We identified a double dissociation between reward learning and impulsivity in PD patients with and without ICBs. Relative to both PD−ICB patients and controls, PD+ICB patients exhibited greater delay discounting, suggesting decreased tolerance for delayed gratification, consistent with a recent finding (Voon et al, 2010
). By contrast, PD+ICB patients exhibited similar scores on the explicit adaptive salience measure derived from the SAT to controls, with both these groups scoring higher than PD−ICB patients. This finding is consistent with other previous studies demonstrating impaired stimulus-reward learning in PD patients without ICBs (Bodi et al, 2009
; Peterson et al, 2009
; Rutledge et al, 2009
; Swainson et al, 2000
). The difference between the groups seemed to be driven by learning pertaining to high-probability-reinforced stimuli, as the groups rated low-probability-reinforced stimuli equally likely to be associated with reward. These data are consistent with an explanation of ICBs in terms of elevated impulsivity, but contradict the hypothesis that ICBs result from overvaluation of CS+ (Evans et al, 2006
; Holden, 2001
; Isaias et al, 2008
; Tamminga and Nestler, 2006
An explanation for decreased explicit adaptive salience in PD patients without ICBs may relate to depleted dopaminergic release to reward in the ventral striatum compared with healthy controls and PD+ICB (Evans et al, 2006
; Steeves et al, 2009
). This finding replicates past research that has shown reward-learning deficits in PD patients without ICBs (Czernecki et al, 2002
; Swainson et al, 2000
). In PD, the phasic release of dopamine that is normally associated with the presentation of a conditioned stimulus may be reduced, impairing the ability to form associations between stimuli and rewards (Rescorla and Wagner, 1972
). Consistent with this suggestion, experimental studies in healthy volunteers have shown that administration of haloperidol (a dopamine antagonist) impairs the ability to learn to choose rewarding actions and decreases ventral striatal responses elicited by reward prediction errors (Pessiglione et al, 2006
; Pleger et al, 2009
). By contrast, patients with ICBs were able to learn the dissociation between high-probability and low-probability reward predicting stimuli, as well as controls. This suggests that elevated dopamine transmission in these patients may have ameliorated the reward-learning deficit. Therefore, although PD+ICB patients experience disruptive drug-induced behavioral symptoms, they can perform as well as controls at learning stimulus–reward associations, and importantly did not seem to overvalue stimuli associated with reward.
PD+ICB patients were significantly more impulsive than the other two groups, with greatly elevated k
values on the Kirby delay discounting questionnaire. This result is consistent with another recent study that used a different method to assess delay discounting over shorter timescales in PD patients with and without ICBs (Voon et al, 2010
). Similar findings have also been reported in several studies involving substance abusers, including cigarette smokers (Mitchell, 1999
; Odum et al, 2002
), alcoholics (Petry, 2001
), and heroin addicts (Kirby et al, 1999
; Madden et al, 1999
). Rates of delay discounting correlate with personality questionnaires of impulsivity (Kirby et al, 1999
; Petry, 2001
), suggesting that delay discounting taps some construct of impulsivity.
It is possible that high levels of impulsivity in these patients result from excessive dopaminergic transmission in the orbitalfrontal cortex (OFC: Cools, 2006
), to which the ventral striatum projects. Frontal lesions in experimental animals result in a preference for immediate small rewards over delayed larger rewards (Mobini et al, 2002
), as well as with other disinhibited behaviors (Roesch et al, 2007
). Studies using in vivo
microdialysis have shown increases in dopamine release in the OFC during the performance of delay-discounting tasks in rats (Winstanley et al, 2006
). Interestingly, a recent study reported that administration of pramipexole diminished OFC sensitivity during a decision-making task in PD patients, specifically reducing de-activations elicited by negative prediction errors (van Eimeren et al, 2009
). However, this study only included PD patients without ICBs, limiting comparison with this study.
Alternatively, the preference for sooner, smaller rewards in PD patients with ICBs could be mediated by altered dopaminergic transmission in the ventral striatum itself, as studies in both rats and humans suggest that this structure has a central role in encoding information relating to delays (Cardinal et al, 2001
; Pine et al, 2009
). Either way, the pattern of decreased tolerance to delay in tandem with intact stimulus-reinforcement learning suggests that ICBs are mediated by increased impulsivity and not overvaluation of rewards. Future studies using functional neuroimaging measures would help to elucidate the neurobiological mechanisms underpinning this pattern of results; specifically, it would be of great interest to use tasks that can dissociate neural responses associated with delay discounting from those associated with reward magnitude (see Pine et al, 2009
). As impulsivity is a multidimensional construct (Evenden, 1999
), it would also be of interest in future studies to examine whether PD patients with ICBs are more impulsive in other domains behavior, such as motor inhibitory control and risky decision making (Rossi et al, 2010
The usual treatment for ICBs is to reduce the dose of dopaminergic medication as there is some evidence to suggest that this is associated with the development of ICBs (Lee et al, 2010
). However, this is often unsatisfactory for the patient because it can worsen motor control. The finding that this group was specifically more impulsive has potentially important clinical implications as recent findings in other impulse-control disorders suggest that this can be treated with nondopaminergic medication. For example, impulsivity in these patients might be ameliorated by drugs with noradrenergic actions, such as atomoxetine, which have been shown to improve response inhibition (Chamberlain et al, 2009
) and effectively treat impulsive features in certain psychiatric disorders, such as ADHD (Chamberlain et al, 2007
Consistent with the notion that cognitive impairment and maladaptive behavior after DRT in PD result from an ‘overdosing' effect on dopaminergic transmission in the ventral striatum, PD patients exhibited significantly increased scores on the O-LIFE schizotypy scales, which tap personality traits relating to psychosis. Although psychotic reactions after DRT have been reported previously in some patients with PD, to our knowledge this is the first report of generally elevated schizotypy scores. This elevation was exaggerated in PD+ICB patients, who scored significantly higher than the PD−ICB group on the ‘Introvertive Anhedonia' and ‘Impulsive Non-conformity' subscales (although this effect was nonsignificant when age and IQ were included as covariates), with a trend toward greater scores on the ‘Unexplained Experiences' subscale. The latter of these assesses experiences relating to psychotic symptoms, such as hallucinations and delusions. These data suggest that low-level psychotic experiences may be a relatively common side effect of DRT, especially in PD+ICB patients, even in those who do not experience full-blown psychosis. This finding is also consistent with the dopamine hypothesis of schizophrenia and related theories, which propose a central role for dysregulated dopaminergic transmission in the instantiation and maintenance of both positive and negative symptoms of schizophrenia (Juckel et al, 2006
; Kapur, 2003
; Roiser et al, 2009
; Ziauddeen and Murray, 2010
). However, despite elevated schizotypy in PD+ICB patients, aberrant salience scores derived from the SAT did not differ between the groups, although schizotypy and aberrant salience were correlated across all participants, replicating our previous findings (Roiser et al, 2009
; Schmidt and Roiser, 2009
). The relationship between psychotic symptoms in PD, DRT, and aberrant salience warrants further investigation in future studies.
