Impulse control behaviours (ICBs) are a spectrum of behaviours characterised by repetitive, reward-based actions and subjective loss of control. In their severe form, they are known as impulse control disorders (ICDs). In Parkinson’s disease (PD), ICBs can occur as adverse effects of dopamine replacement therapies (DRTs), particularly dopamine agonists (DAs),1 with potentially devastating and long-lasting psychosocial effects on patients and caregivers.2 There is a strong clinical need to understand the pathophysiology underlying PD-ICB and how to evaluate and manage PD-ICBs in clinical practice.
PD-ICBs are common, although the reported prevalence varies widely (14%–40%).1 3–6 This is likely due to differences in diagnostic criteria, assessment tools7 and possibly cultural differences between populations.3 6
The most common PD-ICBs described are as follows:
Pathological gambling (PG).S1
Compulsive shopping (CS).S2
Compulsive or binge eating (BE).S1
Dopamine dysregulation syndrome (DDS)/compulsive medication use.11
Other behaviours, sometimes considered subsets of those above, include hoarding12 and walkabout.13 ICBs more specific to the technological era, such as internet overuse and gaming,S3 are increasingly recognised. This is unsurprising, as a fundamental nature of these technologies is to encourage repetitive use leading to an increased recognition of excessive use as a problem in the wider population.14 The specific form of internet overuse varies, and, depending on the driver, may include aspects of hobbyism, punding or even gambling and hypersexuality. The incidence of these behaviours is likely to increase with time as technology is increasingly used by older generations.
A phenomenon that specialists may recognise from clinical practice is that PD-ICBs exists on a broad spectrum of severity, ranging from a change in premorbid traits that are still considered within the limits of normal behaviour to severe disorders of high frequency and intensity (see figure 1).
Conceptual illustration of the Parkinson’s disease with impulse control behaviour (PD-ICB) spectrum. PD-ICBs exist on a spectrum of severity determined by both the frequency of the impulsive activity and the intensity or magnitude of that activity, ie, the impact that it has on patients and caregivers. At the mild end of the spectrum, these behaviours may be considered ‘within normal limits’ and have little or no impact. At the severe end of the spectrum, behaviours typically meet the diagnostic criteria for PD-ICD. In the middle of the spectrum, however, behaviours may not meet these criteria and be considered ‘subsyndromal’, but still have a significant impact that may require monitoring or intervention. Adapted from Parkinson’s Impulse Control Scale rating scale and results from Baig and colleagues (2019).29
Diagnostic Statistical Manual (DSM-) and DSM-aligned diagnostic criteria exist for individual ICBs (see references listed above). While this specificity is useful in research, these criteria risk giving the erroneous impression that PD-ICBs exist as a binary entity—in which patients either meet or do not meet criteria for ‘caseness’. Conversely, screening questionnaires such as the Questionnaire for Impulse Control Disorders in Parkinson’s Disease (QUIP)13 and the Minnesota Impulsive Disorders Interview (MIDI)15 may capture most cases without differentiating the innocuous from the dangerous. This categorical approach fails to recognise the dimensional nature of the spectrum where ‘subsyndromal’ PD-ICDs may have an observable impact on patients without meeting strict diagnostic criteria. This has additionally been a limitation of many PD-ICB prevalence studies.1 7 Attempts to address this dimensional approach include the QUIP rating scale (QUIP-RS) and Parkinson’s Impulse Control Scale (PICS).16 17 Recognition of PD-ICB as a spectrum will help to risk-stratify patients, target interventions appropriately and assess response to such interventions in future observational and treatment studies.
The objectives of this review are to as follows:
Update clinicians on what is known about the pathophysiology of PD-ICB, presenting a comprehensive and interpretable biopsychosocial model that takes into account neurochemical, neural-network, psychological and epidemiological evidence to date.
Highlight how to assess and manage PD-ICBs using pharmacological and non-pharmacological approaches, underpinned through application of this model.
