SSRI versus SNRI Initiation and Incident Bipolar Disorder in Tertiary Psychiatric Care: An Active-Comparator Cohort Study from the United Arab Emirates

SSRI versus SNRI Initiation and Incident Bipolar Disorder in Tertiary Psychiatric Care: An Active-Comparator Cohort Study from the United Arab Emirates | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article SSRI versus SNRI Initiation and Incident Bipolar Disorder in Tertiary Psychiatric Care: An Active-Comparator Cohort Study from the United Arab Emirates Syed Ali Bokhari, Abdulhameed Omer Elnour, Muhanad Elnoor This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8580658/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 28 Apr, 2026 Read the published version in International Journal of Bipolar Disorders → Version 1 posted 8 You are reading this latest preprint version Abstract Background: Whether antidepressant class influences the risk of subsequent bipolar disorder diagnosis remains clinically relevant, though direct active-comparator studies are scarce. We compared incident bipolar disorder risk between selective serotonin reuptake inhibitor (SSRI) and serotonin-norepinephrine reuptake inhibitor (SNRI) initiators using tiered outcome definitions. Methods: We conducted a retrospective cohort study from 2018 to 2025 using a 90-day landmark design at a tertiary psychiatric hospital in the United Arab Emirates. Adults aged 18–60 years initiating SSRIs or SNRIs for depressive disorders were followed for incident bipolar disorder. The primary outcome was anchored bipolar disorder, defined as two or more diagnoses at least 30 days apart plus initiation of a mood stabiliser. Secondary outcomes included confirmed bipolar disorder (two or more diagnoses) and any bipolar diagnosis. Cox proportional hazards models adjusted for age, sex, schizophrenia spectrum disorder, substance use disorder, and prior psychotropic use. Exploratory analyses examined individual antidepressants. Results: Among 1,095 antidepressant initiators (818 SSRI; 277 SNRI), 72 (6.6%) developed anchored bipolar disorder over 1,610.6 person-years. SNRI versus SSRI initiation was not associated with increased risk of anchored bipolar disorder (adjusted hazard ratio [aHR] 1.07, 95% CI 0.63–1.84, p = 0.80). Findings were consistent across secondary outcomes (confirmed: aHR 1.24, 95% CI 0.96–1.60; any diagnosis: aHR 1.12, 95% CI 0.92–1.37) and drug-level comparisons, including venlafaxine versus pooled SSRIs (aHR 1.27, 95% CI 0.62–2.60). Event rates varied seven-fold by outcome stringency (6.6% to 46.3%). Conclusions: Antidepressant class was not associated with incident bipolar disorder using stringent outcome definitions. The marked variation in event rates across outcome tiers suggests that high conversion rates reported in prior literature may partly reflect diagnostic revision rather than pharmacological effects. Bipolar disorder antidepressants selective serotonin reuptake inhibitor serotonin-norepinephrine reuptake inhibitor pharmacoepidemiology cohort study diagnostic conversion United Arab Emirates Figures Figure 1 Introduction Risk of treatment-emergent mood switch and diagnostic conversion from unipolar depression to bipolar disorder has been widely discussed in psychiatry (1). Bipolar disorder represents a significant diagnostic challenge, with approximately half of initial presentations occurring as depressive episodes rather than mania or hypomania (2,3). This characteristic presentation pattern frequently results in initial diagnoses of major depressive disorder, with subsequent diagnostic conversion to bipolar disorder occurring as the illness course unfolds. Prospective registry studies have documented cumulative conversion rates from unipolar depression to bipolar disorder ranging from 6.5% to 11.1% over observation periods of 10 to 15 years (4–6). The highest conversion rates typically occur within the first year following initial depression diagnosis, with the rate declining to approximately 0.8% to 1.0% annually thereafter (7). Older antidepressants, especially tricyclic antidepressants, were long implicated in inducing mania (Tondo et al., 2010), but modern agents such as SSRIs and SNRIs are now most commonly prescribed for depressive episodes, and their relative propensity to unmask bipolarity remains unclear (8,9). The question of whether antidepressant treatment influences the risk of subsequent bipolar diagnosis has generated substantial controversy. Three competing hypotheses exist: first, that antidepressants may precipitate manic or hypomanic episodes in individuals with latent bipolar vulnerability, thereby unmasking the condition; second, that observed associations reflect confounding by indication, whereby individuals with unrecognised bipolar depression are more likely to receive antidepressant treatment and subsequently manifest overt bipolarity; and third, that apparent associations arise from diagnostic revision processes in psychiatric care, particularly in specialty settings where sustained clinical attention increases the probability of detecting mood instability (1,10). Experts have often cautioned that SNRIs, notably venlafaxine, may carry higher switch risk than SSRIs (11). The seminal work by Post et al. (2006), comparing venlafaxine, sertraline, and bupropion as adjunctive treatments in bipolar depression, reported significantly elevated switch rates with venlafaxine (29%) compared with sertraline (9%) and bupropion (10%). The most recent network meta-analysis by Oliva et al. (2025) confirmed that whilst no individual antidepressant demonstrated statistically significantly increased switch risk compared with placebo, venlafaxine consistently showed the highest relative risk estimates across sensitivity analyses (8). The 2018 Canadian Network for Mood and Anxiety Treatments and International Society for Bipolar Disorders (CANMAT/ISBD) guidelines accordingly recommend that SSRIs (particularly sertraline) and bupropion be preferred over SNRIs when antidepressants are considered for bipolar depression (12,13). However, observational evidence comparing SSRI and SNRI classes in depression cohorts is inconsistent. Large electronic health record-based cohorts have found that prior antidepressant exposure in general is associated with increased rates of new mania or bipolar diagnosis, but direct comparisons between SSRI and SNRI classes have been limited. Patel et al. (2015), in a South London and Maudsley NHS Trust cohort study, reported that both SSRI (HR 1.34, 95% CI 1.18 to 1.52) and venlafaxine exposure (HR 1.35, 95% CI 1.07 to 1.70) were associated with subsequent mania or bipolar disorder diagnosis compared with no antidepressant exposure (14). A nationwide Korean register study by Kim et al. (2020) found that the conversion rate was dramatically higher among patients prescribed SNRIs, with SNRI users having more than double the risk versus no antidepressant (15). Conversely, other studies suggest that class differences may be less pronounced than hypothesised. Pradier et al. (2021), using multi-site US health records (n = 67,807), found that short-term conversion rates within 3 months were numerically higher for SNRIs (1.93%) than SSRIs (1.23%), although this was a preliminary finding in a predictive modelling context (16). In contrast, a recent multi-centre South Korean analysis by Yoo et al. (2025) reported no significant difference between SSRI and SNRI initiators in conversion to bipolar disorder using rigorous new-user design methodology. Similarly, Zhu et al. (2025) reported from a Chinese inpatient cohort that patients who eventually converted to bipolar had commonly received either an SNRI (venlafaxine) or an SSRI (paroxetine), without clear class differentiation (17). Thus, evidence is inconsistent and may depend on population (psychiatric specialty versus general practice), follow-up period, and how outcomes are defined (narrow mania or hypomania diagnoses versus broader bipolar disorder codes). The methodological quality of observational studies examining antidepressant effects on bipolar conversion has been variable. Two design features merit particular attention. First, confounding by indication represents a fundamental challenge: individuals prescribed antidepressants differ systematically from those not treated, and those prescribed SNRIs may differ from those prescribed SSRIs in ways that independently influence bipolar risk (18). Active-comparator designs, which compare initiators of one treatment with initiators of an alternative treatment for the same indication rather than with untreated individuals, substantially attenuate this bias by ensuring that all included patients have crossed the threshold for treatment initiation (19,20). Second, immortal time bias can substantially distort effect estimates in pharmacoepidemiological studies. This bias arises when follow-up time during which the outcome cannot occur is misattributed between comparison groups (21). The landmark analysis approach, which defines a fixed time point at which exposure is assigned and from which follow-up begins, represents an established solution to this problem (22). A third consideration specific to psychiatric outcomes concerns the definition and validation of the outcome itself. In specialty psychiatric settings where patients receive ongoing diagnostic assessment, the detection of any single bipolar diagnosis may reflect provisional clinical impressions, diagnostic speculation, or billing necessities rather than definitive diagnostic conclusions. The use of stringent, anchored outcome definitions that require diagnostic persistence and treatment concordance may more accurately identify true diagnostic conversions. The present study addresses these methodological considerations by employing an active-comparator landmark cohort design to compare SSRI versus SNRI initiators for incident bipolar disorder in a specialty psychiatric setting. We implemented a tiered outcome hierarchy to evaluate how outcome specificity influences observed event rates and treatment associations. Additionally, we conducted exploratory drug-level analyses to examine whether within-class heterogeneity, particularly regarding venlafaxine, might be obscured by class-level comparisons. Based on prior literature and pharmacological reasoning, we hypothesised that if antidepressant class itself confers differential risk, SNRIs might show higher rates of bipolar conversion than SSRIs. We also considered the possibility that observed associations could reflect confounding by indication or disease severity rather than true pharmacological effects. Methods Study Design and Setting We conducted a retrospective cohort study using electronic health records from a tertiary psychiatric hospital in the United Arab Emirates (1 January 2018 to 22 December 2025). We employed an active-comparator design among antidepressant initiators with a 90-day landmark to minimise immortal time bias. The study was approved by the institutional ethics committee. Data Source Data were extracted from the General Adult Psychiatry service, which provides outpatient and inpatient care for patients aged 18 to 60 years. The database includes demographic information, encounter records with International Classification of Diseases, 10th Revision (ICD-10) diagnoses, and medication dispensing records with dose information. Study Population The index condition was a new depression diagnosis, defined as the first encounter with any qualifying ICD-10 code: F32 (depressive episode), F33 (recurrent depressive disorder), F34.1 (dysthymia), or F41.2 (mixed anxiety and depressive disorder). Inclusion criteria required age 18 to 60 years at index, at least one encounter after the 90-day landmark, and first SSRI or SNRI dispensed within the exposure window (index date to landmark). Patients with bipolar disorder diagnosis (F30 or F31) documented before the landmark were excluded. Exposure Definition Patients were classified according to their first antidepressant dispensed within the 90-day exposure window. SSRIs included sertraline, escitalopram, fluoxetine, paroxetine, citalopram, and fluvoxamine. SNRIs included venlafaxine, duloxetine, and desvenlafaxine. Exposure was assigned at the 90-day landmark based on the first antidepressant dispensed within the index-to-landmark window and remained fixed throughout follow-up (intention-to-treat). This approach captures the clinical decision at treatment initiation and avoids informative censoring from treatment changes. Outcome Definitions We employed a tiered hierarchy to assess outcome specificity. The primary outcome, Tier 1 (anchored) bipolar disorder, required two or more bipolar diagnoses (F30 or F31) occurring 30 or more days apart, with the first occurring after the landmark, plus new initiation of a mood stabiliser (lithium, valproate, carbamazepine, or lamotrigine) within 0 to 90 days after the first qualifying bipolar diagnosis, with no mood stabiliser dispensing before the index date. The Tier 2 (confirmed) outcome required two or more bipolar diagnoses 30 or more days apart. The Tier 3 (any) outcome required one or more bipolar diagnoses. Follow-up began at the 90-day landmark and ended at outcome occurrence, last encounter, or study end. Covariates Baseline characteristics were assessed from 365 days before index through the index date. Covariates included age (continuous), sex, schizophrenia spectrum disorders (F20 to F29), substance use disorder (F10 to F19), prior antipsychotic dispensing, and prior SSRI or SNRI dispensing in the 365 days before index. A baseline bipolar proxy variable, defined as any recorded bipolar spectrum diagnosis in the 365 days before index, was used in prespecified sensitivity analyses. Statistical Analysis The primary analysis employed Cox proportional hazards regression comparing hazards between SSRI and SNRI initiators, adjusting for all covariates. The proportional hazards assumption was tested using Schoenfeld residuals. Sensitivity analyses included 180-day landmark, exclusion of baseline schizophrenia spectrum diagnoses, exclusion of baseline bipolar proxy diagnoses, and restriction to a washout cohort with no SSRI or SNRI use in the 365 days before index. Exploratory drug-level analyses examined individual antidepressants to assess within-class heterogeneity. Prespecified comparisons included venlafaxine versus sertraline, venlafaxine versus escitalopram, venlafaxine versus pooled SSRIs, duloxetine versus pooled SSRIs, and desvenlafaxine versus pooled SSRIs. A within-SSRI comparison of sertraline versus escitalopram served as a negative control. These analyses were considered feasible if sample size exceeded 80 per arm or combined Tier 1 events exceeded 20. P-values were not adjusted for multiple comparisons, and these results are interpreted as exploratory. Complete case analysis was used; all key analytic variables had no missing data among the 1,095 patients in the final cohort. As a sensitivity analysis, we estimated propensity scores for SSRI versus SNRI assignment using logistic regression