Assessing Representativeness

# Load required libraries
library(tidyverse)
library(readxl)
library(knitr)
library(kableExtra)
library(gtsummary)
library(patchwork)

## Original metabolomics_dataset values:

## 
##   No  Yes <NA> 
## 3631  228    0

## Data loaded successfully!

## Total participants: 3859

## Group distribution:

## 
## Excluded Included     <NA> 
##     3631      228        0

Overview

This report assesses the representativeness of the selected Metabolomics dataset cohort relative to the overall cohort (hereafter referred to as the “Metabolomics” and “overall” cohorts).

Tables and plots are presented to assess the frequency, distribution, etc. for each variable.

Assess representativeness according to demographic/baseline variable set: - Outlined in the “Variable Category” column of the data dictionary sheet in the data set.

Summary of Results

We find the distributions for most “primary” variables broadly similar between the Metabolomics and overall cohorts, with some notable differences:

Primary Variables: - Maternal age at birth differs significantly (p<0.001), with the Metabolomics cohort being slightly older (mean 32.9 vs 31.8 years) - DASS domains show significant differences, with the Metabolomics cohort having better mental health scores at 36 weeks (p=0.049) - Other demographic variables (gender, birth weight/length, BMI, diabetes, asthma, etc.) are well-balanced

Sample Availability: - The Metabolomics cohort shows significantly higher engagement across all questionnaire measures (all p<0.001) - Much higher completion rates for ASQ questionnaires at all time points - Greater participation in follow-up assessments

Key Differences: - Twin births are significantly over-represented in the Metabolomics cohort (8.3% vs 3.2%, p<0.001) - Child age differs significantly, with Metabolomics children being older (mean 4.6 vs 3.8 years, p<0.001) - BMI at 3 years is significantly lower in the Metabolomics cohort (15.5 vs 15.9, p=0.006) - 3-year asthma is significantly higher in the Metabolomics cohort (3.1% vs 0.6%, p=0.009) - Previous pregnancies are significantly higher in the Metabolomics cohort (p=0.039)

Data & Methods

DEIDENTIFIED Full Participant list Perron - Final dataset July 2025.xlsx
- Contains data and data dictionary.
After some cleaning and the assignment of variable names, we get the following dimensions (rows, columns):

dim(dat)

## [1] 3859   88

We assess the variable similarity between Metabolomics cohort (N = 228) and overall cohort (N = 3859) using:

Summary tabulations
Distributional plots
Simple statistical tests
- Note: the p-values presented can be interpreted with a grain of salt. Often in cases with large sample sizes, non-meaningful differences (in reality) return “significant” p-values.

Variable Breakdown

There are ~79 candidate variables that can be used to assess the similarity between the Metabolomics and overall cohorts. - These variables are broadly “sample availability”, “child/maternal demographic”, “child/maternal characteristics”.

Firstly, let’s select just a handful candidate variables with the aim of getting an overall “snapshot” of the similarity between the sub-cohort and overall cohort. - For example, we want to ensure the sub-cohort is not entirely female, born in a single year, of a single ethnic origin, etc.

“Primary” variables (amended to include the 19 variables outlined) - Gender of child - Maternal age at birth - Maternal pre-pregnancy weight - Maternal pre-pregnancy height - Maternal pre-pregnancy BMI - Infant weight - Infant length - Infant BMI at birth - Infant ethnic origin - Indigenous status of baby - Vaginal or C-section birth - Maternal gestational diabetes status - Maternal Type 1 or Type 2 Diabetes - Maternal mental health diagnosis (Depression, Anxiety disorder, Bipolar, Schizophrenia, OCD, Anorexia Nervous, Specific Phobias, Behavioural Disorders) - Individual disorder breakdown: Each mental health condition is now analyzed separately, allowing participants with multiple conditions to be counted in each relevant category - Any mental health diagnosis: Overall indicator of any mental health condition - Depression: Depressive disorders - Anxiety: Anxiety disorders
- Bipolar: Bipolar affective disorder - Schizophrenia: Schizophrenia spectrum disorders - OCD: Obsessive-compulsive disorder - Anorexia: Anorexia nervosa - Phobias: Specific phobias - Behavioural: Behavioural disorders - Maternal Asthma - Number of 18wk DASS domains with “Severe” or “Extremely Severe” - Number of 18wk DASS domains with “Normal” - Number of 36wk DASS domains with “Severe” or “Extremely Severe” - Number of 36wk DASS domains with “Normal”