Terminology regarding PD-ICBs differs between studies and specialist centres. For example, the term ‘impulsive and compulsive behaviours’18 is also used. While harm avoidance is included in the psychological formulation of PD-ICB, there is little evidence to suggest the behaviours denote ‘compulsions’, which are defined as behaviours specifically targeted to relieve anxiety.S1 Furthermore, the strongest evidence to date favours ICBs representing a ‘devaluation of future rewards’ or immediate gratification (also known as myopia for the future).19 Therefore, in this review, the term ‘impulse control behaviour’ and the following terminologies are used:
Impulse control disorder (ICD/PD-ICD): An ICB that fits DSM or published DSM-aligned criteria, that is, ‘Syndromal’ ICD. The cornerstone of such criteria is that the ICB sufficiently impacts on social and/or occupational functioning.
Subsyndromal impulse control disorder: An ICB that falls short of the criteria for ICD. In the case of PD, this specifically refers to a change in behaviour; either an exacerbation of previous behaviours or their novel occurrence, thought related to the introduction of DRT.
ICB/PD-ICB: The full spectrum of ICB, as defined above, including not only syndromal ICDs but also subsyndromal ICDs. N.B. this includes the full range of disinhibitory psychopathologies listed above including hobbyism and punding.
Pathophysiology Animal models, imaging, neuropsychometric and observational clinical studies provide evidence on mechanisms underpinning PD-ICB. Evidence from the study of ICBs in the healthy population, particularly PG, have been extrapolated to and investigated in PD. While describing all the complex theories proposed to explain impulsivity would be beyond the scope of this review, we aim to examine some of the more comprehensive neurochemical, neural-network, psychological and clinical models of PD-ICB.
The role of dopamine While impulsivity is a complex and heterogeneous concept (see the Psychological modelling section), reward and punishment learning, mediated primarily by dopamine, are recognised as key components of impulsivity in PD-ICB,S4 providing a useful model to examine the role of dopamine in PD-ICB. Dopamine receptor subtypes have varying effects and patterns of distribution throughout the basal ganglia. D3 receptors are expressed preferentially in the ventral striatum (vs), linked to reward processing.S5 D1 and D2 receptors have been proposed, in a computational model, to mediate motor response inhibition, demonstrated by procedural learning tasks.20 Phasic dopaminergic bursts in the D1-mediated direct (excitatory, ‘Go’) pathway are released in response to positive stimuli. Dips or pauses in dopamine release in the D2-mediated indirect (inhibitory, ‘NoGo’) pathway occur in response to negative outcomes. This model has been consistently replicated clinically: medicated patients with PD demonstrate impaired negative-reinforcement learning and an enhanced response to positive reinforcement, with the opposite observed in individuals with unmedicated PD.21 22 S6 Modern DAs demonstrate relative selectivity for D2-type (D2, D3, D4) receptors over D1 types (D1, D5).S7 DAs, therefore, may overstimulate D3-mediated reward circuits and excessively disinhibit the D2-mediated indirect pathway. These effects may be less pronounced with levodopa therapy, which demonstrates a less selective, and more physiological receptor activation pattern. In animal models of PD-ICB, pramipexole enhances delay discounting, a marker of impulsive decision-making.S8 D2 and D3 agonists stimulate reward-seeking behaviour in rodents at intermediate doses.S9 However, in these experimental models, these effects also occur in control animals, not just PD models. Furthermore, in the latter study,S9 D1-specific agonists and higher doses of D2 and D3 agonists actually induced reward aversion, inconsistent with the clinical models described above. It is thus likely that preclinical models cannot yet recreate all the complex neurochemical processes underlying PD-ICB.S10 Nuclear imaging studies demonstrate lower levels of D2/D3 receptor binding23–25 and reduced dopamine transporter (DAT) availability26 in the vs of PD-ICB subjects. The latter has been shown in longitudinal analyses to predate the introduction of DRT.27 S11 These findings may be explained by greater levels of vs dopamine activity (either due to enhanced release and/or reduced uptake). Thus, D2/D3 receptors are ‘overdosed’ by the addition of DRT, increasing the potential for reward-driven behaviours.S12 Conversely, PD-ICB patients may have a greater burden of dopaminergic circuit denervation, potentially explaining why patients with more advanced disease are at greater risk of PD-ICB.28 29
The role of genetics in dopamine signalling Kraemmer and colleagues30 found the heritability of PD-ICB to be 57%, suggesting a genetic predisposition in some individuals. Although, only a screening questionnaire (QUIP) was used for diagnosis. Heterogeneity in neurotransmitter metabolism may explain why some individuals are at greater risk of PD-ICB. The findings of seven candidate gene association studies comparing PD patients with and without PD-ICB are shown in table 1. Polymorphisms of DRD1, DRD2, DRD3, SLC22A1 and DDC, all involved in dopamine signalling and metabolism, have been associated with increased risk of PD-ICB. A polymorphism of DRD4 was associated with changes in gambling task performance on levodopa.31 Genetic panels may be useful in predicting PD-ICB risk.30 However, results are variable and contradictory, likely due to differences in diagnostic criteria, ethnicity and candidate gene selection. Large-scale, multicentre, genome wide-analysis studies are warranted to investigate the role of genetic risk in PD-ICBs.
