Parametric-Regression-Based Causal Mediation Analysis of Binary Outcomes and Binary Mediators: Moving Beyond the Rareness or Commonness of the Outcome

Associations between endometriosis and adverse pregnancy and perinatal outcomes: a population-based cohort study

PURPOSE

To examine the association between endometriosis and adverse pregnancy and perinatal outcomes (preeclampsia, placenta previa, and preterm birth).

METHODS

A population-based retrospective cohort study was conducted among 468,778 eligible women who contributed 912,747 singleton livebirths between 1980 and 2015 in Western Australia (WA). We used probabilistically linked perinatal and hospital separation data from the WA data linkage system's Midwives Notification System and Hospital Morbidity Data Collection databases. We used a doubly robust estimator by combining the inverse probability weighting with the outcome regression model to estimate adjusted risk ratios (RR) and 95% confidence intervals (CIs).

RESULTS

There were 19,476 singleton livebirths among 8874 women diagnosed with endometriosis. Using a doubly robust estimator, we found pregnancies in women with endometriosis to be associated with an increased risk of preeclampsia with RR of 1.18, 95% CI 1.11-1.26, placenta previa (RR 1.59, 95% CI 1.42-1.79) and preterm birth (RR 1.45, 95% CI 1.37-1.54). The observed association persisted after stratified by the use of Medically Assisted Reproduction, with a slightly elevated risk among pregnancies conceived spontaneously.

CONCLUSIONS

In this large population-based cohort, endometriosis is associated with an increased risk of preeclampsia, placenta previa, and preterm birth, independent of the use of Medically Assisted Reproduction. This may help to enhance future obstetric care among this population.

An empirical evaluation of approximate and exact regression-based causal mediation approaches for a binary outcome and a continuous or a binary mediator for case-control study designs

BACKGROUND

In the causal mediation analysis framework, several parametric regression-based approaches have been introduced in past years for decomposing the total effect of an exposure on a binary outcome into a direct effect and an indirect effect through a target mediator. In this context, a well-known strategy involves specifying a logistic model for the outcome and invoking the rare outcome assumption (ROA) to simplify estimation. Recently, exact estimators for natural direct and indirect effects have been introduced to circumvent the challenges prompted by the ROA. As for the approximate approaches relying on the ROA, these exact approaches cannot be used as is on case-control data where the sampling mechanism depends on the outcome.

METHODS

Considering a continuous or a binary mediator, we empirically compare the approximate and exact approaches using simulated data under various case-control scenarios. An illustration of these approaches on case-control data is provided, where the natural mediation effects of long-term use of oral contraceptives on ovarian cancer, with lifetime number of ovulatory cycles as the mediator, are estimated.

RESULTS

In the simulations, we found few differences between the performances of the approximate and exact approaches when the outcome was rare, both marginally and conditionally on variables. However, the performance of the approximate approaches degraded as the prevalence of the outcome increased in at least one stratum of variables. Differences in behavior were also observed among the approximate approaches. In the data analysis, all studied approaches were in agreement with respect to the natural direct and indirect effects estimates.

CONCLUSIONS

In the case where a violation of the ROA applies or is expected, approximate mediation approaches should be avoided or used with caution, and exact estimators favored.

Influence of preterm birth on the association between gestational diabetes mellitus and childhood developmental vulnerability: a causal mediation analysis

BACKGROUND

Epidemiological studies examining the direct and indirect effects of gestational diabetes mellitus (GDM) on offspring early childhood developmental vulnerability are lacking. Therefore, the aims of this study were to estimate the direct and indirect effects of GDM (through preterm birth) on early childhood developmental vulnerability.

METHODS

We conducted a retrospective population-based cohort study on the association between gestational diabetes mellitus and early childhood developmental vulnerability in children born in Western Australia (WA) using maternal, infant and birth records from the Midwives Notification, Hospitalizations, Developmental Anomalies, and the Australian Early Development Census (AEDC) databases. We used two aggregated outcome measures: developmentally vulnerable on at least one AEDC domain (DV1) and developmentally vulnerable on at least two AEDC domains (DV2). Causal mediation analysis was applied to estimate the natural direct (NDE), indirect (NIE), and total (TE) effects as relative risks (RR).

RESULTS

In the whole cohort (n = 64,356), approximately 22% were classified as DV1 and 11% as DV2 on AEDC domains. Estimates of the natural direct effect suggested that children exposed to GDM were more likely to be classified as DV1 (RR = 1.20, 95% CI: 1.10-1.31) and DV2 (RR = 1.34, 95% CI: 1.19-1.50) after adjusting for potential confounders. About 6% and 4% of the effect of GDM on early childhood developmental vulnerability was mediated by preterm birth for DV1 and DV2, respectively.

