Effects of low-dose aspirin in bipolar disorder: study protocol for a randomised controlled trial (the A-Bipolar RCT)

Abstract

Introduction

Accumulating data support the association between increased inflammation and bipolar disorder (BD), and preliminary data suggest that augmentation with low-dose aspirin (LDA) may protect against the onset and deterioration of BD via anti-inflammatory pathways. The A-bipolar randomised controlled trial (RCT) aims to investigate whether adding LDA to standard treatment improves day-to-day mood instability (MI) in BD.

Methods and analysis

A two-arm, triple-blind, parallel-group, superiority RCT including 250 patients with newly diagnosed BD treated at the Copenhagen Affective Disorder Clinic, Denmark. Participants are randomised 1:1 to either 150 mg of acetylsalicylic acid daily (LDA) or a placebo for six months in addition to their regular treatment. Mood instability, calculated from daily smartphone-based mood evaluations, is the primary outcome measure due to its internal validity as a real-life measure for patients and external validity as it reflects patients’ illness severity and functioning. Analyses will be conducted as intention-to-treat analyses using a linear mixed model including time (categorical) and the time–treatment interaction as fixed effects and with an unstructured covariance pattern to account for repeated measurements on each study participant. The trial is Good Clinical Practice monitored.

Ethics and dissemination

The Danish Research Ethics Committee (H-21014515) and the data agency, Capital Region of Copenhagen (P-2021-576) approved the trial. Results will be published in peer-reviewed journals.

Trial registration number

NCT05035316.

STRENGTHS AND LIMITATIONS OF THIS STUDY.

• A triple-blind, placebo-controlled randomised trial comparing augmentation treatment with low-dose aspirin versus placebo in bipolar disorder including analysis of inflammatory biomarkers.

• Mood instability is chosen as the primary outcome measure due to its internal validity as a real-life measure for patients and external validity as it reflects patients’ illness severity and functioning.

• Possible selection of participants favourably disposed towards pharmacological treatment.

• The trial’s requirements may result in the selection of stable patients and thus impair the detection of a true intervention effect.

Introduction

Bipolar disorder (BD) is a common psychiatric disorder with a 1%–2% prevalence. With an early age of onset and a lifelong course characterised by a high risk of recurrence of manic and depressive episodes, BD ranks among the six leading causes of disability worldwide.¹ BD is associated with a lifelong elevated risk of suicide, and a decreased life expectancy of 8–12 years primarily due to physical illnesses.^{2 3} Further, BD is associated with decreased socioeconomic functioning,⁴ with 30%–60% of patients having impaired psychosocial functioning even during remission.^{5 6} Cognitive dysfunction is present in 30%–70% of remitted patients,⁷,⁹ and available treatments fail to alleviate this key contributor to patients’ psychosocial disability.^{10 11} Current treatments do not provide adequate affective symptom control, as the affective symptoms persist half of the time,¹² alternating between depressive episodes and manic episodes three times per year on average,¹³ with many patients experiencing a progressive worsening.¹¹⁴,¹⁷ Thus, more effective treatment strategies are needed for the prevention and treatment of affective episodes.

Bipolar disorder is increasingly conceived as a multisystem disorder with pathophysiological abnormalities involving immune dysregulation with low-grade inflammation, enhanced oxidative stress, neurotrophic deficiencies and telomere shortening.¹⁸ Several meta-analyses have confirmed that levels of proinflammatory mediators in blood and cerebrospinal fluid are elevated in at least a subset of BD patients.¹⁹,²³ Other inflammation-related biomarkers are abnormal in BD compared with healthy controls (HCs), including higher oxidative stress,²⁴,³¹ systemic cortisol exposure³² and gut dysbiosis with altered short-chain fatty acids (SCFAs).³²,³⁶ Neuroinflammation, indicated by microglia activation and elevated cytokine levels centrally, may underpin the cognitive impairments in BD.^{37 38} Consequently, several conceptual models have been proposed by our group³⁹ and others^{40 41} to explain the involvement of inflammatory pathways in BD. Although there is no consensus on a biomarker signature sensitive to illness phase, some inflammatory mediators exhibit trait characteristics, while others vary with affective states.^{1922 23 42},⁴⁶