with all covariates and applied stabilised inverse probability of treatment weights (IPTW). Weights were truncated at the 1st and 99th percentiles. Covariate balance was assessed using standardised mean differences before and after weighting. Analyses were conducted through Python version 3.9. Statistical significance was defined as alpha = 0.05 for the primary outcome; secondary and exploratory analyses were hypothesis-generating. Results Study Population From 14,583 General Adult Psychiatry patients, 4,166 had qualifying depression diagnoses. After applying the 90-day landmark and exposure criteria, 1,095 antidepressant initiators comprised the final analytic cohort. SSRI initiators (n = 818, 74.7%) and SNRI initiators (n = 277, 25.3%) differed in mean age (34.4 versus 37.7 years, standardised mean difference [SMD] = 0.31) and prevalence of baseline schizophrenia spectrum diagnoses (35.8% versus 44.0%, SMD = 0.17). Prior antidepressant use in the year before index was common in both groups (52.8% versus 58.1%, SMD = 0.11). Median follow-up was 0.90 years (interquartile range 0.31 to 2.21), with total follow-up of 1,610.6 person-years. Baseline characteristics are presented in Table 1 . Table 1 . Baseline Characteristics of the Study Cohort. Characteristic SSRI (N = 818) SNRI (N = 277) SMD Age, years, mean (SD) 34.4 (11.1) 37.7 (10.2) 0.31 Female sex, n (%) 371 (45.4) 111 (40.1) 0.11 Schizophrenia spectrum, n (%) 293 (35.8) 122 (44.0) 0.17 Substance use disorder, n (%) 219 (26.8) 92 (33.2) 0.14 Anxiety disorder, n (%) 161 (19.7) 52 (18.8) 0.02 Bipolar disorder markers, n (%) 74 (9.0) 26 (9.4) 0.01 Prior antipsychotic use, n (%) 345 (42.2) 130 (46.9) 0.10 Prior SSRI/SNRI use (365 days), n (%) 432 (52.8) 161 (58.1) 0.11 Follow-up, years, mean (SD) 1.4 (1.4) 1.7 (1.7) 0.17 Note. SMD = standardised mean difference. Values above 0.10 suggest meaningful imbalance. Baseline and prior medication covariates were assessed in the 365 days before index. Primary Outcome During follow-up, 72 patients (6.6%) developed Tier 1 anchored bipolar disorder, with an incidence rate of 4.47 per 100 person-years. SSRI versus SNRI initiation was not associated with this outcome in adjusted analyses (aHR 1.07, 95% CI 0.63 to 1.84, p = 0.80; Table 2 ). In the full model, younger age was associated with increased risk (HR 0.97 per year, 95% CI 0.95 to 0.99, p = 0.016), and prior antidepressant use was associated with reduced risk of anchored bipolar disorder (HR 0.42, 95% CI 0.23 to 0.76, p = 0.004). Cumulative incidence of anchored bipolar disorder was 1.3% at 1 year, 4.0% at 2 years, 6.0% at 3 years, and 11.8% at 5 years. Secondary Outcomes For Tier 2 confirmed bipolar disorder (328 events, 30.0%), no significant difference was observed between groups (aHR 1.24, 95% CI 0.96 to 1.60, p = 0.095). For Tier 3 any bipolar diagnosis (507 events, 46.3%), results were similarly null (aHR 1.13, 95% CI 0.92 to 1.38, p = 0.24). The proportional hazards assumption was satisfied for all models (Schoenfeld test p > 0.05). Table 2. Primary and Secondary Outcomes: SSRI versus SNRI Initiation and Incident Bipolar Disorder. Outcome N Events aHR 95% CI p Tier 1: Anchored (Primary) 1,095 72 1.07 0.63 to 1.84 0.80 Tier 2: Confirmed 1,095 328 1.24 0.96 to 1.60 0.10 Tier 3: Any diagnosis 1,095 507 1.13 0.92 to 1.38 0.24 Note. aHR = adjusted hazard ratio; CI = confidence interval. Hazard ratios are adjusted for age, sex, baseline schizophrenia spectrum diagnosis, baseline substance use disorder, prior antipsychotic use, and prior SSRI or SNRI use in the 365 days before index. Reference group is SNRI initiators. Kaplan-Meier survival curves for anchored bipolar disorder (Tier 1) and any bipolar diagnosis (Tier 3) are presented in Figure 1 . The curves demonstrated overlapping survival probabilities across antidepressant classes for both outcome tiers, consistent with the null hazard ratio findings. Sensitivity Analyses Results were consistent across all prespecified sensitivity analyses ( Table 3 ). In the washout-restricted cohort (n = 502) without SSRI or SNRI exposure in the prior year, the Tier 1 aHR was 0.91 (95% CI 0.46 to 1.81, p = 0.79). Excluding patients with baseline schizophrenia spectrum diagnoses (n = 680) yielded a Tier 1 aHR of 0.74 (95% CI 0.39 to 1.41, p = 0.36). Excluding patients with baseline bipolar proxy diagnoses (n = 995) produced a Tier 1 aHR of 0.96 (95% CI 0.53 to 1.71, p = 0.88). The 180-day landmark analysis (n = 1,027) showed a Tier 1 aHR of 1.44 (95% CI 0.78 to 2.67, p = 0.24). No significant interactions were observed between antidepressant class and baseline schizophrenia spectrum or bipolar markers for any outcome tier. Table 3. Sensitivity Analyses: SSRI versus SNRI Initiation Across Alternative Cohort Definitions. Analysis Tier 1 aHR (95% CI) p Main Analysis (N = 1,095) 1.07 (0.63 to 1.84) 0.80 Washout-Restricted (N = 502) 0.91 (0.46 to 1.81) 0.79 Excluding Schizophrenia (N = 680) 0.74 (0.39 to 1.41) 0.36 Excluding Bipolar Markers (N = 995) 0.96 (0.53 to 1.71) 0.88 180-Day Landmark (N = 1,027) 1.44 (0.78 to 2.67) 0.24 Note. aHR = adjusted hazard ratio; CI = confidence interval. All models adjusted for age, sex, baseline schizophrenia spectrum diagnosis, baseline substance use disorder, prior antipsychotic use, and prior SSRI/SNRI use (except washout-restricted analysis which excludes prior SSRI/SNRI covariate). Propensity Score-Weighted Analyses IPTW achieved excellent covariate balance, with all standardised mean differences reduced from 0.10–0.31 to less than 0.04 (Supplementary Table S1). In IPTW-weighted analyses, Tier 1 anchored bipolar disorder remained non-significant (HR 1.02, 95% CI 0.87–1.21, p = 0.78), consistent with covariate-adjusted estimates. For secondary outcomes, IPTW analyses showed statistically significant associations for Tier 2 (HR 1.23, 95% CI 1.13–1.33, p < 0.001) and Tier 3 (HR 1.11, 95% CI 1.04–1.18, p = 0.002), although effect sizes remained modest (Supplementary Table S2). This divergence between primary and secondary outcomes in fully balanced analyses further supports the interpretation that less specific outcome definitions capture diagnostic processes distinct from clinically meaningful bipolar conversion. Drug-Level Exploratory Analyses Among individual antidepressants, escitalopram (n = 373, 34.1%) and sertraline (n = 199, 18.2%) were the most commonly prescribed SSRIs, whilst venlafaxine (n = 106, 9.7%) was the most common SNRI. Drug-level comparisons for the primary outcome (Tier 1 anchored bipolar disorder) are presented in Table 4. No individual drug comparison showed a statistically significant association with anchored bipolar disorder. Venlafaxine versus pooled SSRIs yielded an aHR of 1.27 (95% CI 0.62 to 2.60, p = 0.52). Direct comparisons of venlafaxine with sertraline (aHR 1.69, 95% CI 0.61 to 4.70, p = 0.32) and escitalopram (aHR 1.17, 95% CI 0.54 to 2.51, p = 0.69) were similarly null. Duloxetine versus pooled SSRIs (aHR 0.87, 95% CI 0.31 to 2.43, p = 0.79) and desvenlafaxine versus pooled SSRIs (aHR 0.82, 95% CI 0.35 to 1.90, p = 0.64) showed no association with the primary outcome. For secondary outcomes, one notable finding emerged: sertraline showed significantly lower risk than escitalopram for Tier 2 confirmed bipolar disorder (aHR 0.60, 95% CI 0.42 to 0.88, p = 0.009) and Tier 3 any bipolar diagnosis (aHR 0.69, 95% CI 0.53 to 0.92, p = 0.01). This within-SSRI difference was not hypothesised a priori and should be interpreted as hypothesis-generating. Sensitivity analyses for drug-level comparisons showed consistent null findings for the primary outcome across all cohort restrictions. Table 4. Drug-Level Exploratory Analyses: Individual Antidepressant Comparisons for Tier 1 Anchored Bipolar Disorder. Comparison Events aHR (95% CI) p Venlafaxine vs Sertraline 9 / 8 1.69 (0.61 to 4.70) 0.32 Venlafaxine vs Escitalopram 9 / 29 1.17 (0.54 to 2.51) 0.69 Venlafaxine vs Pooled SSRIs 9 / 53 1.27 (0.62 to 2.60) 0.52 Duloxetine vs Pooled SSRIs 4 / 53 0.87 (0.31 to 2.43) 0.79 Desvenlafaxine vs Pooled SSRIs 6 / 53 0.82 (0.35 to 1.90) 0.64 Sertraline vs Escitalopram (within-SSRI) 8 / 29 0.73 (0.33 to 1.64) 0.45 Note. aHR = adjusted hazard ratio; CI = confidence interval. Events column shows exposed/comparator. All models adjusted for age, sex, baseline schizophrenia spectrum diagnosis, baseline substance use disorder, prior antipsychotic use, and prior SSRI/SNRI use. P-values are not adjusted for multiple comparisons; results are exploratory. Discussion In this large psychiatric cohort using an active-comparator landmark design, we found no evidence that initiating an SSRI rather than an SNRI altered the long-term incidence of new bipolar diagnoses. Across multiple outcome definitions, drug-level analyses, and adjusted models, the hazard of bipolar conversion was essentially identical for SSRIs and SNRIs. This null finding was consistent across three hierarchically defined outcome tiers, multiple sensitivity analyses, and exploratory drug-level analyses including specific examination of venlafaxine. Importantly, our results demonstrate that outcome definition exerts profound influence on observed event rates in specialty psychiatric settings, with anchored bipolar disorder occurring in 6.6% of patients compared with 46.3% for any bipolar diagnosis over the same follow-up period. Comparison with Prior Literature Our findings align closely with the recent multi-site study by Yoo et al. (2025), who also observed no class difference in bipolar conversion risk when comparing SSRI and SNRI initiators using rigorous new-user design methodology across multiple Korean centres (23). This convergence of null findings from independent cohorts employing similar methodological approaches strengthens the case that class-level differences may not exist in populations initially treated for presumed unipolar depression. Our null findings contrast with some earlier signals in the literature. The seminal work by Post et al. (2006) reported switch rates of 29% for venlafaxine compared with 9% for sertraline and 10% for bupropion in patients with established bipolar disorder receiving adjunctive antidepressant treatment (11). However, this population (established bipolar disorder) differs fundamentally from ours (new depression without prior bipolar diagnosis), representing a higher-risk stratum. Patel et al. (2015) found that both SSRI (HR 1.34, 95% CI 1.18 to 1.52) and venlafaxine exposure (HR 1.35, 95% CI 1.07 to 1.70) were associated with subsequent mania or bipolar disorder diagnosis, but did not directly compare SSRI versus SNRI classes. Kim et al. (2020), in a nationwide Korean register, reported that the conversion rate was dramatically higher among patients prescribed SNRIs, with SNRI users having more than double the risk versus no antidepressant (15). However, comparison against no antidepressant, rather than an active comparator, may reflect confounding by indication. Pradier et al. (2021), using multi-site US health records (n = 67,807), found that short-term conversion rates within 3 months were numerically higher for SNRIs (1.93%) than SSRIs (1.23%) (16). This timeframe differs substantially from our longer follow-up, and the difference was not the primary focus of their predictive modelling study. Zhu et al. (2025) reported from a Chinese inpatient cohort that patients who eventually converted to bipolar had commonly received either venlafaxine or paroxetine without clear class differentiation, consistent with our findings (17). The most recent network meta-analysis by Oliva et al. (2025) evaluated switch to mania across randomised controlled trials of acute antidepressant treatment in bipolar depression. Whilst no individual antidepressant demonstrated statistically significantly elevated switch risk compared with placebo, venlafaxine consistently showed the highest relative risk estimates (8). Our exploratory drug-level analyses specifically examined venlafaxine and found no excess risk compared with pooled SSRIs (aHR 1.27, 95% CI 0.62 to 2.60) or individual SSRIs. This null finding persisted across sensitivity analyses, suggesting that when patients are initially treated for presumed unipolar depression with stringent outcome definitions, venlafaxine may not confer the differential risk observed in established bipolar populations. Interpretation of Tiered Outcome Findings Perhaps the most striking finding of this study concerns the seven-fold difference in event rates between anchored (6.6%) and any (46.3%) bipolar diagnosis outcomes. This pattern suggests diagnostic reassignment or clarification over time, rather than direct pharmacological induction of mania (24). In other words, many patients initially labelled with depressive episodes may later be found to meet bipolar criteria when additional symptoms emerge (3). Antidepressant exposure may coincide with, but not necessarily cause, this unfolding bipolarity. In tertiary psychiatric settings where patients receive ongoing longitudinal assessment, the documentation of any bipolar diagnosis may reflect provisional clinical impressions, diagnostic speculation, or billing necessities rather than definitive diagnostic conclusions (25). Our anchored definition, requiring both diagnostic persistence and treatment concordance, identifies patients for whom the clinical team has acted on the bipolar diagnosis through mood stabiliser initiation. This operational definition more closely approximates the clinically meaningful outcome of interest: patients whose illness trajectory has substantively changed to warrant bipolar-specific treatment. The observation that neither antidepressant class nor individual drugs were associated with anchored bipolar disorder suggests that, at least in this tertiary setting, class-specific pharmacological effects on mood destabilisation are not the primary driver of documented bipolar conversions. The divergence between covariate-adjusted and IPTW results for secondary outcomes merits comment. After achieving near-perfect covariate balance through propensity score weighting, Tier 2 and Tier 3 outcomes showed modest but statistically significant associations with SNRI initiation, whilst the primary anchored outcome remained robustly null. This pattern suggests that SNRI initiators may receive more frequent bipolar diagnoses overall, but these do not translate to clinically anchored conversions requiring mood stabiliser therapy. Possible explanations include differential diagnostic attention to patients receiving SNRIs, channelling of diagnostically complex patients to SNRIs, or heightened clinician vigilance for mood instability in SNRI-treated patients (26). Regardless of mechanism, this finding reinforces the importance of using clinically anchored outcome definitions when evaluating antidepressant-associated bipolar conversion. Drug-Level Findings and Venlafaxine The drug-level analyses provide additional reassurance regarding venlafaxine specifically. Despite theoretical concern about its noradrenergic effects and prior evidence of elevated switch risk in bipolar