Evolution of clinical trials in multiple sclerosis

Changes in diagnostic criteria

The MS diagnostic criteria and its subsequent revisions have made it easier to diagnose the disease and this has resulted in many patients receiving an earlier diagnosis.^5,6 Since the 2010 revision of the McDonald criteria, a diagnosis of MS can be made in a single time frame, in which lesions on a single magnetic resonance imaging (MRI) scan can fulfill both criteria for dissemination in time (DIT) and dissemination in space (DIS). This contrasts with previous criteria in which the diagnosis required subsequent clinical presentations or MRI to fulfill DIT. The latest revision of the McDonald criteria in 2017 further broadened the diagnostic parameters and included the presence of oligoclonal bands in the cerebrospinal fluid to substitute for DIT when the initial MRI findings did not fulfill this requirement. ⁷ In addition, cortical lesions, and symptomatic lesions excluding the optic nerve, may now be counted as MRI evidence of DIS or DIT. By broadening the diagnostic criteria, more patients are likely to be diagnosed earlier whereas they previously would not be diagnosed by an earlier version of the criteria.

A consequence of earlier diagnosis resulting from changes in diagnostic criteria is that the overall clinical course of MS appears to improve. The average MS patient today has milder disease compared with those diagnosed using the older criteria. However, this may have resulted in lead-time bias, in which earlier diagnosis appears to improve the disease course as patients take longer to accumulate neurological disability. This would lower the ARR by providing an observed time with fewer relapses relative to the older definitions, such as the Poser criteria. This lead-time bias would also lead to the appearance of less virulent disease, at least in the short term. Whether this is occurring is difficult to separate from potential confounders such as improved availability and efficacy of therapies.

Likewise, a portion of patients with clinically isolated syndrome (CIS) who are at high risk of developing MS will have a diagnosis of MS under the new criteria. The reassignment of patients who are at a higher risk within the CIS group to those having milder disease in the MS group results in the improved overall prognosis of both groups, an effect known as the Will Rogers phenomenon. ⁸ This has been demonstrated since the first iteration of the McDonald criteria by a retrospective analysis showing that 50% of patients with CIS would progress into definite MS within a year by the 2001 McDonald criteria compared with only 20% under the previously used Poser criteria. ⁹ Currently, there is no evidence as to whether the 2017 revision will select patients with less active disease. Recently, a retrospective analysis on reaching the diagnosis of MS under the latest revision of the McDonald criteria compared with the 2010 revision, concluded that the 2017 revision expedites diagnosis without compromising diagnostic accuracy. ¹⁰

Availability of treatment

The increasing availability of treatment for RRMS has also affected patient enrollment in trials. With multiple DMTs available, there are motivations for high-risk patients not to enroll in a clinical trial as these patients may opt to select a proven therapy rather than enroll in a trial and risk receiving a placebo or a medication with unknown safety or efficacy profiles. Likewise, patients currently involved in trials who experience a relapse may be rescued and removed from the trial to start a proven therapy, thus leaving those whose disease activity is relatively well-controlled during the trial period to complete the study. There are suggestions that over time, withdrawals from therapy have been increasing and even mandated on the occurrence of relapses or progression, thus reducing the opportunity to have subsequent outcome events reducing the ARR and other event rates over the course of the study.

On the other spectrum, patients who are treatment nonresponders may be interested in seeking new options by enrolling in a trial. However, some trials exclude enrollment of patients with extensive exposure to prior DMTs.^11,12

Additional considerations

Other factors have been proposed to account for the changing MS trial population. For example, increasing physician and patient awareness of the disease and accessibility to MRI scanning has led to earlier diagnosis and initiation of treatment.^13,14 Improved standards of care for MS and accessibility to DMTs have resulted in better overall management and thus, lower disease activity in the overall MS population. Modern-day trials allow patients to enroll closer to their time of diagnosis by reducing the requirement of the number of relapses prior to screening. Whereas most early trials required at least two relapses over the course of 2 years, many contemporary trials allow for enrollment with just one relapse over the past year. In addition, recent trials use more stringent and objective definitions of relapse requiring evaluation by a neurologist and measurable increases in the Expanded Disability Status Scale (EDSS) functional system subscores. ¹⁵ Together, these changes further substantiate the shift of the MS trial population towards milder disease activity.

