Among people with HIV (PWH), initiating antiretroviral therapy (ART) is associated with weight gain that might increase the risk of cardiometabolic outcomes. We estimated the effect of hypothetical interventions to restrict increases in body mass index (BMI) in the two years after ART initiation on the 7-year risk of incident type-2 diabetes mellitus (T2DM) among PWH in the North American AIDS Cohort Collaboration on Research and Design (NA-ACCORD) clinical cohorts using the parametric extended g-formula. Among 13,822 normal weight or overweight PWH (BMI ≥18.5 and ≤29.9 kg/m2) who initiated ART between 2012-2021, the 7-year risk of T2DM was 5.6% (95% confidence interval [CI]: 4.8%, 6.3%). Compared to no intervention on BMI, if BMI increase during the first 2 years was capped at 5% of baseline BMI, the risk ratio (RR) for T2DM risk was 0.94 (95% CI, 0.86, 1.03); if BMI was capped at 29.9 kg/m2, RR=0.88 (95% CI, 0.80, 0.97); and if PWH were prevented from gaining any weight over the first 2 years, RR=0.81 (95% CI, 0.72, 0.92). These results provide preliminary evidence to promote studies of real weight maintenance interventions among PWH initiating ART.

Dynamic treatment regimes (DTRs) formalize sequences of treatment decision rules that adapt to a patient's evolving clinical status. In observational studies, estimation of the value of a DTR (i.e., the expected population-outcome of a treatment strategy) using inverse probability weighting (IPW) and augmented IPW can be unstable when strict regime compatibility yields few valid participant trajectories, leading to highly variable weights and noisy estimates. In this paper, we introduce two adaptive weighting frameworks: generalized compatibility weighting (GCW) and bootstrap compatibility windowing (BCW), that allow near-compatible trajectories to contribute information in limited samples, with diminishing information-borrowing in larger samples. We illustrate how these estimators reduce noise in estimation and stabilize empirical optimal treatment threshold identification. Using data from adults living with HIV in one site of the TREAT Asia HIV Observational Database in India observed from 2013-2024, we evaluate clinically interpretable threshold-based decision rules based on baseline CD4 count and weight. In this application, GCW and BCW produced more stable estimated regime values than standard IPW-based approaches, with reduced variability across candidate thresholds and more reliable identification of empirically optimal treatment rules. These methods provide practical tools for improving treatment rule evaluation in observational studies.

Racial discrimination has been associated with an increased risk of preterm birth (PTB) among Black women. Few studies have examined whether this association differs by clinical presentation of PTB. We examined the association between racial discrimination and PTB subtype using data from a 2010-2013 clinic-based prospective cohort study of nulliparous women among women who self-reported Black race. Lifetime racial discrimination was measured with the revised Experiences of Discrimination scale. We used inverse probability of exposure and post-stratification weights with the cumulative incidence function to estimate associations between racial discrimination and spontaneous PTB and indicated PTB within a competing risk framework. We estimated standard errors via bootstrapping. We stratified by education and age group to examine effect modification. Compared to none, racial discrimination in ≥3 situations was associated with 1.66 (95% CI, 0.72, 3.82) times the risk of spontaneous PTB and 0.37 (95% CI, 0.15, 0.93) times the risk of indicated PTB. Point estimates between racial discrimination and spontaneous PTB were strongest among women with some college education and ≥20 year-olds. Future studies should consider clinical subtype in assessments of the association between racism and preterm birth and include in the study design late prenatal care initiators.

The g-formula is a foundational tool for identifying causal effects in observational data. This tool is based on the law of iterated expectation, a key mathematical identity in statistics. However, the notation with which the law of iterated expectation and the g-formula is expressed can be opaque to those with little background in statistics. We provide a primer introducing the law of iterated expectation, the integration notation used to express it, and its role for causal effect identification via the g-formula. Combined with the assumptions of causal consistency, positivity, and conditional exchangeability, the law of iterated expectation yields a causal standardization formula (the g-formula) in two nonparametrically equivalent forms: a non-iterative conditional expectation (NICE) form involving a single weighted average of conditional outcome means, and an iterative conditional expectation (ICE) form involving nested expectations. We illustrate both forms using three progressively complex numerical examples: a time-fixed example with a single binary confounder, a time-fixed example with discrete and continuous confounders, and a time-varying example with two timepoints. We provide clarity on what the law of iterated expectation is, how it is related to the g-formula, and how to gain intuition of its mathematical formulations in actual data examples that can be generalized to a range of settings.