Aspirin belongs to the class of non-steroid anti-inflammatory drugs (NSAIDs) and is on The WHO Model List of Essential Medicines.⁴⁷ With the ability to inhibit platelet aggregation, low-dose aspirin (LDA) is the cornerstone in secondary prevention of atherosclerotic cardiovascular disease (ASCVD) worldwide.⁴⁸ Other indications include the acute treatment and primary prevention of ASCVD and preeclampsia.⁴⁹ In recent years, LDA has gained new ground in the prevention and treatment of various cancer types.⁵⁰,⁵² The evidence base for the effects of LDA in BD includes the following observations:

1. Mode of action: LDA effects the neuroimmune system via inhibition of the cyclooxygenase (COX) enzymes^{53 54} and through non-COX-dependent pathways.⁵⁵ Specifically, LDA inhibition of COX-1 is considered neuroprotective while acetylation of COX-2 generates local anti-inflammatory mediators⁵⁶ and immunoresolvents such as Aspirin-Triggered Lipoxins.^{57 58} The anti-inflammatory effects of LDA have been confirmed in clinical trials.⁵³⁵⁹,⁶³ LDA may also influence neurotransmission through COX-2 acetylation and the modification of arachidonate-specific signalling in neurons.⁶⁴

2. Effects of aspirin in animal models: Animal models of depression indicate that aspirin may enhance the antidepressant effect of lithium⁶⁵ and fluoxetine⁶⁶,⁶⁸ or have an independent antidepressant effect.⁶⁹ One rodent study found manic-like states could be reversed by aspirin administration.⁷⁰

3. Brain imaging studies: Patients with BD have shown widespread white matter abnormalities on brain imaging, which are thought to represent microstructural demyelination,⁷¹,⁷³ while LDA has been found to stimulate myelination.⁷⁴,⁷⁸

4. Pharmacoepidemiological studies: In two Danish nationwide register-based longitudinal studies, the continued use of LDA (50–175 mg/day) was associated with decreased rates of incident depression⁷⁹ and BD,⁸⁰ whereas high-dose aspirin (≥500 mg/day) and other NSAIDs were not. Concordantly, a Dutch register study found that in patients with BD on lithium, LDA was associated with a duration-independent reduction in the relative risk of clinical deterioration when compared with patients not using aspirin, whereas other NSAIDs and glucocorticoids were not.⁸¹

5. Pilot clinical studies: A short-duration randomised controlled trial (RCT) with add-on LDA found a substantial effect on treatment response compared with placebo for bipolar depression.⁸² Another small RCT found that coadministration of N-acetylcysteine and high-dose aspirin (1000 mg) in BD was associated with a reduction in depressive symptoms.⁸³ In a short-duration open-label trial with a mixed sample of bipolar and unipolar treatment-resistant depression, the addition of LDA to selective serotonin reuptake inhibitor (SSRI) treatment was found to improve treatment response and remission rates.⁸⁴

In summary, emerging data suggest that LDA may protect against the onset and progression of BD. However, there are limited data on the antimanic potential and long-term effects of LDA on the clinical course, and the hypothesised working mechanism requires further investigation. Thus, we are currently conducting a pragmatic large-scale RCT to clarify the therapeutic role of LDA across the different clinical stages of BD and its relationship with the inflammatory phenotype. Specifically, we wish to test the following hypotheses: Adding LDA versus placebo to standard treatment for BD will reduce: (1) mood instability (MI) and (2) other critical outcomes such as activity instability and severity of depression. Finally, we hypothesise that the reduction in MI is higher in patients with systemic inflammation at baseline indexed with interleukin (IL) 6.

