Article Text
PDF
Abstract
Background Impulse control disorders (ICDs) are known psychiatric conditions in Parkinson’s disease (PD), especially as a side effect of antiparkinsonian therapy. Screening for vulnerable patients and avoiding high-risk treatments can be an effective approach to reduce the ICD burden in patients with PD. Thus, our goal was to identify risk factors for ICDs in PD in the Swedish total population.
Methods Our longitudinal study was based on records of all patients with PD in the Swedish National Patient Registries and the Prescribed Drug Register (n=55 235). Patients with incident gambling disorder and other ICDs were compared with a control group on demographic factors, psychiatric comorbidity, antiparkinsonian dopaminergic treatment and therapies for advanced disease. Potential risk factors were analysed using logistic regressions and relative frequency comparisons (Fisher’s exact test).
Results Main predictors for incident gambling disorder were treatment with dopamine agonists (Frequency ratio 1.4, p=0.058), monoamine oxidase B (MAO-B) inhibitors (Frequency ratio 1.8, p=0.006) and a prescription for drugs used in addictive disorders (OR 5.85, 95% CI 2.00 to 17.10). Main predictors for other ICDs were dopamine agonist treatment (frequency ratio 1.6, p=0.003), anxiety disorders (OR 7.04, 95% CI 2.96 to 16.71) and substance use disorders other than alcohol (OR 5.66, 95% CI 1.75 to 18.23).
Conclusions Our results support possible risk factors for incident ICDs that had previously been identified, like dopamine agonist treatment and raise additional attention for risk factors like MAO-B inhibitor treatment and specific psychiatric comorbidities. These findings enable tailoring antiparkinsonian therapy to individual patient-specific risk profiles.
- IMPULSE CONTROL
- PARKINSON'S DISEASE
- EPIDEMIOLOGY
- NEUROPSYCHIATRY
- BEHAVIOURAL DISORDER
Data availability statement
Data may be obtained from a third party and are not publicly available. Data cannot be shared publicly because they contain confidential patient information, and the authors hold no permission to share them. Data are available from the Swedish National Board of Health and Welfare, SE-106 30 Stockholm, Sweden (contact via +46 (0)75 247 30 00 or socialstyrelsen@socialstyrelsen.se) for researchers who meet the criteria for access to confidential data.
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
WHAT IS ALREADY KNOWN ON THIS TOPIC
While some risk factors for impulse control disorders (ICDs) in Parkinson’s disease (PD) have been identified, we still lack full knowledge about vulnerable patients and high-risk treatments.
WHAT THIS STUDY ADDS
Monoamine oxidase B inhibitors seem less safe regarding ICD development than originally assumed, and treatment with ropinirole might be more disadvantageous than with pramipexole.
Gambling disorder and other ICDs present with distinct profiles of psychiatric comorbidity.
HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY
Recognising at-risk patients and avoiding or monitoring high-risk medications in their therapy closely is a crucial step to address the issue of ICDs in PD.
Introduction
Impulse control disorders (ICDs) are one type of non-motor symptoms in Parkinson’s disease (PD), with gambling disorder (GD), compulsive sexuality, compulsive buying and binge eating being the most common behavioural manifestations.1 In addition to motor symptoms, ICDs have been described to severely impact the quality of life in PD, potentially leading to serious consequences for patients and their families.1–3 Clinical features and risk factors for ICDs in PD have been studied in different populations, pointing out dopamine agonists, especially pramipexole and ropinirole, as high-risk treatments.4–8 Other antiparkinsonian therapies like levodopa or monoamine oxidase B (MAO-B) inhibitors have been found to correlate with ICDs to a lesser extent. Regarding the effect of advanced therapies including deep brain stimulation (DBS), levodopa-carbidopa intestinal gel (LCIG) and apomorphine pumps on ICDs, consensus has not yet been reached.9–11 Additionally, certain patient groups have been identified as vulnerable towards developing ICDs, such as patients of male sex, with a young PD onset or high novelty-seeking and risk-taking personality traits.
