Antipsychotics and obsessive–compulsive disorder/obsessive–compulsive symptoms: A pharmacovigilance study of the FDA adverse event reporting system

Antipsychotics have conflicting data with respect to obsessive–compulsive disorder/symptoms (OCD/OCS), with some reporting causality and some reporting treatment benefits. This pharmacovigilance study aimed to investigate reporting of OCD/OCS in association with the use of antipsychotics in comparison to one another, as well as treatment failure using data derived from the FDA Adverse Event Reporting System (FAERS).

of pharmacovigilance studies they should ideally be validated through alternative prospective research studies involving direct comparisons of antipsychotic agents.

K E Y W O R D S
antipsychotic, FDA adverse event reporting system, obsessive-compulsive disorder, pharmacovigilance, reporting odds ratio

| INTRODUCTION
Obsessive-compulsive disorder (OCD) is characterized by recurrent, intrusive thoughts neutralized only by routinely performing a behavior in response to the urge. 1 These obsessive or compulsive symptoms gravely impair an individuals' ability to function. The prevalence rate of OCD is around 2%-3%. Obsessive and/or compulsive symptoms (OCS), while characterized by the same symptoms as OCD, are subsyndromal, implying a lower level of severity as compared to OCD. 2 Pathophysiologically OCD appears complex, with some research supporting anomalies in the dopaminergic, serotonergic, and even glutamatergic systems, possibly within the orbitofrontal cortex, prefrontal cortex, cortico-striato-thalamo-cortical (CSTC) pathways, reward circuit, and the limbic regions. [3][4][5][6][7] Neuroimaging findings also point to the CSTC model by demonstrating hyperactivity in the prefrontal cortex (primarily orbitofrontal cortex), anterior cingulate cortex and caudate nucleus. Emerging imaging findings have shown involvement of widespread associative networks including parietal cortex areas, limbic areas and cerebellum. 8 Dysfunction in these aforementioned areas of the brain may lead to over-valuation of ideas with inhibited behavioral control.
Even though prolonged administration of selective serotonin reuptake intake inhibitors (SSRIs) is the most effective psychopharmacological treatment of OCD, around 40%-60% of patients fail to satisfactorily respond to monotherapy with serotonergic drugs. 9,10 The major focus of studies on refractory OCD have emphasized the study of psychotropic medications as adjuncts. Refractory OCD can be addressed with different approaches including a switch to another SSRI or clomipramine, or augmentation with a second-or first-generation antipsychotic. 9 Aripiprazole, haloperidol, and risperidone have all shown positive results on refractory OCD as adjuvants. The evidence is less promising with regards to quetiapine, olanzapine, and paliperidone. [11][12][13][14] Interestingly, despite the apparent benefit of some antipsychotics in the treatment of refractory OCD, some antipsychotics such as clozapine and olanzapine have classically been thought to worsen or induce OCD or OCS, based on the large number of published cases in comparison to other antipsychotics. 15 The prevalence of developing de novo OCS or experiencing an exacerbation of OCS secondary to clozapine was estimated at 20%-28% and 10%-18%, respectively. 15 While other antipsychotics have published reports on presumed OCD/OCS causality, it remains unclear if specific agents are truly more highly correlated with this adverse drug reaction (ADR). Although the mechanism underlying antipsychotic-related OCD/OCS ADRs is unclear, one possibility is that since serotonin and dopamine often counterbalance one another through feedback mechanisms, any inhibition of serotonin or

Significant outcomes
• OCD was reported more frequently with aripiprazole than with any other antipsychotic evaluated including those like clozapine, which have historically been considered to be at greatest risk for this adverse effect. • Risperidone, paliperidone, and haloperidol had the lowest reporting of de novo or exacerbated OCD/OCS, while risperidone and quetiapine had the lowest reporting of failure when used for OCD/OCS. • Spearman's correlation depicted a significant strength of association for OCD/OCS with the 5-HT 1A receptor as well as the 5-HT 1A /D 2 receptor ratio.

