Ten-year follow-up of three different initial treatments in de-novo PD

Patients and methods.

Results.

Figure 1 shows the number of patients randomized to each treatment group, the numbers subsequently randomized again to another treatment arm, and the status of patients at the end of the study. Survival time was censored at date last known to be alive for 13 patients lost to follow-up after withdrawal from randomized treatment. Although patients tended to be rated less frequently in later years, the median number of assessments and the average intervals between assessments were similar in treatment arms 1 and 3. Early adverse reactions were more common in the bromocriptine group than in the patients randomized to levodopa alone, and all those rerandomized to another treatment arm came from arm 3.3 By the end of the trial, only 44 patients from arm 1 and eight patients from arm 3 remained on their original randomized treatment.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1. Patient flow during the study. Follow-up of rerandomized patients shown under their original arm (intention-to-treat).

Almost all those withdrawn from the bromocriptine arm (regardless of whether or not rerandomized) went onto levodopa treatment either alone or in combination with other drugs. The median duration before introduction of levodopa in arm 3 was 2.1 years (95% CI 1.3, 2.5). Sixty-two percent were on bromocriptine monotherapy after 1 year, 20% after 5 years. Five years after initial randomization, the median daily levodopa dose in the bromocriptine arm 3 was 250 mg (mean daily dose was 328 mg); compared to arm 1, in which the median daily levodopa dose was 500 mg (mean daily dose was 609 mg). After 8 years, the median daily levodopa dose in arm 3 was 500 mg (mean daily dose was 516 mg); compared with arm 1, in which the median daily levodopa dose was 625 mg (mean daily dose was 663 mg). The mean dose of bromocriptine during the first year of treatment was 36 mg/d (range 7.5 to 120).

The proportion of patients who were reported as taking amantadine at some stage during the trial was around 6% overall, and this was similar in the three arms.

The baseline characteristics of patients have already been reported; on entry to the study, the three treatment groups were similar with respect to age, sex, duration of PD, and disability scores.1

A total of 49 patients (16 in arm 1, 16 in arm 2, and 17 in arm 3) had their diagnosis revised during the course of the trial.

Mortality.

By December 1999, the average follow up of all patients randomized (including those who had died) was 9.2 years. The standardized mortality ratio for patients in the study compared to the general population of the United Kingdom was 1.78 (95% CI 1.62, 1.96). Table 1 gives the number of deaths and total person-years since randomization by the end of December 1999 for each treatment group. The cumulative mortality rate is slightly higher in the group initially randomized to bromocriptine, but this difference is not significant (figure 2). The hazard ratio for bromocriptine vs levodopa alone is 1.15 (95% CI 0.90, 1.47). After adjustment for age, sex, duration of disease before randomization, baseline disability, and year of randomization, the hazard ratio is hardly changed at 1.12. A similar hazard ratio is obtained for mortality in the first 5 years after initial randomization (see table 1), and there is no evidence that the hazard ratios vary by time of follow-up.

View this table:

• View inline
• View popup

Table 1.

Causes of death, death rates, and hazard ratios (95% CI) in arm 1 (levodopa alone), arm 2 (levodopa and selegiline), and arm 3 (bromocriptine)

• Download figure
• Open in new tab
• Download powerpoint

Figure 2. Mortality by treatment arm.

Table 1 also gives results comparing both arm 1 and arm 3 with the early-terminated arm 2 (levodopa plus selegiline) group. By the end of December 1999, there were 148 deaths among patients initially randomized to arm 2. The hazard ratio for arm 2 vs arm 3 is 1.06 (95% CI 0.84, 1.34). Including all deaths before December 1999, the hazard ratio for arm 2 vs arm 1 is 1.22 (95% CI 0.95, 1.55). This hazard ratio is slightly lower than that given in our earlier report3 when, after an average 6.8 years follow-up, the ratio of mortality in arm 2 compared with arm 1 was 1.32 (95% CI 0.98, 1.79). We have also reported that excess mortality in arm 2 (levodopa plus selegiline) seemed to be greatest in the third and fourth year of follow up. In contrast, there was no greater excess in the early years when comparing mortality in the bromocriptine arm 3 and the levodopa alone arm 1.

Disability.

Previous analyses have shown that the average improvement in disability scores during the first year after randomization was significantly less for those patients initially randomized to bromocriptine arm 3, compared with those randomized to arm 1 or arm 2.1 In subsequent years, average disability scores have worsened, but over the first 5 years, the difference between arm 1 and arm 3 remained fairly constant at around one point on the Webster scale (figure 3). After 5 years of follow-up, the difference in Webster score (adjusted for baseline score) between arm 1 and arm 3 was 1.0 (95% CI 0.2, 2.1), showing that those in arm 3 had slightly worse disability scores on average (table 2).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3. Disability by treatment arm.

View this table:

• View inline
• View popup

Table 2.

