Antidepressant drug-specific prediction of depression treatment outcomes from genetic and clinical variables

Association between polygenic risk scores combined with clinical characteristics and antidepressant efficacy

BACKGROUND

While millions of people suffer from major depressive disorder (MDD), research has shown that individual differences in antidepressant efficacy exist, potentially attributable to various factors. Polygenic risk scores (PRSs) carry clinical potential, but associations with treatment response are seldom reported. Here, we examined whether PRSs for MDD and schizophrenia (SCZ) are associated with antidepressant effectiveness and the influence of other factors.

METHODS

A total of 999 patients were included, and the PRSs for the MDD and SCZ were calculated. The main outcome was a change in the 17-item Hamilton Depression Rating Scale (HAMD17) scores from before to after 2-week treatment. The Mann-Whitney test, Spearman correlation analysis, multiple stepwise linear regression analysis, and interaction analysis were used for statistical analysis.

RESULTS

In the 912 subjects passing quality control, a difference in the HAM-D17 score reduction rate between the MDD phenotype PRS (MDD-PRS) high-risk and the low-risk groups was discovered (P = 0.009), and a correlation was found between the MDD-PRS and the HAM-D17 score reduction rate (r = -0.075, P = 0.024). Moreover, antidepressant efficacy was related to MDD-PRS (β = -4.086, P = 0.039), the Snaith-Hamilton Pleasure Scale-total score (β = -0.009, P = 0.005), and non-first episode (β = -0.039, P < 0.001). However, the result of the interaction analysis was nonsignificant.

LIMITATIONS

The main limitation was that only 1309 targeted genes were selected based on pathways known to be involved in MDD and/or antidepressant effects.

CONCLUSION

These findings suggest a difference in antidepressant efficacy between patients in different MDD-PRS groups. Moreover, the MDD-PRS combined with clinical characteristics partially explained inter-individual differences in antidepressant efficacy.

Pharmacogenomic augmented machine learning in electronic health record alerts: A health system-wide usability survey of clinicians

Pharmacogenomic (PGx) biomarkers integrated using machine learning can be embedded within the electronic health record (EHR) to provide clinicians with individualized predictions of drug treatment outcomes. Currently, however, drug alerts in the EHR are largely generic (not patient-specific) and contribute to increased clinician stress and burnout. Improving the usability of PGx alerts is an urgent need. Therefore, this work aimed to identify principles for optimal PGx alert design through a health-system-wide, mixed-methods study. Clinicians representing multiple practices and care settings (N = 1062) in urban, rural, and underserved regions were invited to complete an electronic survey comparing the usability of three drug alerts for citalopram, as a case study. Alert 1 contained a generic warning of pharmacogenomic effects on citalopram metabolism. Alerts 2 and 3 provided patient-specific predictions of citalopram efficacy with varying depth of information. Primary outcomes included the System's Usability Scale score (0-100 points) of each alert, the perceived impact of each alert on stress and decision-making, and clinicians' suggestions for alert improvement. Secondary outcomes included the assessment of alert preference by clinician age, practice type, and geographic setting. Qualitative information was captured to provide context to quantitative information. The final cohort comprised 305 geographically and clinically diverse clinicians. A simplified, individualized alert (Alert 2) was perceived as beneficial for decision-making and stress compared with a more detailed version (Alert 3) and the generic alert (Alert 1) regardless of age, practice type, or geographic setting. Findings emphasize the need for clinician-guided design of PGx alerts in the era of digital medicine.

Predicting the outcome of psychotherapy for chronic depression by person-specific symptom networks

Psychotherapies are efficacious in the treatment of depression, albeit only with a moderate effect size. It is hoped that personalization of treatment can lead to better outcomes. The network theory of psychopathology offers a novel approach suggesting that symptom interactions as displayed in person-specific symptom networks could guide treatment planning for an individual patient. In a sample of 254 patients with chronic depression treated with either disorder-specific or non-specific psychotherapy for 48 weeks, we investigated if person-specific symptom networks predicted observer-rated depression severity at the end of treatment and one and two years after treatment termination. Person-specific symptom networks were constructed based on a time-varying multilevel vector autoregressive model of patient-rated symptom data. We used statistical parameters that describe the structure of these person-specific networks to predict therapy outcome. First, we used symptom centrality measures as predictors. Second, we used a machine learning approach to select parameters that describe the strength of pairwise symptom associations. We found that information on person-specific symptom networks strongly improved the accuracy of the prediction of observer-rated depression severity at treatment termination compared to common covariates recorded at baseline. This was also shown for predicting observer-rated depression severity at one- and two-year follow-up. Pairwise symptom associations were better predictors than symptom centrality parameters for depression severity at the end of therapy and one year later. Replication and external validation of our findings, methodological developments, and work on possible ways of implementation are needed before person-specific networks can be reliably used in clinical practice. Nevertheless, our results indicate that the structure of person-specific symptom networks can provide valuable information for the personalization of treatment for chronic depression.

