Earlier Dopaminergic Treatment in Parkinson's Disease Is Not Associated With Improved Outcomes

ABSTRACT

Background

The appropriate timing of dopaminergic treatment initiation in Parkinson's disease (PD) remains a matter of debate. The primary objective of this study was to determine whether earlier initiation of treatment was associated with less worsening of total UPDRS scores over 48 months.

Methods

We performed a secondary analysis of data from the CALM‐PD (Comparison of the Agonist Pramipexole With Levodopa on Motor Complications of Parkinson's Disease) trial to examine the associations between years since diagnosis and 48‐month changes in total and component UPDRS scores, Parkinson's Disease Quality of Life Scale (PDQUALIF) score, and the EuroQol‐5D visual analogue scale (VAS) score.

Results

There were no associations between years since PD diagnosis and 48‐month changes in total UPDRS, component UPDRS scores, PDQUALIF score, or EuroQol‐5D VAS score.

Conclusion

Earlier treatment was not associated with improved long‐term outcomes in this secondary analysis. Prospective studies are required to determine the appropriate timing of initiation of dopaminergic treatment to inform clinical practice.

Despite numerous advances in the therapeutics of Parkinson's disease (PD), the question of when to initiate dopaminergic treatment remains unanswered.1 Disease progression is faster earlier in the course of PD,2, 3 and compensatory mechanisms, which may forestall the development of motor symptoms, have been hypothesized to be detrimental to nigral neurons.4, 5 Therefore, prompt and early dopaminergic treatment may slow the progression of disease by normalizing basal ganglia dysfunction.3 Several clinical trials have suggested that early dopaminergic treatment may have a neuroprotective effect; however, interpretation of these studies has been confounded by the inability to differentiate symptomatic effects from neuroprotective effects, discordant functional neuroimaging results, the absence of expected dose‐response relationships, and the short duration of clinical trials.6, 7, 8, 9, 10 Initiation of pharmacological treatment may be reasonably delayed in the setting of mild symptoms that do not impair function, concern regarding the potential for adverse effects, and lack of convincing evidence supporting a neuroprotective effect of any pharmacological treatment.11 We analyzed data from CALM‐PD (Comparison of the Agonist Pramipexole With Levodopa on Motor Complications of Parkinson's Disease), a large, randomized, double‐blind, placebo‐controlled trial comparing initial treatment with pramipexole and levodopa,12 to examine whether earlier initiation of dopaminergic treatment is associated with improved long‐term outcomes.

Patients and Methods

In CALM‐PD, 301 participants with early PD requiring symptomatic treatment were randomized to either pramipexole or l‐dopa for an initial duration of 23.5 months that was subsequently extended to 48 months.12, 13 A full description of the methodology and study results has been previously published.12, 13, 14 The primary outcome was time to development of motor complications. Eligible participants were those aged ≥30 years diagnosed <7 years ago with idiopathic PD who were H & Y stage I to III and requiring symptomatic treatment at study entry and had not recently been treated with l‐dopa or a dopamine agonist. At study entry, mean (standard deviation) years since PD diagnosis was 1.6 (1.6), with range 0.1 to 7.2, and the mean total UPDRS score was 31.8 (12.8).12

In the primary analysis, we examined the association between years since diagnosis and 48‐month change in total UPDRS score using a mixed‐model repeated‐measures (MMRM) analysis.15 The model included treatment group, week, the interaction between treatment group and week, years since diagnosis, the interaction between years since diagnosis and week, center, baseline total UPDRS score, and the interaction between baseline total UPDRS score and week. The correlation matrix for the within‐subject observations across time was modeled using an unstructured pattern.

