Abstract
Abstract
Our knowledge of the cerebral bases of decision making has grown considerably in the past decade. The dopamine system is closely involved in many aspects of the decisional process. It is therefore not surprising that the dysfunctions that occur in Parkinson’s disease (PD) can alter some patients’ decisions. Put simply, a decision is the final step of a process in which a subject weighs up the potential benefits and costs associated with each of the different options available for a given choice. The option that appears to have the best ratio of benefits to costs is chosen. In some PD patients, dopamine agonists destabilize the balance: the benefits are given an inappropriately high weighting relative to the costs, leading patients to take decisions they would not otherwise have taken. This might be one of the explanations for impulse control disorders observed in some PD patients. Dysfunction of the subthalamic nucleus (STN) induced by dopamine replacement or by deep brain stimulation is another mechanism that can alter decision making. The STN plays an active role in the decisional process, especially by slowing down the process when the difference between the options to be considered in a given choice is small (e.g. a win-win choice). Deep brain stimulation applied to the STN may interfere with its monitoring role and lead to an impulsive choice. Attention disorders and frontal lobe dysfunction, highly prevalent in the course of PD, are other factors that may alter a patient’s decision making. Patients and caregivers need to be aware of this, since the consequences can sometimes be detrimental.
Keywords
ABBREVIATIONS
Parkinson’s disease impulse control disorders positron emission tomography subthalamic nucleus deep brain stimulation
INTRODUCTION
Decision making has reached a particularly high level of sophistication in the human brain
[1, 2]. Among
the many cerebral structures involved, the basal ganglia have a key position. It is
therefore not surprising that decision making can easily be impaired in Parkinson’s disease
(PD), as a consequence of the disease
A decision is an executive process consisting of choosing among different options of action the one to perform (“action”, as used here, also includes non-action or the planning of an action that may or may not be executed). We are constantly taking decisions. Many decisions are taken automatically and we are not fully aware of them (e.g. the decision to stop walking when one sees a snake or the decision to choose a given sandwich for lunch). Others are more under control: we are well aware that we are taking a decision (e.g. the decision to buy a car or the decision to attend a meeting abroad). These different types of decision involve different neural networks.
Schematically, for all types of decision three main steps can be considered: i) an
evaluation of the different options available; ii) a comparison between these options; and
iii) the choice of the best option to execute, whether immediately or later (Fig. 1). Valuation. The first step in the decision
making process is to attribute a value to each option considered in the choice. A
scale of measurement exists in the brain. It allows any option to be positioned and
thus compared with any another option regardless of its nature (e.g. a chocolate cake
or an ice cream, attending a meeting abroad or staying at home). The value scale is,
however, individual. In a given person, it changes constantly with time. In economics
it has been defined as the “utility” measure [3]. The construction of the value scale is linked to the person’s previous
positive or negative experiences. Comparison. Positioned on the unique scale, the different options can be easily
compared to find which one is, for the person taking the decision and at the time of
the decision, the best one to consider. Once the best option has been determined, the decision leads to the immediate or
deferred execution of the action linked to the option.
DOPAMINE AND THE VALUATION PROCESS
The valuation process involves two major brain structures: the orbitofrontal cortex and the ventral striatum. The orbitofrontal cortex is the structure where the value scale is managed. Cell recordings in monkeys have provided good evidence for this. For example, some cells have a level of activity associated with the value attributed by the animal to a given beverage. The level of activity is ranked according to the degree of preference for each of the available beverages. It codes for the value that the animal assigns to the beverage [4]. The value, thus the cell activity level, is relative since it changes according to the choice of beverages presented (e.g. apple juice may be the most preferred beverage in one experiment and the least preferred in another; the cell activity coding for the value will be the highest in the first experiment and the lowest in the second one). The value of a beverage can also change with time, for example if the animal has drunk a large quantity of a given beverage before the experiment, the value of the beverage will decrease when presented in a choice, compared to the same choice when the animal has not reached satiety for the beverage [5]. In humans, the activation of this region is associated with the so-called willingness to pay (i.e. the amount of money one is ready to spend to acquire a certain good or service) [6].
