Ophthalmologic Baseline Characteristics and 2-Year Ophthalmologic Safety Profile of Pramipexole IR Compared with Ropinirole IR in Patients with Early Parkinson's Disease

Background. Parkinson's disease (PD) progressively affects dopaminergic neurotransmission and may affect retinal dopaminergic functions and structures. Objective. This 2-year randomized, open-label, parallel-group, flexible-dose study, NCT00144300, evaluated ophthalmologic safety profiles of immediate-release (IR) pramipexole and ropinirole in patients with early idiopathic PD with ≤6 months' prior dopamine agonist exposure and without preexisting major eye disorders. Methods. Patients received labeled IR regimens of pramipexole (n = 121) or ropinirole (n = 125) for 2 years. Comprehensive ophthalmologic assessments (COA) included corrected acuity, Roth 28-color test, slit-lamp biomicroscopy, intraocular pressure, computerized visual field test, fundus photography, and electroretinography. Results. At baseline, we observed retinal pigmentary epithelium (RPE) hypopigmentation not previously reported in PD patients. The estimated relative risk of 2-year COA worsening with pramipexole versus ropinirole was 1.07 (95% CI: 0.71–1.60). Mean changes from baseline in Unified Parkinson's Disease Rating System parts II+III total scores (pramipexole: 1 year, −4.1 ± 8.9, and 2 years, −0.7 ± 10.1, and ropinirole: 1 year, −3.7 ± 8.2, and 2 years, −1.7 ± 10.5) and Hoehn–Yahr stage distribution showed therapeutic effects on PD symptoms. Safety profiles were consistent with labeling. Conclusions. The risk of retinal deterioration did not differ in early idiopathic PD patients receiving pramipexole versus ropinirole. RPE hypopigmentation at baseline was not previously reported in this population. This trial is registered with NCT00144300.


Introduction
Ocular manifestations of Parkinson's disease (PD) include visual field defects, electroretinographic changes, defective color vision and motion perception, impaired acuity (letters correct), and/or visual hallucinations [1,2], possibly reflecting retinal dopaminergic defects and morphologic changes [3][4][5]. Ophthalmologic findings in early PD and their changes during progression and therapy are understudied.
Retinal safety of dopamine agonists required reinvestigation in humans because albino rat studies showed retinal degeneration with ropinirole, pramipexole, and rotigotine; this effect was not seen in albino mice or pigmented rats, monkeys, or pigs [9][10][11] but involved the pan-vertebrate process of disk shedding. Few full peer-reviewed papers exist on the potential effects of pramipexole on the human retina. The US Food and Drug Administration (FDA) required a postmarketing study on eye safety of pramipexole in humans with PD, which supported a labeling revision that remains in the current label [9]. This 2-year study (NCT00144300) compared ophthalmologic safety of immediate-release (IR) pramipexole and ropinirole in early idiopathic PD, adapting a comprehensive ophthalmologic assessment (COA) previously developed for vigabatrin [12,13].
To our knowledge, to date, this is the largest cohort of early PD subjects undergoing a prospective and complete visual assessment.

Objectives
Primary. The primary objective is to determine the presence or absence of retinal change or other ophthalmologic deterioration from baseline to 2 years of pramipexole compared with ropinirole in subjects with idiopathic PD.
Secondary. The secondary objective is to assess and monitor safety profile and tolerability of pramipexole versus ropinirole in PD and to assess progression of PD during 2 years' treatment.
Post Hoc Exploratory. Post hoc exploratory objective is to analyse baseline ophthalmologic parameters in relation to age, sex, PD stage, and PD duration and to model their 2-year longitudinal changes with respect to age, sex, PD severity, duration, and treatment effects.

Study Design, Ethics, and
Structure. This open-label, randomized, flexible-dose, active-controlled, parallel-group, phase IV study (NCT00144300) included 21 neurology and 19 ophthalmology sites in the United States and was compliant with Declaration of Helsinki and Good Clinical Practice. The study was approved by Schulman Associates Institutional Review Board (IRB; Cincinnati, OH), Western Institutional Review Board (Olympia, WA), and/or local sites' IRBs (See Acknowledgments). All patients signed informed consent.
Electroretinography (ERG), visual field, Roth 28, and fundus photography were evaluated by central reading centers masked to treatment.

