Abstract
Several studies of the effects on cognition of selective serotonin reuptake inhibitors (SSRI), administered either acutely or sub-chronically in healthy volunteers, have found changes in learning and reinforcement outcomes. In contrast, to our knowledge, there have been no studies of chronic effects of escitalopram on cognition in healthy volunteers. This is important in view of its clinical use in major depressive disorder (MDD) and obsessive-compulsive disorder (OCD). Consequently, we aimed to investigate the chronic effect of the SSRI, escitalopram, on measures of ‘cold’ cognition (including inhibition, cognitive flexibility, memory) and ‘hot cognition’ including decision-making and particularly reinforcement learning. The study, conducted at the University of Copenhagen between May 2020 and October 2021, used a double-blind placebo-controlled design with 66 healthy volunteers, semi-randomised to receive either 20 mg of escitalopram (n = 32) or placebo (n = 34), balanced for age, sex and intelligence quotient (IQ) for at least 21 days. Questionnaires, neuropsychological tests and serum escitalopram measures were taken. We analysed group differences on the cognitive measures using linear regression models as well as innovative hierarchical Bayesian modelling of the Probabilistic Reversal Learning (PRL) task. The novel and important finding was that escitalopram reduced reinforcement sensitivity compared to placebo on both the Sequential Model-Based/Model-Free task and the PRL task. We found no other significant group differences on ‘cold’ or ‘hot’ cognition. These findings demonstrate that serotonin reuptake inhibition is involved in reinforcement learning in healthy individuals. Lower reinforcement sensitivity in response to chronic SSRI administration may reflect the ‘blunting’ effect often reported by patients with MDD treated with SSRIs.
Introduction
Serotonin or 5-hydroxytryptamine (5-HT) is a monoamine neurotransmitter implicated in several cognitive and affective brain functions [1]. Drugs that target serotonin transmission, such as selective serotonin reuptake inhibitors (SSRIs) are the first-line pharmacological treatments for many neuropsychiatric disorders such as major depressive disorder (MDD), obsessive-compulsive disorder (OCD) and anxiety [2]. Understanding the modulatory role of serotonin on cognition and reinforcement learning is particularly important [3].
Many studies examining the modulatory effects of serotonin on cognition have been conducted in experimental animals [4,5,6]. In rats, impairing serotonin function disrupted reversal learning, whereas enhancing serotonin function improved reversal learning [4]. In marmoset monkeys, targeted neurotoxic serotonin depletion of the orbito-frontal cortex, but not of the caudate nucleus, consistently produced reversal deficits [5, 6]. Marmosets have also shown reduced reinforcement sensitivity following serotonin depletion [7].
In humans, the modulatory effects of serotonin on cognition have largely been examined through acute dietary tryptophan depletion (ATD) [8,9,10,11,12,13,14,15,16,17,18], or through acute SSRI administration [19,20,21,22,23]. ATD has been shown to affect measures of both ‘cold’ (rational and non-emotional), and ‘hot’ (social and emotional) cognition [13,14,15,16,17,18]. Specifically, ATD induces ‘waiting’ impulsivity and impulsive behaviours, impairs goal-directed behaviour and shifts behavioural control toward habitual responding in appetitive conditions, but goal-directed in aversive conditions [10,11,12]. Effects of ATD have also been seen on reinforcement behaviour [13], reversal learning [9, 14], learning and memory [15], affective and social cognition [9, 16, 17] and moral judgement [18]. Studies examining the acute administration of SSRIs have shown impaired probabilistic learning [19, 20], and impaired cognitive flexibility [20], but increased long-term memory recall [21], emotion recognition [22], and harm aversion for moral judgements [23]. One study showed that response inhibition improved with SSRI administration [20], whereas another showed no effect [19]. Taken together, a wide range of cognitive functions is affected by serotonin modulation in healthy volunteers.
Given that SSRIs are administered chronically in the treatment of neuropsychiatric disorders, it is particularly important to understand the long-term effects of SSRI administration on cognition. Currently, only a few studies have examined SSRIs administered sub-chronically, over approximately 7 days [24, 25]. Short-term administration of antidepressants may ameliorate the negative biases in information processing that are often present in mood and anxiety disorders [24]. A recent study examined both the acute and short-term effects of SSRIs. The results showed that acute administration did not affect reinforcement learning, but short-term administration resulted in increased learning from punishment, with reduced learning from reward [25]. However, there was no statistical difference in performance between the acute and short-term administration, and therefore these results must be interpreted with caution. In addition, studies with patients with MDD have shown that SSRIs impair learning from negative feedback, while having negligible effect on learning from positive feedback [26]. These findings demonstrate the difficulty in understanding the modulatory role of SSRI on various cognitive and motivational processes. One study gave a tryptophan-rich diet to middle-aged healthy volunteers for 19 days and showed that emotional bias to negative stimuli was reduced [27].