While some risk factors for ICDs in PD have been confirmed convincingly, others are still the subject of discussion. Thus, our goal was to provide a retrospective cohort study on the Swedish total population to support the evidence for known risk factors and gain new insights into other features possibly associated with ICDs. Working with ICD-10 diagnoses codes,12 we were able to select patients with GD and other, non-specified ICDs and chose to present these two groups separately. We hope to contribute to a clearer picture and provide clinical guidance concerning ICDs in PD, which ultimately could enable an effective screening and adapted treatment for vulnerable patients, preventing and alleviating ICD symptoms under antiparkinsonian therapy.
Materials and methods
Register data
Demographic variables, as well as records on operations, neurological and psychiatric diagnoses, were obtained from the Swedish National Patient Registries of inpatient and specialised outpatient care while information on dispensed drugs was based on the Swedish Prescribed Drug Register. Diagnoses followed the ICD-10 classification,12 and drug prescriptions were classified according to ATC codes of the WHO.13 DBS treatment was identified with the procedure code AAG20 and LCIG therapy with the procedure code GA014 in combination with a prescription for levodopa and carbidopa.
The nationwide data screening on all patients in Sweden was performed with the inclusion criteria of any antiparkinsonian dopaminergic drug prescription (N04B) and a PD diagnosis (G20) at any time during the study period 2005–2022, resulting in 55 235 studied individuals. The study population was divided into a control group without GD or another ICD (n=55 170), a GD group diagnosed after the first dopaminergic prescription (F63.0, n=29) and an ICD group, excluding GD, diagnosed after the first dopaminergic prescription (F63.1–9, n=27). One patient diagnosed both with GD and another ICD was included in both groups while patients diagnosed prior to their dopaminergic therapy were excluded from the analysis (n=10).
Statistical analysis
Statistical tests were performed in IBM SPSS Statistics (V.28.0.0.0), graphs were created in GraphPad Prism (V.9.3.1). All diagnoses and drug prescriptions were treated as binary variables, screening for any occurrence during the time period of interest for each group and analysis. For the control group, the full study period of maximal 17 years (2005–2022) was included for all variables obtained. In the GD and the other ICD group, the analysed time period was split into before and after the diagnosis when analysing parameters of psychiatric comorbidity. For assessing the prescription frequencies of dopaminergic drugs, the time period included for GD and other ICD patients was 6 months before the diagnosis, and frequency ratios GD group/control or other ICD group/control were calculated for comparability between different drugs.
Separate logistic regression analyses were conducted for the outcome GD and other ICDs, with consideration given to the covariates: (1) sex and age at first dopaminergic prescription, (2) diagnoses indicating psychiatric comorbidity before any ICD diagnosis and (3) prescriptions indicating psychiatric comorbidity before any ICD diagnosis. For comparing if the frequencies of psychiatric comorbidity diagnoses or dopaminergic drug prescriptions in the GD or other ICD group were larger than in the control group, a one-sided Fisher’s exact test was used. The frequencies of levodopa prescriptions before an ICD diagnosis with or without cotreatment of dopamine agonists or MAO-B inhibitors were compared using a two-tailed binomial test.
Results
Demographic factors
In general, more of the patients with PD were men, constituting 59.3% of the control group (32 698 of 55 170, figure 1A). Additionally, men were over-represented in the GD and the other ICD group, with 82.8% (24 of 29) and 81.5% (22 of 27), respectively. In a logistic regression together with the age at first dopaminergic prescription as a covariate, male sex was a predictor for GD (OR 3.07, 95% CI 1.16 to 8.15, figure 2) and other ICDs (OR 2.71, 95% CI 1.02 to 7.19). The mean age at the first dopaminergic prescription in the control group was 72.2±9.7 years (figure 1B). Both GD (49.9±7.9 years) and other ICD patients (55.0±14.6 years) received their first dopaminergic medication on average younger than other patients with PD. Younger age at the first dopaminergic prescription was a predictor for both GD (OR 1.14, 95% CI 1.12 to 1.17) and other ICDs (OR 1.12, 95% CI 1.09 to 1.15) in a logistic regression with sex as a covariate (figure 2).