Limitations
• Disproportionality studies using voluntarily reported data have inherent issues, including reporting biases or under-reporting. • Misclassification of OCD for alternative diagnoses, including impulse control disorders, may have skewed analysis. • We did not evaluate the dose of antipsychotic as it relates to the reporting of de-novo or exacerbated OCD/OCS. serotonergic receptors, as seen with depression or through use of certain antipsychotic medications, may elicit OCD symptoms, perhaps through upregulation of dopamine. Still, because OCD/OCS are frequently comorbid with conditions like schizophrenia and other mood disorders treated with antipsychotics, it is important to rule out iatrogenic causes to ensure harm reduction. 16 Because of the seemingly rare occurrence of OCD/OCS ADRs and the lack of primary clinical trial data related to these symptoms, it is difficult to compare the relative risk of treatment-emergent OCD/OCS with regard to individual antipsychotics. This is where postmarketing data may prove useful in determining a disproportionality signal for potential correlation with specific antipsychotic agents, and provide evidence for a potential iatrogenic cause of de novo OCD/OCS. In this pharmacovigilance study, we examined the signal of association of OCD/OCS among antipsychotic medications by using reports to the Food and Drug Administration Adverse Event Reporting System (FAERS) database to calculate the Information Component (IC), a measure of signal strength between a drug and an ADR. We also calculated the reporting odds ratio (ROR) for all antipsychotics reported to be associated with OCD/OCS in the FAERS system, to discern any major differences between the evaluated agents. The purpose of this study was to evaluate pre-existing international data in order to provide directionality on OCD/OCS associated with antipsychotics, laying the groundwork for future studies on this topic.

| Data collection
All data were obtained from the FAERS Public Dashboard, a worldwide pharmacovigilance database containing consumer and clinician reported ADRs, which are voluntarily reported. 17 The FAERS Public Dashboard database utilizes coding terms which adhere to the Medical Dictionary for Regulatory Activities (MedDRA) per the International Council for Harmonization (ICH). 17 This database contains 25 variables, including demographic, reporter, medication, adverse reaction, and outcome data. It is important to note that reports to the FAERS system do not impute causality. Further, it should be stated that information within the FAERS system is not verified by a third party. Any report to FAERS of [("obsessive-compulsive disorder") or ("obsessive-compulsive symptom")] from January 1st, 2010 to December 31st, 2020 was queried, as well as all reports of any ADR for the evaluated antipsychotics for case/non-case analysis. These cases were downloaded, and the data were manipulated in Microsoft Excel ® . All FDAapproved antipsychotics found within FAERS with at least 1 report for OCD/OCS were included. From there, all ADRs for the included antipsychotics of interest (i.e., those having at least 1 report for OCD/OCS) were individually queried for the same time period and downloaded into Microsoft Excel ® . For the purpose of this study we excluded any other class of medications outside of antipsychotics from analysis (e.g., amphetamines, antidepressants, or antiepileptics), in order to examine specific medication effects, as well as control for the effect of the underlying disease state. In addition, only the primary suspected antipsychotic was included in the analysis, to avoid any influence from dual antipsychotic use. Lastly deduplication was performed manually based on strong case similarities (e.g., patient age, sex, reaction terms, concomitant product names, FDA received date all the same). When duplicates were identified, one case was deleted from the analysis.

| Statistical analysis
We utilized the IC to measure any disproportionality between the observed number (N OBS ) and expected number (N EXP ) of notifications for the drug-ADR combination. 18 The IC analysis (IC = log 2 ((N OBS + 0.5)/(N EXP + 0.5))) was run across all FAERS ADR reports for the defined time range correlated with the number of reports of OCD for each antipsychotic for that same time range. A positive IC value indicates a higher-than-expected reporting of the ADR to FAERS. Safety signaling was calculated utilizing the IC 025 , the lower threshold for statistical significance with an IC 025 greater than 0 being considered significant. The quotient of the N OBS to the calculated N EXP , which is also known as the relative reporting ratio (RRR), was computed for any antipsychotic with a significant IC 025 in efforts to remove bias from reporting, such as historical bias with older or newer antipsychotics that may be prescribed less frequently (e.g., perphenazine and cariprazine, respectively). The N EXP was computed using the equation (N DRUG Â N REACTION )/N TOTAL , where N DRUG is the number of notifications for the drug, regardless of the ADR, N REACTION is the number of notifications for the ADR, regardless of the drug, and N TOTAL is the total number of ADRs reported to FAERS over the examined time period. The ROR and 95% confidence interval (CI) for each antipsychotic with at least 1 report of OCD/OCS was calculated using the 2 Â 2 contingency table ((A Â D)/(B Â C)) formulation ( Figure 1) to determine if one specific antipsychotic was more highly correlated with OCD/OCS than another through drug-event pairing for intra-class analysis. 19 While the case threshold for calculating ROR is sometimes considered 3 reports, our threshold of 1 report was used to reduce bias for less commonly used antipsychotics, increase power of the analysis, and formulate a more insightful opinion on first vs. second-generation antipsychotic OCD-reporting. The ROR provides information on the odds of ADR occurrence with a medication relative to reference medications. In our study, a ROR >1 suggested OCD/OCS is more frequently reported with the drug of interest versus the comparators (i.e., all other antipsychotics of interest). All RORs with respective confidence intervals and p-values were calculated using a free online tool (https://www.medcalc.org/calc/odds_ratio.php), while the IC and IC 025 were computed in Microsoft Excel ® . 20,21 Any p-value <0.05 was considered significant.
