Adult outpatients (18–65 years) diagnosed with bipolar I (BD-I) or bipolar II (BD-II) disorder were recruited from Ankara University Psychiatry Clinic. Two board-certified psychiatrists confirmed diagnoses per DSM-V criteria. Patients needed to be in euthymia, defined as two months without significant mood symptoms (HDRS-17 score <10 and YMRS score <7). Additional inclusion criteria was stable pharmacotherapy (no medication changes for ≥6 weeks) and sufficient literacy (minimum level of education must be eight years) for testing, having an established diagnosis of BD-I or BD-II through clinical interview, participants must have enough hours of sleeping before testing.

Exclusion criteria were: neurological illness (e.g., stroke, epilepsy), intellectual disability (IQ<70), current substance use, other primary DSM V comorbid disorders (notably borderline personality disorders), recent electroconvulsive therapy or long-acting antipsychotic use (within past 6 months), taking of any kind of benzodiazepine and other CNS inhibituary medications effecting cognitive functions, severe uncontrolled medical disease (e.g., diabetes with complications, cardiovascular disease), and significant sensory or motor deficits, any transient physiological and endocrinological condition that have potential to effect cognitive test performances (etc. fasting, hypoglycemia, physical exercise, menstrual period, alcohol or caffeine intake). All participants had normal or corrected vision and could comply with testing procedures.

Depressive and manic symptoms at screening were rated with the 17-item Hamilton Depression Rating Scale (HDRS-17) and Young Mania Rating Scale (YMRS), respectively. Both scales are widely used clinician-administered measures of depression and mania severity. Cognitive function was assessed using a comprehensive battery of neuropsychological tests, all of which were Turkish-validated versions. General intellectual ability, working memory, and processing speed were estimated from four subtests of the Wechsler Adult Intelligence Scale–Revised (WAIS-R): Information, Arithmetic, Digit Symbol, and Block Design. Verbal episodic memory was assessed by the Logical Memory subtest of the Wechsler Memory Scale–Revised (WMS-R) (13, 14). Executive function and cognitive flexibility were evaluated using the Wisconsin Card Sorting Test (WCST), which scores perseverative errors and the number of categories completed. All cognitive measures employed standard administration procedures and normative references, and their Turkish adaptations have demonstrated adequate reliability and validity (15).

The neuropsychological battery was administered by a trained psychologist in a quiet testing room. Testing sessions took place during morning hours to minimize fatigue and diurnal variation, and each session lasted approximately 2–3 hours (including breaks as needed). Mood ratings (HDRS-17, YMRS) were again recorded at testing to document euthymia. All instructions and scoring followed the standardized manuals for each instrument.

Data were analyzed using IBM SPSS Statistics, version 22. Demographic and clinical variables were compared between BD-I and BD-II groups: categorical variables by chi-square tests and continuous variables by independent-samples t-tests (normal data) or Mann–Whitney U tests (non-normal data). Normality was assessed with the Shapiro–Wilk test, and variances with Levene’s test. Comparison of three groups’ cognitive outcomes were evaluated with analysis of variance (ANOVA) across neuropsychological measures. If the overall ANOVA was significant, follow-up univariate ANCOVAs were performed for each cognitive score. ANCOVA covariates included age, education, illness duration, psychiatric hospitalizations, major depressive episodes, chlorpromazine-equivalent antipsychotic daily dose, and cumulative mood stabilizers (lithium and valproate) – based on their known influence on cognition. Statistical significance was p<0.05.

The study protocol was approved by the Ankara University Faculty of Medicine Clinical Research Ethics Committee. All participants provided written informed consent after receiving a complete description of study procedures. The research was conducted in accordance with the ethical standards of the Declaration of Helsinki.

A series of one-way analyses of variance (ANOVAs) was conducted to compare the three groups – Bipolar I (BD-I), Bipolar II (BD-II), and healthy controls – on each cognitive measure. Post hoc pairwise comparisons were conducted using the Fisher LSD test or the Games–Howell test (when variance homogeneity assumptions were violated) to determine which group differences were significant. Descriptive statistics and ANOVA outcomes for each cognitive domain are presented in Table 2. Detailed pairwise comparison results are provided in Table 3. A final ANCOVA model was also tested for verbal memory, incorporating key clinical covariates, and its summary is given in Table 4.