Inclusion and Exclusion
Criteria. Eligible patients were ≥30 years old with idiopathic PD of ≤7 years' duration, with modified Hoehn-Yahr stages I-III, and with ≤6 months' cumulative lifetime dopamine agonist exposure. Exclusion criteria included nonidiopathic PD, prior stereotactic brain surgery, and existing eye abnormalities (retinopathy, dense cataracts, and glaucoma; best-corrected visual acuity [ [14] and Unified Parkinson's Disease Rating Scale [15] parts II and III (UPDRS II+III) were administered at screening/baseline and after 1 year and 2 years of treatment. Baseline clinical assessments were performed off the designated study medication. Posttreatment assessments were performed on the study medication.

Ocular Status and Retinal Function.
Board-certified or eligible retinal ophthalmologists evaluated corrected acuity, ocular status, and retinal function at baseline and at 6, 12, and 24 months. Early Treatment Diabetic Retinopathy Study (ETDRS) acuity was determined at 4 meters for each eye and analyzed as "number of letters correct." Spherical and cylindrical refraction was measured.
Computerized visual field testing used the Zeiss-Humphrey Visual Field Analyzer with white-on-white, 30-2 SITAstandard threshold strategy with age-corrected normal comparisons.
Standard ERG (ISCEV Protocol) was measured for amplitudes and implicit times: dim flash dark-adapted b-wave; bright flash dark-adapted response, a-and b-wave; darkadapted oscillatory potentials (OPs)-sum of amplitudes of all individual OP wavelets; light-adapted 30 HZ cone flicker b-waves; and light-adapted single flash a-and b-wave. Roth 28-color test (R 28) axis and error score were determined for each eye.
Fundus photographic slides were evaluated for overall impression in the ETDRS-standard 7 designated fields for retinal hemorrhage or microaneurysm, hard exudates, retinal edema or whitening, RPE hyperpigmentation, or RPE hypopigmentation/drusen/pigment epithelial detachment.
Continuous data were dichotomized by the following thresholds from the authors' clinical experience: for IOP, a clinically established value of 21 mmHg was regarded as upper bound of normal. Similar transformations were done for acuity (<20/25, regarded as "fair" acuity by Gittings and Fozard [16]), cup-to-disc ratio (>0.75), mean deviation (<−2 dB), PSD (>5 dB), and Roth error score (>128-in comparison, 126 was the maximum error score for the youngest adults in Erb et al. [17]). ERGs were also converted into dichotomous normal/abnormal values using site-specific lower bounds of normal from 9 age-matched control ERGs per site [18].

Nonophthalmic Safety Evaluations.
Physical exams and electrocardiograms were performed at baseline; dermatologic exams were performed at baseline and 6 months; blood pressure, pulse, and incidence of orthostatic hypotension were tested at all visits; and routine blood tests were performed at baseline and months 12 and 24. Treatment-emergent adverse events (AEs) were summarized per body system and drug.

Outcomes
3.5.1. Prespecified. The primary outcome measure was presence or absence of COA change/deterioration from baseline to 2 years, adjudicated by the Expert Panel. Secondary outcomes included 1-year COA changes and prespecified 2year subgroup analyses of COA changes, efficacy (UPDRS parts II+III total; Hoehn and Yahr), and nonophthalmic safety profile.

Additional ERG
Analysis. An FDA-required reanalysis evaluated treatment effects on the differences between the 2year and baseline log 10 values of each ERG parameter. Change scores for both eyes were averaged per visit per patient (assuming bilateral systemic effects). Frequency distributions of changes were plotted and compared between the 2 drugs.

