Open questions — bipolar disorder¶
Last curated: 2026-09-03
These are research questions, not treatment recommendations. Stable IDs should be retained when a question is refined or answered.
Tier 1 — practice-changing and designable now¶
OQ-1 — Prospective polarity prediction at first depression¶
Can a preregistered model using longitudinal clinical features, family history and repeated sleep/activity measures predict later mania or hypomania with useful calibration across community, outpatient and inpatient settings? Conversion estimates range from 5.84% over 13 years in Swedish registers to 12.4% over three years in a Brazilian cohort, showing why transportability matters (Rhee 2023, PMID 37427550; Oliveira 2021, PMID 33493732).
OQ-2 — The hypomania threshold¶
What minimum duration, change-from-baseline and impairment rule for hypomania best predicts recurrence, familial loading and treatment response without materially increasing false-positive diagnosis? The four-day boundary has limited empirical support, but shorter-duration phenotypes are not yet tied to a validated replacement threshold (Miller 2016, PMID 26830885; Vieta 2008, PMID 18199235).
OQ-3 — A harmonized mixed-state definition¶
Can DSM-5, ICD-11 and broader opposite-polarity definitions be compared prospectively against relapse, suicide attempt and treatment response? Updated prevalence estimates vary from 3.6% to 23.8% in MDD depending on the rule, while mixed features in bipolar mania/hypomania correlate with rapid cycling and suicide-attempt history (Grasso 2026, PMID 41448395; Bartoli 2020, PMID 32697704).
OQ-4 — Acute mania: efficacy versus total clinical value¶
Which antimanic strategy maximizes rapid response while minimizing sedation, akathisia, extrapyramidal effects, metabolic harm and subsequent depression? Network meta-analysis ranks multiple effective agents, but trials average about four weeks and cannot establish long-term net benefit (Kishi 2022, PMID 34642461; Cipriani 2011, PMID 21851976).
OQ-5 — Mixed presentations as a prospective trial population¶
Do purpose-designed trials in mixed mania and mixed bipolar depression reproduce benefits inferred from post-hoc subgroups? Only two older trials prospectively focused on mixed mania/hypomania, and most depressive evidence is antipsychotic-centered (Takeshima 2017, PMID 28004626; Xiao 2025, PMID 40808264).
OQ-6 — Clinically meaningful bipolar-depression benefit¶
Which acute therapies produce durable functional recovery, not only a small mean symptom-scale separation at six to eight weeks? A 101-trial network meta-analysis found standardized mean differences from 0.16 to 0.41 for moderate-confidence agents, leaving individual absolute benefit and persistence uncertain (Yildiz 2023, PMID 37595997).
OQ-7 — Antidepressant responder and harm subgroups¶
Can a trial prospectively identify patients who benefit from adjunctive antidepressants without switch, mixed-symptom worsening or cycle acceleration? STEP-BD found durable recovery of 23.5% versus 27.3% with placebo, while BEAM-BD found that 52-week continuation reduced depressive recurrence (HR 0.43, 95% CI 0.25–0.75) but numerically increased mania/hypomania (HR 2.28, 0.86–6.08) without significantly improving the composite outcome (Sachs 2007, PMID 17392295; Yatham 2023, PMID 37530824).
OQ-8 — Maintenance by polarity¶
Can long-term pragmatic randomization match lithium, lamotrigine, valproate and antipsychotics to prospectively defined predominant polarity while measuring both manic and depressive recurrence? Current network evidence ranks average effects, and enriched trials may favor the stabilization drug (Miura 2014, PMID 26360999; Bowden 2003, PMID 12695317).
OQ-9 — Safe maintenance discontinuation¶
Who can taper maintenance pharmacotherapy with acceptably low recurrence risk, and what taper rate and monitoring strategy is safest? Maintenance reduced six-month recurrence to RR 0.61, yet 47.3% of discontinuers remained recurrence-free, indicating heterogeneity that group averages cannot resolve (Kishi 2021, PMID 33046156).
OQ-10 — Lithium target concentration¶
What serum concentration maximizes recurrence prevention for a given age, polarity, renal risk and adverse-effect burden? A dose-response synthesis modeled different concentration requirements for depression and mania prevention, while kidney synthesis found lower eGFR and faster annual decline but highly uncertain chronic-kidney-disease event risk (Hsu 2022, PMID 35158229; Macaron 2026, PMID 41727809).
OQ-11 — Lithium and suicide causality¶
How large is lithium’s suicide-preventive effect when tested with enough events and retention to be conclusive? A 2013 synthesis estimated suicide OR 0.13, whereas newer randomized syntheses remain statistically imprecise and nonsignificant (Cipriani 2013, PMID 23814104; Nabi 2022, PMID 36111461; Wang 2025, PMID 40441661).