Understanding the acute effects of SSRIs on cognitive processes in healthy volunteers and patients with MDD is complex. This may be due to the differing possible pre- and post-synaptic actions [28]. In addition, there is some evidence that the neuroplasticity effects of SSRIs emerge only after more chronic administration (14–21 days) [29, 30]. As such, the chronic administration of SSRI may provide more robust results. Importantly, chronic SSRI administration is an experimental model that better mimics a treatment model of MDD. In addition, to our knowledge, no studies have examined the more chronic effects of SSRIs on a wide range of cognitive measures.
Escitalopram is the active S-enantiomer of the racemic SSRI citalopram (RS-citalopram) [31]. By removing the R enantiomer and only containing the pure active S enantiomer the effects of the drug are improved [28]. For example, there are no higher dose restrictions, and it also makes the lowest dose more efficacious [28]. In addition, Escitalopram shows very high selectivity for the serotonin transporter and is thus the best choice for testing pharmacologic actions of SSRIs [28, 31]. Moreover, escitalopram is an effective treatment for moderate-to-severe major depressive disorder (MDD) and is one of the best-tolerated SSRIs [28, 31].
In the present study, we used a double-blind placebo-controlled design to examine the effects of the SSRI escitalopram administered on average for 26 days, on a comprehensive set of measures of ‘cold’ and, ‘hot’ cognition, including decision-making and computational measures of reinforcement learning. We hypothesised that SSRI treatment would affect reinforcement-related behaviour, probabilistic reversal learning, and response inhibition.
Discussion
To our knowledge, this is the first study to determine the effects of chronic escitalopram administration on a broad range of measures of ‘cold’, and ‘hot’ cognition, including reinforcement learning in healthy volunteers. In this double-blind placebo-controlled study, a relatively large group of healthy volunteers received either escitalopram or placebo for an average of 26 days. The novel and important finding was that escitalopram had the specific effect of reducing reinforcement sensitivity in two independent tests, but had no effects on other measures of ‘cold’ or ‘hot’ cognition.
Reinforcement behaviour
The reinforcement sensitivity parameter, as modelled here, governs the degree to which a participant is driven by their reinforcement history [35]. Using this innovative approach, we found reduced reinforcement sensitivity in the escitalopram group in two different test paradigms, one on model-based vs model-free behaviour (Fig. 2) and the other in a standard PRL task. A previous study examining how reinforcement is influenced by serotonergic modulation showed that acute tryptophan depletion decreased reinforcement sensitivity, by impairing the representation of reward outcome value [13]. This was only the case for reward sensitivity, and there was no effect on punishment sensitivity [13]. In the present study, we did not find any effects on reward or punishment learning rates, whereas one study showed increased reward learning neural signals, specifically related to prediction error, following 2 week SSRI administration [39]. However, it is important to note the different methodologies used in these studies, which makes a direct comparison of the results difficult.
Importantly, our results are of considerable relevance when considering the patients’ experience of taking SSRIs chronically. Patients’ often report experiencing a ‘blunting’ effect [40,41,42]. This blunting effect has also been demonstrated for rewarding and punishing stimuli. Specifically, participants receiving 7 days of SSRI had lower neural processing of both rewarding and aversive stimuli [43]. In light of our own results, it is possible that the clinical effectiveness of SSRIs for MDD is due to this reduced negative affect. However, if indeed positive affect is also reduced, then this would lead to a more general blunting effect, as often reported by patients taking chronic SSRIs. This is supported by the present study, in which lower reinforcement sensitivity would suggest decreased control over behaviour by both rewarding and punishing stimuli. This may also be further supported by our findings that the escitalopram group had significantly higher dysfunction on the dimensions/phases corresponding to orgasm/completion on the CSFQ-14. It is possible that participants taking escitalopram experience greater sexual dysfunction due to experiencing less pleasure, which has been supported by previous reports [44]. However, this is speculative as there are other mechanisms that may explain this effect [45].
‘Hot’ cognition
Our results showed no effects on other measures of ‘hot’ cognition. Studies have previously shown that acute and sub-chronic SSRI intervention affects emotion recognition, specifically for recognition of fear and happiness [22]. In our chronic administration study, we did not examine emotion recognition in each emotion, but rather examined affective bias, which was not affected significantly by escitalopram. We also did not find any effects of escitalopram on moral judgements as previously reported following acute treatment [18, 23].