Sensitivity analyses were performed post-hoc on the primary outcome of ROR in order to account for factors that may have potentially skewed our primary analysis. This included analyzing the ROR in three separate ways: (1) Without cases where non-implicated antipsychotics were used concomitantly, (2) Exclusion of cases where OCD may have been present at baseline, and (3) Exclusion of an antipsychotic with an RRR an order of magnitude greater than any other antipsychotics. These analyses helped to mitigate the risk of underlying disease effects or alternative medication synergistic/antagonistic interactions on the outcome.

| Antipsychotic treatment in OCD
Because antipsychotics have been utilized in the treatment of OCD, or in the treatment of OCD comorbid with other conditions, we simultaneously evaluated the ROR and percent of failures with each primary suspected antipsychotic when used for OCD treatment. Here, the ROR was again evaluated via intra-class analysis including only the primary suspected medication. We queried all ADRs for each antipsychotic of interest within the FAERS system during the same timeframe as primary analysis and subsequently searched for terminology consistent with treatment failure or exacerbation. Therapeutic failure was dictated by the reaction terms "obsessivecompulsive symptom", "obsessive-compulsive disorder", or "drug ineffective", while exacerbation was dictated by the reaction term "condition aggravated", all when use was listed for OCD treatment within FAERS. In order to rule out the effects of other mental illnesses, ROR analyses for treatment failure were run including and excluding comorbidities. These data may be taken as a proxy until more formal studies can be conducted on OCD treatment with antipsychotics.

| Spearman correlation analysis of antipsychotic-OCD mechanism
After RRR analysis was complete, we assessed the potential mechanism behind OCD secondary to antipsychotics using Spearman correlation analysis. Human Ki data for the most relevant receptors (5-HT 1A , 5-HT 2A , and D 2 ) on each antipsychotic of interest was compiled from the Psychoactive Drug Screening Program (PDSP) Ki Database. 22 If multiple Ki values were reported, the values were summed and divided by the total count to create the average Ki. The inverse of Ki (i.e., higher Ki value equals higher affinity) was used for Spearman correlation analysis against the RRR.

| RESULTS
Across all drugs, there were 16,388,063 total reports to the FAERS for suspected ADRs between January 1st, 2010 and December 31st, 2020, and this was used to calculate the N EXP and IC 025 ( Figure 2). In terms of OCD/OCS reported ADRs, a total of 4506 were reported to FAERS during the same evaluated period. Of the reported OCD/OCS ADRs, antipsychotics were implicated in 2629 (58.3%) cases. After including only the primary suspected antipsychotic and excluding duplicates (236 cases), a total of 1454 reported OCD/OCS cases were found (Table 1) and used in the primary analysis. The characteristics of patients with OCD/OCS secondary to antipsychotics are listed in Table 2. When looking at all other reported ADRs for the evaluated antipsychotics after excluding OCD/OCS cases a total of 385,972 ADRs were reported to the FAERS during the same time period, and these were used as noncases. OCD/OCS constituted 0.4% of all antipsychotic reports to FAERS during the evaluated time period. Reports of antipsychotic treatment-emergent OCD/OCS ranged from 0% of the total chlorpromazine reports to 2.1% of the total aripiprazole reports. Interestingly, clozapine was on the lower end with OCD/OCS only making up 0.12% of its total reported ADRs. Four firstgeneration antipsychotics were evaluated, and constituted only 9 (0.59%) of the total primary antipsychotic OCD/OCS cases. Of the second-generation antipsychotics, cariprazine comprised 6 (0.41%) cases, clozapine comprised 72 (5%) cases, while aripiprazole represented 1158 (79.6%) cases.