Implications

The changing MS trial population complicates the comparison of results from different studies, especially those from different times. ¹⁶ Assessing whether one DMT is superior to another is challenging without a head-to-head comparison of the two agents in a randomized controlled trial, yet it is impractical to conduct head-to-head studies of all available agents. As trial patients present with milder disease, it becomes more difficult to generate on-study changes with traditional designs, and therefore larger group sizes and longer study duration are required. At the same time, it is no longer practical to design studies based on assumptions drawn from historic trials. Studies that are designed and statistically powered according to findings from earlier trials may overestimate disease activity and underestimate the number of patients enrolled to be able to detect treatment differences. There is a need for more efficient designs and sensitive outcome measurements.

The ethics of placebo in RRMS trials

New treatment options for RRMS have led to the decline of placebo-controlled trials due to the ethics and marketing concerns of performing such studies when multiple effective therapies exist. The 1990s saw the approval of the first MS therapies, including interferon β preparations and glatiramer acetate.^17–19 These trials relied on the use of a placebo group to demonstrate the efficacy of the investigated agent. Subsequent pivotal phase III trials of new therapies incorporated a placebo group until the approval of alemtuzumab, which in 2014 became the first US-approved agent without comparison with a placebo.^20,21 However, the ethics of using a placebo in MS trials have been in discussion since much earlier.

Following the availability of the first injectable DMTs, the National MS Society organized an international task force in 2000 to discuss and publish proceedings on the ethical use of placebos in MS trials, and its recommendations were further revised in 2008.^22,23 The group concluded that participation in placebo trials is still ethical for patients who decline or fail available agents or when established therapy is not available at the time, such as in PMS. In countries without access to any DMTs, placebos may also remain ethical, but active comparator trials still seem more appropriate. Some countries, such as Brazil, will not allow placebo-controlled trials if the trial is not conducted in the country originating the trial. In all cases, appropriate informed consent and information on available therapies must be provided.

Controversies exist on whether placebo trials are justified, and despite recommendations restricting the use of placebo in RRMS trials, subsequent phase III trials of fingolimod, teriflunomide, and dimethyl fumarate had approximately 1500 patients exposed to placebo treatment. ²⁴ Many researchers and organizations have taken a middle ground with regard to using a placebo when proven therapies are available. ²⁵ The continued use of a placebo is backed by claims of the changing natural history of MS and indication that data from previous placebo groups are no longer valid. ²⁶ However, as growing evidence surmounts to the detrimental effects of delaying initiation of DMT for RRMS, the use of a placebo in RRMS trials have ended. ²⁷

Active comparator designs

The past decade saw a gradual replacement of placebo designs by active comparator trials. The pivotal phase III trial of peginterferon β-1a marked the last placebo-controlled trial of a therapeutic agent for RRMS. ²⁸ Subsequent published phase III trials of daclizumab and ocrelizumab, as well as the recently concluded phase III trial for ozanimod, have been conducted with active comparator arms.^29,30 These trials have shown that active comparator studies are feasible to detect both the efficacy and superiority of new agents. On the other hand, the use of an active comparator arm instead of a placebo reduces on-study changes between both groups, resulting in more patients needed to observe a significant and meaningful treatment effect. Modern-day trials have reflected this change. Whereas early trials typically enrolled around 200 patients, modern trials have frequently enrolled more than 1000 patients. The increasing costs of conducting large-scale trials have put a strain on existing resources, and there is a need for more efficient designs.

Combination trials

Combination trials seek to answer the question of whether multiple therapies administered together have greater efficacy than either alone. Studies are designed on the premise of using different agents with complementary or synergistic mechanisms of action. ³¹ In this regard, the most rigorous study to date examined the effects of glatiramer acetate and interferon β-1a administered together, which did not show a benefit in using both drugs over the more effective agent, glatiramer acetate.^32,33 Other studies combining DMTs with hormonal therapy ³⁴ or statins ³⁵ have shown mixed results. Significant drawbacks of employing combination trials include the potential antagonistic mechanisms of action among study agents, compounded costs and side effects, and increased size and complexity of the trial design. ³⁶ While these reasons have deterred combination trials in MS, there may be a role in the future as we progress in understanding the disease pathophysiology and developing more therapeutic options. ³⁷

