Rasagiline, Parkinson neuroprotection, and delayed-start trials

Abstract

Rasagiline has been studied as a Parkinson disease (PD) neuroprotective agent in 2 major clinical trials, utilizing the delayed-start design in an attempt to separate symptomatic drug benefits from a disease-modifying effect. The ostensibly positive outcomes of these studies, however, are obscured by potential confounding factors that seem intrinsic to this trial design, including 1) very small changes in clinical outcome measures that could easily be overshadowed by other influences; 2) probable incomplete blinding to study end; 3) subjective components of the Unified Parkinson's Disease Rating Scale (UPDRS) scoring system; and 4) practice influences from repeated scoring. Interpretation of the recent Attenuation of Disease Progression with Azilect Given Once-daily (ADAGIO) trials is especially problematic given 1) divergent results with the 2 symptomatically beneficial doses and 2) variability in UPDRS scores with active rasagiline, which was twice the magnitude of the major finding of the study. These studies further illustrate the difficulty in documenting a disease-modifying effect when considering a PD drug with symptomatic benefit.

GLOSSARY

ADAGIO

= Attenuation of Disease Progression with Azilect Given Once-daily trial;

PD

= Parkinson disease;

TEMPO

= TVP-1012 in Early Monotherapy for PD Outpatients study;

UPDRS

= Unified Parkinson's Disease Rating Scale.

A fundamental goal of Parkinson disease (PD) research is development of drugs to halt or at least slow disease progression.¹ We have efficacious drugs for treating dopamine-deficiency PD symptoms, but no drug is proven to attenuate PD causative/pathogenic factors.² This primarily relates to the fact that we do not know what causes most cases of PD.

PD PROGRESSION IS DIFFICULT TO MEASURE

Compounding this problem has been the difficulty simply measuring PD progression. PD is a complex disorder affecting not only motor, but cognitive, behavioral, and autonomic systems. Measurement of progression in any of these domains might be meaningful. Because PD is primarily clinically defined as an extrapyramidal motor disorder with a dopaminergic substrate, this has been the measurement focus in neuroprotective trials.

Many of the drugs proposed to slow progression improve dopaminergic neurotransmission and treat PD symptoms. This has confounded clinical assessments, where it has proven very difficult to separate symptomatic effects from a true effect on disease progression. This is exemplified by the prior experience with the MAO-B inhibitor, selegiline, in the DATATOP trial, the largest and most expensive NIH-sponsored drug trial of its time.^3,4 The DATATOP results were initially interpreted as demonstrating neuroprotection; only later was the confounding symptomatic benefit recognized, associated with a pharmacologic effect exceeding the duration of study-drug washout (40-day half-life of brain MAO-B inhibition).⁵ Subsequent follow-up studies in this cohort cast doubt on a true disease-modifying effect from selegiline,^6,7 although this topic remains controversial.

Most drugs being contemplated as PD-slowing agents have potential symptomatic properties, including nondopaminergic drugs such as creatine, which may have some nonspecific effects improving energy or sense of well-being. A valid, measurable biomarker of the biologic process causing PD would be the ideal outcome measure, but since we understand little about the pathogenic substrates for PD, this is not currently possible.

Dopaminergic brain imaging as a biomarker of PD progression.

Theoretically, brain imaging of nigrostriatal dopaminergic integrity should be an ideal strategy for documenting PD progression in clinical studies. This outcome measure was employed in 2 large clinical trials assessing whether the dopamine agonists, pramipexole⁸ or ropinirole,⁹ had a favorable effect on PD progression compared to levodopa therapy. Striatal dopaminergic imaging changes in these 2 studies suggested disease-modifying effects, although clinical outcomes were opposite to the imaging. The confounding influences of the study drugs on the radioligand binding or metabolism were subsequently recognized.^10,11 Consensus opinion therefore concluded that dopaminergic brain imaging is unproven as a strategy for measuring PD progression in clinical trials using agents interacting with dopaminergic neurotransmission.¹²

Inhibitors of apoptosis: No symptomatic effect, but not neuroprotective in PD clinical trials.

Apoptosis has been proposed as fundamental to the PD neurodegenerative process. This has led to clinical trials of apoptosis inhibitors, which have been devoid of PD symptomatic effects. Absence of symptomatic benefit allowed clinical batteries to be used for outcome measurements (Unified Parkinson's Disease Rating Scale [UPDRS]), a more straightforward assessment strategy. Unfortunately, independent clinical trials of 2 different antiapoptosis drugs revealed no evidence of any disease-modifying effects in PD.^13,14

Delayed-start trial design to assess PD progression.