Difference (95% CI) in mean Webster disability scores

Figure 3 indicates that the difference in disability between arm 1 and arm 3 diminishes after the fifth year of follow-up. By the ninth year of follow-up, the adjusted difference was 0.2 (95% CI 1.5, 1.5). We have also analyzed ‘final’ disability scores based on the average of the most recent two ratings before death or the end of 1999. This gave an adjusted difference of 0.8 (95% CI 0.3, 1.9) between arm 3 and arm 1. Similar findings were obtained from an analysis of the Northwestern University disability scale.

Figure 3 indicates that, on average, patients in the bromocriptine arm returned to pretreatment levels of disability after 3 years, about 1 year earlier than patients in the levodopa arm.

Discussion.

The PDRGUK study is the largest randomized study to date comparing the effect of different initial treatment regimes on long-term outcome of patients with early, mild PD. This is also the only report on a 10-year follow-up of motor complications in patients initially randomized either to a dopamine agonist or to levodopa. The only other published study comparing levodopa and bromocriptine as initial treatment of PD over 10 years7 had far fewer patients and focused on disability and mortality data.

The PDRGUK trial was designed to be pragmatic by recruiting patients with PD who would broadly reflect everyday practice. Inclusion criteria were much wider than in many recently published studies where stringent exclusion criteria were obligatory. For instance, there was no age cut-off and patients with stable concomitant illnesses were not excluded. This is reflected by the standardized mortality ratio (1.78) observed for the participants, which is much more typical of the mortality experience of recent population-based observational studies8,9,10⇓⇓ than trials such as the Parkinson Study Group–DATATOP where a mortality ratio of less than 1 was reported.11 Patients whose diagnosis was subsequently revised to one of the atypical parkinsonian disorders (all of which have a worse prognosis than idiopathic PD) remained in the study and were included in the analysis. This group of patients was excluded from analysis in the other 10-year study of levodopa and bromocriptine,7 which may at least partly account for their slightly lower mortality experience (standardized mortality ratio: 1.58).

We found no significant difference in mortality between the arms randomized to levodopa and to bromocriptine. There is thus no evidence that life expectancy is improved by initiating treatment with an agonist. This is in accordance with results reported from another 10-year study7 and does not confirm one previous report of reduced mortality on a combination regime of levodopa and bromocriptine.12

Our results show a higher incidence of dyskinesias in the group of patients randomized to levodopa, but no significant difference when results for the more clinically relevant moderate and severe grades are analyzed separately. No significant difference was found in wearing-off, and in fact, the incidence of moderate or severe wearing-off was marginally higher in the group of patients randomized to bromocriptine. Motor function as measured by the Webster scale was significantly better in the levodopa group for the first years of follow-up and continued to be better throughout the study. Similar results were found in the Northwestern University Disability Scale, which includes scores for activities of daily living. Following an initial improvement, the time until patients returned to their pretreatment levels of disability was 3 years in the bromocriptine arm and 4 years in the levodopa arm.

At the time when this study was designed, bromocriptine was the most widely used dopamine agonist for the treatment of PD. A 5-year study13 reported a lower incidence of dyskinesias, but not of wearing-off, in patients initiated on bromocriptine. Nevertheless, no significant difference in motor scores between levodopa and bromocriptine arms after 5 years of follow-up was shown. Since the initiation of our trial, several new agonists with different pharmacokinetic and receptor affinity profiles have been introduced into clinical practice. However, we believe that our findings are unlikely to be specific to bromocriptine. There is little evidence of a relevant clinical difference among individual dopamine agonists. In the few studies directly comparing individual dopamine agonists, no14 or little15,16⇓ difference was reported. A systematic review by the Cochrane Movement Disorders Group17 led to the conclusion that there was a small advantage in the efficacy of pergolide over bromocriptine on motor function, but no conclusion regarding levodopa-induced motor complications could be drawn. Another systematic review comparing lisuride and bromocriptine18 failed to reach a firm conclusion on the comparative efficacy of these two drugs.

Recently, several studies comparing levodopa to other agonists as initial treatment in early PD have been published. Although they all differ in design (including methods of dyskinesia assessment) and none spanned a follow-up period of 10 years, the results are remarkably similar to ours: all show some delay in the onset of dyskinesias in the agonist group, and that the agonist had a weaker effect on motor function than levodopa.

The controlled, double-blind 5-year study of ropinirole vs levodopa19 showed a higher rate of dyskinesias, but significantly greater motor improvement in the levodopa arm, with persistently better unified PD rating scale (UPDRS) scores throughout the follow-up period. The cabergoline trial20 showed a significantly lower incidence of dyskinesias in the agonist arm over 3 to 5 years, with a borderline difference in end-of-dose failures. Patients randomized to cabergoline had worse scores on UPDRS factor II and III throughout the follow-up period; the significance of the difference was not reported. Three-year results of the PELMOPET study (pergolide vs levodopa)21 and 2-year results of the CALM-PD study22 (pramipexole vs levodopa) showed a similar pattern. The latter study involved quality of life scores, which were significantly in favor of levodopa.

