Validation of a Composite Medical Device Using a Blood Biomarker-based Algorithm and MDQ for the Diagnosis of Bipolar Disorder

Introduction Bipolar disorder (BD) is a chronic and disabling condition affecting between 1% to 2.5% of the population, corresponding to approximately 650,000 to 1,650,000 individuals in France. Onset typically occurs between the ages of 15 and 25, with symptoms persisting throughout life. Bipolar disorder ranks as the sixth leading cause of disability globally, according to the World Health Organization (WHO). Additionally, individuals with bipolar disorder face a reduced life expectancy of around 10 years compared with the general population. This increased mortality is partly attributed to suicide, as approximately 20% of untreated BD patients die by suicide. In addition, BD is associated with a high prevalence of comorbidities such as alcohol or substance use disorders, diabetes, dyslipidaemia, which significantly heighten the risk of developing other diseases, notably cardiovascular pathologies. Risky behaviors, including impulsive spending, substance use, and unsafe sexual practices, further contribute to the functional impairments experienced by those with bipolar disorder. Clinically, bipolar disorder is characterized by alternating episodes of depression and mania or hypomania. While mania and hypomania are hallmark features of bipolar disorder, depressive episodes are more frequent and often serve as the initial clinical manifestation. This overlap with Major Depressive Disorder presents diagnostic challenges, particularly since patients often seek treatment during depressive phases, usually from general practitioners. Current diagnostic approaches for bipolar disorder rely heavily on clinical assessment, necessitating detailed knowledge of mood disorders and extensive patient interviews-both of which are often impractical in primary care settings. Consequently, many bipolar disorder patients are misdiagnosed with major depressive disorder and receive inappropriate treatment with antidepressants, which can exacerbate the condition. Studies indicate that the prevalence of undiagnosed bipolar disorder in patients treated with antidepressants ranges around 10% in primary care to 16%-47% in psychiatric settings. In this context, the average diagnostic delay of bipolar disorder is 8 years in France. Early diagnosis and treatment could significantly improve the outcomes for bipolar disorder patients. Self-administered tools such as the Mood Disorder Questionnaire (MDQ) offer potential solutions to improve bipolar disorder screening. The MDQ evaluates symptoms of mania or hypomania through a structured questionnaire divided into three sections: (1) assessment of 13 symptoms related to mania/hypomania, (2) evaluation of symptom temporality, and (3) determination of symptom impact on daily life. Despite its simplicity and utility, the MDQ has notable limitations, including low sensitivity (43%) and high specificity (95%), which restricts its standalone diagnostic reliability in primary care settings. In this context, the development of an in vitro diagnostic test based on blood biomarkers, enabling differential diagnosis between bipolar disorder and major depressive disorder in patients with major depressive episode, represents a critical advancement. Recent research underscored the potential utility of biomarkers to complement clinical evaluations and enhance diagnostic accuracy. A predictive model was developed incorporating routinely measured blood-based biomarkers, such as eosinophil counts, prolactin levels, total cholesterol (TC), and low-density lipoprotein (LDL) cholesterol, to differentiate bipolar disorder from major depressive disorder. Their nomogram demonstrated robust discriminatory power with an area under the curve (AUC) of 0.858, highlighting the promise of biomarker-based diagnostics. Another notable development in this area is the Edit-B® test, which utilizes RNA editing markers and was introduced in France in 2024. However, this test has faced regulatory challenges, emphasizing the need of alternative diagnostic solutions. Building on these advancements, the principal investigator and collaborators have developed a novel diagnostic algorithm combining biomarker analysis with the MDQ. Preliminary data suggest that this composite tool can enhance diagnostic precision, particularly in primary care settings where traditional clinical assessments are constrained. The proposed algorithm integrates the strengths of the biomarker-based approach, notably its high sensitivity, with the MDQ's specificity, aiming to achieve a sensitivity exceeding 85% and a specificity above 80%. This research seeks to validate the efficacy of this composite diagnostic tool, hypothesizing that its implementation will significantly improve bipolar disorder screening and diagnosis in primary care. By leveraging biomarker insights and patient-reported data, the study aims to reduce diagnostic delays, optimize treatment strategies, and ultimately improve outcomes for individuals with bipolar disorder. The results of this study have the potential to bridge the gap between research and clinical practice, offering a scalable and accessible diagnostic tool for psychiatric disorders. This is a multicenter, prospective, with no comparator group interventional clinical trial aiming to evaluate the diagnostic performance of a blood-based medical device for the detection of bipolar disorder in adult patients presenting with a current major depressive episode in primary care. The study will be conducted in the psychiatry departments of the different hospitals in the project. Participants will be recruited in primary care settings during a consultation for a depressive episode. All eligible participants will provide written informed consent. Objectives The primary aim of this study is to assess the diagnostic performance of a composite medical device that combines a blood biomarker-based algorithm and the MDQ for diagnosing bipolar disorder in patients presenting with a major depressive episode in primary care