The delayed-start clinical trial design has been proposed to overcome confounding by drug symptomatic effects in PD progression trials.^15–17 With this scheme, untreated patients with PD are randomized to receive the study drug for (a) the full study duration, or (b) only the last half of the trial. With trials spanning a year or more, the presumption is that drug symptomatic effects will stabilize and be equivalent in both groups by study end. A clinical rating scale, the UPDRS, is utilized to document changes over time. If there is a disease-slowing effect, the group administered placebo for the first half of the study should never catch up to the other group.

Obviously, a sufficiently long trial is necessary to allow measurable clinical decline to accrue in this slowly progressive disorder. This requires selection of patients who have a high likelihood of remaining in the investigation despite being untreated during the first half of the study (placebo phase). Moreover, if the study drug, itself, provides only limited symptomatic benefit, patient selection is additionally crucial, so that patients with PD in the active arm of the study will not drop out in order to start levodopa or dopamine agonist therapy. Thus, patients in such a study may be restricted to those with mild and early PD. However, with mild PD, and only very slow progression, the modest changes in measurable parameters challenge this study design, even with long study durations.

RASAGILINE

Reminiscent of the early selegiline experience, the newer PD drug, rasagiline, is proposed to have a neuroprotective effect.¹⁶ Rasagiline and selegiline are, in fact, structurally and pharmacologically very similar, including selectively blocking brain MAO-B. Both drugs inhibit apoptosis in vitro, which appears independent of MAO inhibition.¹⁶ Unlike selegiline, however, rasagiline does not generate l-amphetamine metabolites; apoptosis blockade by these drugs tends to be reversed by such l-amphetamines.¹⁸ Like selegiline, rasagiline mildly improves PD symptoms.^19–21

Rasagiline and delayed-start clinical trial outcomes.

The delayed-start design has been utilized in 2 major clinical trials to assess whether rasagiline has disease-modifying effects. The initial trial compared 1 year to 6 months of rasagiline in the TVP-1012 in Early Monotherapy for PD Outpatients (TEMPO) study.¹⁵ Data analysis revealed that patients with PD receiving rasagiline for 1 year were statistically superior at study end to those administered rasagiline for only the last 6 months of that trial. Thus, one interpretation was that the findings “. . . may be due to a disease-modifying activity of the drug.”

The second rasagiline trial, designated ADAGIO (Attenuation of Disease Progression with Azilect Given Once-daily),^16,22 has been the stimulus for recent publicity. It employed the same design as the TEMPO study but a larger N (total of 1,176, vs 404 initial subjects in the TEMPO trial); it was also longer, with 9 months in each of the 2 phases, vs 6 months in each of the 2 phases of the TEMPO trial. In this 72-week ADAGIO trial, there were 4 study arms: 1) rasagiline, 1 mg daily during the entire study; 2) placebo during the first phase (36 weeks), then 1 mg rasagiline daily in the second phase (weeks 36–72); 3) rasagiline, 2 mg daily during the entire study; 4) placebo during the first phase, then 2 mg rasagiline daily in the second phase.

The outcome of the ADAGIO study, however, differed from TEMPO: only the group administered 1 mg daily for the full study had significantly better UPDRS scores at study end than the delayed-start group. Unlike the TEMPO study, the group receiving 2 mg daily for the entire ADAGIO trial was no different at study end than the group starting this dose 9 months into the study. The authors concluded that a disease-modifying effect is “possible,” at least in the 1 mg group.

The ADAGIO study specified 2 additional primary endpoints that were not utilized in the TEMPO study: the graphed slopes of UPDRS changes were compared between groups during the first, and also the second half of the study. However, the fundamental, intuitive rationale for using the delayed-start design relates to comparison of scores over the entire trial, study end vs baseline; this was the sole outcome measure in the TEMPO trial. Hence, we primarily focus our discussion on this, but later address the slope comparisons.

IS THE DELAYED-START TRIAL METHODOLOGY ROCK-SOLID?