The ventral striatum and dopamine are closely involved in the constant construction of the value scale. For each performed action, dopamine “indicates” to the orbitofrontal cortex whether the action is associated with a positive or a negative outcome. Experience after experience, the orbitofrontal cortex “learns” the value of a given action.
A positive outcome of a given action is accompanied by a phasic burst of dopamine cells located within the ventral tegmental area, which leads to a transient increase in the tonic (permanent) release of dopamine within the ventral striatum. More specifically, the intensity of the burst, thus of the amount of released dopamine, is linked to the a priori expectation of the positive outcome (reward), as illustrated in the following example. Based on previous experiences (e.g. bad weather, boring time, etc.), a given action (e.g. the action of going to the beach) may be considered to be associated with a high probability of a negative outcome. If the execution of the action is unexpectedly associated with a positive outcome (e.g. a sunny afternoon with good friends), the release of dopamine is very high. In contrast, a low amount of dopamine is released in the case of a positive outcome of an action regularly associated with a positive outcome (e.g. if every time one attends the International Congress on Non-Motor Dysfunctions in Parkinson’s Disease and Related Disorders the lectures are great, then listening to a great lecture at this Congress will not be associated with a high release of dopamine in the ventral striatum!). Dopamine is said to code for the expected value of an action.
Most striatal cells are gabaergic. Among these, two major populations can be differentiated: gabaergic cells containing D1 dopamine receptors that project directly to the internal part of the globus pallidus and the substantia nigra pars reticulata (the so-called direct pathway), and gabaergic cells containing D2 dopamine receptors that project to the external part of the globus pallidus. These latter cells project to the subthalamic nucleus, which contains cells projecting to the internal part of the globus pallidus and the substantia nigra pars reticulata (the so-called indirect pathway). On the D1 receptor, dopamine leads to cell activation. On the D2 receptor, dopamine leads to cell inhibition. Subsequently, the dopamine release that accompanies the positive outcome of an action activates the direct pathway in the motivational part of the basal ganglia. Through this activation, the orbitofrontal cortex receives information about the outcome of the action and further adjusts the value scale. In this way, a positive outcome increases the value of the action, i.e. the action will be considered as having a higher probability of a positive outcome in the future than before. This sums up the simplified neural basis of positive reinforcement learning (Fig. 2A).
In contrast, a negative outcome of a given action is accompanied by a transient decrease – a dip – in the baseline tonic activity of dopamine cells in the ventral tegmental area, resulting in a transient decrease in the basal level of dopamine released continuously in the ventral striatum. Again, the effect is amplified in the event of a negative outcome of an action that is usually expected to have a positive outcome. The decrease of dopamine in the ventral striatum subsequently activates the indirect pathway in the motivational part of the basal ganglia. In this way, the orbitofrontal cortex receives information about the outcome of the action and adjusts the value scale accordingly. This sums up the neural basis of negative reinforcementlearning (Fig. 2B).
The learning process is associated with a succession of positive and negative experiences that activate the direct or the indirect pathway in the motivational part of the basal ganglia, updating continuously the value scale in the orbitofrontal cortex. Hence, a normal person learns, through positive (‘carrot’) and negative (‘stick’) reinforcements, the value of actions [7–10].
In PD, the functioning of the dopamine system is disturbed by both the lack of dopamine due to the disease and the non-physiologic replacement of dopamine provided by the antiparkinsonian treatment. How does it affect the learning process described above? To address this issue, Frank and colleagues trained PD patients and age-matched control subjects to choose between pairs of symbols [11]. Each symbol was associated with a given probability of reward (i.e. high, low or intermediate probability of reward). After each choice, patients and subjects received immediatefeedback (win or lose). Hence, trial by trial, patients and subjects learned through positive and negative reinforcements the value (i.e. the probability of reward) of the different symbols. In a second part of the experiment, patients and subjects had to choose between new pairs of the same symbols, but did not receive any feedback after their choice. The first group of pairs systematically contained the symbol that was associated with the highest probability of gain (80% ) in the first part of the experiment. The second group of pairs systematically contained the symbol associated with the lowest probability of gain (20% ). The analysis of the results helped to determine whether the patients and the subjects had learned preferentially by positive or negative reinforcement (depending on whether they preferentially recognized the symbol with a high probability of gain or the symbol with a low probability of gain). A normal subject learned in an equal fashion by positive and negative reinforcement. In PD patients, the learning process is altered. The learning process is linked to the amount of brain dopamine: in a state in which they were deprived of brain dopamine (‘off’ state), patients mainly learned by negative reinforcements; in a state in which their lack of dopamine was compensated by the antiparkinsonian treatment (‘on’ state), they mainly learned by positive reinforcement.