Sample availability variables - Availability of maternal/child urine/blood/stool samples (20 weeks, 2 months, 6 months, 12 months, 3 years) - ASQ completion (4 month, 9 month, 1 year, 3 year, 5 year) - Early Connors assigned and completed - REDCap questionnaires assigned and completed - Availability of MNS data - Total questionnaires completed

Outcome variables - 1yr child wheeze - 1 year BMI - 1 year Ferritin results - 1 year count of positive SPT wheals - 1 year any positive food SPT wheals - 1 year any positive airborne/enviro SPT wheals - 3 year count of positive SPT wheals - 3 year any positive food SPT wheals - 3 year any positive airborne/enviro SPT wheals - 3 year BMI - 3 year wheeze - 3 year asthma - 3 year ferritin - 5 year BMI - 5 year asthma - 5 year Ferritin - 5 year any positive food SPT wheals - 5 year any positive airborne/enviro SPT wheals - 3 year count of Connors domains equal to, or above 65 - 3 year count of other clinical indicators parent reported as “3” highest

Remaining variables - Variables not contained in the primary nor sample availability variable set.

1) Primary Variables

Broadly, the distribution “primary” variable set between the Metabolomics and overall cohorts is similar.

Notes - Child sex is well-balanced between groups (45.6% vs 47.9% female, p=0.517) - Maternal age at birth differs significantly, with the Metabolomics cohort being older (mean 32.9 vs 31.8 years, p<0.001) - Maternal pre-pregnancy characteristics (weight, height, BMI) are well-balanced between groups (all p>0.69) - Infant birth characteristics (weight, length, BMI) show no significant differences (all p>0.11) - Ethnic origin and indigenous status are well-balanced (p=0.407 and p=1.000 respectively) - Birth type and maternal diabetes show no significant differences (p=0.234 and p=0.912) - Mental health diagnosis patterns: - Overall mental health diagnoses are similar between groups (11.4% vs 14.5%, p=0.198) - Individual disorders show similar distributions across both cohorts - No participants in the Metabolomics cohort had bipolar disorder, schizophrenia, OCD, anorexia, phobias, or behavioural disorders - Maternal asthma shows no significant difference (5.7% vs 8.1%, p=0.223) - DASS mental health scores: - 18-week assessments show the Metabolomics cohort trends toward better mental health (more “Normal” scores, p=0.074) - 36-week assessments show significantly more “Normal” scores in the Metabolomics cohort (p=0.049) - Generally fewer severe mental health symptoms in the Metabolomics group

# Primary continuous variables
primary_continuous <- list(
  list("gender_child", "Child sex assigned at birth"),
  list("maternal_age_birth", "Maternal Age (Birth)"),
  list("maternal_prepreg_weight", "Maternal pre-pregnancy weight"),
  list("maternal_prepreg_height", "Maternal Pre-pregnancy height"),
  list("maternal_prepreg_bmi_calc", "Maternal Pre-pregnancy BMI (Calc.)"),
  list("infant_weight", "Infant Weight"),
  list("infant_length", "Infant Length at birth"),
  list("infant_bmi_calc", "Infant BMI at birth (Calc.)"),
  list("ethnic_origin", "Ethnic Origin"),
  list("indigenous_status", "Indigenous Status of Baby"),
  list("birth_type_derived", "Vaginal or C section birth (Deriv.)"),
  list("maternal_gest_diabetes_derived", "Maternal Gestational Diabetes? (Deriv.)"),
  list("maternal_diabetes_t1t2_derived", "Maternal Type 1 or Type 2 Diabetes? (Deriv.)"),
  list("mh_any", "Any Maternal Mental Health Diagnosis"),
  list("mh_depression", "Maternal Depression"),
  list("mh_anxiety", "Maternal Anxiety Disorder"),
  list("mh_bipolar", "Maternal Bipolar Disorder"),
  list("mh_schizophrenia", "Maternal Schizophrenia"),
  list("mh_ocd", "Maternal OCD"),
  list("mh_anorexia", "Maternal Anorexia"),
  list("mh_phobias", "Maternal Specific Phobias"),
  list("mh_behavioural", "Maternal Behavioural Disorders"),
  list("maternal_asthma_derived", "Maternal Asthma? (Deriv.)"),
  list("dass21_18w_severe_count", "Number of 18wk DASS domains with \"Severe\" or \"Extremely Severe\""),
  list("dass21_18w_normal_count", "Number of 18wk DASS domains with \"Normal\""),
  list("dass21_36w_severe_count", "Number of 36wk DASS domains with \"Severe\" or \"Extremely Severe\""),
  list("dass21_36w_normal_count", "Number of 36wk DASS domains with \"Normal\"")
)