Both PD groups were impaired on the implicit adaptive salience measure derived from the SAT, indicating that they failed to speed responses on trials in which reward was very likely relative to when it was improbable. Implicit adaptive salience scores were similar in the PD−ICB and PD+ICB groups, despite the relatively normal explicit adaptive scores in the PD+ICB group. This latter finding is evidence against the explanation that impaired implicit adaptive salience in PD patients resulted from a learning deficit. Instead, a more likely explanation is that, at least in the PD+ICB group, patients were unable to use cue values to guide speeded responding. Importantly, studies in experimental animals have shown that dopamine transmission in the ventral striatum has a central role in the invigoration of responding after the presentation of CS+ (known as PIT: Berridge and Robinson, 1998
). Impairment in reward-elicited speeding in PD patients, in whom dopamine transmission is disrupted, may thus reflect dysfunction of a similar mechanism. However, it should be acknowledged that both PD groups were impaired on other indices on the SAT: neither group improved on explicit adaptive salience from block 1 to block 2, which could be considered evidence of a learning deficit; and both made more premature responses, although this finding may simply reflect a motor deficit.
PD patients were also impaired on the digit-span test of working memory independent of ICB status, consistent with previous findings (Brown and Marsden, 1988
; Cooper et al, 1991
; Dubois and Pillon, 1997
; Lewis et al, 2005
; Owen et al, 1992
). Experiments in animals have demonstrated a critical role for prefrontal cortex dopamine release in working memory, raising the possibility that this deficit is similarly related to disrupted dopamine transmission in PD (Floresco and Phillips, 2001
; Sawaguchi and Goldman-Rakic, 1994
; Williams and Goldman-Rakic, 1995
; Zahrt et al, 1997
). As we only tested participants while they were on medication, it is difficult to know whether these deficits are related to the reduced dopaminergic transmission associated with PD per se
, or to an overdosing effect of DRT on components of the dopamine system spared by PD (Cools, 2006
). Either way, our data underscore the difficulty of ameliorating cognitive deficits in PD with DRT, and emphasize the need for novel treatment strategies in this domain.
A limitation of our study is that PD+ICB patients were younger and of lower IQ than the other two groups, raising the possibility that the differences between PD patients with and without ICBs could be explained by nonspecific cognitive impairments. In addition, it is possible that other clinical concomitants of PD may have affected our results. However, we consider these explanations to be unlikely for two reasons. First, working memory, depression, anxiety, PD symptoms, duration of illness, and medication level were all well matched between the two PD groups. Second, and most importantly, statistically accounting for differences in age and IQ did not change our results, other than on one subscale of the O-LIFE schizotypy questionnaire.
A further limitation is that we did not recruit a sufficiently large sample of PD patients to allow meaningful statistical analysis of specific ICB subgroups. It is not yet known whether different kinds of ICBs may have different underlying neural and behavioral mechanisms. For example, some patterns of behavior included in the definition of ICBs may reflect compulsiveness (eg, punding) rather than impulsiveness (eg, DDS). Future studies should aim to recruit larger samples of PD patients with ICBs to explore this question further.
In summary, we found that PD patients without ICBs were impaired at learning stimulus–reward associations, replicating previous findings (Peterson et al, 2009
; Rutledge et al, 2009
; Swainson et al, 2000
). Strikingly, however, PD patients with ICBs were unimpaired at learning stimulus–reward associations compared with those without ICBs, which may relate to relatively preserved striatal dopamine transmission. Rather, increased impulsivity as demonstrated by elevated delay discounting, seems to be a prominent feature of PD+ICB patients. Impulsive symptoms identified by neuropsychological tasks, such as the stop signal task, have previously been related to impulsive behaviors in everyday life (Chamberlain and Sahakian, 2007
). It is important to note that our PD patients with and without ICBs were well matched for daily DRT used, and that DRT amounts did not correlate with delay discounting or aberrant salience. This suggests that a subgroup of patients are more sensitive to the behavioral effects of DRT and go on to develop ICBs, as opposed to a dose effect of DRT influencing behavior in PD. Therefore, further investigations of impulsivity may lead to better diagnostic classification systems for ICBs and novel treatments for these medication-induced disorders.