compute total brain volumes for ROIs
# basal ganglia
HF_ida_final$total_putamen <- HF_ida_final$hf_mm_left_putamen + HF_ida_final$hf_mm_right_putamen
HF_ida_final$total_caudate <- HF_ida_final$hf_mm_left_caudate + HF_ida_final$hf_mm_right_caudate
# corpus callosum
HF_ida_final$total_cc <- HF_ida_final$hf_mm_posterior_callosum + HF_ida_final$hf_mm_mid_posterior_callosum + HF_ida_final$hf_mm_central_callosum + HF_ida_final$hf_mm_mid_anterior_callosum + HF_ida_final$hf_mm_anterior_callosum
```
**Create New Age Variable (Days to Months):**
```{r}
# convert child age at scan from days to months
# average number of days in a month (accounting for leap years): 30.44
HF_ida_final <- HF_ida_final %>%
mutate(scan_age_months = hf_age / 30.44)
```
**Creating Meaningful Categorical Variables**
```{r}
# recode categorical variables from numeric to factor
# maternal anaemia status
HF_ida_final$maternal_preg_anaemia_status <- factor(
HF_ida_final$maternal_preg_anaemia_status,
levels = c(0, 1),
labels = c("no_mat_anaemia", "mat_anaemia")
)
# maternal ID status
HF_ida_final$maternal_preg_adj_brinda_id <- factor(
HF_ida_final$maternal_preg_adj_brinda_id,
levels = c(0, 1),
labels = c("no_mat_id", "mat_id")
)
# maternal ID status
HF_ida_final$maternal_preg_unadjusted_id <- factor(
HF_ida_final$maternal_preg_unadjusted_id,
levels = c(0, 1),
labels = c("no_mat_id", "mat_id")
)
# maternal trimester Hb
HF_ida_final$maternal_trimester_hb <- factor(
HF_ida_final$maternal_trimester_hb,
levels = c(1, 2, 3),
labels = c("first", "second", "third")
)
# child sex
HF_ida_final$child_sex <- factor(
HF_ida_final$child_sex,
levels = c(0, 1),
labels = c("girl", "boy")
)
# maternal pae
HF_ida_final$maternal_pae <- factor(
HF_ida_final$maternal_pae,
levels = c(0, 1),
labels = c("no_pae", "pae")
)
#maternal pte
HF_ida_final$maternal_pte <- factor(
HF_ida_final$maternal_pte,
levels = c(0, 1),
labels = c("no_pte", "pte")
)
# maternal hiv
HF_ida_final$maternal_hiv <- factor(
HF_ida_final$maternal_hiv,
levels = c(0, 1),
labels = c("no_hiv", "hiv")
)
# maternal dep
HF_ida_final$maternal_dep <- factor(
HF_ida_final$maternal_dep,
levels = c(0, 1),
labels = c("no_dep", "dep")
)
# maternal employment
HF_ida_final$mom_employment_en_dichotomous <- factor(
HF_ida_final$mom_employment_en_dichotomous,
levels = c(0, 1),
labels = c("unemployed", "employed")
)
# household income
HF_ida_final$income_household_en <- factor(
HF_ida_final$income_household_en,
levels = c(1, 2, 3, 4),
labels = c("Less than R1000 per month", "R1000-R5000 per monthy", " R5000-R10 000 per month", "More than R10 000 per month" )
)
#maternal education
HF_ida_final$mom_edu_en <- factor(
HF_ida_final$mom_edu_en,
levels = c(2, 3, 4, 5, 6),
labels = c("primary", "some_secondary", "complete_secondary", "some_tertiary", "completed_tertiary" )
)
# baby hiv
HF_ida_final$baby_hiv <- factor(
HF_ida_final$baby_hiv,
levels = c(0, 1),
labels = c("no_child_hiv", "child_hiv")
)
```
**Sample Distribution: Availability of Data Across Study Timepoints**
```{r}
# sample over 12 months with maternal anaemia data
sample_over_12M <- HF_ida_final %>%
filter(scan_age_months > 12,
!is.na(maternal_preg_anaemia_status))
# sample under 12 months with maternal anaemia data
sample_under_12M <- HF_ida_final %>%
filter(scan_age_months < 12,
!is.na(maternal_preg_anaemia_status))
# sample over 18 months with maternal anaemia data
sample_over_18 <- HF_ida_final %>%
filter(scan_age_months > 18,
!is.na(maternal_preg_anaemia_status))
# sample over 24 months with maternal anaemia data
sample_over_24 <- HF_ida_final %>%
filter(scan_age_months > 24,
!is.na(maternal_preg_anaemia_status))
```
```{r}
# HF sample size
# unique observations in HF dataset
# excluding repeated measures
library(dplyr)
HF_ida_final %>%
summarise(n_unique_subjects = n_distinct(subject_id))
```
------------------------------------------------------------------------
**Primary Analysis: Antenatal Maternal Anaemia Status (HF Sample)**
------------------------------------------------------------------------
**Sample Size: For infants with Antenatal Maternal Anemia Data**
Note: Sample sizes were calculated with 1) the inclusion of repeated measures (multiple scans across study timepoints), and for 2) unique subject IDs excluding repeated measures.
- The sample of unique participants was used to report on sample characteristics
- The full sample with repeated measures was used for analyses including linear mixed effects (LME) models to include children with multiple scans from different study timepoints.
```{r}
# total sample size
# repeated measures included
# sample with antenatal maternal anaemia data
HF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status)) %>% # only include PIDs with antenatal maternal anaemia data
summarise(n = n())
```
```{r}
# unique sample size
# repeated measures excluded
# number of unique participants in sample
library(dplyr)
HF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status)) %>% # only include subjects with anaemia data
distinct(subject_id, .keep_all = TRUE) %>% # keep unique subject IDs
summarise(n_total = n()) %>% # count unique subjects
print()
```
```{r}
# prevalence of antenatal maternal anaemia
# excluding repeated measures
library(dplyr)
HF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status)) %>% # only include PIDs with antenatal maternal anaemia data
distinct(subject_id, .keep_all = TRUE) %>% # keep only one record per mother
group_by(maternal_preg_anaemia_status) %>%
summarise(n = n()) %>%
mutate(percent = round(100 * n / sum(n), 2))
```
```{r}
# antenatal maternal anaemia severity
# unique subject IDs: Based on true prevalence
# excluding repeated measures
library(dplyr)
HF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_severity)) %>% # exclude missing severity
group_by(subject_id) %>% # group by mother
summarise(
maternal_preg_anaemia_severity = first(
maternal_preg_anaemia_severity[maternal_preg_anaemia_severity != ""]
)
) %>%
group_by(maternal_preg_anaemia_severity) %>% # count unique mothers by severity
summarise(n = n()) %>%
mutate(percent = round(100 * n / sum(n), 2))
```
**Repeated Measures Data Availability**
```{r}
# repeated measures count
# unique subject IDs
library(dplyr)
HF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status)) %>% # keep only rows with anaemia status
count(subject_id) %>% # count rows per subject
count(n) %>% # count how many subjects had n rows
rename(times_measured = n, number_of_subjects = nn) %>%
mutate(percent = round(100 * number_of_subjects / sum(number_of_subjects), 1)) # add % column
```
```{r}
# HF Figure 2
# plot of data availability with repeated measures
library(ggplot2)
library(dplyr)
Figure_2_HF<- HF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status)) %>%
mutate(
timepoint = factor(timepoint, levels = c("3M", "6M", "12M", "18M", "24M")),
subject_id = factor(subject_id),
anaemia_group = as.factor(maternal_preg_anaemia_status)
) %>%
ggplot(aes(x = timepoint, y = subject_id, group = subject_id, color = anaemia_group)) +
geom_line(alpha = 0.5, linewidth = 0.5) + # updated for ggplot2 ≥3.4.0
geom_point(size = 2) +
scale_color_manual(
values = c("red", "blue"),
labels = c("Maternal Anaemia", "No Maternal Anaemia")
) +
scale_x_discrete(expand = expansion(mult = 0.05)) +
scale_y_discrete(expand = expansion(mult = 0.02)) +
coord_cartesian(clip = "off") + # ensures no points are clipped at the edges
labs(
title = "Participant Data Availability by Timepoint",
x = "Timepoint",
y = "Participants",
color = "Maternal Anaemia"
) +
theme_minimal(base_size = 13) +
theme(
axis.text.y = element_blank(),
axis.ticks.y = element_blank(),
panel.grid.major.y = element_blank()
)
print(Figure_2_HF)
```
**Sample Characteristics and Group Differences: Antenatal Maternal Anaemia Status**
Sample characteristics were assessed based on the number of unique subjects in each subsample, excluding repeated measures data for children with multiple scans. This approach was used to avoid skewing results with duplicated demographic and clinical observations for the same infants. The only exception to this approach was child age at scan which was considered for the full repeated measures subsample, with a proportion of children having multiple scans acquired at different ages.
However, given the repeated measures study design with individual infant data across multiple study visits in both HF and ULF subsamples, linear mixed effects (LME) models were fitted to account for individual differences in assessing the association between antenatal maternal anaemia status and absolute regional brain volume trajectories
- **Categorical Variables**
*Trimester of Maternal Hb Measurement*
```{r}
# trimester Hb was collected from mothers
# excluding repeated measures
# unique subject IDs: Based on true prevalence
library(dplyr)
HF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status)) %>% # exclude missing anaemia
group_by(subject_id) %>% # keep one row per mother
summarise(maternal_trimester_hb = first(maternal_trimester_hb)) %>%
group_by(maternal_trimester_hb) %>%
summarise(n = n()) %>%
mutate(percent = round(100 * n / sum(n), 2))
```
```{r}
# group differences in trimester of Hb collection
# chi-squared: trimester Hb
# unique subject IDs: based on true prevalence
library(dplyr)
# Filter unique subjects with non-missing anaemia status
unique_data <- HF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status)) %>%
distinct(subject_id, .keep_all = TRUE)
# Create contingency table: Trimester by Maternal Anaemia Status
tbl <- table(unique_data$maternal_trimester_hb, unique_data$maternal_preg_anaemia_status)
# Print contingency table
cat("Contingency Table: Trimester by Maternal Anaemia Status\n")
print(format(tbl, nsmall = 0))
# Print column percentages
cat("\nColumn Percentages (% of each anaemia status group):\n")
print(round(100 * prop.table(tbl, margin = 2), 2))
# Run appropriate test based on expected cell counts
expected <- chisq.test(tbl)$expected
if (any(expected < 5)) {
cat("\nFisher's Exact Test (due to small expected cell counts):\n")
print(fisher.test(tbl))
} else {
cat("\nChi-squared Test:\n")
print(chisq.test(tbl))
}
```
*Child Sex at Birth*
```{r}
# child sex at birth (full sample)
# unique subject IDs: based on true prevalence
library(dplyr)
HF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status)) %>% # only include PIDs with antenatal maternal anaemia data
distinct(subject_id, .keep_all = TRUE) %>% # keep only one record per mother/child
group_by(child_sex) %>%
summarise(n = n()) %>%
mutate(percent = round(100 * n / sum(n), 2))
```
```{r}
# group differences in child sex by antenatal maternal anaemia status
# chi-squared: child sex
# unique subject IDs: based on true prevalence
library(dplyr)
# keep only unique subjects and non-missing anaemia status
unique_data <- HF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status)) %>%
distinct(subject_id, .keep_all = TRUE)
# create contingency table
tbl <- table(
unique_data$child_sex,
unique_data$maternal_preg_anaemia_status
)
# print contingency table
cat("Contingency Table: Child Sex by Maternal Anaemia Status\n")
print(format(round(tbl, 2), nsmall = 2))
cat("\nColumn Percentages (% of each anaemia status group):\n")
print(round(100 * prop.table(tbl, margin = 2), 2))
# run appropriate test
expected <- chisq.test(tbl)$expected
if (any(expected < 5)) {
cat("\nFisher's Exact Test (due to small expected cell counts):\n")
print(fisher.test(tbl))
} else {
cat("\nChi-squared Test:\n")
print(chisq.test(tbl))
}
```
*Prenatal Alcohol Exposure (PAE)*
```{r}
# pae prevalence (full sample)
# unique subject IDs: based on true prevalence
HF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status)) %>% # only include subjects with anaemia data
distinct(subject_id, .keep_all = TRUE) %>% # keep only unique subjects
group_by(maternal_pae) %>%
summarise(n = n()) %>%
mutate(percent = round(100 * n / sum(n), 2))
```
```{r}
# group differences in pae by antenatal maternal anaemia status
# chi-squared: pae
# unique subject IDs: based on true prevalence
# filter unique subjects and exclude missing anaemia status
unique_data <- HF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status)) %>%
distinct(subject_id, .keep_all = TRUE)
# create contingency table
tbl <- table(
unique_data$maternal_pae,
unique_data$maternal_preg_anaemia_status
)
# print contingency table
cat("Contingency Table: Maternal PAE by Maternal Anaemia Status\n")
print(tbl)
cat("\nColumn Percentages (% of each anaemia status group):\n")
print(round(100 * prop.table(tbl, margin = 2), 2))
# run appropriate test
expected <- chisq.test(tbl)$expected
if (any(expected < 5)) {
cat("\nFisher's Exact Test (due to small expected cell counts):\n")
print(fisher.test(tbl))
} else {
cat("\nChi-squared Test:\n")
print(chisq.test(tbl))
}
```
*Prenatal Tobacco Exposure (PTE)*
```{r}
# pte prevalence (full sample)
# unique subject IDs: based on true prevalence
unique_data <- HF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status)) %>%
distinct(subject_id, .keep_all = TRUE)
# summarise maternal pte
unique_data %>%
group_by(maternal_pte) %>%
summarise(n = n()) %>%
mutate(percent = round(100 * n / sum(n), 2))
```
```{r}
# group differences in pte by antenatal maternal anaemia status
# chi-squared: pte
# unique subject IDs: based on true prevalence
# filter unique subjects and exclude missing anaemia status
unique_data <- HF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status)) %>%
distinct(subject_id, .keep_all = TRUE)
# create contingency table
tbl <- table(
unique_data$maternal_pte,
unique_data$maternal_preg_anaemia_status
)
# print contingency table
cat("Contingency Table: Maternal PTE by Maternal Anaemia Status (Unique Subjects)\n")
print(tbl)
cat("\nColumn Percentages (% of each anaemia status group):\n")
print(round(100 * prop.table(tbl, margin = 2), 2))
# run appropriate test based on expected counts
expected <- chisq.test(tbl)$expected
if (any(expected < 5)) {
cat("\nFisher's Exact Test (due to small expected cell counts):\n")
print(fisher.test(tbl))
} else {
cat("\nChi-squared Test:\n")
print(chisq.test(tbl))
}
```
*Maternal HIV Infection*
```{r}
# materal hiv prevalence (full sample)
# unique subject IDs: based on true prevalence
HF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status)) %>%
distinct(subject_id, .keep_all = TRUE) %>%
group_by(maternal_hiv) %>%
summarise(n = n()) %>%
mutate(percent = round(100 * n / sum(n), 2))
```
```{r}
# group differences in maternal hiv by antenatal maternal anaemia status
# chi-squared: hiv
# unique subject IDs: based on true prevalence
unique_data <- HF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status)) %>%
distinct(subject_id, .keep_all = TRUE)
# create contingency table
tbl <- table(
unique_data$maternal_hiv,
unique_data$maternal_preg_anaemia_status
)
# print contingency table
cat("Contingency Table: Maternal HIV by Maternal Anaemia Status\n")
print(tbl)
cat("\nColumn Percentages (% of each anaemia status group):\n")
print(round(100 * prop.table(tbl, margin = 2), 2))
# run appropriate test
expected <- chisq.test(tbl)$expected
if (any(expected < 5)) {
cat("\nFisher's Exact Test (due to small expected cell counts):\n")
print(fisher.test(tbl))
} else {
cat("\nChi-squared Test:\n")
print(chisq.test(tbl))
}
```
*Maternal Depression*
```{r}
# maternal depression prevalence (full sample)
# unique subject IDs: based on true prevalence
HF_ida_final %>%
filter(!is.na(maternal_dep)) %>%
distinct(subject_id, .keep_all = TRUE) %>%
group_by(maternal_dep) %>%
summarise(n = n()) %>%
mutate(percent = round(100 * n / sum(n), 2))
```
```{r}
# group differenced by maternal depression status
# chi-squared: maternal depression
# unique subject IDs: based on true prevalence
unique_data <- HF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status)) %>%
distinct(subject_id, .keep_all = TRUE)
# create contingency table
tbl <- table(
unique_data$maternal_dep,
unique_data$maternal_preg_anaemia_status
)
# print contingency table
cat("Contingency Table: Maternal DEP by Maternal Anaemia Status\n")
print(tbl)
cat("\nColumn Percentages (% of each anaemia status group):\n")
print(round(100 * prop.table(tbl, margin = 2), 2))
# run appropriate test
expected <- chisq.test(tbl)$expected
if (any(expected < 5)) {
cat("\nFisher's Exact Test (due to small expected cell counts):\n")
print(fisher.test(tbl))
} else {
cat("\nChi-squared Test:\n")
print(chisq.test(tbl))
}
```
*Maternal Employment*
```{r}
# maternal employment (full sample)
# unique subject IDs: based on true prevalence
HF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status)) %>% # only include PIDs with antenatal maternal anaemia data
distinct(subject_id, .keep_all = TRUE) %>% # keep unique subjects
group_by(mom_employment_en_dichotomous) %>%
summarise(n = n()) %>%
mutate(percent = round(100 * n / sum(n), 2))
```
```{r}
# group differences in maternal employment by antenatal maternal anaemia status
# chi-squared: maternal employment
# unique subject IDs: based on true prevalence
unique_data <- HF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status)) %>%
distinct(subject_id, .keep_all = TRUE)
# create contingency table
tbl <- table(
unique_data$mom_employment_en_dichotomous,
unique_data$maternal_preg_anaemia_status
)
# print contingency table
cat("Contingency Table: Maternal Employment by Maternal Anaemia Status\n")
print(tbl)
cat("\nColumn Percentages (% of each anaemia status group):\n")
print(round(100 * prop.table(tbl, margin = 2), 2))
# run appropriate test
expected <- chisq.test(tbl)$expected
if (any(expected < 5)) {
cat("\nFisher's Exact Test (due to small expected cell counts):\n")
print(fisher.test(tbl))
} else {
cat("\nChi-squared Test:\n")
print(chisq.test(tbl))
}
```
*Maternal Education*
```{r}
# maternal education (full sample)
# unique subject IDs: based on true prevalence
unique_data <- HF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status)) %>%
distinct(subject_id, .keep_all = TRUE)
# summarise maternal education
unique_data %>%
group_by(mom_edu_en) %>%
summarise(n = n()) %>%
mutate(percent = round(100 * n / sum(n), 2))
```
```{r}
# group differences in maternal education by antenatal maternal anaemia status
# chi-squared: maternal education
# unique subject IDs: based on true prevalence
unique_data <- HF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status)) %>%
distinct(subject_id, .keep_all = TRUE)
# create contingency table
tbl <- table(
unique_data$mom_edu_en,
unique_data$maternal_preg_anaemia_status
)
# print contingency table
cat("Contingency Table: Maternal Education by Maternal Anaemia Status\n")
print(tbl)
cat("\nColumn Percentages (% of each anaemia status group):\n")
print(round(100 * prop.table(tbl, margin = 2), 2))
# run appropriate test
expected <- chisq.test(tbl)$expected
if (any(expected < 5)) {
cat("\nFisher's Exact Test (due to small expected cell counts):\n")
print(fisher.test(tbl))
} else {
cat("\nChi-squared Test:\n")
print(chisq.test(tbl))
}
```
*Household Income*
```{r}
# household income (full sample)
# unique subject IDs: based on true prevalence
unique_data <- HF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status)) %>%
distinct(subject_id, .keep_all = TRUE)
# summarise household income
unique_data %>%
group_by(income_household_en) %>%
summarise(n = n()) %>%
mutate(percent = round(100 * n / sum(n), 2))
```
```{r}
# group differences in household income by antenatal maternal anaemia status
# chi-squared: household income
# unique subject IDs: based on true prevalence
unique_data <- HF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status)) %>%
distinct(subject_id, .keep_all = TRUE)
# create contingency table
tbl <- table(
unique_data$income_household_en,
unique_data$maternal_preg_anaemia_status
)
# print contingency table
cat("Contingency Table: Household Income by Maternal Anaemia Status\n")
print(tbl)
cat("\nColumn Percentages (% of each anaemia status group):\n")
print(round(100 * prop.table(tbl, margin = 2), 2))
# run appropriate test
expected <- chisq.test(tbl)$expected
if (any(expected < 5)) {
cat("\nFisher's Exact Test (due to small expected cell counts):\n")
print(fisher.test(tbl))
} else {
cat("\nChi-squared Test:\n")
print(chisq.test(tbl))
}
```
*Infant HIV Infection*
```{r}
# infant hiv prevalence (full sample)
# unique subject IDs: based on true prevalence
unique_data <- HF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status)) %>%
distinct(subject_id, .keep_all = TRUE)
# summarise counts and percentages
unique_data %>%
group_by(baby_hiv) %>%
summarise(n = n()) %>%
mutate(percent = round(100 * n / sum(n), 2))
```
```{r}
# group differences in infant hiv by antenatal maternal anaemia status
# chi-squared infant hiv
# unique subject IDs: based on true prevalence
unique_data <- HF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status)) %>%
distinct(subject_id, .keep_all = TRUE)
# create contingency table
tbl <- table(
unique_data$baby_hiv,
unique_data$maternal_preg_anaemia_status
)
# print contingency table
cat("Contingency Table: Baby HIV by Maternal Anaemia Status\n")
print(tbl)
cat("\nColumn Percentages (% of each anaemia status group):\n")
print(round(100 * prop.table(tbl, margin = 2), 2))
# run appropriate test
expected <- chisq.test(tbl)$expected
if (any(expected < 5)) {
cat("\nFisher's Exact Test (due to small expected cell counts):\n")
print(fisher.test(tbl))
} else {
cat("\nChi-squared Test:\n")
print(chisq.test(tbl))
}
```
- **Continuous Variables**
```{r}
#install packages if needed
#install.packages('car')
library("car")
```
*Maternal Age at Enrolment*
```{r}
# maternal age at enrolment (full sample)
# unique subject IDs: based on true prevalence
library(dplyr)
HF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status),
!is.na(mom_age_en)) %>% # exclude missing anaemia or age
group_by(subject_id) %>% # keep unique mothers only
summarise(
mom_age_en = first(mom_age_en) # assume age is constant per mother
) %>%
summarise(
mean_mom_age = mean(mom_age_en, na.rm = TRUE),
sd_mom_age = sd(mom_age_en, na.rm = TRUE),
min_mom_age = min(mom_age_en, na.rm = TRUE),
max_mom_age = max(mom_age_en, na.rm = TRUE),
n = n()
)
```
```{r}
# group differences in maternal age at enrolment by antenatal maternal anaemia status
# ttest: maternal age at enrolment
# unique subject IDs: based on true prevalence
library("dplyr")
library("car")
# keep only unique subjects with non-missing anaemia status
unique_data <- HF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status)) %>%
distinct(subject_id, .keep_all = TRUE)
# summary stats with exactly 2 decimals
summary_stats <- unique_data %>%
group_by(maternal_preg_anaemia_status) %>%
summarise(
n = n(),
mean_age = format(round(mean(mom_age_en, na.rm = TRUE), 2), nsmall = 2),
sd_age = format(round(sd(mom_age_en, na.rm = TRUE), 2), nsmall = 2),
min_age = format(round(min(mom_age_en, na.rm = TRUE), 2), nsmall = 2),
max_age = format(round(max(mom_age_en, na.rm = TRUE), 2), nsmall = 2),
.groups = "drop"
)
# print summary stats table
print(summary_stats)
# levene's Test for equality of variances
levene_result <- leveneTest(
mom_age_en ~ maternal_preg_anaemia_status,
data = unique_data
)
print(levene_result)
# extract p-value from Levene’s Test
levene_p <- levene_result$`Pr(>F)`[1]
# automatically select Student's or Welch's t-test
if (levene_p > 0.05) {
cat("\nLevene’s test not significant (p =", round(levene_p, 4),
") → Using Student's t-test (equal variances assumed)\n\n")
t_result <- t.test(
mom_age_en ~ maternal_preg_anaemia_status,
data = unique_data,
var.equal = TRUE
)
} else {
cat("\nLevene’s test significant (p =", round(levene_p, 4),
") → Using Welch's t-test (equal variances NOT assumed)\n\n")
t_result <- t.test(
mom_age_en ~ maternal_preg_anaemia_status,
data = unique_data,
var.equal = FALSE
)
}
# round t-test output (except p-value)
t_result$estimate <- round(t_result$estimate, 2)
t_result$statistic <- round(t_result$statistic, 2)
t_result$conf.int <- round(t_result$conf.int, 2)
# keep p-value as is, or round to 4 decimals for clean display
t_result$p.value <- round(t_result$p.value, 4)
# print t-test result
print(t_result)
```
*Child Age at Scan*
Include repeated measures to account for different ages for scans acquired across study timepoints from the same infant
```{r}
# child age at scan for full group
# include repeated measures: different ages at every scan acquired (full sample)
HF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status)) %>% # only include PIDs with antenatal maternal anaemia data
summarise(
mean_child_age = mean(scan_age_months, na.rm = TRUE),
sd_child_age = sd(scan_age_months, na.rm = TRUE),
min_child_age = min(scan_age_months, na.rm = TRUE),
max_child_age = max(scan_age_months, na.rm = TRUE),
n = n()
)
```
```{r}
# child age at scan
# full sample including repeated measures
# t-test: child age at scan
library(dplyr)
library(car)
# filter once and store
analysis_data <- HF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status))
# summary statistics with exactly 2 decimals displayed
summary_stats <- analysis_data %>%
group_by(maternal_preg_anaemia_status) %>%
summarise(
n = n(),
mean_age = format(round(mean(scan_age_months, na.rm = TRUE), 2), nsmall = 2),
sd_age = format(round(sd(scan_age_months, na.rm = TRUE), 2), nsmall = 2),
min_age = format(round(min(scan_age_months, na.rm = TRUE), 2), nsmall = 2),
max_age = format(round(max(scan_age_months, na.rm = TRUE), 2), nsmall = 2),
.groups = "drop"
)
# print summary stats
print(summary_stats)
# levene's Test for equality of variances
levene_result <- leveneTest(
scan_age_months ~ maternal_preg_anaemia_status,
data = analysis_data
)
print(levene_result)
# extract Levene p-value (rounded just for reporting)
levene_p <- round(levene_result$`Pr(>F)`[1], 4)
cat("\nLevene’s test p-value =", levene_p, "\n")
# automatically choose correct t-test based on Levene
if (levene_p > 0.05) {
cat("Levene’s test not significant → Using Student's t-test (equal variances assumed)\n\n")
t_result <- t.test(
scan_age_months ~ maternal_preg_anaemia_status,
data = analysis_data,
var.equal = TRUE
)
} else {
cat("Levene’s test significant → Using Welch's t-test (equal variances NOT assumed)\n\n")
t_result <- t.test(
scan_age_months ~ maternal_preg_anaemia_status,
data = analysis_data,
var.equal = FALSE
)
}
# print t-test result in full precision
print(t_result)
```
*Child Gestational Age (GA) at Birth*
```{r}
# child ga at birth for full group
# unique subject IDs: based on true prevalence
library(dplyr)
# filter for unique subjects with non-missing maternal anaemia status
unique_data <- HF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status)) %>%
distinct(subject_id, .keep_all = TRUE)
# summary statistics for the full sample
unique_data %>%
summarise(
