Changes in Prescribing Practices of Dopaminergic Medications in Individuals with Parkinson's Disease by Expert Care Centers from 2010 to 2017: The Parkinson's Foundation Quality Improvement Initiative

ABSTRACT

Background

During the past decade, there has been increasing awareness of the side effects of dopamine agonists (DAs), including impulse control disorders. We hypothesized that there may be a shift toward more conservative use of DAs.

Objective

To explore the change in prescribing practices for dopaminergic medications in Parkinson's disease between 2010 and 2017.

Methods

Data were collected from the Parkinson's Foundation Quality Improvement Initiative registry. Baseline characteristics were compared between the 2010 and 2017 cohorts using chi‐squared tests for discrete and t tests for continuous variables. Logistic regressions were conducted for each class of medications to assess the effect of time points (2010 vs. 2017) and prespecified covariates on the probability of prescribing.

Results

A total of 2,717 participants from 2010 and 2,900 participants from 2017 were included in the analysis. Mean (standard deviation) age was 67.4 (10) and 68.7 (9.3) for the 2010 and 2017 cohorts, respectively (P < 0.0001). After controlling for baseline characteristics, DA use was unchanged (P = 0.1172). The odds of using monoamine oxidase B inhibitors was 52% higher in 2017 than in 2010 (P < 0.0001), 38% lower for catechol‐O‐methyltransferase inhibitors (P < 0.0001), 25% lower for amantadine (P < 0.0001), and 31% lower for anticholinergics (P = 0.0153). There was no difference in the utilization of levodopa in the 2 cohorts (86.1% vs. 86.2%; P = 0.5783).

Conclusions

Despite increasing awareness of impulse control disorders, there has been no reduction in the use of DAs during the past decade. Overall, there is less utilization of adjunctive classes of drugs except for an increase in the use of monoamine oxidase B inhibitors.

Levodopa continues to be the gold standard for treatment of Parkinson's disease (PD); however, the armamentarium of treatment options for PD has grown in recent years, allowing for more nuanced symptomatic management of the disease. This includes a better understanding of the risks and benefits of adjunctive medications as well as the development of longer acting formulations of levodopa and other classes of medications. There have been 4 evidence‐based medicine reviews for the treatments of PD published by the Movement Disorder Society to guide practitioners about which pharmacological, surgical, and psychosocial interventions are useful.1 Of course, one of the major advancements in our treatment of the motor symptoms of PD has been the use of deep brain stimulation. With more detailed studies of surgical outcomes,2 we have begun to better select the site of stimulation based on each individual patient's disease characteristics. Similarly, as more research studies have focused on the pharmacologic management of PD motor symptoms, it is likely that the landscape of medical management styles has changed.

Previous studies, including a retrospective database analysis in Taiwan comparing prescribing practices from 2004 to 2011, have demonstrated a trend of increased nonergot dopamine agonist (DA) use in younger aged participants.3 Recently, however, several studies have been published recognizing impulse control disorders (ICDs) as a side effect of DAs.4, 5, 6 In a large cross‐sectional study of more than 3,000 participants, it was demonstrated that there was a 2‐fold to 3.5‐fold increased odds of having an ICD with DA therapy.4 Several years later, a retrospective analysis was done using the U.S. Food and Drug Administration's Adverse Event Reporting System supporting that DAs have a strong signal associated with ICDs.5 Moreover, in a post hoc analysis of 6 open‐label extension studies up to 6 years in duration it was demonstrated that there was a 9% frequency of ICD behavior as an adverse effect of the use of the rotigotine transdermal patch.6 With increased recognition of the adverse effects of DAs, we suspected that there may be a shift toward more conservative use. In younger participants, this shift was suggested in a study analyzing inpatient data acquired from the Cerner electronic medical record database (Culver, CA) in the United States between 2001 and 2012, although DA use in older individuals remained stable between those 2 time points.7

More recent studies have shown persistent benefits for patient‐related mobility scores in patients treated initially with levodopa as compared with DAs or monoamine oxidase type B (MAO‐B) inhibitors,8 likely further demonstrating to providers the benefits of levodopa therapy even early in the disease. In addition, currently there is convincing preclinical and clinical data refuting prior concerns regarding a potentially toxic long‐term effect of levodopa.9, 10 Although the most recent study did not demonstrate disease‐modifying effect of the earlier initiation of levodopa, there was also no deleterious effect.10 Furthermore, recent work has demonstrated risks with the use of anticholinergic medications.11, 12, 13, 14

Despite a large number of data on the efficacy of various pharmacological interventions in PD derived from well‐designed controlled studies, there is still a paucity of information on the real‐life use of various classes of PD therapeutics. Using the Parkinson's Foundation Quality Improvement Initiative registry collected at participating expert care centers, we analyzed the recent changes in the prescribing practices for the motor symptoms of PD as well as the present‐day prescribing landscape.