Demographic parameters in control group, GD and other ICD patients. (A) Sex distribution. (B) Age at first dopaminergic prescription, arithmetic mean±SD. GD, gambling disorder; ICD, impulse control disorder.
Forest plot for developing GD or another ICD related to demographic and psychiatric comorbidity parameters (ORs with 95% CIs). GD, gambling disorder; ICD, impulse control disorder.
Psychiatric comorbidity
Substance use, mood and anxiety disorders
For 16 out of 29 GD patients (55.2%), and for 14 out of 27 other ICD patients (51.9%), the ICD diagnosis was their first psychiatric diagnosis compared with substance use, mood and anxiety disorders. Nevertheless, those disorders were overall more common in patients with GD or another ICD already before their diagnosis (figure 3A). GD patients had an especially high frequency of alcohol use disorders (10.3% vs 1.6%, p=0.011), mood disorders (27.6% vs 12.3%, p=0.021) and anxiety disorders (24.1% vs 9.2%, p=0.014) in comparison with the control group. Other ICD patients, on the other hand, showed the largest deviation in frequency from the control group in other substance use disorders than alcohol (14.8% vs 1.6%, p<0.001), mood disorders (22.2% vs 12.3%, p=0.041) and anxiety disorders (44.4% vs 9.2%, p<0.001). When combining all four comorbidity diagnoses as covariates in one logistic regression, having an alcohol use disorder before a GD diagnosis was found to be a predictor for GD (OR 4.19, 95% CI 1.12 to 15.59, figure 2). Regarding psychiatric diagnoses before other ICDs, especially anxiety disorders (OR 7.04, 95% CI 2.96 to 16.71), but also other substance use disorders than alcohol (OR 5.66, 95% CI 1.75 to 18.23) were associated with an increased risk.
Relative frequencies of psychiatric comorbidity parameters, (A) Common psychiatric diagnoses. (B) Psychotropic drug prescriptions. *p (Fisher’s exact) <0.05 for comparing prescriptions before a diagnosis to controls. GD, gambling disorder; ICD, impulse control disorder.
Psychotropic medication
Investigating prescriptions for anxiolytic and antidepressant drugs before an ICD diagnosis disclosed a similar, though weaker, trend compared with common psychiatric diagnoses, with mildly, non-significantly elevated frequencies in relation to the control group (figure 3B). Antipsychotics had a trend to be more frequently prescribed before an ICD diagnosis (33.3% vs 25.2 %, p=0.222), but less before a GD diagnosis (13.8% vs 25.2 %, p=0.110). However, drugs used in addictive disorders, like nicotine, naltrexone or buprenorphine, were more frequent before a GD diagnosis than in the control group (13.8% vs 2.3 %, p=0.004), but not prescribed in any other ICD patient before their diagnosis. Psychostimulants were slightly more common before a GD diagnosis (3.4% vs 1.4 %, p=0.340) and not prescribed before any other ICD diagnosis. A logistic regression including all five drug classes indicated a prescription for drugs used in addictive disorders as a predictor for GD (OR 5.85, 95% CI 2.00 to 17.10, figure 2) and did not identify any prescriptions as predictors for another ICD.
Increased mental health disorder treatment following ICD diagnoses
In addition to certain diagnoses and prescriptions constituting risk factors for ICDs, patients presented with further increased psychiatric comorbidity after their GD or other ICD diagnosis (figure 3). Notably, a GD diagnosis was followed overly often by a mood or anxiety disorder (31.0% and 17.3% of newly diagnosed patients, respectively). Furthermore, the frequency of first antipsychotic prescriptions was extraordinarily high in both GD and other ICD patients after their diagnosis (44.8% and 37.1 %, respectively).