for(var_info in primary_continuous) {
  var_name <- var_info[[1]]
  title <- var_info[[2]]
  
  if(var_name %in% numeric_vars) {
    analyze_continuous(dat, var_name, title)
  } else {
    analyze_categorical(dat, var_name, title)
  }
}

Child sex assigned at birth

**P-value:** 0.517 (Chi-squared test)
Characteristic	Included N = 228	Overall N = 3859
Female	104 (45.6%)	1849 (47.9%)
Male	124 (54.4%)	2010 (52.1%)
Total	228 (100.0%)	3859 (100.0%)

Maternal Age (Birth)

**P-value:** <0.001 (Wilcoxon rank sum test)
Characteristic	N	Mean (Min, Max)	Median (Q1, Q3)
Included	228	32.9 (24, 44)	32 (30, 35)
Overall	3859	31.8 (17, 50)	32 (29, 35)

Unknown: Included = 0 , Overall = 0

Maternal pre-pregnancy weight

**P-value:** 0.718 (Wilcoxon rank sum test)
Characteristic	N	Mean (Min, Max)	Median (Q1, Q3)
Included	216	69.7 (47, 128)	67 (60, 75.2)
Overall	3633	70.5 (38, 134)	68 (60, 79)

Unknown: Included = 12 , Overall = 226

Maternal Pre-pregnancy height

**P-value:** 0.714 (Wilcoxon rank sum test)
Characteristic	N	Mean (Min, Max)	Median (Q1, Q3)
Included	219	1.7 (1.5, 1.8)	1.7 (1.6, 1.7)
Overall	3698	1.7 (1.4, 1.9)	1.6 (1.6, 1.7)

Unknown: Included = 9 , Overall = 161

Maternal Pre-pregnancy BMI (Calc.)

**P-value:** 0.693 (Wilcoxon rank sum test)
Characteristic	N	Mean (Min, Max)	Median (Q1, Q3)
Included	216	25.5 (17.3, 42.3)	24.2 (21.9, 28.7)
Overall	3623	25.7 (14.7, 47.3)	24.7 (21.9, 28.6)

Unknown: Included = 12 , Overall = 236

Infant Weight

**P-value:** 0.380 (Wilcoxon rank sum test)
Characteristic	N	Mean (Min, Max)	Median (Q1, Q3)
Included	209	3331.1 (1480, 4500)	3360 (3070, 3655)
Overall	3578	3369.2 (1095, 5410)	3390 (3062.8, 3695)

Unknown: Included = 19 , Overall = 281

Infant Length at birth

**P-value:** 0.112 (Wilcoxon rank sum test)
Characteristic	N	Mean (Min, Max)	Median (Q1, Q3)
Included	209	50.1 (39, 57)	50 (48, 52)
Overall	3574	50.3 (31, 60)	50 (49, 52)

Unknown: Included = 19 , Overall = 285

Infant BMI at birth (Calc.)

**P-value:** 0.976 (Wilcoxon rank sum test)
Characteristic	N	Mean (Min, Max)	Median (Q1, Q3)
Included	209	13.2 (7.6, 17.6)	13.2 (12.4, 14.1)
Overall	3574	13.3 (7.6, 32.3)	13.2 (12.3, 14.2)