With the exception of chlorpromazine, loxapine, and perphenazine the IC 025 was greater than 0 for all other antipsychotics, indicating a significant disproportionality signal (Table 1). However, when comparing against all other antipsychotics, the only antipsychotic with a significantly greater ROR for OCD/OCS was aripiprazole, which had a ROR of 23.87 (95% CI: 21.01-27.13; p < 0.0001). Similarly, aripiprazole had the largest RRR for OCD/OCS reports of any antipsychotic evaluated, being an order of magnitude greater than the next nearest antipsychotic, cariprazine. Apart from aripiprazole, the other two D 2 partial agonists, cariprazine and brexpiprazole, as well as lurasidone and asenapine, had the highest RRR for OCD/OCS reports while haloperidol, risperidone, and paliperidone had the lowest ( Table 1). Given that the ROR and RRR for OCD/OCS were much higher for aripiprazole relative to all other antipsychotics, there was concern that it may be skewing or minimizing the differences between the other antipsychotics. For this reason, a sensitivity analysis was completed with aripiprazole excluded (Table 3). Overall, the results were largely similar and consistent with the analysis that included aripiprazole though there were many more antipsychotics that significantly separated from all other antipsychotics based on the ROR for OCD/OCS, with brexpiprazole, cariprazine, clozapine, lurasidone, and olanzapine having a significantly higher ROR for OCD/OCS and haloperidol, paliperidone, and risperidone having a significantly lower ROR for OCD/OCS. Additional sensitivity analyses included the removal of concomitant medications and separately the removal of baseline OCD diagnoses. Importantly, the ROR results in these instances were predominantly in line with the primary analysis wherein these factors were included (Table 3).
We identified a total of 220 reports of primary antipsychotic use for OCD excluding other concomitant suspected products and comorbid conditions (Table 4). Of these cases, 31 (14.1%) were reported to be therapeutic failures. The rates of therapeutic failure did not appear to be different between the various antipsychotics, however some medications were used more commonly than others. When other comorbid conditions were included, a total of 1234 cases of antipsychotic use for OCD were found, and 248 failures were noted (20.1%). Here, therapeutic failure from aripiprazole use was noted 46.2% of the time with a significantly increased ROR of 7.9 for OCD treatment failure relative to all other antipsychotics (Table 4). Of note, paliperidone, quetiapine, and risperidone all had a statistically significantly lower ROR of OCD treatment failure relative to all other antipsychotics, indicating a lesser likelihood of reports for treatment failure.
Because of this large discrepancy with aripiprazole we examined the specific reports of OCD/OCS with aripiprazole and also found an increased propensity for other  reported compulsive/impulse control reactions. OCD/OCS were co-reported with gambling disorder, compulsive shopping, eating disorders, trichotillomania/ dermatillomania, and compulsive sexual behaviors 1090, 592, 367, 133, and 357 times, respectively. Further, to investigate the potential impact of specific psychiatric disorders on the rates of OCD/OCS cases, we evaluated the total number of OCD/OCS cases with that diagnosis (including and excluding comorbid disorders). As depicted in Table 2, including comorbidities, where OCD/OCS was reported with antipsychotic use (only primary suspected agent), depression, followed by bipolar disorder, anxiety, then schizophrenia/schizoaffective disorder were the most common indications (comprising 720 (49.5%), 621 (42.7%), 544 (37.4%), and 335 (23%) comorbidities, respectively). At baseline however, OCD was reported in 104 of the 720 depression cases (14.4%), 93 of the 621 bipolar disorder cases (15%), 104 of the 544 anxiety cases (19.1%), and 48 of the 335 schizophrenia/schizoaffective disorder cases (14.3%). Using the RRR as the correlative value against the inverse Ki affinity values for the evaluated receptors (D 2 , 5-HT 2A , and 5-HT 1A ) and receptor ratios (5-HT 2A /D 2 and 5-HT 1A /D 2 ), Spearman's rank analysis identified a significant strength of association for OCD/OCS reporting with stronger 5HT 1A receptor affinity (r s = 0.77; p = 0.0008) as well as the 5-HT 1A /D 2 receptor ratio (r s = 0.66; p = 0.008). 22 None of the other correlations were significant.