OQ-12 — Renal risk stratification¶
Which baseline and time-varying measures predict irreversible kidney decline during lithium treatment, and which monitoring or dose changes preserve benefit? Impaired kidney function prevalence was 25.5% across heterogeneous studies, with OR 2.09 versus non-lithium groups (Schoretsanitis 2022, PMID 34783413).
OQ-13 — Psychotherapy component and dose¶
Which components—relapse-signature work, family communication, rhythm stabilization, cognitive strategies or medication support—drive benefit, for whom, and at what session dose? Group psychoeducation reduced relapse (OR 0.43, 95% CI 0.28–0.62), yet a head-to-head trial found structured skills therapy and supportive emotion-focused therapy nearly identical (49% versus 46% relapse), making common factors and dose plausible alternatives to branded technique (Tan 2022, PMID 36421612; Hautzinger 2024, PMID 38837133).
OQ-14 — Digital self-management that changes relapse¶
Can a privacy-preserving digital intervention reduce syndromal recurrence in an unselected bipolar population? LiveWell’s primary relapse HR was 0.65 (95% CI 0.39–1.09), SmartBipolar found no six-month benefit, and a 2026 review found no bipolar-specific symptom or readmission benefit from monitoring alone; a positive circadian-feedback trial combined bipolar and major depressive disorders and therefore does not settle the bipolar-specific question (Goulding 2023, PMID 36542401; Faurholt-Jepsen 2026, PMID 41865316; Astill Wright 2026, PMID 41499681; Yeom 2026, PMID 42337416).
OQ-15 — Cardiovascular prevention as a bipolar outcome¶
Does an integrated cardiometabolic pathway reduce myocardial infarction, stroke, heart failure and mortality rather than only weight or laboratory surrogates? Bipolar disorder is associated with stroke incidence HR 1.43, incident heart-failure RR 1.95 (four pooled cohorts; I²=80.1%), and systemic shortfalls in cardiovascular screening/treatment (Yuan 2022, PMID 34052937; Dong 2026, PMID 42668911; Solmi 2021, PMID 34256605).
OQ-16 — Perinatal comparative effectiveness¶
What strategy best balances maternal recurrence against fetal/neonatal risk across pregnancy, delivery and postpartum? Discontinuation doubled pregnancy recurrence in a prospective cohort; postpartum relapse was 66% without medication versus 23% with prophylaxis, while lithium cardiac-malformation risk was dose-dependent (Viguera 2007, PMID 18056236; Wesseloo 2016, PMID 26514657; Patorno 2017, PMID 28591541).
OQ-17 — Postpartum lithium dosing¶
Should lithium routinely be reduced or withheld near delivery, or should dosing instead be individualized using maternal levels, hydration and neonatal monitoring? One cohort found no delivery-related rise in level/dose ratio, while another linked higher neonatal levels with transient hypotonia (Molenaar 2021, PMID 32526071; Imaz 2024, PMID 39187196).
OQ-18 — Closing the mortality gap¶
Which service-level intervention reduces both unnatural and natural-cause mortality? Meta-analysis found all-cause SMR 2.05, suicide SMR 14.44 and circulatory SMR 1.73, implying that suicide prevention alone is insufficient (Hayes 2015, PMID 25735195).
Tier 2 — enabling science and longer-horizon questions¶
OQ-19 — From polygenic association to treatment selection¶
Can polygenic and rare-variant information predict phase-specific response beyond clinical history across ancestries? GWAS has identified 64 loci, European-derived scores explained only 1.27% of liability variance in one Han Chinese study, and a 3,915-case Han Chinese CNV analysis found rare-deletion enrichment with replication of only 3q29 and 15q11.2 among 12 European CNV loci (Mullins 2021, PMID 34002096; Li 2021, PMID 33263727; Wu 2026, PMID 42661060).
OQ-20 — State, trait and treatment in neuroimaging¶
Which large-scale brain differences persist across mood states and medication exposures, and do they predict outcomes at the individual level? ENIGMA and multimodal meta-analyses find distributed group differences but not a diagnostic classifier (Ching 2022, PMID 32725849; Chen 2022, PMID 36093787).
OQ-21 — Inflammation as mechanism or correlate¶
Does an inflammatory subgroup causally contribute to episodes, or do CRP differences reflect mood state, adiposity, smoking and treatment? CRP elevation was largest in mania (SMD 0.73), small in euthymia and nonsignificant in depression (Dargél 2015, PMID 25742201).
OQ-22 — Cross-disorder biology versus diagnostic specificity¶
How should extensive genetic overlap with schizophrenia, MDD and ADHD be converted into clinically useful dimensional models without erasing episode-based treatment distinctions? Cross-disorder analysis found 109 pleiotropic loci, and ADHD–bipolar SNP correlation was 0.64 (Cross-Disorder Group 2019, PMID 31835028; van Hulzen 2017, PMID 27890468).