This analysis uses years since diagnosis as a measure of the timing of initiation of dopaminergic treatment, with the hypothesis being that a shorter time since diagnosis (i.e., early initiation of treatment) would be associated with better long‐term outcome. The association between years since diagnosis and outcome could be confounded, however, by those entering the trial with a long time since diagnosis possibly having a more slowly progressive form of PD and, as a result, better long‐term outcomes. To attempt to account for this confounding, we also included in the primary analysis model the generalized propensity score (GPS),16 which serves to balance covariate distributions, and the interaction between the GPS and week. The GPS was constructed by regressing log‐transformed years since diagnosis on the following baseline variables: age, sex, H & Y stage ≥2, presence of orthostatic hypotension, selegiline use, past l‐dopa use, Mini‐Mental State Examination score ≥28, and greater than five body systems with a comorbid illness.17 A total of 40 of the 301 observations were excluded from the analyses either because of missing covariate information (n = 13) or to impose the “common support” condition18 for GPS values in the quartiles of the distribution of years since diagnosis (n = 27). Similar analyses were performed for the 48‐month changes in the UPDRS components (mental, activities of daily living, and motor), the total score on the Parkinson's Disease Quality of Life Scale (PDQUALIF),19 and the EuroQol visual analogue scale (EuroQol VAS),20 except that the model for the MMRM analysis substituted the baseline value of the outcome variable for the baseline total UPDRS score. An analysis of time to development of a motor complication was also performed using a Cox proportional hazards model that included treatment group, years since diagnosis, and GPS as covariates and center as a stratification variable. Given the ability of amantadine to modulate dyskinesias, baseline amantadine use was also included in the regression model used to construct the GPS for the time to motor complications analysis.

The above analyses were used to estimate the so‐called dose‐response functions16 relating years since diagnosis to 48‐month outcome (or, in the case of time to development of motor complications, the hazard function) along with point‐wise 95% confidence bands computed using the bootstrap. The change in the dose‐response function between 0.4 and 2.3 years since diagnosis (interquartile range), along with its 95% confidence interval computed using the bootstrap, is also reported for each outcome.

Results

No association between years since PD diagnosis and 48‐month change in total UPDRS score was detected in either the unadjusted or GPS‐adjusted analyses (Table 1; Fig. 1). Similarly, no association between years since PD diagnosis and 48‐month change in any of the secondary outcomes, including the UPDRS mental, UPDRS activities of daily living, and UPDRS motor scores, PDQUALIF score, and EuroQol VAS score, was detected in either unadjusted or GPS‐adjusted analyses (Table 1). Moreover, no association between years since PD diagnosis and time to development of motor complications was detected (Table 1). Graphs of the dose‐response functions for each of the secondary outcomes can be found in the Supporting Information.

Table 1.

Associations between years since diagnosis and 48‐month outcome

	No GPS Adjustment*			Dose‐Response Function**
Outcome	Slope	95% CI	P Value	Difference	95% CI	P Value
UPDRS
Total	–0.47	(–1.81, 0.88)	0.50	–0.60	(–2.09, 0.89)	0.43
Mental	–0.02	(–0.14, 0.11)	0.79	–0.10	(–0.23, 0.02)	0.10
ADL	–0.27	(–0.67, 0.12)	0.18	–0.22	(–0.59, 0.15)	0.25
Motor	0.01	(–0.95, 0.98)	0.98	–0.24	(–1.07, 0.60)	0.58
PDQUALIF	–0.37	(–1.30, 0.57)	0.44	–0.15	(–0.91, 0.60)	0.69
EuroQol	0.01	(–0.01, 0.02)	0.56	0.01	(–0.01, 0.03)	0.35
Time to first dopaminergic complicationa	0.97	(0.87, 1.08)	0.60	0.98	(0.86, 1.11)	0.70

^* Reported is the coefficient of years since diagnosis (slope) obtained from the MMRM analysis that only adjusts for treatment group, center, and the baseline value of the outcome variable.

^** Reported is the difference in mean outcome between those enrolled at the 25th percentile (0.4 years) and those enrolled at the 75th percentile (2.3 years) of the distribution of years since diagnosis, obtained from the dose‐response function adjusting for GPS.

^a Reported is the hazard ratio associated with each additional year since diagnosis (no GPS adjustment) or with an increase from the 25th percentile (0.4 years) to the 75th percentile (2.3 years) of the distribution of years since diagnosis (dose response function).

Abbreviations: ADL, activities of daily living; CI, confidence interval.

Figure 1.

Dose‐response function for total UPDRS score. The middle curve is the mean total UPDRS score at month 48 as a function of time since diagnosis derived from an MMRM analysis. The upper and lower curves are point‐wise 95% confidence bands for the dose‐response function calculated using the bootstrap (see text for details).