Unknown: Included = 19 , Overall = 285

Ethnic Origin

**P-value:** 0.407 (Fisher’s exact test)
Characteristic	Included N = 228	Overall N = 3859
Unknown	19 (8.3%)	285 (7.4%)
1	178 (78.1%)	2951 (76.5%)
3	8 (3.5%)	177 (4.6%)
4	2 (0.9%)	116 (3%)
5	2 (0.9%)	18 (0.5%)
8	19 (8.3%)	288 (7.5%)
10	0 (0%)	8 (0.2%)
6	0 (0%)	4 (0.1%)
7	0 (0%)	12 (0.3%)
Total	228 (100.0%)	3859 (100.0%)

Indigenous Status of Baby

**P-value:** 1.000 (Fisher’s exact test)
Characteristic	Included N = 228	Overall N = 3859
Unknown	19 (8.3%)	283 (7.3%)
4	209 (91.7%)	3563 (92.3%)
1	0 (0%)	12 (0.3%)
2	0 (0%)	1 (0%)
Total	228 (100.0%)	3859 (100.0%)

Vaginal or C section birth (Deriv.)

**P-value:** 0.242 (Fisher’s exact test)
Characteristic	Included N = 228	Overall N = 3859
Unknown	19 (8.3%)	281 (7.3%)
Caesarean Elective	67 (29.4%)	964 (25%)
Caesarean Emergency	44 (19.3%)	804 (20.8%)
Vaginal	98 (43%)	1810 (46.9%)
Total	228 (100.0%)	3859 (100.0%)

Maternal Gestational Diabetes? (Deriv.)

**P-value:** 0.912 (Chi-squared test)
Characteristic	Included N = 228	Overall N = 3859
Unknown	19 (8.3%)	281 (7.3%)
FALSE	192 (84.2%)	3271 (84.8%)
TRUE	17 (7.5%)	307 (8%)
Total	228 (100.0%)	3859 (100.0%)

Maternal Type 1 or Type 2 Diabetes? (Deriv.)

**P-value:** 1.000 (Fisher’s exact test)
Characteristic	Included N = 228	Overall N = 3859
Unknown	19 (8.3%)	281 (7.3%)
FALSE	209 (91.7%)	3572 (92.6%)
TRUE	0 (0%)	6 (0.2%)
Total	228 (100.0%)	3859 (100.0%)

Any Maternal Mental Health Diagnosis

**P-value:** 0.198 (Chi-squared test)
Characteristic	Included N = 228	Overall N = 3859
FALSE	202 (88.6%)	3298 (85.5%)
TRUE	26 (11.4%)	561 (14.5%)
Total	228 (100.0%)	3859 (100.0%)

Maternal Depression

**P-value:** 0.551 (Chi-squared test)
Characteristic	Included N = 228	Overall N = 3859
FALSE	213 (93.4%)	3557 (92.2%)
TRUE	15 (6.6%)	302 (7.8%)
Total	228 (100.0%)	3859 (100.0%)

Maternal Anxiety Disorder

**P-value:** 0.159 (Chi-squared test)
Characteristic	Included N = 228	Overall N = 3859
FALSE	210 (92.1%)	3437 (89.1%)
TRUE	18 (7.9%)	422 (10.9%)
Total	228 (100.0%)	3859 (100.0%)

Maternal Bipolar Disorder

**P-value:** 1.000 (Fisher’s exact test)
Characteristic	Included N = 228	Overall N = 3859
FALSE	228 (100%)	3845 (99.6%)
TRUE	0 (0%)	14 (0.4%)
Total	228 (100.0%)	3859 (100.0%)

Maternal Schizophrenia

**P-value:** NA
Characteristic	Included N = 228	Overall N = 3859
FALSE	228 (100%)	3859 (100%)
Total	228 (100.0%)	3859 (100.0%)

Maternal OCD

**P-value:** NA
Characteristic	Included N = 228	Overall N = 3859
FALSE	228 (100%)	3859 (100%)
Total	228 (100.0%)	3859 (100.0%)

Maternal Anorexia

**P-value:** 0.620 (Fisher’s exact test)
Characteristic	Included N = 228	Overall N = 3859
FALSE	228 (100%)	3843 (99.6%)
TRUE	0 (0%)	16 (0.4%)
Total	228 (100.0%)	3859 (100.0%)

Maternal Specific Phobias

**P-value:** NA
Characteristic	Included N = 228	Overall N = 3859
FALSE	228 (100%)	3859 (100%)
Total	228 (100.0%)	3859 (100.0%)

Maternal Behavioural Disorders

**P-value:** 1.000 (Fisher’s exact test)
Characteristic	Included N = 228	Overall N = 3859
FALSE	228 (100%)	3849 (99.7%)
TRUE	0 (0%)	10 (0.3%)
Total	228 (100.0%)	3859 (100.0%)

Maternal Asthma? (Deriv.)