Exploratory Post Hoc.
Additional exploratory post hoc analyses characterized baseline, 1-year, and 2-year ophthalmologic data, as described below.
3.6. Statistical Analyses. The FDA agreed to an empirical minimum sample size of 100 subjects/group completing ≥12 months' treatment. Recruitment goal was 300 subjects.
Descriptive statistics were applied to primary and secondary results.
Cross-sectional baseline analyses modelled continuous outcomes by univariate and multivariate linear regression (acuity, cup/disk ratio horizontal and vertical, refraction spherical and cylindrical, IOP, mean deviation, pattern standard deviation, Roth error score, axis, and ERG parameters) for left and right eyes separately. Linear mixed all-eyes models included random effects for subjects with a compound symmetry correlation structure to account for within-subject dependency.
Logistic regression for dichotomous outcomes modelled abnormal response probability for left and right eyes separately. A generalized linear mixed model combined left-and right-eye data. Random effects for subjects were included in the logistic models, assuming a compound symmetry correlation structure. Covariates were age, sex, Hoehn-Yahr stage, and disease duration. Age and disease duration were categorized if empirical logit plots evidenced nonlinearity. Goodness-of-fit was assessed using the Hosmer-Lemeshow test. Odds ratios, 95% confidence intervals (CIs), and P values were derived for each model and covariate.
Longitudinal (on-treatment) analysis assessed ophthalmologic characteristics from baseline to 2 years similarly to the baseline assessment, but including drug assignment as a covariate and including baseline values as predictors of outcomes' change from baseline. Changes over 2 years were modelled against PD severity (baseline UPDRS parts II+III total); ERG amplitude changes from baseline were analyzed on a log 10 scale.

Treatment Parameters.
During the maintenance phase, 113 pramipexole recipients received a mean ± SD daily dose of 3.00 ± 1.21 mg, median 3 mg; 116 ropinirole recipients received a mean ± SD daily dose of 9.57 ± 5.21 mg, median 9.00 mg.

Baseline Ophthalmologic Findings.
Baseline best-corrected ETDRS acuity averaged 56 ± 6 letters (approximately 20/20 Snellen). To facilitate comparisons, we dichotomized continuous variables using clinically established normal cutoff values shown in Table 2 along with prevalence of baseline eye abnormalities.
At baseline, the NORDIC Fundus Photography Reading Center of the University of Rochester (Rochester, NY) graded 30.0% of patients as having retinal hypopigmentation in Table 2: Percentages of patients ( = 246) with abnormal ophthalmologic or ERG values at baseline (see text of baseline cross-sectional analyses). Normal cutoff values for non-ERG parameters are shown; patients' ERG parameter abnormalities were scored against site-specific lower bounds of normal determined from 9 healthy control subjects per site. The Roth error score threshold for abnormality used in baseline analyses was lower than the Expert Panel's threshold used for the on-treatment COA.

Measure
Normal Baseline ERG measures (Table 2) showed that >33.3% of patients had abnormal mixed rod-cone b-wave implicit times and 24.4% abnormal cone flicker amplitudes.

Exploratory Baseline Analyses.
Age significantly predicted 19/81 baseline ophthalmologic outcomes; sex predicted only 3/81 outcomes. Hoehn-Yahr stage was associated with visual acuity, horizontal and vertical cup-todisc ratios, motility smooth pursuit, and single flash cone response amplitude a-and b-waves. PD duration predicted lens condition, overall fundus photo clinical opinion, retinal hemorrhage or microaneurysm in inferior/superior nasal fields, RPE hyperpigmentation in the macula-centered field, and rod-cone mixed response implicit time b-wave. Effect sizes were generally small and deemed not clinically meaningful (Table 3).

Comprehensive Ophthalmologic Assessment (COA).
Expert Panel evaluations for the 2-year posttreatment COA (primary outcome) and the 1-year COA (secondary outcome) are shown in Table 4. Percentages of patients with 2-year COA deterioration, including those deemed clinically meaningful, did not differ significantly between drugs. No clinically meaningful ophthalmologic changes were deemed probably or definitely drug-related. Prespecified subgroup analyses (Table 5) were descriptive only.
UPDRS parts II+III sum scores improved in both groups after 1 year (mean changes from baseline: pramipexole −4.1; ropinirole −3.7) and 2 years (pramipexole −0.7; ropinirole −1.7). Table 6. Proportions with AEs or serious AEs (SAEs) were similar between groups. More ropinirole than pramipexole recipients had ≥1 severe or drug-related AE. There were no clinically significant changes in laboratory values.

