Idiopathic Parkinson's disease (PD) is the second most common neurodegenerative disorder. Although it is mainly classified as a movement disorder, non-motor symptoms such as psychiatric complications are a regular phenomenon during disease progression 12. Although less common than depression 3, psychosis is well-known as a highly relevant psychiatric symptom in PD. Psychotic symptoms occur only in a minority (less than 10%) of untreated PD patients but are known to be associated with dopaminergic treatment 456. Hallucinations or illusions are observed in up to 40% of PD patients treated with dopaminergic drugs 789101112. Psychosis in PD patients is associated with excessive disability, worse quality of life, poor outcome, caregiver distress 12, cognitive impairment, and sleep disturbances 10. There are multiple factors which are considered to elevate the risk for the development of psychotic syndromes, i.e. higher age, duration of disease, microvascular changes, brain atrophy, polypharmacotherapy, cognitive impairment, depression, and sleep disturbances 781012131415. In addition, the application of anticholinergic drugs, other reasons for a cholinergic deficit or temporal Lewy bodies can contribute to the development of psychosis in PD patients. In patients with hypokinetic movement disorders distinct from PD, such as Dementia with Lewy bodies (DLB) or Progressive Supranunclear Palsy (PSP), psychosis is also a common phenomenon 161718, but this has not been addressed in this study.

Dopaminergic treatment seems to be a particular risk factor for psychoses in PD. Several reports have described an induction of psychosis by most of the common dopamine agonists (DA) 1920212223242526. Papapetropoulos et al. concluded that virtually all anti-Parkinsonian drugs are able to induce psychotic symptoms 27. The question whether certain dopaminergic drugs do have a higher potential to induce psychosis than others is discussed in a controversial manner. In a study of Aarsland et al., the authors found no correlation at all between anti-Parkinsonian drugs (levodopa and DA) and hallucinations 12. In contrast, a prospective comparison of patients treated with either levodopa or ropinirole demonstrated a higher incidence of hallucinations in the ropinirole group (17% vs. 6%) 2428. In the present retrospective study, possible correlations between dopaminergic treatment with different drugs and the risk for PD patients to develop psychosis were analyzed.

The occurrence of psychotic episodes in PD patients was shown to correlate with the presence of dementia and the numbers of concomitant diseases and of concomitant drugs, whereas MRI-based factors of cortical atrophy and white matter lesions could not be finally assessed due to limited data. With respect to dopaminergic treatment, pergolide was associated with a significantly increased risk for the development of psychotic symptoms, whereas levodopa was found to bear the lowest risk of all investigated dopaminergic drugs. In the literature, there is a large variability in the reported incidence of PD-associated psychosis among different authors, particularly depending on the method used 78910111231. Prospective studies mostly show higher incidence rates than retrospective investigations. In one retrospective study with outpatients, the incidence was as low as 3% 31. Some investigators found no clear association between psychosis in PD and dopaminergic treatment itself 32, but reported a correlation between psychosis in PD and polypharmacotherapy 33. In the present study, a positive correlation between the number of overall used drugs per patient and the risk to develop psychotic symptoms was observed.

Aarsland et al. described a correlation between higher age and the incidence of psychosis in PD patients 12. The peak age of onset of PD is 55 to 66 years 34, while the median age of our PD patient sample with psychotic episodes was 72 ± 9 years. Whether this indicates a correlation of psychotic symptoms with age itself or with the duration of the disease, as described previously by Fenelon et al. 9 cannot be distinguished as no reliable data concerning disease duration are available in all patients (see "Statistics")

Based upon the consideration that a main risk factor for the development of psychosis in PD is the application of dopaminergic medication itself 5912, it was our aim to investigate the potential of different dopaminergic drugs to induce psychotic episodes. In randomized double-blind trials investigating different DAs the incidence of hallucinations was 8.1% for priribedil, 2.8% for bromocriptine, 4% for rotigotine, 5% for ropinirole, 7%–9.3% for pramipexole, 4.8% cabergoline, 3.4 for pergolide, 0% – 4.4% for levodopa 353637383940. As in these studies often early PD stages (Hoehn and Yahr stage 1–2.5) without severe concomitant diseases are enrolled, these values represent not necessarily the situation in "real life".

This study was performed in the assumption that each included PD patient was treated with an individually optimized DA dose with respect to the efficacy on motor symptoms. By far the highest association for the induction of psychotic episodes was observed for pergolide. Since this association was much lower for cabergoline, i.e. also an ergoline DA, the high risk for psychosis in pergolide treatment cannot solely be explained as a group effect.

All investigated DAs behave as highly effective D₂ agonists 41 and, to a lower extent, D₃ agonists 42. Therefore actions at other receptors, as for example agonism at D₄ receptors may play a role. Remarkably many antipsychotic agents show marked antagonism at D₄ receptors 43. Levodopa – the drug with the lowest risk for psychosis in the present investigation – has only low affinity to D₄ receptors 43. Also cabergoline and pramipexole, the two DAs which showed the lowest adjusted odds ratios in this study (0.58 and 0.64, respectively) have a lower agonistic efficacy (E_max 49 and 42) at the D₄ receptor 41 than pergolide (E_max 56, adjusted odds ratio 2.8). Another noticeable point is the investigated DA's α₂ receptor affinity: Only pergolide shows a significant agonistic effect at α₂ receptors. In contrast, cabergoline behaves as a strong antagonist at α₂ receptors, whereas ropinirole and pramipexole are mostly inactive at α receptors 41. Further, both cabergoline and pergolide are 5-HT_2A-C receptor agonists 44 and especially 5-HT_2A receptors are suspect to be related to the development of hallucinations 4546. In the comparison of all hereby investigated DAs, pergolide shows the strongest agonistic effect at the 5-HT_2A receptor 44. This profile of pergolide as a strong agonist at 5-HT₂ and α₂ receptors resembles the pharmacological profile of amphetamine derivates 45. Therefore we conclude that the unique pharmacological profile of pergolide (strong D₂ and D₄ receptor agonism in combination with strong 5-HT_2a and α₂ receptor agonism) might contribute to the high association of pergolide with the induction of psychosis.

Our analysis confirmed that levodopa shows the lowest risk of all investigated drugs for the development of psychotic episodes in PD. In accordance with this result, Holroyd et al. found no association between dose and duration of levodopa treatment and the development of hallucinations at all 32. Thus, the common clinical practice to use levodopa in psychotic PD patients rather than DAs is plausible.

Because of the design as a monocentric, retrospective, cross sectional study, there are certainly intrinsic limitations to the results. We are aware that the influence of different dopaminergic drugs is difficult to interpret. The dosages are hard to compare due to different pharmacokinetic and pharmacodynamic characteristics of the different DA. There is also no generally accepted "levodopa equivalent dose". The levodopa equivalent doses of different dopaminergic drugs vary substantially since they result in general from the consensus of an expert committee and refer only to the effect on the motor symptoms. Nevertheless we included the calculated levodopa equivalence doses as a confounder to the logistic regression model to exclude significantly higher dosage in the psychosis group.

Being well aware of the intrinsic methodological limitations which this retrospective, cross sectional study has, we could nevertheless demonstrate by systematic data analysis in a well defined patient sample that PD-associated psychosis correlates both with epidemiological factors and with the choice of different dopaminergic drugs. These factors might be included in differential therapeutic considerations in order to minimize the risk of psychosis.

The authors declare that they have no competing interests.

DE, AU and MS performed the chart review and the data collection. DE performed the statistical analysis. DE, AU, JK and MS performed the discussion of the manuscript ED drafted the manuscript, JK and MS finalized the manuscript. All authors read and approved the manuscript.