Sleep, sedentary behaviour (SB), standing, light-intensity physical activity (LPA) and moderate-to-vigorous physical activity (MVPA) compose the 24-hour day, and reallocating time between these behaviours is associated with premature mortality; however, whether these associations differ by sleep characteristics remains unclear. We examined associations between time reallocations across these device-measured behaviours and all-cause, cardiovascular disease (CVD) and physical activity-related cancer mortality, and potential effect modification by sleep duration and regularity. We conducted a population-based prospective cohort study among UK Biobank participants and applied compositional Cox regression. In 68 995 adults, 2668 deaths occurred over a mean follow-up of 8.0 years. Among participants meeting sleep duration guidelines, reallocating 30-minutes from sleep to standing, LPA or MVPA was favourably associated with all-cause mortality, with HRs of 0.85 (95%CI 0.79, 0.92), 0.87 (0.81, 0.94), and 0.80 (0.74, 0.87), respectively. Among short sleepers, reallocating 30-minutes from sleep to SB, standing or LPA was adversely associated with CVD mortality, with HRs of 1.18 (1.04, 1.34), 1.20 (1.04, 1.38), and 1.18 (1.03, 1.35), respectively. Beneficial associations of reallocating SB to sleep were evident among both regular sleepers and irregular sleepers. These observational findings suggest that associations between 24-hour behavioural reallocations and mortality may vary according to sleep characteristics.

In the article by Fox et al. (Am J Epidemiol. 0000;000(00):0000-0000), they argue that research questions in epidemiology are too often unclear leading, in turn, to confusion regarding correlation and causation. I build on points made by Fox et al, arguing that the research question is equivalent to the destination on a train ticket: both tell us where we want to go. Just as we would be upset if we were issued a train ticket with an ambiguous destination, we should similarly be upset when reading an epidemiologic article with an unclear research question. I further leverage the analogy to show that research questions are statements about "where we want to go" not "where we will end up". I end with a call that given the fundamental status of research questions in the research process, they deserve their own section in our research articles.

We described how to adapt the prevalent new user (PNU) study design to estimate the effect of drug discontinuation among patients who discontinue and clone-censor-weighting (CCW) approach to estimate the effect of drug discontinuation among an eligible population at a pre-specified time point, focusing on the differences between the target trials the two approaches emulate. While both approaches produced unbiased estimates of the effect of drug discontinuation when all confounders were adjusted for in simulations, the true estimands differed, particularly when treatment effect heterogeneity was introduced. We also applied the approaches to study the effect of statin discontinuation among United States Medicare beneficiaries who discontinued statin therapy after at least 5 years (PNU) or at approximately 5 years (CCW) of continuous use. The PNU analysis estimated a mortality rate difference (MRD) of 3.6 (95% CI, 3.3-3.9) per 100 person-years and a mortality rate ratio (MRR) of 1.41 (95% CI, 1.37-1.45), while the CCW analysis estimated a MRD of 3.4 (95% CI, 2.8-4.0) per 100 person-years and MRR of 1.55 (1.45-1.66). Researchers studying drug discontinuation should carefully consider their causal question and target population when choosing between these two newer approaches.

We present a novel application of the SCCS method for estimating the effectiveness and waning of the third dose of COVID-19 monovalent vaccine in France. We used data on individuals hospitalised for COVID-19 between September 2021 and February 2022 from the Système National des Données de Santé. Only cases having received a 3rd dose of COVID-19 vaccine prior to or on the day of hospitalisation were included in the analysis. We developed a new SCCS model including only time post-dose 3 vaccination, with reference period 24-65 days post-vaccination to capture maximum vaccine protection. Time was split into short-term and long-term post-vaccination risk periods, including days 4-9 corresponding to minimum protection, and successive 7-day windows from day 66 post vaccination. The model included two exposure variables, representing short-term and long-term effects of vaccination. The analysis confirms the relative effectiveness of the third-dose COVID-19 vaccination, the waning of vaccine immunity though of limited extent, and possible effect modification related to virus strain, vaccine, vaccination history, age, and sex. The strengths and weaknesses of this approach are discussed. Suitably applied, the SCCS method may provide a useful method of analysis of vaccine effectiveness which automatically controls for time-invariant confounders.

