Paediatric and perinatal epidemiology最新文献

Background: Pregnancy loss is recognised as a competing event in studies of prenatal medication use. A healthy live birth also precludes subsequent pregnancy outcomes, yet is often censored in time-to-event analyses.

Objectives: Using a Monte Carlo simulation, we examined bias resulting from censoring versus accounting for healthy live birth as a competing event in estimates of the total effect of prenatal medication use on pregnancy outcomes.

Methods: We simulated 2000 cohorts of 7500 conceptions with chronic hypertension under 12 treatment profiles. Ongoing pregnancies were indexed into the trial and randomly assigned to initiate or not initiate antihypertensives. Using time-to-event methods, we estimated absolute risks, risk differences (RD) per 100 pregnancies, and risk ratios (RR) for two outcomes, mirroring a prior trial: (i) composite fetal death or severe prenatal preeclampsia and (ii) small for gestational age (SGA) live birth. For the composite outcome, we conducted analyses where non-preeclamptic live birth was: (1) a censoring event and (2) a competing event. For SGA live birth, we conducted analyses where fetal death and non-SGA live birth were: (1) censoring events; (2) a competing event and censoring event, respectively; and (3) competing events.

Results: For the composite outcome, censoring non-preeclamptic live births overestimated the absolute risk by 42.3 to 49.1 percentage points; RD and RR estimates were biased (e.g., RD bias range -6.18 to 0.46). For SGA live birth, analyses censoring non-SGA live births (with or without fetal death as a competing event) overestimated absolute risk by 30.0 to 37.7 and 40.9 to 52.4 percentage points on average; RD and RR estimates were biased (e.g., RD bias range -7.45 to 0.79 and -9.62 to 1.81, respectively). Analyses in which healthy live births were modelled as competing events produced unbiased risks, RDs and RRs.

Conclusions: Censoring healthy live births resulted in overestimated risks as well as biased and imprecise total treatment effect estimates. Such inaccuracies about risks undermine informed patient-provider decision-making.

Background: Longitudinal perinatal studies that study the effects of preconception or prenatal treatments on pregnancy outcomes can have inherent forms of selection bias. For example, these studies often restrict analyses to those who had a livebirth, those with a specified gestation duration or those with complete follow-up. These selection factors are often associated with the treatment and have shared causes with the outcome, which may induce bias in estimating causal effects. Though such selection bias can affect all causal inference approaches, what is unknown is how this bias compares in direction and magnitude across different approaches.

Objectives: We conducted a simulation study to assess and compare the direction and magnitude of bias due to censoring across three common analytic approaches: inverse probability weighting (IPW), instrumental variable (IV) and sibling-comparison design.

Methods: We simulated data for various scenarios under two censoring mechanisms (loss to follow-up; and competing events) with a null true causal treatment effect. The simulated scenarios varied in the probability of the censoring mechanism or its strength of association with treatment or outcome. For each scenario, we generated 500 datasets (sample size = 10,000) and calculated the mean bias in risk difference estimates obtained from the three analytic approaches.

Results: Across all approaches, the proportion of censoring had no specific effect on mean bias. However, increasing the association of censoring with treatment or outcome increased the mean bias. The mean bias in all approaches was generally away from the null in the same direction and often to a similar extent (e.g., 0.5 percentage points away from the null in simulated scenarios with moderate association between treatment and censoring). However, in simulated scenarios with strong association between treatment and censoring, IV analyses were meaningfully more biased than IPW and sibling-comparison design analyses, with mean bias reaching two percentage points.

Conclusions: Across the simulated scenarios, the mean bias in all three approaches was generally away from the null in the same direction and often to a similar extent. Thus, triangulating effect estimates from different analytic approaches in perinatal studies is challenging and may lead to invalid interpretations in the presence of selection processes.

Background: Compared to undergoing prenatal care with a physician, care with a midwife reduces the risk of medical interventions and complications during labor and delivery among low-risk pregnant individuals. However, many analyses that assess the relationship between midwifery-led care and birth outcomes condition on live births, potentially inducing a type of collider bias.

Objective: The objective was to analyse the change in prevalence of caesarean deliveries, primary and secondary postpartum haemorrhage, obstetric trauma, and maternal infection under hypothetical scenarios where midwifery-led prenatal care was increased.

Methods: Our sample included commercially insured, midwifery-eligible pregnant people in an insurance claims data source. We used g-computation to assess the change in prevalence of caesarean deliveries, primary and secondary postpartum haemorrhage, obstetric trauma, and maternal infection if 10%, 20%, and 50% more pregnant people enrolled in prenatal care with a midwife rather than a physician, among a cohort of low-risk pregnant people with commercial insurance in the U.S. between 2004 and 2015.

Results: With a 50% increase in midwifery-led care compared with no increase, we found the prevalence of caesarean deliveries was reduced by 5.4 percentage points (95% confidence interval [CI] -5.7, -5.1) and of maternal infection decreased by 1.3 percentage points (95% CI -1.6, -1.0), while the prevalence of primary postpartum haemorrhage increased by 0.5 percentage points (95% CI 0.4, 0.6) and of secondary postpartum haemorrhage increased by 0.6 percentage points (95% CI 0.4, 0.7).

Conclusions: Increasing midwifery-led prenatal care reduced the prevalence of caesarean deliveries and maternal infections and slightly increased the prevalence of primary and secondary postpartum haemorrhage. Our results were similar to those of studies among live birth cohorts.