**P-value:** 0.223 (Chi-squared test)
Characteristic	Included N = 228	Overall N = 3859
Unknown	19 (8.3%)	281 (7.3%)
FALSE	196 (86%)	3264 (84.6%)
TRUE	13 (5.7%)	314 (8.1%)
Total	228 (100.0%)	3859 (100.0%)

Number of 18wk DASS domains with “Severe” or “Extremely Severe”

**P-value:** 0.102 (Wilcoxon rank sum test)
Characteristic	N	Mean (Min, Max)	Median (Q1, Q3)
Included	196	0.1 (0, 2)	0 (0, 0)
Overall	2781	0.1 (0, 3)	0 (0, 0)

Unknown: Included = 32 , Overall = 1078

Number of 18wk DASS domains with “Normal”

**P-value:** 0.074 (Wilcoxon rank sum test)
Characteristic	N	Mean (Min, Max)	Median (Q1, Q3)
Included	196	2.7 (0, 3)	3 (3, 3)
Overall	2781	2.6 (0, 3)	3 (2, 3)

Unknown: Included = 32 , Overall = 1078

Number of 36wk DASS domains with “Severe” or “Extremely Severe”

**P-value:** 0.235 (Wilcoxon rank sum test)
Characteristic	N	Mean (Min, Max)	Median (Q1, Q3)
Included	127	0 (0, 3)	0 (0, 0)
Overall	1570	0.1 (0, 3)	0 (0, 0)

Unknown: Included = 101 , Overall = 2289

Number of 36wk DASS domains with “Normal”

**P-value:** 0.049 (Wilcoxon rank sum test)
Characteristic	N	Mean (Min, Max)	Median (Q1, Q3)
Included	127	2.8 (0, 3)	3 (3, 3)
Overall	1570	2.6 (0, 3)	3 (3, 3)

Unknown: Included = 101 , Overall = 2289

2) Sample Availability

In general, sample availability and engagement in the Metabolomics cohort is substantially higher relative to the overall cohort, with significant differences across multiple measures.

# Sample availability variables
sample_vars <- list(
  list("mns_data_available", "MNS Data Available?"),
  list("asq_assigned", "ASQ Questionnaires Assigned"),
  list("asq_completed", "ASQ Questionnaires Completed"),
  list("early_connors_assigned", "Early Connors Assigned"),
  list("early_connors_completed", "Early Connors Completed"),
  list("aes_assigned", "AES Questionnaires Assigned"),
  list("aes_completed", "AES Questionnaires Completed"),
  list("redcap_assigned", "RedCap Questionnaires Assigned"),
  list("redcap_completed", "RedCap Questionnaires completed"),
  list("questionnaires_total_completed", "Total Questionnaires Completed"),
  list("asq_4m_completed", "ASQ 4 Month Completed"),
  list("asq_4m_paed_review", "ASQ 4 Month Review with Paediatrician"),
  list("asq_9m_completed", "ASQ 9 Month Completed"),
  list("asq_9m_paed_review", "ASQ 9 Month Review with Paediatrician"),
  list("asq_1yr_completed", "ASQ 1 Year Completed"),
  list("asq_1yr_paed_review", "ASQ 1 Year Review with Paediatrician"),
  list("asq_3yr_completed", "ASQ 3 Year Completed"),
  list("asq_3yr_paed_review", "ASQ 3 Year Review with Paediatrician"),
  list("asq_5yr_completed", "ASQ 5 Year Completed"),
  list("asq_5yr_paed_review", "ASQ 5 Year Review with Paediatrician"),
  list("asq_paed_review_count", "Number of times ASQ has prompted review with PAED")
)

for(var_info in sample_vars) {
  var_name <- var_info[[1]]
  title <- var_info[[2]]
  
  if(var_name %in% numeric_vars) {
    analyze_continuous(dat, var_name, title)
  } else {
    analyze_categorical(dat, var_name, title)
  }
}