Methods and analyses

Study design, setting and recruitment

The trial is a two-arm, triple-blind, parallel-group, superiority RCT including 250 patients with newly diagnosed BD. Patients are randomised to either 150 mg of acetylsalicylic acid daily or a placebo for 6 months in addition to their regular treatment (see figure 1). The sample consists of patients from The Copenhagen Affective Disorder Clinic, a specialised mood disorder clinic that provides treatment for patients with newly diagnosed/first-episode BD.⁸⁵ The clinic receives patients with BD from the entire Capital Region of Denmark covering a catchment area of 1.8 million people and all psychiatric centres in the region.⁸⁵ All patients referred to as newly diagnosed/first-episode patients, that is, onset of first manic or hypomanic episode or when the International Classification of Diseases 10th Revision diagnosis of BD is made for the first time, are routinely asked for participation. Enrolment of patients started on 20 January 2022, with the last follow-up visit in November 2024.

Figure 1. Study flow chart.

The trial is monitored by the Good Clinical Practice (GCP) unit of Copenhagen University. The monitoring is independent of the sponsor and the investigators and is carried out as on-site visits with approximately four visits a year.

Inclusion and exclusion criteria

Patients newly diagnosed with BD (type 1 or 2) aged 18–65 years are eligible (see additional criteria in table 1). Eligibility is evaluated by medical doctors, who also obtain informed consent. Initial diagnostic assessment is done by a senior psychiatrist, and the diagnosis is confirmed on inclusion by medical doctors or psychologists with psychiatric expertise using the semistructured Schedules for Clinical Assessment in Neuropsychiatry (SCAN) interview.⁸⁶

Table 1. Inclusion and exclusion criteria.

Inclusion criteria	Exclusion criteria
Primary diagnoses of bipolar disorder (type 1 or 2), confirmed with SCAN interview	Chronic kidney disease with GFR 0–10 mL/min
Age 18–65 years	Severe cardiac insufficiency
Informed consent	History of gastric ulcers, gastrointestinal bleeding or other pathological bleeding tendency; thrombocytopaenia
	Asthma or other allergic symptoms developed after intake of aspirin, paracetamol or other NSAIDs
	Concomitant daily use of aspirin, another NSAID or antithrombotics; SSRI treatment
	Female patients: Pregnancy (positive pregnancy test), breast feeding or if reluctant to use contraception during the trial period

GFR, glomerular filtration rate; NSAID, non-steroid anti-inflammatory drug; SCAN, Schedules for Clinical Assessment in NeuropsychiatrySSRI, selective serotonin reuptake inhibitor

Randomisation and blinding

The Hospital Pharmacy of the Capital Region of Denmark manages the randomisation process. The randomisation is carried out as block randomisation with changing predefined block sizes and 1:1 allocation to either placebo or LDA. The trial medication is encapsulated and prepacked from the pharmacy in sequentially numbered plastic bins with identical labelling. When a patient is enrolled, a sequential randomisation number and a corresponding treatment bin are irreversibly assigned. Participants, investigators and data analysts remain blind to treatment allocation throughout the entire trial duration. Allocation concealment and blinding are secured as follows:

1. Treatment allocation and randomisation is outsourced to a third, independent, party (ie, the pharmacy).

2. Trial medication is encapsulated and prepacked with identical labelling.

3. Thromboxane B2 (TXB2) levels, which will reveal whether the participant ingests LDA⁸⁷ (see below), will not be analysed until the end of the trial, and the results will not be available until after the data analyses.

Interventions

The participants are randomised to either LDA (one daily oral capsule of 150 mg acetylic salicylic acid) or placebo (calcium) during a trial period of 6–12 months (see figure 1). Participation is supplementary to the usual treatment at the mood disorder clinic and does not limit any other psychopharmacological or non-pharmacological treatment. Typically, the doses for secondary ASCVD prevention ranges from 75 to 100 mg daily⁸⁸ but can go up to 162 mg daily for patients with diabetes.⁸⁹ Thus, the dose of 150 mg aspirin daily is on the higher end of the commonly used doses for prevention in clinical practice. The rationale for this dose is a trade-off between a dose expected to exert a measurable, anti-inflammatory/immunomodulatory effect and the minimisation of side effects. The 6-month treatment period allows for the assessment of changes in MI (see later) and other outcomes. After the trial period, there is no further follow-up due to the short half-life of LDA.