mean_child_ga = mean(ga_weeks, na.rm = TRUE),
sd_child_ga = sd(ga_weeks, na.rm = TRUE),
min_child_ga = min(ga_weeks, na.rm = TRUE),
max_child_ga = max(ga_weeks, na.rm = TRUE),
n = n()
)
```
```{r}
# group differences in child ga at birth
# ttest: child ga
# unique subject IDs: based on true prevalence
library(dplyr)
library(car)
# keep only unique subjects with non-missing anaemia status
unique_data <- HF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status)) %>%
distinct(subject_id, .keep_all = TRUE)
# summary statistics with exactly 2 decimals displayed
summary_stats <- unique_data %>%
group_by(maternal_preg_anaemia_status) %>%
summarise(
n = n(),
mean_ga = format(round(mean(ga_weeks, na.rm = TRUE), 2), nsmall = 2),
sd_ga = format(round(sd(ga_weeks, na.rm = TRUE), 2), nsmall = 2),
min_ga = format(round(min(ga_weeks, na.rm = TRUE), 2), nsmall = 2),
max_ga = format(round(max(ga_weeks, na.rm = TRUE), 2), nsmall = 2),
.groups = "drop"
)
# print summary stats
print(summary_stats)
# levene's test
levene_result <- leveneTest(
ga_weeks ~ maternal_preg_anaemia_status,
data = unique_data
)
# round levene's p-value just for display
levene_p <- round(levene_result$`Pr(>F)`[1], 4)
cat("\nLevene’s test p-value =", levene_p, "\n")
# automatically choose correct t-test
if (levene_p > 0.05) {
cat("Levene’s test not significant → Using Student's t-test (equal variances assumed)\n\n")
t_result <- t.test(
ga_weeks ~ maternal_preg_anaemia_status,
data = unique_data,
var.equal = TRUE
)
} else {
cat("Levene’s test significant → Using Welch's t-test (equal variances NOT assumed)\n\n")
t_result <- t.test(
ga_weeks ~ maternal_preg_anaemia_status,
data = unique_data,
var.equal = FALSE
)
}
# print t-test result in full precision
print(t_result)
```
*Child Birth Weight (BW)*
```{r}
# child bw (full sample)
# unique subject IDs: based on true prevalence
library(dplyr)
# filter for unique subjects with non-missing maternal anaemia status
unique_data <- HF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status)) %>%
distinct(subject_id, .keep_all = TRUE)
# summary statistics for the full sample
unique_data %>%
summarise(
mean_bw = mean(bw, na.rm = TRUE),
sd_bw = sd(bw, na.rm = TRUE),
min_bw = min(bw, na.rm = TRUE),
max_bw = max(bw, na.rm = TRUE),
n = n()
)
```
```{r}
# group differences in child bw
# ttest: child bw
# unique subject IDs: based on true prevalence
library(dplyr)
library(car)
# filter for unique subjects with non-missing maternal anaemia status
unique_data <- HF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status)) %>%
distinct(subject_id, .keep_all = TRUE)
# summary statistics by maternal anaemia status with exactly 2 decimals displayed
summary_stats <- unique_data %>%
group_by(maternal_preg_anaemia_status) %>%
summarise(
n = n(),
mean_bw = format(round(mean(bw, na.rm = TRUE), 2), nsmall = 2),
sd_bw = format(round(sd(bw, na.rm = TRUE), 2), nsmall = 2),
min_bw = format(round(min(bw, na.rm = TRUE), 2), nsmall = 2),
max_bw = format(round(max(bw, na.rm = TRUE), 2), nsmall = 2),
.groups = "drop"
)
# print summary stats
print(summary_stats)
# levene's Test for homogeneity of variance
levene_result <- leveneTest(bw ~ maternal_preg_anaemia_status, data = unique_data)
print(levene_result)
# extract Levene p-value (rounded for reporting)
levene_p <- round(levene_result$`Pr(>F)`[1], 4)
cat("\nLevene’s test p-value =", levene_p, "\n")
# t-test comparing birthweight by anaemia status
# Automatically use equal variances if Levene p > 0.05
if (levene_p > 0.05) {
cat("Levene’s test not significant → Using Student's t-test (equal variances assumed)\n\n")
t_result <- t.test(
bw ~ maternal_preg_anaemia_status,
data = unique_data,
var.equal = TRUE
)
} else {
cat("Levene’s test significant → Using Welch's t-test (equal variances NOT assumed)\n\n")
t_result <- t.test(
bw ~ maternal_preg_anaemia_status,
data = unique_data,
var.equal = FALSE
)
}
# Print t-test result in full precision
print(t_result)
```
*Child Birth Length (BL)*
```{r}
# child bl (full sample)
# unique subject IDs: based on true prevalence
library(dplyr)
# filter for unique subjects with non-missing maternal anaemia status
unique_data <- HF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status)) %>%
distinct(subject_id, .keep_all = TRUE)
# summary statistics for birth length
unique_data %>%
summarise(
mean_birthlength = mean(birth_length, na.rm = TRUE),
sd_birthlength = sd(birth_length, na.rm = TRUE),
min_birthlength = min(birth_length, na.rm = TRUE),
max_birthlength = max(birth_length, na.rm = TRUE),
n = n()
)
```
```{r}
# group differences in child birthlength
# ttest: child bl
# exclude repeated measures: unique id
library(dplyr)
library(car)
# keep only unique subjects with non-missing anaemia status
unique_data <- HF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status)) %>%
distinct(subject_id, .keep_all = TRUE)
# summary statistics with exactly 2 decimals displayed
summary_stats <- unique_data %>%
group_by(maternal_preg_anaemia_status) %>%
summarise(
n = n(),
mean_bl = format(round(mean(birth_length, na.rm = TRUE), 2), nsmall = 2),
sd_bl = format(round(sd(birth_length, na.rm = TRUE), 2), nsmall = 2),
min_bl = format(round(min(birth_length, na.rm = TRUE), 2), nsmall = 2),
max_bl = format(round(max(birth_length, na.rm = TRUE), 2), nsmall = 2),
.groups = "drop"
)
# print summary stats
print(summary_stats)
# levene's Test
levene_result <- leveneTest(
birth_length ~ maternal_preg_anaemia_status,
data = unique_data
)
print(levene_result)
# extract Levene p-value (rounded for reporting only)
levene_p <- round(levene_result$`Pr(>F)`[1], 4)
cat("\nLevene’s test p-value =", levene_p, "\n")
# automatically choose correct t-test
if (levene_p > 0.05) {
cat("Levene’s test not significant → Using Student's t-test (equal variances assumed)\n\n")
t_result <- t.test(
birth_length ~ maternal_preg_anaemia_status,
data = unique_data,
var.equal = TRUE
)
} else {
cat("Levene’s test significant → Using Welch's t-test (equal variances NOT assumed)\n\n")
t_result <- t.test(
birth_length ~ maternal_preg_anaemia_status,
data = unique_data,
var.equal = FALSE
)
}
# print t-test result in full precision
print(t_result)
```
*Birth Head Circumference (HC)*
```{r}
# child birth hc (full sample)
# unique subject IDs: based on true prevalence
library(dplyr)
# filter for unique subjects with non-missing maternal anaemia status
unique_data <- HF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status)) %>%
distinct(subject_id, .keep_all = TRUE)
# summary statistics for birth head circumference
unique_data %>%
summarise(
mean_hc = mean(birth_hc, na.rm = TRUE),
sd_hc = sd(birth_hc, na.rm = TRUE),
min_hc = min(birth_hc, na.rm = TRUE),
max_hc = max(birth_hc, na.rm = TRUE),
n = n()
)
```
```{r}
# group differences in child birth hc
# ttest: child hc
# unique subject IDs: based on true prevalence
library(dplyr)
library(car)
# filter for unique subjects with non-missing maternal anaemia status
unique_data <- HF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status)) %>%
distinct(subject_id, .keep_all = TRUE)
# summary statistics with exactly 2 decimals displayed
summary_stats <- unique_data %>%
group_by(maternal_preg_anaemia_status) %>%
summarise(
n = n(),
mean_hc = format(round(mean(birth_hc, na.rm = TRUE), 2), nsmall = 2),
sd_hc = format(round(sd(birth_hc, na.rm = TRUE), 2), nsmall = 2),
min_hc = format(round(min(birth_hc, na.rm = TRUE), 2), nsmall = 2),
max_hc = format(round(max(birth_hc, na.rm = TRUE), 2), nsmall = 2),
.groups = "drop"
)
# print summary stats
print(summary_stats)
# levene's Test
levene_result <- leveneTest(
birth_hc ~ maternal_preg_anaemia_status,
data = unique_data
)
print(levene_result)
# extract Levene p-value (rounded for reporting only)
levene_p <- round(levene_result$`Pr(>F)`[1], 4)
cat("\nLevene’s test p-value =", levene_p, "\n")
# automatically choose correct t-test
if (levene_p > 0.05) {
cat("Levene’s test not significant → Using Student's t-test (equal variances assumed)\n\n")
t_result <- t.test(
birth_hc ~ maternal_preg_anaemia_status,
data = unique_data,
var.equal = TRUE
)
} else {
cat("Levene’s test significant → Using Welch's t-test (equal variances NOT assumed)\n\n")
t_result <- t.test(
birth_hc ~ maternal_preg_anaemia_status,
data = unique_data,
var.equal = FALSE
)
}
# print t-test result in full precision
print(t_result)
```
*Gestational Age at Maternal Hb Measurement*
```{r}
# ga at Hb measurement (full sample)
# unique subject IDs: based on true prevalence
HF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status),
!is.na(maternal_min_hb_ga_roundown)) %>%
distinct(subject_id, .keep_all = TRUE) %>%
summarise(
n = n(),
mean_ga_hb = round(mean(maternal_min_hb_ga_roundown, na.rm = TRUE), 2),
sd_ga_hb = round(sd(maternal_min_hb_ga_roundown, na.rm = TRUE), 2),
min_ga_hb = round(min(maternal_min_hb_ga_roundown, na.rm = TRUE), 2),
q1_ga_hb = round(quantile(maternal_min_hb_ga_roundown, 0.25, na.rm = TRUE), 2),
median_ga_hb = round(median(maternal_min_hb_ga_roundown, na.rm = TRUE), 2),
q3_ga_hb = round(quantile(maternal_min_hb_ga_roundown, 0.75, na.rm = TRUE), 2),
max_ga_hb = round(max(maternal_min_hb_ga_roundown, na.rm = TRUE), 2)
)
```
```{r}
# group differences in ga at Hb measurement by antenatal maternal anaemia status
# ttest: ga at Hb measurement
# unique subject IDs: based on true prevalence
library(dplyr)
library(car)
# filter non-missing and keep one row per subject
HF_ida_final_unique <- HF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status),
!is.na(maternal_min_hb_ga_roundown)) %>%
distinct(subject_id, .keep_all = TRUE)
# summary statistics with exactly 2 decimals displayed
summary_stats <- HF_ida_final_unique %>%
group_by(maternal_preg_anaemia_status) %>%
summarise(
n = n(),
mean_ga = format(round(mean(maternal_min_hb_ga_roundown, na.rm = TRUE), 2), nsmall = 2),
sd_ga = format(round(sd(maternal_min_hb_ga_roundown, na.rm = TRUE), 2), nsmall = 2),
min_ga = format(round(min(maternal_min_hb_ga_roundown, na.rm = TRUE), 2), nsmall = 2),
max_ga = format(round(max(maternal_min_hb_ga_roundown, na.rm = TRUE), 2), nsmall = 2),
.groups = "drop"
)
# print summary stats
print(summary_stats)
# levene's test
levene_result <- leveneTest(
maternal_min_hb_ga_roundown ~ maternal_preg_anaemia_status,
data = HF_ida_final_unique
)
print(levene_result)
# extract Levene p-value (rounded for reporting)
levene_p <- round(levene_result$`Pr(>F)`[1], 4)
cat("\nLevene’s test p-value =", levene_p, "\n")
# automatically choose correct t-test
if (levene_p > 0.05) {
cat("Levene’s test not significant → Using Student's t-test (equal variances assumed)\n\n")
t_result <- t.test(
maternal_min_hb_ga_roundown ~ maternal_preg_anaemia_status,
data = HF_ida_final_unique,
var.equal = TRUE
)
} else {
cat("Levene’s test significant → Using Welch's t-test (equal variances NOT assumed)\n\n")
t_result <- t.test(
maternal_min_hb_ga_roundown ~ maternal_preg_anaemia_status,
data = HF_ida_final_unique,
var.equal = FALSE
)
}
# print t-test result in full precision
print(t_result)
```
*Minimum Maternal Hb measurement in Trimester 1*
```{r}
# min Hb across groups in trimester 1
# unique subject IDs: based on true prevalence
library(dplyr)
library(car)
# filter for trimester 1, non-missing anaemia status and min Hb, keep unique subjects
trim1_data_unique <- HF_ida_final %>%
filter(
maternal_trimester_hb == "first",
!is.na(maternal_preg_anaemia_status),
!is.na(maternal_preg_min_hb)
) %>%
distinct(subject_id, .keep_all = TRUE)
# summary statistics by anaemia status with exactly 2 decimals displayed
summary_stats <- trim1_data_unique %>%
group_by(maternal_preg_anaemia_status) %>%
summarise(
n = n(),
mean_hb = format(round(mean(maternal_preg_min_hb, na.rm = TRUE), 2), nsmall = 2),
sd_hb = format(round(sd(maternal_preg_min_hb, na.rm = TRUE), 2), nsmall = 2),
min_hb = format(round(min(maternal_preg_min_hb, na.rm = TRUE), 2), nsmall = 2),
max_hb = format(round(max(maternal_preg_min_hb, na.rm = TRUE), 2), nsmall = 2),
.groups = "drop"
)
# print summary stats
print(summary_stats)
# levene's test for homogeneity of variance
levene_result <- leveneTest(
maternal_preg_min_hb ~ maternal_preg_anaemia_status,
data = trim1_data_unique
)
print(levene_result)
# extract Levene p-value (rounded for reporting only)
levene_p <- round(levene_result$`Pr(>F)`[1], 4)
cat("\nLevene’s test p-value =", levene_p, "\n")
# automatically choose correct t-test
if (levene_p > 0.05) {
cat("Levene’s test not significant → Using Student's t-test (equal variances assumed)\n\n")
t_result <- t.test(
maternal_preg_min_hb ~ maternal_preg_anaemia_status,
data = trim1_data_unique,
var.equal = TRUE
)
} else {
cat("Levene’s test significant → Using Welch's t-test (equal variances NOT assumed)\n\n")
t_result <- t.test(
maternal_preg_min_hb ~ maternal_preg_anaemia_status,
data = trim1_data_unique,
var.equal = FALSE
)
}
# print t-test result in full precision
print(t_result)
```
*Minimum Maternal Hb in trimester 2*
```{r}
# min Hb across groups in trimester 2
# unique subject IDs: Based on true prevalence
library(dplyr)
library(car)
# filter for trimester 2, non-missing anaemia status and min Hb, keep unique subjects
trim2_data_unique <- HF_ida_final %>%
filter(
maternal_trimester_hb == "second",
!is.na(maternal_preg_anaemia_status),
!is.na(maternal_preg_min_hb)
) %>%
distinct(subject_id, .keep_all = TRUE)
# summary statistics by anaemia status with exactly 2 decimals displayed
summary_stats <- trim2_data_unique %>%
group_by(maternal_preg_anaemia_status) %>%
summarise(
n = n(),
mean_hb = format(round(mean(maternal_preg_min_hb, na.rm = TRUE), 2), nsmall = 2),
sd_hb = format(round(sd(maternal_preg_min_hb, na.rm = TRUE), 2), nsmall = 2),
min_hb = format(round(min(maternal_preg_min_hb, na.rm = TRUE), 2), nsmall = 2),
max_hb = format(round(max(maternal_preg_min_hb, na.rm = TRUE), 2), nsmall = 2),
.groups = "drop"
)
# print summary stats
print(summary_stats)
# levene's test for homogeneity of variance
levene_result <- leveneTest(
maternal_preg_min_hb ~ maternal_preg_anaemia_status,
data = trim2_data_unique
)
print(levene_result)
# extract Levene p-value (rounded for reporting only)
levene_p <- round(levene_result$`Pr(>F)`[1], 4)
cat("\nLevene’s test p-value =", levene_p, "\n")
# automatically choose correct t-test
if (levene_p > 0.05) {
cat("Levene’s test not significant → Using Student's t-test (equal variances assumed)\n\n")
t_result <- t.test(
maternal_preg_min_hb ~ maternal_preg_anaemia_status,
data = trim2_data_unique,
var.equal = TRUE
)
} else {
cat("Levene’s test significant → Using Welch's t-test (equal variances NOT assumed)\n\n")
t_result <- t.test(
maternal_preg_min_hb ~ maternal_preg_anaemia_status,
data = trim2_data_unique,
var.equal = FALSE
)
}
# print t-test result in full precision
print(t_result)
```
*Minimum Maternal Hb in Trimester 3*
```{r}
# min Hb across groups in trimester 3
# unique subject IDs: based on true prevalence
library(dplyr)
library(car)
# filter for trimester 3, non-missing anaemia status and min Hb, keep unique subjects
trim3_data_unique <- HF_ida_final %>%
filter(
maternal_trimester_hb == "third",
!is.na(maternal_preg_anaemia_status),
!is.na(maternal_preg_min_hb)
) %>%
distinct(subject_id, .keep_all = TRUE)
# summary statistics by anaemia status with exactly 2 decimals displayed
summary_stats <- trim3_data_unique %>%
group_by(maternal_preg_anaemia_status) %>%
summarise(
n = n(),
mean_hb = format(round(mean(maternal_preg_min_hb, na.rm = TRUE), 2), nsmall = 2),
sd_hb = format(round(sd(maternal_preg_min_hb, na.rm = TRUE), 2), nsmall = 2),
min_hb = format(round(min(maternal_preg_min_hb, na.rm = TRUE), 2), nsmall = 2),
max_hb = format(round(max(maternal_preg_min_hb, na.rm = TRUE), 2), nsmall = 2),
.groups = "drop"
)
# print summary stats
print(summary_stats)
# levene's test for homogeneity of variance
levene_result <- leveneTest(
maternal_preg_min_hb ~ maternal_preg_anaemia_status,
data = trim3_data_unique
)
print(levene_result)
# extract levene p-value (rounded for reporting only)
levene_p <- round(levene_result$`Pr(>F)`[1], 4)
cat("\nLevene’s test p-value =", levene_p, "\n")
# automatically choose correct t-test
if (levene_p > 0.05) {
cat("Levene’s test not significant → Using Student's t-test (equal variances assumed)\n\n")
t_result <- t.test(
maternal_preg_min_hb ~ maternal_preg_anaemia_status,
data = trim3_data_unique,
var.equal = TRUE
)
} else {
cat("Levene’s test significant → Using Welch's t-test (equal variances NOT assumed)\n\n")
t_result <- t.test(
maternal_preg_min_hb ~ maternal_preg_anaemia_status,
data = trim3_data_unique,
var.equal = FALSE
)
}
# print t-test result in full precision
print(t_result)
```
**Exploratory Analyses: Plotted Data using LOESS Curves**
Absolute Volumes for ROIs plotted against child age at scan. Includes repeated measures to capture all scans collected across study timepoints.
1. Full sample
2. By group (Key: Antenatal Maternal Anaemia Status)
*LOESS Curves: ICV*
```{r}
# icv (full sample)
# including repeated measures
library(ggplot2)
icv_hf_sample <- ggplot(HF_ida_final, aes(x = scan_age_months, y = hf_mm_icv)) +
geom_point(alpha = 0.6) +
geom_smooth(method = "loess", se = TRUE, color = "blue") +
scale_x_continuous(
breaks = seq(0, max(HF_ida_final$scan_age_months, na.rm = TRUE), by = 3) # x-axis every 3 months
) +
labs(
title = "ICV vs. Child Age",
x = "Child Age at Scan (Months)",
y = "Total ICV"
) +
theme_minimal()
icv_hf_sample
```
```{r}
# icv by group (antenatal maternal anaemia status)
# including repeated measures
library(ggplot2)
library(dplyr)
icv_hf_group <- ggplot(
data = HF_ida_final %>% filter(!is.na(maternal_preg_anaemia_status)),
aes(
x = scan_age_months,
y = hf_mm_icv,
color = maternal_preg_anaemia_status,
fill = maternal_preg_anaemia_status
)
) +
geom_point() +
geom_smooth(
method = "loess",
se = TRUE,
level = 0.95,
alpha = 0.3 # make ribbons slightly transparent
) +
scale_color_manual(
values = c(
"no_mat_anaemia" = "#1f77b4", # blue line
"mat_anaemia" = "#d62728" # red line
),
name = "Maternal Anaemia Status"
) +
scale_fill_manual(
values = c(
"no_mat_anaemia" = "#c6d9f1", # lighter blue ribbon
"mat_anaemia" = "#f7b0b0" # lighter red ribbon
),
guide = "none" # hide fill legend
) +
scale_x_continuous(
breaks = seq(0, max(HF_ida_final$scan_age_months, na.rm = TRUE), by = 3)
) +
labs(
title = "Total ICV Volume vs. Scan Age",
x = "Scan Age (Months)",
y = "Total ICV Volume (mm3)",
color = "Maternal Anaemia Status"
) +
theme_minimal()
icv_hf_group
```
*LOESS Curves: Putamen*
```{r}
# putamen (full sample)
# including repeated measures
library(ggplot2)
putamen_hf_sample <- ggplot(HF_ida_final, aes(x = scan_age_months, y = total_putamen)) +
geom_point(alpha = 0.6) +
geom_smooth(method = "loess", se = TRUE, color = "blue") +
scale_x_continuous(
breaks = seq(0, max(HF_ida_final$scan_age_months, na.rm = TRUE), by = 3) # x-axis every 3 months
) +
labs(
title = "Putamen vs. Child Age",
x = "Child Age at Scan (Months)",
y = "Total Putamen Volume"
) +
theme_minimal()
putamen_hf_sample
```
```{r}
# putamen by group (antenatal maternal anaemia status)
# including repeated measures
library(ggplot2)
library(dplyr)
putamen_hf_group <- ggplot(
data = HF_ida_final %>% filter(!is.na(maternal_preg_anaemia_status)),
aes(
x = scan_age_months,
y = total_putamen,
color = maternal_preg_anaemia_status,
fill = maternal_preg_anaemia_status
)
) +
geom_point() +
geom_smooth(
method = "loess",
se = TRUE,
level = 0.95,
alpha = 0.3 # slightly transparent ribbons
) +
scale_color_manual(
values = c(
"no_mat_anaemia" = "#1f77b4", # blue line
"mat_anaemia" = "#d62728" # red line
),
name = "Maternal Anaemia Status"
) +
scale_fill_manual(
values = c(
"no_mat_anaemia" = "#c6d9f1", # lighter blue ribbon
"mat_anaemia" = "#f7b0b0" # lighter red ribbon
),
guide = "none" # hide fill legend
) +
scale_x_continuous(
breaks = seq(0, max(HF_ida_final$scan_age_months, na.rm = TRUE), by = 3)
) +
labs(
title = "Total Putamen Volume vs. Scan Age",
x = "Scan Age (Months)",
y = "Total Putamen Volume",
color = "Maternal Anaemia Status"
) +
theme_minimal()
putamen_hf_group
```
*LOESS Curves: Caudate Nucleus*
```{r}
# caudate Nucleus (full sample)
# including repeated measures
library(ggplot2)
caudate_hf_sample <- ggplot(HF_ida_final, aes(x = scan_age_months, y = total_caudate)) +
geom_point(alpha = 0.6) +
geom_smooth(method = "loess", se = TRUE, color = "blue") +
scale_x_continuous(
breaks = seq(0, max(HF_ida_final$scan_age_months, na.rm = TRUE), by = 3) # x-axis every 3 months
) +
labs(
title = "Caudate vs. Child Age",
x = "Child Age at Scan (Months)",
y = "Total Caudate Volume"
) +
theme_minimal()
caudate_hf_sample
```
```{r}
# caudate nucleus by group (antenatal maternal anaemia status)
# including repeated measures
library(ggplot2)
library(dplyr)
caudate_hf_group <- ggplot(
data = HF_ida_final %>% filter(!is.na(maternal_preg_anaemia_status)),
aes(
x = scan_age_months,
y = total_caudate,
color = maternal_preg_anaemia_status,
fill = maternal_preg_anaemia_status
)
) +
geom_point() +
geom_smooth(
method = "loess",
se = TRUE,
level = 0.95,
alpha = 0.3 # light, semi-transparent ribbon
) +
scale_color_manual(
values = c(
"no_mat_anaemia" = "#1f77b4", # blue line
"mat_anaemia" = "#d62728" # red line
),
name = "Maternal Anaemia Status"
) +
scale_fill_manual(
values = c(
"no_mat_anaemia" = "#c6d9f1", # slightly lighter blue ribbon
"mat_anaemia" = "#f7b0b0" # slightly lighter red ribbon
),
guide = "none" # hide fill legend
) +
scale_x_continuous(
breaks = seq(0, max(HF_ida_final$scan_age_months, na.rm = TRUE), by = 3)
) +
labs(
title = "Total Caudate Volume vs. Scan Age",
x = "Scan Age (Months)",
y = "Total Caudate Volume",
color = "Maternal Anaemia Status"
) +
theme_minimal()
caudate_hf_group
```
*LOESS Curves: Corpus Callosum*
```{r}
# corpus callosum (full sample)
# including repeated measures
library(ggplot2)
cc_hf_sample <- ggplot(HF_ida_final, aes(x = scan_age_months, y = total_cc)) +
geom_point(alpha = 0.6) +
geom_smooth(method = "loess", se = TRUE, color = "blue") +
scale_x_continuous(
breaks = seq(0, max(HF_ida_final$scan_age_months, na.rm = TRUE), by = 3) # x-axis every 3 months
) +
labs(
title = "CC vs. Child Age",
x = "Child Age at Scan (Months)",
y = "Total Corpus Callosum Volume"
) +
theme_minimal()
cc_hf_sample
```
```{r}
# corpus callosum by group (antenatal maternal anaemia status)
# including repeated measures
library(ggplot2)
library(dplyr)
cc_hf_group <- ggplot(
data = HF_ida_final %>% filter(!is.na(maternal_preg_anaemia_status)),
aes(
x = scan_age_months,
y = total_cc,
color = maternal_preg_anaemia_status,
fill = maternal_preg_anaemia_status
)
) +
geom_point() +
geom_smooth(
method = "loess",
se = TRUE,
level = 0.95,
alpha = 0.3 # make ribbons slightly transparent
) +
scale_color_manual(
values = c(
"no_mat_anaemia" = "#1f77b4", # original blue line
"mat_anaemia" = "#d62728" # original red line
),
name = "Maternal Anaemia Status"
) +
scale_fill_manual(
values = c(
"no_mat_anaemia" = "#c6d9f1", # slightly lighter blue for ribbon
"mat_anaemia" = "#f7b0b0" # slightly lighter red for ribbon
),
guide = "none" # hide fill legend
) +
scale_x_continuous(
breaks = seq(0, max(HF_ida_final$scan_age_months, na.rm = TRUE), by = 3)
) +
scale_y_continuous(
n.breaks = 6
) +
labs(
title = "Total CC Volume vs. Scan Age",
x = "Scan Age (Months)",
y = "Total Corpus Callosum Volume (mm³)",
color = "Maternal Anaemia Status"
) +
theme_minimal()
cc_hf_group
```
**Zero-Order Correlation Matrices: Antenatal Maternal Anaemia Status**
```{r}
# install packages if needed
# install and load Hmisc
if (!require(Hmisc)) {
install.packages("Hmisc")
library(Hmisc)
} else {
library(Hmisc)}
```
Convert variables to numeric for correlation matrices
```{r}
# convert variables back from categorical to numeric
HF_ida_final$income_household_en <- as.numeric(HF_ida_final$income_household_en)
HF_ida_final$maternal_preg_anaemia_status_numeric <- as.numeric(as.character(
factor(HF_ida_final$maternal_preg_anaemia_status,
levels = c("no_mat_anaemia", "mat_anaemia"),
labels = c(0, 1))
))
HF_ida_final$maternal_hiv <- as.numeric(as.character(
factor(HF_ida_final$maternal_hiv,
levels = c("no_hiv", "mat_hiv"),
labels = c(0, 1))
))
HF_ida_final$child_sex <- as.numeric(as.character(
factor(HF_ida_final$child_sex,
levels = c("girl", "boy"),
labels = c(0, 1))
))
```
\*\* The variable for "antenatal maternal anaemia status" remained a categorical variable even after running the code above. Therefore, I recalled the dataset to ensure that this was numeric (as initially defined) for assessing correlations:
```{r}
# recall HF dataset
HF_ida_final <- read_excel("HF_ida_final.xlsx")
# create new variables for total brain volumes across ROIs
# basal ganglia
HF_ida_final$total_putamen <- HF_ida_final$hf_mm_left_putamen + HF_ida_final$hf_mm_right_putamen
HF_ida_final$total_caudate <- HF_ida_final$hf_mm_left_caudate + HF_ida_final$hf_mm_right_caudate
# corpus callosum
HF_ida_final$total_cc <- HF_ida_final$hf_mm_posterior_callosum + HF_ida_final$hf_mm_mid_posterior_callosum + HF_ida_final$hf_mm_central_callosum + HF_ida_final$hf_mm_mid_anterior_callosum + HF_ida_final$hf_mm_anterior_callosum
# create new age variable (days to months):
# convert days to months
# average Number of days in a month (accounting for leap years): 30.44
HF_ida_final <- HF_ida_final %>%
mutate(scan_age_months = hf_age / 30.44)
```
**Correlation Matrices**
```{r}
# correlation matrix for corpus callosum
# filter dataset
filtered_data <- HF_ida_final[!is.na(HF_ida_final$maternal_preg_anaemia_status), ]
# select relevant columns
vars <- c("child_sex", "scan_age_months", "income_household_en", "maternal_hiv", "maternal_preg_anaemia_status", "total_cc")
data_selected <- filtered_data[, vars]
# convert categorical variables to numeric (for correlation)
data_selected$child_sex <- as.numeric(as.factor(data_selected$child_sex))
data_selected$maternal_hiv <- as.numeric(as.factor(data_selected$maternal_hiv))
data_selected$maternal_preg_anaemia_status <- as.numeric(as.factor(data_selected$maternal_preg_anaemia_status))
# compute correlation matrix with significance
rcorr_result <- Hmisc::rcorr(as.matrix(data_selected))
# rcorr_result$r contains correlation coefficients
print("Correlation matrix:")
print(rcorr_result$r)
# rcorr_result$P contains p-values for testing the null hypothesis of no correlation
print("P-value matrix:")
print(rcorr_result$P)
```
```{r}
# correlation matrix for caudate nucleus
# filter dataset
filtered_data <- HF_ida_final[!is.na(HF_ida_final$maternal_preg_anaemia_status), ]
# select relevant columns
vars <- c("child_sex", "scan_age_months", "income_household_en", "maternal_hiv", "maternal_preg_anaemia_status", "total_caudate")
data_selected <- filtered_data[, vars]
# convert categorical variables to numeric (for correlation)
data_selected$child_sex <- as.numeric(as.factor(data_selected$child_sex))