Methods

Data were collected from the Parkinson's Foundation Quality Improvement Initiative registry, which is an international, multicenter, prospective observational study at 23 participating centers of excellence.15 The study protocol was approved by the ethics committee at each participating site. Any person who receives medical care for the diagnosis of PD at participating centers is eligible for participation in the database, and each participant gives informed consent to have his or her data included. The Parkinson's Outcomes Project data used to support the findings of this study were supplied by the Parkinson's Foundation under license. Requests for access to these data should be made to the Research Department at the Parkinson's Foundation. The initial database was launched in 2009 and includes data on participant demographics, PD severity, medications, cognitive function, quality of life, and other variables.15 Individuals who participate are seen annually for the purpose of this registry. At the time of data download, 2017 was the most recent completed year of available registry information.

All participants who had a visit in 2010 and all participants who had a visit in 2017 were initially included. Participants with deep brain stimulation (343 participants in the 2010 cohort and 504 participants in the 2017 cohort) and/or Duopa therapies (AbbVie Inc. Chicago, IL; 45 participants in the 2017 cohort) were excluded. Among the remaining, there were 610 participants who had visits in both 2010 and 2017. These participants were excluded from the 2017 cohort so as to not include the same participant twice.

Baseline characteristics were compared between 2010 and 2017 cohorts using chi‐squared tests for discrete and t tests for continuous variables. A P value of <0.05 was considered significant. Logistic regressions were conducted for each class of medications to assess the effect of time points (2010 vs. 2017) and prespecified covariates on the probability of prescribing. The year 2010 was chosen as the first time point as this was the first complete calendar year during which the registry was in effect and enrolled more than 1000 participants. Considering the large variability in the baseline demographic and clinical characteristics of the participants, our analyses were adjusted for age, disease duration, Hoehn & Yahr (H&Y) stage, TUG (Timed Up and Go) time, and combined cognition status. Combined cognition status was calculated as the average of the delayed 5‐word recall z score and the verbal fluency z score.

In 2016, investigators updated the dataset to include the quantitative data of all dopaminergic therapies, thus allowing for the calculation of levodopa equivalent dosing (LED). A linear regression analysis of the 2017 cohort was performed for the purposes of this analysis, and 6 significant outliers were removed from that analysis. In the linear regression model for the 2017 cohort, LED was the response variable, and the covariates were the same as those used in the generalized linear model. All statistical analyses were performed using R software (R Foundation for Statistical Computing, Vienna, Austria).

Results

A total of 2,717 participants who had a visit in 2010 and 2,900 participants who had a visit in 2017 were ultimately included in the analysis. Demographic and baseline disease characteristics of both cohorts are presented in Table 1. Participants in the 2017 cohort were older, had longer disease durations, were more likely to have a rest tremor, had a shorter Timed Up and Go time, and had a lower Parkinson's Disease Questionnaire–39 total score. The differences, although statistically significant, were small, as seen in Table 1. For example, the average age (standard deviation) was 67.4 (10) and 68.7 (9.3) for the 2010 versus 2017 cohorts, respectively. There were no differences in sex, certainty of diagnosis, verbal fluency, H&Y stage, or motor fluctuations.

Table 1.