Adaptive designs

Adaptive designs offer flexible and efficient trial methods that allow for preplanned modification of the study protocol after an interim analysis. This allows for flexibility in cost and trial duration. While adaptive designs have seen use in other fields of medicine, current experience with adaptive designs in MS is just emerging. The phase II trial of siponimod for PMS was the first and only published MS trial to date that employed an adaptive design to evaluate the safety and tolerability of the study agent and to determine its dose-efficacy response curve. ³⁸ The study randomized one cohort of patients to receive 10 mg, 2 mg, or 0.5 mg of the study agent or placebo and evaluated the outcome of combined unique active MRI lesions at month 3. An interim analysis was performed, and based on an estimation of which doses would be needed for the optimum dose–response characterization, a second cohort of patients were randomized to receive 1.25 mg and 0.25 mg of the study drug. Evidence for the further efficacy of siponimod was recently published in a double-blind randomized placebo phase III trial. ³⁹ The design of adaptive trials demands meticulous planning and a detailed interpretation of the interim results and navigation of a complex review process. It also requires a sensitive short-term outcome measure on which to base interim decisions and long-term outcomes to assess treatment efficacy. Adaptive designs for MS are nascent and their prevalence remains to be seen.

Pragmatic trials

As therapeutic options for MS broaden, there is a growing need for evidence demonstrating applicability of clinical trial results to routine clinical practice. In contrast with explanatory trials, which measure efficacy under controlled settings, pragmatic trials are designed to evaluate treatment effectiveness in the real-world setting, producing results that are more generalizable and applicable to clinical practice. ⁴⁰ Despite the distinction between explanatory and pragmatic trials, actual trials fall on a spectrum between the two categories as characterized the Pragmatic–Explanatory Continuum Index Summary (PRECIS) tool and its subsequent revision (PRECIS-2).^41,42 The tool helps trialists to ensure their design decisions match the trial’s intended purpose.

Just as explanatory trials face the challenge of requiring a large sample size with enough follow-up time to detect significant treatment effects, the issue is even greater in pragmatic trials when selection criteria and interventions are less controlled. A purely pragmatic trial in MS is almost unattainable and will almost certainly include components towards the explanatory end of the spectrum. As adeptly summarized by Ford and Norrie in a publication of pragmatic trials in the New England Journal of Medicine, ‘a pragmatic approach to pragmatism would be to adopt the features of pragmatic trials whenever feasible and sensible and when such features do not compromise trial quality and the ability to answer the clinical question of interest.’ ⁴³

Clinical outcome measures

Clinical outcomes play an important role as the primary endpoint of pivotal phase III trials. Assessments of relapse and changes in the EDSS score have been the oldest and most widely used clinical outcome measures. They each capture distinct pathophysiologic processes of the disease. Whereas relapse rates measure inflammatory activity, changes in the EDSS reflect accumulation of clinical disability from inflammatory or neurodegenerative disease processes. The EDSS is universally familiar to MS clinicians and accepted by regulators,^44,45 but it is marked by shortcomings in its variability between examiners, heavy emphasis on walking, limited impact of cognitive impairment and nonlinearity. The imperfections of the EDSS have led to the development of the MS functional composite (MSFC), ⁴⁶ which covers three major MS domains, can be assessed quickly and reliably, and gives a z-score on a continuous scale. However, z-scores are abstract and difficult to interpret clinically, which has limited the widespread use of the MSFC. Despite its imperfections, the EDSS currently remains the most accepted clinical outcome measure and will likely remain so for the near future. Recently, electronic scoring of the EDSS has been developed and has shown an increased consistency of EDSS scoring, ⁴⁷ and it has seen use in the phase III trials of ocrelizumab ³⁰ and siponimod. ³⁹

Over the past decade, patient-reported outcome (PRO) measures have increasingly been used as complements to objective clinical measures and provide important information on subjective outcomes, such as evaluations of subjective symptoms, satisfaction with treatment, and quality of life.^48–50 They offer an additional facet in the assessment of treatment risks and benefits. Nevertheless, the subjectivity of PROs typically limits their use to a secondary outcome measure for DMT investigations. ⁵¹ Their use as primary outcomes for symptomatic therapies may be appropriate.

Neuroimaging outcome measures

Neuroimaging outcome measures provide additional information to supplement clinical measurements of treatment efficacy. Most commonly used are MRI measurements of lesion load or atrophy, which provide objective and highly sensitive measurements of current and past disease activity. Inflammatory activity can be quantified as focal lesions on T2 MRI sequences or T1 gadolinium-enhancing lesions and have shown a correlation with relapse rate.^52,53 This important metric has allowed for rapid development of DMTs for RRMS due to its use as the primary outcome in short-term phase II trials. The ability to detect reductions in lesions on MRI associated with treatment allows for quick progress into phase III trials. ⁵⁴ The accrual of disability as reflected by neurodegeneration is measured by brain atrophy. This has been commonly assessed either as total brain parenchymal volume or fractional volume, defined as the ratio of brain parenchymal volume to total volume within the brain surface contour.