Study medication adherence is monitored in three ways: (1) Participants are asked to return residual medication or empty bins at each visit; (2) in a random subset of 25% of the participants, plasma TXB2 is measured at 3 and 6-month follow-up to confirm adherence to LDA/placebo and (3) participants daily register compliance to medication in the Monsenso app.

The intervention is discontinued if the participant meets an exclusion criterion (see table 1), in the occurrence of unacceptable side effects, or withdraws informed consent. In the case of major surgery, the trial medication is paused due to the potential bleeding risk. If the treatment is discontinued, we register the reason for the discontinuation and seek to keep the participant in the trial, completing a full follow-up with assessment of all outcomes. In the event of exclusion or drop-out where a full follow-up is not possible, we collect data on the reason for stopping the trial as well as clinical ratings, questionnaires and biological measures, if possible.

Participant timeline

Trial participation includes a baseline visit and two follow-up visits after 3 and 6 months, respectively. Further, if the time frame allows, a fourth follow-up will be conducted after 12 months (see table 2 for an overview of enrolment, interventions, assessments and trial visits).

Table 2. Treatment procedures.

	Screening	Inclusion	3 months	6 months	12 months
Introduction, informed consent, eligibility	X
Randomisation and allocation to LDA/placebo		X
Smartphone-based mood ratings			Daily during the entire RCT
Clinical ratings (Ham-D17, YMRS, FAST)		X	X	X	X
Questionnaires		X	X	X	X
Cognition with SCIP		X		X
Blood-based biomarkers		X		X
Hair cortisol		X	X	X	X
Gut microbiome composition; SCFAs				X
Adherence to treatment by TXB2*		X	X	X
Adherence to treatment by pill count and interview			X	X	X
Monitoring of adherence to mood ratings			Weekly during the entire RCT
Adverse event assessment			X	X	X

^*A random subset of 25% of participants.

FAST, Functional Assessment Short Test; Ham-D17, Hamilton Depression Rating Scale (17-item version); LDA, low-dose aspirin; RCTrandomised controlled trialSCFAs, Short-Chain Fatty Acids; SCIP, Screen for Cognitive Impairment in Psychiatry; TXB2, Thromboxane B2YMRS, Young Mania Rating Scale

Primary outcome measure

During the last decade, there has been a gradual shift from a focus on affective episodes to interepisodic MI.^{90 91} Mood instability has internal validity as a real-life measure for patients and high external validity as it reflects patients’ illness severity and functioning, and extensive evidence shows that MI is of core pathogenetic significance in BD.⁹²,⁹⁶ Thus, a substantial proportion of patients with BD experience subsyndromal daily mood swings associated with increased perceived stress, decreased quality of life and functioning,^{95 96} and increased risk of relapse and hospitalisation.⁹²,⁹⁴ Mood instability is considered a new treatment target in BD and may be a more sensitive outcome measure in RCTs than more conventional outcomes focusing on relapse or recurrence of affective episodes.^{90 91 97 98} Building on this, the primary outcome measure is a MI score which reflects the daily variability in self-monitored mood, that is, the extent to which consecutively measured daily mood scores differ from one another during the 6-month trial duration.

Mood instability data are collected via smartphone using the Monsenso system, which allows for fine-grained real-time remote assessment of mood for ecological momentary assessments (EMAs)^{99 100} in research settings. Ecological momentary assessments allow the timing and compliance of data collection to be verified and eliminate the need for costly and error-prone data entry. Further, EMAs reduce retrospective recall bias, which is a particular problem for mood monitoring because patients need to recall both variation and intensity around a global mean.⁹⁹ The Monsenso system collects time series of data representing daily self-monitoring of mood, activity, sleep, irritability, cognitive problems, alcohol consumption, stress and medication self-administration. Patients are prompted every evening to fill in these data, and the process takes 2 min. We have previously shown that adherence to reporting is over 93% during a 6-month trial period¹⁰¹ and over 72% during a 9-month trial period.¹⁰² Moreover, smartphone-based mood ratings in the Monsenso system correlate highly with clinical ratings of mood according to subitem 1 on the HAMD-17 and the YMRS, respectively.⁹⁶

According to our established methodology, for each participant, we will estimate an MI score by applying the root mean square successive difference (rMSSD) method to the daily mood ratings.¹⁰³,¹⁰⁶ Specifically, participants daily rate their mood on a symmetric 9-point scale with a midpoint of 0 and a range from −3 to +3. Scores between −0.5 and +0.5 reflect normal variations around a neutral mood, whereas scores of +1, +2 or +3 correspond to mildly, moderately and severely increased mood, and scores of −1 to –2 or −3 correspond to mildly, moderately and severely decreased mood. The mood ratings are then used to calculate the MI score (see box 1 for an overview of all outcomes).