This study analyzed neurocognitive profiles of patients with Bipolar I disorder (BD-I), Bipolar II disorder (BD-II), and healthy controls using various neuropsychological assessments. In the unadjusted analysis the study found bipolar groups showed significant cognitive impairments in the several cognitive domains. Specifically, BD-I patients had significant deficits in verbal episodic memory compared to BD-II and controls. In an immediate verbal recall task, BD-I patients performed lowest, BD-II patients were intermediate, and controls performed highest, indicating a gradient of impairment (BD-I < BD-II < Control). Executive functioning, measured by the Wisconsin Card Sorting Test (WCST), also showed group differences. Both bipolar groups demonstrated impaired concept formation and greater error rates compared to healthy controls. BD-I and BD-II patients completed fewer WCST categories on average than controls, indicating reduced overall executive problem-solving success. The two bipolar groups did not significantly differ from each other in number of categories completed. In terms of errors, both BD-I and BD-II groups made a higher total number of non-perseverative (random) errors than controls. Notably, perseverative errors – a specific indicator of set-shifting difficulty – were significantly elevated only in the BD-I group. BD-I patients committed more perseverative errors than healthy controls and also more than BD-II patients, whereas the BD-II group’s perseverative error count was similar to that of controls. This suggests that set-shifting deficits were present in BD-I but not evident in BD-II in the unadjusted analysis. Group differences were further observed in processing speed. On the WAIS Digit Symbol Coding test (psychomotor processing speed), there was a significant main effect of group. BD-I patients had the slowest processing speed, with a mean Coding score significantly lower than both the BD-II and control groups. In contrast, BD-II patients performed comparably to healthy controls on the Coding task, indicating that processing speed was largely intact in BD-II. A similar pattern emerged in the visuospatial domain: BD-I patients scored lowest on the Block Design test, significantly below the high performance of BD-II patients and controls. Short-term working memory performance, measured by the WAIS Arithmetic subtest, showed no significant differences across groups, indicating preserved function regardless of diagnosis. Additionally, premorbid intellectual ability, assessed through general knowledge, was similar among all groups.

In summary, prior to adjustments, BD-I was associated with broad cognitive deficits (especially in memory, executive function, and processing speed), whereas BD-II showed milder impairments and even relative strengths (normal visuo-spatial ability and processing speed).

To determine whether the observed group differences were independent of illness burden and medication effects, we conducted ANCOVA models for each cognitive outcome, controlling for chlorpromazine-equivalent antipsychotic dose, lithium load, valproate load, total number of lifetime psychiatric hospitalizations, and number of past depressive episodes. After adjusting for these covariates, most of the initial group differences were no longer statistically significant, suggesting that many cognitive disparities were attributable to differences in treatment or illness severity rather than inherent to diagnostic groups. Crucially, the only cognitive measure that remained significantly different between BD-I, BD-II, and the control group after full adjustment was verbal memory performance. In the fully adjusted ANCOVA model for the WMS performance (controlling for antipsychotic, lithium, valproate, hospitalizations, and depressive episodes), a robust effect of diagnostic group persisted. After adjustment, many cognitive differences diminished and became insignificant. BD-I, BD-II patients and controls exhibited no notable differences in executive function (e.g., WCST), processing speed, working memory, or visuospatial skills after controlling for medication exposure and clinical factors.

Across cognitive domains, current antipsychotic exposure was the most consistent covariate influencing performance, particularly on executive function tasks assessed by WCST subtests, such as perseverative errors and completed categories. Notably, verbal memory performance remained significantly associated with diagnostic group status after full statistical adjustment, suggesting that the observed cognitive impairments in bipolar groups was not entirely attributable to pharmacological or illness-related confounders.

Our findings align with existing research on cognitive impairments in bipolar disorder, particularly regarding differences between BD-I and BD-II. Prior studies indicate that both subtypes exhibit significant neurocognitive deficits compared to healthy individuals (6). Cognitive deficits span multiple domains—including verbal episodic memory, visual memory, attention, executive function, and processing speed—and persist during remission in both subtypes (16, 17). However, inconsistencies persist regarding the extent and nature of these deficits. Our comprehensive neuropsychological testing and covariate adjustments may help clarify these discrepancies.