populations (11), venlafaxine showed no excess risk for anchored bipolar disorder compared with either pooled SSRIs or individual agents. The point estimates were modestly elevated (aHR 1.27 to 1.69 depending on comparator) but confidence intervals were wide and consistently included the null. These findings suggest that venlafaxine's hypothesised differential risk may be specific to patients with established or latent bipolar disorder rather than a generalised phenomenon in depression populations. The unexpected finding of lower risk with sertraline compared with escitalopram for secondary outcomes (Tier 2 aHR 0.60, p = 0.009; Tier 3 aHR 0.69, p = 0.01) warrants cautious interpretation. This within-SSRI difference was not hypothesised a priori and may reflect channelling bias, whereby escitalopram is preferentially prescribed to patients with features suggesting higher bipolar risk. Alternatively, this finding may represent a type I error given multiple comparisons. We present this result as hypothesis-generating rather than confirmatory. Prior Antidepressant Use and Diagnostic Stability An unexpected finding was the strong protective association between prior antidepressant use and anchored bipolar disorder (HR 0.42, 95% CI 0.23 to 0.76). This association may reflect depletion of susceptibles or diagnostic stability mechanisms. Patients who have previously tolerated antidepressant treatment without mood destabilisation represent a lower-risk subgroup, having been effectively screened by prior exposure. Alternatively, patients with stable antidepressant histories may have more definitively established unipolar illness trajectories, making subsequent bipolar reclassification less likely. Clinically, this observation suggests that prior antidepressant tolerability may provide some reassurance when re-initiating treatment. This finding warrants replication and further investigation regarding potential predictive utility. Strengths and Limitations Methodological strengths of this study include the large sample size, new-user design, and examination of tiered outcomes. The active-comparator design substantially reduces confounding by indication compared with studies comparing antidepressant users to non-users. By synchronising the cohort at antidepressant initiation and controlling for prior illness course, we reduce biases inherent in observational treatment research. The 90-day landmark eliminates immortal time bias that has affected prior pharmacoepidemiological studies of psychiatric outcomes. The tiered outcome hierarchy provides novel insights into how outcome definition influences findings in tertiary psychiatric settings. Drug-level exploratory analyses address concerns about within-class heterogeneity and specifically examine venlafaxine, the agent of greatest theoretical concern. Comprehensive sensitivity analyses, including inverse probability of treatment weighting that achieved excellent covariate balance (all standardised mean differences <0.04), demonstrated robustness of the null finding for the primary outcome. Nonetheless, as an observational study, limitations remain. Unmeasured confounding, such as family history or subtle mixed features, could influence both antidepressant choice and bipolar risk. Our tertiary psychiatric setting may limit generalisability to primary care populations where antidepressants are commonly prescribed. Residual misclassification is possible, though we attempted to maximise diagnostic capture through tiered definitions. Our anchored outcome definition, whilst increasing specificity, may miss true bipolar conversions managed without mood stabiliser initiation. The drug-level analyses were underpowered to detect modest effect sizes, with confidence intervals typically spanning 0.5 to 2.0 or wider, and multiple comparisons increase the probability of spurious findings. Death and loss to follow-up were not explicitly modelled as competing risks; however, in this cohort of adults aged 18–60 years receiving tertiary psychiatric care, mortality is expected to be low, and patients remained under care at our institution until their last recorded encounter. Finally, we used an intention-to-treat approach with exposure fixed at the landmark. This approach mirrors clinical decision-making at treatment initiation and avoids informative censoring from treatment changes, but may dilute exposure contrast if patients subsequently switched treatments, biasing results toward the null. Clinical Implications Our findings suggest that among patients with depression seen in psychiatric care, SSRIs and SNRIs carry comparable risk of subsequent bipolar diagnosis. If anything, the observed high incidence of bipolar coding appears driven by the natural diagnostic evolution of mood disorders rather than by class-specific drug effects. Clinically, this implies that class selection (SSRI versus SNRI) need not be guided by bipolarity concerns alone; focus should remain on symptom profile, tolerability, and patient factors. This does not contradict guideline recommendations to prefer SSRIs over SNRIs for established bipolar depression (12,13), where the evidence base differs substantially. These findings apply to patients receiving tertiary psychiatric care and may not generalise to primary care settings where bipolar features may be less readily identified. Second, the absence of excess risk with venlafaxine in this depression cohort provides some reassurance for its use when clinically indicated. However, caution remains warranted in patients with features suggesting latent bipolar disorder, where guideline recommendations advise preferring agents with lower switch risk, such as SSRIs or bupropion (9). Third, the dramatic influence of outcome definition on event rates should inform interpretation of diagnostic conversion studies. Studies employing any-diagnosis outcomes may substantially overestimate rates of clinically meaningful bipolar conversion. Vigilant assessment for mania remains important for all antidepressant-treated patients (9,13), but these results do not single out one class as uniquely prone to unmask bipolar disorder. Fourth, the protective association observed with prior antidepressant use suggests that patients with a history of antidepressant tolerability may carry lower risk of subsequent bipolar diagnosis, though this requires prospective validation. Conclusions In this active-comparator landmark cohort study, neither SSRI versus SNRI class nor individual antidepressant choice, including venlafaxine, was associated with incident bipolar disorder when stringent, anchored outcome definitions were employed. The seven-fold difference in event rates between anchored and any-diagnosis outcomes highlights the profound influence of outcome definition on findings in specialty psychiatric settings. High event rates for less specific outcomes likely reflect diagnostic revision processes rather than pharmacological effects. These findings, consistent with recent multi-site studies, suggest that antidepressant class and drug selection may be less consequential for bipolar emergence than previously assumed when treating patients with presumed unipolar depression. Declarations Ethics approval and consent to participate This study was approved by the Ministry of Health and Prevention Research Ethics Committee (MOHAP REC), United Arab Emirates (Reference: MOHAP/DXB-REC/M.J.J/No.91/2024). The requirement for individual informed consent was waived by the ethics committee due to the retrospective nature of the study and the use of de-identified data from electronic health records. Consent for publication Not applicable. Availability of data and materials The data that support the findings of this study are available from Emirates Health Services, United Arab Emirates, but restrictions apply to the availability of these data, which were used under license for the current study, and so are not publicly available. Competing interests The authors declare that they have no competing interests. Funding This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors. Authors' contributions SAB conceived and designed the study, acquired and analysed the data, interpreted the results, and drafted the manuscript. AOE contributed to the study design, data interpretation, and critically revised the manuscript for important intellectual content. ME contributed to data interpretation and critically revised the manuscript for important intellectual content. All authors read and approved the final manuscript. Acknowledgements Not applicable. References Gitlin MJ. Antidepressants in bipolar depression: an enduring controversy. Int J Bipolar Disord [Internet]. 2018 Dec [cited 2026 Jan 12];6(1):25. Available from: https://journalbipolardisorders.springeropen.com/articles/10.1186/s40345-018-0133-9 Baldessarini RJ, Faedda GL, Offidani E, Vázquez GH, Marangoni C, Serra G, et al. Antidepressant-associated mood-switching and transition from unipolar major depression to bipolar disorder: A review. Journal of Affective Disorders [Internet]. 2013 May [cited 2026 Jan 12];148(1):129–35. Available from: https://linkinghub.elsevier.com/retrieve/pii/S0165032712007306 Hirschfeld RMA, Lewis L, Vornik LA. Perceptions and Impact of Bipolar Disorder: How Far Have We Really Come? Results of the National Depressive and Manic-Depressive Association 2000 Survey of Individuals With Bipolar Disorder. J Clin Psychiatry [Internet]. 2003 Feb 15 [cited 2026 Jan 12];64(2):161–74. Available from: https://www.psychiatrist.com/jcp/perceptions-impact-bipolar-disorder-far-really-results Baryshnikov I, Sund R, Marttunen M, Svirskis T, Partonen T, Pirkola S, et al. Diagnostic conversion from unipolar depression to bipolar disorder, schizophrenia, or schizoaffective disorder: A nationwide prospective 15‐year register study on 43 495 inpatients. Bipolar Disorders [Internet]. 2020 Sept [cited 2026 Jan 12];22(6):582–92. Available from: https://onlinelibrary.wiley.com/doi/10.1111/bdi.12929 Jo YT, Joo SW, Kim H, Ahn S, Choi YJ, Choi W, et al. Diagnostic conversion from unipolar to bipolar affective disorder-A population-based study. J Affect Disord. 2022 Mar 15;301:448–53. Musliner KL, Østergaard SD. Patterns and predictors of conversion to bipolar disorder in 91 587 individuals diagnosed with unipolar depression. Acta Psychiatr Scand [Internet]. 2018 May [cited 2026 Jan 12];137(5):422–32. Available from: https://onlinelibrary.wiley.com/doi/10.1111/acps.12869 Kessing LV, Willer I, Andersen PK, Bukh JD. Rate and predictors of conversion from unipolar to bipolar disorder: A systematic review and meta‐analysis. Bipolar Disorders [Internet]. 2017 Aug [cited 2026 Jan 12];19(5):324–35. Available from: https://onlinelibrary.wiley.com/doi/10.1111/bdi.12513 Oliva V, De Prisco M, La Spina E, Paolucci S, Fico G, Anmella G, et al. Switch to mania after acute antidepressant treatment for bipolar depression: a systematic review and network meta-analysis of randomised controlled trials. EClinicalMedicine. 2025 Sept;87:103413. Pacchiarotti I, Bond DJ, Baldessarini RJ, Nolen WA, Grunze H, Licht RW, et al. The International Society for Bipolar Disorders (ISBD) Task Force Report on Antidepressant Use in Bipolar Disorders. AJP [Internet]. 2013 Nov [cited 2026 Jan 12];170(11):1249–62. Available from: https://psychiatryonline.org/doi/10.1176/appi.ajp.2013.13020185 Goldberg JF, Truman CJ. Antidepressant‐induced mania: an overview of current controversies. Bipolar Disorders [Internet]. 2003 Dec [cited 2026 Jan 12];5(6):407–20. Available from: https://onlinelibrary.wiley.com/doi/10.1046/j.1399-5618.2003.00067.x Post RM, Altshuler LL, Leverich GS, Frye MA, Nolen WA, Kupka RW, et al. Mood switch in bipolar depression: comparison of adjunctive venlafaxine, bupropion and sertraline. Br J Psychiatry [Internet]. 2006 Aug [cited 2026 Jan 12];189(2):124–31. Available from: https://www.cambridge.org/core/product/identifier/S0007125000170333/type/journal_article Keramatian K, Chithra NK, Yatham LN. The CANMAT and ISBD Guidelines for the Treatment of Bipolar Disorder: Summary and a 2023 Update of Evidence. FOC [Internet]. 2023 Oct [cited 2026 Jan 12];21(4):344–53. Available from: https://psychiatryonline.org/doi/10.1176/appi.focus.20230009 Yatham LN, Kennedy SH, Parikh SV, Schaffer A, Bond DJ, Frey BN, et al. Canadian Network for Mood and Anxiety Treatments ( CANMAT ) and International Society for Bipolar Disorders ( ISBD ) 2018 guidelines for the management of patients with bipolar disorder. Bipolar Disorders [Internet]. 2018 Mar [cited 2026 Jan 12];20(2):97–170. Available from: https://onlinelibrary.wiley.com/doi/10.1111/bdi.12609 Patel R, Reiss P, Shetty H, Broadbent M, Stewart R, McGuire P, et al. Do antidepressants increase the risk of mania and bipolar disorder in people with depression? A retrospective electronic case register cohort study. BMJ Open [Internet]. 2015 Dec [cited 2026 Jan 12];5(12):e008341. Available from: https://bmjopen.bmj.com/lookup/doi/10.1136/bmjopen-2015-008341 Kim EY, Kim NW, Kim MJ, Yang BR, Rhee SJ, Park CHK, et al. Rate of diagnostic conversion to bipolar disorder in adults with unipolar depression and psychopharmacological treatment in the republic of Korea: A nationwide register-based study. J Affect Disord. 2020 Aug 1;273:240–6. Pradier MF, Hughes MC, McCoy TH, Barroilhet SA, Doshi-Velez F, Perlis RH. Predicting change in diagnosis from major depression to bipolar disorder after antidepressant initiation. Neuropsychopharmacol [Internet]. 2021 Jan [cited 2026 Jan 12];46(2):455–61. Available from: https://www.nature.com/articles/s41386-020-00838-x Zhu T, Kou R, Mu D, Hu Y, Yuan C, Yuan M, et al. Predicting Conversion From Unipolar Depression to Bipolar Disorder and Schizophrenia: A 10-Year Retrospective Cohort Study on 12,182 Inpatients. Depress Anxiety. 2025;2025:4048082. Schneeweiss S. Sensitivity analysis and external adjustment for unmeasured confounders in epidemiologic database studies of therapeutics. Pharmacoepidemiology and Drug [Internet]. 2006 May [cited 2026 Jan 12];15(5):291–303. Available from: https://onlinelibrary.wiley.com/doi/10.1002/pds.1200 Lund JL, Richardson DB, Stürmer T. The Active Comparator, New User Study Design in Pharmacoepidemiology: Historical Foundations and Contemporary Application. Curr Epidemiol Rep [Internet]. 2015 Dec [cited 2026 Jan 12];2(4):221–8. Available from: http://link.springer.com/10.1007/s40471-015-0053-5 Yoshida K, Solomon DH, Kim SC. Active-comparator design and new-user design in observational studies. Nat Rev Rheumatol [Internet]. 2015 July [cited 2026 Jan 12];11(7):437–41. Available from: https://www.nature.com/articles/nrrheum.2015.30 Suissa S. Immortal Time Bias in Pharmacoepidemiology. American Journal of Epidemiology [Internet]. 2008 Jan 7 [cited 2026 Jan 12];167(4):492–9. Available from: https://academic.oup.com/aje/article-lookup/doi/10.1093/aje/kwm324 Mi X, Hammill BG, Curtis LH, Lai ECC, Setoguchi S. Use of the landmark method to address immortal person-time bias in comparative effectiveness research: a simulation study: Landmark method for immortal person-time. Statist Med [Internet]. 