The challenge of comorbidity in clinical trials for multiple sclerosis

EFFECT OF COMORBIDITY ON TREATMENT OF MS

Literature regarding the effect of comorbidity on treatment of MS is limited. Findings in other chronic diseases suggest that comorbidity may affect multiple aspects of treatment. First, comorbidity may impede care. Individuals affected by multimorbidity report multiple barriers to self-care, including the compound effects of medications, difficulties in coordinating multiple medications, the total burden of medications, and financial challenges.¹⁰ Second, comorbidity may affect the frequency or intensity of treatment of coexisting conditions.^11,–,14 Third, comorbidity may affect persistence or adherence (defined in Ref. 15) to treatment, further reducing the benefits of therapies, which are only partially effective. For example, depressed individuals with diabetes are 1.5-fold less likely to persist with pharmacotherapy for diabetes after 12 months of follow-up than nondepressed individuals.¹⁶ Although the effect of comorbidity on persistence with MS disease-modifying therapies is unknown, depression is associated with reduced adherence to disease-modifying therapy (odds ratio 0.55; 0.42–0.74).¹⁷ Findings on whether adherence improves after treatment of depression are inconsistent.^17,18 Fourth, comorbidity may affect the effectiveness, safety, and tolerability of treatment, although evidence for these issues is limited in MS. In a secondary analysis of longitudinal data from a randomized controlled trial of a teleconference-delivered fatigue management intervention for MS, comorbid diabetes or arthritis modified the response to the intervention. Individuals with diabetes improved more slowly after intervention than those without diabetes, while individuals with arthritis improved more rapidly than those without arthritis but they had difficulty sustaining improvements.¹⁹ Finally, comorbidity may increase the risk of drug–drug and drug–disease interactions.

COMORBIDITY AND ELIGIBILITY FOR CLINICAL TRIALS

Individuals with comorbidities frequently are underrepresented in clinical trials.²⁰ Therefore, trial findings may not apply to a typical clinic population with comorbidities. Boyd et al.²⁰ reviewed clinical trials identified using Cochrane reviews for diabetes, heart failure, chronic obstructive pulmonary disease, and stroke. These trials frequently excluded individuals with comorbidities, ranging from 0% to 44% of diabetes trials, 0% to 42% of heart failure trials, 0% to 55% of chronic obstructive pulmonary disease trials, and 0% to 39% of stroke trials. Moreover, only 43.5% (70/161) of the trials reported the prevalence of any comorbidity among participants. Information regarding the definition or ascertainment of comorbidity was limited. Only 3.1% (5/161) of trials used comorbidity as a subgroup variable. A review of randomized trials published in the 5 highest-impact general medical journals and specialized journals that focused on the most prevalent chronic conditions found that multimorbidity affected participant eligibility in 95% of trials.²¹ Individuals with multimorbidity were excluded in 63% of the trials examined; this did not change from 1995 to 2010. Only 2.1% of trials explicitly included individuals with multimorbidity. A systematic review of 26 trials or prospective observational studies in cardiovascular disease focused on comorbidity measurement.²² The comorbidities assessed varied across studies and were assessed using varied data sources with 35% not reporting the data source.

The situation is similar in MS where most pharmacologic trials exclude individuals with severe comorbidities or substance use disorders.^23,–,26 We reviewed the published results of 9 sentinel placebo-controlled trials of disease-modifying therapies approved for MS in the United States (table 1).^23,–,31 Five of 9 trials (55.6%) excluded individuals with various comorbidities. In 4 of those 5 trials, the description of the exclusions for comorbidities was vague,^25,29,–,31 making it unclear how they were operationalized. Uniformly, the way in which comorbidity was assessed for the purposes of eligibility was not reported. None of the trials described the comorbidity status of participants at enrollment. Furthermore, none considered the presence of comorbidity as a subgroup variable for a priori or post hoc analyses.

Although we did not review all clinical trials of nonpharmacologic interventions for managing symptoms of MS, an ongoing systematic review of comorbidity measurement in randomized trials of rehabilitation interventions suggests that comorbidities are only considered in relation to exclusion criteria and are rarely reported when participant characteristics are described (M. Finlayson, personal communication, 2015). For example, Oken et al.³² randomized patients with MS to a weekly yoga class, weekly exercise class using a stationary bicycle, or a waiting-list control group. Participants with insulin-dependent diabetes; symptomatic lung disease; uncontrolled hypertension; liver or kidney failure; alcoholism/drug abuse; symptoms or signs of congestive heart failure, ischemic heart disease, or symptomatic valvular disease; or corrected visual acuity worse than 20/50 binocularly were excluded. A controlled trial of physiotherapy as intervention to improve mobility in MS excluded individuals with other major general medical or surgical disorders.³³

While exclusion of participants with comorbidities is intended to ensure participant safety, the consequence is that we lack knowledge about the safety, tolerability, and efficacy of the studied regimens in persons with MS who have common comorbidities. Limited data suggest that safety and tolerability of disease-modifying therapies are affected by comorbidity. An observational study found that individuals with migraine have more difficulty tolerating interferon-β because of worsening headache profiles.³⁴ The risk of fingolimod-associated macular edema is higher among individuals with comorbid diabetes.³⁵ These issues are particularly important for future trials in progressive MS, in which the target population is likely to be older and, therefore, more likely to have comorbidity. Excluding individuals with comorbidities in these trials slows accrual and produces a study population that is even more divergent from clinical populations than trials conducted in relapsing-remitting MS.