settings. Specifically, the focus is on determining the sensitivity and specificity of this composite medical device. Several secondary objectives will be explored in this study. First, the area under the curve, negative predictive value (NPV), and positive predictive value (PPV) of the composite medical device will be determined for diagnosing bipolar disorder in patients with major depressive episode in primary care. Additionally, the characteristics of false negatives (FN) and false positives (FP), identified by the composite medical device, will be described. These characteristics will include demographic factors (age, sex, BMI), depression severity, biological markers of inflammation, clinical and biological variables collected during the study, comorbid addictive conditions, and psychotropic treatments. Another key secondary objective is to assess the composite medical device's ability to predict the onset of hypomanic symptoms at 8 weeks and the improvement of depressive symptoms at 8 weeks. All these objectives will be revisited by taking the re-evaluated diagnosis at 8 weeks as a reference, which will also consider the occurrence of manic or hypomanic symptoms. The study will also compare the diagnostic performance of the composite medical device with the diagnosis proposed by the general practitioner general practitioner. Furthermore, the variability in biomarker measurements across operators, methods, and sites will be measured, as well as the acceptability of the composite medical device among both patients and general practitioners. The medico-economic impact of the composite medical device, if used to diagnose bipolar disorder in primary care, will also be evaluated, considering its sensitivity, specificity, and overall acceptability. Lastly, the study will seek to identify which biomarkers, either previously described in the literature or routinely measured in blood tests, could further enhance the performance of the composite medical device. Interventions and study timeline The first visit will be a pre-screening visit (V1). The pre-screening will be carried out by a general practitioner in their office during routine care. At the end of this pre-screening, and with the patient's consent, the general practitioner at each recruitment center will send the patient's contact information to the designated principal investigator's team. The principal investigator's team will then contact the potential participant by phone to schedule the screening visit. The second visit will be a screening visit (V2). It will be conducted through a telephone interview. The investigator team will provide the patient with all relevant information about the study, including its objectives, benefits, risks, and procedures. The patient will be informed of the relevant legal provisions and their rights. The third visit (V3) will be the inclusion visit. It will take place at one of the participating study centers, and at least 12 hours after the V2 visit to provide the patient with sufficient time to decide whether or not to participate in the study, without unduly delaying the specialized consultation and subsequent therapeutic guidance. A maximum delay of 7 days between patient selection and inclusion will also be respected. At the start of this visit, patients will be asked to sign the informed consent form for participation in the study. The patient will be evaluated by one of the investigators or a member of their team using standardized tools, self-assessment questionnaires (including MDQ, HCL-32, QIDS), and hetero-questionnaires (including MINI, MADRS). An audio recording will be made during the MINI administration. An evaluation of the test's acceptability will be carried out using a Likert scale. The fourth visit (V4) will be the visit during which a blood sample will be taken, between 7:00 AM and 11:00 AM, with the patient required to fast. This visit will take place either on the same day or within 3 days at most after the V3 visit. The collection will include two 2.5 mL Paxgene® tube for whole-blood RNA (ribonucleic acid), two 6 mL dry tube for serum, two 4 mL heparinized tube one for native DNA and one for plasma preparation and a 6 mL EDTA tube for complete blood count (retained for 15 years). These samples will be anonymised and sent to the Biological Resource Centre at Montpellier University Hospital. During V4 or immediately after V4, the general practitioner will be interviewed by the investigation team, either in person, via videoconference, or by email (secure medical messaging) regarding their diagnostic opinion (likelihood of bipolar disorder according to them on the Likert scale) and their view on the acceptability of the test. At the end of V4, the patient will be referred to their general practitioner or to any appropriate healthcare facility based on the diagnosis and the severity of the symptoms, with the patient's consent. A written or verbal care recommendation will be provided to the patient and their general practitioner, considering the evaluations carried out during V3. The 5th visit (V5) will be a follow-up visit. It will take place between 7 and 9 weeks after V4 and will be conducted via telephone interview or teleconsultation. The visits will be conducted by the investigation team from the CHU of Montpellier or the centre where the inclusion took place, depending on team availability. The patient will be assessed by one of the investigators using standardized tools, self-questionnaires (inlcusingincluding the MDQ) completed online or in writing, and hetero-questionnaires. The investigator will document the treatment administered to the patient and their care path since V3 and determine how the depressive symptoms have evolved and whether hypomanic symptoms have appeared since V3. Following this interview, the investigator will make a new diagnosis, confirming or modifying the one made at V3. Biological Sample Collection Biological samples will be collected during the inclusion visit at each of the participating study centers and then temporarily stored in a freezer at -20°C. Every 3 