Baseline demographics and disease characteristics

Variable	2010	2017	P Value
Total patients, n	2717	2900
Age, y, mean ± SD	67.4 ± 10.0	68.7 ± 9.3	<0.0001
Male, n (%)	1649 (60.7)	1814 (62.7)	0.124
Stands unaided, n (%)	2451 (92.1)	2596 (89.9)	0.0048
Age at onset, y, mean ± SD	59.3 ± 10.9	60.2 ± 10.3	0.003
Year of first onset of PD symptoms, mean ± SD	2001.5 ± 5.7	2008.2 ± 5.8	<0.0001
Year of PD diagnosis, mean ± SD	2003.2 ± 5.3	2009.7 ± 5.3	<0.0001
Disease duration, y, mean ± SD	8.5 ± 5.7	8.8 ± 5.8	<0.0001
Certainty of idiopathic PD diagnosis, n (%)
<50%	58 (2.2)	78 (2.7)	0.298
50%–89%	344 (12.8)	345 (12.1)
≥90%	2290 (85.0)	2431 (85.2)
Rest tremor present, n (%)	1903 (70.5)	2195 (76.2)	<0.0001
Motor fluctuations present, n (%)	1290 (47.8)	1316 (45.8)	0.144
H&Y stage, n (%)
1	312 (12.1)	315 (11.4)	0.055
2	1380 (53.4)	1550 (56.2)
3	655 (25.3)	692 (25.1)
4–5	238 (9.2)	201 (7.3)
TUG, seconds, mean ± SD	15.2 ± 8.8	12.7 ± 6.5	<0.0001
Combined cognition status, mean ± SD	−0.1 ± 0.8	0.1 ± 0.8	<0.0001
Immediate 5‐word recall, mean ± SD	4.3 ± 1.0	4.4 ± 0.9	0.002
Verbal fluency, mean ± SD	18 ± 6.7	18.4 ± 6.7	0.07
Delayed 5‐word recall, mean ± SD	2.8 ± 1.4	3.2 ± 1.5	<0.0001
PDQ‐39, mean ± SD
Total	24.6 ± 15.8	22.1 ± 14.5	<0.0001
Mobility	12.8 ± 11.4	11.2 ± 10.7	<0.0001
ADL	7.6 ± 6.1	6.5 ± 5.6	<0.0001
Cognition	4.3 ± 3.3	3.9 ± 3.0	<0.0001
LED, mg/day, mean ± SD		698.9 ± 714.3

Combined cognition status = average of delayed 5‐word recall and verbal fluency.

SD, standard deviation; PD, Parkinson's disease; H&Y, Hoehn & Yahr; TUG, Timed Up and Go; PDQ, Parkinson's Disease Questionnaire; ADL, activities of daily living; LED, levodopa equivalent dosing.

Table 2 provides the descriptive statistics of the percentage of participants taking various pharmacotherapies in the 2010 versus 2017 cohorts. When compared with the 2010 cohort, participants in the 2017 cohort were less likely to be prescribed DAs (43.2% vs. 39.4%; P = 0.004), amantadine (17.7% vs. 15.1%; P = 0.009), catechol‐O‐methyltransferase (COMT) inhibitors (17.5% vs. 11.5%; P < 0.0001), and anticholinergics (4.8% vs. 3.6%; P = 0.03), but there was a higher overall use of MAO‐B inhibitors (22.4% vs. 30.0%; P < 0.0001). In the 2010 cohort, 68 (2.5%) of the patients were not prescribed any PD medication, 588 (21.6%) patients were only prescribed 1 PD medication, and the remaining 2,061 (75.9%) patients were prescribed 2 or more PD medications. In the 2017 cohort, 123 (4.2%) patients were not prescribed any PD medications, 600 (20.7%) patients were only prescribed 1 PD medication, and the remaining 2,177 (75.1%) patients were prescribed 2 or more PD medications.

Table 2.

Comparison in number (%) of prescriptions between 2010 and 2017 and LED in 2017^*

Medication	2010	2017	P Value	LED (mg/day) in 2017
Any form of levodopa	2343 (86.2)	2496 (86.1)	0.888	518.4 ± 435.3
Dopamine agonist	1173 (43.2)	1142 (39.4)	0.004	93.5 ± 533.3
MAO‐B inhibitor	608 (22.4)	871 (30.0)	<0.0001	30.3 ± 58.7
COMT inhibitor	475 (17.5)	334 (11.5)	<0.0001	24.3 ± 77.1
Amantadine	480 (17.7)	437 (15.1)	0.009	32.5 ± 85.2
Anticholinergic	131 (4.8)	105 (3.6)	0.030

^*Missing values of the use of medication in the dataset were categorized as “no medication use” for the analysis.

LED, levodopa equivalent dose; MAO‐B, monoamine oxidase B; COMT, catechol‐O‐methyltransferase.

After controlling for baseline characteristics including age, disease duration, H&Y stage, Timed Up and Go time, and cognition status, the odds of using MAO‐B inhibitors was 52% higher in 2017 than 2010, 38% lower for COMT inhibitors, 25% lower for amantadine, and 31% lower for anticholinergics (Table 3). All of these differences were statistically significant. There was no significant difference in the utilization of DAs (P = 0.1172). There was no significant difference in the utilization of levodopa in the 2 cohorts (P = 0.5783).

Table 3.

