Assessing the Most Vulnerable Subgroup to Type II Diabetes Associated with Statin Usage: Evidence from Electronic Health Record Data

Xinzhou Guo, Waverly Wei, Molei Liu, Tianxi Cai, Chong Wu, Jingshen WangJournal of the American Statistical Association 2022 · DOI

TL;DR

This paper tackles the concern that statins, widely used for heart disease, might increase the risk of Type II Diabetes (T2D). Using electronic health records and a novel causal inference framework, they identify that women with a high genetic predisposition to T2D are most vulnerable to this side effect. This work offers a robust statistical approach to pinpointing at-risk subgroups, crucial for personalized medicine.

What Problem Does This Paper Address?

Statins are a cornerstone in managing cardiovascular disease, but a growing body of evidence suggests a potential link to an increased risk of developing Type II Diabetes (T2D). However, existing research lacks robust evidence to definitively say if and which specific populations are most vulnerable to this T2D risk when taking statins. This is a critical gap because it hinders precise clinical decision-making regarding statin prescription, especially for individuals with pre-existing T2D risk factors. The challenge lies in analyzing observational data to establish causal links while accounting for complex confounding factors and the "winner's curse" bias inherent in subgroup analysis.

Methods

The authors introduce a novel data analysis pipeline and statistical methodology rooted in causal inference. Key components include:

  • Rigorous Causal Framework: They employ a genetic variant (rs12916-T) as a surrogate treatment variable for statin usage. This genetic variant, located in the HMGCR gene (the target of statins), allows them to establish clear temporal precedence (genetics predating T2D onset) and mitigate concerns about reverse causation and unmeasured confounding.
  • Subgroup Analysis: They pre-define six subgroups based on sex and baseline T2D genetic risk profiles (categorized by the number of risk alleles). This approach aims to identify heterogeneous treatment effects.
  • Bootstrap Calibration: To address the "winner's curse" bias and regularization bias in estimating the effect in the most vulnerable subgroup, they develop a bootstrap-assisted R-Split procedure. This method provides asymptotically sharp confidence intervals and debiased estimates, even in the presence of high-dimensional covariates.

Datasets

The study leverages data from the Partner Health System (PHS) biobank and electronic health records (EHR). This combined dataset includes:

  • Electronic Health Records (EHR): Containing patient demographics, diagnoses, and genetic information.
  • Biobank Data: Providing genotype information.

The combined dataset forms an EHR virtual cohort of 17,023 subjects with 337 features. The T2D status was defined using a validated Multimodal Automated Phenotyping (MAP) algorithm.

Key Insights and Ideas

  • Genetic Proxy for Statin Use: Using a genetic variant as a surrogate for statin use is a clever way to overcome limitations in observational data, particularly regarding temporal ordering and unmeasured confounding.
  • Targeting the "Most Vulnerable": The focus on identifying the subgroup with the largest adverse treatment effect is crucial for precision medicine, allowing for targeted interventions and risk mitigation.
  • Debiasing Complex Inference: The development of a bootstrap calibration procedure to simultaneously address regularization and winner's curse bias is a significant methodological contribution, enabling more reliable inference in high-dimensional subgroup analyses.
  • Biological Plausibility: The finding that high genetic risk in females is linked to increased T2D risk from statins aligns with biological understanding and prior research, lending credibility to the results.

Key Results

The study found that while the overall adverse effect of statin usage on developing T2D is marginal, females with a high genetic predisposition to T2D are the most vulnerable subgroup. Specifically, these individuals have the highest risk of developing T2D when taking statins. The proposed bootstrap-assisted R-Split method confirmed the statistical significance of this finding, providing asymptotically sharp confidence intervals.

Limitations and Future Work

  • Treatment Dosage: The study uses a binary indicator for statin use (presence of the rs12916-T allele) and cannot investigate the effect of statin dosage.
  • Generalizability: While the MAP algorithm for T2D status is validated, further confirmation from clinical trials is warranted for generalizing these findings.
  • Data-Adaptive Subgroups: The current study focuses on pre-defined subgroups. Future research could extend the methodology to data-adaptively identified subgroups, which might offer new insights but also introduce additional bias considerations.
  • Other Causal Pathways: The paper discusses potential causal pathways, including those involving statin use as a mediator or collider, but due to data limitations (e.g., lack of temporal information on statin use), these could not be fully explored.

The authors also suggest that their methodology could be extended to investigate other vulnerable subgroups beyond the most vulnerable one.

Summary updated

Abstract

There have been increased concerns that the use of statins, one of the most commonly prescribed drugs for treating coronary artery disease, is potentially associated with the increased risk of new-onset Type II diabetes (T2D). Nevertheless, to date, there is no robust evidence supporting as to whether and what kind of populations are indeed vulnerable for developing T2D after taking statins. In this case study, leveraging the biobank and electronic health record data in the Partner Health System, we introduce a new data analysis pipeline and a novel statistical methodology that address existing limitations by (i) designing a rigorous causal framework that systematically examines the causal effects of statin usage on T2D risk in observational data, (ii) uncovering which patient subgroup is most vulnerable for developing T2D after taking statins, and (iii) assessing the replicability and statistical significance of the most vulnerable subgroup via a bootstrap calibration procedure. Our proposed approach delivers asymptotically sharp confidence intervals and debiased estimate for the treatment effect of the most vulnerable subgroup in the presence of high-dimensional covariates. With our proposed approach, we find that females with high T2D genetic risk are at the highest risk of developing T2D due to statin usage.
BibTeX
@article{guo2022assessing,
  author = {Xinzhou Guo and Waverly Wei and Molei Liu and Tianxi Cai and Chong Wu and Jingshen Wang},
  title = {{Assessing the Most Vulnerable Subgroup to Type II Diabetes Associated with Statin Usage: Evidence from Electronic Health Record Data}},
  journal = {Journal of the American Statistical Association},
  year = {2022},
  doi = {10.1080/01621459.2022.2157727},
}
Figure