Do Parkinson's Disease Patients Have Deficits in Sequential Sampling Tasks?

Abstract

The aim of this study was to assess the neuropsychological behavior of Parkinson's disease (PD) patients with addictive behaviors. Characteristically, these patients have younger onset of PD, higher novelty‐seeking personality traits, jump to conclusions, and often make irrational choices. We assessed whether PD patients with and without addictive behaviors have deficits in a sequential sampling task, often called the secretary problem. In this task, participants needed to pick the best out of multiple offers. Critically, once participants rejected a deal, this option became unavailable. Thus, decisions needed to be balanced not to stop too soon or sample for too long and miss the best deal. We tested 13 PD patients with and 13 patients without addictive behaviors. Results were compared to healthy volunteers. We found that all patients declined fewer options before committing to a deal. There was, however, no difference between the two patient groups. Furthermore, there was no difference in overall choice rank between patients and controls. These results suggest that, compared to controls, PD patients gather less evidence before committing to an offer, but have no deficits in recognizing the best deal out of many options, regardless of whether or not they have addictive behaviors.

When should one stop sampling for more information and take the current option? If one keeps on exploring more options, chances are that one might miss the best deal. This is known as the threshold problem1 or sometimes called the secretary problem or marriage problem. In the secretary problem, a manager wants to employ a secretary and interviews each candidate in a one‐to‐one order. The manager can then hire a person immediately or reject the secretary. Once rejected, however, the manager cannot go back and hire the previously declined person. In sequential sampling tasks, participants therefore continue to gather information until they reach a threshold where the subject is ready to commit to an option. Thus, the optimal strategy is to build up enough evidence by rejecting a certain number of candidates (generally recommended the first 37%) and then accept the next candidate that is better than the rest.2

A functional MRI study in healthy volunteers showed that stopping and committing to an offer engaged the frontal‐parietal cortex, the anterior cingulate, anterior insula, and the ventral striatum,3 all brain areas that are thought to play a critical role in addictive behaviors in Parkinson's disease (PD).4, 5, 6

Impulsive or compulsive behaviours (ICBs) are a common side effect of dopaminergic therapy, affecting at least 14% of treated PD patients.7 Thus far, the most sensitive neuropsychological tasks to distinguish PD patients with ICBs from PD patients without ICBs involve temporal discounting,8 novelty seeking,9, 10 jumping to conclusions,11 and working memory (WM) function.12

Because PD patients with ICBs have significant impairments in classical information sampling tasks,11 we hypothesized that these patients would also perform worse in the sequential sequence task and would therefore end up with poorer choice outcomes than PD patients without ICBs. Furthermore, we predicted that both PD groups would take an offer earlier than healthy volunteers.

Methods

All participants provided written informed consent according to the declaration of Helsinki, and the study was approved by the University College London Hospitals Trust.

Patients

All patients were recruited from the National Hospital for Neurology and Neurosurgery (London, UK), fulfilled the Queen Square Brain Bank criteria for the diagnosis of PD,13 and were taking levodopa (l‐dopa). We recruited 13 PD patients without ICBs (PD−ICBs) and 13 PD patients with ICBs (PD+ICBs). Results were compared to 23 healthy volunteers. All PD patients with ICBs were assessed using a semistructured interview, using accepted diagnostic criteria for pathological gambling,14 compulsive shopping,15 compulsive sexual behavior,16 and punding.17 All patients were tested in their “ON medication” state, usually mid‐mornings when their motor symptoms were best controlled, to minimize “OFF” dysphoria and anxiety.18 Participants who scored under 26 of 30 points on the Mini–Mental State Examination (MMSE)19 were excluded. All ICBs were directly triggered by dopamine agonist (DA) therapy. Seven PD+ICB patients were already weaned off the DA and were taking l‐dopa alone, whereas 9 PD−ICB patients were taking l‐dopa in combination with a DA. l‐dopa equivalent units (LEUs) were calculated as previously described.17

Sequential Sampling Task

The task was performed on a laptop computer in a quiet environment to minimize distractions. We used a version of this task that has been published recently.3 Participants were asked to select the best deal out of multiple options and perform 14 trials of the sequential sampling task. The options contained several different categories, such as buying a digital camera, a diamond ring, a used car, a house, a used truck, a printer, a used motorcycle, or renting an apartment. Participants then had to pick the best value; for example, they should select a truck with the lowest mileage. These options were presented one at a time, and participants had to decide whether they accept immediately or reject the current deal. Once participants decided to reject an offer, this option became unavailable for them. Because participants could not go back and choose a previous rejected deal which, in hindsight, turned out to be the best, they had to balance the risk of stopping too soon against that of continuing for too long and finding that the best deal was declined.20 Thus, a constant trade‐off between sampling more information and optimal stopping is required.