MRI outcome measurements have undergone refinement as the understanding of the disease process continues to develop. Composite imaging outcomes of combined unique active lesions have seen use in a growing number of trials to assess total new or enlarging T2 or gadolinium-enhancing lesions. Regional atrophy is increasingly explored as a more sensitive measurement of changes reflective of associated deficits, and areas of the thalamus and cerebellum may be more sensitive to changes.^55,56 Gray matter atrophy has been shown to have more clinical correlation than white matter atrophy and is more sensitive to changes. ⁵⁷

Optical coherence tomography (OCT) is a noninvasive method to image the retina and characterize pathology of the optic nerve. ⁵⁸ By visualizing the unmyelinated axons of the retinal nerve fiber layer (RNFL), OCT can directly assess axonal injury. In MS patients, the RNFL is thinner than in healthy people, regardless of a previous history of optic neuritis, supporting its role as a unique in vivo model of neurodegeneration.^59–62 OCT has been used as an outcome measure in a few clinical trials in PMS without evidence of neuroprotective effects.^63–66 Despite its current lack of prevalence in clinical trials, OCT as a fast, inexpensive, and noninvasive approach to evaluate neurodegeneration, continues to prompt further investigation into refining the method.

Advancements in imaging techniques have led to the development of new outcome measures. Magnetization transfer ratio (MTR) is an MRI technique used to assess tissue integrity by correlating it with axonal loss in MS lesions. Whole brain MTR has been used in several trials,^{28,63,64,67,68} while regional MTR has been used in the phase III trial of laquinimod for RRMS. ⁶⁹ While imaging outcome measurements are not likely to replace clinical endpoints in trials, they provide important information that supplements data on the efficacy of novel agents.

Fluid biomarkers

Clinical and radiologic outcome measures have served as the cornerstone of assessing the therapeutic efficacy of drugs and disease activity; however, they provide limited assessment of the pathophysiological disease process, such as inflammation, axonal injury, and demyelination. Early search for biomarkers in the spinal fluid of MS patients hoped to identify outcomes that could be used in trials, but these efforts have not yielded promising results. In recent years, several biomarkers for MS have shown potential for diagnosis, prognosis, monitoring disease activity and treatment response. ⁷⁰ Although currently there are no established biomarkers in use in clinical trials, among the most promising is neurofilament light chain (NFL). ⁷¹ NFL is a subunit of neurofilaments that compose the axonal and dendritic cytoskeleton. It is released during axonal injury, making it a marker of axonal damage. ⁷² In addition to its presence in the cerebrospinal fluid, NFL can also be detected in the serum due to using the ultrasensitive single molecule array (Simoa) technology.^73,74 Serum NFL has been shown to correlate with relapses, EDSS scores, MRI lesions, and brain and spinal cord atrophy,^75–78 and its levels have shown a decrease in patients treated with DMTs.^79–82 Its downside is the lack of specificity, and levels may be influenced by many comorbidities or conditions. With an increased availability and understanding of NFL and disease activity, NFL may become the first fluid biomarker to be used to monitor therapeutic effects in future randomized clinical trials and we may eventually see its use in the clinical setting.

Combined outcome measures

Other new outcome measures have integrated domains to achieve a more holistic evaluation. ⁸³ The concept of disease activity-free status in MS was first explored in the natalizumab phase III trial, ⁸⁴ and it was later defined as ‘no evidence of disease activity’ (NEDA) to reflect the absence of relapses and no progression of disability or new MRI activity. NEDA has been used in some phase II^85,86 and phase III trials.^20,21,32 The combination of outcome measures increased the predictive value of disability progression compared with individual clinical or MRI metrics. ⁸⁷ The drawback of using NEDA lies in the inability to assess its effects on specific outcomes, since clinical and radiological activity are all combined into one assessment that is heavily dependent on timing. Currently, NEDA is still undergoing refinement as it remains unclear which optimal components to include and how frequently they should be assessed. In addition, brain volume as a marker of neurodegeneration has been proposed to be included in a revised measure referred to as NEDA-4. ⁸⁸ As the concept of NEDA continues to evolve, it becomes important to establish standardized definitions to facilitate universal interpretation and demonstrate its validity and reliability across studies.