data_selected$maternal_hiv <- as.numeric(as.factor(data_selected$maternal_hiv))
data_selected$maternal_preg_anaemia_status <- as.numeric(as.factor(data_selected$maternal_preg_anaemia_status))
# compute correlation matrix with significance
rcorr_result <- Hmisc::rcorr(as.matrix(data_selected))
# rcorr_result$r contains correlation coefficients
print("Correlation matrix:")
print(rcorr_result$r)
# rcorr_result$P contains p-values for testing the null hypothesis of no correlation
print("P-value matrix:")
print(rcorr_result$P)
```
```{r}
# correlation matrix for putamen
# filter dataset
filtered_data <- HF_ida_final[!is.na(HF_ida_final$maternal_preg_anaemia_status), ]
# select relevant columns
vars <- c("child_sex", "scan_age_months", "income_household_en", "maternal_hiv", "maternal_preg_anaemia_status", "total_putamen")
data_selected <- filtered_data[, vars]
# convert categorical variables to numeric (for correlation)
data_selected$child_sex <- as.numeric(as.factor(data_selected$child_sex))
data_selected$maternal_hiv <- as.numeric(as.factor(data_selected$maternal_hiv))
data_selected$maternal_preg_anaemia_status <- as.numeric(as.factor(data_selected$maternal_preg_anaemia_status))
# compute correlation matrix with significance
rcorr_result <- Hmisc::rcorr(as.matrix(data_selected))
# rcorr_result$r contains correlation coefficients
print("Correlation matrix:")
print(rcorr_result$r)
# rcorr_result$P contains p-values for testing the null hypothesis of no correlation
print("P-value matrix:")
print(rcorr_result$P)
```
```{r}
# correlation matrix for ICV
# filter dataset
filtered_data <- HF_ida_final[!is.na(HF_ida_final$maternal_preg_anaemia_status), ]
# select relevant columns
vars <- c("child_sex", "scan_age_months", "income_household_en", "maternal_hiv", "maternal_preg_anaemia_status", "hf_mm_icv")
data_selected <- filtered_data[, vars]
# convert categorical variables to numeric (for correlation)
data_selected$child_sex <- as.numeric(as.factor(data_selected$child_sex))
data_selected$maternal_hiv <- as.numeric(as.factor(data_selected$maternal_hiv))
data_selected$maternal_preg_anaemia_status <- as.numeric(as.factor(data_selected$maternal_preg_anaemia_status))
# compute correlation matrix with significance
rcorr_result <- Hmisc::rcorr(as.matrix(data_selected))
# rcorr_result$r contains correlation coefficients
print("Correlation matrix:")
print(rcorr_result$r)
# rcorr_result$P contains p-values for testing the null hypothesis of no correlation
print("P-value matrix:")
print(rcorr_result$P)
```
**Multicollinearity**
```{r}
# correlation between child age at scan and ICV
library(dplyr)
correlation_test <- cor.test(
HF_ida_final$scan_age_months[!is.na(HF_ida_final$maternal_preg_anaemia_status)],
HF_ida_final$hf_mm_icv[!is.na(HF_ida_final$maternal_preg_anaemia_status)])
print(correlation_test)
```
Decision to exclude ICV from the models for regional brain volumes due to high levels of multicollinearity (*r* \> 0.80) between child age at scan and ICV. Only child age at scan was included in the models below.
**Linear Mixed Effects (LME) Models**
```{r}
# install packages if needed
# install.packages("lme4")
library(lme4)
# install.packages("lmerTest")
library(lmerTest)
# install.packages("parameters")
library(parameters)
# install.packages("emmeans")
library(emmeans)
```
Based on the exploratory analyses (plotted data; LOESS curves), absolute age was used for models with an observed linear relationship. Log (age) was used for models with an observed non-linear relationship.
In the HF dataset:
- Corpus callosum: Absolute age - Linear growth trajectory on plot
- ICV, Putamen, Caudate Nucleus: Log (age) - Growth trajectory characterised by rapid growth and plateauing on plot
For the HF sample, no interactions (child age at scan \* antenatal maternal anaemia status) were explored due to limited power. We kept the models simple and only assessed main effects.
**LME Model variables**
- Recall the dataset to reset variables
- Create variabels for total ROI volumes and child age at scan in months
- Ensure that all categorical variables (e.g. anaemia status, child sex, maternal HIV etc) are factors
- Create variable for the logarithmic transformation of child age in months
```{r}
# recall dataset
HF_ida_final <- read_excel("HF_ida_final.xlsx")
# create new variables for total brain volumes across ROIs
# basal ganglia
HF_ida_final$total_putamen <- HF_ida_final$hf_mm_left_putamen + HF_ida_final$hf_mm_right_putamen
HF_ida_final$total_caudate <- HF_ida_final$hf_mm_left_caudate + HF_ida_final$hf_mm_right_caudate
# corpus callosum
HF_ida_final$total_cc <- HF_ida_final$hf_mm_posterior_callosum + HF_ida_final$hf_mm_mid_posterior_callosum + HF_ida_final$hf_mm_central_callosum + HF_ida_final$hf_mm_mid_anterior_callosum + HF_ida_final$hf_mm_anterior_callosum
# create new age variable (days to months):
# convert days to months
# average Number of days in a month (accounting for leap years): 30.44
HF_ida_final <- HF_ida_final %>%
mutate(scan_age_months = hf_age / 30.44)
```
```{r}
# recode variables from numeric to factor
# maternal anaemia status
HF_ida_final$maternal_preg_anaemia_status <- factor(
HF_ida_final$maternal_preg_anaemia_status,
levels = c(0, 1),
labels = c("no_mat_anaemia", "mat_anaemia")
)
# maternal ID status
HF_ida_final$maternal_preg_adj_brinda_id <- factor(
HF_ida_final$maternal_preg_adj_brinda_id,
levels = c(0, 1),
labels = c("no_mat_id", "mat_id")
)
# maternal ID status
HF_ida_final$maternal_preg_unadjusted_id <- factor(
HF_ida_final$maternal_preg_unadjusted_id,
levels = c(0, 1),
labels = c("no_mat_id", "mat_id")
)
# maternal trimester Hb
HF_ida_final$maternal_trimester_hb <- factor(
HF_ida_final$maternal_trimester_hb,
levels = c(1, 2, 3),
labels = c("first", "second", "third")
)
# child sex
HF_ida_final$child_sex <- factor(
HF_ida_final$child_sex,
levels = c(0, 1),
labels = c("girl", "boy")
)
# maternal pae
HF_ida_final$maternal_pae <- factor(
HF_ida_final$maternal_pae,
levels = c(0, 1),
labels = c("no_pae", "pae")
)
# maternal pte
HF_ida_final$maternal_pte <- factor(
HF_ida_final$maternal_pte,
levels = c(0, 1),
labels = c("no_pte", "pte")
)
# maternal hiv
HF_ida_final$maternal_hiv <- factor(
HF_ida_final$maternal_hiv,
levels = c(0, 1),
labels = c("no_hiv", "hiv")
)
# maternal dep
HF_ida_final$maternal_dep <- factor(
HF_ida_final$maternal_dep,
levels = c(0, 1),
labels = c("no_dep", "dep")
)
# maternal employment
HF_ida_final$mom_employment_en_dichotomous <- factor(
HF_ida_final$mom_employment_en_dichotomous,
levels = c(0, 1),
labels = c("unemployed", "employed")
)
# household income
HF_ida_final$income_household_en <- factor(
HF_ida_final$income_household_en,
levels = c(1, 2, 3, 4),
labels = c("Less than R1000 per month", "R1000-R5000 per monthy", " R5000-R10 000 per month", "More than R10 000 per month" )
)
# maternal education
HF_ida_final$mom_edu_en <- factor(
HF_ida_final$mom_edu_en,
levels = c(2, 3, 4, 5, 6),
labels = c("primary", "some_secondary", "complete_secondary", "some_tertiary", "completed_tertiary" )
)
# baby hiv
HF_ida_final$baby_hiv <- factor(
HF_ida_final$baby_hiv,
levels = c(0, 1),
labels = c("no_child_hiv", "child_hiv")
)
```
```{r}
# create a log(age) variable for use in models of brain regions with non-linear growth
HF_ida_final$log_age <- log(HF_ida_final$scan_age_months)
```
*LME: Corpus Callosum*
```{r}
# LME for corpus callosum
# no interaction between maternal anaemia and age
# absolute age for linear growth
# model summary
hf_model_cc <- lmer(total_cc ~ maternal_preg_anaemia_status + scan_age_months + child_sex + (1 | subject_id),
data = HF_ida_final)
summary(hf_model_cc)
# beta values
standardize_parameters(hf_model_cc, method="refit")
```
```{r}
# emms for corpus callosum volume by antenatal maternal anaemia status
# based on final adjusted model
library(emmeans)
emm_cc <- emmeans(hf_model_cc, ~ maternal_preg_anaemia_status)
# ----- EMM table with exactly 2 decimals -----
emm_cc_df <- as.data.frame(emm_cc)
emm_cc_df$emmean <- format(round(emm_cc_df$emmean, 2), nsmall = 2)
emm_cc_df$SE <- format(round(emm_cc_df$SE, 2), nsmall = 2)
emm_cc_df$lower.CL <- format(round(emm_cc_df$lower.CL, 2), nsmall = 2)
emm_cc_df$upper.CL <- format(round(emm_cc_df$upper.CL, 2), nsmall = 2)
print(emm_cc_df)
# ----- contrast results -----
contrast_results_cc <- contrast(emm_cc, method = "revpairwise", adjust = "holm")
contrast_df_cc <- as.data.frame(summary(contrast_results_cc, infer = TRUE))
# force 2 decimals for estimate
contrast_df_cc$estimate <- format(round(contrast_df_cc$estimate, 2), nsmall = 2)
print(contrast_df_cc)
```
*LME: ICV*
```{r}
# LME for icv
# no interaction between maternal anaemia and age
# log (scan age) as growth non-linear
# model summary
hf_model_icv <- lmer(hf_mm_icv ~ maternal_preg_anaemia_status + log_age + child_sex + (1 | subject_id),
data = HF_ida_final)
summary(hf_model_icv)
# beta values
standardize_parameters(hf_model_icv)
```
```{r}
# emms for icv by antenatal maternal anaemia status
# based on final adjusted model
library(emmeans)
emm_icv <- emmeans(hf_model_icv, ~ maternal_preg_anaemia_status)
# ----- EMM table with exactly 2 decimals -----
emm_icv_df <- as.data.frame(emm_icv)
emm_icv_df$emmean <- format(round(emm_icv_df$emmean, 2), nsmall = 2)
emm_icv_df$SE <- format(round(emm_icv_df$SE, 2), nsmall = 2)
emm_icv_df$lower.CL <- format(round(emm_icv_df$lower.CL, 2), nsmall = 2)
emm_icv_df$upper.CL <- format(round(emm_icv_df$upper.CL, 2), nsmall = 2)
print(emm_icv_df)
# ----- contrast results -----
contrast_results_icv <- contrast(emm_icv, method = "revpairwise", adjust = "holm")
contrast_df_icv <- as.data.frame(summary(contrast_results_icv, infer = TRUE))
# force 2 decimals for estimate
contrast_df_icv$estimate <- format(round(contrast_df_icv$estimate, 2), nsmall = 2)
print(contrast_df_icv)
```
*LME: Caudate Nucleus*
```{r}
# lme for caudate nucleus
# no interaction between maternal anaemia and age
# log (scan age) as growth non-linear
hf_model_caudate <- lmer(total_caudate ~ maternal_preg_anaemia_status + log_age + child_sex + (1 | subject_id),
data = HF_ida_final)
summary(hf_model_caudate)
# beta values
standardize_parameters(hf_model_caudate)
```
```{r}
# emms for caudate nucleus by antenatal maternal anaemia status
# based on final adjusted model
library(emmeans)
emm_caudate <- emmeans(hf_model_caudate, ~ maternal_preg_anaemia_status)
# ----- emm table with exactly 2 decimals -----
emm_caudate_df <- as.data.frame(emm_caudate)
emm_caudate_df$emmean <- format(round(emm_caudate_df$emmean, 2), nsmall = 2)
emm_caudate_df$SE <- format(round(emm_caudate_df$SE, 2), nsmall = 2)
emm_caudate_df$lower.CL <- format(round(emm_caudate_df$lower.CL, 2), nsmall = 2)
emm_caudate_df$upper.CL <- format(round(emm_caudate_df$upper.CL, 2), nsmall = 2)
print(emm_caudate_df)
# ----- contrast results -----
contrast_results_caudate <- contrast(emm_caudate, method = "revpairwise", adjust = "holm")
contrast_df_caudate <- as.data.frame(summary(contrast_results_caudate, infer = TRUE))
# force 2 decimals for estimate
contrast_df_caudate$estimate <- format(round(contrast_df_caudate$estimate, 2), nsmall = 2)
print(contrast_df_caudate)
```
*LME: Putamen*
```{r}
# lme for putamen
# no interaction between maternal anaemia and age
# log (scan age) as growth non-linear
hf_model_putamen <- lmer(total_putamen ~ maternal_preg_anaemia_status + log_age + child_sex + (1 | subject_id),
data = HF_ida_final)
summary(hf_model_putamen)
# beta values
standardize_parameters(hf_model_putamen)
```
```{r}
# emms for caudate nucleus by antenatal maternal anaemia status
# based on final adjusted model
library(emmeans)
emm_putamen <- emmeans(hf_model_putamen, ~ maternal_preg_anaemia_status)
# ----- emm table with exactly 2 decimals -----
emm_putamen_df <- as.data.frame(emm_putamen)
emm_putamen_df$emmean <- format(round(emm_putamen_df$emmean, 2), nsmall = 2)
emm_putamen_df$SE <- format(round(emm_putamen_df$SE, 2), nsmall = 2)
emm_putamen_df$lower.CL <- format(round(emm_putamen_df$lower.CL, 2), nsmall = 2)
emm_putamen_df$upper.CL <- format(round(emm_putamen_df$upper.CL, 2), nsmall = 2)
print(emm_putamen_df)
# ----- contrast results -----
contrast_results_putamen <- contrast(emm_putamen, method = "revpairwise", adjust = "holm")
contrast_df_putamen <- as.data.frame(summary(contrast_results_putamen, infer = TRUE))
# force 2 decimals for estimate
contrast_df_putamen$estimate <- format(round(contrast_df_putamen$estimate, 2), nsmall = 2)
print(contrast_df_putamen)
```
**Sensitivity Analyses**
Based on the final models for each ROI (established in primary analyses): With the inclusion of maternal HIV (potential confounder)
```{r}
# sensitivity analysis for corpus callosum
# including maternal hiv
# excluding interaction between maternal anaemia and age
hf_model_cc_sens <- lmer(total_cc ~ maternal_preg_anaemia_status + scan_age_months + child_sex + maternal_hiv + (1 | subject_id),
data = HF_ida_final)
summary(hf_model_cc_sens)
standardize_parameters(hf_model_cc_sens)
```
```{r}
# sensitivity analysis for icv
# including maternal hiv
# excluding interaction between maternal anaemia and age
hf_model_icv_sens <- lmer(hf_mm_icv ~ maternal_preg_anaemia_status + log_age + child_sex + maternal_hiv + (1 | subject_id),
data = HF_ida_final)
summary(hf_model_icv_sens)
standardize_parameters(hf_model_icv_sens)
```
```{r}
# sensitivity analysis for putamen
# including maternal hiv
# excluding interaction between maternal anaemia and age
hf_model_putamen_sens <- lmer(total_putamen ~ maternal_preg_anaemia_status + log_age + child_sex + maternal_hiv + (1 | subject_id),
data = HF_ida_final)
summary(hf_model_putamen_sens)
standardize_parameters(hf_model_putamen_sens)
```
```{r}
# sensitivity analysis for caudate nucleus
# including maternal hiv
# excluding interaction between maternal anaemia and age
hf_model_caudate_sens <- lmer(total_caudate ~ maternal_preg_anaemia_status + log_age + child_sex + maternal_hiv + (1 | subject_id),
data = HF_ida_final)
summary(hf_model_caudate_sens)
standardize_parameters(hf_model_caudate_sens)
```
------------------------------------------------------------------------
[***Ultra-Low-Field (ULF) Analyses***]{.underline}
------------------------------------------------------------------------
**Import ULF Dataset**
```{r}
ULF_ida_final <- read_excel("ULF_ida_final.xlsx")
```
**Create New Variables for Total Brain Volumes across ROIs**
```{r}
# basal ganglia
ULF_ida_final$total_putamen <- ULF_ida_final$mm_hyp_left_putamen + ULF_ida_final$mm_hyp_right_putamen
ULF_ida_final$total_caudate <- ULF_ida_final$mm_hyp_left_caudate + ULF_ida_final$mm_hyp_right_caudate
# corpus callosum
ULF_ida_final$total_cc <- ULF_ida_final$mm_hyp_posterior_callosum + ULF_ida_final$mm_hyp_mid_posterior_callosum + ULF_ida_final$mm_hyp_central_callosum + ULF_ida_final$mm_hyp_mid_anterior_callosum + ULF_ida_final$mm_hyp_anterior_callosum
```
**Create New Age Variable (days to months)**
```{r}
# convert days to months
# average number of days in a month (accounting for leap years): 30.44
ULF_ida_final <- ULF_ida_final %>%
mutate(scan_age_months = hyp_age / 30.44)
```
**Create Meaningful Categorical Variables**
```{r}
# convert from numeric to factor
# maternal anaemia status
ULF_ida_final$maternal_preg_anaemia_status <- factor(
ULF_ida_final$maternal_preg_anaemia_status,
levels = c(0, 1),
labels = c("no_mat_anaemia", "mat_anaemia")
)
# maternal anaemia severity
ULF_ida_final$maternal_preg_anaemia_severity <- factor(
ULF_ida_final$maternal_preg_anaemia_severity,
levels = c(1, 2, 3),
labels = c("mild", "moderate", "severe")
)
# child status
ULF_ida_final$cor_child_anaemia <- factor(
ULF_ida_final$cor_child_anaemia,
levels = c(0, 1),
labels = c("no_child_anaemia", "child_anaemia")
)
# maternal ID status
ULF_ida_final$maternal_preg_adj_brinda_id <- factor(
ULF_ida_final$maternal_preg_adj_brinda_id,
levels = c(0, 1),
labels = c("no_mat_id", "mat_id")
)
# child ID status
ULF_ida_final$cor_child_id <- factor(
ULF_ida_final$cor_child_id,
levels = c(0, 1),
labels = c("no_child_id", "child_id")
)
# maternal trimester hb
ULF_ida_final$maternal_trimester_hb <- factor(
ULF_ida_final$maternal_trimester_hb,
levels = c(1, 2, 3),
labels = c("first", "second", "third")
)
# child sex
ULF_ida_final$child_sex <- factor(
ULF_ida_final$child_sex,
levels = c(0, 1),
labels = c("girl", "boy")
)
# maternal pae
ULF_ida_final$maternal_pae <- factor(
ULF_ida_final$maternal_pae,
levels = c(0, 1),
labels = c("no_pae", "pae")
)
# maternal pte
ULF_ida_final$maternal_pte <- factor(
ULF_ida_final$maternal_pte,
levels = c(0, 1),
labels = c("no_pte", "pte")
)
# maternal hiv
ULF_ida_final$maternal_hiv <- factor(
ULF_ida_final$maternal_hiv,
levels = c(0, 1),
labels = c("no_hiv", "hiv")
)
# maternal dep
ULF_ida_final$maternal_dep <- factor(
ULF_ida_final$maternal_dep,
levels = c(0, 1),
labels = c("no_dep", "dep")
)
# maternal employment
ULF_ida_final$mom_employment_en_dichotomous <- factor(
ULF_ida_final$mom_employment_en_dichotomous,
levels = c(0, 1),
labels = c("unemployed", "employed")
)
# household income
ULF_ida_final$income_household_en <- factor(
ULF_ida_final$income_household_en,
levels = c(1, 2, 3, 4),
labels = c("Less than R1000 per month", "R1000-R5000 per monthy", " R5000-R10 000 per month", "More than R10 000 per month" )
)
# maternal education
ULF_ida_final$mom_edu_en <- factor(
ULF_ida_final$mom_edu_en,
levels = c(2, 3, 4, 5, 6),
labels = c("primary", "some_secondary", "complete_secondary", "some_tertiary", "completed_tertiary" )
)
# baby hiv
ULF_ida_final$baby_hiv <- factor(
ULF_ida_final$baby_hiv,
levels = c(0, 1),
labels = c("no_child_hiv", "child_hiv")
)
```
**Data Distribution across Timepoint**
```{r}
# sample over 12 months with maternal anaemia data
sample_over_12M <- ULF_ida_final %>%
filter(scan_age_months > 12,
!is.na(maternal_preg_anaemia_status))
# sample under 12 months with maternal anaemia data
sample_under_12M <- ULF_ida_final %>%
filter(scan_age_months < 12,
!is.na(maternal_preg_anaemia_status))
# sample over 18 months with maternal anaemia data
sample_over_18 <- ULF_ida_final %>%
filter(scan_age_months > 18,
!is.na(maternal_preg_anaemia_status))
# sample over 24 months with maternal anaemia data
sample_over_24 <- ULF_ida_final %>%
filter(scan_age_months > 24,
!is.na(maternal_preg_anaemia_status))
```
**Overlap Between HF and ULF Samples**
No. of unique subjects for full ULF neuroimaging sample:
```{r}
# unique observations in ULF dataset
# excluding repeated measures
library(dplyr)
ULF_ida_final %>%
summarise(n_unique_subjects = n_distinct(subject_id))
```
Overlap in unique subjects between full HF and ULF samples:
```{r}
# no. of children in common between HF and ULF neuroimaging datasets
# find overlapping unique subject_ids
overlap_subjects <- intersect(HF_ida_final$subject_id, ULF_ida_final$subject_id)
# Count how many there are
num_overlap <- length(overlap_subjects)
# Print the number
num_overlap
```
For neuroimaging sample with maternal anaemia data available:
```{r}
# no. of unique children in common between HF and ULF neuroimaging datasets with maternal anaemia data
# filter dataframes to keep only rows with non-missing maternal_preg_anaemia_status
HF_filtered <- HF_ida_final[!is.na(HF_ida_final$maternal_preg_anaemia_status), ]
ULF_filtered <- ULF_ida_final[!is.na(ULF_ida_final$maternal_preg_anaemia_status), ]
# find overlapping subject_ids among the filtered sets
overlap_subjects <- intersect(HF_filtered$subject_id, ULF_filtered$subject_id)
# count the number of overlapping subjects
num_overlap <- length(overlap_subjects)
# print the result
num_overlap
```
------------------------------------------------------------------------
**Primary Analysis: Antenatal Maternal Anaemia Status (ULF Sample)**
------------------------------------------------------------------------
**Sample Size: For infants with Antenatal Maternal Anemia Data**
Note: Sample sizes were calculated with 1) the inclusion of repeated measures (multiple scans across study timepoints), and for 2) unique subject IDs excluding repeated measures.
```{r}
# total sample size
# including repeated measures
ULF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status)) %>% # only include PIDs with antenatal maternal anaemia data
summarise(n = n())
```
```{r}
# repeated measures excluded
# unique subject IDs: based on true prevalence
library(dplyr)
ULF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status)) %>%
summarise(unique_subjects = n_distinct(subject_id))
```
```{r}
# prevalence of antenatal maternal anaemia
# unique subject IDs: based on true prevalence
ULF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status)) %>% # Exclude missing data
group_by(subject_id) %>%
summarise(
maternal_preg_anaemia_status = if_else(
any(maternal_preg_anaemia_status == "mat_anaemia", na.rm = TRUE),
"mat_anaemia",
"no_mat_anaemia"
)
) %>%
count(maternal_preg_anaemia_status) %>%
mutate(
prevalence_percent = n / sum(n) * 100
)
```
```{r}
# antenatal maternal anaemia severity
# unique subject IDs: based on true prevalence
library(dplyr)
ULF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_severity)) %>% # exclude missing severity
group_by(subject_id) %>% # group by mother
summarise(
maternal_preg_anaemia_severity = first(
maternal_preg_anaemia_severity[maternal_preg_anaemia_severity != ""]
)
) %>%
group_by(maternal_preg_anaemia_severity) %>% # count unique mothers by severity
summarise(n = n()) %>%
mutate(percent = round(100 * n / sum(n), 2))
```
**Data Availability with Repeated Measures**
```{r}
# repeated measures count
# unique ID
library(dplyr)
ULF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status)) %>% # keep only rows with anaemia status
count(subject_id) %>% # count rows per subject
count(n) %>% # count how many subjects had n rows
rename(times_measured = n, number_of_subjects = nn) %>%
mutate(percent = round(100 * number_of_subjects / sum(number_of_subjects), 1)) # Add % column
```
```{r}
# figure 2
library(ggplot2)
library(dplyr)
Figure_2_ULF <-ULF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status)) %>%
mutate(
timepoint = factor(timepoint, levels = c("3M", "6M", "12M", "18M", "24M")), # enforce order
subject_id = factor(subject_id),
anaemia_group = as.factor(maternal_preg_anaemia_status)
) %>%
ggplot(aes(x = timepoint, y = subject_id, group = subject_id, color = anaemia_group)) +
geom_line(alpha = 0.5, linewidth = 0.5) + # use linewidth instead of deprecated size
geom_point(size = 2) +
scale_color_manual(
values = c("red", "blue"),
labels = c("Maternal Anaemia", "No Maternal Anaemia")
) +
labs(
title = "Participant Data Availability by Timepoint",
x = "Timepoint",
y = "Participants",
color = "Maternal Anaemia"
) +
theme_minimal(base_size = 13) +
theme(
axis.text.y = element_blank(),
axis.ticks.y = element_blank(),
panel.grid.major.y = element_blank()
) +
# <-- add padding so points are not cut off
scale_y_discrete(expand = expansion(mult = c(0.02, 0.02))) +
scale_x_discrete(expand = expansion(mult = c(0.02, 0.02)))
print(Figure_2_ULF)
```
**Sample Characteristics and Group Differences: Antenatal Maternal Anaemia Status**
Sample characteristics were assessed based on the number of unique subjects in each subsample, excluding repeated measures data for children with multiple scans. This approach was used to avoid skewing results with duplicated demographic and clinical observations for the same infants. The only exception to this approach was child age at scan which was considered for the full repeated measures subsample, with a proportion of children having multiple scans acquired at different ages.
However, given the repeated measures study design with individual infant data across multiple study visits in both HF and ULF subsamples, linear mixed effects (LME) models were fitted to account for individual differences in assessing the association between antenatal maternal anaemia status and absolute regional brain volume trajectories
- **Categorical Variables**
*Trimester of Maternal Hb Measurement*
```{r}
# trimester Hb was collected from mothers
# excluding repeated measures
# unique subject IDs: based on true prevalence
library(dplyr)
ULF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status)) %>% # exclude missing anaemia
group_by(subject_id) %>% # keep unique mothers only
summarise(
maternal_trimester_hb = first(maternal_trimester_hb) # assume first record per mother
) %>%
group_by(maternal_trimester_hb) %>%
summarise(n = n()) %>%
mutate(percent = round(100 * n / sum(n), 2))
```
```{r}
# group differences in trimester of Hb collection
# Chi-squared: trimester Hb
# unique subject IDs: based on true prevalence
library(dplyr)
# run contingency table + test without creating a new dataset
tbl <- ULF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status)) %>% # exclude missing anaemia status
group_by(subject_id) %>% # ensure one row per mother
summarise(
maternal_trimester_hb = first(maternal_trimester_hb),
maternal_preg_anaemia_status = first(maternal_preg_anaemia_status)
) %>%
{table(.$maternal_trimester_hb, .$maternal_preg_anaemia_status)} # build contingency table inline
# print contingency table
cat("Contingency Table: Trimester by Maternal Anaemia Status\n")
print(format(round(tbl, 2), nsmall = 2))
cat("\nColumn Percentages (% of each anaemia status group):\n")
print(round(100 * prop.table(tbl, margin = 2), 2))
# run appropriate test
chisq_result <- suppressWarnings(chisq.test(tbl))
expected <- chisq_result$expected
if (any(expected < 5)) {
cat("\nFisher's Exact Test (due to small expected cell counts):\n")
print(fisher.test(tbl))
} else {
cat("\nChi-squared Test:\n")
print(chisq_result)
}
```
*Child Sex at Birth*
```{r}
# child sex at birth (full sample)
# unique subject IDs: based on true prevalence
library(dplyr)
ULF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status)) %>% # exclude missing anaemia data
group_by(subject_id) %>% # ensure one record per child
summarise(
child_sex = first(child_sex) # assume sex is consistent per child
) %>%
group_by(child_sex) %>%
summarise(n = n()) %>%
mutate(percent = round(100 * n / sum(n), 2))
```
```{r}
# group differences in child sex by antenatal maternal anaemia status
# chi-squared: child sex
# unique subject IDs: based on true prevalence
library(dplyr)
# run contingency table + test without creating a new dataset
tbl <- ULF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status)) %>% # exclude missing anaemia data
group_by(subject_id) %>% # ensure one record per child
summarise(
child_sex = first(child_sex),
maternal_preg_anaemia_status = first(maternal_preg_anaemia_status)
) %>%
{table(.$child_sex, .$maternal_preg_anaemia_status)} # build contingency table inline
# print contingency table
cat("Contingency Table: Child Sex by Maternal Anaemia Status\n")