Declined options were listed on the right side of the screen to reduce working memory load, because previous studies have shown that all PD patients, but particularly those with ICBs, have poorer WM.12 The length of total options varied within the trials between 8 and 16 offers. In order to make sure that all subjects understood the rules, a practice trial and all subsequent trials were performed under the supervision of the researcher. Once participants committed to an option, feedback was given immediately; for example, for the best choice, “Rank 1” was displayed.

Statistical Analyses

Data analyses were performed using SPSS 21 (SPSS, Inc., Chicago, IL). For the demographic variables, age, gender, age of disease onset, Unified Parkinson's Disease Rating Scale (UPDRS) score, and LEU dose were separately used as dependent variables. We used analysis of variance, t test, or chi‐square (χ²) tests, where appropriate.

Demographic and Clinical Features

Demographic characteristics are shown in Table 1. There was a significant difference in age between the groups (P = 0.049), and post‐hoc analysis showed a trend that the PD−ICB group was older than the control group (P = 0.062). There were no other differences. As a consequence, age was included as a covariate in all analysis. As expected, PD+ICB patients had a significantly younger onset of PD symptoms, which is in line with previous studies.7 There was no difference in LEU dose, UPDRS part III “ON” motor scores, or disease duration.

Table 1.

Demographic characteristics

	Controls	PD−ICB	PD+ICB	t Statistic χ2 or F‐Statistic	P Value
Participants, no.	23	13	13
Gender, male	19	11	12	χ2 = 0.4	0.8
Age, years	59.1 ± 8.5	65.6 ± 6.0	59.2 ± 8.2	F = 3.2	0.049*
Age PD of diagnosis		53 ± 6.5	47.1 ± 6.5	t = 2.3	0.033*
Disease duration, years		12.3 ± 4.6	11.3 ± 4.9	t = 0.5	0.6
LEU dose, mg/day		883 ± 360	1016 ± 354	t = 0.9	0.3
Patients on DA (%)		9 (69.2)	7 (53.8)	χ2 = 0.6	0.4
UPDRS on motor scores		12.6 ± 3.6	14.6 ± 4.0	t = 1.2	0.24

LEUs were calculated as l‐dopa dose + pramipexole (mg) × 100, ropinirole (mg) × 20, amantadine × 1, rasagiline × 100, rotigotine (mg) × 30, apomorphine × 8, and l‐dopa × 1/3 if on entacapone. Significant group differences are labeled with an asterisk (*).

Sequential Sampling Task

We performed analyses using a generalized linear model (SPSS) because the dependent variables were counts and not continuous values. We examined the number of draws each participant made and found a significant group difference (Wald's χ² = 28.6; P < 0.001). Furthermore, there was a significant effect of rank (Wald's χ² = 83.4; P < 0.001), but no significant interaction between group and rank (Wald's χ² = 33.0; P = 0.3). Pair‐wise comparison (Bonferroni corrected) showed that both patient groups declined significantly less options than controls (P < 0.001), but there was no difference between the two patient groups (P = 1.0; Fig. 1). There was also no correlation between number of draws and choice rank (r = 0.57; P = 0.15).

Figure 1.

Overall drawing behavior. Box plot showing the median (horizontal line) within a box containing the central 50% of the observations (i.e., the upper and lower limits of the box are the 75th and the 25th percentiles) and extremes of the whiskers containing the central 95% of the ordered observations. Controls, PD–ICB, and PD+ICB. Outliners are shown as circles. Pair‐wise comparisons were Bonferroni corrected.