EXPLANATORY VS PRAGMATIC TRIALS

Randomized clinical trials can be classified as explanatory or pragmatic (table 2).³⁶ Explanatory trials test the efficacy of an intervention, that is, the effects under ideal conditions. An effort is made to enhance the internal validity of the trial by achieving a more homogeneous study population with well-specified inclusion and exclusion criteria, and by controlling all aspects of the intervention and comparator. In contrast, pragmatic trials test the effectiveness of an intervention, that is, the effects in real-world clinical settings where populations are more heterogeneous and differ from those in explanatory trials. Therefore, pragmatic trials produce findings that are potentially more generalizable and applicable to clinical practice. Clinical trials in MS to date largely have been explanatory in nature.

Although this classification suggests trials fall into distinct groups, explanatory and pragmatic trials really exist on a continuum, and several aspects or domains of a trial's design may influence where it falls in the continuum.³⁷ These may include eligibility criteria, the flexibility of the experimental and comparison interventions, required investigator expertise, choice of outcomes, intensity of follow-up and ascertainment of outcomes, and measures to ensure and assess adherence to the interventions by the participants and practitioners.³⁷ The Pragmatic–Explanatory Continuum Index Summary (PRECIS) tool and its subsequent revision (PRECIS-2)^38,39 were developed to assist trialists in considering these factors when designing their trials with the goal of matching designs to the intended uses of the trial results. The tool could also be used to evaluate existing trials (for a more detailed description of PRECIS-2, see appendix e-1 on the Neurology® Web site at Neurology.org).

Pragmatic trials face challenges. Because the intervention(s) is not as tightly controlled, it may not be as well delivered, reducing its benefit. Increased heterogeneity and the dilution of the intervention's effect require larger sample sizes (discussed further in Ref. 40) and longer follow-up time, increasing costs and the risk of attrition. Although a pragmatic trial may be performed in multiple clinical settings, we cannot assume that the findings from one clinical setting translate to other clinical settings. Nonetheless, relaxing criteria regarding age and comorbidity status could enroll a population more similar to that seen in clinic settings.

COMORBIDITY AND SAFETY MONITORING IN CLINICAL TRIALS

All clinical trials need safety monitoring. Typically, large randomized, multisite studies such as those used to evaluate disease-modifying therapies establish a Data Safety Monitoring Committee, which reviews trial conduct and accumulating data and advises the study sponsor regarding the ongoing safety of the trial participants. When participants experience treatment emergent adverse events including comorbidities such as hypertension, the Data Safety Monitoring Committee must decide whether the observed events are occurring at a rate greater than expected and whether they are treatment-related (i.e., a true safety signal) or not. Understanding how adverse events may affect differentially those with comorbidity is also important.

For common adverse events, the observed event rates can simply be compared between intervention groups, that is, using data internal to the trial. However, for rare (i.e., with a low event rate), serious adverse events such as cancers, the use of external comparators is needed. Such assessments require a good understanding of the expected age- and sex-specific incidence of comorbidities in the (untreated) MS population as these may differ from the rates reported for the general population. However, a recent systematic review identified very few studies evaluating the incidence of comorbidity in MS, with even fewer being population-based.⁵ The most commonly studied conditions were cancer and epilepsy; however, the time period over which incidence was being studied was often unclear, the definition of the conditions varied across studies, and age- and sex-specific estimates were not generally reported. Thus, the information necessary to assess the significance of rare but clinically important adverse events is largely lacking.

Often it is not until a drug is released to the market that it is used in a significant number (if any) of individuals with comorbidities, and this is when adverse effects specific to individuals with comorbidity may be detected. However, postapproval, safety is not monitored as systematically at the individual or group level as in clinical trials. Although the strategies used for postmarketing surveillance vary somewhat across nations,⁴¹ the challenges faced are similar. Postmarketing surveillance may involve one or more components including passive reporting of adverse drug reactions, active pharmacovigilance, phase IV studies, and the use of research networks to evaluate specific adverse events of interest. Passive reporting systems underreport adverse drug reactions by as much as 98% as compared to systematic monitoring, and phase IV studies may not be completed as requested.⁴¹ Specific efforts are required to assess the safety of new therapies in individuals with MS and comorbidities in the postmarketing phase.

ETHICAL ISSUES

Consideration of whether to include or exclude participants with comorbidity in a clinical trial illustrates a potential conflict in ethical principles inherent in clinical trials. Because the lifetime prevalence of comorbidity in patients with MS is substantial, this issue is frequently encountered in MS trials. Expanding the age of enrolled participants and including those with comorbidities are ways to make the study population more typical. However, such individuals may have decreased likelihood of benefiting (i.e., violation of the ethical principle of beneficence) and an elevated risk of adverse effects (i.e., violation of the principle of nonmaleficence). There also may be concern that inclusion of participants with comorbidity may complicate interpretation of the trial if, for example, they adhere less well to assigned treatment or differentially discontinue treatment or trial participation prematurely.

These considerations are potentially in conflict with the principle of autonomy that states that unless the safety concerns are more than hypothetical, informed patients should have the opportunity to participate. Moreover, the external validity (generalizability) of the study is reduced if the trial population does not represent the clinical population in whom the therapy will be used. From an ethical perspective, the principle of justice suggests that particular individuals should not bear a disproportionate burden of research participation, nor should they be unfairly excluded from the potential benefits of participation. Thus, excluding individuals with MS from clinical trials solely on the basis of age or comorbidity status requires careful consideration of the reasons and ramifications.

RECOMMENDATIONS

The workshop produced several recommendations to address knowledge gaps regarding the incidence of comorbidity that may interfere with the ability to assess safety in monitored trials, and the limited generalizability of clinical trials to clinical settings where the MS population commonly has comorbid conditions. These are described generally below and summarized in table 3 in more detail for the most common comorbidity, psychiatric comorbidity.