2016 Nov 20 [cited 2026 Jan 12];35(26):4824–36. Available from: https://onlinelibrary.wiley.com/doi/10.1002/sim.7019 Yoo KH, Lee KJ, Lee SM, Han C, Park RW, Jo YT. Comparative effectiveness of selective serotonin reuptake inhibitors versus serotonin-norepinephrine reuptake inhibitors in the risk of diagnostic conversion from unipolar depression to bipolar disorder. International Journal of Psychiatry in Clinical Practice [Internet]. 2025 Dec 14 [cited 2026 Jan 12];1–9. Available from: https://www.tandfonline.com/doi/full/10.1080/13651501.2025.2600083 Baldessarini RJ, Bolzani L, Cruz N, Jones PB, Lai M, Lepri B, et al. Onset-age of bipolar disorders at six international sites. J Affect Disord. 2010 Feb;121(1–2):143–6. Davis KAS, Coleman JRI, Adams M, Allen N, Breen G, Cullen B, et al. Mental health in UK Biobank - development, implementation and results from an online questionnaire completed by 157 366 participants: a reanalysis. BJPsych Open. 2020 Feb 6;6(2):e18. Sendor R, Stürmer T. Core Concepts in Pharmacoepidemiology: Confounding by Indication and the Role of Active Comparators. Pharmacoepidemiol Drug Saf [Internet]. 2022 Mar [cited 2026 Jan 12];31(3):261–9. Available from: https://pmc.ncbi.nlm.nih.gov/articles/PMC9121653/ Additional Declarations No competing interests reported. Supplementary Files TableS1.CovariateBalanceBeforeandAfterInverseProbabilityofTreatmentWeighting.docx TableS2.InverseProbabilityofTreatmentWeightedAnalysisResults.docx TableS3.MissingDataAssessment.docx TableS4.FullModelCoefficientsforAllCovariatesAcrossOutcomeTiers.docx TableS5.ICD10DiagnosticCodesandMedicationCodes.docx TableS6.DrugLevelSensitivityAnalysesFullResults.docx TableS7.AdditionalDrugLevelHeadtoHeadComparisons.docx FigureS1PatientFlow.docx FigureS2StudyDesign.docx FigureS3DirectedAcyclicGraphforCovariateSelection.docx Cite Share Download PDF Status: Published Journal Publication published 28 Apr, 2026 Read the published version in International Journal of Bipolar Disorders → Version 1 posted Editorial decision: Revision requested 02 Apr, 2026 Reviewers agreed at journal 16 Jan, 2026 Reviews received at journal 14 Jan, 2026 Reviewers agreed at journal 14 Jan, 2026 Reviewers invited by journal 14 Jan, 2026 Editor assigned by journal 12 Jan, 2026 Submission checks completed at journal 12 Jan, 2026 First submitted to journal 12 Jan, 2026 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-8580658","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":575842797,"identity":"f754994f-afa4-4064-8d7c-8d43cb259ba8","order_by":0,"name":"Syed Ali Bokhari","email":"","orcid":"","institution":"Al Amal Psychiatric Hospital, Emirates Health Services","correspondingAuthor":false,"prefix":"","firstName":"Syed","middleName":"Ali","lastName":"Bokhari","suffix":""},{"id":575842798,"identity":"bbd60e7c-b432-4b92-8b8e-2395f2e73a7c","order_by":1,"name":"Abdulhameed Omer Elnour","email":"","orcid":"","institution":"Al Ain University","correspondingAuthor":false,"prefix":"","firstName":"Abdulhameed","middleName":"Omer","lastName":"Elnour","suffix":""},{"id":575842799,"identity":"8ba9b741-2d0a-4d1c-89c6-24c519ada5ef","order_by":2,"name":"Muhanad Elnoor","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABAElEQVRIiWNgGAWjYFACHhDBzMAH5lQwJBCvhQ3MOUOyFsY2IrTozsg9+LmgxlqOTbr54OPKeTZ58g3sFx/zMNjJ4dJidiMvWXrGsXRjNpljyYZnt6UVGxzgKTbmYUg2xq0lx0Cah+1wYptEjplk47bDiRsYeNIkZzAwJzbg1mL8m+ff4XqIljn/E+c3gLXU1+PRYibN23Y4gQ2speFAYsMB9mMSHxgO4wwHszNvzKx5+9IN2yTSkg0bjiUnbjjMw2zwweC4IU5bjucY3+b5Zi3PL5F88GFDjV3i/Pb2hw8SKqrlcdmCBTDzGDAwGJCgAQjYH5CmfhSMglEwCoY7AADfUFIpYhsAdgAAAABJRU5ErkJggg==","orcid":"","institution":"Sudan University of Science and Technology","correspondingAuthor":true,"prefix":"","firstName":"Muhanad","middleName":"","lastName":"Elnoor","suffix":""}],"badges":[],"createdAt":"2026-01-12 10:40:08","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8580658/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8580658/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1186/s40345-026-00426-w","type":"published","date":"2026-04-28T15:57:04+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":100560881,"identity":"79552f33-e527-43a0-b473-5e2c416c5b8c","added_by":"auto","created_at":"2026-01-19 08:43:52","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":527639,"visible":true,"origin":"","legend":"","description":"","filename":"SSRIvsSNRIBipolarRiskFinal.docx","url":"https://assets-eu.researchsquare.com/files/rs-8580658/v1/ce494568d2f099c676826f43.docx"},{"id":100560960,"identity":"466a473d-8d73-4bfa-bb4b-c81c1d495c80","added_by":"auto","created_at":"2026-01-19 08:43:54","extension":"json","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":6029,"visible":true,"origin":"","legend":"","description":"","filename":"6006d1d203a649a484c3b7d1c754a402.json","url":"https://assets-eu.researchsquare.com/files/rs-8580658/v1/c1876a461afb394fa2a5a674.json"},{"id":100560978,"identity":"7b7dd7d0-814d-4ee7-b466-fd0cfb2375ab","added_by":"auto","created_at":"2026-01-19 08:43:54","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":58996,"visible":true,"origin":"","legend":"","description":"","filename":"FigureS1PatientFlow.docx","url":"https://assets-eu.researchsquare.com/files/rs-8580658/v1/0af704ce3f80a66a652266ef.docx"},{"id":100560909,"identity":"aff208cf-82ad-49b1-9eb2-da2d96fa5f26","added_by":"auto","created_at":"2026-01-19 08:43:52","extension":"docx","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":31483,"visible":true,"origin":"","legend":"","description":"","filename":"FigureS2StudyDesign.docx","url":"https://assets-eu.researchsquare.com/files/rs-8580658/v1/a643c57d07e2be96be0c7dd8.docx"},{"id":100560848,"identity":"31b9c89d-df46-4b67-b52a-8f21ba775fe2","added_by":"auto","created_at":"2026-01-19 08:43:51","extension":"docx","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":299856,"visible":true,"origin":"","legend":"","description":"","filename":"FigureS3DirectedAcyclicGraphforCovariateSelection.docx","url":"https://assets-eu.researchsquare.com/files/rs-8580658/v1/655a16fee1129dfac01b711c.docx"},{"id":100594558,"identity":"7b628897-3bb9-448e-82d2-b4805d0f996c","added_by":"auto","created_at":"2026-01-19 13:42:45","extension":"docx","order_by":5,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":16204,"visible":true,"origin":"","legend":"","description":"","filename":"TableS1.CovariateBalanceBeforeandAfterInverseProbabilityofTreatmentWeighting.docx","url":"https://assets-eu.researchsquare.com/files/rs-8580658/v1/85b647627309e70db3bdd5aa.docx"},{"id":100560811,"identity":"bfac8ec4-f5b7-4b81-a2bc-6df969ace683","added_by":"auto","created_at":"2026-01-19 08:43:51","extension":"docx","order_by":6,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":16532,"visible":true,"origin":"","legend":"","description":"","filename":"TableS2.InverseProbabilityofTreatmentWeightedAnalysisResults.docx","url":"https://assets-eu.researchsquare.com/files/rs-8580658/v1/c0c8ad5811268a90487c75ed.docx"},{"id":100560897,"identity":"320683ed-79a3-473c-b0eb-cb5bb359692b","added_by":"auto","created_at":"2026-01-19 08:43:52","extension":"docx","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":17638,"visible":true,"origin":"","legend":"","description":"","filename":"TableS3.MissingDataAssessment.docx","url":"https://assets-eu.researchsquare.com/files/rs-8580658/v1/1a13a47efc8237e7cc1a652f.docx"},{"id":100561015,"identity":"ad30cf7d-f255-44b2-bec2-79c972676852","added_by":"auto","created_at":"2026-01-19 08:43:55","extension":"docx","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":17089,"visible":true,"origin":"","legend":"","description":"","filename":"TableS4.FullModelCoefficientsforAllCovariatesAcrossOutcomeTiers.docx","url":"https://assets-eu.researchsquare.com/files/rs-8580658/v1/54c33aa9acd73fcabebc2239.docx"},{"id":100560624,"identity":"a1a5ab03-a5a0-438b-a4b1-d7450e0b3d8f","added_by":"auto","created_at":"2026-01-19 08:43:49","extension":"docx","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":22636,"visible":true,"origin":"","legend":"","description":"","filename":"TableS5.ICD10DiagnosticCodesandMedicationCodes.docx","url":"https://assets-eu.researchsquare.com/files/rs-8580658/v1/9e644ce89c0c90f121e02716.docx"},{"id":100595671,"identity":"9d21afc5-d5d7-4cbc-971c-09b9f57477c7","added_by":"auto","created_at":"2026-01-19 13:49:05","extension":"docx","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":36730,"visible":true,"origin":"","legend":"","description":"","filename":"TableS6.DrugLevelSensitivityAnalysesFullResults.docx","url":"https://assets-eu.researchsquare.com/files/rs-8580658/v1/0e806f5348aaf26a225fe7d0.docx"},{"id":100594825,"identity":"04d0a03c-0476-405a-804d-1817811efb93","added_by":"auto","created_at":"2026-01-19 13:45:26","extension":"docx","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":19630,"visible":true,"origin":"","legend":"","description":"","filename":"TableS7.AdditionalDrugLevelHeadtoHeadComparisons.docx","url":"https://assets-eu.researchsquare.com/files/rs-8580658/v1/b11d800408ed12abf7205321.docx"},{"id":100560779,"identity":"f0770e95-9b49-457a-8d6a-83265f8cc0a4","added_by":"auto","created_at":"2026-01-19 08:43:50","extension":"png","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":467307,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-8580658/v1/03d0c255777d46f51d87fc0f.png"},{"id":100560532,"identity":"32d1c9de-9625-4ee9-84ee-4d3ff09eb09e","added_by":"auto","created_at":"2026-01-19 08:43:42","extension":"png","order_by":13,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":71801,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-8580658/v1/8536bddc267d1c4684b332d6.png"},{"id":100595502,"identity":"8e95cde6-1138-4898-af13-80b0075d5f5a","added_by":"auto","created_at":"2026-01-19 13:48:37","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":142069,"visible":true,"origin":"","legend":"\u003cp\u003eKaplan-Meier Curves for Incident Bipolar Disorder by Antidepressant Class.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eNote. Panel A: Tier 1 anchored bipolar disorder. Panel B: Tier 3 any bipolar diagnosis. Shaded areas represent 95% confidence intervals. aHR = adjusted hazard ratio. Number-at-risk tables are shown below each panel.\u003c/em\u003e\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-8580658/v1/e88d386568da1150654bf2f2.png"},{"id":108437563,"identity":"ff3bb832-3c2c-490e-8718-200874a8b652","added_by":"auto","created_at":"2026-05-04 15:59:20","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":431669,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8580658/v1/fc9ca6a4-9723-41ca-b15d-34573175908a.pdf"},{"id":100560917,"identity":"8b2d9206-d76b-43f7-b13f-1ff025ca7ac6","added_by":"auto","created_at":"2026-01-19 08:43:53","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":16204,"visible":true,"origin":"","legend":"","description":"","filename":"TableS1.CovariateBalanceBeforeandAfterInverseProbabilityofTreatmentWeighting.docx","url":"https://assets-eu.researchsquare.com/files/rs-8580658/v1/508cfcd000a5a73cf46bb3f2.docx"},{"id":100560766,"identity":"a29f7039-91fb-4551-b78b-4e38624023f3","added_by":"auto","created_at":"2026-01-19 08:43:50","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":16532,"visible":true,"origin":"","legend":"","description":"","filename":"TableS2.InverseProbabilityofTreatmentWeightedAnalysisResults.docx","url":"https://assets-eu.researchsquare.com/files/rs-8580658/v1/4d47d2eb94dc490d68d6cd0c.docx"},{"id":100560684,"identity":"472e2d0c-40ac-49a7-ad2e-0ccf976b96d2","added_by":"auto","created_at":"2026-01-19 08:43:49","extension":"docx","order_by":3,"title":"","display":"","copyAsset":false,"role":"supplement","size":17638,"visible":true,"origin":"","legend":"","description":"","filename":"TableS3.MissingDataAssessment.docx","url":"https://assets-eu.researchsquare.com/files/rs-8580658/v1/4107567203e32fd3fbb00f82.docx"},{"id":100560577,"identity":"69c0340e-0ff9-4bb6-bce8-287cdf697e19","added_by":"auto","created_at":"2026-01-19 08:43:48","extension":"docx","order_by":4,"title":"","display":"","copyAsset":false,"role":"supplement","size":17089,"visible":true,"origin":"","legend":"","description":"","filename":"TableS4.FullModelCoefficientsforAllCovariatesAcrossOutcomeTiers.docx","url":"https://assets-eu.researchsquare.com/files/rs-8580658/v1/c51bdbfdb63006e63df22873.docx"},{"id":100561011,"identity":"0f994c48-c57b-4d03-8744-f29218d2c05e","added_by":"auto","created_at":"2026-01-19 08:43:55","extension":"docx","order_by":5,"title":"","display":"","copyAsset":false,"role":"supplement","size":22636,"visible":true,"origin":"","legend":"","description":"","filename":"TableS5.ICD10DiagnosticCodesandMedicationCodes.docx","url":"https://assets-eu.researchsquare.com/files/rs-8580658/v1/68648e6c332c37d437577339.docx"},{"id":100560674,"identity":"dffa060f-a3df-432b-957d-3bf846127531","added_by":"auto","created_at":"2026-01-19 08:43:49","extension":"docx","order_by":6,"title":"","display":"","copyAsset":false,"role":"supplement","size":36730,"visible":true,"origin":"","legend":"","description":"","filename":"TableS6.DrugLevelSensitivityAnalysesFullResults.docx","url":"https://assets-eu.researchsquare.com/files/rs-8580658/v1/052e856118e580784ec15151.docx"},{"id":100560717,"identity":"ec4f6daa-a329-4f5d-a29c-6315e45dbf89","added_by":"auto","created_at":"2026-01-19 08:43:49","extension":"docx","order_by":7,"title":"","display":"","copyAsset":false,"role":"supplement","size":19630,"visible":true,"origin":"","legend":"","description":"","filename":"TableS7.AdditionalDrugLevelHeadtoHeadComparisons.docx","url":"https://assets-eu.researchsquare.com/files/rs-8580658/v1/e6a27bf3a12fdf17378f6a5e.docx"},{"id":100594816,"identity":"77ce9be8-3309-44c8-961e-9a49b1560613","added_by":"auto","created_at":"2026-01-19 13:45:21","extension":"docx","order_by":8,"title":"","display":"","copyAsset":false,"role":"supplement","size":58996,"visible":true,"origin":"","legend":"","description":"","filename":"FigureS1PatientFlow.docx","url":"https://assets-eu.researchsquare.com/files/rs-8580658/v1/f1352c676ed1bb7e7fb9b323.docx"},{"id":100560733,"identity":"e06f0090-e390-4ce7-bc39-2b7eb09e817c","added_by":"auto","created_at":"2026-01-19 08:43:49","extension":"docx","order_by":9,"title":"","display":"","copyAsset":false,"role":"supplement","size":31483,"visible":true,"origin":"","legend":"","description":"","filename":"FigureS2StudyDesign.docx","url":"https://assets-eu.researchsquare.com/files/rs-8580658/v1/cfe53d6e9eebfce6e2c3b1ed.docx"},{"id":100595491,"identity":"47b531bf-f208-4b53-bdf0-d7bc26152b6b","added_by":"auto","created_at":"2026-01-19 13:48:35","extension":"docx","order_by":10,"title":"","display":"","copyAsset":false,"role":"supplement","size":299856,"visible":true,"origin":"","legend":"","description":"","filename":"FigureS3DirectedAcyclicGraphforCovariateSelection.docx","url":"https://assets-eu.researchsquare.com/files/rs-8580658/v1/4b0e41517e273453bbd76fe8.