Risk factors in antiparkinsonian treatment
Type of dopaminergic drug
In the 6 months before a GD diagnosis, MAO-B inhibitors (48.3% vs 26.4 %, frequency ratio=1.8, p=0.006) and dopamine agonists (65.5% vs 47.6 %, frequency ratio=1.4, p=0.058) were prescribed excessively often compared with the control group (figure 4A). Additionally, dopamine agonist prescriptions were increased in the 6 months before any other ICD diagnosis (77.8% vs 47.6 %, frequency ratio=1.6, p=0.003). No other dopaminergic drug classes showed substantially increased prescriptions associated with any ICD. Levodopa was prescribed slightly less before a GD or other ICD diagnosis while amantadine prescriptions were rare in general and lower before ICD diagnoses. Furthermore, other dopaminergic drugs were prescribed with similar frequencies before a GD or other ICD diagnosis and in the control group.
Relative frequencies of antiparkinsonian dopaminergic prescriptions, labelled with frequency ratios in comparison to control, (A) Dopaminergic drug classes (n=55 225). (B) Specific dopamine agonists (n=26 325). **p (Fisher’s exact) <0.05. GD, gambling disorder; ICD, impulse control disorder.
Among patients receiving dopamine agonists in the 6 months before their GD diagnosis, 89.5% were additionally treated with levodopa, compared with 89.7% of all GD patients in the 6 months before their diagnosis (p>0.999). GD patients receiving MAO-B inhibitors before their diagnosis were about as likely to be treated with levodopa in this period as the whole GD group (92.9% vs 89.7 %, p>0.999). Similarly, polytherapy with levodopa and dopamine agonists (85.7% vs 88.9 %, p=0.502) or levodopa and MAO-B inhibitors (100.0% vs 88.9 %, p>0.999) was about as common as observed in other ICD patients.
Specific dopamine agonists
Frequencies of specific dopamine agonist prescriptions were analysed within all patients who had received any dopamine agonist (n=26 325, figure 4B). None of the five dopamine agonists was prescribed with a significant increase before a GD or another ICD diagnosis, and prescription frequencies for ropinirole were similar in all three groups. For pramipexole, cabergoline and rotigotine, the frequency both before GD and any other ICD was about half the frequency of the control group. Apomorphine was not prescribed in the 6 months before a GD diagnosis but was prescribed more often before another ICD diagnosis than in the control group (12.0% vs 6.2 %, frequency ratio=1.9, p=0.183).
Advanced therapies
Considering the whole study period, DBS was found to be much more common in GD patients (7 of 29, 24.1%), and in other ICD patients (2 of 27, 11.1%) compared with 1.5% in the control group. However, only in about half of the cases the treatment was initiated before the ICD diagnosis. Given the large time difference between DBS treatment and ICD diagnosis (2.5, 2.5, 3.5 or 5.6 years prior), no apparent temporal relationship could be concluded from the case studies. Similarly to DBS, but to a smaller extent, LCIG therapy was more frequent in the complete observation period in GD patients (1 of 29, 3.4%) and other ICD patients (3 of 27, 3.7%) compared with 0.1% in the control group. In all four cases, an LCIG pump was inserted right after or just before the ICD diagnosis. The register data did not allow for the analysis of patients with apomorphine pump therapy.
Discussion
Main risk factors for clinical guidance
Male sex and younger age at PD onset
The proportion of men differed largely between around 60% men in the control group and over 80% in the GD and other ICD group. Male sex is a known risk factor for ICDs in the general population and in PD populations,4 5 14–19 with the exception of compulsive buying and binge eating.5 8 16 Furthermore, GD and other ICD patients on average were around 20 years younger when they received their first antiparkinsonian medication compared with other patients with PD in our study, presumably equivalent to the age at PD onset. Even this finding strengthens the previous consensus about early PD onset as a risk factor for ICDs.4 5 8 16 19 20 Nevertheless, it remains unclear if young age at onset is itself the risk factor or if patients are in need of longer and higher-dose dopaminergic treatment when diagnosed young, with the therapy itself leading to a decreased impulse control.