The primary outcome measure of these delayed-start PD trials is the standard UPDRS total score, parts I through III (31 multiple-choice questions). This includes 2 patient-scored batteries (Mentation and Activities of Daily Living subscales; maximum score for both = 68 units) and the investigator-rated motor evaluation (maximum, 108 units). UPDRS outcomes measured in these delayed-start PD trials are subject to confounding influences for 4 reasons. First, the measured changes over the course of the trials are very small and easily overshadowed by other factors; second, blinding may be partially transparent; third, UPDRS scoring is not entirely objective; fourth, UPDRS repetitions may translate into biased influences from practice effects. Each of these concerns deserves discussion.

Slow progression and small changes.

A confounding effect would not need to be substantial to explain the outcomes in either the TEMPO or ADAGIO trials. In each study, the UPDRS difference between the delayed-start and early-start groups was on the order of 2 points. Placed in context, the UPDRS maximum score is 176 points and “2” represents approximately 1% of the maximum. This small change in UPDRS scores reflects that very slow progression of PD, plus selection of less aggressive PD that would allow patients to initially remain untreated if randomized to the delayed-start arm.

Blinding may be broken before the final scoring.

Although delayed-start trials are labeled as double-blind, this is true for only the first half of the study, with the last half, open-label. With transition to known symptomatic drug therapy, clinical responses to the open-label active drug may also disclose the randomization status of the initial double-blind phase; e.g., initial placebo treatment, then a clinical response to the active drug in phase 2 may retrospectively unblind these subjects. Since clinical scoring at study end is crucial, unblinding in phase 2 could compromise the findings.

UPDRS is not completely objective.

UPDRS scoring has a substantial subjective element, and for some entries, the distinctions are subtle. Thus, consider item 14 of the patient-scored ADL scale, where 1 = “rare freezing when walking; may have start hesitation”; 2 = “occasional freezing when walking.” Or consider the clinician scored item 19 (motor scale), facial expression: 1 = “minimal hypomimia; could be normal ‘poker face’”; 2 = “slight but definitely abnormal diminution of facial expression.”

Note that the UPDRS is subject to substantial placebo effects, including among investigator-raters, which has been well-documented.²³ Thus, patients with biologically stable PD potentially have a range of UPDRS scores that could be recorded, depending on subjective factors.

UPDRS response imprinting.

In delayed-start PD trials, subjects are repeatedly scored during the first phase, when one group receives symptomatic benefit from the active drug (rasagiline) while the other group receives placebo. In the TEMPO and ADAGIO trials, a clear symptomatic effect was borne out by the UPDRS scores that diverged right after the drugs were started; the rasagiline groups improved and the placebo groups did not (figure 3 in both the TEMPO¹⁵ and ADAGIO studies²²).

There is substantial potential for UPDRS choices to become somewhat automatic with repeat testing; the more often a task is repeated, the more likely for responses to become imprinted in memory and habit. In other words, there is potential for scores during the placebo-controlled phase to become locked-in; less thought is given as the UPDRS battery continues to be readministered. Thus, UPDRS scoring in the placebo-controlled phase may well influence UPDRS scores in the last half of the study when all subjects receive the study drug; each group may be more likely to retain some of the entries from the first phase when there was differential treatment (rasagiline or placebo).

ADAGIO: DIVERGENT OUTCOMES FROM DIFFERENT RASAGILINE DOSES

The ADAGIO trial generated counterintuitive findings based on dose. For the 1 mg rasagiline dose, the difference between the baseline and end-of-study UPDRS scores declined significantly less in the ADAGIO early-start group compared to the delayed-start group. Surprisingly, this was not the case for the 2 mg arm, where the baseline to end-of-study UPDRS changes were nearly identical in the 2 groups.²² Moreover, note the simple rankings of best to worst total UPDRS score changes in the 4 groups over the 18 months of the trial: 1) early 1 mg (18 months rasagiline), declined by 2.82 points; 2) delayed 2 mg (9 months rasagiline), declined by 3.11 points; 3) early 2 mg (18 months rasagiline), declined by 3.47 points; 4) delayed 1 mg (9 months rasagiline), declined by 4.5 points.

The difference between the early- vs delayed-start 2 mg group favored the delayed-start group by 0.36 points; this actually was slightly more than the 0.29-point difference between the top 2 groups above (early-start 1 mg vs delayed-start 2 mg).