Discussion

In our secondary analysis of the CALM‐PD cohort, we found no association between years since diagnosis and 48‐month change in any of the selected outcome measures. In contrast, the results of several clinical trials, none without limitations, suggest that early treatment may be associated with improved outcomes. In DATATOP (Deprenyl and tocopherol antioxidative therapy of parkinsonism), which utilized a 2 × 2 factorial design to evaluate deprenyl and tocopherol, deprenyl treatment was associated with a significantly delayed time to initiation of l‐dopa.6 However, these results may have reflected unanticipated symptomatic benefit. In ELLDOPA (Earlier vs. Later L‐DOPA), participants with early PD were randomized to receive one of three dosages of l‐dopa or placebo for 40 weeks followed by a 2‐week washout period.7 In a clear dose‐response pattern, there was significantly less worsening of total UPDRS scores from baseline to 42 weeks in the l‐dopa groups compared to the placebo group. However, the washout period may have been insufficient and the β‐CIT single‐photon emission computed tomography data were discordant with clinical data, showing significantly greater decline in striatal dopamine transporter binding in participants receiving l‐dopa. In two studies of rasagiline that used delayed‐start designs (TEMPO [TVP‐1012 in Early Monotherapy for PD Outpatients] and ADAGIO [Attenuation of Disease Progression with Azilect Given Once‐daily]), there was significantly less worsening of total UPDRS scores in the early rasagline groups.9, 10 However, in ADAGIO, which was a much larger trial with longer‐term follow‐up, the results supported a possible neuroprotective benefit for rasagiline 1 mg/day, but not for rasagiline 2 mg/day, which is not consistent with the expected dose‐response relationship.10

Although our results suggest that earlier dopaminergic treatment does not positively affect long‐term outcome, several limitations need to be considered. First, insufficient variability in time since diagnosis may have limited our ability to detect an association. While the range for years since diagnosis was broad, the interquartile range (0.4–2.3 years) was quite narrow and perhaps more heterogeneity in time to initiation of dopaminergic treatment is needed to detect an association. Second, the use of time since diagnosis as a surrogate for disease duration is inherently problematic. Whereas the clinical diagnosis of PD requires the presence of motor signs, the development of motor signs is typically preceded by a prodromal period characterized by nonmotor symptoms. In recognition of this, diagnostic criteria for prodromal PD have been developed.21 Third, individuals who entered the study with longer disease duration may have been fundamentally different with a more slowly progressive form of PD or a subtype with predominantly nonmotor symptoms22 that would not be expected to respond to dopaminergic medications. However, we did attempt to account for this confounding using the GPS. Fourth, whereas we included baseline amantadine use, baseline selegiline use, and past l‐dopa use in as variables in the GPS, we did not include other medications that may have potentially influenced our outcome measures. Fifth, we were limited in our selection of outcome measures to those that were included in the original CALM‐PD trial, and our selected outcome measures may not be the best measures of disease progression, particularly with respect to nonmotor features. Our primary outcome measure, the UPDRS, is influenced by symptomatic treatment, which may mask detection of underlying disease progression, particularly in early disease.23 For example, in the PPMI (Parkinson's Progression Markers Initiative), which is a prospective cohort study of individuals with early untreated PD, change in the International Parkinson and Movement Disorder Society–Unified Parkinson Disease Rating Scale (MDS‐UPDRS) score was largest in the first year and then slowed, reflecting the initiation of dopaminergic therapy.24 Furthermore, whereas we included the PDQUALIF, a validated PD‐specific scale,19 and EuroQol, a validated scale that is not PD specific,25 as measures of health‐related quality of life, we recognize that the disease‐specific 39‐item Parkinson's Disease Questionnaire (PDQ‐39) may be a better measure of health‐related quality of life in PD.26 Supporting the utility of the PDQ‐39 in early PD, in a prospective observational study called PDLIFE, the group in whom PD treatment was delayed, had higher PDQ‐39 scores (indicating worse quality of life) after a mean follow‐up of 18 months, whereas those who were treated early maintained their quality of life.27 Although this study may support the use of the PDQ‐39 as an outcome measure, it should be noted that it was not a randomized clinical trial and did not include the UPDRS as an outcome measure. Novel measures of disease progression are needed.