In PD, there is a tendency for patients, at least those who are globally undercorrected (i.e. in an hypodopaminergic state) most of the day, to attach more weight to negative than to positive outcome results; in other words, they are more likely to see the negative rather than the positive aspects of a given choice. This is the likely explanation for many PD patients being more risk averse than non-PD patients.
WHEN DOPAMINE AGONISTS DISTURB THE VALUATION PROCESS
Most likely due the extensive use of dopamine agonists during the past 20 years, some behavioral disturbances, such as irrepressible gambling, compulsive shopping or hypersexuality, have recently been observed with an unusually high frequency in PD patients (about 13.6% [15]; and, according to a study, gambling is 25 times more frequent in PD patients than in the general population [16]). These behavioral disturbances are now commonly classified under the heading of impulse control disorders (ICDs). ICDs are almost exclusively observed in patients treated with dopamine agonist alone or associated with L-DOPA [15]. A disturbance in the valuation process might be one of the explanations for ICDs. PD patients with ICDs might attach more weight to the positive aspects of an option than to their negative aspects. For example, patients attach more weight to the positive aspect of gambling (i.e. the prospect of winning a large amount of money) than the negative aspects associated with the amount of money spent in participating to the game.
The fact that ICDs are observed with dopamine agonists and are less likely in patients treated with L-DOPA alone lies at the heart of this explanation. One of the differences between dopamine agonists and L-DOPA is that dopamine agonists have a sustained effect on dopamine receptors whereas L-DOPA has only a transient effect. In patients being treated with dopamine agonist, there is a sustained activation of dopamine receptors, which prevents the transient “dip” in tonic dopamine release that normally occurs in the case of a negative outcome of an action. As remarked by Frank et al. [11]: “the dopamine agonists fill the gap”. Since the valuation system loses information about negative outcomes, the valuation will be weighted in favor of the positive aspects of an option. In contrast, L-DOPA treatment leads to the synthesis of brain dopamine, thus a natural compound having less impact than dopamine agonists on the variation in the phasic and tonic dopamine activity during the valuation process. When L-DOPA is used as a single treatment in PD, the valuation process is not disturbed. Recent studies have investigated this issue more specifically. They studied learning from gains and losses in PD patients with ICDs and compared the results to those of PD patients without ICDs, using a protocol similar to that of the study by Frank et al. described above. One study [15] found that in patients with ICDs dopamine agonists increased the rate of learning from positive outcome and led these patients to consider the outcome as “better than expected”. In contrast, another study [16] observed that patients with ICDs clearly took less into account negative prediction errors in the ‘on’ state than the patients without ICDs (whether in the ‘on’ or the ‘off’ state) and healthy controls. Patients with ICDs in the ‘on’ state appeared to have lower learning from negative feedback. Further studies are needed to clarify the exact mechanism. Despite the discrepancies, these studies demonstrate that the valuation process is impaired in PD patients with ICDs, leading them towards expected gains without appreciating the negative aspects of their choice.