print(format(round(tbl, 2), nsmall = 2))
cat("\nColumn Percentages (% of each anaemia status group):\n")
print(round(100 * prop.table(tbl, margin = 2), 2))
# run appropriate test
chisq_result <- suppressWarnings(chisq.test(tbl))
expected <- chisq_result$expected
if (any(expected < 5)) {
cat("\nFisher's Exact Test (due to small expected cell counts):\n")
print(fisher.test(tbl))
} else {
cat("\nChi-squared Test:\n")
print(chisq_result)
}
```
*Prenatal Alcohol Exposure (PAE)*
```{r}
# pae prevalence (full sample)
# unique subject IDs: based on true prevalence
library(dplyr)
ULF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status)) %>% # exclude missing anaemia data
group_by(subject_id) %>% # ensure one record per mother
summarise(
maternal_pae = first(maternal_pae) # assume pae status is consistent per mother
) %>%
group_by(maternal_pae) %>%
summarise(n = n()) %>%
mutate(percent = round(100 * n / sum(n), 2))
```
```{r}
# group differences in pae by antenatal maternal anaemia status
# chi-squared: pae
# unique subject IDs: based on true prevalence
library(dplyr)
# build contingency table and run test inline (unique mothers only)
tbl <- ULF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status)) %>% # exclude missing anaemia data
group_by(subject_id) %>% # ensure one record per mother
summarise(
maternal_pae = first(maternal_pae),
maternal_preg_anaemia_status = first(maternal_preg_anaemia_status)
) %>%
{table(.$maternal_pae, .$maternal_preg_anaemia_status)} # build contingency table inline
# print contingency table
cat("Contingency Table: Maternal PAE by Maternal Anaemia Status\n")
print(tbl)
cat("\nColumn Percentages (% of each anaemia status group):\n")
print(round(100 * prop.table(tbl, margin = 2), 2))
# run appropriate test
chisq_result <- suppressWarnings(chisq.test(tbl))
expected <- chisq_result$expected
if (any(expected < 5)) {
cat("\nFisher's Exact Test (due to small expected cell counts):\n")
print(fisher.test(tbl))
} else {
cat("\nChi-squared Test:\n")
print(chisq_result)
}
```
*Prenatal Tobacco Exposure (PTE)*
```{r}
# pte prevalence (full sample)
# unique subject IDs: based on true prevalence
library(dplyr)
ULF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status)) %>% # exclude missing anaemia data
group_by(subject_id) %>% # ensure one record per mother
summarise(
maternal_pte = first(maternal_pte) # assume pte status is consistent per mother
) %>%
group_by(maternal_pte) %>%
summarise(n = n()) %>%
mutate(percent = round(100 * n / sum(n), 2))
```
```{r}
# group differences in pte by antenatal maternal anaemia status
# chi-squared: pte
# unique subject IDs: based on true prevalence
library(dplyr)
# build contingency table and run test inline (unique mothers only)
tbl <- ULF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status)) %>% # exclude missing anaemia data
group_by(subject_id) %>% # ensure one record per mother
summarise(
maternal_pte = first(maternal_pte),
maternal_preg_anaemia_status = first(maternal_preg_anaemia_status)
) %>%
{table(.$maternal_pte, .$maternal_preg_anaemia_status)} # build contingency table inline
# print contingency table
cat("Contingency Table: Maternal PTE by Maternal Anaemia Status\n")
print(tbl)
cat("\nColumn Percentages (% of each anaemia status group):\n")
print(round(100 * prop.table(tbl, margin = 2), 2))
# run appropriate test
chisq_result <- suppressWarnings(chisq.test(tbl))
expected <- chisq_result$expected
if (any(expected < 5)) {
cat("\nFisher's Exact Test (due to small expected cell counts):\n")
print(fisher.test(tbl))
} else {
cat("\nChi-squared Test:\n")
print(chisq_result)
}
```
*Maternal HIV Infection*
```{r}
# materal hiv prevalence (full sample)
# unique subject IDs: based on true prevalence
ULF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status)) %>% # exclude missing anaemia data
group_by(subject_id) %>% # ensure one record per mother
summarise(
maternal_hiv = first(maternal_hiv) # assume HIV status is consistent per mother
) %>%
group_by(maternal_hiv) %>%
summarise(n = n()) %>%
mutate(percent = round(100 * n / sum(n), 2))
```
```{r}
# group differences in maternal hiv by antenatal maternal anaemia status
# chi-squared: hiv
# unique subject IDs: based on true prevalence
library(dplyr)
# build contingency table and run test inline (unique mothers only)
tbl <- ULF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status)) %>% # exclude missing anaemia data
group_by(subject_id) %>% # ensure one record per mother
summarise(
maternal_hiv = first(maternal_hiv),
maternal_preg_anaemia_status = first(maternal_preg_anaemia_status)
) %>%
{table(.$maternal_hiv, .$maternal_preg_anaemia_status)} # build contingency table inline
# print contingency table
cat("Contingency Table: Maternal HIV by Maternal Anaemia Status\n")
print(tbl)
cat("\nColumn Percentages (% of each anaemia status group):\n")
print(round(100 * prop.table(tbl, margin = 2), 2))
# run appropriate test
chisq_result <- suppressWarnings(chisq.test(tbl))
expected <- chisq_result$expected
if (any(expected < 5)) {
cat("\nFisher's Exact Test (due to small expected cell counts):\n")
print(fisher.test(tbl))
} else {
cat("\nChi-squared Test:\n")
print(chisq_result)
}
```
*Maternal Depression*
```{r}
# maternal depression prevalence (full sample)
# unique subject IDs: based on true prevalence
ULF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status)) %>%
distinct(subject_id, .keep_all = TRUE) %>%
group_by(maternal_dep) %>%
summarise(n = n()) %>%
mutate(percent = round(100 * n / sum(n), 2))
```
```{r}
# group differenced by maternal depression status
# chi-squared: maternal depression
# unique subject IDs: based on true prevalence
library(dplyr)
# build contingency table and run test inline (unique mothers only)
tbl <- ULF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status)) %>% # exclude missing anaemia data
group_by(subject_id) %>% # ensure one record per mother
summarise(
maternal_dep = first(maternal_dep),
maternal_preg_anaemia_status = first(maternal_preg_anaemia_status)
) %>%
{table(.$maternal_dep, .$maternal_preg_anaemia_status)} # build contingency table inline
# print contingency table
cat("Contingency Table: Maternal Depression by Maternal Anaemia Status\n")
print(tbl)
cat("\nColumn Percentages (% of each anaemia status group):\n")
print(round(100 * prop.table(tbl, margin = 2), 2))
# run appropriate test
chisq_result <- suppressWarnings(chisq.test(tbl))
expected <- chisq_result$expected
if (any(expected < 5)) {
cat("\nFisher's Exact Test (due to small expected cell counts):\n")
print(fisher.test(tbl))
} else {
cat("\nChi-squared Test:\n")
print(chisq_result)
}
```
*Maternal Employment*
```{r}
# maternal employment (full sample)
# unique subject IDs: based on true prevalence
library(dplyr)
ULF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status)) %>% # exclude missing anaemia status
group_by(subject_id) %>% # ensure one record per mother
summarise(mom_employment_en_dichotomous = first(mom_employment_en_dichotomous)) %>%
group_by(mom_employment_en_dichotomous) %>%
summarise(n = n()) %>%
mutate(percent = round(100 * n / sum(n), 2))
```
```{r}
# group differences in maternal employment by antenatal maternal anaemia status
# chi-squared: maternal employment
# unique subject IDs: based on true prevalence
# create contingency table inline using first record per mother
tbl <- table(
ULF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status)) %>%
group_by(subject_id) %>%
summarise(mom_employment_en_dichotomous = first(mom_employment_en_dichotomous),
maternal_preg_anaemia_status = first(maternal_preg_anaemia_status)) %>%
pull(mom_employment_en_dichotomous),
ULF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status)) %>%
group_by(subject_id) %>%
summarise(mom_employment_en_dichotomous = first(mom_employment_en_dichotomous),
maternal_preg_anaemia_status = first(maternal_preg_anaemia_status)) %>%
pull(maternal_preg_anaemia_status)
)
# print contingency table
cat("Contingency Table: Maternal Employment by Maternal Anaemia Status\n")
print(tbl)
cat("\nColumn Percentages (% of each anaemia status group):\n")
print(round(100 * prop.table(tbl, margin = 2), 2))
# run appropriate test
expected <- chisq.test(tbl)$expected
if (any(expected < 5)) {
cat("\nFisher's Exact Test (due to small expected cell counts):\n")
print(fisher.test(tbl))
} else {
cat("\nChi-squared Test:\n")
print(chisq.test(tbl))
}
```
*Maternal Education*
```{r}
# maternal education (full sample)
# unique subject IDs: based on true prevalence
ULF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status)) %>% # Include only mothers with anaemia data
group_by(subject_id) %>% # Ensure one record per mother
summarise(mom_edu_en = first(mom_edu_en)) %>% # Take first value per mother
group_by(mom_edu_en) %>%
summarise(n = n()) %>%
mutate(percent = round(100 * n / sum(n), 2))
```
```{r}
# group differences in maternal education by antenatal maternal anaemia status
# chi-squared: maternal education
# unique subject IDs: based on true prevalence
# create contingency table inline, one record per mother
tbl <- table(
ULF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status)) %>%
group_by(subject_id) %>% # Ensure unique mother
summarise(mom_edu_en = first(mom_edu_en)) %>%
pull(mom_edu_en),
ULF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status)) %>%
group_by(subject_id) %>%
summarise(maternal_preg_anaemia_status = first(maternal_preg_anaemia_status)) %>%
pull(maternal_preg_anaemia_status)
)
# print contingency table
cat("Contingency Table: Maternal Education by Maternal Anaemia Status\n")
print(tbl)
cat("\nColumn Percentages (% of each anaemia status group):\n")
print(round(100 * prop.table(tbl, margin = 2), 2))
# run appropriate test
expected <- chisq.test(tbl)$expected
if (any(expected < 5)) {
cat("\nFisher's Exact Test (due to small expected cell counts):\n")
print(fisher.test(tbl))
} else {
cat("\nChi-squared Test:\n")
print(chisq.test(tbl))
}
```
*Household Income*
```{r}
# household income (full sample)
# unique subject IDs: based on true prevalence
ULF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status)) %>% # only include mothers with anaemia data
group_by(subject_id) %>% # ensure each mother counted once
summarise(income_household_en = first(income_household_en)) %>%
group_by(income_household_en) %>%
summarise(n = n()) %>%
mutate(percent = round(100 * n / sum(n), 2))
```
```{r}
# group differences in household income by antenatal maternal anaemia status
# chi-squared: household income
# unique subject IDs: based on true prevalence
# create contingency table counting each subject only once
tbl <- table(
ULF_ida_final$income_household_en[!duplicated(ULF_ida_final$subject_id) &
!is.na(ULF_ida_final$maternal_preg_anaemia_status)],
ULF_ida_final$maternal_preg_anaemia_status[!duplicated(ULF_ida_final$subject_id) &
!is.na(ULF_ida_final$maternal_preg_anaemia_status)]
)
# print contingency table
cat("Contingency Table: Household Income by Maternal Anaemia Status\n")
print(tbl)
cat("\nColumn Percentages (% of each anaemia status group):\n")
print(round(100 * prop.table(tbl, margin = 2), 2))
# run appropriate test
expected <- chisq.test(tbl)$expected
if (any(expected < 5)) {
cat("\nFisher's Exact Test (due to small expected cell counts):\n")
print(fisher.test(tbl))
} else {
cat("\nChi-squared Test:\n")
print(chisq.test(tbl))
}
```
*Infant HIV Infection*
```{r}
# infant hiv prevalence (full sample)
# unique subject IDs: based on true prevalence
ULF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status)) %>%
filter(!duplicated(subject_id)) %>% # count each subject once
group_by(baby_hiv) %>%
summarise(n = n()) %>%
mutate(percent = round(100 * n / sum(n), 2))
```
```{r}
# group differences in infant hiv by antenatal maternal anaemia status
# chi-squared: infant hiv
# unique subject IDs: based on true prevalence
# create contingency table counting each subject only once
tbl <- table(
ULF_ida_final$baby_hiv[!is.na(ULF_ida_final$maternal_preg_anaemia_status) & !duplicated(ULF_ida_final$subject_id)],
ULF_ida_final$maternal_preg_anaemia_status[!is.na(ULF_ida_final$maternal_preg_anaemia_status) & !duplicated(ULF_ida_final$subject_id)]
)
# print contingency table
cat("Contingency Table: Infant HIV by Maternal Anaemia Status\n")
print(tbl)
cat("\nColumn Percentages (% of each anaemia status group):\n")
print(round(100 * prop.table(tbl, margin = 2), 2))
# run appropriate test
expected <- chisq.test(tbl)$expected
if (any(expected < 5)) {
cat("\nFisher's Exact Test (due to small expected cell counts):\n")
print(fisher.test(tbl))
} else {
cat("\nChi-squared Test:\n")
print(chisq.test(tbl))
}
```
- **Continuous Variables**
```{r}
#install packages if needed
#install.packages('car')
library(car)
```
*Maternal Age at Enrolment*
```{r}
# maternal age at enrolment (full sample)
# unique subject IDs: based on true prevalence
library(dplyr)
ULF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status),
!is.na(mom_age_en)) %>% # exclude missing anaemia or age
group_by(subject_id) %>% # keep unique mothers only
summarise(
mom_age_en = first(mom_age_en) # assume maternal age at enrolment is constant
) %>%
summarise(
mean_mom_age = mean(mom_age_en, na.rm = TRUE),
sd_mom_age = sd(mom_age_en, na.rm = TRUE),
min_mom_age = min(mom_age_en, na.rm = TRUE),
max_mom_age = max(mom_age_en, na.rm = TRUE),
n = n()
)
```
```{r}
# group differences in maternal age at enrolment by antenatal maternal anaemia status
# ttest: maternal age at enrolment
# unique subject IDs: based on true prevalence
library(dplyr)
library(car)
# keep only unique subjects with non-missing anaemia status
unique_data <- ULF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status),
!is.na(mom_age_en)) %>%
distinct(subject_id, .keep_all = TRUE)
# summary stats (force exactly 2 decimals for display)
unique_data %>%
group_by(maternal_preg_anaemia_status) %>%
summarise(
n = n(),
mean_age = format(round(mean(mom_age_en, na.rm = TRUE), 2), nsmall = 2),
sd_age = format(round(sd(mom_age_en, na.rm = TRUE), 2), nsmall = 2),
min_age = format(round(min(mom_age_en, na.rm = TRUE), 2), nsmall = 2),
max_age = format(round(max(mom_age_en, na.rm = TRUE), 2), nsmall = 2),
.groups = "drop"
) %>%
print()
# levene's test for homogeneity of variance
levene_result <- leveneTest(
mom_age_en ~ maternal_preg_anaemia_status,
data = unique_data
)
print(levene_result)
# extract p-value from Levene’s Test
levene_p <- levene_result$`Pr(>F)`[1]
# decide automatically which t-test to use
if (levene_p > 0.05) {
cat("\nLevene’s test not significant (p =", round(levene_p, 4),
") → Using Student's t-test (equal variances assumed)\n\n")
t_result <- t.test(
mom_age_en ~ maternal_preg_anaemia_status,
data = unique_data,
var.equal = TRUE
)
} else {
cat("\nLevene’s test significant (p =", round(levene_p, 4),
") → Using Welch's t-test (equal variances NOT assumed)\n\n")
t_result <- t.test(
mom_age_en ~ maternal_preg_anaemia_status,
data = unique_data,
var.equal = FALSE
)
}
# print t-test results (full precision)
print(t_result)
```
*Child Age at Scan:*
Include repeated measures to account for different ages for scans acquired across study timepoints from the same infant
```{r}
# child age at scan for full group
# include repeated measures as child is a different age at every scan acquired (full sample)
ULF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status)) %>% #only include PIDs with antenatal maternal anaemia data
summarise(
mean_child_age = mean(scan_age_months, na.rm = TRUE),
sd_child_age = sd(scan_age_months, na.rm = TRUE),
min_child_age = min(scan_age_months, na.rm = TRUE),
max_child_age = max(scan_age_months, na.rm = TRUE),
n = n()
)
```
```{r}
# child age at scan
# full sample including repeated measures
# t-test: child age at scan
library(dplyr)
library(car)
# keep only unique subjects with non-missing anaemia status and scan age
unique_data <- ULF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status),
!is.na(scan_age_months))
# summary stats (force exactly 2 decimals)
unique_data %>%
group_by(maternal_preg_anaemia_status) %>%
summarise(
n = n(),
mean_age = format(round(mean(scan_age_months, na.rm = TRUE), 2), nsmall = 2),
sd_age = format(round(sd(scan_age_months, na.rm = TRUE), 2), nsmall = 2),
min_age = format(round(min(scan_age_months, na.rm = TRUE), 2), nsmall = 2),
max_age = format(round(max(scan_age_months, na.rm = TRUE), 2), nsmall = 2),
.groups = "drop"
) %>%
print()
# levene's test for homogeneity of variance
levene_result <- leveneTest(
scan_age_months ~ maternal_preg_anaemia_status,
data = unique_data
)
print(levene_result)
# extract p-value from Levene’s Test
levene_p <- levene_result$`Pr(>F)`[1]
# decide automatically which t-test to use
if (levene_p > 0.05) {
cat("\nLevene’s test not significant (p =", round(levene_p, 4),
") → Using Student's t-test (equal variances assumed)\n\n")
t_result <- t.test(
scan_age_months ~ maternal_preg_anaemia_status,
data = unique_data,
var.equal = TRUE
)
} else {
cat("\nLevene’s test significant (p =", round(levene_p, 4),
") → Using Welch's t-test (equal variances NOT assumed)\n\n")
t_result <- t.test(
scan_age_months ~ maternal_preg_anaemia_status,
data = unique_data,
var.equal = FALSE
)
}
# print t-test results (full precision)
print(t_result)
```
*Child Gestational Age (GA) at Birth*
```{r}
# child ga at birth for full group
# unique subject IDs: based on true prevalence
library(dplyr)
ULF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status)) %>%
distinct(subject_id, .keep_all = TRUE) %>% # keep only one row per child
summarise(
mean_child_ga = mean(ga_weeks, na.rm = TRUE),
sd_child_ga = sd(ga_weeks, na.rm = TRUE),
min_child_ga = min(ga_weeks, na.rm = TRUE),
max_child_ga = max(ga_weeks, na.rm = TRUE),
n = n()
)
```
```{r}
# group differences in child ga at birth
# ttest: child ga at birth
# unique subject IDs: based on true prevalence
library(dplyr)
library(car)
# keep only unique subjects with non-missing maternal anaemia status and GA
unique_data <- ULF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status),
!is.na(ga_weeks)) %>%
distinct(subject_id, .keep_all = TRUE)
# summary stats by maternal anaemia status (force 2 decimals)
unique_data %>%
group_by(maternal_preg_anaemia_status) %>%
summarise(
n = n(),
mean_ga = format(round(mean(ga_weeks, na.rm = TRUE), 2), nsmall = 2),
sd_ga = format(round(sd(ga_weeks, na.rm = TRUE), 2), nsmall = 2),
min_ga = format(round(min(ga_weeks, na.rm = TRUE), 2), nsmall = 2),
max_ga = format(round(max(ga_weeks, na.rm = TRUE), 2), nsmall = 2),
.groups = "drop"
) %>%
print()
# levene's test for homogeneity of variance
levene_result <- leveneTest(
ga_weeks ~ maternal_preg_anaemia_status,
data = unique_data
)
print(levene_result)
# extract p-value from Levene’s Test
levene_p <- levene_result$`Pr(>F)`[1]
# Decide automatically which t-test to use
if (levene_p > 0.05) {
cat("\nLevene’s test not significant (p =", round(levene_p, 4),
") → Using Student's t-test (equal variances assumed)\n\n")
t_result <- t.test(
ga_weeks ~ maternal_preg_anaemia_status,
data = unique_data,
var.equal = TRUE
)
} else {
cat("\nLevene’s test significant (p =", round(levene_p, 4),
") → Using Welch's t-test (equal variances NOT assumed)\n\n")
t_result <- t.test(
ga_weeks ~ maternal_preg_anaemia_status,
data = unique_data,
var.equal = FALSE
)
}
# print t-test results (full precision)
print(t_result)
```
*Child Birth Weight (BW)*
```{r}
# child bw (full sample)
# unique subject IDs: based on true prevalence
ULF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status)) %>% # only include records with maternal anaemia data
distinct(subject_id, .keep_all = TRUE) %>% # ensure one record per child
summarise(
mean_bw = mean(bw, na.rm = TRUE),
sd_bw = sd(bw, na.rm = TRUE),
min_bw = min(bw, na.rm = TRUE),
max_bw = max(bw, na.rm = TRUE),
n = n()
)
```
```{r}
# group differences in child bw
# ttest: child bw
# exclude repeated measures:unique ID
library(dplyr)
library(car)
# keep only unique subjects with non-missing maternal anaemia status and birthweight
unique_data <- ULF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status),
!is.na(bw)) %>%
distinct(subject_id, .keep_all = TRUE)
# summary stats by maternal anaemia status (force 2 decimals)
unique_data %>%
group_by(maternal_preg_anaemia_status) %>%
summarise(
n = n(),
mean_bw = format(round(mean(bw, na.rm = TRUE), 2), nsmall = 2),
sd_bw = format(round(sd(bw, na.rm = TRUE), 2), nsmall = 2),
min_bw = format(round(min(bw, na.rm = TRUE), 2), nsmall = 2),
max_bw = format(round(max(bw, na.rm = TRUE), 2), nsmall = 2),
.groups = "drop"
) %>%
print()
# levene's test for homogeneity of variance
levene_result <- leveneTest(
bw ~ maternal_preg_anaemia_status,
data = unique_data
)
print(levene_result)
# extract p-value from Levene’s Test
levene_p <- levene_result$`Pr(>F)`[1]
# decide automatically which t-test to use
if (levene_p > 0.05) {
cat("\nLevene’s test not significant (p =", round(levene_p, 4),
") → Using Student's t-test (equal variances assumed)\n\n")
t_result <- t.test(
bw ~ maternal_preg_anaemia_status,
data = unique_data,
var.equal = TRUE
)
} else {
cat("\nLevene’s test significant (p =", round(levene_p, 4),
") → Using Welch's t-test (equal variances NOT assumed)\n\n")
t_result <- t.test(
bw ~ maternal_preg_anaemia_status,
data = unique_data,
var.equal = FALSE
)
}
# print t-test results (full precision)
print(t_result)
```
*Child Birth Length (BL)*
```{r}
# child bl (full sample)
# unique subject IDs: based on true prevalence
ULF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status)) %>% # only include records with anaemia data
distinct(subject_id, .keep_all = TRUE) %>% # ensure only one record per child
summarise(
mean_birthlength = mean(birth_length, na.rm = TRUE),
sd_birthlength = sd(birth_length, na.rm = TRUE),
min_birthlength = min(birth_length, na.rm = TRUE),
max_birthlength = max(birth_length, na.rm = TRUE),
n = n()
)
```
```{r}
# child bl
# unique subject IDs: based on true prevalence
# t-test: child bl
library(dplyr)
library(car)
# keep only unique subjects with non-missing maternal anaemia status and birth length
unique_data <- ULF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status),
!is.na(birth_length)) %>%
distinct(subject_id, .keep_all = TRUE)
# summary stats by maternal anaemia status (force 2 decimals)
unique_data %>%
group_by(maternal_preg_anaemia_status) %>%
summarise(
n = n(),
mean_bl = format(round(mean(birth_length, na.rm = TRUE), 2), nsmall = 2),
sd_bl = format(round(sd(birth_length, na.rm = TRUE), 2), nsmall = 2),
min_bl = format(round(min(birth_length, na.rm = TRUE), 2), nsmall = 2),
max_bl = format(round(max(birth_length, na.rm = TRUE), 2), nsmall = 2),
.groups = "drop"
) %>%
print()
# levene's test for homogeneity of variance
levene_result <- leveneTest(
birth_length ~ maternal_preg_anaemia_status,
data = unique_data
)
print(levene_result)
# extract p-value from Levene’s Test
levene_p <- levene_result$`Pr(>F)`[1]
# decide automatically which t-test to use
if (levene_p > 0.05) {
cat("\nLevene’s test not significant (p =", round(levene_p, 4),
") → Using Student's t-test (equal variances assumed)\n\n")
t_result <- t.test(
birth_length ~ maternal_preg_anaemia_status,
data = unique_data,
var.equal = TRUE
)
} else {
cat("\nLevene’s test significant (p =", round(levene_p, 4),
") → Using Welch's t-test (equal variances NOT assumed)\n\n")
t_result <- t.test(
birth_length ~ maternal_preg_anaemia_status,
data = unique_data,
var.equal = FALSE
)
}
# print t-test results (full precision)
print(t_result)
```
*Child Birth Head Circumference (HC)*
```{r}
# child birth hc (full sample)
# unique subject IDs: based on true prevalence
ULF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status)) %>% # only include records with anaemia data
distinct(subject_id, .keep_all = TRUE) %>% # keep only one record per unique child
summarise(
mean_hc = mean(birth_hc, na.rm = TRUE),
sd_hc = sd(birth_hc, na.rm = TRUE),
min_hc = min(birth_hc, na.rm = TRUE),
max_hc = max(birth_hc, na.rm = TRUE),
n = n()
)
```
```{r}
# group differences in birth hc by antenatal maternal anaemia status
# t-test: child hc at birth
# unique subject IDs: based on true prevalence
library(dplyr)
library(car)
# keep only unique subjects with non-missing maternal anaemia status and birth HC
unique_data <- ULF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status),
!is.na(birth_hc)) %>%
distinct(subject_id, .keep_all = TRUE)
# summary stats by maternal anaemia status (force 2 decimals)
unique_data %>%
group_by(maternal_preg_anaemia_status) %>%
summarise(
n = n(),
mean_hc = format(round(mean(birth_hc, na.rm = TRUE), 2), nsmall = 2),
sd_hc = format(round(sd(birth_hc, na.rm = TRUE), 2), nsmall = 2),
min_hc = format(round(min(birth_hc, na.rm = TRUE), 2), nsmall = 2),
max_hc = format(round(max(birth_hc, na.rm = TRUE), 2), nsmall = 2),
.groups = "drop"
) %>%
print()
# levene's test for homogeneity of variance
levene_result <- leveneTest(
birth_hc ~ maternal_preg_anaemia_status,
data = unique_data
)
print(levene_result)
# extract p-value from Levene’s Test
levene_p <- levene_result$`Pr(>F)`[1]
# Decide automatically which t-test to use
if (levene_p > 0.05) {
cat("\nLevene’s test not significant (p =", round(levene_p, 4),
") → Using Student's t-test (equal variances assumed)\n\n")
t_result <- t.test(
birth_hc ~ maternal_preg_anaemia_status,
data = unique_data,
var.equal = TRUE
)
} else {
cat("\nLevene’s test significant (p =", round(levene_p, 4),
") → Using Welch's t-test (equal variances NOT assumed)\n\n")
t_result <- t.test(
birth_hc ~ maternal_preg_anaemia_status,
data = unique_data,
var.equal = FALSE
)
}
# print t-test results (full precision)
print(t_result)
```
*Gestational Age (GA) at Maternal Hb Measurement*
```{r}
# ga at min Hb measurement (full sample)
# unique subject IDs: based on true prevalence
ULF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status),
!is.na(maternal_min_hb_ga_roundown)) %>%
distinct(subject_id, .keep_all = TRUE) %>%
summarise(
n = n(),
mean_ga_hb = round(mean(maternal_min_hb_ga_roundown, na.rm = TRUE), 2),
sd_ga_hb = round(sd(maternal_min_hb_ga_roundown, na.rm = TRUE), 2),
min_ga_hb = round(min(maternal_min_hb_ga_roundown, na.rm = TRUE), 2),
q1_ga_hb = round(quantile(maternal_min_hb_ga_roundown, 0.25, na.rm = TRUE), 2),
median_ga_hb = round(median(maternal_min_hb_ga_roundown, na.rm = TRUE), 2),
q3_ga_hb = round(quantile(maternal_min_hb_ga_roundown, 0.75, na.rm = TRUE), 2),
max_ga_hb = round(max(maternal_min_hb_ga_roundown, na.rm = TRUE), 2)
)
```
```{r}
# group differences in ga at minimum maternal Hb by antenatal maternal anaemia status
# t-test: ga at min Hb measurement
# unique subject IDs: based on true prevalence
library(dplyr)
library(car)