CONCLUSION

Children exposed to gestational diabetes mellitus were more likely to be developmentally vulnerable in one or more AEDC domains. The biological mechanism for these associations is not well explained by mediation through preterm birth.

Eliminating children's tobacco smoke exposure: a pathway to bioverified abstinence among low-income maternal smokers in the Babies Living Safe and Smokefree (BLiSS) trial

BACKGROUND

Identifying behavioral pathways to smoking cessation in high-risk populations, such as low-income maternal smokers, could reduce tobacco disparities. The previous "BLiSS" multilevel intervention trial demonstrated efficacy of the BLiSS intervention in facilitating low-income maternal smokers' bioverified abstinence. This present study examined four putative pathways measured at 3-month end of treatment (Time 2) that could account for the observed intervention effect on smoking abstinence through 12 months (Time 2 - Time 3).

METHODS

Nutritionists in community clinics delivering safety net nutrition promotion programs across Philadelphia, Pennsylvania, USA, were trained by trial principal investigators to deliver a brief tobacco intervention informed by the American Academy of Pediatrics best practice guidelines ("Ask, Advise, Refer [AAR]"). After referral, 396 eligible participants were randomized to either a multimodal behavioral intervention (AAR + MBI) or a parallel attention control (AAR + control). Random effects regression analysis tested mediation.

RESULTS

Elimination of children's tobacco smoke exposure (TSE) at Time 2 was the only significant mediator of longitudinal smoking abstinence through Time 3. AAR + MBI mothers were more likely to eliminate their children's TSE by Time 2 (OR = 2.11, 95%CI 1.30, 3.42), which was significantly associated with Time 3 abstinence (OR = 6.72, CI 2.28, 19.80). Modeling showed a significant total effect of AAR + MBI on abstinence (OR = 6.21, CI 1.86, 20.71), a direct effect of AAR + MBI on abstinence (OR = 4.80, CI 1.45, 15.94) and an indirect effect through TSE elimination (OR = 1.29, CI 1.06, 1.57).

CONCLUSIONS

Integrating smoking cessation interventions with counseling prior to the quit attempt that is designed to facilitate adoption of smokefree home policies and efforts to eliminate children's TSE could enhance the likelihood of long-term abstinence in populations of smokers with elevated challenges quitting smoking.

It goes both ways: The relationship between anxiety and mild cognitive impairment

OBJECTIVE

To investigate the relationship between anxiety and mild cognitive impairment (MCI), and whether it is mediated by perceived stress, at the population level.

METHOD AND DESIGN

In a longitudinal study of 368 adults aged 65+ from a population-based cohort, we annually assessed anxiety symptoms (GAD-7), perceived stress (PSS-4), and ratings on the Clinical Dementia Rating (CDR®), where CDR = 0.5 was operationalized as MCI. Examining data from three consecutive assessment waves, we first determined the associations between anxiety at the first wave with MCI at the third wave, and vice versa. We then used mediation analyses to determine whether the pathways in both directions were mediated by perceived stress at the second wave, adjusting for demographics and other relevant covariates.

RESULTS

We confirmed significant bidirectional longitudinal associations between anxiety and MCI. Perceived stress was detected as a significant mediator for both pathways between anxiety and MCI, explaining 37.1% of the total effect (TE) of anxiety on incident MCI while conversely explaining 27.1% of the TE of MCI on anxiety.

CONCLUSIONS

A bidirectional relationship with a 2-year lag between anxiety and MCI was mediated through perceived stress. Clinicians should be sensitive both to potential consequent anxiety when patients present with cognitive impairment, and to potential incipient MCI when the presenting complaint is anxiety. Managing stress may help mitigate adverse outcomes.

An exact regression-based approach for the estimation of natural direct and indirect effects with a binary outcome and a continuous mediator

In the causal mediation framework, a number of parametric regression-based approaches have been introduced in recent years for estimating natural direct and indirect effects for a binary outcome in an exact manner, without invoking simplifying assumptions based on the rareness or commonness of the outcome. However, most of these works have focused on a binary mediator. In this article, we aim at a continuous mediator and introduce an exact approach for the estimation of natural effects on the odds ratio, risk ratio, and risk difference scales. Our approach relies on logistic and linear models for the outcome and mediator, respectively, and uses numerical integration to calculate the nested counterfactual probabilities underlying the definition of natural effects. Formulas for the delta method standard errors for all effects estimators are provided. The performance of our proposed exact estimators was evaluated in simulation studies that featured scenarios with different levels of outcome rareness/commonness, including a marginally but not conditionally rare outcome scenario. Furthermore, we evaluated the merit of Firth's penalization to mitigate the bias in the logistic regression coefficients estimators for the smallest outcome prevalences and sample sizes investigated. Using a SAS macro provided, we implemented our approach to assess the effect of placental abruption on low birth weight mediated by gestational age. We found that our exact natural effects estimators worked properly in both simulated and real data applications.