There is no intuitive reason that 1 mg and 2 mg should have generated different outcomes. The study authors proposed that the symptomatic effect may have been greater with the 2 mg dose and this might have overshadowed a disease-modifying effect. However, the measured symptomatic effect was nearly identical between the 1 mg and 2 mg doses in the initial placebo-controlled phase of this study (as assessed by the secondary endpoint). Moreover, the symptomatic effect from rasagiline is thought to occur via MAO-B inhibition; rasagiline is an irreversible inhibitor of MAO-B (like selegiline) and both doses should have completely inhibited this brain enzyme, with a half-life of 40 days.⁵ Thus, why one dose should have induced a disease-modifying effect but not the other is not obvious.

Analysis of the individual 9-month outcomes suggests potential for confounding influences.

The delayed-start design assumed that the symptomatic benefit would have plateaued early in each of the 2 9-month study phases, presumably by 12 weeks.²² Thus, the primary influence on scores should then be due to the neuroprotective effect, slowing the progressive decline in UPDRS scores. Note, however, the best-to-worst ranking of total UPDRS score changes during the 9-month study phases where active rasagiline was administered: 1) delayed-group, active-phase 2 mg rasagiline: improved by 1.16 points; 2) delayed-group, active-phase 1 mg rasagiline: declined by 0.23 points; 3) early-start group, first active-phase 2 mg rasagiline: declined by 1.11 points; 4) early-start group, first active-phase 1 mg rasagiline: declined by 1.26 points; 5) early-start group, second active-phase 1 mg rasagiline: declined by 1.56 points; 6) early-start group, second active-phase 2 mg rasagiline: declined by 2.36 points.

Consider these UPDRS score changes in the context of the major positive finding of this ADAGIO study where the early-start 1 mg rasagiline group differed from the delayed-start 1 mg group by 1.68 points at study end; this is less than half of the range of 9-month scores listed above (3.52 points). Restated, variability in rasagiline scores in these 9-month epochs was twice the magnitude of the major positive finding of this study. Thus, small influences could easily bias study outcomes not only in this trial, but in delayed-start PD trials in general.

Other ADAGIO primary outcome measures: Slope comparisons.

ADAGIO phase 1 primary endpoint: Initial slopes.

If rasagiline has a neuroprotective effect beyond symptomatic benefit, the ADAGIO authors proposed that this should already be apparent in the initial phase, when 2 of the 4 groups were administered placebo.²² Assuming that the symptomatic effect would be fully developed by 12 weeks, they compared the UPDRS rate-of-change slopes for the last 24 weeks of the 36-week placebo-controlled phase. Indeed, the rate of change during this 24-week phase was significantly better with rasagiline.

Visual inspection of the actual curves for this 24-week phase, however, gives a different impression (figure 3 in the ADAGIO article²²). Whereas the slopes do diverge during the first 12 weeks of this 24-week placebo-controlled phase, the opposite is apparent during the second half of this phase. During the last 12 weeks of this phase, the 1 mg and placebo slopes start to converge, not diverge. Correspondingly, the 2 mg arm and placebo slopes no longer diverge, but run parallel during the second half of this 24-week phase. Thus, whereas the authors assumed that any symptomatic effect should have fully plateaued by 12 weeks, this graphic appearance suggests otherwise.

ADAGIO phase 2 primary endpoint: Terminal slopes.

The TEMPO trial results indicated that even if rasagiline is neuroprotective, it does not halt progression; UPDRS scores continue to deteriorate and the progression slopes never plateau despite rasagiline. However, a partial neuroprotective effect from early-start rasagiline should translate into UPDRS rate-of-change slopes that do not converge with the early-placebo curves at the end of the study. In other words, the slope analysis should confirm that the early-placebo group never “caught up” with the early rasagiline group. In fact, this was not the outcome in the early-start 2 mg rasagiline analysis; the early- and delayed-start curves converged to exactly the same data point at study end (figure 3B in the ADAGIO article²²). The 1 mg analysis did demonstrate persistent separation of the early- and late-start curves at study end (figure 3A in the ADAGIO article²²); whether this reflects the potential confounding influences discussed above is open to speculation.

Long-term follow-up of the TEMPO trial.

Additionally arguing for a rasagiline neuroprotective effect was the outcome of the TEMPO open-label extension study, which also generated recent publicity.²⁴ Thus, subjects in the original TEMPO trial were subsequently monitored for up to 6.5 years while continuing rasagiline treatment (but allowed to add other PD drugs). Interestingly, the group whose rasagiline was delayed 6 months had significantly poorer PD scores at every point in time thereafter. This was a striking finding indeed, given that only a 6-month delay of rasagiline still had an impact more than 5 years later.