Compromised mental health
In general, the GD and other ICD patients in our study had worse mental health than the control group already before their GD or other ICD diagnosis and deteriorating additionally afterwards. Thus, our results point out a compromised mental health state both as a risk factor and a consequential comorbidity of GD and other ICDs, in line with many previous studies.4 21–24
On a more detailed level, especially alcohol use disorder and drugs used in addictive disorders were more common before a GD diagnosis. In other ICD patients, anxiety disorders were a strong risk factor in addition to substance use other than alcohol. Interestingly, the other ICD group presented with a different profile of psychiatric comorbidities than GD patients, suggesting that they are two distinct patient types. This could support the reclassification of GD as a behavioural addiction rather than an ICD in the latest diagnostic manuals.25 26
Regarding psychiatric comorbidity following a GD diagnosis, mood disorders doubled in comparison to before the diagnosis. Antipsychotic prescriptions were four times as high after a GD, and double as high after another ICD diagnosis, perhaps in an attempt to treat the ICD, or due to an increase in psychotic symptoms related to dopaminergic therapy.27 28 This highlights the severe consequences of GD and other ICDs for the general mental health of affected patients with PD and underlines the necessity of improving treatment options for ICDs, given that antipsychotics have no evident effect on ICDs.
About half of the patients who developed GD or another ICD had a prior psychiatric diagnosis, making it possible to screen for a compromised mental health as a risk factor in PD healthcare. For the other half, on the other hand, GD or another ICD was their first psychiatric diagnosis, pointing out the challenges in identifying vulnerable patients.
Dopamine agonist and MAO-B inhibitor treatment
In line with the majority of previous studies, we identified dopamine agonist treatment to be a strong risk factor for ICDs in our study population.6 7 16 29 30 Notably, we also find an increased frequency of MAO-B inhibitor therapy during the six months before a GD diagnosis, which has rarely been reported before.29 31–33 While dopamine agonists are widely recognised as a high-risk treatment, it is important to highlight that even other antiparkinsonian dopaminergics can increase the ICD risk by altering dopaminergic signalling more indirectly. The lower rates of levodopa and amantadine prescriptions in the six months before any ICD diagnosis can possibly be explained by the much shorter included time period compared with up to 17 years in the control group. Even if the study design did not allow conclusions about the effect of polytherapy with different dopaminergic drugs, cotreatment with levodopa and dopamine agonists or MAO-B inhibitors seemed not excessively common before an ICD diagnosis.
Among patients with a dopamine agonist prescription, ropinirole was about as common in the six months before any ICD diagnosis as in the whole study period for controls, often including 17 years of medication records. In comparison, pramipexole, cabergoline and rotigotine were only about half as frequently prescribed before a GD or other ICD diagnosis in relation to the control group. Even if no absolute increase in prescription frequency could be detected for ropinirole, it is more strongly associated with GD and other ICDs in our study than other dopamine agonists. Especially the discrepancy to pramipexole is noteworthy, since both ropinirole and pramipexole are generally acknowledged as the dopamine agonists with the highest risk for ICDs while rotigotine is attributed a lower risk in comparison.7 16 34
Regarding advanced therapies, DBS and LCIG infusion have been implemented remarkably often in GD and other ICD patients. Nevertheless, no temporal relationship between DBS and ICD diagnoses could be detected and a possible confounder might be the young age at PD onset in GD and other ICD patients. LCIG pumps had been inserted very close to the date of the ICD diagnosis, possibly as a reaction to ICD symptoms with the goal of reducing or discontinuing dopamine agonist treatment. Based on our data, we find a correlation between advanced therapies and a GD or other ICD diagnosis in relative frequencies. However, considering a temporal relationship, we do not identify them as risk factors, contributing to a controversial status quo in the literature to date.10 35–37
Strengths and limitations
The main strength of our data material is the full coverage of all diagnosed patients with PD in Sweden who received treatment at any point between 2005 and 2022, making a retrospective, longitudinal design possible. Thus, we hope to provide a complete and representative analysis of the Swedish population. Another advantage are the dates recorded for every event, making it possible to put diagnoses and treatments into a temporal context to each other rather than only studying associations. Additionally, we expect the national register data to be of high reliability regarding diagnosis codes and dispensed medication.