A subsequent question is why only a subgroup of PD patients treated with dopamine agonists develops ICDs. An analysis of the level of dopamine released in the ventral striatum during gambling using [11C] raclopride positron emission tomography (PET) has provided some clues [17]. PD patients with pathological gambling have a greater release of dopamine than PD patients without such a disorder. There are therefore some similarities between PD patients with pathological gambling and patients with chemical addiction, in whom an increased release of dopamine in response to their drug of choice has been observed. Excessive gambling can be viewed as a behavioral addiction [18]. Interestingly, during a control task, the binding of raclopride to the D2/D3 dopamine receptors was decreased in the ventral striatum in PD patients with pathological gambling compared to PD patients without such impulsive behavior. Such a difference could be due either to a higher basal release of dopamine or to a lower density in D2/D3 receptors in theformer compared to the latter. It might be a marker of vulnerability to addiction that may exist even before PD develops. When vulnerable patients start the disease and receive a dopamine agonist treatment to correct their parkinsonism, an addiction-like behavior might emerge. The fact that patients with ICDs often have a history of higher novelty seeking, impulsivity, or drug or alcohol abuse [19], is in line with such a vulnerability hypothesis. A genetic susceptibility to ICDs in PD patients is probable. Although several studies have searched for polymorphisms of genes involved in dopamine metabolism, no clear conclusion can yet be drawn [20].
THE KEY ROLE OF THE SUBTHALAMIC NUCLEUS WHEN DEALING WITH A DIFFICULT DECISION
Some decisions are more difficult to take than others. Decision making is particularly difficult when two of the options available will lead to a very similar level of benefits, namely a win-win situation. For example, whether to attend an outstanding lecture by a recent Nobel prize winner or enjoy a great time on a sunny beach! In such cases, the decision-making process needs to be slowed down to analyse in more detail the advantages and drawbacks of each option and carefully compare them. The basal ganglia, and especially the subthalamic nucleus, appear to have an important role in this situation. To demonstrate the place of the subthalamic nucleus in the slowing down process, Frank et al. [21] confronted PD patients treated by subthalamic deep brain stimulation (STN DBS) with a win-win situation. The protocol was similar to the one described above, but after the learning phase the subjects faced either a high conflict choice, with pairs of symbols associated with a similar probability of gain (e.g. 80% and 70% chance to win), or low conflict choice, with pairs of symbols associated with a more different probability of gain. They compared the responses of these patients while they were in the ‘on stimulation’ condition and while they were in the ‘off stimulation’ condition, and also compared them to patients not treated with STN DBS but ‘on’ and ‘off’ L-DOPA. They found that PD patients in the ‘on stimulation’ condition failed to slow down the decision-making process when faced with a high conflict choice when compared to their performance in the ‘off stimulation’ condition, and when compared to other PD patients not treated by STN DBS and either ‘on’ or ‘off’ L-DOPA. The STN appears to play a crucial role in slowing the process when a difficult decision has to be made, at least in a win-win situation. This may explain the impulsivity that may be observed in PD patients treated bySTN DBS.
OTHER FACTORS THAT MAY ALTER THE DECISION PROCESS IN PARKINSON’S DISEASE
Apart from difficulties in valuing the available options and in slowing down the valuation
process when faced with a difficult decision, many other factors may affect the decision
process in PD patients. Attention is an important but limited resource in the
decision-making process, especially in decision under control (i.e. a decision the subject
is fully aware of and has explicitly assessed the advantages and the drawbacks associated
with the consequences of the decision). Due to the disease
The frontal lobe is one of the main sites of dysfunction when the disease spreads and leads to cognitive decline. Due to the crucial role of this part of the brain in the decision-making process, PD patients are at a high risk of dysfunction. Frontal lobe dysfunction is not always easy to detect at an early stage of cognitive decline [24]. An alteration of the decision-making process might therefore be an early manifestation.
Conclusions
For many reasons PD patients are at a high risk of making an erroneous choice when they are faced with making a decision. The involvement of dopamine and basal ganglia in the decision-making process is one of the main explanations. Moreover, the antiparkinsonian treatment, especially dopamine agonists and STN DBS, might lead, through various mechanisms, lead to impulsive and biased decisions with consequences that may be highly prejudicial to the patients. At the early stage of PD cognitive decline, even when the decline has not yet been detected, the dysfunction of the frontal lobe may also alter the decisional process. Patients and caregivers should be made aware of this risk, especially when important decisions (e.g. financial investment) need to be made. Further studies, including the analysis of the social impact of the disease, are still much-needed in order to clarify the impact of the alteration of the decision-making process in PD patients’ daily life.
CONFLICT OF INTEREST
The author declares having no specific conflict of interest in relation to the writing of this review.