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"SSRI versus SNRI Initiation and Incident Bipolar Disorder in Tertiary Psychiatric Care: An Active-Comparator Cohort Study from the United Arab Emirates","fulltext":[{"header":"Introduction","content":"\u003cp\u003eRisk of treatment-emergent mood switch and diagnostic conversion from unipolar depression to bipolar disorder has been widely discussed in psychiatry (1). Bipolar disorder represents a significant diagnostic challenge, with approximately half of initial presentations occurring as depressive episodes rather than mania or hypomania (2,3). This characteristic presentation pattern frequently results in initial diagnoses of major depressive disorder, with subsequent diagnostic conversion to bipolar disorder occurring as the illness course unfolds. Prospective registry studies have documented cumulative conversion rates from unipolar depression to bipolar disorder ranging from 6.5% to 11.1% over observation periods of 10 to 15 years\u0026nbsp;(4–6). The highest conversion rates typically occur within the first year following initial depression diagnosis, with the rate declining to approximately 0.8% to 1.0% annually thereafter\u0026nbsp;(7).\u003c/p\u003e\n\u003cp\u003eOlder antidepressants, especially tricyclic antidepressants, were long implicated in inducing mania (Tondo et al., 2010), but modern agents such as SSRIs and SNRIs are now most commonly prescribed for depressive episodes, and their relative propensity to unmask bipolarity remains unclear (8,9). The question of whether antidepressant treatment influences the risk of subsequent bipolar diagnosis has generated substantial controversy. Three competing hypotheses exist: first, that antidepressants may precipitate manic or hypomanic episodes in individuals with latent bipolar vulnerability, thereby unmasking the condition; second, that observed associations reflect confounding by indication, whereby individuals with unrecognised bipolar depression are more likely to receive antidepressant treatment and subsequently manifest overt bipolarity; and third, that apparent associations arise from diagnostic revision processes in psychiatric care, particularly in specialty settings where sustained clinical attention increases the probability of detecting mood instability (1,10).\u003c/p\u003e\n\u003cp\u003eExperts have often cautioned that SNRIs, notably venlafaxine, may carry higher switch risk than SSRIs (11). The seminal work by Post et al. (2006), comparing venlafaxine, sertraline, and bupropion as adjunctive treatments in bipolar depression, reported significantly elevated switch rates with venlafaxine (29%) compared with sertraline (9%) and bupropion (10%). The most recent network meta-analysis by Oliva et al. (2025) confirmed that whilst no individual antidepressant demonstrated statistically significantly increased switch risk compared with placebo, venlafaxine consistently showed the highest relative risk estimates across sensitivity analyses (8). The 2018 Canadian Network for Mood and Anxiety Treatments and International Society for Bipolar Disorders (CANMAT/ISBD) guidelines accordingly recommend that SSRIs (particularly sertraline) and bupropion be preferred over SNRIs when antidepressants are considered for bipolar depression (12,13).\u003c/p\u003e\n\u003cp\u003eHowever, observational evidence comparing SSRI and SNRI classes in depression cohorts is inconsistent. Large electronic health record-based cohorts have found that prior antidepressant exposure in general is associated with increased rates of new mania or bipolar diagnosis, but direct comparisons between SSRI and SNRI classes have been limited. Patel et al. (2015), in a South London and Maudsley NHS Trust cohort study, reported that both SSRI (HR 1.34, 95% CI 1.18 to 1.52) and venlafaxine exposure (HR 1.35, 95% CI 1.07 to 1.70) were associated with subsequent mania or bipolar disorder diagnosis compared with no antidepressant exposure (14). A nationwide Korean register study by Kim et al. (2020) found that the conversion rate was dramatically higher among patients prescribed SNRIs, with SNRI users having more than double the risk versus no antidepressant (15).\u003c/p\u003e\n\u003cp\u003eConversely, other studies suggest that class differences may be less pronounced than hypothesised. Pradier et al. (2021), using multi-site US health records (n = 67,807), found that short-term conversion rates within 3 months were numerically higher for SNRIs (1.93%) than SSRIs (1.23%), although this was a preliminary finding in a predictive modelling context (16). In contrast, a recent multi-centre South Korean analysis by Yoo et al. (2025) reported no significant difference between SSRI and SNRI initiators in conversion to bipolar disorder using rigorous new-user design methodology. Similarly, Zhu et al. (2025) reported from a Chinese inpatient cohort that patients who eventually converted to bipolar had commonly received either an SNRI (venlafaxine) or an SSRI (paroxetine), without clear class differentiation (17). Thus, evidence is inconsistent and may depend on population (psychiatric specialty versus general practice), follow-up period, and how outcomes are defined (narrow mania or hypomania diagnoses versus broader bipolar disorder codes).\u003c/p\u003e\n\u003cp\u003eThe methodological quality of observational studies examining antidepressant effects on bipolar conversion has been variable. Two design features merit particular attention. First, confounding by indication represents a fundamental challenge: individuals prescribed antidepressants differ systematically from those not treated, and those prescribed SNRIs may differ from those prescribed SSRIs in ways that independently influence bipolar risk (18). Active-comparator designs, which compare initiators of one treatment with initiators of an alternative treatment for the same indication rather than with untreated individuals, substantially attenuate this bias by ensuring that all included patients have crossed the threshold for treatment initiation (19,20). Second, immortal time bias can substantially distort effect estimates in pharmacoepidemiological studies. This bias arises when follow-up time during which the outcome cannot occur is misattributed between comparison groups (21). The landmark analysis approach, which defines a fixed time point at which exposure is assigned and from which follow-up begins, represents an established solution to this problem (22).\u003c/p\u003e\n\u003cp\u003eA third consideration specific to psychiatric outcomes concerns the definition and validation of the outcome itself. In specialty psychiatric settings where patients receive ongoing diagnostic assessment, the detection of any single bipolar diagnosis may reflect provisional clinical impressions, diagnostic speculation, or billing necessities rather than definitive diagnostic conclusions. The use of stringent, anchored outcome definitions that require diagnostic persistence and treatment concordance may more accurately identify true diagnostic conversions.\u003c/p\u003e\n\u003cp\u003eThe present study addresses these methodological considerations by employing an active-comparator landmark cohort design to compare SSRI versus SNRI initiators for incident bipolar disorder in a specialty psychiatric setting. We implemented a tiered outcome hierarchy to evaluate how outcome specificity influences observed event rates and treatment associations. Additionally, we conducted exploratory drug-level analyses to examine whether within-class heterogeneity, particularly regarding venlafaxine, might be obscured by class-level comparisons. Based on prior literature and pharmacological reasoning, we hypothesised that if antidepressant class itself confers differential risk, SNRIs might show higher rates of bipolar conversion than SSRIs. We also considered the possibility that observed associations could reflect confounding by indication or disease severity rather than true pharmacological effects.\u003c/p\u003e"},{"header":"Methods","content":"\u003cp\u003e\u003cstrong\u003eStudy Design and Setting\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe conducted a retrospective cohort study using electronic health records from a tertiary psychiatric hospital in the United Arab Emirates (1 January 2018 to 22 December 2025). We employed an active-comparator design among antidepressant initiators with a 90-day landmark to minimise immortal time bias. The study was approved by the institutional ethics committee.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData Source\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eData were extracted from the General Adult Psychiatry service, which provides outpatient and inpatient care for patients aged 18 to 60 years. The database includes demographic information, encounter records with International Classification of Diseases, 10th Revision (ICD-10) diagnoses, and medication dispensing records with dose information.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eStudy Population\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe index condition was a new depression diagnosis, defined as the first encounter with any qualifying ICD-10 code: F32 (depressive episode), F33 (recurrent depressive disorder), F34.1 (dysthymia), or F41.2 (mixed anxiety and depressive disorder). Inclusion criteria required age 18 to 60 years at index, at least one encounter after the 90-day landmark, and first SSRI or SNRI dispensed within the exposure window (index date to landmark). Patients with bipolar disorder diagnosis (F30 or F31) documented before the landmark were excluded.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eExposure Definition\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003ePatients were classified according to their first antidepressant dispensed within the 90-day exposure window. SSRIs included sertraline, escitalopram, fluoxetine, paroxetine, citalopram, and fluvoxamine. SNRIs included venlafaxine, duloxetine, and desvenlafaxine. Exposure was assigned at the 90-day landmark based on the first antidepressant dispensed within the index-to-landmark window and remained fixed throughout follow-up (intention-to-treat). This approach captures the clinical decision at treatment initiation and avoids informative censoring from treatment changes.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eOutcome Definitions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe employed a tiered hierarchy to assess outcome specificity. The primary outcome, Tier 1 (anchored) bipolar disorder, required two or more bipolar diagnoses (F30 or F31) occurring 30 or more days apart, with the first occurring after the landmark, plus new initiation of a mood stabiliser (lithium, valproate, carbamazepine, or lamotrigine) within 0 to 90 days after the first qualifying bipolar diagnosis, with no mood stabiliser dispensing before the index date. The Tier 2 (confirmed) outcome required two or more bipolar diagnoses 30 or more days apart. The Tier 3 (any) outcome required one or more bipolar diagnoses. Follow-up began at the 90-day landmark and ended at outcome occurrence, last encounter, or study end.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCovariates\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eBaseline characteristics were assessed from 365 days before index through the index date. Covariates included age (continuous), sex, schizophrenia spectrum disorders (F20 to F29), substance use disorder (F10 to F19), prior antipsychotic dispensing, and prior SSRI or SNRI dispensing in the 365 days before index. A baseline bipolar proxy variable, defined as any recorded bipolar spectrum diagnosis in the 365 days before index, was used in prespecified sensitivity analyses.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eStatistical Analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe primary analysis employed Cox proportional hazards regression comparing hazards between SSRI and SNRI initiators, adjusting for all covariates. The proportional hazards assumption was tested using Schoenfeld residuals. Sensitivity analyses included 180-day landmark, exclusion of baseline schizophrenia spectrum diagnoses, exclusion of baseline bipolar proxy diagnoses, and restriction to a washout cohort with no SSRI or SNRI use in the 365 days before index.\u003c/p\u003e\n\u003cp\u003eExploratory drug-level analyses examined individual antidepressants to assess within-class heterogeneity. Prespecified comparisons included venlafaxine versus sertraline, venlafaxine versus escitalopram, venlafaxine versus pooled SSRIs, duloxetine versus pooled SSRIs, and desvenlafaxine versus pooled SSRIs. A within-SSRI comparison of sertraline versus escitalopram served as a negative control. These analyses were considered feasible if sample size exceeded 80 per arm or combined Tier 1 events exceeded 20. P-values were not adjusted for multiple comparisons, and these results are interpreted as exploratory.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eComplete case analysis was used; all key analytic variables had no missing data among the 1,095 patients in the final cohort. As a sensitivity analysis, we estimated propensity scores for SSRI versus SNRI assignment using logistic regression with all covariates and applied stabilised inverse probability of treatment weights (IPTW). Weights were truncated at the 1st and 99th percentiles. Covariate balance was assessed using standardised mean differences before and after weighting.\u003c/p\u003e\n\u003cp\u003eAnalyses were conducted through Python version 3.9. Statistical significance was defined as alpha = 0.05 for the primary outcome; secondary and exploratory analyses were hypothesis-generating.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003e\u003cstrong\u003eStudy Population\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFrom 14,583 General Adult Psychiatry patients, 4,166 had qualifying depression diagnoses. After applying the 90-day landmark and exposure criteria, 1,095 antidepressant initiators comprised the final analytic cohort. SSRI initiators (n = 818, 74.7%) and SNRI initiators (n = 277, 25.3%) differed in mean age (34.4 versus 37.7 years, standardised mean difference [SMD] = 0.31) and prevalence of baseline schizophrenia spectrum diagnoses (35.8% versus 44.0%, SMD = 0.17). Prior antidepressant use in the year before index was common in both groups (52.8% versus 58.1%, SMD = 0.11). Median follow-up was 0.90 years (interquartile range 0.31 to 2.21), with total follow-up of 1,610.6 person-years. Baseline characteristics are presented in \u003cstrong\u003eTable 1\u003c/strong\u003e.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTable 1\u003c/strong\u003e. Baseline Characteristics of the Study Cohort.\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"100%\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003eCharacteristic\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003eSSRI (N = 818)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003eSNRI (N = 277)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003eSMD\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eAge, years, mean (SD)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e34.4 (11.