4.1 Summary and Key Results

We investigated aripiprazole and pramipexole reports to capture the syndromes and sub-syndromes related to drug-induced impulsivity with dopamine partial agonists and dopamine agonists, respectively.

The event-event disproportionality analysis revealed signs and symptoms commonly reported alongside impulsivity, thus delineating an impulsivity syndrome separately for aripiprazole and pramipexole recipients. The impulsivity syndrome encompassed mainly psychosocial events but also organic conditions.

The network analysis identified meaningful clusters, such as delusional jealousy (also known as Othello syndrome [60]) and dopamine dysregulation syndrome (i.e., the excessive use of levodopa) in pramipexole recipients, and obesity-hypoventilation syndrome (historically Pickwickian syndrome) and social issues in aripiprazole recipients. In particular, employing more sensitive network analyses like Phi and PPMI generated more interconnected networks, which identified potential macro-clusters combining several smaller clusters identified by the Ising model.

Crucially, potential causal mechanisms and secondary consequences of drug-induced impulsivity can be highlighted by Bayesian Network methods, providing targets for potential interventions, for example targeting anxiety and economic problems.

In order to better contextualize and qualify these findings, we now discuss in detail the results of case retrieval, disproportionality analysis, network analysis, Bayesian network, and the limits and conclusions of the study.

4.2 Case Retrieval

Our findings align with established risk factors for impulsivity, including male gender and younger age [61, 62], Parkinson’s Disease (PD) [63, 64] and depression [65]. Commonly reported impulsivity manifestations included the “four knights” [9] (i.e., gambling, shopping, hyperphagia, and hypersexuality), garnering special attention due to their pronounced impact on QoL. Other manifestations were body-focused repetitive behaviors, paraphilic disorders, and hoarding.

4.3 Disproportionality Analysis: The Drug-Induced Impulsivity Syndrome

By performing the event–event disproportionality analysis within each drug population separately, comparing reports involving impulsivity with those encompassing various reactions—other than impulsivity—to the same drug, we addressed indication bias and other confounding factors. This comparative analysis served as a rigorous filter, allowing us to sift through the complex data and unveil the genuine characteristics associated with impulsivity, as well as those arising from the dynamic interaction between impulsivity and the underlying drug or disease, excluding traits tied solely to the underlying drug or disease.

This approach revealed a complex syndrome, characterized by psychosocial, cognitive, psychosomatic, and metabolic events. The syndromes identified for impulsivity within pramipexole and aripiprazole recipients differ significantly. Multiple factors may contribute to the seemingly higher burden on QoL observed with aripiprazole-induced impulsivity, with functional (or psychosomatic) manifestations and social issues impacting work, relationships, and economics. Pramipexole is primarily administered to older patients with hypodopaminergic conditions, characterized by motor impairment and reduced motivational drive. These patients, well managed and supported by caregivers because of the later onset and clear neurologic origin of the disease, may experience a mitigated drug-induced impulsivity burden. Conversely, aripiprazole is prescribed to younger patients with mood and psychotic disorders, often linked to hyperdopaminergic states and a pre-existing diathesis for impulsivity. Challenges for caregivers and social support are heightened in these cases due to earlier onset, psychiatric origins, and stigma, potentially leading to a greater burden. Further, over one-third of aripiprazole cases were submitted by lawyers. This may be explained either by a potentially malicious overreporting for legal compensation (cfr., Abilify lawsuit) [66] or by a reaction to underdiagnosis by physicians, who may be hesitant to attribute behavioral changes to the medication when underlying psychiatric conditions are present. In reports filed by lawyers, the desire to win legal compensation may have prompted more detailed descriptions of the impact of impulsivity on QoL, or possibly even exaggerated the impact, thus contributing to the observed differences between the two drugs. Intriguingly, there could also be an ascertainment bias, as neurologists prescribing pramipexole may be less aware of psychiatric issues compared to psychiatrists prescribing aripiprazole.