| DISCUSSION
Our analysis of the FAERS reports points to disproportionality safety signals for all examined second-generation antipsychotics in the development of OCD/OCS, with varying signal strengths, based on the IC 025 . By contrast, most first-generation antipsychotics were not found to have a disproportionality safety signal. Along with haloperidol, risperidone and paliperidone were associated with low overall OCD/OCS reporting rates. Although the number of overall reports for brexpiprazole and cariprazine were low, when evaluating the RRR our results demonstrated that all three currently marketed D 2 partial agonists were among the top five agents for OCD/OCS reporting. While clozapine has historically been presumed to be the most likely antipsychotic to cause or exacerbate OCD/OCS, the rates of reporting with clozapine in this study were not dramatically higher than other secondgeneration antipsychotics, being comparable to olanzapine and quetiapine, and were far below that of aripiprazole. Contrary to our findings, a study of 209 Korean patients screened for OCS while on antipsychotics found 44 patients (21.2%) displayed symptoms, most prevalently with clozapine and olanzapine (76.9% and 11.5%, respectively). 23 Exact percentages from specific medications cannot be compared to our study, however, as OCD/OCS incidence rates are not calculable from FAERS data. A potential explanation for the beliefs surrounding clozapine and OCD/OCS is the notoriety effect, wherein one published ADR on a medication leads to accelerated reporting or recognition of the ADR for that medication. However, the notoriety bias did not appear to be a major factor in the present study as it would have been expected to result in higher rates of OCD/OCS reporting for clozapine within FAERS. Thus, the fact that we did not identify clozapine as the antipsychotic most associated with OCD/OCS may, in fact, strengthen the findings of our study.
Compared with all other antipsychotics, only aripiprazole had a significantly higher ROR. However, we found only 5 published case reports representing 6 patients, which identified aripiprazole as the likely cause or exacerbation of OCD/OCS. [24][25][26][27][28] In four cases, including three females and one male, none of whom had a history of OCD, all developed OCS while on 5-15 mg/day of aripiprazole. Two of these four aripiprazole cases described patients with schizophrenia and baseline minor OCD symptoms, one who experienced a worsening of OCS while on aripiprazole 10 mg/day but not while on previously trialed clozapine or risperidone at doses of 400 and 8 mg/day, respectively. 26 Further, when risperidone was re-initiated and aripiprazole was stopped the patient's OCS improved. In the other case the patient suffered intensifying of OCS when aripiprazole was titrated to 30 mg/day that subsequently remitted when the medication was switched to sulpiride. 21 As previously noted, OCD/OCS from aripiprazole was commonly co-reported with gambling disorder, compulsive shopping, hyperphagia, and hypersexuality. In a separate study by Fusaroli et al., impulse control disorders (ICD) from D 2 partial agonists were similarly investigated through the FAERS. 29 Here, the authors found D 2 partial agonists constituted 2708 (23%) of the 11,629 total ICD reports. Aripiprazole constituted 2545 of these reports (94%) with the majority of time the indication for use was a mood disorder, rather than a psychotic disorder. ICD and OCD are often assumed to be highly comorbid due to their similar etiology and symptom profile. Obsessive-Compulsive and Related Disorders (OCRDs) represent a spectrum of disorders with overlapping phenomenology and symptomatology ranging from impulse control and eating disorders to somatoform and neurological disorders such as Tourette's syndrome. OCRD has a prevalence rate of up to 9.5% in the general population. 30 However, 57.6% of those with OCD will additionally meet criteria for another OCRD. 31 One distinguishing feature between ICD, tic disorder, and OCD may be whether the compulsive/impulsive behavior is ego-syntonic or dystonic. While compulsions are performed voluntarily in both OCD and ICD/tic disorders, those with OCD are performing compulsions in response to an ego-dystonic obsession to avoid perceived bad consequences, whereas those with ICD may sometimes, but not always, be performing compulsions to satisfy an urge. Overall, the similarities between these disorders may have led to diagnostic confusion, increasing OCD reporting.
