Causal inference in the face of competing events

Influence of incomplete death information on cumulative risk estimates in US claims data

Administrative claims databases often do not capture date or fact of death, so studies using these data may inappropriately treat death as a censoring event-equivalent to other withdrawal reasons-rather than a competing event. We examined 1-, 3-, and 5-year inverse-probability-of-treatment weighted cumulative risks of a composite cardiovascular outcome among 34 527 initiators of telmisartan (exposure) and ramipril (referent), who were aged ≥55 years, in Optum (United States) claims data from 2003 to 2020. Differences in cumulative risks of the cardiovascular endpoint due to censoring of death (cause-specific), as compared with treating death as a competing event (subdistribution), increased with greater follow-up time and older age, where event and mortality risks were higher. Among ramipril users, 5-year cause-specific and subdistribution cumulative risk estimates per 100, respectively, were 16.4 (95% CI, 15.3-17.5) and 16.2 (95% CI, 15.1-17.3) among ages 55-64 (difference = 0.2) and were 43.2 (95% CI, 41.3-45.2) and 39.7 (95% CI, 37.9-41.4) among ages ≥75 (difference = 3.6). Plasmode simulation results demonstrated the differences in cause-specific versus subdistribution cumulative risks to increase with increasing mortality rate. We suggest researchers consider the cohort's baseline mortality risk when deciding whether real-world data with incomplete death information can be used without concern. This article is part of a Special Collection on Pharmacoepidemiology.

Navigating the Complexities of Artificial Intelligence-Enabled Real-World Data Collection for Oncology Pharmacovigilance

This new editorial discusses the promise and challenges of successful integration of natural language processing methods into electronic health records for timely, robust, and fair oncology pharmacovigilance.

Differential Participation, a Potential Cause of Spurious Associations in Observational Cohorts in Environmental Epidemiology

Differential participation in observational cohorts may lead to biased or even reversed estimates. In this article, we describe the potential for differential participation in cohorts studying the etiologic effects of long-term environmental exposures. Such cohorts are prone to differential participation because only those who survived until the start of follow-up and were healthy enough before enrollment will participate, and many environmental exposures are prevalent in the target population and connected to participation via factors such as geography or frailty. The relatively modest effect sizes of most environmental exposures also make any bias induced by differential participation particularly important to understand and account for. We discuss key points to consider for evaluating differential participation and use causal graphs to describe two example mechanisms through which differential participation can occur in health studies of long-term environmental exposures. We use a real-life example, the Canadian Community Health Survey cohort, to illustrate the non-negligible bias due to differential participation. We also demonstrate that implementing a simple washout period may reduce the bias and recover more valid results if the effect of interest is constant over time. Furthermore, we implement simulation scenarios to confirm the plausibility of the two mechanisms causing bias and the utility of the washout method. Since the existence of differential participation can be difficult to diagnose with traditional analytical approaches that calculate a summary effect estimate, we encourage researchers to systematically investigate the presence of time-varying effect estimates and potential spurious patterns (especially in initial periods in the setting of differential participation).

Accounting for Competing Events When Evaluating Long-Term Outcomes in Survivors of Critical Illness

The clinical trajectory of survivors of critical illness after hospital discharge can be complex and highly unpredictable. Assessing long-term outcomes after critical illness can be challenging because of possible competing events, such as all-cause death during follow-up (which precludes the occurrence of an event of particular interest). In this perspective, we explore challenges and methodological implications of competing events during the assessment of long-term outcomes in survivors of critical illness. In the absence of competing events, researchers evaluating long-term outcomes commonly use the Kaplan-Meier method and the Cox proportional hazards model to analyze time-to-event (survival) data. However, traditional analytical and modeling techniques can yield biased estimates in the presence of competing events. We present different estimands of interest and the use of different analytical approaches, including changes to the outcome of interest, Fine and Gray regression models, cause-specific Cox proportional hazards models, and generalized methods (such as inverse probability weighting). Finally, we provide code and a simulated dataset to exemplify the application of the different analytical strategies in addition to overall reporting recommendations.