Box 1. Outcomes.

Primary outcome

• Daily mood instability (self-reported, smartphone based).

Secondary outcomes

• Daily activity instability (self-reported, smartphone-based).

• Depressive symptoms (Hamilton Depression Rating Scale-6 items).

Tertiary outcomes

• Manic symptoms (Young Mania Rating Scale).

• Functioning (Functional Assessment Short Test).

• Cognition (Screen for Cognitive Impairment in Psychiatry).

• Quality of life (WHO Quality of Life-BREF questionnaire).

• Perceived stress (Cohen’s Perceived Stress Scale).

• Physical activity (International Physical Activity Questionnaire).

• Sleep quality (Pittsburgh Sleep Quality Index).

• Hours of sleep (self-reported, smartphone based).

• Daily level of physical activity (smartphone accelerometer data).

• Social activity (numbers of outgoing and incoming calls and text messages/24 hours and time spent on the smartphone).

Secondary outcome measures

Daily self-reported activity instability collected via the Monsenso app, as previously reported by Stanislaus et al^{107 108}; severity of depression measured by the Hamilton Depression Scale (HAM-D)-6 items,¹⁰⁹ which is more sensitive in RCTs than the HAM-D17,¹¹⁰ at baseline, 3-month and 6-month follow-up.

Tertiary outcome measures

Clinical ratings of manic symptoms and functioning

Manic symptoms were assessed at baseline, 3-month and 6-month follow-up using the Young Mania Rating Scale (YMRS)¹¹¹; functioning was assessed at baseline, 3-month and 6-month follow-up using the Functional Assessment Short Test (FAST), a 24-item rating scale concerning autonomy, occupational functioning, cognitive functioning, financial issues, interpersonal relationships and leisure time.¹¹²

Cognition

Cognition was assessed at baseline and at 6- month follow-up according to the clinician-administered Screen for Cognitive Impairment in Psychiatry (SCIP) tool.

Questionnaire data

Self-assessed scores at baseline, 3-month and 6-month follow-up on the following questionnaires: Quality of life according to WHO Quality of Life-BREF¹¹³; perceived stress according to Cohen’s Perceived Stress Scale¹¹⁴; physical activity according to the International Physical Activity Questionnaire¹¹⁵ ; and sleep quality according to Pittsburgh Sleep Quality Index.¹¹⁶

Self-reported hours of sleep

Daily self-reported hours of sleep were collected via the Monsenso app.

Automatically smartphone-generated data

Level of physical activity measured by an accelerometer¹¹⁷; social activity expressed as numbers of outgoing and incoming calls; and text messages/24 hours and time spent on the smartphone.¹¹⁸

Biological measures

Fasting state blood samples at inclusion and at 6 month follow-up; hair cortisol at baseline, midway and at 6-month follow-up; and stool samples at 6-month follow-up (box 2).

Box 2. Biological measures.

Fasting state blood sampling

• Blood-based markers of inflammation: Eotaxin, IFN-γ, IL-10, IL-12p70, IL-17A, IL-1RA, IL-1β, IL-2, IL-4, IL-6, IL-8, MCP-1, MCP-4, MDC, TARC, and TNF-α.

Hair sampling

• Systemic cortisol exposure.

Stool sampling

• Microbiome characterisation; short-chain fatty acid levels.