As it relates to schizophrenia, when OCD/OCS have been reported, they occur in 40% of patients preceding and in another 40% following psychosis onset, with the remaining 20% reported concurrently. 16 Additionally, meta-analyses reveal that up to 12%-14% of schizophrenia patients meet DSM-5 criteria for OCD. 16 Multiple hypotheses have been formed surrounding the mechanism of their overlap including neuropsychological impairment, familial aggregation, and antipsychotic use. 16 Because of the high prevalence of OCD/OCS in schizophrenia patients before initiation of antipsychotics and the current gap in literature surrounding the oftentimes multifactorial pathogenesis, correlation versus causation is difficult to discern. The overlap between OCD/OCS in conjunction with bipolar disorder is much more apparent, as the prevalence rate of comorbid OCD in this population ranges from 11.1%-21%. 16,32 As with schizophrenia, OCD may also be considered as a risk factor for bipolar disorder. The comorbidity of OCD with anxiety and depression is commonly described as a causal relationship, as one serves as a risk factor for the other. 33 One study found 53% of patients with depression reported ≥1 OCS, while 14% reported ≥4 OCS, with higher rates of OCS appearing correlative to depression severity. 34 In comparison to these reported percentages, our study found that of the OCD/OCS ADR reports to FAERS, 42.4% had comorbid depression, 36.3% had comorbid bipolar disorder, and 19.7% had comorbid schizophrenia. Differences in our percentages may be accounted for by OCS not being reported within studies, or by antipsychotic treatment. We cannot fully rule out the possibility of disease causality, although these disease correlations also do not fully explain differences in rates of OCD/OCS reporting with individual antipsychotics. However, because aripiprazole may be utilized more commonly than other antipsychotics as an adjunctive treatment for depression or anxiety this may have skewed reporting.
While debated, these data taken together do appear to align with current understanding of OCD pathophysiology. Dopamine hyperactivity or serotonergic underactivity are thought to elicit or exacerbate OCD symptoms. Although outside the scope of this paper, dopaminergic agonists (e.g., pramipexole and ropinirole) constituted a substantial number of OCD/OCS reports to FAERS. Dopaminergic partial agonists may increase dopamine signaling within the caudate and putamen, while serotonin-2A (5-HT 2A ) and 5-HT 2C antagonists may indirectly increase dopamine in the cortex and striatum through an inhibition of GABA release. 35 Thus, for antipsychotics where 5-HT 2A affinity is greater than D 2 affinity (i.e., most secondgeneration antipsychotics) the risk of OCD/OCS development may be greater. Conversely, D 2 antagonism may mitigate OCD symptoms, hence the lower reporting of OCD/OCS with all of the first-generation antipsychotics as well as the second-generation antipsychotics risperidone and paliperidone, which have high affinity for the D 2 receptor. 36 Further demonstrating the influence of 5-HT 2A receptors on OCD, psilocybin and lysergic acid diethylamide (LSD), both 5-HT 2A receptor agonists, have shown benefit for reducing compulsive-type behaviors in both mice and humans. [37][38][39] Separately, in a recent pharmacovigilance-pharmacodynamic study of FAERS, authors found significant positive associations between D 3 -receptor agonism and impulsivity which may have relevance to the present study. 40 Alternatively, serotonin-1A (5-HT 1A ) partial agonism has been postulated as causative of OCD, through possible dopaminergic increases in the raphe nucleus and cortex. 40,41 Indeed, many antipsychotics are also partial agonists at the 5-HT 1A receptor. Alterations in the glutamatergic systems may also play a role in OCD, given the profound efficacy of memantine in some preliminary data. 42 In total, a possible scenario outside of D 2 -receptor partial agonism is multiple receptors acting to counterbalance one another, possibly relating to 5-HT 1A /D 2 or 5-HT 2A /D 2 ratios. Spearman's correlation analysis (Table 5) indicated that higher affinity for the 5-HT 1A receptor was associated with greater rates of OCD/OCS reporting. While we cannot discern if this association is secondary to partial agonistic effects for this receptor, it is worth noting that all of the antipsychotic agents with the highest affinity at the 5-HT1A receptor are also partial agonists at this receptor (Table 5). Further, while the D 2 receptor was not significantly correlated with OCD/OCS risk, we cannot rule out the influence of D 2 receptor partial agonism since all three D 2 receptor partial agonists (aripiprazole, brexpiprazole, and cariprazine) were among the top four agents associated with OCD/OCS reporting based on the RRR (Table 5). Still, our finding of a significant association between the 5-HT 1A /D 2 receptor ratio for individual antipsychotics and their RRR for OCD/OCS with agents having higher affinity for D 2 receptors relative to 5-HT 1A receptors being associated with lower RRR may support the previous rationale that greater D 2 antagonism may offset the risk of OCD/OCS from 5-HT 1A agonism/partial agonism.