4.4 Network Analysis: Sub-syndromes

Network analysis in pharmacovigilance, complementary to other unsupervised approaches such as vigiGroup [67], is a promising tool to detect potential syndromes and sub-syndromes, to help the characterization of signals. Employing three estimation methods, the network analysis revealed potential sub-syndromes associated with specific impulsivity expressions in the two populations (Fig. 7). The Ising delineated well-defined clusters, while PPMI and \(\phi\) emphasized inter-clusters relationships. By incorporating various expressions of impulsivity, we anticipated that the central cluster would encompass the key features of impulsivity regardless of its form, whether they act as risk factors or consequences of impulsivity. In both populations, cognitive and mood disorders (e.g., cognitive and memory impairment, bipolar disorder, depression) were included in the central cluster. Notably, they have been recorded as frequently associated with drug-induced impulsivity and contributing to disability development [68]. Obesity-hypoventilation syndrome [69], which involves weight gain, cognitive and sleep disorders, and sedation, was consistent in both populations but seemingly heavier in aripiprazole recipients, which also reported obesity, sleep apnea syndrome, hypertension, and metabolic blood alterations (increased lipids, transaminases, and glucose in the blood), supporting the observed link between hyperphagia and diabetes onset [70].

Fig. 7

Drug-induced impulsivity syndrome, aripiprazole and pramipexole. The main syndrome, representing one or more strongly interconnected central clusters of symptoms and signs identified through network analysis, is depicted as the central figure. Other potential sub-syndromes are shown on the sides highlighted with a colored square

For aripiprazole recipients (Fig. 4), the central cluster also included panic attack and auditory hallucinations, sleep disorders, decreased appetite, and stress. Stress was further connected to a psychosomatic sub-syndrome involving irritability, migraine, back and abdominal pain, reflux, diarrhea, constipation, and hyperidrosis. Gambling and shopping were linked to pervasive social issues (hoarding, unemployment, homeless, bankruptcy, divorce), theft, and suicidal ideation and attempts (already observed during hyperdopaminergic impulsive states [71]). Hypersexuality was linked to unintended pregnancy, sexually transmitted diseases, and sexual dysfunction.

Among pramipexole recipients (Fig. 5), the central cluster also included apathy, delusion, and economic problems. The dopamine dysregulation sub-syndrome (a manifestation of pathological impulsivity marked by excessive levodopa use [72,73,74], co-administered with dopamine agonists to better control motor symptoms), involved on and off phenomenon (oscillations in effectiveness and motor and motivational symptoms), excessive levodopa use to avoid off phases, and dopamine agonist withdrawal syndromes (DAWS) upon discontinuation [75]. A cluster aligned with paranoid delusional jealousy (false and unwavering belief in the partner’s unfaithfulness), often seen in PD with drug-induced hypersexuality [76] and here characterized by delusional jealousy, hallucinations, irritability, crying, and marital problems. We also found a cluster with fear, pain, stress, anxiety, depression, and suicidal ideation, indicative of the transformation of reward-driven impulsivity into stressful risk-averting compulsivity over time [77]. Finally, the co-reporting of two archetypal compulsive symptoms—body-focused repetitive behaviors and stealing behaviors—was evident.

4.5 Bayesian Network: The Secondary Impact of Drug-Induced Impulsivity

This interplay of events within the context of drug-induced impulsivity is intricate and multifaceted. Events reported alongside drug-induced impulsivity may result from impulsivity itself (like financial problems from gambling) or may predispose individuals to impulsivity (e.g., bipolar disorder). Sometimes, events can both trigger and be exacerbated by drug-induced impulsivity (e.g., anxiety [78, 79]). Sometimes events are concomitantly mentioned for precision, such as in cases of semantic overlap (e.g., theft and shoplifting, or injury and brain injury). Events associated with drug-induced impulsivity may even be synonyms for well-known impulsivity expressions (e.g., restlessness, referring to excessive wandering and poriomania), or could be the very reason for prescribing the drug, as seen in the off-label use of aripiprazole and dopamine partial agonists to prevent behavioral and cognitive decline in brain injury [80] or to address dependence [81,82,83,84,85]. We implemented a Bayesian Network to obtain insights into potential directional associations to attempt the formulation on clinically plausible causal sequences. Anxiety emerged as a central factor, preceding insomnia, irritability, cognitive impairment, stress, injury, pain (linked to disability and economic problems), depression, and even suicidal ideation. Drug-induced impulsivity manifestations appeared to exacerbate each other. Economic problems had the highest out-degree centrality among aripiprazole recipients, preceding theft, relationship difficulties, and suicidal ideation.

The Bayesian Network provides researchers with valuable insights on the pivotal nodes that could be targeted by interventions to disrupt the cascade of events and ameliorate the secondary impact of drug-induced impulsivity. It also highlighted secondary ramifications of main impulsivity manifestations: hypersexuality precedes marital problems through delusional jealousy in pramipexole recipients, while it precedes unintended pregnancy and sexually transmitted diseases in aripiprazole recipients; hyperphagia precedes weight increase in pramipexole recipients and obesity, somnolence, and cognitive impairment in aripiprazole. Finally, the Bayesian Network seems to support the higher secondary impact of drug-induced impulsivity in aripiprazole recipients.