OQ-23 — Passive relapse prediction and privacy¶
Can activity, sleep, communication and speech models prospectively predict a transition early enough to change care, with external validation and acceptable data governance? A 52-study review reported AUC 0.70–0.88 one to four weeks ahead, but 75% of studies were high risk of bias and predominantly internally validated; qualitative synthesis also documents burden, worsened mood/anxiety, privacy and clinician-liability concerns (Fang 2026, PMID 42106724; Astill Wright 2025, PMID 41105870).
OQ-24 — Comorbidity-aware trials¶
Do integrated treatments outperform parallel single-disorder care for bipolar disorder with ADHD, anxiety or substance use? ADHD is present in an estimated 17.11% of adults with bipolar disorder, while adjunctive drug effects in bipolar–substance-use trials are small and low certainty (Schiweck 2021, PMID 33515607; Radua 2024, PMID 37689524).
Dots not yet connected¶
The table records junctions not joined by a study retrieved in the audit searches through the 2026-09-03 evidence-broadening pass. It is a search agenda, not proof that no such study exists.
| # | Dot A | Dot B | The missing junction | Powers |
|---|---|---|---|---|
| D-1 | Mixed features predict rapid cycling and suicide-attempt history (PMID 32697704) | Antidepressant outcomes differ by mixed symptoms and drug (PMIDs: 17728419, 16880481) | Prospective mixed-definition × randomized antidepressant interaction with switch and function endpoints | OQ-3, OQ-7 |
| D-2 | Polygenic architecture differs partly between bipolar I and II (PMID 34002096) | Maintenance benefit differs by polarity (PMIDs: 12695317, 14628976) | Ancestry-diverse pharmacogenomic maintenance trial stratified by episode polarity | OQ-8, OQ-19 |
| D-3 | Mania-linked CRP elevation is state-dependent (PMID 25742201) | Acute antimanic agents differ substantially in tolerability (PMID 34642461) | Repeated inflammatory sampling within randomized mania treatment tied to response and metabolic harm | OQ-4, OQ-21 |
| D-4 | Digital tools measure sleep/activity continuously (PMID 40408762) | Maintenance discontinuation produces heterogeneous recurrence (PMID 33046156) | Sensor-guided randomized taper with predefined rescue thresholds | OQ-9, OQ-23 |
| D-5 | Executive impairment predicts later function (PMID 33792890) | Psychoeducation reduces recurrence (PMID 19252157) | Cognitive phenotype × psychotherapy-component trial with occupational recovery endpoint | OQ-13 |
| D-6 | Metabolic syndrome affects about one-third of bipolar cohorts (PMID 40921519) | Maintenance drugs differ in relapse efficacy and metabolic burden (PMID 26360999) | Pragmatic maintenance randomization powered for major cardiovascular events and recurrence jointly | OQ-8, OQ-15 |
| D-7 | Lithium benefit varies with serum concentration (PMID 34227095) | Renal impairment risk rises with duration (PMID 34783413) | Individualized concentration algorithm using recurrent-event and kidney-slope co-primary outcomes | OQ-10, OQ-12 |
| D-8 | Pregnancy discontinuation sharply increases recurrence (PMID 18056236) | Lithium malformation risk shows a dose gradient (PMID 28591541) | Target-concentration trial/registry emulation spanning conception through postpartum | OQ-16 |
| D-9 | Cardiovascular care is delivered less often to people with mental disorders (PMID 34256605) | Occupational function is impaired even below syndromal thresholds (PMID 36090375) | Integrated physical-health intervention measuring employment and cardiovascular events together | OQ-15, OQ-18 |
| D-10 | ADHD–bipolar comorbidity is common (PMID 33515607) | Bipolar digital phenotyping targets episodic change (PMID 40408762) | Within-person digital study separating persistent ADHD signal from emerging mood episodes | OQ-23, OQ-24 |
| D-11 | Patient/family psychoeducation reduces recurrence (PMIDs: 12695318, 18452447) | Perinatal relapse clusters in a short high-risk window (PMID 26514657) | Perinatal family-focused relapse-prevention trial integrated with pharmacologic planning | OQ-13, OQ-16 |
| D-12 | Observational mortality strongly favors suicide prevention efforts (PMID 25735195) | Randomized lithium suicide estimates are imprecise (PMID 40441661) | Multi-dataset target-trial emulation with prespecified confounding controls and cause-specific mortality | OQ-11, OQ-18 |
| D-13 | Lithium concentrations show polarity-dependent recurrence gradients (PMID 35158229) | Lithium exposure is associated with lower eGFR and faster decline (PMID 41727809) | Adaptive concentration trial with depressive/manic recurrence and kidney-slope co-primary outcomes | OQ-10, OQ-12 |
| D-14 | Passive sensing studies report internally validated relapse AUCs of 0.70–0.88 (PMID 42106724) | Users report both insight and monitoring-related distress/privacy concerns (PMID 41105870) | Patient-governed alert trial measuring clinical utility, false alarms, burden and autonomy together | OQ-14, OQ-23 |