Selective serotonin reuptake inhibitors and clomipramine have commonly been the mainstays of treatment for OCD. Benefit from these medications again supports the dopamine-serotonin imbalance theory, as increases in serotonin may decrease dopamine hyperactivity. However, between 25% and 60% of patients with OCD have residual symptoms with first-line treatment options. 42,43 Antipsychotics have been trialed as augmenting agents for those failing serotonergic agonist treatment, with varying success. The FGAs haloperidol and pimozide have been shown to be beneficial for refractory OCD when used in combination with an antidepressant. 44,45 A 2006 meta-analysis by Bloch et al. found the number needed to treat with antipsychotic augmentation for OCD patients failing 12 weeks of maximal therapy with a serotonergic agent was 4.5 (95% CI: 3.2-7.7), but nearly one-third of patients still exhibited refractory symptoms. 46 While aripiprazole was not included in this analysis, risperidone and haloperidol appeared efficacious, but olanzapine and quetiapine did not. 46 In a separate metaanalysis, Dold et al. found risperidone to be the treatment of choice for antipsychotic augmentation in OCD. 47 In more recent meta-analyses, aripiprazole has also been shown to have significant impact on OCD symptoms. 12,48,49 Of note, Zhou et al. found comorbid depression to be a possible predictor of antipsychotic failure in their network meta-analysis. 49 Perhaps in depressed patients, where there is not elevated striatal dopamine, using aripiprazole can increase dopamine signaling, inducing OCS, but in patients with schizophrenia where there is already elevated striatal dopamine the propensity is less likely.
We only found minimal data comparing antipsychotics to one-another for OCD treatment. Comparative studies on aripiprazole versus other antipsychotics have found mixed results. One trial found olanzapine was as effective as aripiprazole for OCD treatment, while another found quetiapine to be as effective. 50,51 Only a 2011 study by Selvi et al. found risperidone to be more effective than aripiprazole for SSRI-refractory OCD. 52 Unfortunately, most other data compare antipsychotics to placebo, therefore more head-to-head studies are required to determine efficacy from individual antipsychotics for treatment of OCD. Ultimately, in patients with primary OCD, it is highly conceivable that strong blockade of dopamine D 2 receptors with an antipsychotic could mitigate symptoms. This is in line with our finding of lower rates of antipsychotic therapeutic failure reporting for FGAs and high D 2 affinity SGAs (i.e., risperidone and paliperidone) when used for primary OCD treatment T A B L E 5 Antipsychotic receptor profiles with corresponding OCD reporting as arranged by 5-HT 1A /D 2 receptor affinity (Ki ( Table 4). Indeed, a meta-regression analysis found higher affinity for the D 2/3 receptor was more highly associated with treatment success. 53 Paradoxically, those without OCD may experience de novo or exacerbated OCD/OCS from antipsychotics. However, these data do not account for comorbid conditions. In those with comorbid conditions, our data showed that OCD treatment failure reporting was most prevalent with aripiprazole. This appears in direct contrast to published reports where aripiprazole was successful in the treatment of OCS comorbid with schizophrenia. 54,55 While publication bias may be an explanation for this discrepancy it cannot be discounted that our findings might reflect the high rates of aripiprazole prescribing overall, especially in depression, which encompassed a larger proportion of comorbidities present in the reports for OCD than did schizophrenia. Even fewer data exist for treatment of OCD secondary to antipsychotics. A case series of 7 patients with clozapine-induced OCS were treated with a mean dose of about 23 mg aripiprazole for 9.7 weeks with significant decreases in Yale Brown Obsessive Compulsive Rating Scale (YBOCS) scores. 56 Similarly, aripiprazole 15 mg was shown to be of benefit in reducing YBOCS scores in a female with schizophrenia with OCS induced by olanzapine. It has been hypothesized that low antipsychotic doses, where 5-HT 2A receptor antagonism may outweigh D 2 receptor antagonism, may precipitate OCD. 36 Thus, one proposed strategy has been to increase the dose of the antipsychotic, possibly by shifting the balance in favor of D 2 antagonism rather than 5-HT 1A/2A . 57 Alternatively, addition of an SSRI may be of benefit. 58

| Limitations