4.6 Clinical Considerations

This study underscores the significant burden of drug-induced impulsivity, which encompasses biological, psychological, and social consequences. Clinical outcomes related to such impulsivity are broad and multifaceted, often manifesting as complex syndromes rather than isolated events. These syndromes can have prognostic and therapeutic implications. For instance, impulsivity in the form of hyperphagia can lead to metabolic syndrome and sleep disorders, in the form of hypersexuality to sexually transmitted diseases and unintended pregnancies, and in the form of pathological gambling and compulsive shopping to severe social issues such as job loss, bankruptcy, and divorce. Furthermore, the impact of aripiprazole-induced impulsivity appears more severe compared to pramipexole, potentially due to differences in the patient populations using these medications.

Given the substantial burden associated with drug-induced impulsivity, it is crucial to meticulously review patients’ medical histories. Factors such as young age, male gender, pre-existing mood disorders, and family history of dependencies should be considered red flags, necessitating heightened vigilance. When impulsivity is detected, the recommended course of action is to taper the offending medication and switch to an alternative. However, this pharmacological switch is not always feasible or sufficient, and in this case there is no consensus among experts on the optimal strategy [86]. In such cases, the Bayesian Network can be an invaluable causal discovery tool for identifying critical events that might be targeted by interventions to prevent the chronicization and exacerbation of drug-induced impulsivity. For example, effective management strategies might include monitoring and addressing anxiety, providing financial guidance, or appointing legal guardianship to prevent wasteful spending, thus targeting critical events in the evolution of the syndrome. Additionally, addressing marital issues is crucial, as they are linked to early placement in nursing homes and poorer prognoses [24, 87]. Specifically, tackling delusional jealousy and economic problems, which often precede marital issues, may be vital for preserving the well-being of pramipexole recipients.

In conclusion, this study highlights the necessity for comprehensive clinical evaluations and individualized management plans for patients at risk of or exhibiting drug-induced impulsivity. Proactive measures and targeted interventions, driven by the conceptualization of ADRs as networks of causally interacting events, can mitigate the adverse outcomes associated with these medications, improving overall patient prognosis and QoL.

4.7 Limitations and Further Developments

While this study provides valuable insights into the intricate interplay of events related to drug-induced impulsivity and its subsequent implications, it is crucial to acknowledge its limitations.

Individual case safety reports, while uniquely granting access to a patient’s perspective, are susceptible to biases such as under-reporting, missing data, and unverified reliability, preventing reliable incidence or prevalence estimates. The high contribution of reports from lawyers may have influenced the higher psychosocial impact attributed to aripiprazole-induced impulsivity. Nonetheless, this study sets the foundation for further studies and a potential score to assess the impact of ADRs on QoL.

Limitations in network analysis methodologies adopted include the Ising estimation’s assumptions (pairwise interaction, linear effects, and binary variables), and the inability to account for time and severity in symptom manifestation. The incorporation of negative links could facilitate a more nuanced separation of symptoms that infrequently co-occur. The Bayesian Network lacks bidirectional relationships and cyclic feedback loops and would require the inclusion of all shared causes between any two events (causal Markov condition) to achieve its full capacity to illuminate causality. These limitations could be rectified by integrating clinical longitudinal data and embedding temporal aspects into the network analysis (see Fusaroli et al for similar, more detailed, considerations [88]). Moreover, the used models assume that an event can only be part of a sub-syndrome, while we know that in fact the same event can be a manifestation of multiple sub-syndromes. In the future, it could be an opportunity for the integration of network analysis with vigiGroup [67], a latent class expectation maximization model developed by Uppsala Monitoring Centre, which instead allows for one event to be part of multiple sub-syndromes (but not for one report to describe multiple sub-syndromes).

Looking ahead, a broader definition of drug-induced impulsivity could improve sensitivity in case retrieval. Conditions such as suicide attempts, hypersomnia, obsessive-compulsive symptoms, explosive anger, personality changes, disturbance in attention, and drug dependence might represent different expressions of this underdefined condition, warranting further exploration [10, 74]. Additionally, aripiprazole and pramipexole may not fully represent their entire drug classes. This could be due to their use in different populations or their distinct pharmacological activities. In the future, it would be valuable to conduct network analyses on newer drugs (e.g., brexpiprazole and cariprazine) as more reports become available.