Additional ERG Analysis.
In the FDA-requested additional ERG analysis, the plotted frequency distributions of log 10 change values between baseline and final treatment visits were similar for pramipexole and ropinirole.

Exploratory Post Hoc Longitudinal Analyses.
The prevalence of RPE hypopigmentation on fundus photographs was unchanged from baseline to 2 years (30% baseline; 28.9% at 2 years). The percent of patients with drusen increased from 47.4% at baseline to 58.2% at 2 years.
Longitudinal multivariate models probed relationships of baseline PD severity (UPDRS) with ERG changes over 2 years on treatment (Table 7). Adjustment for baseline ophthalmologic outcomes in these models was essential, as they were the strongest predictors of 2-year changes in their outcomes.

Interpretation of Prespecified Analysis Results.
Retinal deterioration over 2 years' treatment in subjects with PD did not differ between pramipexole and ropinirole, as assessed by masked independent Expert Panel COA review. Modified Hoehn-Yahr stages and UPDRS II+III scores indicated therapeutic effects of study drug dosages given for PD signs and symptoms. The drugs had similar incidences of overall AEs and SAEs and individual AEs; safety profiles were consistent with labeling.

Study Limitations.
Protocol violations among treated subjects included refractive errors exceeding −6 D (with waivers) in 2 pramipexole and 2 ropinirole patients, bilateral abnormal ERG and retinal infarction in 1 pramipexole patient each, macular degeneration and retinal hemorrhage in 1 ropinirole patient each, diabetes mellitus in 1 pramipexole and 1 ropinirole patient, orthostatic hypertension in 1 pramipexole and 1 ropinirole patient, and malignant melanoma in 1 ropinirole patient.
Noting the limitations of statistical power and nonplacebo design, the resulting FDA-approved pramipexole IR label revision [9] stated, "There was no statistical difference in retinal deterioration between the treatment arms; however, the study was only capable of detecting a very large difference between treatments. In addition, because the study did not include an untreated comparison group (placebo treated), it is unknown whether the findings reported in patients treated with either drug are greater than the background rate in an aging population."

Generalizability.
These findings may not be applicable to other dopamine agonists.

Prespecified Analysis Conclusions.
The COA results suggest no significant difference in risk for retinal deterioration between pramipexole and ropinirole in our study a For this analysis Roth error scores were dichotomized at the median value of 12, which was also the published minimum score for the youngest adult subjects in the study of Erb et al. 1998 [17]. b For this analysis Roth axis scores were dichotomized at the median value of 45. CI: confidence interval; PD: Parkinson's disease; RPE: retinal pigment epithelium. Table 4: Summary of Expert Panel's COA (full analysis set, LOCF). The Expert Panel defined worsening as any of the following (thresholds determined from clinical experience and reading centers' assessment of control subjects): Roth 28 error scores >295 right eye, >271 left; MD change of −2 dB confirmed by a later test; field defect clusters of 3 locations at < 0.05 or 2 at < 0.01; ERG parameter change scores (log 10 differences from log 10 baseline values) <2.56 times standard deviation of repeatability; fundus changes from "absent" to "obvious" for any finding; acuity loss ≥10 letters; pupil change from normal to abnormal; IOP >22 mmHg. Prespecified subgroup analyses of the COA results are shown in Table 5.

Interpretation of FDA-Required Additional ERG Analysis.
Frequency distributions of ERG log 10 change values between baseline and final visits were similar for pramipexole and ropinirole, suggesting no differential effects on retinal electrophysiology. For both drugs, the general symmetry of the distribution reflected the variability in the test-retest reliability.  [19] and inverse correlation of central minimum thickness with Hoehn-Yahr stage [20]). Retinal thickness distinguishes advanced PD patients from healthy persons [21]. Linear depigmenting RPE lesions occur in the Guam ALS/parkinsonism/dementia complex but are distinct from RPE hypopigmentation [22]. RPE hypopigmentation may be pathophysiologically related to PD because RPE cells produce dopamine [23]. Experimental chimeric loss of the retinal cell survival factor Ranbp in mice induced both RPE hypopigmentation and juvenile parkinsonism [24]. Drusen baseline frequency in our patients (47.4%) was generally similar to reported frequencies in geographically diverse studies in similar age groups [25][26][27][28].

