Journal of Neurology

, Volume 257, Supplement 2, pp 309–313

Evidence-based initiation of dopaminergic therapy in Parkinson's disease


    • The Movement Disorders Centre, Toronto Western HospitalUniversity Health Network

DOI: 10.1007/s00415-010-5718-x

Cite this article as:
Miyasaki, J.M. J Neurol (2010) 257: 309. doi:10.1007/s00415-010-5718-x


The mainstay of Parkinson's disease (PD) therapy is levodopa. The crucial question is when should levodopa be initiated? Levodopa provides the most potent motor benefit for PD, but longer term use is marked by the development of motor complications such as fluctuations in response and involuntary motor movements. Dopamine agonists reduce the risk of development of motor complications in the 5-year term. However, side effects may change the risk-benefit of dopamine agonist first strategies. In the following, the evidence for levodopa and dopamine agonists as initial monotherapy for PD is examined.


Parkinson diseaseInitiation of therapyEvidence-based practiceDopamine agonistsLevodopa


The introduction of levodopa (LD) to Parkinson's disease (PD) therapy was hailed as miraculous, yet even initially was associated with dyskinesias in some patients with post-encephalitis parkinsonism [32]. In fact, LD was known to result in motor complications in as many as 80% of young patients and 44% in older patients after 5 years of use [13]. At times, dyskinesias are so severe that patients are more disabled by dyskinesias than by the symptoms of PD itself. Therefore, various surgeries were developed to address these motor complications such as pallidotomy of the globus pallidus interna and later deep brain stimulation (DBS) of the same target and the subthalamic nucleus (STN).

Investigators suggested LD may be “toxic,” resulting in faster deterioration of the substantia nigra [5]. Preclinical studies supported that LD was both toxic and neuroprotective [3, 16]. Several clinical studies examined LD compared to dopamine agonists (DA) as initial monotherapy.

Initiating dopaminergic therapy

Based on evidence cited in the practice parameter sponsored by the American Academy of Neurology (AAN) [17], either LD or a DA could be used to initiate therapy. The evidence used to make this recommendation included the results of pramipexole compared with levodopa as initial monotherapy in early PD (CALM-PD) [22]. This was a Class I study according to AAN rating schemes. In their online Clinical Practice Guideline Process Manual (, Class I studies should be prospective, randomized, controlled clinical trials with masked outcome assessment in a representative population. The primary outcome must be clearly defined with exclusion and inclusion criteria clearly defined, adequate accounting for drop-outs and cross-overs with numbers sufficiently low to have minimal potential for bias, relevant baseline characteristics are presented and substantially equivalent among treatment groups or there is appropriate statistical adjustment for differences. The CALM-PD study was a 2-year study of 301 early onset subjects randomized to either levodopa or pramipexole groups. The primary endpoint was the onset of motor complications, either motor fluctuations or dyskinesias. At the end of 2 years, 51% of all LD subjects developed motor fluctuations and 31% developed dyskinesias. By comparison, 28% of the DA subjects developed motor complications, and only 10% developed dyskinesias. At the end of the study, only 32% of subjects remained on pramipexole monotherapy. However, there was a significant difference in the improvement of motor signs as measured by the unified Parkinson's disease rating scale (UPDRS). At the end of 23.5 months, the LD group was on average 7.3 points improved compared with 3.4 for pramipexole on part III of the motor UPDRS (P < 0.001). Part II of the UPDRS (ADLs) was “significantly” different in favor of LD (P < 0.001); however, the difference between LD and pramipexole was only 1 point and likely not clinically significant. There was significantly more leg edema, somnolence and hallucinations in the pramipexole-initiated group compared with the LD group. The number needed to harm (NNH) for somnolence was 3, whereas the NNH was 15 for hallucinations (that is, three subjects need to be exposed to pramipexole for one additional person to experience somnolence, whereas 15 subjects need to be exposed to pramipexole for one additional person to experience hallucinations).

After a mean follow-up of 6 years, the same cohort had similar Schwab and England Activities of Daily Living Scale scores in the on and off states [21]. Mean changes from baseline as a measure by the UPDRS part III did not differ for the pramipexole-first compared with the LD-first cohort. However, a significant difference continued with 68% of initial LD subjects having dopaminergic motor complications compared with 50% of the DA-first group (P < 0.002). The mean Epworth Sleepiness Scale score was significantly higher in the pramipexole group (11.3 mean score) compared with LD-first subjects (8.6). Other complications such as impulse control disorders were not explored in this study.