# keep only unique subjects with non-missing maternal anaemia status and GA at min Hb
unique_data <- ULF_ida_final %>%
filter(!is.na(maternal_preg_anaemia_status),
!is.na(maternal_min_hb_ga_roundown)) %>%
distinct(subject_id, .keep_all = TRUE)
# summary stats by maternal anaemia status (force 2 decimals)
unique_data %>%
group_by(maternal_preg_anaemia_status) %>%
summarise(
n = n(),
mean_ga_hb = format(round(mean(maternal_min_hb_ga_roundown, na.rm = TRUE), 2), nsmall = 2),
sd_ga_hb = format(round(sd(maternal_min_hb_ga_roundown, na.rm = TRUE), 2), nsmall = 2),
min_ga_hb = format(round(min(maternal_min_hb_ga_roundown, na.rm = TRUE), 2), nsmall = 2),
max_ga_hb = format(round(max(maternal_min_hb_ga_roundown, na.rm = TRUE), 2), nsmall = 2),
.groups = "drop"
) %>%
print()
# Levene's test for homogeneity of variance
levene_result <- leveneTest(
maternal_min_hb_ga_roundown ~ maternal_preg_anaemia_status,
data = unique_data
)
print(levene_result)
# extract p-value from Levene’s Test
levene_p <- levene_result$`Pr(>F)`[1]
# decide automatically which t-test to use
if (levene_p > 0.05) {
cat("\nLevene’s test not significant (p =", round(levene_p, 4),
") → Using Student's t-test (equal variances assumed)\n\n")
t_result <- t.test(
maternal_min_hb_ga_roundown ~ maternal_preg_anaemia_status,
data = unique_data,
var.equal = TRUE
)
} else {
cat("\nLevene’s test significant (p =", round(levene_p, 4),
") → Using Welch's t-test (equal variances NOT assumed)\n\n")
t_result <- t.test(
maternal_min_hb_ga_roundown ~ maternal_preg_anaemia_status,
data = unique_data,
var.equal = FALSE
)
}
# print t-test results (full precision)
print(t_result)
```
*Minimum Maternal Hb measurement in Trimester 1*
```{r}
# maternal min Hb across groups in trimester 1: by antenatal maternal anaemia status
# unique subject IDs: based on true prevalence
library(dplyr)
library(car)
# filter for trimester 1, non-missing anaemia status and min Hb, keep unique subjects
trim1_data_unique <- ULF_ida_final %>%
filter(
maternal_trimester_hb == "first",
!is.na(maternal_preg_anaemia_status),
!is.na(maternal_preg_min_hb)
) %>%
distinct(subject_id, .keep_all = TRUE)
# summary statistics by anaemia status with exactly 2 decimals displayed
summary_stats <- trim1_data_unique %>%
group_by(maternal_preg_anaemia_status) %>%
summarise(
n = n(), # ← missing comma fixed here
mean_hb = format(round(mean(maternal_preg_min_hb, na.rm = TRUE), 2), nsmall = 2),
sd_hb = format(round(sd(maternal_preg_min_hb, na.rm = TRUE), 2), nsmall = 2),
min_hb = format(round(min(maternal_preg_min_hb, na.rm = TRUE), 2), nsmall = 2),
max_hb = format(round(max(maternal_preg_min_hb, na.rm = TRUE), 2), nsmall = 2),
.groups = "drop"
)
# print summary stats
print(summary_stats)
# levene's test for homogeneity of variance
levene_result <- leveneTest(
maternal_preg_min_hb ~ maternal_preg_anaemia_status,
data = trim1_data_unique
)
print(levene_result)
# extract Levene p-value (rounded for reporting only)
levene_p <- round(levene_result$`Pr(>F)`[1], 4)
cat("\nLevene’s test p-value =", levene_p, "\n")
# automatically choose correct t-test
if (levene_p > 0.05) {
cat("Levene’s test not significant → Using Student's t-test (equal variances assumed)\n\n")
t_result <- t.test(
maternal_preg_min_hb ~ maternal_preg_anaemia_status,
data = trim1_data_unique,
var.equal = TRUE
)
} else {
cat("Levene’s test significant → Using Welch's t-test (equal variances NOT assumed)\n\n")
t_result <- t.test(
maternal_preg_min_hb ~ maternal_preg_anaemia_status,
data = trim1_data_unique,
var.equal = FALSE
)
}
# print t-test result in full precision
print(t_result)
```
*Minimum Maternal Hb measurement in Trimester 2*
```{r}
# maternal minimum Hb in trimester 2 by antenatal anaemia status
# t-test
# unique subject IDs: based on true prevalence
library(dplyr)
library(car)
# filter for trimester 2, non-missing anaemia status and min Hb, keep unique subjects
trim2_data_unique <- ULF_ida_final %>%
filter(
maternal_trimester_hb == "second",
!is.na(maternal_preg_anaemia_status),
!is.na(maternal_preg_min_hb)
) %>%
distinct(subject_id, .keep_all = TRUE)
# summary statistics by anaemia status with exactly 2 decimals displayed
summary_stats <- trim2_data_unique %>%
group_by(maternal_preg_anaemia_status) %>%
summarise(
n = n(),
mean_hb = format(round(mean(maternal_preg_min_hb, na.rm = TRUE), 2), nsmall = 2),
sd_hb = format(round(sd(maternal_preg_min_hb, na.rm = TRUE), 2), nsmall = 2),
min_hb = format(round(min(maternal_preg_min_hb, na.rm = TRUE), 2), nsmall = 2),
max_hb = format(round(max(maternal_preg_min_hb, na.rm = TRUE), 2), nsmall = 2),
.groups = "drop"
)
# print summary stats
print(summary_stats)
# levene's test for homogeneity of variance
levene_result <- leveneTest(
maternal_preg_min_hb ~ maternal_preg_anaemia_status,
data = trim2_data_unique
)
print(levene_result)
# extract levene p-value (rounded for reporting only)
levene_p <- round(levene_result$`Pr(>F)`[1], 4)
cat("\nLevene’s test p-value =", levene_p, "\n")
# automatically choose correct t-test
if (levene_p > 0.05) {
cat("Levene’s test not significant → Using Student's t-test (equal variances assumed)\n\n")
t_result <- t.test(
maternal_preg_min_hb ~ maternal_preg_anaemia_status,
data = trim2_data_unique,
var.equal = TRUE
)
} else {
cat("Levene’s test significant → Using Welch's t-test (equal variances NOT assumed)\n\n")
t_result <- t.test(
maternal_preg_min_hb ~ maternal_preg_anaemia_status,
data = trim2_data_unique,
var.equal = FALSE
)
}
# print t-test result in full precision
print(t_result)
```
*Minimum Maternal Hb measurement in Trimester 3*
```{r}
# maternal minimum Hb in trimester 3 by antenatal anaemia status
# t-test
# unique subject IDs: based on true prevalence
library(dplyr)
library(car)
# filter for trimester 3, non-missing anaemia status and min Hb, keep unique subjects
trim3_data_unique <- ULF_ida_final %>%
filter(
maternal_trimester_hb == "third",
!is.na(maternal_preg_anaemia_status),
!is.na(maternal_preg_min_hb)
) %>%
distinct(subject_id, .keep_all = TRUE)
# summary statistics by anaemia status with exactly 2 decimals displayed
summary_stats <- trim3_data_unique %>%
group_by(maternal_preg_anaemia_status) %>%
summarise(
n = n(),
mean_hb = format(round(mean(maternal_preg_min_hb, na.rm = TRUE), 2), nsmall = 2),
sd_hb = format(round(sd(maternal_preg_min_hb, na.rm = TRUE), 2), nsmall = 2),
min_hb = format(round(min(maternal_preg_min_hb, na.rm = TRUE), 2), nsmall = 2),
max_hb = format(round(max(maternal_preg_min_hb, na.rm = TRUE), 2), nsmall = 2),
.groups = "drop"
)
# print summary stats
print(summary_stats)
# levene's test for homogeneity of variance
levene_result <- leveneTest(
maternal_preg_min_hb ~ maternal_preg_anaemia_status,
data = trim3_data_unique
)
print(levene_result)
# extract Levene p-value (rounded for reporting only)
levene_p <- round(levene_result$`Pr(>F)`[1], 4)
cat("\nLevene’s test p-value =", levene_p, "\n")
# automatically choose correct t-test
if (levene_p > 0.05) {
cat("Levene’s test not significant → Using Student's t-test (equal variances assumed)\n\n")
t_result <- t.test(
maternal_preg_min_hb ~ maternal_preg_anaemia_status,
data = trim3_data_unique,
var.equal = TRUE
)
} else {
cat("Levene’s test significant → Using Welch's t-test (equal variances NOT assumed)\n\n")
t_result <- t.test(
maternal_preg_min_hb ~ maternal_preg_anaemia_status,
data = trim3_data_unique,
var.equal = FALSE
)
}
# print t-test result in full precision
print(t_result)
```
**Exploratory Analyses: Plotted Data using LOESS Curves**
Absolute Volumes for ROIs plotted against child age at scan. Includes repeated measures to capture all scans collected across study timepoints.
1. Full sample
2. By group (Key: Antenatal Maternal Anaemia Status)
*LOESS Curves: ICV*
```{r}
# icv (full sample)
# including repeated measures
library(ggplot2)
icv_ulf_sample <- ggplot(ULF_ida_final, aes(x = scan_age_months, y = mm_hyp_icv)) +
geom_point(alpha = 0.6) +
geom_smooth(method = "loess", se = TRUE, color = "blue") +
scale_x_continuous(
breaks = seq(0, max(ULF_ida_final$scan_age_months, na.rm = TRUE), by = 3) # x-axis every 3 months
) +
labs(
title = "ICV vs. Child Age",
x = "Child Age at Scan (Months)",
y = "Total ICV"
) +
theme_minimal()
icv_ulf_sample
```
```{r}
# icv by group (antenatal maternal anaemia status)
# including repeated measures
icv_ulf_group <- ggplot(
data = ULF_ida_final %>% filter(!is.na(maternal_preg_anaemia_status)),
aes(
x = scan_age_months,
y = mm_hyp_icv,
color = maternal_preg_anaemia_status,
fill = maternal_preg_anaemia_status
)
) +
geom_point() +
geom_smooth(
method = "loess",
se = TRUE,
level = 0.95,
alpha = 0.3 # slightly transparent ribbons
) +
scale_color_manual(
values = c(
"no_mat_anaemia" = "#1f77b4",
"mat_anaemia" = "#d62728"
),
name = "Maternal Anaemia Status"
) +
scale_fill_manual(
values = c(
"no_mat_anaemia" = "#c6d9f1", # slightly lighter blue ribbon
"mat_anaemia" = "#f7b0b0" # slightly lighter red ribbon
),
guide = "none"
) +
scale_x_continuous(
breaks = seq(0, max(ULF_ida_final$scan_age_months, na.rm = TRUE), by = 3)
) +
scale_y_continuous(
limits = c(600000, NA),
breaks = seq(600000, max(ULF_ida_final$mm_hyp_icv, na.rm = TRUE), by = 200000)
) +
labs(
title = "Total ICV Volume vs. Scan Age",
x = "Scan Age (Months)",
y = "Total ICV Volume (mm³)",
color = "Maternal Anaemia Status"
) +
theme_minimal()
icv_ulf_group
```
*LOESS Curve: Putamen*
```{r}
# putamen (full sample)
# including repeated measures
library(ggplot2)
putamen_ulf_sample <- ggplot(ULF_ida_final, aes(x = scan_age_months, y = total_putamen)) +
geom_point(alpha = 0.6) +
geom_smooth(method = "loess", se = TRUE, color = "blue") +
scale_x_continuous(
breaks = seq(0, max(ULF_ida_final$scan_age_months, na.rm = TRUE), by = 3) # x-axis every 3 months
) +
labs(
title = "Putamen vs. Child Age",
x = "Child Age at Scan (Months)",
y = "Total Putamen Volume"
) +
theme_minimal()
putamen_ulf_sample
```
```{r}
# putamen by group (antenatal maternal anaemia status)
# including repeated measures
putamen_ulf_group <- ggplot(
data = ULF_ida_final %>% filter(!is.na(maternal_preg_anaemia_status)),
aes(
x = scan_age_months,
y = total_putamen,
color = maternal_preg_anaemia_status,
fill = maternal_preg_anaemia_status
)
) +
geom_point() +
geom_smooth(
method = "loess",
se = TRUE,
level = 0.95,
alpha = 0.3 # slightly transparent ribbons
) +
scale_color_manual(
values = c(
"no_mat_anaemia" = "#1f77b4",
"mat_anaemia" = "#d62728"
),
name = "Maternal Anaemia Status"
) +
scale_fill_manual(
values = c(
"no_mat_anaemia" = "#c6d9f1", # slightly lighter blue ribbon
"mat_anaemia" = "#f7b0b0" # slightly lighter red ribbon
),
guide = "none"
) +
scale_x_continuous(
breaks = seq(0, max(ULF_ida_final$scan_age_months, na.rm = TRUE), by = 3)
) +
labs(
title = "Total Putamen Volume vs. Scan Age",
x = "Scan Age (months)",
y = "Total Putamen Volume",
color = "Maternal Anaemia Status"
) +
theme_minimal()
putamen_ulf_group
```
*LOESS Curves: Caudate Nucleus*
```{r}
# caudate nucleus (full sample)
# including repeated measures
library(ggplot2)
caudate_hf_sample <- ggplot(ULF_ida_final, aes(x = scan_age_months, y = total_caudate)) +
geom_point(alpha = 0.6) +
geom_smooth(method = "loess", se = TRUE, color = "blue") +
scale_x_continuous(
breaks = seq(0, max(ULF_ida_final$scan_age_months, na.rm = TRUE), by = 3) # x-axis every 3 months
) +
labs(
title = "Caudate vs. Child Age",
x = "Child Age at Scan (Months)",
y = "Total Caudate Volume"
) +
theme_minimal()
caudate_hf_sample
```
```{r}
# caudate nucleus by group (antenatal maternal anaemia status)
# including repeated measures
caudate_ulf_group <- ggplot(
data = ULF_ida_final %>% filter(!is.na(maternal_preg_anaemia_status)),
aes(
x = scan_age_months,
y = total_caudate,
color = maternal_preg_anaemia_status,
fill = maternal_preg_anaemia_status
)
) +
geom_point() +
geom_smooth(
method = "loess",
se = TRUE,
level = 0.95,
alpha = 0.3 # slightly transparent ribbons
) +
scale_color_manual(
values = c(
"no_mat_anaemia" = "#1f77b4",
"mat_anaemia" = "#d62728"
),
name = "Maternal Anaemia Status"
) +
scale_fill_manual(
values = c(
"no_mat_anaemia" = "#c6d9f1", # slightly lighter blue ribbon
"mat_anaemia" = "#f7b0b0" # slightly lighter red ribbon
),
guide = "none"
) +
scale_x_continuous(
breaks = seq(0, max(ULF_ida_final$scan_age_months, na.rm = TRUE), by = 3)
) +
labs(
title = "Total Caudate Volume vs. Scan Age",
x = "Scan Age (months)",
y = "Total Caudate Volume",
color = "Maternal Anaemia Status"
) +
theme_minimal()
caudate_ulf_group
```
*LOESS Curves: Corpus Callosum*
```{r}
# corpus callosum (full sample)
# including repeated measures
library(ggplot2)
cc_ulf_sample <- ggplot(ULF_ida_final, aes(x = scan_age_months, y = total_cc)) +
geom_point(alpha = 0.6) +
geom_smooth(method = "loess", se = TRUE, color = "blue") +
scale_x_continuous(
breaks = seq(0, max(ULF_ida_final$scan_age_months, na.rm = TRUE), by = 3) # x-axis every 3 months
) +
labs(
title = "CC vs. Child Age",
x = "Child Age at Scan (Months)",
y = "Total Corpus Callosum Volume"
) +
theme_minimal()
cc_ulf_sample
```
```{r}
# corpus callosum by group (antenatal maternal anaemia status)
# including repeated measures
cc_ulf_group <- ggplot(
data = ULF_ida_final %>% filter(!is.na(maternal_preg_anaemia_status)),
aes(
x = scan_age_months,
y = total_cc,
color = maternal_preg_anaemia_status,
fill = maternal_preg_anaemia_status
)
) +
geom_point() +
geom_smooth(
method = "loess",
se = TRUE,
level = 0.95,
alpha = 0.3 # slightly transparent ribbons
) +
scale_color_manual(
values = c(
"no_mat_anaemia" = "#1f77b4",
"mat_anaemia" = "#d62728"
),
name = "Maternal Anaemia Status"
) +
scale_fill_manual(
values = c(
"no_mat_anaemia" = "#c6d9f1", # slightly lighter blue ribbon
"mat_anaemia" = "#f7b0b0" # slightly lighter red ribbon
),
guide = "none"
) +
scale_x_continuous(
breaks = seq(0, max(ULF_ida_final$scan_age_months, na.rm = TRUE), by = 3)
) +
scale_y_continuous(
n.breaks = 6
) +
labs(
title = "Total CC Volume vs. Scan Age",
x = "Scan Age (Months)",
y = "Total Corpus Callosum Volume (mm³)",
color = "Maternal Anaemia Status"
) +
theme_minimal()
cc_ulf_group
```
**Zero-Order Correlation Matrices: Antenatal Maternal Anaemia Status**
```{r}
# install packages if needed
# install and load Hmisc
if (!require(Hmisc)) {
install.packages("Hmisc")
library(Hmisc)
} else {
library(Hmisc)}
```
Convert variables to numeric for correlations
```{r}
# convert variables back from categorical to numeric in ULF_ida_final
# household income
ULF_ida_final$income_household_en <- as.numeric(ULF_ida_final$income_household_en)
# maternal anaemia status: 0 = no anaemia, 1 = anaemia
ULF_ida_final$maternal_preg_anaemia_status_numeric <- as.numeric(as.character(
factor(ULF_ida_final$maternal_preg_anaemia_status,
levels = c("no_mat_anaemia", "mat_anaemia"),
labels = c(0, 1))
))
# maternal HIV status: 0 = no HIV, 1 = HIV
ULF_ida_final$maternal_hiv <- as.numeric(as.character(
factor(ULF_ida_final$maternal_hiv,
levels = c("no_hiv", "mat_hiv"),
labels = c(0, 1))
))
# child sex: 0 = girl, 1 = boy
ULF_ida_final$child_sex <- as.numeric(as.character(
factor(ULF_ida_final$child_sex,
levels = c("girl", "boy"),
labels = c(0, 1))
))
```
The variable for "antenatal maternal anaemia status" remained a categorical variable even after running the code above. Therefore, I recalled the dataset to ensure that this was numeric (as initially defined) for modelling:
```{r}
# recall dataset
# load dataset
ULF_ida_final <- read_excel("ULF_ida_final.xlsx")
# create New Variables for Total Brain Volumes across ROIs
# basal ganglia
ULF_ida_final$total_putamen <- ULF_ida_final$mm_hyp_left_putamen + ULF_ida_final$mm_hyp_right_putamen
ULF_ida_final$total_caudate <- ULF_ida_final$mm_hyp_left_caudate + ULF_ida_final$mm_hyp_right_caudate
# corpus callosum
ULF_ida_final$total_cc <- ULF_ida_final$mm_hyp_posterior_callosum + ULF_ida_final$mm_hyp_mid_posterior_callosum + ULF_ida_final$mm_hyp_central_callosum + ULF_ida_final$mm_hyp_mid_anterior_callosum + ULF_ida_final$mm_hyp_anterior_callosum
# convert days to months
# average Number of days in a month (accounting for leap years): 30.44
ULF_ida_final <- ULF_ida_final %>%
mutate(scan_age_months = hyp_age / 30.44)
```
**Correlation Matrices**
```{r}
# correlation matrix for corpus callosum
# filter dataset
filtered_data <- ULF_ida_final[!is.na(ULF_ida_final$maternal_preg_anaemia_status), ]
# select relevant columns
vars <- c("child_sex", "scan_age_months", "income_household_en", "maternal_hiv", "maternal_preg_anaemia_status", "total_cc")
data_selected <- filtered_data[, vars]
# convert categorical variables to numeric (for correlation)
data_selected$child_sex <- as.numeric(as.factor(data_selected$child_sex))
data_selected$maternal_hiv <- as.numeric(as.factor(data_selected$maternal_hiv))
data_selected$maternal_preg_anaemia_status <- as.numeric(as.factor(data_selected$maternal_preg_anaemia_status))
# compute correlation matrix with significance
rcorr_result <- Hmisc::rcorr(as.matrix(data_selected))
# rcorr_result$r contains correlation coefficients
print("Correlation matrix:")
print(rcorr_result$r)
# rcorr_result$P contains p-values for testing the null hypothesis of no correlation
print("P-value matrix:")
print(rcorr_result$P)
```
```{r}
# correlation matrix for caudate nucleus
# filter dataset
filtered_data <- ULF_ida_final[!is.na(ULF_ida_final$maternal_preg_anaemia_status), ]
# select relevant columns
vars <- c("child_sex", "scan_age_months", "income_household_en", "maternal_hiv", "maternal_preg_anaemia_status", "total_caudate")
data_selected <- filtered_data[, vars]
# convert categorical variables to numeric (for correlation)
data_selected$child_sex <- as.numeric(as.factor(data_selected$child_sex))
data_selected$maternal_hiv <- as.numeric(as.factor(data_selected$maternal_hiv))
data_selected$maternal_preg_anaemia_status <- as.numeric(as.factor(data_selected$maternal_preg_anaemia_status))
# compute correlation matrix with significance
rcorr_result <- Hmisc::rcorr(as.matrix(data_selected))
# rcorr_result$r contains correlation coefficients
print("Correlation matrix:")
print(rcorr_result$r)
# rcorr_result$P contains p-values for testing the null hypothesis of no correlation
print("P-value matrix:")
print(rcorr_result$P)
```
```{r}
# correlation matrix for putamen
# filter dataset
filtered_data <- ULF_ida_final[!is.na(ULF_ida_final$maternal_preg_anaemia_status), ]
# select relevant columns
vars <- c("child_sex", "scan_age_months", "income_household_en", "maternal_hiv", "maternal_preg_anaemia_status", "total_putamen")
data_selected <- filtered_data[, vars]
# convert categorical variables to numeric (for correlation)
data_selected$child_sex <- as.numeric(as.factor(data_selected$child_sex))
data_selected$maternal_hiv <- as.numeric(as.factor(data_selected$maternal_hiv))
data_selected$maternal_preg_anaemia_status <- as.numeric(as.factor(data_selected$maternal_preg_anaemia_status))
# compute correlation matrix with significance
rcorr_result <- Hmisc::rcorr(as.matrix(data_selected))
# rcorr_result$r contains correlation coefficients
print("Correlation matrix:")
print(rcorr_result$r)
# rcorr_result$P contains p-values for testing the null hypothesis of no correlation
print("P-value matrix:")
print(rcorr_result$P)
```
```{r}
# correlation matrix for icv
# filter dataset
filtered_data <- ULF_ida_final[!is.na(ULF_ida_final$maternal_preg_anaemia_status), ]
# select relevant columns
vars <- c("child_sex", "scan_age_months", "income_household_en", "maternal_hiv", "maternal_preg_anaemia_status", "mm_hyp_icv")
data_selected <- filtered_data[, vars]
# convert categorical variables to numeric (for correlation)
data_selected$child_sex <- as.numeric(as.factor(data_selected$child_sex))
data_selected$maternal_hiv <- as.numeric(as.factor(data_selected$maternal_hiv))
data_selected$maternal_preg_anaemia_status <- as.numeric(as.factor(data_selected$maternal_preg_anaemia_status))
# compute correlation matrix with significance
rcorr_result <- Hmisc::rcorr(as.matrix(data_selected))
# rcorr_result$r contains correlation coefficients
print("Correlation matrix:")
print(rcorr_result$r)
# rcorr_result$P contains p-values for testing the null hypothesis of no correlation
print("P-value matrix:")
print(rcorr_result$P)
```
**Multicollinearity**
```{r}
library(dplyr)
correlation_test <- cor.test(
ULF_ida_final$scan_age_months[!is.na(ULF_ida_final$maternal_preg_anaemia_status)],
ULF_ida_final$mm_hyp_icv[!is.na(ULF_ida_final$maternal_preg_anaemia_status)])
print(correlation_test)
```
Decision to exclude ICV from the models for regional brain volumes due to high levels of multicollinearity between child age at scan. Only child age at scan was included in the models below.
**Linear Mixed Effects (LME) Models**
```{r}
# install packages if needed
#install.packages("lme4")
library(lme4)
#install.packages("lmerTest")
library(lmerTest)
#install.packages("parameters")
library(parameters)
#install.packages("emmeans")
library(emmeans)
```
Based on the exploratory analyses (LOESS curves; plotted data), absolute age was used for models with an observed linear relationship. Log (age) was used for models with an observed non-linear relationship.
- ICV, Putamen, Caudate Nucleus, Corpus Callosm: Log (age) - Growth trajectory characterised by rapid growth and plateauing on plot
For the ULF sample, interactions (child age at scan \* antenatal maternal anaemia status) were explored as there was sufficient power. If the interaction was not significant, the model was re-run without an interaction effect and only main effects were interpreted.
**LME Model variables**
- Ensure that all categorical variables (anaemia status, child sex, maternal HIV status etc) are factors
- Create variable for the logarithmic transformation of child age in months
```{r}
# recall dataset so that all input variables are categorical
# load dataset
ULF_ida_final <- read_excel("ULF_ida_final.xlsx")
# create New Variables for Total Brain Volumes across ROIs
# basal ganglia
ULF_ida_final$total_putamen <- ULF_ida_final$mm_hyp_left_putamen + ULF_ida_final$mm_hyp_right_putamen
ULF_ida_final$total_caudate <- ULF_ida_final$mm_hyp_left_caudate + ULF_ida_final$mm_hyp_right_caudate
# corpus callosum
ULF_ida_final$total_cc <- ULF_ida_final$mm_hyp_posterior_callosum + ULF_ida_final$mm_hyp_mid_posterior_callosum + ULF_ida_final$mm_hyp_central_callosum + ULF_ida_final$mm_hyp_mid_anterior_callosum + ULF_ida_final$mm_hyp_anterior_callosum
# create age variable
# convert days to months
# average Number of days in a month (accounting for leap years): 30.44
ULF_ida_final <- ULF_ida_final %>%
mutate(scan_age_months = hyp_age / 30.44)
```
```{r}
# convert categorical variables form numeric to factor
# maternal anaemia status
ULF_ida_final$maternal_preg_anaemia_status <- factor(
ULF_ida_final$maternal_preg_anaemia_status,
levels = c(0, 1),
labels = c("no_mat_anaemia", "mat_anaemia")
)
# maternal anaemia severity
ULF_ida_final$maternal_preg_anaemia_severity <- factor(
ULF_ida_final$maternal_preg_anaemia_severity,
levels = c(1, 2, 3),
labels = c("mild", "moderate", "severe")
)
# child status
ULF_ida_final$cor_child_anaemia <- factor(
ULF_ida_final$cor_child_anaemia,
levels = c(0, 1),
labels = c("no_child_anaemia", "child_anaemia")
)
# maternal ID status
ULF_ida_final$maternal_preg_adj_brinda_id <- factor(
ULF_ida_final$maternal_preg_adj_brinda_id,
levels = c(0, 1),
labels = c("no_mat_id", "mat_id")
)
# child ID status
ULF_ida_final$cor_child_id <- factor(
ULF_ida_final$cor_child_id,
levels = c(0, 1),
labels = c("no_child_id", "child_id")
)
# maternal trimester hb
ULF_ida_final$maternal_trimester_hb <- factor(
ULF_ida_final$maternal_trimester_hb,
levels = c(1, 2, 3),
labels = c("first", "second", "third")
)
# child sex
ULF_ida_final$child_sex <- factor(
ULF_ida_final$child_sex,
levels = c(0, 1),
labels = c("girl", "boy")
)
# maternal pae
ULF_ida_final$maternal_pae <- factor(
ULF_ida_final$maternal_pae,
levels = c(0, 1),
labels = c("no_pae", "pae")
)
# maternal pte
ULF_ida_final$maternal_pte <- factor(
ULF_ida_final$maternal_pte,
levels = c(0, 1),
labels = c("no_pte", "pte")
)
# maternal hiv
ULF_ida_final$maternal_hiv <- factor(
ULF_ida_final$maternal_hiv,
levels = c(0, 1),
labels = c("no_hiv", "hiv")
)
# maternal dep
ULF_ida_final$maternal_dep <- factor(
ULF_ida_final$maternal_dep,
levels = c(0, 1),
labels = c("no_dep", "dep")
)
# maternal employment
ULF_ida_final$mom_employment_en_dichotomous <- factor(
ULF_ida_final$mom_employment_en_dichotomous,
levels = c(0, 1),
labels = c("unemployed", "employed")
)
# household income
ULF_ida_final$income_household_en <- factor(
ULF_ida_final$income_household_en,
levels = c(1, 2, 3, 4),
labels = c("Less than R1000 per month", "R1000-R5000 per monthy", " R5000-R10 000 per month", "More than R10 000 per month" )
)
# maternal education
ULF_ida_final$mom_edu_en <- factor(
ULF_ida_final$mom_edu_en,
levels = c(2, 3, 4, 5, 6),
labels = c("primary", "some_secondary", "complete_secondary", "some_tertiary", "completed_tertiary" )
)
#baby hiv
ULF_ida_final$baby_hiv <- factor(
ULF_ida_final$baby_hiv,
levels = c(0, 1),
labels = c("no_child_hiv", "child_hiv")
)
```
```{r}
# create a log(age) variable for use in models of brain regions with non-linear growth
ULF_ida_final$log_age <- log(ULF_ida_final$scan_age_months)
```
*LME: Corpus Callosum*
```{r}
# lme for corpus callosum
# including interaction between maternal anaemia and age
# log (age): non-linear growth
#model summary
ulf_model_cc_interaction <- lmer(total_cc ~ maternal_preg_anaemia_status * log_age +
child_sex +
(1 | subject_id),
data = ULF_ida_final)
summary(ulf_model_cc_interaction)
# beta values
standardize_parameters(ulf_model_cc_interaction)
```
| |
|-----|
| |
```{r}
# emms for corpus callosum volume by antenatal maternal anaemia status
# based on final adjusted model with interaction between maternal anaemia and age
# by timepoint for significant interaction
library(emmeans)
library(dplyr)
# step 1: Define raw ages and their logs for EMMs
raw_ages <- c(3, 6, 12, 18, 24)
log_ages <- log(raw_ages)
# step 2: Get estimated marginal means at these log ages
emm <- emmeans(
ulf_model_cc_interaction,
specs = ~ maternal_preg_anaemia_status | log_age,
at = list(log_age = log_ages)
)
# step 3: Convert to data frame and add back-transformed age for interpretability
emm_cc <- as.data.frame(emm)
emm_cc$age_months <- round(exp(emm_cc$log_age), 0)
print(emm_cc)
# step 4: Run pairwise contrasts (differences between anaemia groups) at each age
contrast_cc <- contrast(
emm,
method = "revpairwise",
adjust = "holm",
infer = TRUE
) %>%
as.data.frame()
# step 5: Add back-transformed age for contrasts
contrast_cc$age_months <- round(exp(contrast_cc$log_age), 0)
print(contrast_cc)
```
*LME: ICV*
```{r}
# lme for icv
# including nteraction between maternal anaemia and age
# log (age): non-linear growth
# model summary
ulf_model_icv_interaction <- lmer(mm_hyp_icv ~ maternal_preg_anaemia_status * log_age + child_sex + (1 | subject_id),