Based on the results from the BIO-study,¹¹⁹ we will analyse the following blood-based biomarkers: Eotaxin, interleukin-1 receptor antagonist (IL-1RA), IL-8, IL-10, monocyte chemoattractant protein (MCP) 1, MCP-4; macrophage-derived chemokine; and thymus and activation-regulated chemokine (TARC). Further, we will analyse an ultrasensitive assay of biomarkers associated with the inflammatory response and immune system regulation, including: interferon-gamma (IFN-γ), IL-1β, IL-2, IL-4, IL-6, IL-12p70, IL-17A; and tumour necrosis factor-alpha (TNF-α). Hair cortisol will reflect systemic cortisol exposure from 4 months prior to the intervention and until discontinuation. The stool samples will be analysed to investigate the effects of LDA on microbiome composition and SCFAs.

Safety monitoring and adverse events

The most common side effects of LDA are abdominal pain, gastro-oesophageal reflux, pyrosis and bleeding tendency.¹²⁰ Long-duration LDA treatment is associated with an increased bleeding risk,¹²¹ and current guidelines preserve the use of LDA for primary prevention to individuals at high risk of ASCVD with a low bleeding risk.¹²²,¹²⁴ Predictive factors for gastrointestinal (GI) bleeding in relation to long-term LDA use are increased age, smoking, hypertension and chronic kidney disease.¹²⁵ Combined antithrombotic treatment is associated with an increased risk of upper GI bleeding¹²⁶ as well as the concomitant use of other NSAIDs.¹²⁷ Further, concomitant SSRI therapy presents an increased risk of bleeding compared with LDA monotherapy.^{127 128} Consequently, patients with former major bleeding episodes, haemophilia or other conditions with bleeding diathesis are excluded, as well as patients treated with NSAIDs, SSRIs or antithrombotics (table 1).

In summary, when used in low doses in a low-risk population, LDA is well tolerated and safe. Thus, in the present single-centre, low-risk trial, a data monitoring committee is not necessary.

Participants are routinely asked about side effects at each trial visit. If a participant seeks medical help at a public hospital, trial investigators are automatically notified via the electronic health record software. Adverse events are reviewed by study physicians and assessed for relatedness to trial medication. Severe adverse events are reported to the relevant authorities following GCP guidelines and local legislation.

Statistical considerations

Sample size and power

According to prior analyses,⁹⁵ MI varies on a scale from close to 0–10 with an average MI of 4.1 (SD=2.6). To detect a clinically relevant decrease of 0.7 in MI with treatment compared with placebo with a power of 80% and a significance level of 0.05, a total of 217 patients needs to be randomised. To further accommodate an attrition rate of 10%, we will include 250 patients in the study.

Analyses of the primary outcome measure

Mood instability will be computed for each participant by applying the rMSSD method, taking the square root of the sum of the squared differences between daily and previous day mood scores as done by our group⁹⁵ and others.^{92 129} Analyses will be conducted in the intention-to-treat population including data from all randomised patients regardless of treatment adherence and study completion. We will use a linear mixed model^{130 131} including time and the time-treatment interaction as fixed effects and with an unstructured covariance pattern to account for repeated measurements on each study participant. Exploratory subgroup analyses of drug responses will be carried out with the patients stratified according to biomarkers of systemic inflammation at inclusion indexed with IL-6 in accordance with suggestions for biomarker-guided anti-inflammatory therapies.¹³²

Analyses of secondary and tertiary outcome measures

Secondary and tertiary outcomes will be analysed with linear mixed-effect models similar to those used in the primary analysis. To avoid spurious findings, the p values will be adjusted for multiple testing using the method of Benjamini and Hochberg¹³³ which controls the false discovery rate. The p values from the secondary and tertiary analyses will be adjusted separately.

Analyses of biological data

We will conduct moderation analyses by further including fixed effects of inflammatory markers and their interaction with time and treatment in the linear mixed model from the primary analyses and, if relevant, investigate whether the treatment effect in the primary analysis was mediated by changes in inflammatory marker levels. Additionally, using data from HC from the ongoing BIO study¹³⁴, we will compare baseline inflammatory marker levels of BD patients and HC adjusted for sex, age, smoking status and body mass index.

