Selegiline and mortality in subjects with Parkinson’s disease

Methods.

The study was made possible by a record linkage database at the Medicines Monitoring Unit (MEMO), University of Dundee, which is described in detail elsewhere.13 In brief, MEMO uses the Community Health Indexes (CHI) for Scotland (supplied by permission of the Chief Area Medical Officers for each region of Scotland) to allocate the Community Health Number (CHNo) to health care data. Each resident is allocated a unique patient identifier (CHNo) that contains date of birth and a gender indicator. This number allows the accurate record-linkage of healthcare data. The Tayside Script Facility (TSF) within MEMO contains prospectively gathered information on all dispensed community prescriptions. After being dispensed at Tayside pharmacies, the original prescriptions are sent by the Common Services Agency to MEMO. By using custom-written, menu-driven software, the CHNo is ascribed from the details on the prescription. The date the prescription was written is recorded, and the individual drug code is entered from a drug dictionary developed by the Pharmacy Pricing Division and mapped to the British National Formulary.14 In addition, dose of drug, length of prescription, total number of tablets dispensed, and instructions to the patient are recorded.

This study was carried out using deidentified data. The Tayside committee on Medical Research Ethics has ruled that such anonymised studies do not require formal approval.

The main study population was identified from a fixed population of approximately 30,000 individuals from four practices (Dundee and Carnoustie) resident in the Tayside region of Scotland and registered with these practices between January 1989 and December 1995 or who had died in the intervening period. Patients with PD were identified by a first prescription for an anti-Parkinson’s drug classified in the British National Formulary (BNF)14 during this period. Individuals who had previously taken a neuroleptic drug were excluded as neuroleptic drugs may induce Parkinson’s-like extrapyramidal symptoms. Subjects aged 40 years or under were also excluded as PD is rare below this age. A further exclusion criterion was short-term use of bromocriptine (<60 days), which can be prescribed for reasons other than PD. Incident cases received their first PD drug within the study period of July 1989 to December 1995, with no prescriptions in the previous 6 months. The date of the first prescription was the index date for patients with PD. A cohort of patients without PD in the ratio of 9:1 was then randomly generated from the community of Tayside, frequency-matched by age in 5-year bands and gender to be used as a comparator group to the PD cohort.

In addition, a smaller cohort with a longer follow-up period was selected from the two practices in Carnoustie. This cohort had prescription data available from 1985 onwards and so potentially had 11 years of follow-up for mortality analysis. A comparator group for this cohort was generated using the same criteria as the main cohort.

The daily dose for each of the Parkinson’s drugs was determined and age of the patient at the first prescription for an anti-Parkinson’s drug was also obtained.

Survival of the Parkinson’s cohort relative to the comparators was plotted using the method of Kaplan–Meier.15 For the initial regression analyses, a Cox proportional hazards model16 was used to analyze survival in the two cohorts, stratified by age and gender in order to take the matching into account in the analysis. As comorbidity may be an important confounder in relation to mortality, two indicators for use of cardiovascular drugs and diabetic drugs before entry into the cohort were created and added to the regression model.

Secondly, an analysis was carried out comparing mortality according to the pattern of drug exposure over the length of follow-up. Three classes of exposure for the PD cohort were created: those who received levodopa-based drugs alone (co-careldopa or co-beneldopa) over the entire study period, those who received selegiline alone, and those who received selegiline in combination with other medication at any time (mainly co-careldopa and co-beneldopa). Those who received less than a 90-day prescription were excluded, leaving 56 subjects for analysis by drug exposure. The primary causes of death were obtained from death certificates and compared using exact tests because of sparse data. All data analyses were carried out using the SAS statistical package (version 6.12; Cary, NC).

Results.

A total of 103 incident cases of PD were identified from the population during the study period (1989–1995) after excluding those receiving neuroleptic drugs and those who were not true PD subjects. We also excluded a further six subjects who had short-term bromocriptine prescriptions, giving a total of 97 subjects. The whole cohort had a mean age of 72.5 (SD 10.3) years and consisted of 55% men and 45% women. The age and gender distribution in PD subjects and comparators are shown in table 1. There were no significant differences in age group or gender, because frequency matching was carried out by these variables.

A total of 28 incident cases of PD was identified in the Carnoustie cohort from 1985 onwards using the same selection criteria. The age and gender distribution of this cohort was similar to that of the total four-practice cohort consisting of incident cases from July 1989 onwards.

Table 2 shows the initial prescriptions for anti-PD drugs for the 97 subjects with PD. It is clear that the most common initial prescription was for co-beneldopa, with 62.7% receiving this drug, followed by selegiline (18.1%) and co-careldopa (14.5%). In addition, there were 14 patients whose initial prescription consisted of two drugs, with co-beneldopa and selegiline being the most frequent combination (eight cases).