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e37.7 (10.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.31\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eFemale sex, n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e371 (45.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e111 (40.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.11\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eSchizophrenia spectrum, n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e293 (35.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e122 (44.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.17\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eSubstance use disorder, n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e219 (26.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e92 (33.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.14\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eAnxiety disorder, n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e161 (19.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e52 (18.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.02\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eBipolar disorder markers, n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e74 (9.0)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e26 (9.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.01\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003ePrior antipsychotic use, n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e345 (42.2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e130 (46.9)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.10\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003ePrior SSRI/SNRI use (365 days), n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e432 (52.8)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e161 (58.1)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.11\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eFollow-up, years, mean (SD)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e1.4 (1.4)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e1.7 (1.7)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.17\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u003cem\u003eNote. SMD = standardised mean difference. Values above 0.10 suggest meaningful imbalance. Baseline and prior medication covariates were assessed in the 365 days before index.\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePrimary Outcome\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eDuring follow-up, 72 patients (6.6%) developed Tier 1 anchored bipolar disorder, with an incidence rate of 4.47 per 100 person-years. SSRI versus SNRI initiation was not associated with this outcome in adjusted analyses (aHR 1.07, 95% CI 0.63 to 1.84, p = 0.80; \u003cstrong\u003eTable 2\u003c/strong\u003e). In the full model, younger age was associated with increased risk (HR 0.97 per year, 95% CI 0.95 to 0.99, p = 0.016), and prior antidepressant use was associated with reduced risk of anchored bipolar disorder (HR 0.42, 95% CI 0.23 to 0.76, p = 0.004). Cumulative incidence of anchored bipolar disorder was 1.3% at 1 year, 4.0% at 2 years, 6.0% at 3 years, and 11.8% at 5 years.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSecondary Outcomes\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFor Tier 2 confirmed bipolar disorder (328 events, 30.0%), no significant difference was observed between groups (aHR 1.24, 95% CI 0.96 to 1.60, p = 0.095). For Tier 3 any bipolar diagnosis (507 events, 46.3%), results were similarly null (aHR 1.13, 95% CI 0.92 to 1.38, p = 0.24). The proportional hazards assumption was satisfied for all models (Schoenfeld test p \u0026gt; 0.05).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTable 2.\u0026nbsp;\u003c/strong\u003ePrimary and Secondary Outcomes: SSRI versus SNRI Initiation and Incident Bipolar Disorder.\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"100%\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003eOutcome\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003eN\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003eEvents\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003eaHR\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003e95% CI\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003e\u003cem\u003ep\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eTier 1: Anchored (Primary)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e1,095\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e72\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e1.07\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.63 to 1.84\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.80\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eTier 2: Confirmed\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e1,095\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e328\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e1.24\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.96 to 1.60\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.10\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eTier 3: Any diagnosis\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e1,095\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e507\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e1.13\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.92 to 1.38\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.24\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u003cem\u003eNote. aHR = adjusted hazard ratio; CI = confidence interval. Hazard ratios are adjusted for age, sex, baseline schizophrenia spectrum diagnosis, baseline substance use disorder, prior antipsychotic use, and prior SSRI or SNRI use in the 365 days before index. Reference group is SNRI initiators.\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eKaplan-Meier survival curves for anchored bipolar disorder (Tier 1) and any bipolar diagnosis (Tier 3) are presented in \u003cstrong\u003eFigure 1\u003c/strong\u003e. The curves demonstrated overlapping survival probabilities across antidepressant classes for both outcome tiers, consistent with the null hazard ratio findings.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSensitivity Analyses\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eResults were consistent across all prespecified sensitivity analyses (\u003cstrong\u003eTable 3\u003c/strong\u003e). In the washout-restricted cohort (n = 502) without SSRI or SNRI exposure in the prior year, the Tier 1 aHR was 0.91 (95% CI 0.46 to 1.81, p = 0.79). Excluding patients with baseline schizophrenia spectrum diagnoses (n = 680) yielded a Tier 1 aHR of 0.74 (95% CI 0.39 to 1.41, p = 0.36). Excluding patients with baseline bipolar proxy diagnoses (n = 995) produced a Tier 1 aHR of 0.96 (95% CI 0.53 to 1.71, p = 0.88). The 180-day landmark analysis (n = 1,027) showed a Tier 1 aHR of 1.44 (95% CI 0.78 to 2.67, p = 0.24). No significant interactions were observed between antidepressant class and baseline schizophrenia spectrum or bipolar markers for any outcome tier.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTable 3.\u0026nbsp;\u003c/strong\u003eSensitivity Analyses: SSRI versus SNRI Initiation Across Alternative Cohort Definitions.\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"100%\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003eAnalysis\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003eTier 1 aHR (95% CI)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003e\u003cem\u003ep\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eMain Analysis (N = 1,095)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e1.07 (0.63 to 1.84)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.80\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eWashout-Restricted (N = 502)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.91 (0.46 to 1.81)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.79\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eExcluding Schizophrenia (N = 680)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.74 (0.39 to 1.41)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.36\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eExcluding Bipolar Markers (N = 995)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.96 (0.53 to 1.71)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.88\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e180-Day Landmark (N = 1,027)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e1.44 (0.78 to 2.67)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.24\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u003cem\u003eNote. aHR = adjusted hazard ratio; CI = confidence interval. All models adjusted for age, sex, baseline schizophrenia spectrum diagnosis, baseline substance use disorder, prior antipsychotic use, and prior SSRI/SNRI use (except washout-restricted analysis which excludes prior SSRI/SNRI covariate).\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePropensity Score-Weighted Analyses\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eIPTW achieved excellent covariate balance, with all standardised mean differences reduced from 0.10\u0026ndash;0.31 to less than 0.04 (Supplementary Table S1). In IPTW-weighted analyses, Tier 1 anchored bipolar disorder remained non-significant (HR 1.02, 95% CI 0.87\u0026ndash;1.21, p = 0.78), consistent with covariate-adjusted estimates. For secondary outcomes, IPTW analyses showed statistically significant associations for Tier 2 (HR 1.23, 95% CI 1.13\u0026ndash;1.33, p \u0026lt; 0.001) and Tier 3 (HR 1.11, 95% CI 1.04\u0026ndash;1.18, p = 0.002), although effect sizes remained modest (Supplementary Table S2). This divergence between primary and secondary outcomes in fully balanced analyses further supports the interpretation that less specific outcome definitions capture diagnostic processes distinct from clinically meaningful bipolar conversion.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDrug-Level Exploratory Analyses\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAmong individual antidepressants, escitalopram (n = 373, 34.1%) and sertraline (n = 199, 18.2%) were the most commonly prescribed SSRIs, whilst venlafaxine (n = 106, 9.7%) was the most common SNRI. Drug-level comparisons for the primary outcome (Tier 1 anchored bipolar disorder) are presented in Table 4.\u003c/p\u003e\n\u003cp\u003eNo individual drug comparison showed a statistically significant association with anchored bipolar disorder. Venlafaxine versus pooled SSRIs yielded an aHR of 1.27 (95% CI 0.62 to 2.60, p = 0.52). Direct comparisons of venlafaxine with sertraline (aHR 1.69, 95% CI 0.61 to 4.70, p = 0.32) and escitalopram (aHR 1.17, 95% CI 0.54 to 2.51, p = 0.69) were similarly null. Duloxetine versus pooled SSRIs (aHR 0.87, 95% CI 0.31 to 2.43, p = 0.79) and desvenlafaxine versus pooled SSRIs (aHR 0.82, 95% CI 0.35 to 1.90, p = 0.64) showed no association with the primary outcome.\u003c/p\u003e\n\u003cp\u003eFor secondary outcomes, one notable finding emerged: sertraline showed significantly lower risk than escitalopram for Tier 2 confirmed bipolar disorder (aHR 0.60, 95% CI 0.42 to 0.88, p = 0.009) and Tier 3 any bipolar diagnosis (aHR 0.69, 95% CI 0.53 to 0.92, p = 0.01). This within-SSRI difference was not hypothesised a priori and should be interpreted as hypothesis-generating. Sensitivity analyses for drug-level comparisons showed consistent null findings for the primary outcome across all cohort restrictions.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTable 4.\u0026nbsp;\u003c/strong\u003eDrug-Level Exploratory Analyses: Individual Antidepressant Comparisons for Tier 1 Anchored Bipolar Disorder.\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"100%\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003eComparison\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003eEvents\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003eaHR (95% CI)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003e\u003cem\u003ep\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eVenlafaxine vs Sertraline\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e9 / 8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e1.69 (0.61 to 4.70)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.32\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eVenlafaxine vs Escitalopram\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e9 / 29\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e1.17 (0.54 to 2.51)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.69\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eVenlafaxine vs Pooled SSRIs\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e9 / 53\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e1.27 (0.62 to 2.60)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.52\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eDuloxetine vs Pooled SSRIs\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e4 / 53\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.87 (0.31 to 2.43)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.79\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eDesvenlafaxine vs Pooled SSRIs\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e6 / 53\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.82 (0.35 to 1.90)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.64\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eSertraline vs Escitalopram (within-SSRI)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e8 / 29\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.73 (0.33 to 1.64)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.45\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u003cem\u003eNote. aHR = adjusted hazard ratio; CI = confidence interval. Events column shows exposed/comparator. All models adjusted for age, sex, baseline schizophrenia spectrum diagnosis, baseline substance use disorder, prior antipsychotic use, and prior SSRI/SNRI use. P-values are not adjusted for multiple comparisons; results are exploratory.\u003c/em\u003e\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eIn this large psychiatric cohort using an active-comparator landmark design, we found no evidence that initiating an SSRI rather than an SNRI altered the long-term incidence of new bipolar diagnoses. Across multiple outcome definitions, drug-level analyses, and adjusted models, the hazard of bipolar conversion was essentially identical for SSRIs and SNRIs. This null finding was consistent across three hierarchically defined outcome tiers, multiple sensitivity analyses, and exploratory drug-level analyses including specific examination of venlafaxine. Importantly, our results demonstrate that outcome definition exerts profound influence on observed event rates in specialty psychiatric settings, with anchored bipolar disorder occurring in 6.6% of patients compared with 46.3% for any bipolar diagnosis over the same follow-up period.