Target Trial Emulation of a Non-Pharmaceutical Intervention Within the STAGE Project: Smoking Cessation During Pregnancy as a Case Study

Abstract

Introduction: Randomized controlled trials (RCTs) are the gold standard for causal inference, but observational studies can also provide robust causal evidence. Target Trial Emulation (TTE) applies the principles of RCTs to observational data by explicitly specifying the protocol of a hypothetical trial before conducting the analysis. While this framework has been applied to pharmaceutical interventions, non-pharmaceutical interventions present additional challenges: exposures often correspond to behaviors rather than clearly defined interventions capable of inducing those behaviors.

Objective: To describe the development and implementation of a TTE study within the European STAGE project investigating the effect of smoking cessation during pregnancy on maternal and child outcomes.

Methods: We developed a target trial protocol specifying eligibility criteria, treatment strategies, assignment procedures, outcomes, and follow-up for both the hypothetical target trial and its observational emulation. The emulation used data from the NINFEA study, an internet-based birth cohort including approximately 7,500 mother-child pairs, with repeated questionnaires administered from pregnancy until 16 years after delivery.

Results: The target trial includes adult pregnant women who smoked before pregnancy and compares two treatment strategies: smoking cessation (0 cigarettes/day) versus continued smoking (>0 cigarettes/day). Randomization is stratified by maternal age, educational level, parity, recruitment center, smoking history, and pre-pregnancy morbidities. Maternal outcomes include smoking habits after delivery and hospitalization within 18 months postpartum; child outcomes include birth weight and respiratory diseases. Follow-up begins at conception and ends at outcome occurrence, loss to- or end of follow-up. Both intention-to-treat and per-protocol effects are estimable.

The observational emulation includes all adult women in the NINFEA cohort who smoked before pregnancy. The availability of repeated smoking assessments allowed the emulation of two nested target trials, considering the assignment to treatment strategies at two time points: smoking cessation either at conception or during the first trimester. Randomization is emulated through adjustment for baseline covariates corresponding to the stratification variables. Outcomes are assessed using questionnaires administered 6 and 18 months, and 7 years after delivery.

Smoking cessation at conception substantially reduced the odds of postpartum smoking relapse at both 6 months (OR = 0.13; 95% CI: 0.10–0.18) and 18 months (OR = 0.21; 95% CI: 0.15–0.28), and was associated with a higher birth weight (β = 97.6 g; 95% CI: 42.5–152.7 g). Similar results were obtained with smoking cessation within the first trimester.

Conclusions: Our results showed positive effects of smoking cessation during pregnancy on both maternal and child health. The study, demonstrating the feasibility of emulating RCTs of non-pharmaceutical interventions using observational data, also highlights methodological considerations specific to behavioral interventions. Unlike pharmacological treatments, smoking cessation can be achieved through multiple interventions (e.g. counselling or financial incentives), which were not directly observed in the cohort. Consequently, the emulation focused on well-defined behavioral strategies using explicit cessation thresholds rather than the underlying intervention. Furthermore, identifying pregnancy onset as the trial time zero ensured alignment between the target trial and the observational data, while repeated exposure measurements enabled the assessment of adherence to the assigned treatment strategy over time.