Estimating Breast Cancer Overdiagnosis After Screening Mammography Among Older Women in the United States

BACKGROUND

Overdiagnosis is increasingly recognized as a harm of breast cancer screening, particularly for older women.

OBJECTIVE

To estimate overdiagnosis associated with breast cancer screening among older women by age.

DESIGN

Retrospective cohort study comparing the cumulative incidence of breast cancer among older women who continued screening in the next interval with those who did not. Analyses used competing risk models, stratified by age.

SETTING

Fee-for-service Medicare claims, linked to the SEER (Surveillance, Epidemiology, and End Results) program.

PATIENTS

Women 70 years and older who had been recently screened.

MEASUREMENTS

Breast cancer diagnoses and breast cancer death for up to 15 years of follow-up.

RESULTS

This study included 54 635 women. Among women aged 70 to 74 years, the adjusted cumulative incidence of breast cancer was 6.1 cases (95% CI, 5.7 to 6.4) per 100 screened women versus 4.2 cases (CI, 3.5 to 5.0) per 100 unscreened women. An estimated 31% of breast cancer among screened women were potentially overdiagnosed. For women aged 75 to 84 years, cumulative incidence was 4.9 (CI, 4.6 to 5.2) per 100 screened women versus 2.6 (CI, 2.2 to 3.0) per 100 unscreened women, with 47% of cases potentially overdiagnosed. For women aged 85 and older, the cumulative incidence was 2.8 (CI, 2.3 to 3.4) among screened women versus 1.3 (CI, 0.9 to 1.9) among those not, with up to 54% overdiagnosis. We did not see statistically significant reductions in breast cancer-specific death associated with screening.

LIMITATIONS

This study was designed to estimate overdiagnosis, limiting our ability to draw conclusions on all benefits and harms of screening. Unmeasured differences in risk for breast cancer and differential competing mortality between screened and unscreened women may confound results. Results were sensitive to model specifications and definition of a screening mammogram.

CONCLUSION

Continued breast cancer screening was associated with greater incidence of breast cancer, suggesting overdiagnosis may be common among older women who are diagnosed with breast cancer after screening. Whether harms of overdiagnosis are balanced by benefits and for whom remains an important question.

PRIMARY FUNDING SOURCE

National Cancer Institute.

High-Dimensional Survival Analysis: Methods and Applications

In the era of precision medicine, time-to-event outcomes such as time to death or progression are routinely collected, along with high-throughput covariates. These high-dimensional data defy classical survival regression models, which are either infeasible to fit or likely to incur low predictability due to over-fitting. To overcome this, recent emphasis has been placed on developing novel approaches for feature selection and survival prognostication. We will review various cutting-edge methods that handle survival outcome data with high-dimensional predictors, highlighting recent innovations in machine learning approaches for survival prediction. We will cover the statistical intuitions and principles behind these methods and conclude with extensions to more complex settings, where competing events are observed. We exemplify these methods with applications to the Boston Lung Cancer Survival Cohort study, one of the largest cancer epidemiology cohorts investigating the complex mechanisms of lung cancer.

A practical guide to handling competing events in etiologic time-to-event studies

Competing events are events that preclude the occurrence of the primary outcome. Much has been written on mainly the statistics behind competing events analyses. However, many of these publications and tutorials have a strong statistical tone and might fall short in providing a practical guide to clinician researchers as to when to use a competing event analysis and more importantly which method to use and why. Here we discuss the different target effects in the Fine-Gray and cause-specific methods using simple causal diagrams and provide strengths and limitations of both approaches for addressing etiologic questions. We argue why the Fine-Gray method might not be the best approach for handling competing events in etiological time-to-event studies.