Explainable multimodal prediction of treatment-resistance in patients with depression leveraging brain morphometry and natural language processing

Although 20 % of patients with depression receiving treatment do not achieve remission, predicting treatment-resistant depression (TRD) remains challenging. In this study, we aimed to develop an explainable multimodal prediction model for TRD using structured electronic medical record data, brain morphometry, and natural language processing. In total, 247 patients with a new depressive episode were included. TRD-predictive models were developed based on the combination of following parameters: selected tabular dataset features, independent components-map weightings from brain T1-weighted magnetic resonance imaging (MRI), and topic probabilities from clinical notes. All models applied the extreme gradient boosting (XGBoost) algorithm via five-fold cross-validation. The model using all data sources showed the highest area under the receiver operating characteristic of 0.794, followed by models that used combined brain MRI and structured data, brain MRI and clinical notes, clinical notes and structured data, brain MRI only, structured data only, and clinical notes only (0.770, 0.762, 0.728, 0.703, 0.684, and 0.569, respectively). Classifications of TRD were driven by several predictors, such as previous exposure to antidepressants and antihypertensive medications, sensorimotor network, default mode network, and somatic symptoms. Our findings suggest that a combination of clinical data with neuroimaging and natural language processing variables improves the prediction of TRD.

Development of a multivariate prediction model for antidepressant resistant depression using reward-related predictors

Introduction

Individuals with depression who do not respond to two or more courses of serotonergic antidepressants tend to have greater deficits in reward processing and greater internalizing symptoms, yet there is no validated self-report method to determine the likelihood of treatment resistance based on these symptom dimensions.

Methods

This online case-control study leverages machine learning techniques to identify differences in self-reported anhedonia and internalizing symptom profiles of antidepressant non-responders compared to responders and healthy controls, as an initial proof-of-concept for relating these indicators to medication responsiveness. Random forest classifiers were used to identify a subset from a set of 24 reward predictors that distinguished among serotonergic medication resistant, non-resistant, and non-depressed individuals recruited online (N = 393). Feature selection was implemented to refine model prediction and improve interpretability.

Results

Accuracies for full predictor models ranged from .54 to .71, while feature selected models retained 3-5 predictors and generated accuracies of .42 to .70. Several models performed significantly above chance. Sensitivity for non-responders was greatest after feature selection when compared to only responders, reaching .82 with 3 predictors. The predictors retained from feature selection were then explored using factor analysis at the item level and cluster analysis of the full data to determine empirically driven data structures.

Discussion

Non-responders displayed 3 distinct symptom profiles along internalizing dimensions of anxiety, anhedonia, motivation, and cognitive function. Results should be replicated in a prospective cohort sample for predictive validity; however, this study demonstrates validity for using a limited anhedonia and internalizing self-report instrument for distinguishing between antidepressant resistant and responsive depression profiles.

The Burnout PRedictiOn Using Wearable aNd ArtIficial IntelligEnce (BROWNIE) study: a decentralized digital health protocol to predict burnout in registered nurses

BACKGROUND

When job demand exceeds job resources, burnout occurs. Burnout in healthcare workers extends beyond negatively affecting their functioning and physical and mental health; it also has been associated with poor medical outcomes for patients. Data-driven technology holds promise for the prediction of occupational burnout before it occurs. Early warning signs of burnout would facilitate preemptive institutional responses for preventing individual, organizational, and public health consequences of occupational burnout. This protocol describes the design and methodology for the decentralized Burnout PRedictiOn Using Wearable aNd ArtIficial IntelligEnce (BROWNIE) Study. This study aims to develop predictive models of occupational burnout and estimate burnout-associated costs using consumer-grade wearable smartwatches and systems-level data.

METHODS

A total of 360 registered nurses (RNs) will be recruited in 3 cohorts. These cohorts will serve as training, testing, and validation datasets for developing predictive models. Subjects will consent to one year of participation, including the daily use of a commodity smartwatch that collects heart rate, step count, and sleep data. Subjects will also complete online baseline and quarterly surveys assessing psychological, workplace, and sociodemographic factors. Routine administrative systems-level data on nursing care outcomes will be abstracted weekly.

DISCUSSION

The BROWNIE study was designed to be decentralized and asynchronous to minimize any additional burden on RNs and to ensure that night shift RNs would have equal accessibility to study resources and procedures. The protocol employs novel engagement strategies with participants to maintain compliance and reduce attrition to address the historical challenges of research using wearable devices.

