Digital phenotyping in bipolar disorder: Using longitudinal Fitbit data and personalized machine learning to predict mood symptomatology.

BACKGROUND: Effective treatment of bipolar disorder (BD) requires prompt response to mood episodes. Preliminary studies suggest that predictions based on passive sensor data from personal digital devices can accurately detect mood episodes (e.g., between routine care appointments), but studies to date do not use methods designed for broad application. This study evaluated whether a novel, personalized machine learning approach, trained entirely on passive Fitbit data, with limited data filtering could accurately detect mood symptomatology in BD patients. METHODS: We analyzed data from 54 adults with BD, who wore Fitbits and completed bi-weekly self-report measures for 9 months. We applied machine learning (ML) models to Fitbit data aggregated over two-week observation windows to detect occurrences of depressive and (hypo)manic symptomatology, which were defined as two-week windows with scores above established clinical cutoffs for the Patient Health Questionnaire-8 (PHQ-8) and Altman Self-Rating Mania Scale (ASRM) respectively. RESULTS: As hypothesized, among several ML algorithms, Binary Mixed Model (BiMM) forest achieved the highest area under the receiver operating curve (ROC-AUC) in the validation process. In the testing set, the ROC-AUC was 86.0% for depression and 85.2% for (hypo)mania. Using optimized thresholds calculated with Youden's J statistic, predictive accuracy was 80.1% for depression (sensitivity of 71.2% and specificity of 85.6%) and 89.1% for (hypo)mania (sensitivity of 80.0% and specificity of 90.1%). CONCLUSION: We achieved sound performance in detecting mood symptomatology in BD patients using methods designed for broad application. Findings expand upon evidence that Fitbit data can produce accurate mood symptomatology predictions. Additionally, to the best of our knowledge, this represents the first application of BiMM forest for mood symptomatology prediction. Overall, results move the field a step toward personalized algorithms suitable for the full population of patients, rather than only those with high compliance, access to specialized devices, or willingness to share invasive data.

Perceived loneliness and social support in bipolar disorder: relation to suicidal ideation and attempts.

BACKGROUND: The suicide rate in bipolar disorder (BD) is among the highest across all psychiatric disorders. Identifying modifiable variables that relate to suicidal thoughts and behaviors (STBs) in BD may inform prevention strategies. Social connectedness is a modifiable variable found to relate to STBs in the general population, but differences exist across subgroups of the general population and findings specifically in BD have been equivocal. We aimed to clarify how perceived social connectedness relates to STBs in BD. METHOD: 146 adults (86 BD, 60 healthy controls) completed clinical interviews (Hamilton Depression Rating Scale; Structured Clinical Interview for DSM-5) and self-report measures of loneliness (UCLA Loneliness Scale) and social support (Interpersonal Support Evaluation List). Analyses explored differences in indicators of social connectedness (loneliness and social support) between BD participants and healthy controls, and explored relationships between STBs (lifetime suicide attempts and current suicidal ideation) and indicators of social connectedness in BD participants. RESULTS: BD participants reported significantly higher loneliness and lower social support than healthy controls. In BD participants, perceived social support was significantly related to both ever having attempted suicide and number of lifetime attempts. Interestingly, perceived loneliness, but not social support, was significantly associated with current suicidal ideation. CONCLUSIONS: Findings expand the evidence base supporting a relationship between perceived social connectedness and STBs in BD. They suggest that this modifiable variable could be a fruitful treatment target for preventing STBs in BD.

Modafinil's effects on cognition and sleep quality in affectively-stable patients with bipolar disorder: a pilot study.