We observed pronounced verbal episodic memory impairments, with BD-I patients exhibiting the most severe deficits, followed by BD-II, and healthy controls performing best (BD-I < BD-II < HC). This gradient aligns with literature identifying verbal memory as a core deficit in bipolar disorder, especially in BD-I (1, 5, 6, 10, 18–20). The meta-analytic evidence shows large effect sizes for verbal memory measures in BD-I and smaller but significant impairments in BD-II (6). The study results were in line with these findings, as bipolar groups showed significantly poorer performance on the WMS Logical Memory task, even after adjusting for clinical and treatment variables. Longitudinal evidence suggests that memory deficits in BD-I may worsen over time, potentially due to neurotoxic effects of manic episodes or psychosis (21), pointing to a neuropathological basis involving hippocampal or fronto-hippocampal dysfunction. This interpretation is supported by neurobiological evidence linking verbal memory deficits to structural and functional alterations in hippocampal-prefrontal circuits that appear to be present across bipolar subtypes, though potentially more pronounced in BD-I (11, 20).

Bipolar I and Bipolar II differ in the pattern and magnitude of executive dysfunction, but findings vary across samples and measures. Meta-analytic and comparative studies have indicated greater and more widespread executive deficits in BD-I, whereas BD-II shows smaller impairments (6, 11, 22). The meta-analyses have reported that euthymic bipolar patients had cognitive impairment in executive domains (23, 24). The present study's findings were in line with previous studies on executive function in BD-I and BD-II patients, as demonstrated by impaired Wisconsin Card Sorting Test (WCST) performance, with fewer categories completed and more non-perseverative errors compared to healthy controls. Only BD-I patients exhibited significant perseverative errors, suggesting greater cognitive inflexibility, whereas BD-II patients performed similarly to controls in this domain. However, after adjusting for clinical covariates—such as illness duration, hospitalizations, and antipsychotic exposure—executive function differences between BD-I and BD-II disappeared, indicating that these disparities reflect illness severity and treatment effects rather than intrinsic subtype differences (10–12, 24, 25).

The broader literature has provided mixed findings on the processing speed performance of bipolar patients. While some studies reported processing speed deficits in BD-II (26), others have found relatively preserved performance, particularly in euthymic patients with shorter illness duration [27]. The meta-analytic evidence suggested that processing speed impairments may be more consistently present in BD-I, with effect sizes varying considerably across studies for BD-II [6]. In the present study, processing speed, assessed via the Digit Symbol Coding task, was impaired in both BD-I and BD-II compared to controls, with BD-I initially slower than BD-II. This difference diminished after adjusting for covariates, notably antipsychotic dose and clinical factors. For working memory, both BD-I and BD-II performed within normative ranges on the WAIS Arithmetic subtest, and differences on the Digit Span test were non-significant post-adjustment, contrasting with prior reports of working memory impairment in bipolar patients (27, 28).

One of the most striking aspects of our findings was the dramatic reduction in cognitive differences between bipolar subtypes and controls after controlling for medication exposure and clinical variables. This observation has profound implications for interpreting the existing literature and aligns with emerging evidence suggesting that many apparent cognitive differences between BD-I and BD-II may be confounded by differential treatment patterns and illness characteristics.

The systematic review evidence supports this interpretation, with several studies reporting that controlling for antipsychotic exposure, lithium use, and illness severity substantially attenuates or eliminates cognitive differences between bipolar subtypes (12, 29). Our finding that antipsychotic exposure was the most consistent predictor of cognitive performance across domains is particularly important, as BD-I patients in our sample were more frequently treated with antipsychotic medications compared to BD-II patients, who more commonly received lamotrigine or lower-dose antipsychotic treatments.

This pattern reflects real-world clinical practice, where BD-I patients typically require more intensive pharmacological intervention due to the severity of manic episodes and higher rates of psychotic features. However, it raises important questions about whether cognitive differences attributed to diagnostic subtypes may actually reflect differential treatment effects. The meta-analytic evidence showing that lithium use was associated with larger effect sizes for cognitive flexibility and episodic memory further supports the critical importance of considering medication effects in cognitive research (6).

