Many studies, but little certainty about the effects of statin discontinuation on outcomes

Abstract

In this issue of the Journal of the American Geriatrics Society, Piexoto et al. conducted a systematic review of studies of statin discontinuation on clinical outcomes.¹ They identified only one randomized trial of statin discontinuation, conducted in people near the end of life, which found no difference in 60-day mortality or 1-year cardiovascular mortality among people who discontinued statins compared with those who continued statins. In contrast, among 35 nonrandomized studies among people not near the end of life, statin discontinuation was associated with a higher risk of all-cause mortality, cardiovascular mortality, and cardiovascular events. However, the authors noted concerns around bias of confounding by indication, along with concerns about imprecision, inconsistency, and heterogeneity. Together, the findings from observational studies were evaluated as having a high degree of uncertainty and bias, leaving providers and their patients with little useful information outside of the end-of-life setting.

There is discordance among the major Northern American and European guidelines on the evaluation of the benefit of statins in older adults, especially for primary prevention.² This is compounded by the challenge of accurate prediction of cardiovascular events in older adults, as a risk-based prevention strategy is the cornerstone of many of the guideline recommendations. Adding to the complexity, there is also insufficient evidence to capture potential harms of statin use due to the limited representation of older adults in large randomized statin trials. Further, limited evidence on statins and patient-centered outcomes such as frailty or statin-associated physical or cognitive changes exists, although one modestly sized trial demonstrated worsening decline in energy and exertional fatigue among persons randomized to statins.³ Taken collectively, these factors have contributed to a growing interest in medication discontinuation or dose reduction, otherwise referred to as deprescribing, of statins in older adults.

Piexoto aimed to address this evidence gap by synthesizing the evidence for statin discontinuation, but their systematic review only highlights the challenges in estimating medication effects in observational studies.¹ Despite recent advancements in observational research methodologies, the majority of studies included in the review were assessed to have a serious risk of bias. The primary limitation of observational studies, in contrast to randomized trials, is that of confounding, or which occurs when the populations who discontinue statins are systematically different than those who continue statins. This limitation is especially challenging for studies of medication deprescribing and mortality, as limited life expectancy is a common reason for medication review and deprescribing.⁴ The ACC/AHA guideline notes that, “it may be reasonable to stop statin therapy when functional decline (physical or cognitive), multimorbidity, frailty, or reduced life expectancy limits the potential benefits of statin therapy.”⁵ Methods research has demonstrated that the confounding structure can be stronger when the risk factors for the outcome of interest are also the indicating reason for the medication change.⁶ Thus, it would be expected that studies of deprescribing and death would be strongly confounded by the providers’ perception of limited life expectancy. This can be challenging to capture in an electronic health record, and thus control for in a statistical model, as many providers who care for older adults use clinical intuition to predict survial.^{7, 8} In addition to internal validity concerns, the included studies highlight significant research gaps in understanding statin deprescribing among multimorbid or frail patients, or its subsequent impact on patient-centered endpoints. Moreover, while these studies predominantly focused on characterizing the cardiovascular risks associated with deprescribing statins, the potential benefits, such as prevention of adverse reactions (e.g., myopathy, diabetes) or improved quality of life, remain unexamined.

Despite ongoing efforts, clinical trials are unlikely to fully resolve all questions pertaining to statin deprescribing in older adults. Evidentiary uncertainties will likely persist concerning primary versus secondary prevention, effects on geriatric-specific endpoints like cognitive and physical functioning, and applicability to a clinically diverse population across varying levels of multimorbidity, frailty, and conditions that are often excluded from trials, like dementia and severe chronic kidney disease. Thus, for such research questions in which trial evidence is limited, investigators have turned to other contemporary methods to better estimate causal effects. These powerful tools can address some of the common limitations of observational studies, but should be interpreted within the larger body of evidence. Natural experiments combined with econometric methods such as instrumental variables, regression discontinuity, or difference-in-difference studies leverage a change or threshold at which prescribing decisions are made. For example, we have demonstrated that leveraging a UK guideline to prescribe statins among patients with a 10-year cardiovascular disease risk of 20% or higher can be used to estimate trial-like effects of statins on myocardial infarction based on a regression discontinuity model.⁹ This method works because people slightly above and below 20% are similar with regard to other confounding factors. Another method, target trial emulation, works by specifying a hypothetical trial, and using observational data to mimic this design.¹⁰ Newer statistical methods that can better account for confounding, such as the doubly robust methods, can be applied when strong confounding by indication is a concern.¹¹ The target trial approach has shown strong concordance with randomized controlled trials. When evaluated in 16 studies with close emulation of design and measurements, the Pearson correlation between actual trials and target trials was 0.93 (95% CI: 0.79, 0.97).¹²

A valiant effort, the systematic review by Piexoto et al. sheds faint light on the decision of whether statin discontinuation is safe. Their article highlights the limited evidence and a variety of issues with observational studies. The evidence remains limited for statin deprescribing, and patients and their providers have little reliable evidence to inform their discussions around this important issue. As we wait for the results of ongoing deprescribing trials, which may not fully address all outstanding questions, robust observational studies using state-of-the-art design and statistical methods may better inform shared decision-making.

Dr. Odden drafted the manuscript and Dr. Dave provided critical revision of the manuscript for important intellectual content.

The authors declare no conflicts of interest.

The funder had no role in the design of the study, interpretation of results, or preparation of the manuscript. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health or the Veterans' Health Association.

This work was supported in part by the National Institute on Aging (2R01AG062568-02).