Other neurotransmitters Serotonin, norepinephrine, glutamate and opioids have been implicated in addictive behaviours such as PG and substance use disorder (SUD). For a detailed review, see previous work.32 Genes involved in neurotransmitter signalling (GRIN2B; glutamate, HTR2A; serotonin and OPRK1; opioids) have been implicated in PD-ICB risk (see table 1). Other than these gene association studies, however, there has been little study into the role of other neurotransmitters in PD-ICB. Modest and sometimes conflicting results for the use of selective serotonin reuptake inhibitors (SSRIs),S13 selective norepinephrine reuptake inhibitors (SNRIs),S14 opioid receptor antagonistsS15 and N-methyl-D-aspartate (NMDA-) glutamate receptor antagonists33 S16-S18 are discussed in more detail under the Management section.
Neural-network modelling Impulsivity and reward-based decision-making are mediated by a complex neural-network interconnected by mesocortical and mesolimbic circuits. Clinical imaging studies have attempted to demonstrate whether anatomical and/or functional changes in neural-networks predispose certain individuals to ICB within the general population and in PD. Reward circuits project from the ventral tegmental area (VTA) to cortical and subcortical areas.S19 Disruption of these circuits has been implicated in ICDs such as PG and SUD.32 The specific areas implicated vary according to imaging modalities used. Such studies have also been used to investigate a neural-network correlate for increased risk of PD-ICB. Voxel-based morphometry has identified patterns such as cortical thickening of the anterior cingulate cortex (ACC)S20 and cortical thinning of the orbitofrontal cortex (OFC),S21 associated with PD-ICBs. However, such findings are inconsistently replicated and this modality may not be sensitive to subtle or functional changes.34 Brain activity studies such as functional MRI (fMRI) may be more sensitive to functional changes in reward circuitry. These studies are small in size, few in number and vary in their methods and results. A recent meta-analysis35 of such studies found that the most consistently implicated brain activity changes between PD-ICB and PD-non-ICB individuals were the following:
Hyperactivity of the vs. vs dopamine release is triggered by unexpected reward stimuli, an important initial step in reward learning. An intact vs, with greater dopaminergic activity, may be more sensitive to the ‘overdose’ of DRT,S12 as mentioned in the Neurochemical modelling section.
Hyperactivity of the OFC. Dopaminergic activity of the OFC is thought to mediate reward processing and decision-making in goal-directed learning.S22 The OFC may be particularly vulnerable to the tonic stimulation of dopamine receptors by DAs.S23
Hypoactivity of the ACC. Reduced ACC activity may impair the perseveration and premeditation required to mitigate impulsive decision-making.S24
Most imaging studies are cross-sectional and unable to infer causality in the correlations identified. Longitudinal studies are required to identify whether changes in reward circuitry represent a predisposition towards impulsivity or a response to chronic exposure to DA or PD-ICB itself. Comparison of baseline volumetric MRI in drug-naïve patients did not differentiate between those who later developed PD-ICB (n=42) and PD controls in a case–control study nested within the longitudinal Parkinson Progressive Markers Initiative (PPMI) cohort.S25 However, another longitudinal study36 using resting-state fMRI found several differences in baseline connectivity that were later associated with greater risk of PD-ICB (increased salience network (OFC) connectivity, reduced central executive network (left supramarginal gyrus) and central executive network (left precuneus, right middle temporal gyrus) connectivity). Coupling between salience networks and central executive networks also correlated with greater PD-ICB severity. This study was small (n=30, 15 with PD-ICB) and relatively short in duration (36-month follow-up). Only resting-state fMRI was assessed, whereas fMRI during reward-based tasks may be more sensitive to reward-circuit changes. Nevertheless, this and the DAT studies discussed previously27 S11 suggest that at least some neural-network changes precede the introduction of DRT and onset of PD-ICB. DTI studies postulate a role for white-matter denervation in impulsivity.S26 S27 Larger, prospective functional imaging studies are needed to better investigate the role of heterogeneity in neural-networks and the clinical utility of imaging in predicting PD-ICB risk.