Our findings, integrated with the broader literature, have several important clinical implications. First, the identification of verbal memory as a core deficit that persists despite controlling for medication and clinical factors suggests that memory assessment should be a priority in both BD-I and BD-II patients. The meta-analytic evidence supports the use of specific instruments, including the Trail Making Test, Hayling Test, Digit Span Total, and Category Fluency, as sensitive measures for cognitive assessment in bipolar disorder (6) Second, the substantial impact of medication on cognitive performance, particularly antipsychotic exposure, emphasizes the need for careful consideration of treatment effects when interpreting cognitive test results. Clinicians should be aware that cognitive deficits may partially reflect medication side effects rather than intrinsic illness features, particularly in BD-I patients who more commonly receive antipsychotic treatment. Third, the evidence for more selective cognitive impairments in BD-II suggests that cognitive rehabilitation interventions may need to be tailored to specific deficit profiles. While BD-I patients may benefit from comprehensive cognitive remediation targeting multiple domains, BD-II patients might benefit from more focused interventions targeting specific areas of impairment while leveraging preserved cognitive strengths.

The substantial heterogeneity observed in meta-analytic studies of bipolar cognition highlights the importance of methodological factors in interpreting cognitive differences between subtypes. Our study's comprehensive approach to controlling potential confounders may explain why we observed fewer significant differences between BD-I and BD-II compared to some previous studies. Several methodological factors likely contribute to inconsistent findings in the literature. the smaller representation of BD-II patients in most studies (our BD-II n=24 vs. BD-I n=56 reflects this common imbalance) may limit power to detect subtle differences or lead to unstable effect size estimates. Differences in diagnostic criteria, mood state at testing, and medication status across studies create substantial heterogeneity, complicating the interpretation of pooled analyses. The systematic review evidence indicates that studies testing patients during euthymic states tend to show smaller effect sizes and less pronounced differences between subtypes (8, 30). This is consistent with our findings, as all patients were rigorously assessed to ensure euthymic status at the time of testing. Additionally, the duration of euthymia may be critical, as some cognitive deficits may improve with sustained mood stability (31).

This study's strength is its focus on euthymic patients under monotherapy, reducing confounding factors from mood state and polypharmacy. We utilized validated cognitive tests and maintained strict statistical control for clinical variables—antipsychotic dose, lithium/valproate load, hospitalizations, and depressive episodes. Only verbal memory impairment in BD-I remained significant after adjustment, indicating a possible trait-like deficit.

This is the first study to systematically include chlorpromazine-equivalent antipsychotic dosage, lithium and valproate load, prior hospitalizations, and lifetime depressive episodes as covariates in analyzing neurocognitive performance in BD-I and BD-II populations. By uniformly applying this adjustment across domains like verbal memory, executive function, processing speed, and visuospatial construction, we distinguished robust cognitive deficits from those likely influenced by treatment exposure or illness burden.

Our study’s limitations comprise its cross-sectional design, medication-related confounding, moderate sample size (particularly for BD-II), and a cognitive test battery that could not cover all aspects of cognition. These limitations temper our conclusions but highlight areas for careful interpretation and inquiry. Future longitudinal studies with larger, stratified samples and broader assessments (e.g., neuroimaging, social cognition) are necessary to establish the stability and causality of cognitive deficits in bipolar subtypes (32, 33).

To clarify the trajectory and causes of cognitive decline in bipolar disorder, future studies should adopt longitudinal designs with repeated cognitive assessments, detailed medication tracking, and neurobiological measures (e.g., imaging, inflammation). This approach will determine whether deficits like verbal memory impairment in BD-I are stable traits or consequences of illness progression and treatment, while identifying critical intervention periods.

This study highlights cognitive impairment as a core feature of bipolar disorder, demanding clinical attention alongside mood symptoms. While BD-I shows more severe deficits—especially in verbal episodic memory—many differences between BD-I and BD-II diminish after accounting for illness severity and treatment. Verbal memory impairment in BD-I suggests a trait-level vulnerability. Clinically, personalized care is essential: BD-I may require targeted cognitive interventions, while BD-II patients benefit from depression management to prevent cognitive decline. Treating cognition as a vital sign can enhance recovery and quality of life. Future research should explore biological mechanisms and interventions to improve cognitive outcomes.