Confounding interpretation of this follow-up study is the number of patients dropping out. By 1.5 years, approximately 20% of the original 404 subjects were lost and by 3 years, this jumped to 37%, with 59% lost by 5.5 years. Clinical trial analyses are notoriously sabotaged by high dropout rates, opening the potential for biased outcomes. Perhaps this explains the marked and otherwise inexplicable graphic divergence of the 2 groups beginning after the fourth year of the study (shown in figure 2 from that study²⁴). Early-start patients completing the extension study were also more likely to have been treated with levodopa (69%) than the delayed-start group (56%), which obviously could account for the later-developing UPDRS differences. Finally, in an open-label study with the blind broken, rating bias is possible; investigators knew the study hypothesis and might have been consistently more sympathetic to the early-start group. In summary, so many potential sources of confounding were present in this long-term follow-up study that interpretation is impossible.

But rasagiline is neuroprotective in the laboratory . . .

Numerous in vitro and in vivo studies have documented evidence of neuroprotective effects with rasagiline,¹⁶ although one might question whether these models truly replicate the disease process. Arguably, they support the initial interpretation of these clinical trials as demonstrating a disease-modifying effect. However, nearly all the dopamine-active drugs used to treat PD have similarly been reported to demonstrate in vivo and in vitro evidence of neuroprotective influences, including all the dopamine agonists²⁵ and even levodopa.^26,27 In fact, it has been proposed that early treatment with any dopaminergic drug may have a long-term favorable effect in PD.²⁵

Practical problems with prescribing rasagiline in clinical practice.

One might argue that rasagiline should be prescribed to all patients with PD on the chance that it might be neuroprotective (hedging one's bets, so to speak). However, balanced against this are considerations of potential drug interactions and expense.

The package insert lists numerous drugs that are contraindicated with rasagiline, including most antidepressants. Many of the listed drugs are likely to be considered in patients with PD. At the very least, this has medical–legal implications, whereby the drug combination might well be blamed for a variety of coincidental problems. A second issue is the considerable expense of rasagiline. Retail price is approximately $10 per tablet, not inconsequential even with pharmaceutical plans requiring copayments.

DISCUSSION

Twenty years ago, selegiline was prescribed to nearly all patients with PD because of faith in or hope for a possible neuroprotective effect. Now, the very similar drug, rasagiline, is being touted for the same purpose. However, like the earlier DATATOP trial assessing a possible selegiline neuroprotective effect, the current TEMPO and ADAGIO investigations raise more questions than provide definitive answers. This delayed-start study design came under the scrutiny of the American Academy of Neurology Quality Standard Subcommittee after the TEMPO trial and they concluded then that “no treatment has been shown to be neuroprotective.”² This still appears to be an appropriate conclusion.

These studies have implications beyond rasagiline. Unfortunately, they illustrate the collective frustrations with measures to assess PD progression in clinical drug trials. “Unmet needs” has become a buzzword for pharmaceutical companies touting PD drugs in the last few years. Clearly, an unmet need for the PD community is a valid and reliable means of simply assessing PD progression.

DISCLOSURE

Dr. Ahlskog received the Fred Springer Award from the American Parkinson's Disease Association; receives royalties from publishing The Parkinson's Disease Treatment Book (Oxford University Press, 2005) and Parkinson's Disease Treatment Guide for Physicians (Oxford University Press, 2009), Parkinson's Disease and Movement Disorders (Humana Press, 2000), and Surgical Treatment of Parkinson's Disease and other Movement Disorders (Humana Press, 2003); has received honoraria for lectures or educational activities not funded by industry; and receives research support from NIH/NINDS [P50 NS 40256-R (Co-I)]. Dr. Uitti serves as an Associate Editor of Neurology®; has received research support from Advanced Neuromodulations Systems and from the NIH/NINDS (P50NS 40256 [Co-I]); and his institution receives annual royalties from the licensing of the technology related to PARK8/LRRK2.

Address correspondence and reprint requests to Dr. J. Eric Ahlskog, Department of Neurology, Mayo Clinic, 200 First Street SW, Rochester, MN 55905 eahlskog@mayo.edu

See pages 1149 and 1151

Disclosure: Author disclosures are provided at the end of the article.