Using monoamine oxidase type B inhibitors in Parkinson's disease

Despite the growth in new drugs for the treatment of idiopathic Parkinson's disease, high quality evidence from randomised controlled trials to guide clinicians on choosing the most appropriate therapy is scarce. Monoamine oxidase type B inhibitors (MAOBIs) have been available for about 30 years. Their popularity was boosted in the 1980s when researchers proposed that they may have a neuroprotective effect and hence alter the natural history of disease rather than merely provide symptomatic relief. Subsequently one report of increased mortality in patients given selegiline with levodopa dented this popularity.¹ In this issue, Ives et al provide us with a useful summary of the trials on MAOBIs for the treatment of Parkinson's disease (p 593).²

Of the 17 trials, most are short term studies—only seven have follow up data of over 18 months. In addition, most of the trials compare a MAOBI with placebo. These trials show that MAOBIs reduce symptoms compared with placebo and allow patients to be treated with lower doses of levodopa or delay the need to start taking levodopa. However, they provide no evidence that the risk of dyskinesias is lesser or lower with MAOBIs and only modest evidence for some reduction in the fluctuations seen in motor function. This latter finding contradicts an earlier systematic review, which concluded that evidence showing that MAOBIs prevent motor complications was insufficient.³ Moreover, the studies in these reviews²,³ were against either levodopa or placebo rather than against dopamine agonists, which are believed to be the most efficacious drugs in preventing motor complications.⁴

Use of MAOBIs for early Parkinson's disease in the United Kingdom declined possibly as a result of the initial report from the UK-Parkinson's disease research group (UK-PDRG) study, which noted an increased risk of mortality for selegiline and levodopa used in combination compared with levodopa alone.¹ For those who always doubted the validity of this finding, this meta-analysis will provide further reassurance. However, we doubt that it has totally closed this debate. This meta-analysis pools studies that use MAOBIs as monotherapy with those that use it as combination therapy as in one arm of the UK-PDRG study. This may be misleading. For example, both aspirin and angiotensin converting enzyme (ACE) inhibitors used alone reduce cardiovascular events in high risk patients, yet patients taking both drugs paradoxically have a greater risk of myocardial infarction than those treated with ACE inhibitors alone.⁵ If the increased mortality seen in the UK-PDRG study was due to a pharmacological interaction between MAOBIs and levodopa then no effect on mortality would be expected with monotherapy.

We need to check for publication bias with any meta-analysis, and this can be done by producing a funnel plot; this plots the precision of the results from each study against its effect estimate.⁶ If all the studies lie equally around the pooled estimate this inspires some confidence that the analysis is not biased. We used the mortality odds ratios and confidence intervals provided by Ives et al in their paper¹ to generate a funnel plot (see figure on bmj.com). This shows six imprecise studies to the left of the pooled estimate and none to the right, which implies to us that the analysis may underestimate the adverse effect on mortality (P = 0.07 for small study bias).

We also need to consider whether on clinical grounds the studies should be pooled if like is not compared with like. Although Ives et al failed to find any evidence for heterogeneity overall (P = 0.6), they found evidence of heterogeneity between the initial results from the UK-PDRG study (hypothesis generating report) and subsequent data (P = 0.05).²

Why is this important? The UK-PDRG investigators postulated that the observed increased mortality may have been a chance phenomenon or the effect of selective mortality; where the adverse effects of selegiline and levodopa influenced only a subgroup of older Parkinson's patients with comorbidity and, as these were removed from the population over the first five years of follow up, mortality in the remaining patients was not increased.⁷ Almost all the data for mortality in the meta-analysis come from just two studies; the US DATATOP study (selegiline monotherapy) and the UK-PDRG study (selegiline and levodopa therapy). However, the patient populations in these two studies were dissimilar. The US patients were better educated and healthier than expected, as patients with comorbidity were selectively excluded at recruitment. This effect was so strong that in the US patients mortality was, if anything, lower than in the general population⁸ in contrast to the UK-PDRG population, which had an 80% increase in standardised mortality ratio,⁹ which is consistent with results from unselected observational studies of mortality in Parkinson's disease.¹⁰

So should clinicians use MAOBIs for newly diagnosed patients? These drugs clearly provide symptomatic benefit and probably entail no risk of increased mortality if they are used as monotherapy and in younger and otherwise healthy patients. However, as Ives et al conclude, data on comparative efficacy with other first line drugs, particularly dopamine agonists, is lacking.² In general, drugs used to treat Parkinson's disease show a seesaw effect—those that are better at improving motor disability tend to have greater risks of complications and vice versa. The best choice of treatment for an individual patient will be determined by the patient's own preferences in balancing motor benefits against the onset and degree of motor complications.