Antenatal Corticosteroid Administration in Late Preterm Singleton Pregnancies

Reproducibility package for the major-revision resubmission of:

Antenatal Corticosteroid Administration in Late Preterm Singleton Pregnancies: A Propensity Score-Weighted Analysis of Neonatal Outcomes

Version

v2.2-ga-sensitivity-update

This version adds reviewer-requested gestational-age sensitivity analyses to the existing pipeline, alongside AMA-style reference cleanup and updated terminology (PBV → PPV with first-use clarification).

The previous release v2.1-delivery-mode-sensitivity added the delivery-mode sensitivity analysis. The current release builds on that by additionally including gestational age in the propensity score model, both as a continuous variable and as a completed-week categorical variable.

Main analytic cohort

• Source records: 1,091
• Singleton late preterm deliveries: 1,087
• Excluded: ACS before 34 weeks: 75
• Final analytic cohort: 1,012
• ACS exposed: 126
• No ACS: 886

Primary outcome

The manuscript primary outcome is oxygen_any: documented initial respiratory support at birth/early neonatal stabilization, derived from the oxygen-support/neonatal-resuscitation fields. It includes hood oxygen, CPAP, positive-pressure ventilation (PPV; coded as PBV in the source dataset), or intubation.

resp_any is diagnosis-based pulmonary morbidity and is not the primary outcome.

Key revised IPTW results

Outcome	IPTW estimate
Initial respiratory support at birth	OR 1.40, 95% CI 0.90–2.18, P=.130
NICU admission	OR 1.10, 95% CI 0.69–1.75, P=.685
Neonatal hypoglycemia	OR 2.14, 95% CI 0.63–7.28, P=.224
Pulmonary morbidity (resp_any)	OR 0.87, 95% CI 0.48–1.56, P=.632
Higher-intensity initial support	OR 0.96, 95% CI 0.59–1.57, P=.877
Hospital-course respiratory support	OR 1.02, 95% CI 0.59–1.76, P=.939

Sensitivity analyses for the primary outcome

Analysis	n	OR	95% CI	P
Unadjusted	1,012	1.54	1.03–2.29	.034
Adjusted	1,012	1.34	0.87–2.06	.184
IPTW (primary)	1,012	1.40	0.90–2.18	.130
IPTW + doubly robust	1,012	1.41	0.87–2.27	.159
Include early ACS (IPTW)	1,087	1.37	0.89–2.09	.150
Include early ACS (IPTW + DR)	1,087	1.40	0.89–2.21	.148
Exclude delivery mode from PS (IPTW)	1,012	1.54	1.01–2.35	.046
Exclude delivery mode from PS + DR	1,012	1.57	1.02–2.42	.042
Include GA in PS (continuous, days)	1,012	1.28	0.82–2.00	.282
Include GA in PS (completed-week category)	1,012	1.31	0.84–2.04	.241

Gestational-age sensitivity analyses (new in v2.2)

Gestational age at delivery was not included in the primary propensity score because it may reflect delivery timing after the treatment decision rather than a conventional baseline confounder. Effect modification by gestational age was evaluated separately through completed-week stratified analyses, and sensitivity analyses including gestational age in the propensity score were used to assess robustness.

In response to reviewer feedback, we additionally re-fit the propensity score model with gestational age included, both as a continuous variable in days at birth and as a completed-week categorical variable.

When gestational age was added to the propensity score model:

• IPTW odds ratio attenuated from 1.40 (no GA) to 1.28 (continuous GA) or 1.31 (categorical GA).
• All confidence intervals continued to span the null.
• The qualitative conclusion (no significant association with respiratory support) was unchanged.

This pattern indicates that gestational age at delivery is an influential timing and baseline-risk proxy rather than a conventional baseline confounder, supporting the choice to handle it outside the primary propensity-score model and to assess robustness using completed-week stratification and GA-inclusive sensitivity analyses.

Delivery-mode sensitivity analysis

Delivery mode was included in the primary propensity score. Because planned versus intrapartum cesarean delivery and absence of labor were unavailable, we also ran a sensitivity analysis excluding delivery mode from the propensity score.

Sensitivity analysis	Primary outcome estimate
Exclude delivery mode from PS (IPTW)	OR 1.54, 95% CI 1.01–2.35, P=.046
Exclude delivery mode from PS + doubly robust adjustment	OR 1.57, 95% CI 1.02–2.42, P=.042

This sensitivity analysis supports delivery mode as an important measured confounder. It does not resolve residual confounding from absence of labor, planned versus intrapartum cesarean delivery, or indication severity.

Run command

From the repository root:

Rscript code/ACS_late_preterm_statistical_script.R data/ACS_Late_Preterm_deidentified.csv outputs

The script will produce all CSVs listed below, plus diagnostic plots in outputs/plots/. The new GA sensitivity output is outputs/ga_sensitivity_analyses.csv.

Important output files

• outputs/outcome_results_main.csv
• outputs/outcome_results_continuous_main.csv
• outputs/outcome_results_by_gaweek.csv
• outputs/table3_gaweek_primary_outcome.csv
• outputs/outcome_results_sensitivity_exclude_delivery_mode.csv
• outputs/outcome_results_sensitivity_include_early.csv
• outputs/ga_sensitivity_analyses.csv (new in v2.2)
• outputs/dose_timing_summary_overall.csv
• outputs/initial_respiratory_support_modality.csv
• outputs/respiratory_support_descriptive_counts.csv
• outputs/key_manuscript_results.csv
• outputs/comparison_to_revised_manuscript_targets_v6.csv
• outputs/balance_smd_before_after.csv
• outputs/ps_boundary_and_weight_diagnostics_main.csv

Figures

The figures/ folder contains manuscript-ready TIFF files with corrected metadata/captions. Diagnostic plots generated by the R script are in outputs/plots/.

In v2.2, Figure 2 has been relabeled for publication: variable-code labels (e.g., bmi, gravida, type_of_delivery) replaced with clinical labels (Body mass index, Gravidity, Delivery mode, etc.), and the legend now reads "Before IPTW" and "After IPTW" rather than "Unweighted" and "IPTW".

Data use

The dataset is de-identified and provided for reproducibility of the associated manuscript. Do not attempt re-identification or linkage to external data sources.