Clinical Trial Participation: Are we Studying the Patients we are Trying to Treat?

This editorial refers to ‘Is taking part in clinical trials good for your health? A cohort study’ by Andrew L. Clark et al. on page 1078 A cardiologist reads an article in a leading journal which reports positive findings for a randomized clinical trial testing a novel pharmaceutical agent for use in patients with heart failure. The trial is well designed and the results suggest a clinically relevant improvement in mortality and morbidity. Later that day, during a routine office visit, a patient reports worsening symptoms related to her heart failure, despite being on appropriate medications. She meets all of the inclusion criteria and none of the exclusion criteria listed in the study. Is this patient an appropriate candidate for this novel drug? Many of the improvements in outcomes for the treatment of cardiovascular disease are based on results of randomized clinical trials over the past 30 years. The scenario described above is an example of what most would describe as a reasonable situation in which to apply the results of a trial to a patient seen in clinical practice. An important benchmark in the ability to generalize findings from clinical trials is whether the patient being treated is similar to the patients who were studied. This benchmark has traditionally been defined by the explicit inclusion and exclusion criteria, which typically include characteristics such as disease severity, co-morbidities, age, and occasionally even race and gender. But even when this benchmark is met, can we assume that we are caring for patients who are similar to those in the trial population? In clinical practice, careful and judicious application of study findings is often difficult to achieve in the broader population. Study findings are often generalized to patients who do not even meet the explicit inclusion and exclusion criteria of the trial. For example, in a study examining the use of spironolactone soon after the publication of RALES, more than 75% of a representative sample of patients 65 years and older who were hospitalized for heart failure and met enrolment criteria were not prescribed the drug and over 30% of those who did not meet enrolment criteria were prescribed the medication. As a result, there is an obvious issue of whether trial results are applied to proper populations. In the case of spironolactone, there are concerns that the use of the medication in patients with renal dysfunction, who were excluded from the trial, may have led to an increase in the hospitalizations for hyperkalaemia.1 Even when explicit enrolment criteria are satisfied, as in the scenario above, concerns remain regarding the profile of patients selected to participate in clinical trials and how they may differ from the greater population of all patients eligible for the trial. Investigators have discretion in patient selection and may steer away from older, more complicated patients. We generally have little information about the recruitment process in clinical trials.2 Nevertheless, studies have shown that certain groups tend to be underrepresented in clinical trials. The persistent underrepresentation of older patients, women, and minorities in cardiovascular clinical trials is well-documented and continues, despite policies attempting to address the issue.3,4 The underrepresentation may also be a result of the reluctance of some patients to consent to join clinical trials. There is some evidence that women and minorities may have a decreased willingness to participate in clinical trials due to concerns about experiencing harm as a subject. Another study found that black participants expressed a higher degree of distrust towards medical researchers than white participants.5,6 These factors can result in a form of implicit exclusion criteria, which can result in a trial population that differs in clinical risk from patients in the general population.7 More evidence about the effect of patient preference is presented by Clark et al., who show that patients differ in their baseline characteristics and outcome solely based on whether or not they express a willingness to participate in a study.8 In their single-centre, retrospective study, 2332 consecutive patients presenting to a single heart failure clinic were asked if they would be willing to participate in future clinical trials. The 1867 patients who said that they would be willing to participate in future clinical trials had, as a group, worse left ventricular function, greater likelihood of ischaemic heart disease, and a higher likelihood of taking heart failure medications. In their multivariate model, agreeing to participate in a clinical trial was the strongest predictor of survival over the follow-up period of almost 5 years, regardless of whether or not a patient was eventually enrolled in a trial. The hazard ratio of 0.33 (CI 95% 0.26–0.40) for patients agreeing to take part in a clinical trial suggests that the relative risk of death was one-third the risk in the group not willing to take part in a clinical trial. The reason for the association between willingness to participate and survival is unclear. Perhaps, the explanation for this finding may be found in the field of health psychology wherein individuals who believe they have a greater locus of control, defined as one's perceived degree of control over events affecting them, may be more likely to participate in clinical trials. Prior work has suggested that a greater locus of control is associated with improved health outcomes and may be a modifiable risk factor.9 Regardless of the explanation, patients who are willing to participate in clinical trials appear to differ from those who are not willing to participate in ways that affect their outcomes. Although this truly underrepresented group of people unwilling to participate in clinical trials comprised ~20% of the subjects in this study, others have reported that >50% of potential subjects would not be willing to participate in a cardiovascular clinical trial.10 The data from Clark et al. have consequences for the ability to generalize treatment effects from clinical trials to the broader population. The increasing time, cost, and complexity of conducting clinical trials have created more urgency to streamline study protocols,11 which may be at odds with recruiting a representative population, particularly in groups with a documented decreased willingness to participate. Addressing the tension between these two needs has gained increasing importance in the USA, given the newfound emphasis on comparative effectiveness research which places a premium on studying patients who more closely mirror those typically seen in day-to-day clinical care.12 We propose four possible strategies to meet the goal of selecting a more representative population in clinical trials. Pursuing alternative approaches in trial design, which includes improving operational, structural and statistical efficiency. For example, incorporating well-designed prior research into a Bayesian analytic approach can reduce the patient sample size, cost, and time to complete a study.13 Continuing to characterize patient-level factors, such as willingness to participate in clinical trials, will help to develop interventions to improve the underrepresentation of women and minorities as well as increase the overall yield in clinical trial enrolment. Comprehensive reporting in peer-reviewed journals of salient patient recruitment data, such as the number of patients evaluated and enrolled, will aid in the process of judging the generalizability of findings. Increased surveillance of interventions as they become adopted into practice. Deviations from the trial-reported effect and safety profile will provide insight on the appropriate application of interventions and its effect in the broader population. From the individual clinician making a decision for a clinic patient to governmental organizations developing performance measures, the balance between risks and benefits of study findings, even in the best of circumstances, is not uniform to all patients. Studies are suggesting that trials may not represent typical populations. The results from Clark et al. extend our understanding about how patient preference may affect enrolment and the risk profile of the study sample. Increasing attention from governmental agencies and policy makers can provide the momentum to move forward with clinical trial design and strategies for enrolment that will best address the at-risk population. Central to this agenda will be the need to improve our understanding of traditionally underrepresented groups and populations that express a decreased willingness to participate in clinical trials with the goal of obtaining a sample more representative of the target population. We need further progress in this research and its application to ensure that we are truly studying the patients we are trying to treat.