Double-blinded, randomized, controlled trials (RCT) are the gold standard to evaluate the efficacy of new treatments. To generalize RCT findings and to implement them in clinical practice, it is of utmost importance that the study population is representative of the general patient population. Previous reports have indicated that especially elderly adults and patients with comorbid conditions have been underrepresented in RCTs. Also, women have been underrepresented in cardiovascular (CV) trials, partly because of higher age and a larger burden of comorbidities.
In the current issue of Circulation, Lau et al address the study drug discontinuation problem in RCTs conducted by the TIMI (Thrombolysis in Myocardial Infarction) group. The most remarkable finding is that more than 1 in 4 men (26%; range, 11%–37%) and almost 1 in 3 women (30%; range, 16%–49%) discontinued the study drug prematurely. In addition, the study shows that in 10 of 11 of these phase 3 to 4 RCTs, women discontinued the study drug to a higher degree than men, which was also true for study withdrawal in 7 of these RCTs. In pooled patient-level data, women had 29% higher overall odds of study drug discontinuation versus men (odds ratio [OR], 1.29 [95% CI, 1.26–1.32]). Female sex was still associated with discontinuation after adjusting for age, comorbidities, and other possible confounders (OR, 1.18 [95% CI, 1.16–1.21]). The study adds an important piece of information to current knowledge, showing that there are similar sex differences in drug discontinuation in RCTs as are in real-life observational studies. Unique aspects of the current study include a very large study population (135 879 men and 51 812 women) and a thorough evaluation of the association between sex and study drug discontinuation.
Importantly, the sex difference in discontinuation was not explained by adverse events. Only one-third of the patients who stopped their treatment prematurely did so because of adverse events (33% of women and 34% of men), and the risk of discontinuation (women vs men) was similar in the placebo group (OR, 1.20 [95% CI, 1.13–1.27]) and in the active study drug group (OR, 1.23 [95% CI, 1.17–1.30]). Moreover, the observed sex difference was not restricted to a certain class of medication, because the effect of sex was consistent through a wide range of study drugs such as antiplatelets, anticoagulants, lipid-lowering drugs, and antidiabetic agents. Using this large dataset, the authors were also able to show that other previously suggested explanations, sex differences in age and comorbidity, did not explain the difference in drug discontinuation, as it remained after multivariable adjustment.
Another interesting finding that merits further studies was the effect modification by geographic area from which study participants were recruited. There was a large sex difference in drug adherence in North America (OR, 1.35 [95% CI, 1.30–1.41]), whereas the difference was moderate in Europe/Middle East/Africa (OR, 1.13 [95% CI, 1.09–1.17]) and Asia/Pacific (OR, 1.13 [95% CI, 1.03–1.23]), and no sex difference at all was found among study participants from South/Central America (OR, 1.00 [95% CI, 0.92–1.08]). Because countries within these areas have very different health care and insurance systems, socioeconomic as well as cultural factors may have influenced the results. Nonstudy related costs, transportation facilities, and family obligations, as well as concerns about study drug safety and confidence in health care systems, may differ between the geographic areas.
In several observational studies, women have been less likely than men to adhere to prescribed medication, including CV drugs.[3,4] This may contribute to worse prognoses, as the associations between adherence to CV medications and outcomes are known. A meta-analysis by Chowdhury et al of 44 prospective epidemiological studies found 45% (hazard ratio [HR], 0.55 [95% CI, 0.46–0.67]) and 29% (HR, 0.71 [95% CI, 0.64–0.78]) lower risks of mortality associated with adherence to statin and antihypertensive therapy, respectively. In a study from the SWEDEHEART (Swedish Web-system for Enhancement and Development of Evidence-based Care in Heart Disease Evaluated According to Recommended Therapies) registry on Swedish coronary artery bypass surgery patients, adherence to statins (HR, 0.56 [95% CI, 0.52–0.60]), renin-angiotensin-aldosterone system inhibitors (HR, 0.78 [95% CI, 0.73–0.84]), and platelet inhibitors (HR, 0.74 [95% CI, 0.69–0.81]) was associated with lower mortality after multivariable adjustments, whereas no association between β-blockers and outcome was found. Women adhered less to statin therapy but more to β-blockers. In the CHARM (Candesartan in Heart Failure: Assessment of Reduction in Mortality and Morbidity) program, 80% adherence to study medication was associated with 45% lower risk of death during follow-up. Female sex was associated with nonadherence to candesartan. Interestingly, nonadherence was associated with poor prognosis in both the active drug and in the placebo arms. Thus, this association is probably partly because of residual confounding as adherence to study drug is linked to adherence to other prescribed medications and healthy lifestyle factors.