Interpretation of Baseline Findings and Additional
Among other ERG findings, 33.3% of patients had abnormal rod-cone mixed response b-wave implicit times, 26.5% had abnormal single flash cone response a-wave implicit times, and 22.4% had abnormal cone flicker response amplitudes. Baseline UPDRS score was also a multivariate predictor of 2-year change in implicit time b-wave rodcone mixed response in our patients. Retinal dopamine may participate in cone flash responsiveness [29] and rod signaling [30,31]. A previous ERG study in treatmentnaïve PD patients observed "subtle increase in the latency of their short-wavelength sensitive cone response" similar to our implicit time results; conversely, levodopa-experienced patients had worse ERG responses in levodopa withdrawal, improving after intravenous levodopa [29].
Age was a multivariate predictor of these baseline continuous parameters: baseline acuity, echoing the Beaver Dam Eye Study's significant acuity decline between the 43-52-year a Eight subjects were assessed as "worse from baseline" based on responses carried forward (2 subjects in the pramipexole group and 6 in the ropinirole group). and ≥75-year age groups [32], spheroid, ERG amplitudes of cone flicker and b-wave rod responses and oscillatory potential, ERG a-wave, rod-cone mixed response implicit time (age effect also seen in [33]), Roth axis (Erb et al. cite increasing blue-yellow axis errors in seniors [17]), mean deviation [34], and pattern standard deviation. Age predicted these dichotomous parameters: abnormalities of cornea, lens, vitreous body, retinal hemorrhage or microaneurysm (stereo fields 3 and 7), RPE hyperpigmentation (stereo fields 5 and 7), visual field (reflecting published [35,36] but not universally reported [37] age patterns), dichotomized Roth error score and axis [17]. Baseline Roth scores were predicted only by age, which is not surprising in view of reported increases in Roth error rates with age [17]; however, 2-year Roth score change was significantly associated only with baseline Roth score. A different color test yielded significantly higher error scores in PD subjects than controls, and error correlated with disease severity [38,39], although another study found color errors in only 3 of 14 PD patients [40].
Age was a multivariate predictor of 2-year change in log 10 oscillatory potential, consonant with general ERG age patterns [33]. Gender was a multivariate predictor of 2-year change in log 10 cone flicker response and in Roth error score (the latter as in [17]). Other ERG 2-year changes were predicted only by their baseline values.

Limitations and Generalizability of Additional Analyses.
These results represent baseline ophthalmology findings and their longitudinal relationships only in the early idiopathic PD patients qualifying for inclusion.

Conclusions
The prespecified primary outcome, COA deterioration from baseline to 2 years scored by a masked Expert Panel, indicated no significant difference in risk for retinal deterioration in subjects with early idiopathic PD treated with pramipexole compared with ropinirole.
This study's finding of baseline RPE hypopigmentation was not previously reported in early PD and merits further study. Older age and more advanced Hoehn-Yahr stage significantly predicted lower visual acuity. Baseline ERG values and to some degree age significantly predicted most ERG changes at 2 years. Baseline Roth error score significantly predicted 2-year change in Roth error score, as did gender. Extensive ophthalmologic evaluation in this prospective cohort of early idiopathic PD patients suggests a possible Fatal SAE (nondrug related) 1 (0.8%): stab wound 0 (0%) a As defined by the investigator. b Marked hematological and other laboratory abnormalities (other than those meeting the definition of serious) and any events that led to an intervention, including withdrawal of test drug/investigational product treatment, dose reduction, or significant additional concomitant therapy, other than those reported as serious adverse events (SAE).  relation of aging and PD to the observed ophthalmologic findings and could be further evaluated in patients with PD and in age-matched controls.