‘Cold’ cognition
Our results showed no significant effects on any measures of ‘cold’ cognition, thus contrasting with some of our previous data obtained following acute administration [20]. Previous studies manipulating serotonin acutely have shown alterations in both ‘cold’ and ‘hot’ cognitive measures [8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23]. However, it is important to note that our study examined SSRI administration over a longer time period, which has not previously been much studied in the context of human cognition. Contrasting with the present findings Skandali and colleagues [20], who used similar neuropsychological tests, showed that participants administered escitalopram acutely made more errors to criterion during Stage 1 and exhibited increased lose-shifting after misleading negative feedback in the PRL task. However, they did not conduct the same hierarchical Bayesian modelling employed in the present study. Such an analysis of the Skandali et al. [20] similarly shows reduced reinforcement sensitivity in the escitalopram group compared to placebo controls (unpublished findings). In addition, in the present study there was no effect of escitalopram on performance of the 3D-IED task, suggesting that this result may be specific to learning when there is greater uncertainty, as the 3D-IED is deterministic and the PRL is probabilistic in nature. We showed no effect of escitalopram on response inhibition, whereas Skandali et al. [20] showed that acute escitalopram improved stop-signal reaction time. However, in line with our results, Chamberlain et al. [19] showed no acute effect of the SSRI citalopram on response inhibition.
There are a number of points to note when interpreting the results from the present study. First, it is important to acknowledge the differences between acute and chronic SSRI administration. Previous literature has suggested that neuro-adaptive changes might represent homoeostatic mechanisms by which the brain regulates neurotransmission in response to the drug [46, 47]. This may suggest that the acute effects are not as robust as longer-term effects where this mechanism would have stabilised. In addition, acute administration of SSRIs does not seem to affect neuroplasticity, which does occur when administered chronically [29, 30]. Moreover, the synaptic mechanism of action for acute and chronic SSRI administration differs [28, 48]. A meta-analysis showed that within the first week of SSRI administration, 5-HT concentrations drop, which then increases over the following two weeks of administration, although this does vary slightly in different regions of the brain [48]. We chose the duration of 3 weeks because this duration is associated with clinical benefits in patients with MDD and with translational studies of neuroplasticity effects. However, we cannot rule out that neuroplasticity effects might be greater with a longer duration of escitalopram.
Second, the approaches for induction of changes in serotonin vary and this could result in the inconsistent findings. For example, previous studies manipulating serotonin acutely with different methods and using the same PRL task, showed inconsistent results using conventional behavioural measures [9, 19, 20]. It should be noted that currently there is no way to reliably determine interstitial serotonin concentrations non-invasively in humans, which means that interpretation of the manipulations must be inferential. Finally, it is likely that the escitalopram effects are less discernible in our cognitively high-performing (average IQ > 110) healthy volunteers than in patients with MDD. Studies on MDD have found that SSRI intervention often normalises abnormal neural processing [49,50,51,52], which in turn improves cognitive functioning and at a later time point, mood [52]. As healthy individuals are cognitively intact, it is possible that the effects may be different from those in patients with MDD. Differential effects on cognition and mood can be seen when studies are conducted with healthy volunteers or patients with MDD [8, 53]. As such the mechanism of SSRIs may be more restorative in MDD, which is unnecessary in healthy individuals. Given that SSRIs are chronically administered to patients with neuropsychiatric disorders, the present results are more clinically relevant than those of acute studies.
One possible limitation of the study was that there was a significant difference in guessing group allocation. However, the escitalopram group were at chance level for guessing group allocation (53% guessing correctly) and over 15% of the placebo group guessed they were on active substance. It is difficult to know how our results on guessing group allocation compares with other studies, as this measure is frequently not reported in the literature. The results are unlikely to have been affected by this, given the lack of cognitive changes and specificity of the effect.
Our results, importantly, showed a specific significant effect on reinforcement sensitivity, where escitalopram reduced reinforcement sensitivity, which may in part be explanatory for the blunting effect often reported by patients receiving chronic SSRI treatments. This study also highlights the need for future studies to examine chronic administration of SSRIs beyond 21 days. In addition, future studies should examine the chronic effects of SSRI administration on a similar extensive battery including ‘cold’, and ‘hot’ cognition, particularly reinforcement behaviour in patients with neuropsychiatric disorders such as MDD or OCD.
Conclusion
In contrast with previous reports on the acute effects of SSRI administration, we did not find any significant effects on ‘cold’ cognitive measures after more chronic administration (mean 26 days). Using an innovative computational modelling approach, we did find significant effects specific to reinforcement learning; chronic escitalopram reduced reinforcement sensitivity compared to placebo. These novel findings provide strong evidence for a key role of serotonin in reinforcement learning. The results have important clinical implications as they may reflect the blunting effect often reported by patients with neuropsychiatric disorders receiving chronic SSRI treatment.
Source: Nature