Table 3 shows the average daily doses of each anti-Parkinson’s drug, which was similar to the recommended daily dose of each drug as described in the BNF.10 Out of the total of 97 cases of PD, 63 received a single drug formulation over the whole length of follow-up, with the most frequent monotherapy being co-beneldopa followed by selegiline. Co-prescribing was defined as having prescriptions on the same date for different drugs. The most common co-prescribed combination was co-beneldopa along with selegiline. The next most common was co-careldopa with selegiline. There were nine subjects who had co-prescribing of three drugs with the most frequent trio of drugs being co-beneldopa with selegiline and benzhexol. The median initial dose of levodopa was 150 mg, which gradually rose to a median of 200 mg at the end of follow-up.

After stratifying by age and gender in the proportional hazards model the risk of death was highly significantly greater for subjects with PD relative to the controls (rate ratio [RR] 1.76, 95% CI 1.11, 2.81). In the secondary analysis, when the PD cohort was classified by drug exposure, there was significantly greater mortality in those who received levodopa monotherapy (RR = 2.45, 95% CI 1.42, 4.23) relative to the comparators ( table 4⇓). However, there was no significant difference in mortality between subjects with PD who received selegiline in addition to other medication (mainly levodopa-based prescriptions) and the community comparators (RR = 0.92, 95% CI 0.37, 2.31). There was also no significant difference in mortality in those who received selegiline plus other medication compared with those who received selegiline monotherapy (RR = 0.47, 95% CI 0.05, 4.34), although the latter was a very small group. The mortality in relation to selegiline exposure is illustrated in the figure. We also showed a significant decrease in mortality for those receiving selegiline in combination with co-careldopa or co-beneldopa relative to levodopa alone (RR = 0.33, 95% CI 0.13, 0.83). All of these analyses were adjusted for previous cardiovascular drug use and diabetes on entry to the study, which were both significantly associated with increased mortality (table 4). The death certificate primary causes of death in subjects with PD are presented in table 5. There were no significant differences in any cause of death between the drug exposure groups, although the power to detect differences was low because of the small numbers.

• Download figure
• Open in new tab
• Download powerpoint

Figure. Kaplan–Meier plot of survival of the primary Parkinson’s cohort from date of first prescription by drug exposure compared with the comparator cohort.

In the Carnoustie subcohort with a much greater length of follow-up (up to 11 years), the results of the proportional hazards model were similar in that after stratifying by age and gender the mortality rate ratio of PD to comparators was 2.67 (95% CI 1.70, 4.21). There were insufficient prescription data to analyze by drug exposure for this cohort.

Discussion.

The mean age of incidence defined by initial treatment of PD in this study was 73, which was consistent with that found in other studies.1 More men than women were found to have this condition. Daily doses of l-dopa and other anti-Parkinson’s drugs in this study showed good agreement with the suggested daily doses as described in the BNF14 and these gradually increased over the length of follow-up. Anti-muscarinic drugs such as benzhexol and orphenadrine were rarely used.

There was an approximate twofold increase in risk of death in subjects with PD compared to the community comparators. This was consistent with other observational studies, which found increases in mortality of between 1.7 and 2.5 in PD relative to controls.1,3-6⇓⇓⇓⇓ The increased risk of death in subjects with PD could be attributable to their greater frailty or reduced mobility when compared with healthy comparators of the same age. Intermittent institutionalization of subjects with PD for longer periods may also leave them more prone to infections compared to comparators.

This study also showed a significant decrease in mortality for those receiving selegiline in combination with co-careldopa or co-beneldopa relative to levodopa alone (RR = 0.33, 95% CI 0.13, 0.83), despite the low power. In fact, the mortality in the combination therapy group of subjects with PD was similar and not statistically different from the comparators. The RR was close to one and therefore indicated no clinically important difference; thus, there is little evidence from this study that selegiline increased mortality. These results were consistent with randomized controlled trials and a meta-analysis of selegiline mortality in patients with PD10-12⇓⇓ and a retrospective study to evaluate treatment-related mortality using data from the General Practice Research Database.17

Selegiline in combination therapy as opposed to selegiline monotherapy appeared to have a beneficial effect on mortality, although this could be an erroneous conclusion, as the number of subjects in the selegiline alone group was too small to draw any valid conclusions. It is also likely that there was confounding in that long term survivors were more likely to receive combination therapy—a healthy survivor effect. There was some evidence for this in that the median time to first prescription for selegiline from entry to the study was 967 days in those who received combination therapy, whereas the median length of follow-up for those receiving selegiline monotherapy was 706 days. In addition, those who received levodopa alone may have more severe disease on entry to the study. Our data were drawn from an electronic prescribing database and we have no information on the severity of PD for our study. We were careful to only include incident cases and so we would not expect large differences in severity on entry, although we recognize that differences in severity at diagnosis may still exist. Another possibility may be that there was misdiagnosis of, for example, cerebrovascular disease as PD, with overrepresentation in the monotherapy groups. In our analysis by exposure we excluded individuals with less than 90 days of prescriptions and, in addition, there was no evidence of differences in cause of death between exposure groups, although these comparisons lacked power. Further studies in which severity is recorded would be necessary to determine whether there may be a true effect of more aggressive combination therapy or confounding explains the apparent benefit of combination therapy.