Considering the negative aspects of our register material, a major limitation is our selection of GD and other ICD patients purely based on their diagnosis code. Using the Swedish version of ICD-10,12 compulsive sexuality, compulsive buying and binge eating could not be identified. Working with ICD-11 in the registers should partially address this issue in the future.25 Additionally, far from every patient with PD who develops symptoms related to a decreased impulse control is given a specific diagnosis. Therefore, we were only able to detect severe ICD cases, where the patient most likely came in contact with a specialised psychiatrist. On the one hand, this leads to a strong underestimation of the incidence in our population of roughly 1 in 1000 during 17 years compared with incidence and prevalence rates in other studies.16 23 30 33 34 38 On the other hand, a selection bias for patients with other mental disorders who have already established a contact with psychiatric healthcare is likely, possibly affecting our results on psychiatric comorbidity.
Another limitation is the different time periods analysed in the control group (full study period), and the ICD groups (any time before the diagnosis or six months before the diagnosis, respectively), possibly leading to an underestimation of events detected in the ICD groups. Additionally, an ICD diagnosis may be given long after symptom onset, thus compromising the comparability of the time points between events. Lastly, we were unable to include treatment details like dosage or mode of drug administration in our analysis and could, therefore, only give a broad overview of the dopaminergic treatments received.
Conclusion
In summary, our study supports the general consensus about male sex, younger onset age of PD, a compromised mental health state and the treatment with dopamine agonists being risk factors for developing ICDs. Additionally, our findings point out the risk of developing GD under MAO-B inhibitor therapy and a possibility of ropinirole being more disadvantageous than pramipexole. Regarding the psychiatric comorbidity, GD and other ICD patients are characterised by a distinct profile each, and the average worsening in mental health after an ICD diagnosis was remarkable. Taken together, our findings highlight an individual risk assessment for each patient with PD regarding ICDs as a necessary step to decrease the ICD burden in PD healthcare. If possible, high-risk medications like dopamine agonists, especially ropinirole, and perhaps MAO-B inhibitors, could be avoided, or the patient at risk could be monitored closely and offered help in coping with an ICD immediately.
Data availability statement
Data may be obtained from a third party and are not publicly available. Data cannot be shared publicly because they contain confidential patient information, and the authors hold no permission to share them. Data are available from the Swedish National Board of Health and Welfare, SE-106 30 Stockholm, Sweden (contact via +46 (0)75 247 30 00 or socialstyrelsen@socialstyrelsen.se) for researchers who meet the criteria for access to confidential data.
Ethics statements
Patient consent for publication
Ethics approval
The study was approved by the Swedish Ethical Review Authority (file number 2021-05752-01) and data were provided by the Swedish Board of Health and Welfare.
Acknowledgments
The preliminary version of the results discussed in this paper was presented orally and as a scientific poster in context of the Young Scientist Symposium at the 61st Annual Meeting of the Scandinavian College of Neuropsychopharmacology (SCNP) in 2023 under the title 'Risk factors for developing gambling disorder and impulse-control disorders under dopaminergic medication'. The book of abstracts will be published in Acta Neuropsychiatrica by Cambridge University Press and is available under the following link: https://scnp.org/fileadmin/SCNP/SCNP2023/Book_of_abstract_2023_.pdf.
References
Footnotes
Contributors AH, PO, DW and MW were responsible for concept and design of the study. AH obtained the data and is guarantor for the article. MW and GCW performed the analysis. MW wrote the manuscript and all authors revised and approved it.
Funding AH has a position at Lund University that is sponsored by the state-owned gambling operator AB Svenska Spel and has funding acquired from the research council of Svenska Spel. MW is part of AH’s research group and has obtained funding from the research council of Svenska Spel (grant number FO2022-0008).
Disclaimer No specific funding has been obtained for the present study and no other organisation than the authors themselves were involved in the idea or content of the present study.
Competing interests DW, PO and GCW declare no conflicts of interest.
Provenance and peer review Not commissioned; externally peer reviewed.