A Class II study compared ropinirole with LD [27]. In this study, 268 subjects were randomized to ropinirole or levodopa for 5 years. In the CALM-PD study, subjects had an initial titration phase for the original drug assignment and then added levodopa if further motor benefit was required. In comparison, the ropinirole study allowed subjects to continue on the original drug assignment for a longer period of time with levodopa rescue later added. This may have contributed to the higher drop-out rate (49%). The study was downgraded due to the higher drop-out rate since the results may be difficult to replicate in the clinical setting. Similar to the pramipexole study, Rascol et al. [27] found that the motor UPDRS improved more for the LD initiated subjects (4.8 improvement on part III UPDRS compared with 0.8 improvement for the ropinirole group), and this difference was significant (P < 0.008). However, the ADL portion of the UPDRS was not significantly different for the LD- and ropinirole-initiated group. The primary endpoint for the ropinirole study was the appearance of dyskinesias either reported by the patient or observed by the physician. Dyskinesias were present in 45% of the LD-initiated subjects compared with 20% in the ropinirole group. This was significant. However, in this group only 16% of subjects were able to maintain a ropinirole monotherapy regimen at 5 years. A significant increase in hallucinations (17 vs. 6%), edema (14 vs. 6%) and somnolence (27 vs. 19%) occurred in the ropinirole group. This results in a NNH for hallucinations of 9 and somnolence NNH = 10.

The titration phase for the initial drug assignment and duration of the study were the only differences in these studies, and they showed remarkably similar findings. A further study of cabergoline compared with LD replicated these findings.

Both the pramipexole and ropinirole studies incorporated imaging as a secondary endpoint. The results of the studies were conflicting—even within a single study—due to changes in software analysis packages [23, 24]. Both fluorodopa PET and βCIT-SPECT imaging using dopamine transporters measure pre-synaptic conditions and dopaminergic nerve terminals. They do not measure the state of D2 receptors that may be affected by dopaminergic therapy. Further, dopamine transporters are affected by dopaminergic therapy and may result in the spurious interpretation of disease progression rather than the mere presence of dopaminergic therapy. To date, disease progression cannot be measured by one technique alone and likely will require at the minimum pre- and post-synaptic imaging modalities.

Additional information comes from the UK Parkinson's Disease Research Group [11]. Subjects (N = 782) were randomized to levodopa, levodopa plus selegiline or bromocriptine, and followed between 1985 and 1990. Subjects could be “re-randomized” if intolerance occurred. If subjects were initially assigned to levodopa and selegiline, in 1995 they were reassigned to levodopa or bromocriptine alone. Subjects were assessed using the Webster and Northwestern University Disability scale. Statistical adjustment for differences in baseline characteristics was performed.

By 14 years, 496 deaths had occurred (63%). A small but significant difference in Webster scores indicated LD resulted in superior benefit (3-point difference in scores). However, the rate of motor complication development including dyskinesias was not significantly different for LD compared to bromocriptine initial assignment. There was a trend towards improved cognition and less dementia as measured by the mini-mental status examination and components of the quality of life scores for the LD group. Dyskinesias were associated with lower quality of life as measured by the SF-36 and poorer general health perception, while fluctuations were associated with lower general mental health scores.

In terms of the impact of dyskinesias, other investigators found that dyskinesias were associated with improved motor benefit and therefore improved quality of life [15]. Marras et al. [15] examined the data from the above-mentioned CALM-PD cohort and found that Quality of Life as measured by the PDQUALIF was 3 points (of 100 possible points with higher scores reflecting improved quality of life) than those without dyskinesias. This effect was even more marked for older patients. Motor fluctuations, by comparison, did not result in changes in quality of life. Therefore, at least within the first 4 years of dopaminergic therapy, quality of life was not adversely affected by dyskinesias. Intuitively, these results make sense since elderly patients without dyskinesias do not have as much benefit from LD and may also have cognitive decline. However, Montel et al. [18] found dyskinesias resulted in lower quality of life. Subjects also reported coping strategies for their illness. Avoiding situations or activities due to dyskinesias was associated with poor quality of life.

Levodopa toxicity

Virtually from the introduction of LD, investigators explored the possibility of LD toxicity. Preclinical studies have conflicting results although some continue to search for mechanisms of LD neurotoxicity [5]. At least in cell cultures, levodopa may be toxic, probably through dopamine autoxidation and semiquinone formation to produce hydrogen peroxide [1, 9]. Preclinical studies demonstrated free radical damage in the substantia nigra that may be potentiated by dopamine oxidation [20, 29]. Other studies show a putative protective effect of levodopa through upregulation of glutathione levels or neurotrophic factors [3, 16]. Dopamine agonists may exert neuroprotection through antioxidant properties and decreased turnover of levodopa, and may reduce free radical formation by its metabolism and diminish excitotoxicity by reducing subthalamic nucleus activity. By restoring dopamine levels, LD may avoid the changes in basal ganglia circuitry that can result in dyskinesias. Further, cell models indicate that LD may also have antiapoptotic effects [14, 19].

In the face of conflicting preclinical data, the germane question is what have clinical trials concluded? The ELLDOPA trial exposed subjects not yet requiring dopaminergic therapy to placebo or three doses of levodopa. Subjects were randomized 1:1:1:1 to placebo or three LD doses 150/32.5 mg/day, 300/75 mg/day and 600/150 mg/day with 40 weeks’ exposure and then a 2-week washout [6]. The primary outcome measure was the change in the UPDRS between baseline and 42 weeks. The placebo group deteriorated by 7.8 points, 1.9 points in the 150 mg group, 1.9 points for the 300 mg daily group and an improvement of 1.4 points for the 600 mg daily group. However, the levodopa 600 mg group had more dyskinesia, hypertonia and nausea compared with placebo. The wash-out period was likely not long enough to completely discount a prolonged effect of levodopa. Although the evidence is not sufficient to infer a neuroprotective benefit of LD, the ELLDOPA study confirms that LD is not toxic in the short term. The many studies of LD versus DA as initial monotherapy or in combination therapy also do not demonstrate a toxic effect. Therefore, LD is not toxic in PD and continues to be the mainstay of treatment.

