Summary for General Readers
Given it was first introduced to treat Parkinson’s disease in the 1960’s (see the accompanying background information on Parkinson’s disease for general readers), it is surprising that it was not until four decades later that a major study took place looking at outcome versus placebo of levodopa therapy from the point of view of its long-term neurotoxic or neuroprotective effects. At the turn of the 21st century, it was considered a fashionable view that levodopa therapy primed the development of dyskinesia and on-off fluctuations; it was almost a necessary evil in treating Parkinson’s disease and to be delayed as far as possible into the illness.
Then came the ELLDOPA (Earlier vs Later LevoDOPA) study, which confirmed the accepted view that levodopa did lead later on in therapy to dyskinesia, but more importantly showed that treating a patient adequately over nine months with efficacious medication rendered them in a better clinical state than those starved of medication, even after stopping the treatment for two weeks. In other words, treated patients were better even when the drug was temporarily “washed-out” of their systems. Did this mean that the treatment was somehow slowing the deterioration of the disease? Not according to a parallel brain scan study; radioactively labelling the amount of surviving nerve endings of the degenerating dopaminergic nerve cells revealed that patients who had received levodopa had worse scans than those who had received nothing, despite being clinically better off.
To many “jobbing” (which is what some call those who spend their time just managing patients on a practical basis rather than leading opinion) neurologists, this simply suggested that such imaging is perhaps not such a reliable marker of disease progression, and confirmed their suspicions that fears over the dangers of levodopa therapy had been over-played. They would have seen many of their patients do really quite well on levodopa therapy, improving significantly over their prior untreated state, and remaining better than that level for a long time and without complications, especially if they had been dosed cautiously. Keeping a patient under good control, thus maintaining their activities of daily living as best as possible, might easily render the patient in a better state even after a temporary withdrawal than one left untreated to become chronically disabled. This did not necessarily imply neuroprotection.
However debate remained intense over the possible neuroprotective effect and over the study’s methodology. A series of other neuroprotection studies were carried out. The one reviewed here, called ADAGIO – an acronym which, if you can believe it, comes from “Attenuation of Disease progression with Azilect GIven Once daily” – is an example of one that employs an elegant design called “delayed start”. (Azilect is the trade name of rasagiline.)
The problem of studying a neuroprotective effect of a drug that also helps symptoms is that, when the only way you can measure the disease is by symptom severity, you don’t know if the patients are better because their disease course has improved or if they simply feel better from symptom control. The drug “masks” the state of the disease. The obvious solution, which was employed by the ELLDOPA study, is to stop the drug temporarily so that the treated and untreated groups are back on a level playing field. But another solution is to delay the start in one group compared to another. At the end of the study, both groups are on the same treatment, but one group had enjoyed the treatment for a longer time, and therefore had more time over which to have the cumulative neuroprotective effect. (One assumes symptomatic benefits are relatively short-lasting.)
The ADAGIO study investigated the drug rasagiline, a monoamine oxidase inhibitor which works by preserving more dopamine within the synapse, so making surviving dopamine nerve cells work harder (see the Primer in Parkinson’s Disease for details). It was found that those patients given the drug earlier were indeed better off by the end of the study, presumably because they had had a longer time receiving neuroprotection.
But doubling the normal dose of the drug made the patients not even better off when treated earlier, but worse off, and this was not simply because of symptomatic side effects of the higher dose. It is therefore not surprising that the conclusion over neuroprotection was muted, and that the debate still continues.
It may be that there are no “short-cuts” to studying neuroprotection. Fortunately, in most patients PD progresses slowly over many years. A neuroprotective agent should therefore be given the chance to work over 10+ years to measure its benefit, and patient groups on or not on the agent should be on similar best symptomatic therapy throughout, just as you would do if you were using the neuroprotective agent in real life.
In the meantime, what do we do? There is merit in the argument that we should give every PD patient the “good” dose of rasagiline, because the study suggested neuroprotection. When the results come out from a 10 year study, it will be too late for today’s patients. But many neurologists do not do this.
The first reason is economic. In health care economies that are free at source, like that in the UK, costs are limited by a model that requires proof and quantification of efficacy (though there are always “political” exceptions). Clearly, there is some scientific evidence for neuroprotection from rasagiline, but it is a judgement call whether this is enough to extrapolate that patients will be better off after 10 years on the drug because of its neuroprotection than those on other treatment regimes. In these grey areas, the drug is essentially competing with a number of other agents of uncertain cost:benefit ratio and with varying strengths of claim. Even in other health economies, and indeed in advertising in general, there are strict rules about what claims may be made about a product.