Background: Uterine fibroids, a common gynaecologic condition, are often underdiagnosed, potentially biasing results in epidemiologic studies due to measurement error.

Objectives: To examine how varying sensitivity in detecting uterine fibroids impacts effect estimates, using the association with hypertension onset as an example.

Methods: Three simulation studies were conducted (N = 100,000), considering true population prevalences of uterine fibroids of 5%, 20% and 60%. The first study varied detection sensitivity between 0% and 100%. The second examined differential sensitivity by symptom status (asymptomatic vs. symptomatic). The third assessed differential sensitivity by racialised groups. Specificity remained fixed at 90%, and true risk ratios (RRs) for the association with hypertension were set at 1.3 and 1.8.

Results: Decreasing sensitivity biased results towards the null, with low-sensitivity methods (e.g., self-report) showing the largest bias and high-sensitivity methods (e.g., transvaginal ultrasonography) the least bias. At low fibroid prevalence (5%), even gold-standard ascertainment introduced bias due to imperfect specificity, whereas this concern diminished at higher prevalence. Assuming a dose-response relationship between fibroids and hypertension based on symptom status, results remained biased towards the null unless sensitivity was 100% and prevalence was high (60%); bias was most pronounced at low prevalence. When only symptomatic fibroids were associated with hypertension, increasing sensitivity biased results away from the null by capturing more asymptomatic cases. Studies using low-sensitivity methods may fail to identify a true effect among Black females while identifying it among White females, potentially exacerbating disparities. Detection bias, where those with fibroids are more likely to have hypertension detected, could result in bias away from the null.

Conclusions: Underdiagnosis of uterine fibroids can bias results towards the null, particularly with self-report or modest effect estimates, potentially obscuring true effects. When only symptomatic fibroids were associated with the outcome, the bias was away from the null. Results varied by symptom status and race, highlighting the need to prioritise sensitive ascertainment methods, employ sensitivity analyses and improve reliability across diverse gynecologic conditions and health disparities.

Background: The occurrence of reproductive or pregnancy events, such as severe maternal morbidity (SMM), may reveal a predisposition to chronic disease and premature mortality. However, most studies have examined these exposures without considering their timing, severity, or recurrence.

Objectives: We propose using a weighted cumulative exposure (WCE) modelling approach to flexibly describe the relationship between reproductive events and longer-term health outcomes in a longitudinal cohort of pregnant women.

Methods: Application of the WCE modelling approach is accomplished in three steps. First, relative weights are estimated from a multivariable Cox proportional hazards model corresponding to the association of each reproductive risk factor with a given health outcome. Then, a longitudinal dataset is constructed in which all reproductive predictors are recorded at regular intervals (every 3 months), beginning 42 days after each woman's first birth in the cohort and ending at an outcome or censoring event. A new multivariable Cox model applied to this longitudinal dataset, incorporating time-varying WCE-derived reproductive risk scores along with simple time-varying reproductive and non-reproductive predictors, is estimated. Finally, adjusted WCE-based hazard ratios (HR) associated with different reproductive event exposure histories are calculated.

Results: In the cohort of 1,992,972 births in Canada (excluding Quebec), 2008-2021, with mean (SD) follow-up time in the longitudinal dataset of 7.3 ± 3.8 years, we propose to use the WCE approach to predict outcomes such as premature cardiovascular disease (16,846 cardiovascular hospitalisations observed, or 1.19 per 1000 person-years).

Conclusions: Use of flexible WCE modelling to quantify risks of pregnancy events such as SMM, adjusted for reproductive and non-reproductive CVD risk factors, will account for variation in timing and severity of these events and will capture their cumulative effects across a woman's reproductive trajectory. This approach can refine estimates of etiologic associations and inform novel clinical prediction models with the potential to predict postpartum long-term health outcomes for a given woman based on her unique reproductive history.

Background: Physical activity has time-varying associations with injury risk among children. While previous activity may predispose to injury through tissue damage, fatigue and insufficient recovery, it may protect against injury by strengthening tissues and improving fitness and skills. It is unclear what the relevant time window and relative importance of past activity are with regard to current injury risk in children.

Objectives: The objectives of this study were to assess how previous activity patterns are associated with injury risk among children.

Methods: Our data source was the Childhood Health, Activity, and Motor Performance School Study Denmark (CHAMPS-DK), a prospective cohort study of Danish school children conducted between 2008 and 2014. We applied flexible weighted cumulative exposure methods within a Cox proportional hazards model to estimate the time-varying association between the number of weekly activity sessions and time-to-first injury in each school year. We estimated several models with varying time windows and compared goodness-of-fit.

Results: Out of 1667 study participants, 986 (59.1%) were injured at least once, with a total of 1752 first injuries across school years. The best-fitting model included 20 weeks of past physical activity. Higher levels of activity performed 10-20 weeks ago were associated with decreased injury risk, while higher levels of activity performed 2-9 weeks ago were associated with higher injury risks. Compared to those who remained minimally active for the entire past 20-week period, children who were highly active in the past 10 weeks after being minimally active 11-20 weeks ago had an injury hazard ratio of 1.63 (95% confidence interval 1.18, 2.23).

Conclusions: Flexible weighted cumulative exposure methods suggest a complex temporal relationship between past physical activity history and injury in children.