Longitudinal effects of lipid indices on incident cardiovascular diseases adjusting for time-varying confounding using marginal structural models: 25 years follow-up of two US cohort studies

Background

This study assesses the effect of blood lipid indices and lipid ratios on cardiovascular diseases (CVDs) using inverse probability-of-exposure weighted estimation of marginal structural models (MSMs).

Methods

A pooled dataset of two US representative cohort studies, including 16736 participants aged 42-84 years with complete information at baseline, was used. The effect of each lipid index, including low-density lipoprotein cholesterol (LDL-C), high-density lipoprotein cholesterol (HDL-C), triglyceride (TG), ratios of TC/HDL-C, LDL-C/HDL-C, and TG/HDL-C on coronary heart disease (CHD) and stroke were estimated using weighted Cox regression.

Results

There were 1638 cases of CHD and 1017 cases of stroke during a median follow-up of 17.1 years (interquartile range: 8.5 to 25.7). Compared to optimal levels, the risk of CVD outcomes increased substantially in high levels of TC, LDL-C, TC/HDL-C, and LDL-C/HDL-C. If everyone had always had high levels of TC (≥240 mg/dL), risk of CHD would have been 2.15 times higher, and risk of stroke 1.35 times higher than if they had always had optimal levels (<200 mg/dL). Moreover, if all participants had been kept at very high (≥190 mg/dL) levels of LDL-C, risk of CHD would have been 2.62 times higher and risk of stroke would have been 1.92 times higher than if all participants had been kept at optimal levels, respectively. Our results suggest that high levels of HDL-C may be protective for CHD, but not for stroke. There was also no evidence of an adverse effect of high triglyceride levels on stroke.

Conclusions

Using MSM, this study highlights the effect of TC and LDL-C on CVD, with a stronger effect on CHD than on stroke. There was no evidence for a protective effect of high levels of HDL-C on stroke. Besides, triglyceride was not found to affect stroke.

Interaction Contrasts and Collider Bias

Previous papers have mentioned that conditioning on a binary collider would introduce an association between its causes in at least 1 stratum. In this paper, we prove this statement and, along with intuitions, formally examine the direction and magnitude of the associations between 2 risk factors of a binary collider using interaction contrasts. Among level one of the collider, 2 variables are independent, positively associated, and negatively associated if multiplicative risk interaction contrast is equal to, more than, and less than 0, respectively; the same results hold for the other level of the collider if the multiplicative survival interaction contrast, equal to multiplicative risk interaction contrast minus the additive risk interaction contrast, is compared with 0. The strength of the association depends on the magnitude of the interaction contrast: The stronger the interaction is, the larger the magnitude of the association will be. However, the common conditional odds ratio under the homogeneity assumption will be bounded. A figure is presented that succinctly illustrates our results and helps researchers to better visualize the associations introduced upon conditioning on a collider.

Incremental Propensity Score Effects for Time-fixed Exposures

When causal inference is of primary interest, a range of target parameters can be chosen to define the causal effect, such as average treatment effects (ATEs). However, ATEs may not always align with the research question at hand. Furthermore, the assumptions needed to interpret estimates as ATEs, such as exchangeability, consistency, and positivity, are often not met. Here, we present the incremental propensity score (PS) approach to quantify the effect of shifting each person's exposure propensity by some predetermined amount. Compared with the ATE, incremental PS may better reflect the impact of certain policy interventions and do not require that positivity hold. Using the Nulliparous Pregnancy Outcomes Study: monitoring mothers-to-be (nuMoM2b), we quantified the relationship between total vegetable intake and the risk of preeclampsia and compared it to average treatment effect estimates. The ATE estimates suggested a reduction of between two and three preeclampsia cases per 100 pregnancies for consuming at least half a cup of vegetables per 1,000 kcal. However, positivity violations obfuscate the interpretation of these results. In contrast, shifting each woman's exposure propensity by odds ratios ranging from 0.20 to 5.0 yielded no difference in the risk of preeclampsia. Our analyses show the utility of the incremental PS effects in addressing public health questions with fewer assumptions.