It has previously been shown that 40% to 60% of CV patients have poor adherence to their prescribed drugs.[5,6] However, reasons for nonadherence are complex and multifactorial. Socioeconomic status is a predictor of nonadherence, impacting affordability, health literacy, and health care access. Socioeconomic status may play a role regarding sex differences in drug adherence, because women generally have lower income than men. Supporting this, significant but modest improvement of adherence was observed when the individual patient's drug cost was waived. In contrast, a meta-analysis by Lewey et al found that sex differences persisted in statin adherence after adjusting for income and insurance status. Likewise, in a Finnish study performed within a health care system with economic coverage for medication, low socioeconomic status—regardless of indicator (income, education or labor market status)—was associated with statin nonadherence in men, whereas associations were weak and mainly nonsignificant in women.
Another reason for sex differences in adherence could be perceived side effects, although this was not shown in the study by Lau et al. The sexes differ in pharmacodynamic and pharmacokinetics, and women experience an average 1.5- to 1.7-fold higher incidence of adverse drug reactions on CV drugs, which tend to be more severe than in men. Drug-induced torsade de pointes on antiarrhythmics, bleeding complications on anticoagulants and antiplatelet drugs, and myalgia on statins are all shown to be more common in women.
There is a lack of data regarding differences in drug adherence between women and men in RCTs. Drug adherence in a trial testing the efficacy of a certain drug is obviously pivotal. A high degree of study drug discontinuation in a clinical trial increases the risk of a type II error, or failure to detect a true treatment effect, as well as the risk of overestimating drug safety. Supporting side effects as important factors to explain sex differences in nonadherence, discontinuation rates attributable to adverse events were numerically higher in women than men in 4 of 6 RCTs on atorvastatin treatment, even if a formal treatment-by-gender interaction test was significant only in the open label IDEAL (the Incremental Decrease in End Points Through Aggressive Lipid Lowering study) study. Factors associated with discontinuation were also evaluated in IMPROVE-IT (Improved Reduction of Outcomes: Vytorin Efficacy International Trial) from randomization through study end (median, 71.9 months). In addition to smoking, previous revascularization, hypertension, non-White race, and US location, female sex was independently associated with higher discontinuation rates after multivariable adjustment.
Although the study by Lau et al confirms sex differences in drug treatment discontinuation in RCTs, reasons behind the differences remain largely unknown. Several factors that could be associated with poor adherence, including education, income, occupation, health literacy, cultural aspects, care-giving, and social support were not registered in the included studies, and their impact on the results cannot be evaluated. These aspects should be further explored in future trials, to increase drug adherence and thereby improve study validity in both men and women. The Non-adherence Academic Research Consortium proposes that future cardiovascular RCTs use a framework for capturing nonadherence regarding type, decision maker, and reasons. We agree that the quality of future RCTs would be enhanced if nonadherence were better captured. The information of timing of permanent discontinuation is vital, as is the distinction between nonadherence (adherence often defined as taking the drug >80% of the time, based on pill count or pharmacy refill databases) and discontinuation (stopping the medication temporarily or permanently). We know the associations between nonadherence and clinical outcomes. The next step should be to better understand the underlying reasons for, as well as consistent reporting of, nonadherence and discontinuation in RCTs. This will help clinicians to understand the results and the impact on clinical outcomes in a real-world population. In parallel, novel digital solutions used in ongoing clinical trials with dosing reminders and possibilities for patients to instantly report side effects or outcomes, facilitating a more direct contact with the study team, may have the potential to improve adherence in both sexes. The authors of the current article in the Journal should be congratulated on their effort, revealing a potentially important problem regarding drug adherence, including significant differences between women and men.
Circulation. 2021;143(7):696-698. © 2021 American Heart Association, Inc.