Adverse birth outcomes have been linked to "unintended" pregnancy; however, research that relies on retrospective measurement of "intentions" is vulnerable to differential recall bias. This study estimates associations between prospectively measured pregnancy preferences and preterm birth (<37 weeks) and low birthweight (<5.5 pounds). We analyzed longitudinal, population-based data from the Surveys of Women, conducted in nine US states (2016-2023), linking births reported across annual waves with pregnancy preferences measured before pregnancy using the 14-item Desire to Avoid Pregnancy (DAP) scale. We fit weighted logistic regression models to estimate associations between DAP categories (lower DAP: more open to pregnancy; higher DAP: stronger preference to avoid pregnancy) and outcomes using multiple imputation (primary analysis) and complete case analysis excluding participants whose DAP scores were imputed (sensitivity analysis). Although the adjusted association between DAP grouping and low birthweight was not statistically significant in analyses using imputation (15% high vs. 7% low DAP), it was significant treating DAP scores continuously (aOR=1.56 [1.01-2.43]) and in complete case analysis (20% vs. 6%, p=0.01). DAP scores were not associated with preterm birth in imputation analyses (18% high vs. 14% low DAP) or using continuous scores but were in complete case analysis (25% vs. 12%, p=0.02). While the magnitudes of associations were suggestive of a relationship between a strong prospective preference to avoid pregnancy and low birthweight, analyses were limited by low precision. Supporting individuals in achieving their reproductive preferences may have child health benefits.

Accurate tumor growth rate estimation is vital for linking disease progression to patient characteristics and outcomes, especially in rare diseases where limited data increase bias risk. We evaluated two potential sources of bias in linear growth rate (LGR) estimation using data from a prior natural history (NH) study of von Hippel-Lindau (VHL) renal cell carcinoma (RCC), a rare hereditary disorder. A Quantitative bias analysis (QBA) assessed selection bias from differential follow-up, while a Bayesian analysis (BA) examined shrinkage bias from a frequentist linear mixed-effects approach. In the NH cohort (244 patients; 601 target tumors), the original median tumor-level LGR was 1.6 mm/year. QBA produced a bias-corrected LGR of 1.7 mm/year, suggesting minimal impact of selection bias. Bayesian modeling produced an LGR estimate of 3.93 mm/year, aligning more closely with external reports and suggesting that surgically removed tumors had substantially higher growth rates than the mixed cohort implied by frequentist methods. QBA showed limited bias from differential follow-up, while BA provided a complementary estimate that was less influenced by shrinkage bias inherent to mixed-effects modeling. These findings underscore the importance of appropriate study design and method selection to diagnose and adjust for biases when characterizing tumor growth in VHL RCC.

Serology is increasingly used to monitor infection transmission and disease elimination. Embedding dried blood spot collection in trachoma surveys allows transmission intensity inference through population-level seroconversion rate (SCR) estimation, but no formal assessment of required sample size has been published. Using serology data from 40 trachoma prevalence surveys, we estimated seroprevalence intra-cluster correlation coefficient, a key design parameter, and assessed survey design considerations for incorporating serological monitoring. Design scenarios focused on proposed operational SCR thresholds (2.2 [no action needed] and 4.5 [action needed] per 100 child years) for interpretation of serological data in low-transmission and post-elimination settings. We evaluated SCR estimation in 42 two-stage designs by calculating precision (confidence interval width around an SCR value) and power (the measure of deviation of suggested thresholds to the SCR value). When the SCR is ≤1.5 and >5.7 per 100 person-years, sample sizes between 300-2000 allowed good precision of SCR estimation. We anticipate a standard prevalence survey with 30 clusters and 30 households per cluster should provide ≥80% power to determine that an SCR ≤1.2 falls below a 2.2 threshold and SCR ≥5.7 falls above a 4.5 threshold. Our results support collection of serological data via the recommended survey design for trachoma surveillance.