This consistency of the finding of less initial motor improvement and persistently worse motor scores in the patient groups randomized to the agonist arm—even during periods of open-label levodopa addition—is remarkable, particularly because many of the agonists are capable of inducing motor improvement comparable to levodopa in the 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP)-lesioned animal model.23 It has been argued24 that the difference in motor scores may be clinically irrelevant. However, this seems unlikely because it is consistent, independent of the motor scales used, and is reflected in activities of daily living in several studies as well as in quality of life scores in one study.22

The most plausible explanation seems to be undertreatment with levodopa, after levodopa rescue in the agonist arms. Clinicians may be reluctant to increase the levodopa dosage to the same degree in the presence of a treatment which they expected to have a levodopa-sparing effect. This is reflected in the lower levodopa dosages in the agonist treatment arms. Why this should have occurred in studies with a double-blind design may be explained by a hypothesis proposed by Holloway,22 who suggested that the initial motor improvement during the first few weeks may have been used as a benchmark and target for the future treatment effect both by patients and clinicians. As patients on agonist therapy show less initial response, clinicians may tend to be less aggressive in their subsequent medical management.

Our study design had included a levodopa/selegiline arm because of a previous report25 of reduced mortality in patients on this combination. Increased mortality in arm 2 led to a decision to inform all patients in this arm of the results and to advise discontinuation of selegiline in 1995. Patients continued to be assessed annually. The hazard ratio of the levodopa/selegiline arm vs the levodopa arm was 1.57 (1.49 when adjusted for age and sex) during the third and fourth year of follow-up.2 It gradually decreased over the following years and was 1.22 by the end of 1999. There is thus no evidence of excess mortality in later years, after discontinuation of selegiline, in the patients who had initially been randomized to levodopa/selegiline.

We conclude from the similarity of the mortality rates in the levodopa and the bromocriptine arm that the excess mortality observed in the levodopa/selegiline arm was a phenomenon of that arm, rather than the effect of an unusually low mortality rate in the levodopa arm. The cause of the increased mortality in the levodopa/selegiline arm remains undetermined, despite further investigations,3,26,27⇓⇓ and despite other recent reports showing increased signs of autonomic dysregulation in patients with PD who are taking selegiline28 or raising the possibility of increased mortality during selegiline treatment in subpopulations of patients with PD.29,30⇓

We concur that there is no proven causal link between selegiline and increased mortality in patients with PD.31,32⇓

Based on previous medium term (2 to 5 years) studies, it has been claimed that initial treatment with an agonist is the optimum approach to the management of early PD. However, these studies covered only the early phase of the disease, when disability, impairment, and handicap are relatively mild. Clinical practice clearly suggests that there are more early treatment failures on agonists, due to either suboptimal response or adverse events. In our study, the withdrawal rate due to lack of effect was 15.6% in the bromocriptine arm vs 0.8% in the levodopa arm.1 Although the ropinirole study19 reported that only hallucinations occurred more frequently in the agonist arm, most studies, including ours,1 found side effects or drop-outs due to adverse events to be more frequent in patients on agonists.33,20,22⇓⇓

In our study, only 20% of patients randomized to bromocriptine were on monotherapy at 5 years, and 3.1% (8 patients) at 9 years. Similar figures have been reported by others: in the Sydney study,33 there were no patients left on bromocriptine alone in 5 years; and in the study by Rascol et al.19, only 16% of the patients initially randomized to ropinirole remained on agonist monotherapy.

It remains difficult to disentangle whether the lower risk of motor complications truly is an effect of delaying levodopa or an effect of lower cumulative levodopa dosage. It is also unclear how crucial the delay of motor complications is with respect to the patients’ overall quality of life, if this delay is achieved at the expense of less motor improvement. Mild wearing-off and mild dyskinesias are straightforward to manage and dyskinesias may go unnoticed by patients. This is likely to be reflected in the significantly better quality of life scores in the levodopa group in the pramipexole study.22 Our study did not show a significant difference in moderate to severe dyskinesias and fluctuations between the arms randomized to different initial treatments after 10 years of follow-up. The overall difference in wearing-off, which is often perceived by patients as more troubling than dyskinesias, was also not significant.

We are currently extending the follow-up of our patients to one final assessment which we hope will help answer the question to what extent patients’ long-term quality of life in advanced disease is influenced by initial treatment.

We conclude that at the present time, there may not be a universally applicable answer to the question how to best treat early PD. We recommend that physicians, together with their fully informed patients, should attempt to tailor drug treatment to individual needs, balancing optimal motor control and tolerability against the risk of delayed motor complications. Although initial dopamine agonist represents a reasonable therapeutic strategy particularly for younger onset PD, early treatment with low doses of l-dopa (150 to 400 mg/d) would seem the best initial treatment choice for the majority of patients.

Appendix