Psychological modelling Increased understanding of the neuropsychology of ICBs such as PG and SUD may shed light on the formulation of ICB, the role of dopamine and the predilection towards ICB among certain patients with PD. Here, we outline psychological models of (1) reward seeking and impulsivity and (2) harm avoidance.
Impulsivity Impulsivity can be divided into several, not mutually exclusive domains, according to the psychometric tests through which they measured.37 In general, these are conceptualised as:
Motor impulsivity (response inhibition): Acting without thinking. Most commonly measured by Go/NoGoS28 and Start-Stop Reaction tasks.S29
Decision-making impulsivity: Decision-making involves several cognitive domains including executive function, reward processing and punishment sensitivity. Impulsive individuals appear to make riskier decisions, disproportionately incentivised by positive reward and desensitised to negative consequences. Measured, in part, by gambling tasks.S30 S31
Choice impulsivity: Closely related to decision-making; impulsive individuals have difficulty in delaying gratification despite greater rewards in Delay-Discounting tasks.S32
Reflection impulsivity: Insufficient information gathering when making decisions. Measured by information sampling tasks.S33
PD-ICBs can be modelled using these domains, although questions still remain regarding their direct translation into pathological behaviours (see review by Dawson and colleagues18). While the sensitivity and specificity of such tests appear valid, their role as discriminators of severity remains unclear. While some inconsistencies may be explained by variations in methodology, there currently appear limitations to the real-world applicability (or ‘ecological validity’)S34 of such paradigms. Such challenges are replicated in other conditions where executive dysfunction may pose a clinical problem and require psychometric evaluation, for example, traumatic brain injury.S35 PD-ICB patients score higher than PD controls for motor impulsivity, non-planning (choice) impulsivity and attention (reflection impulsivity) on the Barratt Impulsiveness ScaleS36 and on the Delay-Discounting Task, specific to choice impulsivity.28 As described in the Neurochemical modelling section, untreated patients with PD demonstrate impaired reward learning and intact reversal learning, whereas dopaminergic treatment reverses this pattern.21 Drug-naïve PD patients also demonstrate reduced novelty-seeking behaviours compared with healthy controls.38 While this might suggest that patients with PD, in the absence of DRT, are less impulsive than the general population, there is little evidence to indicate a difference in PD-ICB prevalence of between drug-naïve PD patients and healthy controls.39 DAs have been shown to interfere with reward prediction error in patients with PD. Excessive dopaminergic stimulation will interpret rewards as ‘better than expected’, encouraging behaviour through incentive sensitisation.40 It is possible that psychological profiles differ between PD-ICBs. For instance, hypersexuality was associated with poorer performance than PG and BE on Stroop tests believed to represent inhibitory control.S37 PG and CS have been associated with greater novelty-seeking and CS with greater choice impulsivity than other PD-ICBs.28 Harm avoidance Impulsivity is generally seen as a novelty-seeking behaviour.S38 However, it is increasingly recognised that actions in PD-ICB are also driven by harm avoidance. While often considered dimensional opposites, it is likely the relationship is more nuanced than this, with both concepts sharing a loss of inhibition and ‘lack of control’, with an inability to learn from adverse outcomes.37 PD-ICB patients perform more poorly on a temporal binding task, indicating loss of ‘sense of agency’, also known as sense of control,S39 and demonstrate more harm avoidance than healthy controls.21 S40 It has been theorised that the goal of a PD-ICB is not only to seek a reward or ‘high’ but relieve dysphoria or a ‘low’ (eg, feelings of guilt, shame, loneliness, low mood). Various models have been proposed including the idea of (hedonic) homeostatic dysregulation,11 applicable to DDS, where evidence suggests the majority take additional medication in anticipatory fear of an ‘off period dysphoria’.19 Other psychological domains have been implicated in PD-ICB with mixed evidence. For example, some studies have found deficits in executive functions compared with PD controls,41 S37 while others have not.42 43