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eComparison with Prior Literature\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eOur findings align closely with the recent multi-site study by Yoo et al. (2025), who also observed no class difference in bipolar conversion risk when comparing SSRI and SNRI initiators using rigorous new-user design methodology across multiple Korean centres (23). This convergence of null findings from independent cohorts employing similar methodological approaches strengthens the case that class-level differences may not exist in populations initially treated for presumed unipolar depression.\u003c/p\u003e\n\u003cp\u003eOur null findings contrast with some earlier signals in the literature. The seminal work by Post et al. (2006) reported switch rates of 29% for venlafaxine compared with 9% for sertraline and 10% for bupropion in patients with established bipolar disorder receiving adjunctive antidepressant treatment (11). However, this population (established bipolar disorder) differs fundamentally from ours (new depression without prior bipolar diagnosis), representing a higher-risk stratum. Patel et al. (2015) found that both SSRI (HR 1.34, 95% CI 1.18 to 1.52) and venlafaxine exposure (HR 1.35, 95% CI 1.07 to 1.70) were associated with subsequent mania or bipolar disorder diagnosis, but did not directly compare SSRI versus SNRI classes. Kim et al. (2020), in a nationwide Korean register, reported that the conversion rate was dramatically higher among patients prescribed SNRIs, with SNRI users having more than double the risk versus no antidepressant (15). However, comparison against no antidepressant, rather than an active comparator, may reflect confounding by indication.\u003c/p\u003e\n\u003cp\u003ePradier et al. (2021), using multi-site US health records (n = 67,807), found that short-term conversion rates within 3 months were numerically higher for SNRIs (1.93%) than SSRIs (1.23%) (16). This timeframe differs substantially from our longer follow-up, and the difference was not the primary focus of their predictive modelling study. Zhu et al. (2025) reported from a Chinese inpatient cohort that patients who eventually converted to bipolar had commonly received either venlafaxine or paroxetine without clear class differentiation, consistent with our findings (17).\u003c/p\u003e\n\u003cp\u003eThe most recent network meta-analysis by Oliva et al. (2025) evaluated switch to mania across randomised controlled trials of acute antidepressant treatment in bipolar depression. Whilst no individual antidepressant demonstrated statistically significantly elevated switch risk compared with placebo, venlafaxine consistently showed the highest relative risk estimates (8). Our exploratory drug-level analyses specifically examined venlafaxine and found no excess risk compared with pooled SSRIs (aHR 1.27, 95% CI 0.62 to 2.60) or individual SSRIs. This null finding persisted across sensitivity analyses, suggesting that when patients are initially treated for presumed unipolar depression with stringent outcome definitions, venlafaxine may not confer the differential risk observed in established bipolar populations.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eInterpretation of Tiered Outcome Findings\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003ePerhaps the most striking finding of this study concerns the seven-fold difference in event rates between anchored (6.6%) and any (46.3%) bipolar diagnosis outcomes. This pattern suggests diagnostic reassignment or clarification over time, rather than direct pharmacological induction of mania (24). In other words, many patients initially labelled with depressive episodes may later be found to meet bipolar criteria when additional symptoms emerge (3). Antidepressant exposure may coincide with, but not necessarily cause, this unfolding bipolarity.\u003c/p\u003e\n\u003cp\u003eIn tertiary psychiatric settings where patients receive ongoing longitudinal assessment, the documentation of any bipolar diagnosis may reflect provisional clinical impressions, diagnostic speculation, or billing necessities rather than definitive diagnostic conclusions (25). Our anchored definition, requiring both diagnostic persistence and treatment concordance, identifies patients for whom the clinical team has acted on the bipolar diagnosis through mood stabiliser initiation. This operational definition more closely approximates the clinically meaningful outcome of interest: patients whose illness trajectory has substantively changed to warrant bipolar-specific treatment. The observation that neither antidepressant class nor individual drugs were associated with anchored bipolar disorder suggests that, at least in this tertiary setting, class-specific pharmacological effects on mood destabilisation are not the primary driver of documented bipolar conversions.\u003c/p\u003e\n\u003cp\u003eThe divergence between covariate-adjusted and IPTW results for secondary outcomes merits comment. After achieving near-perfect covariate balance through propensity score weighting, Tier 2 and Tier 3 outcomes showed modest but statistically significant associations with SNRI initiation, whilst the primary anchored outcome remained robustly null. This pattern suggests that SNRI initiators may receive more frequent bipolar diagnoses overall, but these do not translate to clinically anchored conversions requiring mood stabiliser therapy. Possible explanations include differential diagnostic attention to patients receiving SNRIs, channelling of diagnostically complex patients to SNRIs, or heightened clinician vigilance for mood instability in SNRI-treated patients\u0026nbsp;(26). Regardless of mechanism, this finding reinforces the importance of using clinically anchored outcome definitions when evaluating antidepressant-associated bipolar conversion.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDrug-Level Findings and Venlafaxine\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe drug-level analyses provide additional reassurance regarding venlafaxine specifically. Despite theoretical concern about its noradrenergic effects and prior evidence of elevated switch risk in bipolar populations (11), venlafaxine showed no excess risk for anchored bipolar disorder compared with either pooled SSRIs or individual agents. The point estimates were modestly elevated (aHR 1.27 to 1.69 depending on comparator) but confidence intervals were wide and consistently included the null. These findings suggest that venlafaxine's hypothesised differential risk may be specific to patients with established or latent bipolar disorder rather than a generalised phenomenon in depression populations.\u003c/p\u003e\n\u003cp\u003eThe unexpected finding of lower risk with sertraline compared with escitalopram for secondary outcomes (Tier 2 aHR 0.60, p = 0.009; Tier 3 aHR 0.69, p = 0.01) warrants cautious interpretation. This within-SSRI difference was not hypothesised a priori and may reflect channelling bias, whereby escitalopram is preferentially prescribed to patients with features suggesting higher bipolar risk. Alternatively, this finding may represent a type I error given multiple comparisons. We present this result as hypothesis-generating rather than confirmatory.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePrior Antidepressant Use and Diagnostic Stability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAn unexpected finding was the strong protective association between prior antidepressant use and anchored bipolar disorder (HR 0.42, 95% CI 0.23 to 0.76). This association may reflect depletion of susceptibles or diagnostic stability mechanisms. Patients who have previously tolerated antidepressant treatment without mood destabilisation represent a lower-risk subgroup, having been effectively screened by prior exposure. Alternatively, patients with stable antidepressant histories may have more definitively established unipolar illness trajectories, making subsequent bipolar reclassification less likely. Clinically, this observation suggests that prior antidepressant tolerability may provide some reassurance when re-initiating treatment. This finding warrants replication and further investigation regarding potential predictive utility.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eStrengths and Limitations\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eMethodological strengths of this study include the large sample size, new-user design, and examination of tiered outcomes. The active-comparator design substantially reduces confounding by indication compared with studies comparing antidepressant users to non-users. By synchronising the cohort at antidepressant initiation and controlling for prior illness course, we reduce biases inherent in observational treatment research. The 90-day landmark eliminates immortal time bias that has affected prior pharmacoepidemiological studies of psychiatric outcomes. The tiered outcome hierarchy provides novel insights into how outcome definition influences findings in tertiary psychiatric settings. Drug-level exploratory analyses address concerns about within-class heterogeneity and specifically examine venlafaxine, the agent of greatest theoretical concern. Comprehensive sensitivity analyses, including inverse probability of treatment weighting that achieved excellent covariate balance (all standardised mean differences \u0026lt;0.04), demonstrated robustness of the null finding for the primary outcome.\u003c/p\u003e\n\u003cp\u003eNonetheless, as an observational study, limitations remain. Unmeasured confounding, such as family history or subtle mixed features, could influence both antidepressant choice and bipolar risk. Our tertiary psychiatric setting may limit generalisability to primary care populations where antidepressants are commonly prescribed. Residual misclassification is possible, though we attempted to maximise diagnostic capture through tiered definitions. Our anchored outcome definition, whilst increasing specificity, may miss true bipolar conversions managed without mood stabiliser initiation. The drug-level analyses were underpowered to detect modest effect sizes, with confidence intervals typically spanning 0.5 to 2.0 or wider, and multiple comparisons increase the probability of spurious findings. Death and loss to follow-up were not explicitly modelled as competing risks; however, in this cohort of adults aged 18–60 years receiving tertiary psychiatric care, mortality is expected to be low, and patients remained under care at our institution until their last recorded encounter. Finally, we used an intention-to-treat approach with exposure fixed at the landmark. This approach mirrors clinical decision-making at treatment initiation and avoids informative censoring from treatment changes, but may dilute exposure contrast if patients subsequently switched treatments, biasing results toward the null.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eClinical Implications\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eOur findings suggest that among patients with depression seen in psychiatric care, SSRIs and SNRIs carry comparable risk of subsequent bipolar diagnosis. If anything, the observed high incidence of bipolar coding appears driven by the natural diagnostic evolution of mood disorders rather than by class-specific drug effects. Clinically, this implies that class selection (SSRI versus SNRI) need not be guided by bipolarity concerns alone; focus should remain on symptom profile, tolerability, and patient factors. This does not contradict guideline recommendations to prefer SSRIs over SNRIs for established bipolar depression (12,13), where the evidence base differs substantially.\u0026nbsp;These findings apply to patients receiving tertiary psychiatric care and may not generalise to primary care settings where bipolar features may be less readily identified.\u003c/p\u003e\n\u003cp\u003eSecond, the absence of excess risk with venlafaxine in this depression cohort provides some reassurance for its use when clinically indicated. However, caution remains warranted in patients with features suggesting latent bipolar disorder, where guideline recommendations advise preferring agents with lower switch risk, such as SSRIs or bupropion (9).\u003c/p\u003e\n\u003cp\u003eThird, the dramatic influence of outcome definition on event rates should inform interpretation of diagnostic conversion studies. Studies employing any-diagnosis outcomes may substantially overestimate rates of clinically meaningful bipolar conversion. Vigilant assessment for mania remains important for all antidepressant-treated patients (9,13), but these results do not single out one class as uniquely prone to unmask bipolar disorder.\u003c/p\u003e\n\u003cp\u003eFourth, the protective association observed with prior antidepressant use suggests that patients with a history of antidepressant tolerability may carry lower risk of subsequent bipolar diagnosis, though this requires prospective validation.\u003c/p\u003e"},{"header":"Conclusions","content":"\u003cp\u003eIn this active-comparator landmark cohort study, neither SSRI versus SNRI class nor individual antidepressant choice, including venlafaxine, was associated with incident bipolar disorder when stringent, anchored outcome definitions were employed. The seven-fold difference in event rates between anchored and any-diagnosis outcomes highlights the profound influence of outcome definition on findings in specialty psychiatric settings. High event rates for less specific outcomes likely reflect diagnostic revision processes rather than pharmacological effects. These findings, consistent with recent multi-site studies, suggest that antidepressant class and drug selection may be less consequential for bipolar emergence than previously assumed when treating patients with presumed unipolar depression.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eEthics approval and consent to participate\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study was approved by the Ministry of Health and Prevention Research Ethics Committee (MOHAP REC), United Arab Emirates (Reference: MOHAP/DXB-REC/M.J.J/No.91/2024). The requirement for individual informed consent was waived by the ethics committee due to the retrospective nature of the study and the use of de-identified data from electronic health records.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent for publication\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAvailability of data and materials\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe data that support the findings of this study are available from Emirates Health Services, United Arab Emirates, but restrictions apply to the availability of these data, which were used under license for the current study, and so are not publicly available.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare that they have no competing interests.