Disproportionality studies have inherent issues, including reporting biases or under-reporting. A masking-effect may have resulted in specific antipsychotics not demonstrating a significant ROR for OCD/OCS due to over-reporting of other non-OCD/OCS ADRs. 59 We cannot comment on the prevalence of de novo OCD or OCS exacerbation, nor can we comment on the severity of de-novo or exacerbated OCS as the data provided via the FAERS Public Dashboard is categorical in nature and does not report on any formal rating scales, if such scales were even used. While the majority of the reports were from consumers and not healthcare practitioners we do not believe this strongly influenced the outcome of the study, as aripiprazole still constituted the largest number of antipsychotic reports, even from healthcare practitioners. However, we cannot fully discount that a misclassification of ICD as OCD may have occurred through consumer unawareness of the nuances between these disorders. We did not evaluate the influence of multiple medications on OCD/OCS, in favor of solely investigating the primary antipsychotic implicated. As noted previously, interactions between antipsychotics may lessen or increase OCD/OCS rates, thus further research is needed. However, it is worth noting that when concomitant medications were removed in a sensitivity analysis, the ROR findings were largely in keeping with the results from the primary analysis (Table 3) suggesting that use of multiple medications was unlikely to have confounded our primary findings. While doses and treatment durations are reported to FAERS within the quarterly data, as we assessed data from the Public Dashboard which does not include these parameters we were not able to assess the potential relationships between these factors and the reporting of antipsychotic treatmentemergent OCD/OCS.
In this study we did not conduct sensitivity analyses on the influence of the underlying disease on antipsychotic-emergent OCD/OCS (save the influence of baseline OCD/OCS). Rather, we chose to broadly assess the rates of OCD/OCS reporting based on all indications combined in order to maximize the power of our study to detect disproportionality signals. Furthermore, sensitivity analyses would likely be more informative if conducted in a longitudinal correlative study rather than voluntarily submitted pharmacovigilance data. It cannot be disregarded that an element of unreported baseline OCD/OCS or reporter error of misclassifying side effect with indication may have skewed our findings. Another potential confound is the relative rates that specific antipsychotic agents were prescribed for OCD/OCS. However, it should be noted that the number of FAERS reports of risperidone used for treatment of OCD/OCS is very similar to the number of reports of aripiprazole for this indication despite very divergent RORs of OCD treatment failure for these two agents (Table 4) suggesting that these confounds are less likely to have influenced these findings. Lastly, while certain illicit substances, such as cocaine or methamphetamine, can induce OCD/OCS and are more commonly abused by those with schizophrenia, bipolar disorder, or depression, use of these substances are not communicated to FAERS, so we cannot rule these out as the cause for some reports. However, we have no reason to believe that any potential effect of concomitant stimulant abuse would not be evenly distributed across all antipsychotics examined in this study, and thus, we would not expect this to significantly impact our findings.
To conclude, our findings are likely to be of critical importance for not only furthering the understanding of OCD pathophysiology, but enhancing patient and clinician awareness of OCD/OCS secondary to antipsychotic treatment. Our analyses identify D 2 partial agonism, 5-HT 1A potency and partial agonism, and potency of D 2 receptor blockade relative to 5-HT 1A blockade as key features associated with increased reporting of antipsychoticinduced or exacerbated OCD/OCS. Contrary to previously published reports showing greater numbers of de novo or exacerbated OCD/OCS with clozapine and olanzapine, we did not find a substantially increased reporting of OCD/OCS with these agents relative to other SGAs. Rather, aripiprazole and other D 2 receptor partial agonists had the most reports of de novo OCD/OCS. These findings from FAERS will hopefully lay the groundwork for future studies on this topic and should ideally be validated through prospective research studies involving direct comparisons of antipsychotic agents.

FUNDING INFORMATION
This work (research, review, or writing) required no funding of any type from any agency in the public, commercial, or not-for-profit sectors.