The second reason for caution over wholesale use of an agent for neuroprotection is historical. In some quarters, until around twenty years ago, there was wholesale use of selegeline, a drug similar to rasagiline, as a neuroprotective agent. This was reversed by a study (DATATOP) that suggested increased mortality from the drug. Many patients were dismayed at this news, and when they were taken off the drug they were even more dismayed because it is actually quite good as a symptomatic agent. Now the findings of DATATOP have been refuted again, and it is back to being used as one of a number of reasonable choices for symptom control. Of course this cannot be directly extrapolated to rasagiline, but there is natural concern over losing the trust of another generation of neurologists and patients.
That is why it may be prudent to steer a middle course. Levodopa is not desperately neurotoxic, and it good for controlling symptoms. On the other hand, it does cause dyskinesia and the doubts over neuroprotection are such that most would not use it until symptoms warranted it. Similarly rasagiline has a good role in symptom control, and is officially recommended for such use (even in “rationed” health economies), but many neurologists are cautious about using it specifically for a special long-term neuroprotective benefit.
After the uncertainty surrounding the disease modifying effects of selegeline, where the final conclusion is that it probably is neither neuroprotective nor increases mortality, in recent years a number of studies have explored again the neuroprotective or neurotoxic properties of symptomatic therapies for Parkinson’s disease. The ELLDOPA (Earlier vs Later LevoDOPA) study (Fahn et al., 2005) took treatment naive patients who had had symptoms for at least two years and measured Unified Parkinson’s Disease Rating Scale (UPDRS) scores during 40 weeks treatment either with levodopa or placebo. Measurements were then taken after a two-week washout period off medication. Patients had the expected dose-dependent improvement on treatment, and the expected deterioration off treatment, but they were still significantly better than those who had been on placebo throughout. Did this mean that levodopa had been partially neuroprotective over those 40 weeks? Functional imaging performed on a subgroup in the same study gave the opposite picture. There was worse deterioration in the treated group as measured by beta-CIT SPECT dopamine transporter levels. The study therefore cast doubt on the notion that such functional imaging is a reliable biomarker of disease progression. However, a number of uncertainties remain concerning interpretation of the study’s findings: i) there could be compensatory transporter up-regulation in untreated patients, ii) transporter levels might otherwise not be an accurate marker of neurodegeneration, iii) the washout period might not have been long enough to remove residual symptomatic benefit, iv) there were some patients who might not in fact have had Parkinson’s disease in this study as they had no functional imaging abnormalities.
The TEMPO study (Parkinson Study Group, 2002) conducted at around the same time employed a different design to look at the possible neuroprotective effects of the Monoamine Oxidase (MAO-B) inhibitor rasagiline. Instead of a washout at the end, there was a delayed start in one group at the beginning. Thus in the first phase, one group was given placebo and the other rasagiline. In the second phase the placebo patients were given rasagiline, and the treated patients carried on with their existing rasagiline.
The study reviewed here, the ADAGIO study (Olanow et al., 2009), employed the same design for the same drug and used three hierarchical statistical tests for inferiority of delayed versus immediate start. The measure of disease severity was total UPDRS ( ie motor aspects, non-motor aspects and disability all combined). First, after the initial improvement over the 12-week wash-in period in both treated and placebo groups following commencement of therapy, the slope of subsequent deterioration in the rasagiline group over week 12 to the delayed start point at week 36 had to be less steep than in the placebo group. Second, the final UPDRS scores after the end of the delayed start period at 72 weeks had to be better in the initial start group than in the delayed start group. Finally, the slope of deterioration in the initial treatment group during the delayed start period from the end of the second wash-in at 48 weeks to the 72-week end point had to be no worse than in the delayed start group. In other words, initial deterioration had to be slowed in earlier treated patients, they needed to remain better off than delayed treated patients even after the latter group had started treatment, and there needed to be no suggestion that the initial treatment patients were catching-up in terms of disease progression with delayed treated patients so that they would eventually become as badly symptomatic as the latter if the trial had gone on longer.
All three of these statistical criteria were met – but only for 1 mg rasagiline (the current standard dose) not 2 mg. The authors commented that there appeared to be no difference in side effects or in drop-out rates. They suggested that a greater symptomatic benefit might mask the neuroprotective effect in mildly affected patients, and found that all was well in a post-hoc analysis of the worst affected patients. In other words, the delayed start patients improved very well initially and in comparison there was little neurodegeneration to act upon.