Moreover, we found no relationship between years since diagnosis and time to first dopaminergic complication, suggesting that earlier initiation of dopaminergic treatment is not associated with earlier development of dopaminergic complications. In pre‐l‐dopa era studies, median disease duration at treatment initiation was 6 to 10 years and dyskinesias developed much sooner than in modern studies, in which the median disease duration at treatment initiation is 2 to 3 years.28 Similarly, a recent observational study found that disease duration at time of development of dopaminergic complications did not differ between two cohorts, even though l‐dopa initiation was significantly delayed in one of them.29 Moreover, a previous secondary analysis of CALM‐PD found that incidence of dyskinesias after starting l‐dopa did not differ based on initial treatment assignment after adjusting for disease duration.30 In contrast, some studies have found that earlier initiation of treatment is associated with earlier onset of dopaminergic complications,31, 32 and that longer duration of l‐dopa treatment is associated with dopaminergic complications.27, 28, 33

Prospective studies are needed to definitively address the question of when to initiate dopaminergic treatment in PD. Results of the recently completed prospective, randomized, placebo‐controlled LEAP‐PD (Levodopa in Early Parkinson's disease) study, which was designed to examine whether early l‐dopa treatment has a disease‐modifying effect, may provide some answers.34, 35 In the meantime, initiation of dopaminergic treatment will likely continue to be delayed until symptomatic treatment is warranted.

Author Roles

(1) Research Project: A. Conception, B. Organization, C. Execution; (2) Statistical Analysis: A. Design, B. Execution, C. Review and Critique; (3) Manuscript Preparation: A. Writing of the First Draft, B. Review and Critique.

R.B.S.: 1B, 2C, 3A

X.L.: 1C, 2B

K.B.: 1A, 3B

M.M.: 1C, 2A, 2C, 3B

Disclosures

Ethical Compliance Statement: The authors confirm that the approval of an institutional review board and the obtainment of informed consent were not required for this secondary analysis of data from the CALM‐PD trial. We confirm that we have read the Journal's position on issues involved in ethical publication and affirm that this work is consistent with those guidelines.

Funding Sources and Conflicts of Interest: This research was supported by the Parkinson Study Group and the Parkinson's Foundation's Advancing Parkinson's Treatments Innovations Grant. Dr. Biglan is a full‐time employee of Eli Lilly and Company.

Financial Disclosures for previous 12 months: Dr. Schneider has received grant support from NINDS, Biohaven Pharma, CHDI Foundation, Teva Pharmaceuticals, Pfizer, Nuredis Inc, The Michael J. Fox Foundation, and the Canadian Institute of Health Research. Dr. Biglan is a full time employee of Eli Lilly. Dr. McDermott has received grant support from NIH, FDA, NYSTEM, SMA Foundation, and PTC Therapeutics. He has served on data and safety monitoring boards for Eli Lilly and Company, Catabasis Pharmaceuticals, Inc., Vaccinex, Inc., Cynapsus Therapeutics, and NIH. He has served as a Consultant for Voyager Therapeutics.

Supporting information

Figure S1. Dose‐response function for EuroQol VAS score. The middle curve is the mean EuroQol VAS score at month 48 as a function of time since diagnosis derived from an MMRM analysis. The upper and lower curves are point‐wise 95% confidence bands for the dose‐response function calculated using the bootstrap (see text for details).

Figure S2. Dose‐response function for total PDQUALIF score. The middle curve is the mean total PDQUALIF score at month 48 as a function of time since diagnosis derived from an MMRM analysis. The upper and lower curves are point‐wise 95% confidence bands for the dose‐response function calculated using the bootstrap (see text for details).

Figure S3. Dose‐response function for UPDRS ADL score. The middle curve is the mean UPDRS ADL score at month 48 as a function of time since diagnosis derived from an MMRM analysis. The upper and lower curves are point‐wise 95% confidence bands for the dose‐response function calculated using the bootstrap (see text for details). ADL, activities of daily living.

Figure S4. Dose‐response function for UPDRS Mental score. The middle curve is the UPDRS Mental score at month 48 as a function of time since diagnosis derived from an MMRM analysis. The upper and lower curves are point‐wise 95% confidence bands for the dose‐response function calculated using the bootstrap (see text for details).

Figure S5. Dose‐response function for UPDRS Motor score. The middle curve is the mean UPDRS Motor score at month 48 as a function of time since diagnosis derived from an MMRM analysis. The upper and lower curves are point‐wise 95% confidence bands for the dose‐response function calculated using the bootstrap (see text for details).

Figure S6. Dose‐response function for time to first dopaminergic complication. The middle curve is the log hazard of developing a dopaminergic complication as a function of time since diagnosis derived from a Cox proportional hazards regression analysis. The upper and lower curves are point‐wise 95% confidence bands for the dose‐response function calculated using the bootstrap (see text for details).