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthors' contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eSAB conceived and designed the study, acquired and analysed the data, interpreted the results, and drafted the manuscript. AOE contributed to the study design, data interpretation, and critically revised the manuscript for important intellectual content. ME contributed to data interpretation and critically revised the manuscript for important intellectual content. All authors read and approved the final manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgements\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eGitlin MJ. Antidepressants in bipolar depression: an enduring controversy. Int J Bipolar Disord [Internet]. 2018 Dec [cited 2026 Jan 12];6(1):25. Available from: https://journalbipolardisorders.springeropen.com/articles/10.1186/s40345-018-0133-9\u003c/li\u003e\n\u003cli\u003eBaldessarini RJ, Faedda GL, Offidani E, V\u0026aacute;zquez GH, Marangoni C, Serra G, et al. Antidepressant-associated mood-switching and transition from unipolar major depression to bipolar disorder: A review. Journal of Affective Disorders [Internet]. 2013 May [cited 2026 Jan 12];148(1):129\u0026ndash;35. Available from: https://linkinghub.elsevier.com/retrieve/pii/S0165032712007306\u003c/li\u003e\n\u003cli\u003eHirschfeld RMA, Lewis L, Vornik LA. Perceptions and Impact of Bipolar Disorder: How Far Have We Really Come? Results of the National Depressive and Manic-Depressive Association 2000 Survey of Individuals With Bipolar Disorder. J Clin Psychiatry [Internet]. 2003 Feb 15 [cited 2026 Jan 12];64(2):161\u0026ndash;74. Available from: https://www.psychiatrist.com/jcp/perceptions-impact-bipolar-disorder-far-really-results\u003c/li\u003e\n\u003cli\u003eBaryshnikov I, Sund R, Marttunen M, Svirskis T, Partonen T, Pirkola S, et al. Diagnostic conversion from unipolar depression to bipolar disorder, schizophrenia, or schizoaffective disorder: A nationwide prospective 15‐year register study on 43 495 inpatients. Bipolar Disorders [Internet]. 2020 Sept [cited 2026 Jan 12];22(6):582\u0026ndash;92. Available from: https://onlinelibrary.wiley.com/doi/10.1111/bdi.12929\u003c/li\u003e\n\u003cli\u003eJo YT, Joo SW, Kim H, Ahn S, Choi YJ, Choi W, et al. Diagnostic conversion from unipolar to bipolar affective disorder-A population-based study. J Affect Disord. 2022 Mar 15;301:448\u0026ndash;53.\u003c/li\u003e\n\u003cli\u003eMusliner KL, \u0026Oslash;stergaard SD. Patterns and predictors of conversion to bipolar disorder in 91 587 individuals diagnosed with unipolar depression. Acta Psychiatr Scand [Internet]. 2018 May [cited 2026 Jan 12];137(5):422\u0026ndash;32. Available from: https://onlinelibrary.wiley.com/doi/10.1111/acps.12869\u003c/li\u003e\n\u003cli\u003eKessing LV, Willer I, Andersen PK, Bukh JD. Rate and predictors of conversion from unipolar to bipolar disorder: A systematic review and meta‐analysis. Bipolar Disorders [Internet]. 2017 Aug [cited 2026 Jan 12];19(5):324\u0026ndash;35. Available from: https://onlinelibrary.wiley.com/doi/10.1111/bdi.12513\u003c/li\u003e\n\u003cli\u003eOliva V, De Prisco M, La Spina E, Paolucci S, Fico G, Anmella G, et al. Switch to mania after acute antidepressant treatment for bipolar depression: a systematic review and network meta-analysis of randomised controlled trials. EClinicalMedicine. 2025 Sept;87:103413.\u003c/li\u003e\n\u003cli\u003ePacchiarotti I, Bond DJ, Baldessarini RJ, Nolen WA, Grunze H, Licht RW, et al. The International Society for Bipolar Disorders (ISBD) Task Force Report on Antidepressant Use in Bipolar Disorders. AJP [Internet]. 2013 Nov [cited 2026 Jan 12];170(11):1249\u0026ndash;62. Available from: https://psychiatryonline.org/doi/10.1176/appi.ajp.2013.13020185\u003c/li\u003e\n\u003cli\u003eGoldberg JF, Truman CJ. Antidepressant‐induced mania: an overview of current controversies. Bipolar Disorders [Internet]. 2003 Dec [cited 2026 Jan 12];5(6):407\u0026ndash;20. Available from: https://onlinelibrary.wiley.com/doi/10.1046/j.1399-5618.2003.00067.x\u003c/li\u003e\n\u003cli\u003ePost RM, Altshuler LL, Leverich GS, Frye MA, Nolen WA, Kupka RW, et al. Mood switch in bipolar depression: comparison of adjunctive venlafaxine, bupropion and sertraline. Br J Psychiatry [Internet]. 2006 Aug [cited 2026 Jan 12];189(2):124\u0026ndash;31. Available from: https://www.cambridge.org/core/product/identifier/S0007125000170333/type/journal_article\u003c/li\u003e\n\u003cli\u003eKeramatian K, Chithra NK, Yatham LN. The CANMAT and ISBD Guidelines for the Treatment of Bipolar Disorder: Summary and a 2023 Update of Evidence. FOC [Internet]. 2023 Oct [cited 2026 Jan 12];21(4):344\u0026ndash;53. Available from: https://psychiatryonline.org/doi/10.1176/appi.focus.20230009\u003c/li\u003e\n\u003cli\u003eYatham LN, Kennedy SH, Parikh SV, Schaffer A, Bond DJ, Frey BN, et al. Canadian Network for Mood and Anxiety Treatments ( CANMAT ) and International Society for Bipolar Disorders ( ISBD ) 2018 guidelines for the management of patients with bipolar disorder. Bipolar Disorders [Internet]. 2018 Mar [cited 2026 Jan 12];20(2):97\u0026ndash;170. Available from: https://onlinelibrary.wiley.com/doi/10.1111/bdi.12609\u003c/li\u003e\n\u003cli\u003ePatel R, Reiss P, Shetty H, Broadbent M, Stewart R, McGuire P, et al. Do antidepressants increase the risk of mania and bipolar disorder in people with depression? A retrospective electronic case register cohort study. BMJ Open [Internet]. 2015 Dec [cited 2026 Jan 12];5(12):e008341. Available from: https://bmjopen.bmj.com/lookup/doi/10.1136/bmjopen-2015-008341\u003c/li\u003e\n\u003cli\u003eKim EY, Kim NW, Kim MJ, Yang BR, Rhee SJ, Park CHK, et al. Rate of diagnostic conversion to bipolar disorder in adults with unipolar depression and psychopharmacological treatment in the republic of Korea: A nationwide register-based study. J Affect Disord. 2020 Aug 1;273:240\u0026ndash;6.\u003c/li\u003e\n\u003cli\u003ePradier MF, Hughes MC, McCoy TH, Barroilhet SA, Doshi-Velez F, Perlis RH. Predicting change in diagnosis from major depression to bipolar disorder after antidepressant initiation. Neuropsychopharmacol [Internet]. 2021 Jan [cited 2026 Jan 12];46(2):455\u0026ndash;61. Available from: https://www.nature.com/articles/s41386-020-00838-x\u003c/li\u003e\n\u003cli\u003eZhu T, Kou R, Mu D, Hu Y, Yuan C, Yuan M, et al. Predicting Conversion From Unipolar Depression to Bipolar Disorder and Schizophrenia: A 10-Year Retrospective Cohort Study on 12,182 Inpatients. Depress Anxiety. 2025;2025:4048082.\u003c/li\u003e\n\u003cli\u003eSchneeweiss S. Sensitivity analysis and external adjustment for unmeasured confounders in epidemiologic database studies of therapeutics. Pharmacoepidemiology and Drug [Internet]. 2006 May [cited 2026 Jan 12];15(5):291\u0026ndash;303. Available from: https://onlinelibrary.wiley.com/doi/10.1002/pds.1200\u003c/li\u003e\n\u003cli\u003eLund JL, Richardson DB, St\u0026uuml;rmer T. The Active Comparator, New User Study Design in Pharmacoepidemiology: Historical Foundations and Contemporary Application. Curr Epidemiol Rep [Internet]. 2015 Dec [cited 2026 Jan 12];2(4):221\u0026ndash;8. Available from: http://link.springer.com/10.1007/s40471-015-0053-5\u003c/li\u003e\n\u003cli\u003eYoshida K, Solomon DH, Kim SC. Active-comparator design and new-user design in observational studies. Nat Rev Rheumatol [Internet]. 2015 July [cited 2026 Jan 12];11(7):437\u0026ndash;41. Available from: https://www.nature.com/articles/nrrheum.2015.30\u003c/li\u003e\n\u003cli\u003eSuissa S. Immortal Time Bias in Pharmacoepidemiology. American Journal of Epidemiology [Internet]. 2008 Jan 7 [cited 2026 Jan 12];167(4):492\u0026ndash;9. Available from: https://academic.oup.com/aje/article-lookup/doi/10.1093/aje/kwm324\u003c/li\u003e\n\u003cli\u003eMi X, Hammill BG, Curtis LH, Lai ECC, Setoguchi S. Use of the landmark method to address immortal person-time bias in comparative effectiveness research: a simulation study: Landmark method for immortal person-time. Statist Med [Internet]. 2016 Nov 20 [cited 2026 Jan 12];35(26):4824\u0026ndash;36. Available from: https://onlinelibrary.wiley.com/doi/10.1002/sim.7019\u003c/li\u003e\n\u003cli\u003eYoo KH, Lee KJ, Lee SM, Han C, Park RW, Jo YT. Comparative effectiveness of selective serotonin reuptake inhibitors versus serotonin-norepinephrine reuptake inhibitors in the risk of diagnostic conversion from unipolar depression to bipolar disorder. International Journal of Psychiatry in Clinical Practice [Internet]. 2025 Dec 14 [cited 2026 Jan 12];1\u0026ndash;9. Available from: https://www.tandfonline.com/doi/full/10.1080/13651501.2025.2600083\u003c/li\u003e\n\u003cli\u003eBaldessarini RJ, Bolzani L, Cruz N, Jones PB, Lai M, Lepri B, et al. Onset-age of bipolar disorders at six international sites. J Affect Disord. 2010 Feb;121(1\u0026ndash;2):143\u0026ndash;6.\u003c/li\u003e\n\u003cli\u003eDavis KAS, Coleman JRI, Adams M, Allen N, Breen G, Cullen B, et al. Mental health in UK Biobank - development, implementation and results from an online questionnaire completed by 157 366 participants: a reanalysis. BJPsych Open. 2020 Feb 6;6(2):e18.\u003c/li\u003e\n\u003cli\u003eSendor R, St\u0026uuml;rmer T. Core Concepts in Pharmacoepidemiology: Confounding by Indication and the Role of Active Comparators. Pharmacoepidemiol Drug Saf [Internet]. 2022 Mar [cited 2026 Jan 12];31(3):261\u0026ndash;9. Available from: https://pmc.ncbi.nlm.nih.gov/articles/PMC9121653/\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"international-journal-of-bipolar-disorders","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"ijbd","sideBox":"Learn more about [International Journal of Bipolar Disorders](http://journalbipolardisorders.springeropen.com/)","snPcode":"40345","submissionUrl":"https://submission.nature.com/new-submission/40345/3","title":"International Journal of Bipolar Disorders","twitterHandle":"@SpringerOpen","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"BMC/SO AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Bipolar disorder, antidepressants, selective serotonin reuptake inhibitor, serotonin-norepinephrine reuptake inhibitor, pharmacoepidemiology, cohort study, diagnostic conversion, United Arab Emirates","lastPublishedDoi":"10.21203/rs.3.rs-8580658/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8580658/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e\u003cstrong\u003eBackground:\u003c/strong\u003e Whether antidepressant class influences the risk of subsequent bipolar disorder diagnosis remains clinically relevant, though direct active-comparator studies are scarce. We compared incident bipolar disorder risk between selective serotonin reuptake inhibitor (SSRI) and serotonin-norepinephrine reuptake inhibitor (SNRI) initiators using tiered outcome definitions.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMethods: \u003c/strong\u003e\u0026nbsp;We conducted a retrospective cohort study from 2018 to 2025 using a 90-day landmark design at a tertiary psychiatric hospital in the United Arab Emirates. Adults aged 18–60 years initiating SSRIs or SNRIs for depressive disorders were followed for incident bipolar disorder. The primary outcome was anchored bipolar disorder, defined as two or more diagnoses at least 30 days apart plus initiation of a mood stabiliser. Secondary outcomes included confirmed bipolar disorder (two or more diagnoses) and any bipolar diagnosis. Cox proportional hazards models adjusted for age, sex, schizophrenia spectrum disorder, substance use disorder, and prior psychotropic use. Exploratory analyses examined individual antidepressants.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eResults: \u003c/strong\u003e\u0026nbsp;Among 1,095 antidepressant initiators (818 SSRI; 277 SNRI), 72 (6.6%) developed anchored bipolar disorder over 1,610.6 person-years. SNRI versus SSRI initiation was not associated with increased risk of anchored bipolar disorder (adjusted hazard ratio [aHR] 1.07, 95% CI 0.63–1.84, p = 0.80). Findings were consistent across secondary outcomes (confirmed: aHR 1.24, 95% CI 0.96–1.60; any diagnosis: aHR 1.12, 95% CI 0.92–1.37) and drug-level comparisons, including venlafaxine versus pooled SSRIs (aHR 1.27, 95% CI 0.62–2.60). Event rates varied seven-fold by outcome stringency (6.6% to 46.3%).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConclusions: \u003c/strong\u003e\u0026nbsp;Antidepressant class was not associated with incident bipolar disorder using stringent outcome definitions. The marked variation in event rates across outcome tiers suggests that high conversion rates reported in prior literature may partly reflect diagnostic revision rather than pharmacological effects.\u003c/p\u003e","manuscriptTitle":"SSRI versus SNRI Initiation and Incident Bipolar Disorder in Tertiary Psychiatric Care: An Active-Comparator Cohort Study from the United Arab Emirates","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-01-19 08:27:08","doi":"10.21203/rs.3.rs-8580658/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2026-04-02T14:50:51+00:00","index":"","fulltext":""},{"type":"reviewerAgreed","content":"196297948759082121151934173534891979105","date":"2026-01-16T21:58:46+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-01-14T18:22:27+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"296340857807475962226741591670524983302","date":"2026-01-14T17:10:23+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2026-01-14T16:58:40+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-01-13T04:04:03+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-01-13T04:01:34+00:00","index":"","fulltext":""},{"type":"submitted","content":"International Journal of Bipolar Disorders","date":"2026-01-12T10:17:08+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"international-journal-of-bipolar-disorders","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"ijbd","sideBox":"Learn more about [International Journal of Bipolar Disorders](http://journalbipolardisorders.springeropen.com/)","snPcode":"40345","submissionUrl":"https://submission.nature.com/new-submission/40345/3","title":"International Journal of Bipolar Disorders","twitterHandle":"@SpringerOpen","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"BMC/SO AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"94fd9854-90a2-4c27-9d03-60bfc597e1d4","owner":[],"postedDate":"January 19th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[],"tags":[],"updatedAt":"2026-05-04T15:59:03+00:00","versionOfRecord":{"articleIdentity":"rs-8580658","link":"https://doi.org/10.1186/s40345-026-00426-w","journal":{"identity":"international-journal-of-bipolar-disorders","isVorOnly":false,"title":"International Journal of Bipolar Disorders"},"publishedOn":"2026-04-28 15:57:04","publishedOnDateReadable":"April 28th, 2026"},"versionCreatedAt":"2026-01-19 08:27:08","video":"","vorDoi":"10.1186/s40345-026-00426-w","vorDoiUrl":"https://doi.org/10.1186/s40345-026-00426-w","workflowStages":[]},"version":"v1","identity":"rs-8580658","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8580658","identity":"rs-8580658","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}