Introduction: Despite advances in the treatment of bipolar disorder (BD), most patients do not achieve complete inter-episode recovery and functional disability is common. During periods of relative remission, many patients continue to experience neurocognitive dysfunction, reduced daytime activity levels, and sleep disturbances. This 8-week, randomized, placebo-controlled pilot study evaluated the feasibility, safety and preliminary efficacy of the wake-promoting drug, modafinil (Provigil®), on neurocognitive functioning, daytime sleepiness, and sleep quality in affectively-stable BD patients. Methods: Twelve individuals with affectively-stable BD were recruited and randomized to a flexible dose of modafinil (100 to 200 mg/day) or placebo, adjunctive to a therapeutic dose of a mood stabilizer. Weekly in-person visits tracked sleep quality and daytime sleepiness as well as side effects and mood symptoms. Neurocognitive functioning was assessed at baseline, week 4, and week 8. Results: No serious adverse events were reported. Newly emergent side effects in the modafinil group included heart palpitations, itching, fatigue, and decreased energy. Two patients discontinued modafinil owing to side effects and one of these patients withdrew from the study. One patient discontinued placebo and was withdrawn from the study. Preliminary evaluations of clinical efficacy showed a marginally significant interaction between treatment group and time in two cognitive domains (speed of processing and verbal learning), indicating greater improvement in the modafinil group versus placebo. Additionally, there was a marginally significant effect of treatment group on daytime sleepiness, suggesting lower daytime sleepiness in the modafinil group versus placebo. Counterintuitively, we found a significant treatment group by time interaction effect on sleep quality, suggesting greater improvement in sleep quality in the placebo group versus the modafinil group. Discussion: Results suggest that modafinil is a relatively safe medication for affectively-stable BD patients when given with adjunctive mood stabilizers. Results are suggestive of cognitive benefit and improved daytime sleepiness, but worse sleep quality in those patients prescribed modafinil. A fully powered clinical trial is warranted with specific attention to the characteristics of patients who are most likely to benefit from treatment with modafinil and other methodological lessons learned from this pilot. Clinical trial registration: ClinicalTrials.gov, identifier NCT01965925.

The engagement problem: A review of engagement with digital mental health interventions and recommendations for a path forward.

Purpose of the review: Digital mental health interventions (DMHIs) are an effective and accessible means of addressing the unprecedented levels of mental illness worldwide. Currently, however, patient engagement with DMHIs in real-world settings is often insufficient to see clinical benefit. In order to realize the potential of DMHIs, there is a need to better understand what drives patient engagement. Recent findings: We discuss takeaways from the existing literature related to patient engagement with DMHIs and highlight gaps to be addressed through further research. Findings suggest that engagement is influenced by patient-, intervention- and systems-level factors. At the patient-level, variables such as sex, education, personality traits, race, ethnicity, age and symptom severity appear to be associated with engagement. At the intervention-level, integrating human support, gamification, financial incentives and persuasive technology features may improve engagement. Finally, although systems-level factors have not been widely explored, the existing evidence suggests that achieving engagement will require addressing organizational and social barriers and drawing on the field of implementation science. Summary: Future research clarifying the patient-, intervention- and systems-level factors that drive engagement will be essential. Additionally, to facilitate improved understanding of DMHI engagement, we propose the following: (a) widespread adoption of a minimum necessary 5-element engagement reporting framework; (b) broader application of alternative clinical trial designs; and (c) directed efforts to build upon an initial parsimonious conceptual model of DMHI engagement.

Study protocol for a phase II, double-blind, randomised controlled trial of cannabidiol (CBD) compared with placebo for reduction of brain neuroinflammation in adults with chronic low back pain.

INTRODUCTION: Chronic pain is a debilitating medical problem that is difficult to treat. Neuroinflammatory pathways have emerged as a potential therapeutic target, as preclinical studies have demonstrated that glial cells and neuroglial interactions play a role in the establishment and maintenance of pain. Recently, we used positron emission tomography (PET) to demonstrate increased levels of 18 kDa translocator protein (TSPO) binding, a marker of glial activation, in patients with chronic low back pain (cLBP). Cannabidiol (CBD) is a glial inhibitor in animal models, but studies have not assessed whether CBD reduces neuroinflammation in humans. The principal aim of this trial is to evaluate whether CBD, compared with placebo, affects neuroinflammation, as measured by TSPO levels. METHODS AND ANALYSIS: This is a double-blind, randomised, placebo-controlled, phase II clinical trial. Eighty adults (aged 18-75) with cLBP for >6 months will be randomised to either an FDA-approved CBD medication (Epidiolex) or matching placebo for 4 weeks using a dose-escalation design. All participants will undergo integrated PET/MRI at baseline and after 4 weeks of treatment to evaluate neuroinflammation using [11C]PBR28, a second-generation radioligand for TSPO. Our primary hypothesis is that participants randomised to CBD will demonstrate larger reductions in thalamic [11C]PBR28 signal compared with those receiving placebo. We will also assess the effect of CBD on (1) [11C]PBR28 signal from limbic regions, which our prior work has linked to depressive symptoms and (2) striatal activation in response to a reward task. Additionally, we will evaluate self-report measures of cLBP intensity and bothersomeness, depression and quality of life at baseline and 4 weeks. ETHICS AND DISSEMINATION: This protocol is approved by the Massachusetts General Brigham Human Research Committee (protocol number: 2021P002617) and FDA (IND number: 143861) and registered with ClinicalTrials.gov. Results will be published in peer-reviewed journals and presented at conferences. TRIAL REGISTRATION NUMBER: NCT05066308; ClinicalTrials.gov.