months, the biological samples will be shipped in a package containing dry ice to the CRB (Centre de Ressources Biologiques) in Montpellier, where it will be stored at -80°C. For biomarker measurements in serum, tubes containing 500 µl of serum (1 tube per patient) will be transported to the Institute of Molecular and Cellular Pharmacology, under the responsibility of Professor Nicolas Glaichenhaus before being used for biomarker measurement. For the biomarker measurements necessary for the use of the composite medical device (DM), serum samples will be analysed using the V-PLEX® (Mesoscale Discovery) and Olink Flex® (Olink) platforms, following the protocols described by the manufacturers. Additional analyses on other platforms will also be carried out to address the secondary objectives, in the laboratory of Professor Lehman in Montpellier, Professor Glaichenhaus in Valbonne, Dr. Lutz in Strasbourg, or any other public or private research facility capable of meeting the study's objectives. The collection of biological samples will be created and stored in compliance with regulations. It will be declared to the relevant regulatory authorities. The samples will be stored at -80°C at the CRB of the CHU of Montpellier for 12 years. A participant in the study may refuse for their biological samples to be stored in this collection. The biological samples may be used to explore new blood biomarkers for bipolar disorder in patients with a mood disorder in primary care settings, which could lead to the creation of new medical devices or the exploration of new pathophysiological hypotheses. Data Management and Statistical Methods The study data will be entered through an electronic case report form (e-CRF) on the Ennov Clinical (Clinsight) software, which allows for real-time data quality control. All this patient data comes from source documents which are the original documents from which the data will be added to the e-CRF. The investigator undertakes to allow direct access to the study source data during control, audit or inspection visits. In the data analysis, a brief description of the studied population will be presented. For each diagnostic test, the investigators will estimate the sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), as well as the area under the ROC (Receiver Operating Characteristic) curve and their associated 95% confidence intervals. To address the primary objective, the parameters of interest (sensitivity and specificity) for the primary outcome measure (sequential criterion) will be estimated both pointwise and by the 95% confidence interval. For secondary objectives, paired data comparison tests will be used for the various outcome measures. The grey zones will be estimated according to the method of Cannesson et al. If the performance of the biomarker-based algorithm and the MDQ are known, those of a composite medical device combining the two are difficult to predict, depending on whether the two tests are used in parallel or sequentially. To identify the most effective, the investigators will combine the biomarker-based algorithm and MDQ in four different ways : * Parallel analysis using the biomarker-based algorithm and MDQ : Patients with the composite medical device (MD) will be those who are positive on both tests. * Parallel analysis using the biomarker-based algorithm and MDQ : Positive patients with the composite MD will be those who are positive on at least one of the two tests. * Sequential analysis using the biomarker-based algorithm and MDQ : Positive patients with the composite MD will be those, among those positive on the biomarker-based algorithm, who are positive on the MDQ. * Sequential analysis using the biomarker-based algorithm and MDQ : Positive patients with the composite MD will be those, among those positive on the MDQ, who are positive on the biomarker-based algorithm. The chosen composite MD will be the one that provides the best sensitivity and specificity, or the highest Youden index in the absence of an ordered relationship. The search for factors associated with misclassification errors will be done using multivariate analysis with logistic regression. Variables with a p-value \<0.20 in univariate analysis will be included in the multivariate model. A forward selection procedure will then be used to obtain a final model. A new composite score for the medical device will be determined through logistic regression. An individual score will be established for each patient using the coefficients from the multivariate logistic regression. Sensitivity, specificity, and predictive values will be calculated. The score's ability to predict patient status will be estimated by the AUC of the ROC curve and its associated 95% confidence interval. Various thresholds will be identified via the ROC curve based on the Youden index. The variability in biomarker measurements according to the operator, laboratory, or platform usage will be assessed through the Intraclass Correlation Coefficient (ICC) calculations and the Bland-Altman graphical method. All tests will be two-tailed with an alpha risk set at 5%. A false discovery rate (FDR) penalty will be considered for the analysis of secondary criteria. For quality control purposes, diagnostic interviews will be re-listened to by an expert team, and the concordance between the diagnosis established by the investigator and the experts will be analysed by calculating Cohen's kappa index. It will have two interim analyses, one after the inclusion of 200 patients and the other after the inclusion of 400 patients. During these two interim analyses, the prevalence of TB in the cohort of already included patients will be measured, as well as the sensitivity, specificity, PPV, NPV, and AUC of the composite MD using the diagnosis established at V3 as the reference. The variability of biomarker measurements will be performed at each interim analysis. Biomarkers will be measured in parallel and independently in two different laboratories, specifically in the laboratories of Nicolas Glaichenhaus (CNRS) and Sylvain Lehmann (CHU of Montpellier).