data = ULF_ida_final)
summary(ulf_model_icv_interaction)
# beta values
standardize_parameters(ulf_model_icv_interaction)
```
Interaction was not significant, so re-ran the mdoel without it and interpreted main effects:
```{r}
# lme for icv
# excluding interaction between maternal anaemia and age
# log (age): non-linear growth
# model summary
ulf_model_icv_no_interaction <- lmer(mm_hyp_icv ~ maternal_preg_anaemia_status + log_age + child_sex + (1 | subject_id),
data = ULF_ida_final)
summary(ulf_model_icv_no_interaction)
# beta values
standardize_parameters(ulf_model_icv_no_interaction)
```
```{r}
# emms for icv volume by antenatal maternal anaemia status
# based on final adjusted model with no interaction
library(emmeans)
emm_icv <- emmeans(ulf_model_icv_no_interaction, ~ maternal_preg_anaemia_status)
# ----- EMM table with 2 decimals -----
emm_icv_df <- as.data.frame(emm_icv)
emm_icv_df$emmean <- format(round(emm_icv_df$emmean, 2), nsmall = 2)
emm_icv_df$SE <- format(round(emm_icv_df$SE, 2), nsmall = 2)
emm_icv_df$lower.CL <- format(round(emm_icv_df$lower.CL, 2), nsmall = 2)
emm_icv_df$upper.CL <- format(round(emm_icv_df$upper.CL, 2), nsmall = 2)
print(emm_icv_df)
# ----- contrast results -----
contrast_results <- contrast(emm_icv, method = "revpairwise", adjust = "none")
contrast_df <- as.data.frame(summary(contrast_results, infer = TRUE))
# force 2 decimals for estimate
contrast_df$estimate <- format(round(contrast_df$estimate, 2), nsmall = 2)
print(contrast_df)
```
*LME: Caudate Nucleus*
```{r}
# lme for caudate nucleus
# icluding interaction between maternal anaemia and age
# log (age): non-linear growth
# model summary
ulf_model_caudate_interaction <- lmer(total_caudate ~ maternal_preg_anaemia_status * log_age + child_sex + (1 | subject_id),
data = ULF_ida_final)
summary(ulf_model_caudate_interaction)
# beta values
standardize_parameters(ulf_model_caudate_interaction)
```
```{r}
# emms for caudate volume by antenatal maternal anaemia status
# based on final adjusted model with interaction
# by timepoint for significant interaction
library(emmeans)
library(dplyr)
# step 1: Define raw ages and their logs for EMMs
raw_ages <- c(3, 6, 12, 18, 24)
log_ages <- log(raw_ages)
# step 2: Get estimated marginal means at these log ages
emm <- emmeans(
ulf_model_caudate_interaction,
specs = ~ maternal_preg_anaemia_status | log_age,
at = list(log_age = log_ages)
)
# step 3: Convert to data frame and add back-transformed age for interpretability
emm_caudate <- as.data.frame(emm)
emm_caudate$age_months <- round(exp(emm_caudate$log_age), 0)
print(emm_caudate)
# step 4: Run pairwise contrasts (differences between anaemia groups) at each age
contrast_caudate <- contrast(emm, method = "revpairwise", adjust = "holm", infer = TRUE) %>% as.data.frame()
# step 5: Add back-transformed age for contrasts (will be in 'log_age' column)
contrast_caudate$age_months <- round(exp(contrast_caudate$log_age), 0)
print(contrast_caudate)
```
*LME: Putamen*
```{r}
# lme for putamen
# including interaction between maternal anaemia and age
# log (age): non-linear growth
# model summary
ulf_model_putamen_interaction <- lmer(total_putamen ~ maternal_preg_anaemia_status * log_age + child_sex + (1 | subject_id),
data = ULF_ida_final)
summary(ulf_model_putamen_interaction)
# beta values
standardize_parameters(ulf_model_putamen_interaction)
```
Interaction was not significant, so re-ran the model without it and interpreted main effects:
```{r}
# lme for putamen
# excluding interaction between maternal anaemia and age
# model summary
ulf_model_putamen_no_interaction <- lmer(total_putamen ~ maternal_preg_anaemia_status + log_age + child_sex + (1 | subject_id),
data = ULF_ida_final)
summary(ulf_model_putamen_no_interaction)
# beta values
standardize_parameters(ulf_model_putamen_no_interaction)
```
```{r}
# emms for putamen volume by antenatal maternal anaemia status
# based on final adjusted model with no interaction
library(emmeans)
emm_putamen <- emmeans(ulf_model_putamen_no_interaction, ~ maternal_preg_anaemia_status)
# ----- emm table with exactly 2 decimals -----
emm_putamen_df <- as.data.frame(emm_putamen)
emm_putamen_df$emmean <- format(round(emm_putamen_df$emmean, 2), nsmall = 2)
emm_putamen_df$SE <- format(round(emm_putamen_df$SE, 2), nsmall = 2)
emm_putamen_df$lower.CL <- format(round(emm_putamen_df$lower.CL, 2), nsmall = 2)
emm_putamen_df$upper.CL <- format(round(emm_putamen_df$upper.CL, 2), nsmall = 2)
print(emm_putamen_df)
# ----- contrast results -----
contrast_results_putamen <- contrast(emm_putamen, method = "revpairwise", adjust = "none")
contrast_df_putamen <- as.data.frame(summary(contrast_results_putamen, infer = TRUE))
# force 2 decimals for estimate
contrast_df_putamen$estimate <- format(round(contrast_df_putamen$estimate, 2), nsmall = 2)
print(contrast_df_putamen)
```
**Sensitivity Analyses**
Based on the final models for each ROI (established in primary analyses): With the inclusion of maternal HIV (potential confounder)
- ICV: Final model was one without an interaction
- Putamen: FInal model was without an interaction
- Corpus Callosum: Final model was with an interaction
- Caudate Nucleus: Final model was with an interaction
```{r}
# sensitivity analysis for ICV
# final model: excluding interaction between maternal anaemia and age
# including maternal hiv
# model summary
ulf_model_icv_no_interaction_sens <- lmer(mm_hyp_icv ~ maternal_preg_anaemia_status + log_age + child_sex + maternal_hiv + (1 | subject_id),
data = ULF_ida_final)
summary(ulf_model_icv_no_interaction_sens)
# beta values
standardize_parameters(ulf_model_icv_no_interaction_sens)
```
```{r}
# sensitivity analysis for corpus callosum
# final model: including interaction between maternal anaemia and age
# including maternal hiv
# model summary
ulf_model_cc_interaction_sens <- lmer(total_cc ~ maternal_preg_anaemia_status * scan_age_months + child_sex + maternal_hiv + (1 | subject_id), data = ULF_ida_final)
summary(ulf_model_cc_interaction_sens)
# beta values
standardize_parameters(ulf_model_cc_interaction_sens)
```
```{r}
# sensitivity analysis for putamen
# final model: excluding interaction between maternal anaemia and age
# including maternal hiv
# model summary
ulf_model_putamen_no_interaction_sens <- lmer(total_putamen ~ maternal_preg_anaemia_status + log_age + child_sex + maternal_hiv + (1 | subject_id),
data = ULF_ida_final)
summary(ulf_model_putamen_no_interaction_sens)
# beta values
standardize_parameters(ulf_model_putamen_no_interaction_sens)
```
```{r}
# sensitivity analysis for caudate nucleus
# final model: including interaction between maternal anaemia and age
# including maternal hiv
# model summary
ulf_model_caudate_interaction_sens <- lmer(total_caudate ~ maternal_preg_anaemia_status * log_age + child_sex + maternal_hiv + (1 | subject_id),
data = ULF_ida_final)
summary(ulf_model_caudate_interaction_sens)
# beta values
standardize_parameters(ulf_model_caudate_interaction_sens)
```
------------------------------------------------------------------------
**Secondary Analysis: Child Anaemia (ULF Sample)**
------------------------------------------------------------------------
**Sample Size: For infants with Child Anemia Data**
Note: Sample sizes were calculated with 1) the inclusion of repeated measures (multiple scans across study timepoints), and for 2) unique subject IDs excluding repeated measures.
```{r}
# total sample size
# sample iwth child anaemia data
# repeated measures included
ULF_ida_final %>%
filter(!is.na(cor_child_anaemia)) %>% # only include PIDs with child anaemia data
summarise(n = n())
```
```{r}
# sample with child anaemia data
# repeated measures excluded
# unique subject IDs: based on true prevalence
ULF_ida_final %>%
filter(!is.na(cor_child_anaemia)) %>%
distinct(subject_id, .keep_all = TRUE) %>% #only include PIDs with child anaemia data
summarise(n = n())
length(unique(ULF_ida_final$subject_id)) # includes chidlren with no anaemia data
```
**Child Anaemia Variables**
An overall classification for child anaemia status was determined based on the diagnosis of anaemia at least once across study visits. This was used for the assessment of group differences in sample characteristics between infants who had ever been anaemic versus infants who had never been anaemic between 3 and 24 months of age. Minimum child haemoglobin and corresponding age across study visits were also determined for each infant and reported on across both groups. For modelling with the repeated measures subsample (multiple scans per child in some instances), child anaemia status was assessed based on the most severe presentation of anaemia as indicated by the minimum haemoglobin measure observed between birth and the time of scan.
Each infant has multiple haemoglobin measurements from across study timepoints.
- "Overall Anaemia" variable: For the purpose of assessing group differences in sample characteristics based on unqiue subject IDs (excluding repeated measures), we created a binary variable for child anaemia status based on being anaemic at least once between 3 and 24 months. i.e. "overall anaemia" = "anaemic at any timepoint"
- Child anaemia status in repeated measures models (LMEs): Every scan has a corresponding child anaemia status variable (cor_child_anaemia) based on the minimum Hb measurement (cor_child_hb) between birth and the time of scan and the corresponding age (cor_child_hb_age) at the time of this measurement. i.e. In a repeated measures sample:
- 3-month scan: anaemia status based on Hb at 3 months
- 6-month scan: anaemia status based on min Hb between birth and time of scan (i.e. between measures at 3 and 6 months)
- 12-month scan: anaemia status based on min Hb between birth and time of scan (i.e. between measures at 3, 6, and 12 months)
- 18-month scan: anaemia status based on min Hb between birth and time of scan (i.e. between measures at 3, 6, 12, and 18 months)
- 24-month scan: anaemia status based on min Hb between birth and time of scan (i.e. between measures at 3, 6, 12, 18, and 24 months)
- Minimum haemoglobin measurement: This was computed to compare Hb across groups when assessing sample characteristics based on unique subjects (i.e. excluding repeated measures). We identified the minimum Hb measurement (and corresponding age) across study timepoints - i.e. across multiple cor_child_hb measures for children with multiple scans in a repeated measures sample.
**Creating New Binary Variable for Overall Anaemia Status**
```{r}
# overall child anaemia status variable
# step 1: Create new dataframe with only child anaemia data. Select only anaemia variables interested in
# step 2: Pivot to wide format so each child ID appears once only (see all timepoints for each child)
dem_child <- ULF_ida_final %>% #create new dataframe
filter(!is.na(cor_child_anaemia)) %>% #filter so excluding children with no child anaemia data
select(subject_id, timepoint, cor_child_hb, cor_child_anaemia) %>% #select variables interested in
pivot_wider(id_cols = subject_id, names_from = timepoint, values_from = c(cor_child_hb,cor_child_anaemia)) #convert to wide format so can see anaemia data fro all timepoints
```
```{r}
# overall child anaemia status variable
# step 3: create overall anaemia variable - i.e. "if anaemic at any timepoint"
dem_child <- dem_child %>%
mutate(overall_child_anaemia = if_else(if_any(c(cor_child_anaemia_3M, cor_child_anaemia_6M, cor_child_anaemia_12M, cor_child_anaemia_18M, cor_child_anaemia_24M), ~ !is.na(.x) & .x == "child_anaemia"), "overall_child_anaemia", "no_overall_child_anaemia")) %>%
select(subject_id, overall_child_anaemia) #select and save the overall anaemia status
# step 4: Merge the overall_child_anaemia variable/column back into long format of the original dataframe - i.e. "ULF_ida_final"
ULF_ida_final <- left_join(ULF_ida_final, dem_child, by = 'subject_id')
```
```{r}
# overall child anaemia status variable
# sanity check that have the correct IDs
ULF_ida_final %>%
filter(!is.na(overall_child_anaemia)) %>%
distinct(subject_id, .keep_all = TRUE) %>% #only include PIDs with child anaemia data
summarise(n = n())
length(unique(ULF_ida_final$subject_id)) # inlcudes chidlren with no anaemia data
```
**Creating Variable for Minimum Hb Across Study Timepoints**
```{r}
# minimum child Hb variable
# step 1: Create new dataframe
# step 2: Pivot to wide format
dem_child_1 <- ULF_ida_final %>% #create new dataframe
filter(!is.na(cor_child_anaemia)) %>% #filter so excluding children with no child anaemia data
select(subject_id, timepoint, cor_child_hb, cor_child_hb_age, cor_child_anaemia) %>% #select variables interested in
pivot_wider(id_cols = subject_id, names_from = timepoint, values_from = c(cor_child_hb,cor_child_hb_age, cor_child_anaemia)) #convert to wide format so can see anaemia data for all timepoints
```
```{r}
# step 3: Create min hb variable - min hb from across timepoints
library(dplyr)
dem_child_1 <- dem_child_1 %>%
# create a new variable that represents the minimum child Hb across timepoints
mutate(
min_child_hb = pmin(cor_child_hb_3M, cor_child_hb_6M, cor_child_hb_12M, cor_child_hb_18M, cor_child_hb_24M, na.rm = TRUE),
# create a new variable that represents the age at the time of the minimum Hb
min_hb_age = case_when(
min_child_hb == cor_child_hb_3M ~ cor_child_hb_age_3M,
min_child_hb == cor_child_hb_6M ~ cor_child_hb_age_6M,
min_child_hb == cor_child_hb_12M ~ cor_child_hb_age_12M,
min_child_hb == cor_child_hb_18M ~ cor_child_hb_age_18M,
min_child_hb == cor_child_hb_24M ~ cor_child_hb_age_24M
)
) %>%
# select subject_id, min_child_hb, and min_hb_age variables
select(subject_id, min_child_hb, min_hb_age)
# view the resulting dataframe
head(dem_child_1)
```
```{r}
# step 4: merge in the min Hb and cor age at min hb variable back into long format original dataframe "ULF_ida_final"
ULF_ida_final <- left_join(ULF_ida_final, dem_child_1, by = 'subject_id')
```
**Anaemia Prevalence**
```{r}
# child anaemia prevalence
# based on new variable for overall anaemia status
# unique subject IDs: based on true prevalence
ULF_ida_final %>%
filter(!is.na(overall_child_anaemia)) %>% # only include PIDs with child anaemia data
distinct(subject_id, .keep_all = TRUE) %>%
group_by(overall_child_anaemia) %>%
summarise(n = n()) %>%
mutate(percent = round(100 * n / sum(n), 2))
```
```{r}
# maternal anaemia prevalence
# for infants with child anaemia data
# unique subject IDs: based on true prevalence
ULF_ida_final %>%
filter(!is.na(overall_child_anaemia)) %>% # only include PIDs with child anaemia data
distinct(subject_id, .keep_all = TRUE) %>%
group_by(maternal_preg_anaemia_status) %>%
summarise(n = n()) %>%
mutate(percent = round(100 * n / sum(n), 2))
```
**Sample Characteristics and Group Differences: Child Anaemia Status**
Sample characteristics were assessed based on the number of unique subjects with child anaemia (haemoglobin) data, excluding repeated measures data for children with multiple scans. This approach was used to avoid skewing results with duplicated demographic and clinical observations for the same infants. The only exception to this approach was child age at scan which was considered for the full repeated measures subsample, with a proportion of children having multiple scans acquired at different ages.
An overall classification for child anaemia status was determined based on the diagnosis of anaemia at least once across study visits. This was used for the assessment of group differences in sample characteristics between infants who had ever been anaemic versus infants who had never been anaemic between 3 and 24 months of age. Minimum child haemoglobin and corresponding age across study visits were also determined for each infant and reported on across both groups. For modelling with the repeated measures subsample (multiple scans per child in some instances), child anaemia status was assessed based on the most severe presentation of anaemia as indicated by the minimum haemoglobin measure observed between birth and the time of scan.
- **Categorical Variables**
*Maternal Anaemia*
```{r}
# group differences in maternal anaemia by child anaemia status (using overall child anaemia variable created)
# unique subject IDs: based on true prevalence
# filter only for non-missing child anaemia status and keep unique subjects
filtered <- ULF_ida_final[!is.na(ULF_ida_final$overall_child_anaemia), ]
filtered <- filtered[!duplicated(filtered$subject_id), ]
# create contingency table: maternal anaemia vs child anaemia
tbl <- table(filtered$maternal_preg_anaemia_status, filtered$overall_child_anaemia)
# print table
cat("Contingency Table: Maternal anaemia status by Child Anaemia Status\n")
print(tbl)
# column percentages (percent within child anaemia status group)
cat("\nColumn Percentages (% within child anaemia status group):\n")
print(round(100 * prop.table(tbl, margin = 2), 2))
# choose appropriate test
chisq_result <- chisq.test(tbl)
if (any(chisq_result$expected < 5)) {
cat("\nFisher's Exact Test (due to small expected cell counts):\n")
print(fisher.test(tbl))
} else {
cat("\nChi-squared Test:\n")
print(chisq_result)
}
```
*Child Sex at Birth*
```{r}
# child sex at birth (full sample)
# unique subject IDs: Based on true prevalence
ULF_ida_final %>%
filter(!is.na(overall_child_anaemia)) %>% #only include PIDs with child anaemia data
distinct(subject_id, .keep_all = TRUE) %>%
group_by(child_sex) %>%
summarise(n = n()) %>%
mutate(percent = round(100 * n / sum(n), 2))
```
```{r}
# group differences in child sex by child anaemia status
# chi-squared: child sex
# unique subject IDs: based on true prevalence
# 1. filter for non-missing child anaemia status and keep unique subjects
filtered <- ULF_ida_final[!is.na(ULF_ida_final$overall_child_anaemia), ]
filtered <- filtered[!duplicated(filtered$subject_id), ]
# 2. create contingency table: child sex vs child anaemia
tbl <- table(filtered$child_sex, filtered$overall_child_anaemia)
# 3. print contingency table
cat("Contingency Table: Child Sex by Child Anaemia Status\n")
print(tbl)
# 4. column percentages (% within each anaemia status group)
cat("\nColumn Percentages (% of each anaemia status group):\n")
print(round(100 * prop.table(tbl, margin = 2), 2))
# 5. run appropriate test
expected <- chisq.test(tbl)$expected
if (any(expected < 5)) {
cat("\nFisher's Exact Test (due to small expected cell counts):\n")
print(fisher.test(tbl))
} else {
cat("\nChi-squared Test:\n")
print(chisq.test(tbl))
}
```
*Prenatal Alcohol Exposure (PAE)*
```{r}
# pae (full sample)
# unique subject IDs: based on true prevalence
ULF_ida_final %>%
filter(!is.na(overall_child_anaemia)) %>% # only include PIDs with child anaemia data
distinct(subject_id, .keep_all = TRUE) %>%
group_by(maternal_pae) %>%
summarise(n = n()) %>%
mutate(percent = round(100 * n / sum(n), 2))
```
```{r}
# group differences in pae by child anaemia status
# chi-squared: pae
# unique subject IDs: based on true prevalence
# 1. filter for non-missing child anaemia status and keep unique subjects
filtered <- ULF_ida_final[!is.na(ULF_ida_final$overall_child_anaemia), ]
filtered <- filtered[!duplicated(filtered$subject_id), ]
# 2. create contingency table: maternal pae vs child anaemia
tbl <- table(filtered$maternal_pae, filtered$overall_child_anaemia)
# 3. print contingency table
cat("Contingency Table: Maternal PAE by Child Anaemia Status\n")
print(tbl)
# 4. column percentages (% within each child anaemia group)
cat("\nColumn Percentages (% of each anaemia status group):\n")
print(round(100 * prop.table(tbl, margin = 2), 2))
# 5. run appropriate test
expected <- chisq.test(tbl)$expected
if (any(expected < 5)) {
cat("\nFisher's Exact Test (due to small expected cell counts):\n")
print(fisher.test(tbl))
} else {
cat("\nChi-squared Test:\n")
print(chisq.test(tbl))
}
```
*Prenatal Tobacco Exposure (PTE)*
```{r}
# pte (full sample)
# excluding repeated measures: unique id
ULF_ida_final %>%
filter(!is.na(overall_child_anaemia)) %>% #only include PIDs with child anaemia data
distinct(subject_id, .keep_all = TRUE) %>%
group_by(maternal_pte) %>%
summarise(n = n()) %>%
mutate(percent = round(100 * n / sum(n), 2))
```
```{r}
# group differences in pte by child anaemia status
# chi-squared: pte
# unique subject IDs: based on true prevalence
# 1. filter for non-missing child anaemia status and keep unique subjects
filtered <- ULF_ida_final[!is.na(ULF_ida_final$overall_child_anaemia), ]
filtered <- filtered[!duplicated(filtered$subject_id), ]
# 2. create contingency table: maternal PTE vs child anaemia
tbl <- table(filtered$maternal_pte, filtered$overall_child_anaemia)
# 3.print contingency table
cat("Contingency Table: Maternal PTE by Child Anaemia Status\n")
print(tbl)
# 4. column percentages (% within each child anaemia group)
cat("\nColumn Percentages (% of each anaemia status group):\n")
print(round(100 * prop.table(tbl, margin = 2), 2))
# 5. run appropriate test
expected <- chisq.test(tbl)$expected
if (any(expected < 5)) {
cat("\nFisher's Exact Test (due to small expected cell counts):\n")
print(fisher.test(tbl))
} else {
cat("\nChi-squared Test:\n")
print(chisq.test(tbl))
}
```
*Maternal HIV*
```{r}
# maternal hiv (full sample)
# unique subject IDs: based on true prevalence
ULF_ida_final %>%
filter(!is.na(overall_child_anaemia)) %>% # only include PIDs with child anaemia data
distinct(subject_id, .keep_all = TRUE) %>%
group_by(maternal_hiv) %>%
summarise(n = n()) %>%
mutate(percent = round(100 * n / sum(n), 2))
```
```{r}
# group differences in maternal hiv by child anaemia status
# chi-squared: maternal hiv
# unique subject IDs: based on true prevalence
# 1. filter for non-missing child anaemia status and keep unique subjects
filtered <- ULF_ida_final[!is.na(ULF_ida_final$overall_child_anaemia), ]
filtered <- filtered[!duplicated(filtered$subject_id), ]
# 2. create contingency table: maternal HIV vs child anaemia
tbl <- table(filtered$maternal_hiv, filtered$overall_child_anaemia)
# 3. print contingency table
cat("Contingency Table: Maternal HIV by Child Anaemia Status\n")
print(tbl)
# 4. column percentages (% within each child anaemia group)
cat("\nColumn Percentages (% of each anaemia status group):\n")
print(round(100 * prop.table(tbl, margin = 2), 2))
# 5. run appropriate test
expected <- chisq.test(tbl)$expected
if (any(expected < 5)) {
cat("\nFisher's Exact Test (due to small expected cell counts):\n")
print(fisher.test(tbl))
} else {
cat("\nChi-squared Test:\n")
print(chisq.test(tbl))
}
```
*Maternal Depression*
```{r}
# maternal depression (full sample)
# unique subject IDs: based on true prevalence
ULF_ida_final %>%
filter(!is.na(overall_child_anaemia)) %>% # only include PIDs with child anaemia data
distinct(subject_id, .keep_all = TRUE) %>%
group_by(maternal_dep) %>%
summarise(n = n()) %>%
mutate(percent = round(100 * n / sum(n), 2))
```
```{r}
# group differences in maternal depression by child anaemia status
# chi-squared: maternal depression
# unique subject IDs: based on true prevalence
# 1. filter for non-missing child anaemia status and keep unique subjects
filtered <- ULF_ida_final[!is.na(ULF_ida_final$overall_child_anaemia), ]
filtered <- filtered[!duplicated(filtered$subject_id), ]
# 2. create contingency table: maternal DEP vs child anaemia
tbl <- table(filtered$maternal_dep, filtered$overall_child_anaemia)
# 3. print contingency table
cat("Contingency Table: Maternal DEP by Child Anaemia Status\n")
print(tbl)
# 4. column percentages (% within each child anaemia group)
cat("\nColumn Percentages (% of each anaemia status group):\n")
print(round(100 * prop.table(tbl, margin = 2), 2))
# 5. run appropriate test
expected <- chisq.test(tbl)$expected
if (any(expected < 5)) {
cat("\nFisher's Exact Test (due to small expected cell counts):\n")
print(fisher.test(tbl))
} else {
cat("\nChi-squared Test:\n")
print(chisq.test(tbl))
}
```
*Maternal Employment*
```{r}
# maternal employment (full sample)
# unique subject IDs: based on true prevalence
ULF_ida_final %>%
filter(!is.na(overall_child_anaemia)) %>% #only include PIDs with child anaemia data
distinct(subject_id, .keep_all = TRUE) %>%
group_by(mom_employment_en_dichotomous) %>%
summarise(n = n()) %>%
mutate(percent = round(100 * n / sum(n), 2))
```
```{r}
# group differences in maternal employment by child anaemia status
# chi-squared: maternal employment
# unique subject IDs: based on true prevalence
# 1. filter for non-missing child anaemia status and keep unique subjects
filtered <- ULF_ida_final[!is.na(ULF_ida_final$overall_child_anaemia), ]
filtered <- filtered[!duplicated(filtered$subject_id), ]
# 2. create contingency table: maternal employment vs child anaemia
tbl <- table(filtered$mom_employment_en_dichotomous, filtered$overall_child_anaemia)
# 3. print contingency table
cat("Contingency Table: Maternal Employment by Child Anaemia Status\n")
print(tbl)
# 4. column percentages (% within each child anaemia group)
cat("\nColumn Percentages (% of each anaemia status group):\n")
print(round(100 * prop.table(tbl, margin = 2), 2))
# 5. run appropriate test
expected <- chisq.test(tbl)$expected
if (any(expected < 5)) {
cat("\nFisher's Exact Test (due to small expected cell counts):\n")
print(fisher.test(tbl))
} else {
cat("\nChi-squared Test:\n")
print(chisq.test(tbl))
}
```
*Maternal Education*
```{r}
# maternal education (full sample)
# unique subject IDs: based on true prevalence
ULF_ida_final %>%
filter(!is.na(overall_child_anaemia)) %>%
distinct(subject_id, .keep_all = TRUE) %>%
group_by(mom_edu_en) %>%
summarise(n = n()) %>%
mutate(percent = round(100 * n / sum(n), 2))
```
```{r}
# group differences in maternal education by child anaemia status
# chi-squared: maternal education
# unique subject IDs: based on true prevalence
# 1. filter for non-missing child anaemia status and keep unique subjects
filtered <- ULF_ida_final[!is.na(ULF_ida_final$overall_child_anaemia), ]
filtered <- filtered[!duplicated(filtered$subject_id), ]
# 2. create contingency table: maternal education vs child anaemia
tbl <- table(filtered$mom_edu_en, filtered$overall_child_anaemia)
# 3. print contingency table
cat("Contingency Table: Maternal Education by Child Anaemia Status\n")
print(tbl)
# 4. column percentages (% within each child anaemia group)
cat("\nColumn Percentages (% of each anaemia status group):\n")
print(round(100 * prop.table(tbl, margin = 2), 2))
# 5. run appropriate test
expected <- chisq.test(tbl)$expected
if (any(expected < 5)) {
cat("\nFisher's Exact Test (due to small expected cell counts):\n")
print(fisher.test(tbl))
} else {
cat("\nChi-squared Test:\n")
print(chisq.test(tbl))
}
```
*Household Income*
```{r}
# household income (full sample)
# unique subject IDs: based on true prevalence
ULF_ida_final %>%