AOR of prescribing medication comparing 2017 to 2010 after controlling for age, disease duration, H&Y stage, TUG, and combined cognition status

Medication	AOR (95% Confidence Interval)	P Value
Any form of levodopa	0.95 (0.80–1.13)	0.5783
Dopamine agonist	0.91 (0.80–1.03)	0.1170
MAO‐B inhibitor	1.52 (1.33–1.74)	<0.0001
COMT inhibitor	0.62 (0.52–0.73)	<0.0001
Amantadine	0.75 (0.63–0.88)	<0.0001
Anticholinergic	0.69 (0.51–0.93)	0.0153

Combined cognition status = average of delayed 5 word recall and verbal fluency.

AOR, adjusted odds ratio; H&Y, Hoehn & Yahr; TUG, Timed Up and Go; MAO‐B, monoamine oxidase B; COMT, catechol‐O‐methyltransferase.

For the analysis of the cumulative dose of PD medications as expressed by LED in the 2017 cohort, linear regression analysis demonstrated that an increase in age by 1 standard deviation (9.6 years) was associated with 59 units less in LED (P < 0.0001) as seen in Table 4. An increase in disease duration by 1 standard deviation (5.7 years) was associated with 150 units more in LED (P < 0.0001). An H&Y stage increase of 1 was associated with 113 units more in LED (P < 0.0001).

Table 4.

Effect of age, disease duration, H&Y stage, TUG, and combined cognition status on levodopa equivalent dose

Variable	Effect Estimate	P Value
Age, y (SD = 9.6)	−58.84	<0.0001
Disease duration, y (SD = 5.7)	149.97	<0.0001
H&Y stage (stage 1 as reference)
2	178.32	<0.0001
3	300.71	<0.0001
4–5	175.47	0.0281
TUG, seconds (SD = 7.8)	17.15	0.2948
Combined cognition status	4.91	0.7427

Combined cognition status = average of delayed 5‐word recall and verbal fluency.

H&Y, Hoehn & Yahr; TUG, Timed Up and Go; SD, standard deviation.

Discussion

We conducted an analysis of the change in prescribing patterns of various classes of PD medications during the past decade. Overall, these data suggest that despite the data reporting ICDs as a common side effect of DA use, participants in 2017 were just as likely to be prescribed a DA when compared with participants in 2010 after controlling for age and disease characteristics.

It was previously recommended that DAs be used for the treatment of PD, particularly in younger patients, with the rationale that DAs could be used as a levodopa‐sparing strategy to delay the onset of motor complications.16, 17, 18 Recently, multiple prospective studies have demonstrated that initial therapy does not affect later severity of motor fluctuations and dyskinesias.8, 19 As an argument against the use of DAs, there is an increasing recognition of a higher risk of ICD behavior with DA versus levodopa therapy. In 2018, a longitudinal analysis of ICDs in PD was published demonstrating a strong dose‐effect and length‐of‐duration association,20 suggesting that perhaps DAs should be used at lower doses for a shorter period of time. Despite our hypothesis, we did not identify a change in the use of DAs between 2010 and 2017.

We also analyzed the change in the use of other classes of adjunctive PD therapies during a span of 7 years. Interestingly, overall there was higher utilization of MAO‐B inhibitors. This was somewhat unexpected given their low potency and given that studies such as the attenuation of disease progression with Azilect given once‐daily trial were inconclusive regarding a possible disease‐modifying effect.21 Given the conflicting results in that study, an U.S. Food and Drug Administration panel voted against a bid to get extended approval for Azilect (Teva Pharmaceutical Industries, Petah Tikva, Israel) as a medication that could slow the progression of the disease. The precise reasons for increased utilization of MAO‐B inhibitors are unclear and might be driven by their good safety profile and ease of use with once‐daily administration of the drug. In addition, we found that COMT inhibitors were less used in 2017, which is likely because of comparable efficacy to MAOs but that they require multiple dosage times. If newer and more potent formulations (such as Opicapone; BIAL, Portela & Cª., S.A., Spain and Neurocrine Biosciences, San Diego, CA) become available, this may change.

Overall, our analysis demonstrated a reduction in several of the classes of adjunctive treatment. In addition to COMT inhibitors, this included amantadine and anticholinergic medications. Less use of amantadine was unexpected because it is the primary drug used for drug‐induced dyskinesias, but it is possible that given the recent data published regarding AmantadineER (Adamas Pharmaceuticals, Emeryville, CA),22 the use of amantadine will increase in the coming years. In general, it is possible that there has been a reduction in the use of the adjunctive treatment classes because of their overall lower potency and thus higher pill burden and cost for patients. One could also argue that the past decade has been more focused on levodopa optimization and new delivery systems as opposed to the development of new types of adjunctive drug therapies.