But I note that the slopes of deterioration were nevertheless significant in this study in both early and delayed phases. The problem was that at the end of the 72 weeks, the higher strength initial start patients had more deterioration than lower strength initial start patients (3.5 UPDRS points vs 2.8 points) and yet the higher strength delayed start patients had less deterioration than lower strength delayed start patients. The 2 mg initial start end result was basically aberrantly poor.
Given this caveat, the study concluded that there is a possible neuroprotective effect of rasagiline at 1 mg strength but described concerns with the study design. They established that drop out of placebo patients because they were suffering too badly on no treatment was not a factor and admitted the trade-off between the fact that a longer initial phase would yield more potential for neuroprotection but also more placebo dropouts due to uncontrolled symptoms.
Rasagiline has undoubted clinical efficacy (though more modest than levodopa), a good side effect profile and an advantage over the similar agent selegeline in possibly being safer to use concomitantly with certain antidepressants, though it is considered “prudent” not to prescribe them together. Its long duration of action makes it a good choice for reducing nocturnal and early morning symptoms.
Regarding the neuroprotective effect, in my opinion the 2 mg dosing issue remains a major problem in interpretation. A neurotoxic effect of higher doses seems unlikely. Given the different metabolism and body habitus of different subjects and general pharmacological behaviour it is very unlikely that there should be such a narrow range of “special” dose.
I am not convinced that the UPDRS is a true interval scale – in other words a certain change at one level is the equivalent of one at a more severe level. This is important because the analyses in this study have to assume this linearity. Yes, there are studies that indicate UPDRS linearity over time. For example one study showed a linear 3 point annual drop on treatment, but only after treatment had “bedded in” for six months (Guimaraes et al., 2005). But is disease progression itself always linear over time? Clinical experience sometimes suggests otherwise.
So it might just have happened that more patients on 2 mg initial start, while adequately matched in terms of initial clinical severity, were at a stage of disease where they were teetering on the edge of a steeper slope of clinical deterioration. And by the same token, how can we be sure that the delayed start 1 mg patients were not similarly “unlucky”? After all, statistics is only probability and if there are a lot of unknown variables in a complex study there is a fair chance that one of them may throw up an aberration.
Rehabilitationists are very aware that it is much harder to regain lost function than to maintain function. Therefore the likely slope of disease severity vs function when deteriorating is not the same as that when improving; this might be called hysteresis. Could the same apply to a lesser extent to different rates of deterioration, eg allowing deterioration to occur unchecked and then trying to slow it at a later, more resistant, stage?
My arguments are just conjecture, and perhaps I am wrong about the possible non-interval scale behaviour of the UPDRS scale. But I have yet to hear an adequate explanation of why doubling the rasagiline dose statistically significantly removes neuroprotective benefit by the end of the study.
A difficulty with this and other study designs, as mentioned above, is that there is a trade-off between the neuroprotection period and the drop-out rate; a long neuroprotection period may show a bigger effect, but too many placebo patients might drop out of the trial because they were going too long without symptomatic treatment. In addition there is a trade-off regarding the initial disease severity. The ADAGIO patients had to already be diagnosed with two of tremor, rigidity and bradykinesia and then remain untreated for 18 months until baseline. Many of my patients would be desperate for treatment by then! But take patients that are too mild, and there is a risk that many will not have true Parkinson’s disease; that may have been what happened with the normal functional imaging in some of the ELLDOPA study patients.
One sympathises with the difficulties in designing a trial to test the neuroprotective properties of a symptomatic agent, and the ADAGIO study was very carefully designed to address these concerns. But perhaps there is no short cut to a study design where a putative agent is added at an early stage to ongoing best symptomatic treatment (at least six months of such treatment as indicated by the study on predictability of UPDRS behaviour), and continued through in parallel with best symptomatic treatment for 10-20 years. Studies have looked at how much UPDRS change constitutes a clinically important difference (Shulman et al, 2010). Minimal clinically important difference, according to external criteria of disease severity, constitutes a change of at least 4.1 points; the putative neuroprotective effects in ADAGIO did not approach this level. Given that Parkinson’s disease (fortunately) has such a long time course, we would hardly expect that a few months of neuroprotection would result in anything more. Unless we do a trial of this 10-20 year duration, we remain wholly reliant upon extrapolation rather than demonstration.