Missing data

Reasons for non-response and drop-out will be tabulated and the characteristics of dropouts and completers from each randomisation group will be reported in online supplemental table 1. We expect that missing data will be few (<10%) and random, in which case, they will be handled implicitly by maximum likelihood inference in the linear mixed model. In case of larger than expected attrition rates or differential dropout between the randomisation groups, we will use sensitivity analyses based on multiple imputations to assess the robustness of the treatment effect under various scenarios for the missing-data mechanism.

Data collection and management

Mood instability and other self-reported data will be collected via the Monsenso system. Additionally, Monsenso collects sensor data on physical¹¹⁷ and social activity¹¹⁸, which reflect illness activity. Adherence to the Monsenso app is monitored weekly by trial staff, and non-compliant participants are contacted by text message or phone.

At the visits, the clinical ratings and questionnaire data are entered directly into an online Research Electronic Data Capture database.¹³⁵ The procedures for collection of biological specimens (blood, hair and faeces) follow the standard procedures of the ongoing BIO study¹³⁴ . Blood samples are obtained in a fasting state at a 4-hour interval in the morning and stored at –80°C. Blood sampling and laboratory processing are done by technicians blinded for randomisation status.

Patient and public involvement

The design of the trial did not include patient or public involvement.

Discussion

Summary

The A-bipolar RCT is the first pragmatic large-scale RCT investigating the effects of add-on LDA versus placebo to standard treatment using MI as the primary outcome measure in newly diagnosed patients with BD. The study has proven feasible with 250 patients enrolled. Moreover, LDA is well tolerated in the sample, with satisfactory adherence to the primary outcome measure. We expect that the results will illuminate the therapeutic potential of LDA in BD.

Limitations

First, given that the A-bipolar RCT is a drug trial, there might be a selection of participants that are favourably disposed towards pharmacological treatments. In theory, this might lead to a non-representative sample and impair generalisability of the findings. However, as participants are recruited from a regional mood disorder clinic covering a catchment area of 1.8 million people with systematic screening of all newly referred patients, and since the intervention is tested in a real-world context as an add-on to treatment as usual, the external validity is high.

Second, the trial’s requirements—especially the extensive initial assessment (3 hours)—may weed out the most unstable patients (ie, those with higher symptom burden), resulting in selection towards more stable patients. This might impair the detection of a true intervention effect and limit generalisability. However, as MI is characterised by high internal and external validity and appears to be a more sensitive primary outcome measure compared with conventional outcomes, we will be able to detect reductions in the subsyndromal mood swings that BD patients suffer from even in periods of remission.⁹⁸

Strengths

The A-bipolar RCT is a triple-blind, placebo-controlled randomised trial with a representative sample of 250 newly diagnosed patients with BD, hierarchically defined outcome measures and a minimum of a 6-month follow-up period. Participants are included regardless of illness phase (depression, (hypo)mania or euthymia). Taken together, the RCT design and the robust blinding ensures the internal validity and enables the detection of long-term effects of LDA on the clinical course of BD. Further, the exploratory biomarker analyses will permit a mechanistic insight into the hypothesised anti-inflammatory/immunomodulatory and mood-stabilising effects of LDA. Several aspects add to securing the external validity: First, the sample is representative of the target population. Second, the primary outcome measure, MI, sensitively captures clinically relevant mood swings. Third, since the patients are newly diagnosed, the findings will be generalisable to those in the early stages of the disorder when preventing relapse and preservation of quality of life, cognition and functioning is of utmost importance.

Ethics and dissemination

Written informed consent is obtained from participants at enrolment. The study was approved by the Danish Research Ethics Committee (H-21014515) and is conducted in accordance with the principles of the Declaration of Helsinki. Sensitive personal data are handled according to Danish legislation. Important protocol modifications will be communicated to relevant parties and published on ClinicalTrials.gov. Trial investigators will have access to the final dataset. Access to the full protocol, participant-level dataset and statistical code will be granted on request. Rules for authorship follow the ICMJE guidelines, and study results will be reported following the Consolidated Standards of Reporting Trials (CONSORT) 2010 guidelines and published in peer-reviewed journals.

The funding sources have no role in the study design, conduction of the study, or interpretation of the data.