Non-motor complications with dopaminergic therapy

Although DAs clearly result in a lower risk of motor complications, many investigators would suggest that non-motor complications are more disabling [26]. Qin and colleagues reported that motor factors explained only 19% of variance of total Short Form 36 scores (for quality of life measure). In contrast, non-motor variables explained 62% of SF-36 variation. Specifically, they found that depression, sleep disorders and fatigue accounted for most of the reduction in quality of life. Kim et al. [12] found that despite initiation of dopaminergic therapy, non- motor symptoms and non-motor symptoms assessment scale scores did not change in newly diagnosed PD patients. In advanced patients, non-motor symptoms clustered to strongly correlate with stage. These identified symptoms were cognitive impairment, autonomic dysfunction, psychotic symptoms, depression, daytime sleepiness and axial symptoms [30].

All studies of DAs compared to LD reveal a higher rate of somnolence and hallucinations [22, 25, 27]. Recent reports highlight the importance of impulse control disorders (ICDs) in PD. ICDs are defined as a failure to resist an impulse, drive or temptation to perform an act that is harmful to the person or others [2]. The most common behaviors reported are pathological gambling, problem shopping, hypersexuality and binge-eating. Among various investigators the incidence of ICDs ranged from 6.1–14% [7, 31]. Mean losses for pathologic gambling in one centre was $124,000 CD (Canadian dollars) [31].

The risk of ICD development appears to be strongest with dopamine agonist use. The absolute risk increase with DA use was 13% [31]. Weintraub et al. [33] in 46 PD centers in the US and Canada interviewed 2,090 PD patients for the presence of problem gambling, shopping, hypersexuality and binge-eating. Sixty-six percent of patients were taking a DA, and 86.8% of patients were taking LD. At least one ICD was identified in 13.6% of patients while 36% of ICD patients had two or more ICDs. Problem gambling was reported in 5%, compulsive shopping in 5.7%, binge-eating in 4.3% and hypersexuality in 3.5%. For those not taking DAs, only 6.9% reported ICDs. Those currently taking DAs had a 17.1% incidence. For those on pergolide, 22% had DAs.

Excessive daytime somnolence is also common in PD. Suzuki et al. [28] found significantly higher Epworth Sleepiness Scale scores among 188 PD patients compared with 144 age-matched controls. Scores were higher for those on higher doses of dopaminergic agents and more disease severity. A survey of neurologists in Canada found that 4% of neurologists tell patients not to drive when they take DAs [4]. The remainder of neurologists monitor patients more closely when taking DA compared with LD (P < 0.001). A survey of 232 patients with PD followed at 4 and 8 years found that excessive daytime sleepiness correlated with older age, gender and use of DAs [8]. LD was associated with the lowest risk of daytime sleepiness [2.9, 95% confidence interval (CI) 1.7–4%] followed by DA monotherapy (5.3%, CI 1.5–9.2%), and LD and DA combination therapy (7.3%, CI 6.1–8.5%) [26]. The odds ratio for DA therapy causing excessive daytime sleepiness was 2.9 compared with 1.9 for LD. A larger study of 638 consecutive subjects in multiple Canadian Movement Disorders Centres found a 51% of sleepiness when the Epworth Sleepiness Scale score cutoff was 7 or higher [10]. In this study, the majority of subjects were on multiple drugs so that individual drug effects could not be determined. These results mean that 51% of subjects would not be able to drive due to sleepiness.


Based on the evidence discussed, it is clear that LD is not toxic. Initiation of dopaminergic therapy should take both motor benefit and the risk of non-motor complications into account. In terms of motor benefit, LD provides superior motor benefits while being associated with a higher risk of dyskinesias and motor fluctuation development. However, the impact of dyskinesias on quality of life is controversial. Katzenschlager et al. [11] documented no significant differences in motor function between initial bromocriptine compared with initial LD subjects after 14 years (the longest such published follow-up).

Recent research on non-motor complications of PD hints that these symptoms may influence quality of life more than dyskinesias. At least in terms of daytime sleepiness, hallucinations and ICDs, DAs clearly result in a higher risk for the development of these symptoms.

Based on the evidence presented, the DA monotherapy strategy should be re-evaluated. Patients should make their treatment decisions based on a discussion of the potential risks and benefits of both DA and LD. Decisions may be based solely on motor benefit, avoidance of motor complications or minimizing non-motor complications. Each patient will have different values and needs for therapy, and the well-informed neurologist is best situated to guide their therapeutic choices.


The author gratefully acknowledges Ms. Rhoda Ortiz for her administrative assistance.

Conflict of interest

Dr. Miyasaki has received research grants from NIH, Teva, Medivation, speaking honorarium from Teva and an unrestricted educational grant from Teva.

Copyright information

© Springer-Verlag 2010