Discussion

We found that both PD groups performed similarly and declined less deals than healthy volunteers. These results are broadly consistent with our previous studies, showing that both PD groups gather less information than controls.11, 21

There are, however, important differences between the previously used information sampling tasks and the secretary task. In classical information sampling tasks, gathering more information increases the likelihood of correct responses, whereas in this task, continuing to sample information is in anticipation of obtaining a better future deal, rather than guessing a correct outcome. Furthermore, participants have to actively consider, after each offer, whether sampling is ideal given the amount of remaining choices, which is in contrast to classical information sampling tasks, such as the beads task, where sampling more information is beneficial.11 Thus, the same value of a current deal may not result in the same behavioral response, depending on the amount of remaining options.3 Furthermore, this task specifically assesses the trade‐off between declining choices, but, at the same time, not oversampling and rejecting the best option. Therefore, the sequential sampling task is often referred to as the “optimal stopping” task3 to highlight this dilemma.

In contrast to our hypothesis, we did not find any differences between PD patients with and without ICBs. Thus, sequential sampling tasks do not distinguish PD patients with ICBs from those without. Several previous studies have also failed to reliably distinguish PD patients with ICBs from those without ICBs. These include tasks assessing risk‐taking behavior,12 response inhibition measured by the Stroop Test,22, 23 and motor inhibition.24, 25

It is possible that a subgroup of PD−ICB patients who are treated with a DA will develop addictive behaviors in the future. Therefore, comparing results with PD−ICB patients who were never treated with a DA may have shown significant group differences, as has been reported recently.21 However, advanced PD patients are usually treated with l‐dopa in combination with a DA. Therefore, the study was designed to assess whether the secretary task could be a useful screening tool in clinical practice. Another explanation for the lack of group difference may be that this task also assesses an element of novelty seeking (exploring further and potentially better options), which is known to be a characteristic personality trait of PD patients with ICBs.9, 26 Thus, PD+ICB patients may have stopped sampling later than they would have done in other selection tasks11 in anticipation of a novel, better deal.

Interestingly, both PD groups did not choose overall worse options than controls, regardless of whether or not they were diagnosed with an ICB. This implies that similar to feedback learning, which has been found to be intact in PD patients with ICBs,12, 27 PD patients with and without ICBs are not impaired in choosing the best deal out of multiple options.

Only nondemented participants who understood the task were included in this study. However, the MMSE is not specific to assess early cognitive decline. It is therefore possible that the lack of other, more‐suitable screening methods, such as the Parkinson neuropsychometric dementia assessment, may have influenced task results.28

In summary, we found that both PD groups stopped significantly earlier to gather information than healthy volunteers without making overall worse choices. However, there was no difference between the two PD groups, which suggests that sequential sampling tasks cannot distinguish PD patients with ICBs from those without.

Author Roles

(1) Research Project: A. Conception, B. Organization, C. Execution; (2) Statistical Analysis: A. Design, B. Execution, C. Review and Critique; (3) Manuscript: A. Writing of the First Draft, B. Review and Critique.

A.D.: 1A, 1B, 1C, 2A, 2B, 2C, 3A

J.M.: 1C, 2C, 3B

A.T.: 1A, 2B, 3B

G.C.: 1C, 2C, 3B

T.T.W: 1A, 2A, 3B

A.L.: 1A, 2A, 3B

S.S.S.: 1A, 1B, 2A, 3B

B.B.A.: 1A, 1B, 2A, 3B

Disclosures

Funding Sources and Conflicts of Interest: The authors report no sources of funding and no conflicts of interest.

Financial Disclosures for previous 12 months: Atbin Djamshidian received honoraria from UCB. Thomas T. Warner received grants from NHS Innovations, Brain Research Trust, and CHDI. Andrew Lees reports the following: consultancies with Genus; grants from the PSP Association and the Reta Lila Weston Howard Foundation; and honoraria from Roche, Novartis, Boehringer Ingelheim, Lundbeck, GE Healthcare, Servier, Teva, Ipsen, GlaxoSmithKline, Solvay, Allergan, AbbVie, and Orion. Sean S. O'Sullivan received honoraria from Teva, Lundbeck Pharmaceuticals, AbbVie, UCB Pharma, Britannia Pharmaceuticals, Orion Pharma, and Eisai Pharmaceuticals and also received consultancy fees from Britannia Pharmaceuticals.