Regarding anticholinergic use in particular, it has been demonstrated that the anticholinergic burden in patients with PD correlates with poor outcomes. A study in the United States demonstrated that higher anticholinergic use was associated with an increased risk of fractures, delirium, emergency department visits, and readmissions to the hospital.12 Furthermore, a recent study demonstrated that PD patients with a high cumulative dose of anticholinergics had an increased risk of being diagnosed with dementia.11 In addition, a recently published case‐control study demonstrated in a cohort of elderly patients that exposure to several types of strong anticholinergic drugs was associated with an increased risk of dementia.13 This further suggests that providers should be cautious about anticholinergic drug usage in the PD patient population.

Overall, levodopa prescribing did not change significantly from 2010 to 2017. Interestingly, however, our analysis of prescribing practices in 2017 found a significant decrease in LED with increasing age, perhaps because older patients are less tolerant to side effects. As expected, LED increased with disease duration and higher H&Y stages. For many years, physicians have practiced levodopa‐sparing strategies; however, as we have cited there is a growing body of evidence to suggest that there are no negative effects of early initiation of levodopa.9, 10 Importantly, although the recently published Levodopa in EArly Parkinson's disease study of early versus delayed initiation of levodopa therapy failed to demonstrate a disease‐modification effect, it demonstrated that carbidopa‐levodopa treatment did not have a detrimental effect on participants.10 Studies such as this may continue to result in an overall reduction in the use of adjunctive classes of medications and an increase in the use of levodopa.

The strengths of this study include the large sample size, multicenter data collection, and use of a dataset collected during regular clinical visits for treatment and monitoring of the patient's disease. These strengths increase the external validity of the study, although our results are not reflective of nonexpert care center practices. A limitation of this analysis is that quantified PD medication dosing data were not available in the 2010 survey to calculate LED. It is possible that the overall use of DAs as a class in 2017 appeared no different even though the LED might have been lower. Unfortunately, a conclusion regarding this cannot be made with this particular data analysis.

In conclusion, despite our prediction, our analysis demonstrated no change in DA use, but a reduction in the use of several classes of adjunctive treatments. The latter findings potentially reflect an overall trend toward the increased use of levodopa and a shift away from levodopa‐sparing strategies. Certainly, the landscape of prescribing practices will continue to evolve as new therapeutic options become available and as we focus more on the nonmotor manifestations of PD.

Author Roles

(1) Research project: A. Conception, B. Organization, C. Execution; (2) Statistical Analysis: A. Design, B. Execution, C. Review and Critique; (3) Manuscript: A. Writing of the first draft, B. Review and Critique.

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

S.W.: 2A, 2B, 2C, 3B

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

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

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

Disclosures

Ethical Compliance Statement

Each center applied for institutional review board approval for participation in the National Parkinson Foundation (NPF) database as outlined in “Piloting the NPF Data‐Driven Quality Improvement Initiative.” Informed patient consent was not necessary for this work. Informed consent for inclusion in the dataset was obtained by the Parkinson's Foundation Quality Improvement Initiative investigators. 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 Conflict of Interest

The authors report that no specific funding was received for this work. The authors declare that there are no conflicts of interest relevant to this work.

Financial Disclosures for the Previous 12 Months

T.S. has served as a consultant and received consulting fees from Acadia, Abbvie, Allergan, Anavex, Avid, GE Medical, Eli Lilly and Company, Harbor, Ibsen, IMPAX, Lundbeck, Merz, Inc., the National Parkinson Foundation, Navidea, Pfizer, TEVA Pharmaceuticals, Union Chimique Belge (UCB) Pharma, Voyager, US World Meds, and the Michael J. Fox Foundation (MJFF) for Parkinson's Research. T.S. has served as a speaker and received honoraria from Acadia, IMPAX, Lundbeck, TEVA Pharmaceuticals, and UCB Pharma and is on the scientific advisory boards for Anavex, Sanofi, and MJFF. T.S. sits on the advisory board for IMPAX and has received research funding from the National Institute of Neurological Disorders and Stroke (NINDS), MJFF, NPF, TEVA Pharmaceuticals, Auspex, Biotie, Civitas, Acorda, Lundbeck, Neuroderm, NINDS, National Institutes of Health, Northwestern Foundation, and the MJFF for Parkinson's Research. T.S. received funding support for educational programs from GE Medical, TEVA Pharmaceuticals, and Lundbeck. OD has served as a consultant and received consulting fees from Lundbeck.