Long-Acting Injections in Bipolar Disorder: Relapse Polarity Matters

The Psychiatric Record | Clinical Evidence

A maintenance treatment can help keep someone out of hospital for mania while leaving a different question about depression. That distinction should shape how clinicians read an overall relapse result, and how they explain it to a patient.

The useful question is specific: which episodes does this evidence help us prevent?

Deng and colleagues examined Hong Kong records from 2004 through 2023. Of 17,841 people with bipolar disorder, 2,086 received both oral and injectable antipsychotics. Their self-controlled analysis compared treatment periods within individuals; each outcome analysis included people who experienced that outcome. These are selected patients who had been prescribed both formulations, not a representative sample of everyone with bipolar disorder. Results below are expressed as an adjusted incidence rate ratio (aIRR), with a 95% confidence interval (CI).

Patient counts are outcome-specific, not numbers of admissions. There were only 11 mixed-episode admissions during injectable-only periods, compared with 167 during oral-only periods; exposure time differed.

The pooled depressive result was nonsignificant. Separately, injectable aripiprazole was associated with fewer depressive admissions than oral aripiprazole: 0.20 (0.04 to 0.95). That estimate is a signal to investigate, not sufficient evidence for a depression-prevention claim. Its interval is wide, its upper bound is close to 1, and it comes from multiple agent-by-outcome comparisons. The main report does not describe a multiplicity correction.

The manic estimate corresponds to a 49% lower adjusted admission rate. It is an association, not a 49-percentage-point reduction in an individual's probability of relapse. The depressive interval deserves equal attention: it includes both a small reduction and a substantial increase. Describing it simply as “no difference” hides information clinicians need.

Extrapyramidal symptoms were more frequent across the full treatment period: 2.95 (1.40 to 6.22). Beyond 90 days, the estimate was 2.01 (0.85 to 4.75). The point estimate remained approximately doubled, and the interval extended to nearly fivefold. Loss of statistical significance does not establish that the excess risk resolved. Deng et al., 2026.

The Hong Kong study was supported by the National Natural Science Foundation of China; several authors disclosed industry grants, consulting, or other financial relationships. Funding and competing interests.

A 52-week randomized withdrawal trial published in 2017 (266 patients) provides relevant context. Otsuka and Lundbeck funded the trial; the sponsor designed and conducted it. Patients with bipolar I disorder entered during a manic episode, stabilized on oral and then monthly injectable aripiprazole, and were randomized to continued injections or placebo. Mood recurrence occurred in 26.5% with continued injections versus 51.1% with placebo; the hazard ratio was 0.45 (95% CI, 0.30 to 0.68). The benefit was principally in manic recurrence; depressive recurrence did not differ significantly, with 39 depressive episodes overall (P = .864). Calabrese et al., 2017.

This supports maintenance efficacy in a selected population that had already responded to and tolerated treatment. Because the comparator was placebo, it does not establish superiority over continued oral aripiprazole. The observational aripiprazole finding lacks confirmation from this trial, but the results are not a direct contradiction: comparator, patient selection, and depressive outcome definition differed. Neither study establishes adherence as the mechanism of the observed difference.

The Hong Kong models adjusted for concomitant medicines, including mood stabilizers. The manic association persisted beyond 90 days: 0.48 (95% CI, 0.40 to 0.57). In separate analyses, all antipsychotics were associated with fewer manic and depressive admissions during treatment than during the 30 days before initiation. The authors reported that sensitivity analyses generally agreed with the main findings. Severity, adherence, and bipolar subtype were unavailable. Study methods and limitations.

Those distinctions matter. A hospitalization can trigger a treatment change, so the periods being compared may differ in clinical circumstances. Adjustments strengthen the analysis without making treatment assignment random. The pre-exposure comparison uses a period that can be clinically unstable; improvement afterward does not itself establish injectable superiority over oral treatment.

Formulation and medication choice are separate clinical questions. A pooled estimate can inform the first conversation, but the selected medicine still needs evidence that matches the patient's treatment goals. Differences in the medicines used during oral and injectable periods are one possible contributor to the pooled depressive result. The study does not establish that switching formulations removed depression protection or caused worsening.

Hospitalization captures only depressive episodes that lead to admission, not the full burden of depressive relapse. A person can avoid admission while remaining depressed, unable to work, or dissatisfied with treatment.

Consider two fictional examples, offered as communication prompts rather than prescribing recommendations.

For someone whose recent course includes repeated manic admissions after interruptions in medication, the discussion might begin: “Our priority is preventing another manic episode requiring hospital care. We will review whether this formulation makes treatment easier to sustain, alongside its burdens and adverse effects.”

For someone whose main burden is persistent or recurrent depression, the discussion needs another sentence: “We should identify what each part of the maintenance plan is expected to do for depression, and track depressive symptoms and functioning even if hospital admissions decline.”

In both conversations, the patient's reasons for preferring or declining an injection belong in the decision. Convenience, prior experience, tolerability, and the practical burden of appointments deserve explicit discussion.

“Relapse prevention” becomes more useful when it names the relapse being prevented. The maintenance review should do the same.

Reading note: The Hong Kong study's main full text and outcome tables, and the aripiprazole trial's full text, were reviewed. Findings are distinguished from the editorial communication examples above.

Educational Disclaimer: The Psychiatric Record provides general educational information for psychiatric and mental-health professionals. Content does not constitute medical, legal, regulatory, compliance, billing, or other professional advice; does not establish a standard of care; and is not a substitute for independent professional judgment. Appropriate assessment and treatment depend on the individual, setting, and current authoritative guidance. This article does not recommend starting, stopping, or switching any medication or formulation. Approved indications for long-acting injectable antipsychotics vary by product and jurisdiction. The patient conversations are fictional illustrations; no individual is depicted.