filter(!is.na(overall_child_anaemia)) %>%
distinct(subject_id, .keep_all = TRUE) %>%
group_by(income_household_en) %>%
summarise(n = n()) %>%
mutate(percent = round(100 * n / sum(n), 2))
```
```{r}
# group differences in household income by child anaemia status
# chi-squared: household income
# unique subject IDs: based on true prevalence
# 1. filter for non-missing child anaemia status and keep unique subjects
filtered <- ULF_ida_final[!is.na(ULF_ida_final$overall_child_anaemia), ]
filtered <- filtered[!duplicated(filtered$subject_id), ]
# 2. create contingency table: household income vs child anaemia
tbl <- table(filtered$income_household_en, filtered$overall_child_anaemia)
# 3. print contingency table
cat("Contingency Table: Household Income by Child Anaemia Status\n")
print(tbl)
# 4. column percentages (% within each child anaemia group)
cat("\nColumn Percentages (% of each anaemia status group):\n")
print(round(100 * prop.table(tbl, margin = 2), 2))
# 5. run appropriate test
expected <- chisq.test(tbl)$expected
if (any(expected < 5)) {
cat("\nFisher's Exact Test (due to small expected cell counts):\n")
print(fisher.test(tbl))
} else {
cat("\nChi-squared Test:\n")
print(chisq.test(tbl))
}
```
*Infant HIV*
```{r}
# infant hiv (full sample)
# unique subject IDs: based on true prevalence
ULF_ida_final %>%
filter(!is.na(overall_child_anaemia)) %>%
distinct(subject_id, .keep_all = TRUE) %>%
group_by(baby_hiv) %>%
summarise(n = n()) %>%
mutate(percent = round(100 * n / sum(n), 2))
```
```{r}
# group differences in infant hiv by child anaemia status
# chi-squared: infant hiv
# unique subject IDs: based on true prevalence
# 1. filter for non-missing child anaemia status and keep unique subjects
filtered <- ULF_ida_final[!is.na(ULF_ida_final$overall_child_anaemia), ]
filtered <- filtered[!duplicated(filtered$subject_id), ]
# 2. create contingency table: infant HIV vs child anaemia
tbl <- table(filtered$baby_hiv, filtered$overall_child_anaemia)
# 3. print contingency table
cat("Contingency Table: Infant HIV by Child Anaemia Status\n")
print(tbl)
# 4. column percentages (% within each child anaemia group)
cat("\nColumn Percentages (% of each anaemia status group):\n")
print(round(100 * prop.table(tbl, margin = 2), 2))
# 5. run appropriate test
expected <- chisq.test(tbl)$expected
if (any(expected < 5)) {
cat("\nFisher's Exact Test (due to small expected cell counts):\n")
print(fisher.test(tbl))
} else {
cat("\nChi-squared Test:\n")
print(chisq.test(tbl))
}
```
- **Continuous Variables**
```{r}
#install packages if needed
#install.packages('car')
library (car)
```
*Minimum Haemoglobin Measure*
```{r}
# group differences in minimum Hb by child anaemia status
# ttest: min Hb
# unique subject IDs: based on true prevalence
library(dplyr)
library(car) # for Levene's Test
# create cleaned dataset once (avoids repeating code)
child_data <- ULF_ida_final %>%
filter(!is.na(overall_child_anaemia)) %>%
distinct(subject_id, .keep_all = TRUE)
# summary statistics: min child haemoglobin by child anaemia status
summary_stats <- child_data %>%
group_by(overall_child_anaemia) %>%
summarise(
n = n(),
mean_hb = format(round(mean(min_child_hb, na.rm = TRUE), 2), nsmall = 2),
sd_hb = format(round(sd(min_child_hb, na.rm = TRUE), 2), nsmall = 2),
min_hb = format(round(min(min_child_hb, na.rm = TRUE), 2), nsmall = 2),
max_hb = format(round(max(min_child_hb, na.rm = TRUE), 2), nsmall = 2),
.groups = "drop"
)
print(summary_stats)
# levene’s test for homogeneity of variances
levene_result <- leveneTest(min_child_hb ~ overall_child_anaemia, data = child_data)
print(levene_result)
# extract Levene p-value
levene_p <- levene_result$`Pr(>F)`[1]
# run appropriate t-test (full precision output)
if (levene_p < 0.05) {
cat("\nLevene's test is significant — using Welch t-test (unequal variances):\n")
t_result <- t.test(
min_child_hb ~ overall_child_anaemia,
data = child_data,
var.equal = FALSE
)
} else {
cat("\nLevene's test is not significant — using Student's t-test (equal variances):\n")
t_result <- t.test(
min_child_hb ~ overall_child_anaemia,
data = child_data,
var.equal = TRUE
)
}
# print t-test results with full precision
print(t_result)
```
*Child Age at Minimum Haemoglobin Measurement*
```{r}
# group differences in age at Hb measurement by child anaemia status
# ttest: age at Hb measurement
# unique subject IDs: based on true prevalence
library(dplyr)
library(car) # for Levene's Test
# prepare cleaned dataset once
child_data <- ULF_ida_final %>%
filter(!is.na(overall_child_anaemia)) %>%
distinct(subject_id, .keep_all = TRUE)
# summary statistics: child age at Hb measurement by anaemia status
summary_stats <- child_data %>%
group_by(overall_child_anaemia) %>%
summarise(
n = n(),
mean_age = format(round(mean(min_hb_age, na.rm = TRUE), 2), nsmall = 2),
sd_age = format(round(sd(min_hb_age, na.rm = TRUE), 2), nsmall = 2),
min_age = format(round(min(min_hb_age, na.rm = TRUE), 2), nsmall = 2),
max_age = format(round(max(min_hb_age, na.rm = TRUE), 2), nsmall = 2),
.groups = "drop"
)
print(summary_stats)
# levene’s test for homogeneity of variances
levene_result <- leveneTest(min_hb_age ~ overall_child_anaemia, data = child_data)
print(levene_result)
# extract Levene p-value
levene_p <- levene_result$`Pr(>F)`[1]
# run appropriate t-test (full precision output)
if (levene_p < 0.05) {
cat("\nLevene's test is significant — using Welch t-test (unequal variances):\n")
t_result <- t.test(
min_hb_age ~ overall_child_anaemia,
data = child_data,
var.equal = FALSE
)
} else {
cat("\nLevene's test is not significant — using Student's t-test (equal variances):\n")
t_result <- t.test(
min_hb_age ~ overall_child_anaemia,
data = child_data,
var.equal = TRUE
)
}
# Print t-test results in full precision
print(t_result)
```
*Maternal Age at Enrolment*
```{r}
# maternal age at enrolment for full group
# unique subject IDs: based on true prevalence
ULF_ida_final %>%
filter(!is.na(overall_child_anaemia)) %>% #only include PIDs with child anaemia data
distinct(subject_id, .keep_all = TRUE) %>%
summarise(
mean_mom_age = mean(mom_age_en, na.rm = TRUE),
sd_mom_age = sd(mom_age_en, na.rm = TRUE),
min_mom_age = min(mom_age_en, na.rm = TRUE),
max_mom_age = max(mom_age_en, na.rm = TRUE),
n = n()
)
```
```{r}
# group differences in maternal age at enrolment by child anaemia status
# ttest: maternal age at enrolment
# unique subject IDs: based on true prevalence
library(dplyr)
library(car) # for Levene's Test
# deduplicate and filter data once
child_data <- ULF_ida_final %>%
filter(!is.na(overall_child_anaemia)) %>%
distinct(subject_id, .keep_all = TRUE)
# summary statistics: min child Hb by anaemia status (2 decimals)
summary_stats <- child_data %>%
group_by(overall_child_anaemia) %>%
summarise(
n = n(),
mean_hb = format(round(mean(mom_age_en, na.rm = TRUE), 2), nsmall = 2),
sd_hb = format(round(sd(mom_age_en, na.rm = TRUE), 2), nsmall = 2),
min_hb = format(round(min(mom_age_en, na.rm = TRUE), 2), nsmall = 2),
max_hb = format(round(max(mom_age_en, na.rm = TRUE), 2), nsmall = 2),
.groups = "drop"
)
print(summary_stats)
# levene’s test for homogeneity of variances
levene_result <- leveneTest(mom_age_en ~ overall_child_anaemia, data = child_data)
print(levene_result)
# extract Levene p-value
levene_p <- levene_result$`Pr(>F)`[1]
# run t-test (full precision)
if (levene_p < 0.05) {
cat("\nLevene's test is significant — using Welch t-test (unequal variances):\n")
t_result <- t.test(mom_age_en ~ overall_child_anaemia,
data = child_data,
var.equal = FALSE)
} else {
cat("\nLevene's test is not significant — using Student's t-test (equal variances):\n")
t_result <- t.test(mom_age_en ~ overall_child_anaemia,
data = child_data,
var.equal = TRUE)
}
# print t-test results
print(t_result)
```
*Child Age at Scan*
```{r}
# child age at scan (full sample)
# include repeated measures - for all scans across study timepoints
ULF_ida_final %>%
filter(!is.na(cor_child_anaemia)) %>% # only include PIDs with child anaemia data
summarise(
mean_child_age = mean(scan_age_months, na.rm = TRUE),
sd_child_age = sd(scan_age_months, na.rm = TRUE),
min_child_age = min(scan_age_months, na.rm = TRUE),
max_child_age = max(scan_age_months, na.rm = TRUE),
n = n()
)
```
```{r}
# group differences in child age at scan by child anaemia status
# ttest: child age at scan
# all scans included
library(dplyr)
library(car)
# deduplicate and filter data once
child_data <- ULF_ida_final %>%
filter(!is.na(cor_child_anaemia))
# summary statistics: scan age by child anaemia status (2 decimals)
summary_stats <- child_data %>%
group_by(cor_child_anaemia) %>%
summarise(
n = n(),
mean_age = format(round(mean(scan_age_months, na.rm = TRUE), 2), nsmall = 2),
sd_age = format(round(sd(scan_age_months, na.rm = TRUE), 2), nsmall = 2),
min_age = format(round(min(scan_age_months, na.rm = TRUE), 2), nsmall = 2),
max_age = format(round(max(scan_age_months, na.rm = TRUE), 2), nsmall = 2),
.groups = "drop"
)
print(summary_stats)
# levene's Test for homogeneity of variance
levene_result <- leveneTest(scan_age_months ~ cor_child_anaemia, data = child_data)
print(levene_result)
# extract Levene p-value
levene_p <- levene_result$`Pr(>F)`[1]
# automatically choose correct t-test
if (levene_p < 0.05) {
cat("\nLevene's test significant (p =", round(levene_p, 4), ") → Using Welch t-test (var.equal = FALSE)\n\n")
t_result <- t.test(scan_age_months ~ cor_child_anaemia,
data = child_data,
var.equal = FALSE)
} else {
cat("\nLevene's test not significant (p =", round(levene_p, 4), ") → Using Student t-test (var.equal = TRUE)\n\n")
t_result <- t.test(scan_age_months ~ cor_child_anaemia,
data = child_data,
var.equal = TRUE)
}
# print t-test results in full precision
print(t_result)
```
*Child Gestational Age (GA) at Birth*
```{r}
# child ga at birth (full sample)
# unique subject IDs: based on true prevalence
ULF_ida_final %>%
filter(!is.na(overall_child_anaemia)) %>% # only include PIDs with child anaemia data
distinct(subject_id, .keep_all = TRUE) %>%
summarise(
mean_child_ga = mean(ga_weeks, na.rm = TRUE),
sd_child_ga = sd(ga_weeks, na.rm = TRUE),
min_child_ga = min(ga_weeks, na.rm = TRUE),
max_child_ga = max(ga_weeks, na.rm = TRUE),
n = n()
)
```
```{r}
# group differences in ga at birth by child anaemia status
# ttest: ga
# unique subject IDs: based on true prevalence
library(dplyr)
library(car) # for Levene's Test
# deduplicate and filter once
child_data <- ULF_ida_final %>%
filter(!is.na(overall_child_anaemia)) %>%
distinct(subject_id, .keep_all = TRUE)
# summary statistics: GA by child anaemia status (2 decimals)
summary_stats <- child_data %>%
group_by(overall_child_anaemia) %>%
summarise(
n = n(),
mean_ga = format(round(mean(ga_weeks, na.rm = TRUE), 2), nsmall = 2),
sd_ga = format(round(sd(ga_weeks, na.rm = TRUE), 2), nsmall = 2),
min_ga = format(round(min(ga_weeks, na.rm = TRUE), 2), nsmall = 2),
max_ga = format(round(max(ga_weeks, na.rm = TRUE), 2), nsmall = 2),
.groups = "drop"
)
print(summary_stats)
# levene's Test for homogeneity of variance
levene_result <- leveneTest(ga_weeks ~ overall_child_anaemia, data = child_data)
print(levene_result)
# extract Levene p-value
levene_p <- levene_result$`Pr(>F)`[1]
# automatically choose correct t-test
if (levene_p < 0.05) {
cat("\nLevene's test significant (p =", round(levene_p, 4),
") → Using Welch t-test (var.equal = FALSE)\n\n")
t_result <- t.test(
ga_weeks ~ overall_child_anaemia,
data = child_data,
var.equal = FALSE
)
} else {
cat("\nLevene's test not significant (p =", round(levene_p, 4),
") → Using Student t-test (var.equal = TRUE)\n\n")
t_result <- t.test(
ga_weeks ~ overall_child_anaemia,
data = child_data,
var.equal = TRUE
)
}
# print t-test result in full precision
print(t_result)
```
*Infant Birth Weight (BW)*
```{r}
# child bw (full sample)
# unique subject IDs: based on true prevalence
ULF_ida_final %>%
filter(!is.na(overall_child_anaemia)) %>% #only include PIDs with child anaemia data
distinct(subject_id, .keep_all = TRUE) %>%
summarise(
mean_bw = mean(bw, na.rm = TRUE),
sd_bw = sd(bw, na.rm = TRUE),
min_bw = min(bw, na.rm = TRUE),
max_bw = max(bw, na.rm = TRUE),
n = n()
)
```
```{r}
# group differences in bw by child anaemia status
# ttest: bw
# unique subject IDs: based on true prevalence
library(dplyr)
library(car)
# deduplicate and filter once
birth_data <- ULF_ida_final %>%
filter(!is.na(overall_child_anaemia)) %>%
distinct(subject_id, .keep_all = TRUE)
# summary statistics: birth weight by child anaemia status (2 decimals)
summary_stats <- birth_data %>%
group_by(overall_child_anaemia) %>%
summarise(
n = n(),
mean_bw = format(round(mean(bw, na.rm = TRUE), 2), nsmall = 2),
sd_bw = format(round(sd(bw, na.rm = TRUE), 2), nsmall = 2),
min_bw = format(round(min(bw, na.rm = TRUE), 2), nsmall = 2),
max_bw = format(round(max(bw, na.rm = TRUE), 2), nsmall = 2),
.groups = "drop"
)
print(summary_stats)
# levene's test for homogeneity of variance
levene_result <- leveneTest(bw ~ overall_child_anaemia, data = birth_data)
print(levene_result)
# extract p-value
levene_p <- levene_result$`Pr(>F)`[1]
# automatically choose correct t-test
if (levene_p < 0.05) {
cat("\nLevene's test significant (p =", round(levene_p, 4),
") → Using Welch t-test (var.equal = FALSE)\n\n")
t_test_result <- t.test(
bw ~ overall_child_anaemia,
data = birth_data,
var.equal = FALSE
)
} else {
cat("\nLevene's test not significant (p =", round(levene_p, 4),
") → Using Student t-test (var.equal = TRUE)\n\n")
t_test_result <- t.test(
bw ~ overall_child_anaemia,
data = birth_data,
var.equal = TRUE
)
}
# print t-test result in full precision
print(t_test_result)
```
*Infant Birth Length (BL)*
```{r}
# child bl (full sample)
# unique subject IDs: based on true prevalence
ULF_ida_final %>%
filter(!is.na(overall_child_anaemia)) %>% #only include PIDs with child anaemia data
distinct(subject_id, .keep_all = TRUE) %>%
summarise(
mean_birthlength = mean(birth_length, na.rm = TRUE),
sd_birthlength = sd(birth_length, na.rm = TRUE),
min_birthlength = min(birth_length, na.rm = TRUE),
max_birthlength = max(birth_length, na.rm = TRUE),
n = n()
)
```
```{r}
# group differences in bl by child anaemia status
# ttest: bl
# unique subject IDs: based on true prevalence
library(dplyr)
library(car)
# deduplicate and filter once
birth_length_data <- ULF_ida_final %>%
filter(!is.na(overall_child_anaemia)) %>%
distinct(subject_id, .keep_all = TRUE)
# summary statistics: birth length by child anaemia status (2 decimals)
summary_stats <- birth_length_data %>%
group_by(overall_child_anaemia) %>%
summarise(
n = n(),
mean_bl = format(round(mean(birth_length, na.rm = TRUE), 2), nsmall = 2),
sd_bl = format(round(sd(birth_length, na.rm = TRUE), 2), nsmall = 2),
min_bl = format(round(min(birth_length, na.rm = TRUE), 2), nsmall = 2),
max_bl = format(round(max(birth_length, na.rm = TRUE), 2), nsmall = 2),
.groups = "drop"
)
print(summary_stats)
# levene's test for homogeneity of variance
levene_result <- leveneTest(birth_length ~ overall_child_anaemia, data = birth_length_data)
print(levene_result)
# extract p-value
levene_p <- levene_result$`Pr(>F)`[1]
# automatically choose correct t-test
if (levene_p < 0.05) {
cat("\nLevene's test significant (p =", round(levene_p, 4),
") → Using Welch t-test (var.equal = FALSE)\n\n")
t_test_result <- t.test(
birth_length ~ overall_child_anaemia,
data = birth_length_data,
var.equal = FALSE
)
} else {
cat("\nLevene's test not significant (p =", round(levene_p, 4),
") → Using Student t-test (var.equal = TRUE)\n\n")
t_test_result <- t.test(
birth_length ~ overall_child_anaemia,
data = birth_length_data,
var.equal = TRUE
)
}
# print t-test result in full precision
print(t_test_result)
```
*Infant Head Circumference (HC)*
```{r}
# child birth hc (full sample)
# unique subject IDs: based on true prevalence
ULF_ida_final %>%
filter(!is.na(overall_child_anaemia)) %>% #only include PIDs with child anaemia data
distinct(subject_id, .keep_all = TRUE) %>%
summarise(
mean_hc = mean(birth_hc, na.rm = TRUE),
sd_hc = sd(birth_hc, na.rm = TRUE),
min_hc = min(birth_hc, na.rm = TRUE),
max_hc = max(birth_hc, na.rm = TRUE),
n = n()
)
```
```{r}
# group differences in hc at birth by child anaemia status
# ttest: hc at birth
# unique subject IDs: based on true prevalence
library(dplyr)
library(car)
# deduplicate and filter once
birth_hc_data <- ULF_ida_final %>%
filter(!is.na(overall_child_anaemia)) %>%
distinct(subject_id, .keep_all = TRUE)
# summary statistics: birth head circumference by child anaemia status (2 decimals)
summary_stats <- birth_hc_data %>%
group_by(overall_child_anaemia) %>%
summarise(
n = n(),
mean_hc = format(round(mean(birth_hc, na.rm = TRUE), 2), nsmall = 2),
sd_hc = format(round(sd(birth_hc, na.rm = TRUE), 2), nsmall = 2),
min_hc = format(round(min(birth_hc, na.rm = TRUE), 2), nsmall = 2),
max_hc = format(round(max(birth_hc, na.rm = TRUE), 2), nsmall = 2),
.groups = "drop"
)
print(summary_stats)
# levene's test for homogeneity of variance
levene_result <- leveneTest(birth_hc ~ overall_child_anaemia, data = birth_hc_data)
print(levene_result)
# extract p-value
levene_p <- levene_result$`Pr(>F)`[1]
# automatically choose correct t-test
if (levene_p < 0.05) {
cat("\nLevene's test significant (p =", round(levene_p, 4),
") → Using Welch t-test (var.equal = FALSE)\n\n")
t_test_result <- t.test(
birth_hc ~ overall_child_anaemia,
data = birth_hc_data,
var.equal = FALSE
)
} else {
cat("\nLevene's test not significant (p =", round(levene_p, 4),
") → Using Student t-test (var.equal = TRUE)\n\n")
t_test_result <- t.test(
birth_hc ~ overall_child_anaemia,
data = birth_hc_data,
var.equal = TRUE
)
}
# print t-test result in full precision
print(t_test_result)
```
**Exploratory Analyses: Plotted Data using LOESS Curves**
*LOESS Curve: ICV*
```{r}
# icv
library(ggplot2)
library(dplyr)
icv_ulf_child <- ggplot(
data = ULF_ida_final %>% filter(!is.na(cor_child_anaemia)),
aes(
x = scan_age_months,
y = mm_hyp_icv,
color = cor_child_anaemia,
fill = cor_child_anaemia
)
) +
geom_point() +
geom_smooth(
method = "loess",
se = TRUE,
level = 0.95,
alpha = 0.3 # semi-transparent ribbon
) +
scale_color_manual(
values = c(
"no_child_anaemia" = "#1f77b4", # dark blue line
"child_anaemia" = "#d62728" # dark red line
),
name = "Child Anaemia Status"
) +
scale_fill_manual(
values = c(
"no_child_anaemia" = "#c6d9f1", # lighter blue ribbon
"child_anaemia" = "#f7b0b0" # lighter red ribbon
),
guide = "none" # hide fill legend
) +
scale_x_continuous(
breaks = seq(0, max(ULF_ida_final$scan_age_months, na.rm = TRUE), by = 10)
) +
labs(
title = "Total ICV Volume vs. Scan Age",
x = "Scan Age (Months)",
y = "Total ICV Volume (mm3)",
color = "Child Anaemia Status"
) +
theme_minimal()
icv_ulf_child
```
*LOESS Curve: Corpus Callosum*
```{r}
# corpus callosum
cc_ulf_child <- ggplot(
data = ULF_ida_final %>% filter(!is.na(cor_child_anaemia)),
aes(
x = scan_age_months,
y = total_cc,
color = cor_child_anaemia,
fill = cor_child_anaemia
)
) +
geom_point() +
geom_smooth(
method = "loess",
se = TRUE, # show 95% confidence ribbons
level = 0.95,
alpha = 0.3 # make ribbons semi-transparent
) +
scale_color_manual(
values = c(
"no_child_anaemia" = "#1f77b4", # dark blue
"child_anaemia" = "#d62728" # dark red
),
name = "Child Anaemia Status"
) +
scale_fill_manual(
values = c(
"no_child_anaemia" = "#c6d9f1", # lighter blue ribbon
"child_anaemia" = "#f7b0b0" # lighter red ribbon
),
guide = "none" # hide fill legend
) +
scale_x_continuous(
breaks = seq(0, max(ULF_ida_final$scan_age_months, na.rm = TRUE), by = 3)
) +
scale_y_continuous(
breaks = seq(
floor(min(ULF_ida_final$total_cc, na.rm = TRUE) / 10) * 10,
ceiling(max(ULF_ida_final$total_cc, na.rm = TRUE) / 10) * 10,
by = 4000
)
) +
labs(
title = "Total CC Volume vs. Scan Age",
x = "Scan Age (Months)",
y = "Total CC Volume (mm3)",
color = "Child Anaemia Status"
) +
theme_minimal()
cc_ulf_child
```
*LOESS Curve: Caudate Nucleus*
```{r}
# caudate nucleus
caudate_ulf_child <- ggplot(
data = ULF_ida_final %>% filter(!is.na(cor_child_anaemia)),
aes(
x = scan_age_months,
y = total_caudate,
color = cor_child_anaemia,
fill = cor_child_anaemia
)
) +
geom_point() +
geom_smooth(
method = "loess",
se = TRUE, # show 95% confidence ribbons
level = 0.95,
alpha = 0.3 # semi-transparent ribbons
) +
scale_color_manual(
values = c(
"no_child_anaemia" = "#1f77b4", # dark blue
"child_anaemia" = "#d62728" # dark red
),
name = "Child Anaemia Status"
) +
scale_fill_manual(
values = c(
"no_child_anaemia" = "#c6d9f1", # lighter blue ribbon
"child_anaemia" = "#f7b0b0" # lighter red ribbon
),
guide = "none" # hide fill legend
) +
scale_x_continuous(
breaks = seq(0, max(ULF_ida_final$scan_age_months, na.rm = TRUE), by = 3)
) +
scale_y_continuous(
n.breaks = 5 # adjust y-axis ticks
) +
labs(
title = "Total Caudate Volume vs. Scan Age",
x = "Scan Age (Months)",
y = "Total Caudate Volume (mm³)",
color = "Child Anaemia Status"
) +
theme_minimal()
caudate_ulf_child
```
*LOESS Curve: Putamen*
```{r}
# putamen
putamen_ulf_child <- ggplot(
data = ULF_ida_final %>% filter(!is.na(cor_child_anaemia)),
aes(
x = scan_age_months,
y = total_putamen,
color = cor_child_anaemia,
fill = cor_child_anaemia
)
) +
geom_point() +
geom_smooth(
method = "loess",
se = TRUE, # show 95% confidence ribbons
level = 0.95,
alpha = 0.3 # semi-transparent ribbons
) +
scale_color_manual(
values = c(
"no_child_anaemia" = "#1f77b4", # dark blue
"child_anaemia" = "#d62728" # dark red
),
name = "Child Anaemia Status"
) +
scale_fill_manual(
values = c(
"no_child_anaemia" = "#c6d9f1", # lighter blue ribbon
"child_anaemia" = "#f7b0b0" # lighter red ribbon
),
guide = "none" # hide fill legend
) +
scale_x_continuous(
breaks = seq(0, max(ULF_ida_final$scan_age_months, na.rm = TRUE), by = 3)
) +
labs(
title = "Total Putamen Volume vs. Scan Age",
x = "Scan Age (Months)",
y = "Total Putamen Volume (mm³)",
color = "Child Anaemia Status"
) +
theme_minimal()
putamen_ulf_child
```
**Linear Mixed Effects (LME) Models: Child Anaemia Status Only**
First, the LMEs described above for maternal anaemia were replicated with postnatal child anaemia as the primary predictor of interest. This series of models were run to assess the independent associations between child anaemia and regional child brain volumes.
Based on the LOESS curves, there were no observed age-related differences between groups based on child anaemia status. Therefore, interactions were excluded to simplify the models.
--\> Main effects only (with child anaemia as the primary predictor)
Log (age) was used for all models based on non-linear growth trajectories
```{r}
# lme: icv
# main effects: child anaemia
# log (age)
# model summary
ulf_model_icv_child <- lmer(mm_hyp_icv ~ cor_child_anaemia + log_age + child_sex + (1 | subject_id),
data = ULF_ida_final)
summary(ulf_model_icv_child)
# beta values
standardize_parameters(ulf_model_icv_child)
```
```{r}
# lme: cc
# main effects: child anaemia
# log (age)
# model summary
ulf_model_cc_child <- lmer(total_cc ~ cor_child_anaemia + log_age + child_sex + (1 | subject_id), data = ULF_ida_final)
summary(ulf_model_cc_child)
# beta values
standardize_parameters(ulf_model_cc_child)
```
```{r}
# putamen
# main effects: child anaemia
# log (age)
# model summary
ulf_model_putamen_child <- lmer(total_putamen ~ cor_child_anaemia + log_age + child_sex + (1 | subject_id),
data = ULF_ida_final)
summary(ulf_model_putamen_child)
# beta values
standardize_parameters(ulf_model_putamen_child)
```
```{r}
# lme: caudate nucleus
# main effects: child anaemia
# log (age)
# model summary
ulf_model_caudate_child <- lmer(total_caudate ~ cor_child_anaemia + log_age + child_sex + (1 | subject_id),
data = ULF_ida_final)
summary(ulf_model_caudate_child)
#to get beta values
standardize_parameters(ulf_model_caudate_child)
```
**Linear Mixed Effects (LME) Models: Maternal Anaemia and Child Anaemia**
Second, the final models from the primary analyses assessing the impact of antenatal maternal anaemia status on regional brain volumes were repeated with the addition of postnatal child anaemia status as a covariate. This assessed the relative contributions of antenatal versus postnatal exposures, and examined whether the observed associations between maternal anaemia and regional child brain volumes remained after controlling for child anaemia.
Log (age) was used for all models based on non-linear growth trajectories
```{r}
# lme: ICV
# predictors: maternal anaemia and child anaemia
# based on final maternal anaemia model (primary analysis) for icv: exclude interaction between maternal anaemia and age
# model summary
ulf_model_icv_mat_child <- lmer(mm_hyp_icv ~ maternal_preg_anaemia_status + log_age + child_sex + cor_child_anaemia + (1 | subject_id),
data = ULF_ida_final)
summary(ulf_model_icv_mat_child)
# beta values
standardize_parameters(ulf_model_icv_mat_child)
```
```{r}
# lme: corpus callosum
# predictors: maternal anaemia and child anaemia
# based on final maternal anaemia model (primary analysis) for corpus callosum : include interaction between maternal anaemia and age
# model summary
ulf_model_cc_mat_child <- lmer(total_cc ~ maternal_preg_anaemia_status * log_age + child_sex + cor_child_anaemia + (1 | subject_id), data = ULF_ida_final)
summary(ulf_model_cc_mat_child)
# beta values
standardize_parameters(ulf_model_cc_mat_child)
```
```{r}
# lme: putamen
# predictors: maternal anaemia and child anaemia
# based on final maternal anaemia model (primary analysis) for putamen: exclude interaction between maternal anaemia and age
# model summary
ulf_model_putamen_mat_child <- lmer(total_putamen ~ maternal_preg_anaemia_status + log_age + child_sex + cor_child_anaemia + (1 | subject_id),
data = ULF_ida_final)
summary(ulf_model_putamen_mat_child)
# beta values
standardize_parameters(ulf_model_putamen_mat_child)
```
```{r}
# lme: caudate nucleus
# predictors: maternal anaemia and child anaemia
# based on final maternal anaemia model (primary analysis) for caudate nucleus: include interaction between maternal anaemia and age
# model summary
ulf_model_caudate_mat_child <- lmer(total_caudate ~ maternal_preg_anaemia_status * log_age + child_sex + cor_child_anaemia + (1 | subject_id),
data = ULF_ida_final)
summary(ulf_model_caudate_mat_child)
# beta values
standardize_parameters(ulf_model_caudate_mat_child)
```
------------------------------------------------------------------------
**Tertiary Analysis: Maternal Iron Deficiency (ULF sample)**
------------------------------------------------------------------------
**Sample Size: For infants with Maternal ID Data**
```{r}
# total sample size
# including repeated measures
ULF_ida_final %>%
filter(!is.na(maternal_preg_adj_brinda_id)) %>% #only include PIDs with antenatal maternal ID data
summarise(n = n())
```
```{r}
# total sample size
# Excluding repeated measures: unique id
# unique