Correlates of continued cannabis use during pregnancy.

BACKGROUND: Cannabis use is increasingly common among pregnant women despite concern that it may be linked to adverse maternal and infant outcomes. Determining whether variables associated with cannabis use predict whether women continue or quit using during pregnancy may inform strategies to reduce prenatal use. METHODS: Pregnant women who regularly used cannabis before pregnancy (n = 296) were recruited via Facebook. After finding out they were pregnant, 41 % reported quitting, 13 % quit then relapsed, 32 % reduced use, and 15 % continued use at the same rate. Differences among these four cannabis use status groups (quit, relapsed, reduced, continued) in sociodemographics, cannabis use, cigarette use, perceived risk/benefit, delay discounting, and communications about cannabis with their doctor were assessed. RESULTS: Compared to those who quit, continuing use during pregnancy was associated with being unemployed (Relative Risk (RR) = .32, 95 %CI [.13, .78]), using cigarettes pre-pregnancy (RR = 3.43, 95 %CI [1.32, 8.94]), being in an earlier trimester (RR = 4.38, 95 %CI [1.18, 16.23]), less perceived risk (RR = .79, 95 %CI [.74, .85]), and more days per week of use pre-pregnancy (RR = .10, 95 %CI [.01, .84]). Unintended pregnancy, shorter time to cannabis use after waking pre-pregnancy, using cannabis more times per day pre-pregnancy, and greater perceived benefits of use had significant bivariate associations with continued use during pregnancy, but did not retain significance in a multinomial model. CONCLUSIONS: Identification of these correlates provides potential targets for prevention of or intervention for prenatal cannabis use. However, much more research is needed to understand prenatal cannabis use and its effects in order to better educate women and healthcare providers, and to design optimal public health strategies.

Sifting through the weeds: Relationships between cannabis use frequency measures and delay discounting.

BACKGROUND: Delay Discounting (DD) relates to more frequent cannabis use, but results are variable, potentially because of variations in whether integrated or single-item measures are used, and whether the timeframe of measures is narrow or broad. Explicating the relationship between DD and cannabis use may result from comparing use indices that vary on these characteristics. METHODS: This online study of current cannabis users (n = 1,800) assessed DD and three cannabis use frequency items: number of days of use in the past month, times used per day, and weekly-monthly use. A fourth index derived with Latent Class Analysis (LCA) integrated days per month and times per day to try to better characterize frequency patterns. Effect sizes reflecting relations between cannabis use frequency indices and DD were compared. RESULTS: Three frequency classes emerged from the LCA (Low-Moderate-High). DD was significantly associated with times per day (r = 0.11, d = 0.21), days of use (r = 0.09, d = 0.18), and the LCA index (r = 0.06, d = 0.13), but not weekly-monthly use (r = 0.04, d = 0.09). Times per day was more strongly related to DD than LCA classes (p < 0.01) and weekly-monthly use (p < 0.05), but not days of use (p = 0.66). Days of use exhibited a stronger relationship with DD than weekly-monthly use (p < 0.001), but not LCA classes (p = 0.06). CONCLUSIONS: Cannabis use frequency measures with narrower timeframes may demonstrate stronger positive relationships to DD. The LCA index did not improve the relationship between frequency and DD, potentially because of shared variance between use days and times per day. Specific characteristics of cannabis use frequency may be particularly indicative of excessive DD.