Khula_AnaemiaImagingAnalysis.qmd at commit b50e304, no license · at the source

Overview

Authors: Jessica E Ringshaw^1,2,3, Michal R Zieff¹, Niall J Bourke³, Chiara Casella^4,5, Layla E Bradford^1,2, Simone R Williams^1,2, Donna Herr¹, Marlie Miles^1,2, Carly Bennallick³, Khula South Africa Data Collection Team, Sean Deoni⁶, Jonathan O’Muircheartaigh^4,5,7, Dan J Stein^2,8, Daniel C Alexander⁹, Derek K Jones¹⁰, Steven C R Williams³, Kirsten A Donald^1,2

ORCID iDs: Jessica E Ringshaw, Niall J Bourke, Chiara Casella, Layla E Bradford, Jonathan O’Muircheartaigh, Dan J Stein, Daniel C Alexander, Derek K Jones, Steven C R Williams, Kirsten A Donald

Institutions: University of Cape Town (South Africa); King's College London (United Kingdom); Red Cross War Memorial Children's Hospital (South Africa); Gates Foundation (United States); MRC Centre for Neurodevelopmental Disorders (United Kingdom); Medical Research Council (United Kingdom); South African Medical Research Council (South Africa); University College London (United Kingdom); Cardiff University (United Kingdom)

Journal: Brain communications, volume 8, issue 5, article fcag318

Dates: received 8 December 2025; accepted 4 August 2026; published online 8 September 2026

Type: Research article · Language: English

License: CC BY

Identifiers: DOI 10.1093/braincomms/fcag318 · PMID 42713352 · PMCID PMC13553114 · OpenAlex W7212063535

Open access: gold, a free copy (OpenAlex)

Status: code verified

Categories: structural MRI / diffusion (modality), human (organism), developmental (subfield)

Methods: Connectivity, Statistics, Preprocessing, fMRI & imaging

Keywords: neuroimaging, high-field MRI, ultra-low-field MRI, antenatal maternal anaemia, child brain structure

Topic: Advanced MRI Techniques and Applications (Radiology, Nuclear Medicine and Imaging, Medicine), according to OpenAlex

Funding: National Institute for Health and Care Research; Bill & Melinda Gates Foundation (INV-023509); Gates Foundation (INV-090982, INV-047888, INV-023509); UK Foreign, Commonwealth, and Development Office; Wellcome Trust Research Enrichment Public Engagement (224287/Z/21/A); Wellcome Trust International Training Fellowship (224287/Z/21/Z); Developing Excellence in Leadership, Training, and Science in Africa (Del-22-002); European and Developing Countries Clinical Trials Partnership 2; Wellcome Leap 1kD; Wellcome Trust (224287/Z/21/Z, 224287/Z/21/A); The First 1000 Days (222076/Z/20/Z); European Union; Maudsley Biomedical Research Centre; Science for Africa Foundation

Citations: not cited yet (Europe PMC); 57 references in the paper

Abstract

With the evolution of ultra-low-field MRI and the recognition of antenatal maternal anaemia as an important driver of altered neurodevelopment in toddlers and children, it is critical to determine whether these effects are detectable at ultra-low-field (64 mT) in infancy. The aim of this study was to assess the impact of antenatal maternal anaemia on infant brain structure across the first 2 years of life, using high-field (3 T) and ultra-low-field (64 mT) MRI. This neuroimaging substudy was embedded within Khula, an observational population-based birth cohort in South Africa. Pregnant women were enrolled antenatally and postnatally. Mother-child dyads (n = 394) were followed prospectively with a subsample attending neuroimaging at ∼3, 6, 12, 18 and 24 months of age. Anaemia was classified using World Health Organization thresholds, and neuroimaging data were processed using MiniMORPH. Linear mixed-effects models were used to investigate associations between antenatal maternal anaemia status and absolute regional infant brain volumes using high-field and ultra-low-field MRI. In repeated measures high-field (n = 195) and ultra-low-field (n = 341) infant neuroimaging subsamples, the prevalence of antenatal maternal anaemia was 28.24% (37/131) and 29.76% (61/205), respectively. Maternal anaemia in pregnancy was associated with altered child brain structure across both MRI systems, with group differences becoming detectable at ∼12 months. In the ultra-low-field subsample, infants born to anaemic mothers had 3.77% smaller intracranial volume (β = −0.24, P = 0.004) and 3.32% smaller putamen volumes (β = −0.18, P = 0.040) across the first 2 years of life. The interaction between antenatal maternal anaemia and age was significant for the caudate nucleus (β = −0.13, P = 0.038) and corpus callosum (β = −0.15, P = 0.007). Antenatal maternal anaemia was associated with 3.70% and 4.29% smaller caudate nucleus volumes at 18 and 24 months of age, respectively. Similarly, infants born to anaemic mothers had 4.39% smaller corpus callosum volumes by 12 months and 6.27% smaller corpus callosum volumes by 24 months. Postnatal child anaemia and antenatal maternal iron deficiency status were not associated with total or regional child brain volumes in the ultra-low-field subsample from this cohort. Maternal anaemia remained a robust predictor of volume differences in sensitivity analyses. This study is the first to demonstrate that the impact of maternal anaemia in pregnancy on child brain structure is detectable as early as infancy. The implications of this research are 2-fold: (i) informing the feasibility of ultra-low-field MRI in low- and middle-income countries and (ii) the timing and optimization of targeted recommendations for anaemia management in practice and policy.

Reproduced under the paper's license (CC BY), from the paper cited above.

Repositories

Its files are read in the Code ↔ Paper reader above, with 9 matches between paragraphs and lines of code.

UNITY-Physics

License: none: the authors keep all their rights

State: the link answers, verified on 26 September 2026

Evidence: the link answers

Software Heritage: not checked

Found in: the text, “Neuroimaging processing”

Not found: README, license file, CITATION.cff, environment file, tests, continuous integration, documentation

Availability: 1 check, the latest on 26 September 2026: the link answers (HTTP 200)

26 September 2026: the link answers (HTTP 200)

At the source: github.com/UNITY-Physics

UNITY-Physics/fw-minimorph

License: MIT

State: the link answers, verified on 26 September 2026

Evidence: files inventoried

Commit: 08c7398e42c88031da40ae84941214aa6b4a0299, 7 April 2026

Languages: Python (8), Shell (3)

Size: 35 files, 11 scripts

Software Heritage: not archived

Found in: “Data availability”

Holds: README, license file, environment (Dockerfile), tests, documentation

Not found: CITATION.cff, continuous integration

Tools: FSL (2 files), ANTs (1 file), FreeSurfer (1 file), Matplotlib (1 file), NiBabel (1 file), NumPy (1 file), pandas (1 file), Pillow (1 file)

Availability: 1 check, the latest on 26 September 2026: the link answers

26 September 2026: the link answers

13 files

app/main.sh, Shell, 335 lines, 1 match
interactive-run.sh, Shell, 23 lines
run.py, Python, 54 lines
utils/Inspect_segmentations.py, Python, 158 lines
utils/__init__.py, Python, 2 lines
utils/command_line.py, Python, 193 lines
utils/context.py, Python, 874 lines
utils/generate_command.py, Python, 73 lines
utils/join_data.py, Python, 37 lines
utils/parser.py, Python, 69 lines
utils/tests/unit_test.sh, Shell, 45 lines
LICENSE, License, 21 lines
README.md, Text, 173 lines

At the source: github.com/UNITY-Physics/fw-minimorph

JERingshaw/Khula-Anaemia-Imaging-Analysis

License: none: the authors keep all their rights

State: the link answers, verified on 26 September 2026

Evidence: files inventoried

Commit: b50e304453da1c77f7617057f487c5c3334a68b8, 24 March 2026

Languages: Quarto (1)

Size: 3 files, 1 script

Software Heritage: not archived

Found in: “Data availability”

Holds: README, 1 notebook

Not found: license file, CITATION.cff, environment file, tests, continuous integration, documentation

Tools: car (1 file), easystats (1 file), emmeans (1 file), ggplot2 (1 file), lme4 (1 file), lmerTest (1 file), tidyverse (1 file)

Availability: 1 check, the latest on 26 September 2026: the link answers

26 September 2026: the link answers

2 files

Khula_AnaemiaImagingAnalysis.qmd, Quarto, 6,955 lines, 8 matches
README.md, Text, 2 lines

At the source: github.com/JERingshaw/Khula-Anaemia-Imaging-Analysis.git

The paper's code and data availability statement is in the Data section.

Tracing map

Proposed by the machine: these links were found in the paper and verified at the source, without human review. The map will receive a Zenodo DOI once one of the paper's authors has validated it with their ORCID.

What the map holds:

3 repositories of the authors' code, each at its verified commit, with its license and how the link was found in the paper;
12 scripts, each with its path and the digest of its content;
9 matches between paragraphs of the paper and lines of the code (method lexical-v1);
neither the text of the paper nor the code itself.

Its JSON (tracing-map.json) is deposited on Zenodo with its DOI once the map is validated.

Data

No dataset and no data link were found in the paper.

Data availability

The de-identified data that support the findings of this study are available from the authors upon reasonable request as per the Khula South Africa study guidelines. All software used in the development of Flywheel gears for the UNITY project has been shared as an open-source resource on GitHub (https://github.com/UNITY-Physics). This includes the MiniMORPH processing pipeline (https://github.com/UNITY-Physics/fw-minimorph) and the R script used for statistical analyses (https://github.com/JERingshaw/Khula-Anaemia-Imaging-Analysis.git)

Reproduced under the paper's license (CC BY), from the paper cited above.

Versions

The history of this record: each version stored by the harvester or made by a correction of its authors or of the maintainers of its code, and what changed in its facts. The texts of the paper (its abstract, its availability statements) are not part of it; versions that changed only those are not listed.

Version 1, 27 September 2026: the first record

Recorded: type, language, journal, volume, issue, pages, dates, 17 authors, 5 keywords, 14 funders, 48 references.

Cite

This paper

Ringshaw, J. E., Zieff, M. R., Bourke, N. J., Casella, C., Bradford, L. E., Williams, S. R., Herr, D., Miles, M., Bennallick, C., Khula South Africa Data Collection Team, Deoni, S., O’Muircheartaigh, J., Stein, D. J., Alexander, D. C., Jones, D. K., Williams, S. C. R., & Donald, K. A. (2026). Antenatal maternal anaemia and infant brain structure: high-field (3 T) and ultra-low-field (64 mT) MRI findings from South Africa. Brain communications, 8(5), fcag318. https://doi.org/10.1093/braincomms/fcag318

BibTeX

@article{ringshaw2026antenatal, author = {Ringshaw, Jessica E and Zieff, Michal R and Bourke, Niall J and Casella, Chiara and Bradford, Layla E and Williams, Simone R and Herr, Donna and Miles, Marlie and Bennallick, Carly and {Khula South Africa Data Collection Team} and Deoni, Sean and O’Muircheartaigh, Jonathan and Stein, Dan J and Alexander, Daniel C and Jones, Derek K and Williams, Steven C R and Donald, Kirsten A}, title = {{Antenatal maternal anaemia and infant brain structure: high-field (3 T) and ultra-low-field (64 mT) MRI findings from South Africa}}, journal = {Brain communications}, year = {2026}, month = sep, volume = {8}, number = {5}, pages = {fcag318}, publisher = {Oxford University Press}, issn = {2632-1297}, doi = {10.1093/braincomms/fcag318}, url = {https://doi.org/10.1093/braincomms/fcag318}, pmid = {42713352}, pmcid = {PMC13553114} }

RIS

TY - JOUR AU - Ringshaw, Jessica E AU - Zieff, Michal R AU - Bourke, Niall J AU - Casella, Chiara AU - Bradford, Layla E AU - Williams, Simone R AU - Herr, Donna AU - Miles, Marlie AU - Bennallick, Carly AU - Khula South Africa Data Collection Team AU - Deoni, Sean AU - O’Muircheartaigh, Jonathan AU - Stein, Dan J AU - Alexander, Daniel C AU - Jones, Derek K AU - Williams, Steven C R AU - Donald, Kirsten A TI - Antenatal maternal anaemia and infant brain structure: high-field (3 T) and ultra-low-field (64 mT) MRI findings from South Africa T2 - Brain communications J2 - Brain Commun PY - 2026 DA - 2026/09/08 VL - 8 IS - 5 SP - fcag318 SN - 2632-1297 PB - Oxford University Press DO - 10.1093/braincomms/fcag318 UR - https://doi.org/10.1093/braincomms/fcag318 LA - en ER -

CSL-JSON

{ "id": "10.1093/braincomms/fcag318", "type": "article-journal", "title": "Antenatal maternal anaemia and infant brain structure: high-field (3 T) and ultra-low-field (64 mT) MRI findings from South Africa", "container-title": "Brain communications", "author": [ { "family": "Ringshaw", "given": "Jessica E" }, { "family": "Zieff", "given": "Michal R" }, { "family": "Bourke", "given": "Niall J" }, { "family": "Casella", "given": "Chiara" }, { "family": "Bradford", "given": "Layla E" }, { "family": "Williams", "given": "Simone R" }, { "family": "Herr", "given": "Donna" }, { "family": "Miles", "given": "Marlie" }, { "family": "Bennallick", "given": "Carly" }, { "literal": "Khula South Africa Data Collection Team" }, { "family": "Deoni", "given": "Sean" }, { "family": "O’Muircheartaigh", "given": "Jonathan" }, { "family": "Stein", "given": "Dan J" }, { "family": "Alexander", "given": "Daniel C" }, { "family": "Jones", "given": "Derek K" }, { "family": "Williams", "given": "Steven C R" }, { "family": "Donald", "given": "Kirsten A" } ], "container-title-short": "Brain Commun", "volume": "8", "issue": "5", "page": "fcag318", "DOI": "10.1093/braincomms/fcag318", "PMID": "42713352", "PMCID": "PMC13553114", "ISSN": "2632-1297", "publisher": "Oxford University Press", "URL": "https://doi.org/10.1093/braincomms/fcag318", "language": "en", "issued": { "date-parts": [ [ 2026, 9, 8 ] ] } }

The tracing map gets a citation of its own once an author has validated it and it has a DOI.

Contribute

The authors of this paper can claim it, correct its record and validate its tracing map, and the maintainers of its code (its owner, or a public member of its organization) correct what it says of their repository; anyone signed in can ask for its removal. Every request goes to OSCR's own machine, which answers it; your account page follows them.

Sign in with ORCID to claim this paper as one of its authors, correct its record or validate its tracing map: when the paper's metadata lists your ORCID iD, you are recognized at once. Maintainers of its code: sign in with GitHub, then claim the repository on your account page.

Request its removal

To ask OSCR to remove this record, the copies of its authors' scripts or its tracing map, use the removal request page: signed in, you say who you are, what to remove and why, then review and confirm the request. Published rules decide every request (how).

Discussion, reproductions, activity

Discussion: questions and error reports about this paper and its code, from signed-in readers and its authors. It opens with sign-in.

Reproductions: reports from readers who ran the authors' code: what they reproduced, with which environment, commit and data. It opens with sign-in.

Activity: what happens around this paper: new versions of its record, its map's validation, discussions and reproductions. It opens with sign-in.

Antenatal maternal anaemia and infant brain structure: high-field (3 T) and ultra-low-field (64 mT) MRI findings from South Africa.

Code ↔ Paper

Paper

The authors' code

Overview

Abstract

Repositories

UNITY-Physics

UNITY-Physics/fw-minimorph

JERingshaw/Khula-Anaemia-Imaging-Analysis

Tracing map

Data

Data availability

Versions

Version 1, 27 September 2026: the first record

Cite

This paper

Similar papers

Contribute

Claim this paper

Correct its record

Validate its tracing map

Add the badge to its README

Request its removal

Discussion, reproductions, activity