TRIAL REGISTRATION

NCT05481138.

Prediction of depression treatment outcome from multimodal data: a CAN-BIND-1 report

BACKGROUND

Prediction of treatment outcomes is a key step in improving the treatment of major depressive disorder (MDD). The Canadian Biomarker Integration Network in Depression (CAN-BIND) aims to predict antidepressant treatment outcomes through analyses of clinical assessment, neuroimaging, and blood biomarkers.

METHODS

In the CAN-BIND-1 dataset of 192 adults with MDD and outcomes of treatment with escitalopram, we applied machine learning models in a nested cross-validation framework. Across 210 analyses, we examined combinations of predictive variables from three modalities, measured at baseline and after 2 weeks of treatment, and five machine learning methods with and without feature selection. To optimize the predictors-to-observations ratio, we followed a tiered approach with 134 and 1152 variables in tier 1 and tier 2 respectively.

RESULTS

A combination of baseline tier 1 clinical, neuroimaging, and molecular variables predicted response with a mean balanced accuracy of 0.57 (best model mean 0.62) compared to 0.54 (best model mean 0.61) in single modality models. Adding week 2 predictors improved the prediction of response to a mean balanced accuracy of 0.59 (best model mean 0.66). Adding tier 2 features did not improve prediction.

CONCLUSIONS

A combination of clinical, neuroimaging, and molecular data improves the prediction of treatment outcomes over single modality measurement. The addition of measurements from the early stages of treatment adds precision. Present results are limited by lack of external validation. To achieve clinically meaningful prediction, the multimodal measurement should be scaled up to larger samples and the robustness of prediction tested in an external validation dataset.

Multimodal predictions of treatment outcome in major depression: A comparison of data-driven predictors with importance ratings by clinicians

BACKGROUND

Reliable prediction models of treatment outcome in Major Depressive Disorder (MDD) are currently lacking in clinical practice. Data-driven outcome definitions, combining data from multiple modalities and incorporating clinician expertise might improve predictions.

METHODS

We used unsupervised machine learning to identify treatment outcome classes in 1060 MDD inpatients. Subsequently, classification models were created on clinical and biological baseline information to predict treatment outcome classes and compared to the performance of two widely used classical outcome definitions. We also related the findings to results from an online survey that assessed which information clinicians use for outcome prognosis.

RESULTS

Three and four outcome classes were identified by unsupervised learning. However, data-driven outcome classes did not result in more accurate prediction models. The best prediction model was targeting treatment response in its standard definition and reached accuracies of 63.9 % in the test sample, and 59.5 % and 56.9 % in the validation samples. Top predictors included sociodemographic and clinical characteristics, while biological parameters did not improve prediction accuracies. Treatment history, personality factors, prior course of the disorder, and patient attitude towards treatment were ranked as most important indicators by clinicians.

LIMITATIONS

Missing data limited the power to identify biological predictors of treatment outcome from certain modalities.

CONCLUSIONS

So far, the inclusion of available biological measures in addition to psychometric and clinical information did not improve predictive value of the models, which was overall low. Optimized biomarkers, stratified predictions and the inclusion of clinical expertise may improve future prediction models.

A cultural-ecosocial systems view for psychiatry

While contemporary psychiatry seeks the mechanisms of mental disorders in neurobiology, mental health problems clearly depend on developmental processes of learning and adaptation through ongoing interactions with the social environment. Symptoms or disorders emerge in specific social contexts and involve predicaments that cannot be fully characterized in terms of brain function but require a larger social-ecological view. Causal processes that result in mental health problems can begin anywhere within the extended system of body-person-environment. In particular, individuals' narrative self-construal, culturally mediated interpretations of symptoms and coping strategies as well as the responses of others in the social world contribute to the mechanisms of mental disorders, illness experience, and recovery. In this paper, we outline the conceptual basis and practical implications of a hierarchical ecosocial systems view for an integrative approach to psychiatric theory and practice. The cultural-ecosocial systems view we propose understands mind, brain and person as situated in the social world and as constituted by cultural and self-reflexive processes. This view can be incorporated into a pragmatic approach to clinical assessment and case formulation that characterizes mechanisms of pathology and identifies targets for intervention.

Deep phenotyping towards precision psychiatry of first-episode depression - the Brain Drugs-Depression cohort

BACKGROUND

Major Depressive Disorder (MDD) is a heterogenous brain disorder, with potentially multiple psychosocial and biological disease mechanisms. This is also a plausible explanation for why patients do not respond equally well to treatment with first- or second-line antidepressants, i.e., one-third to one-half of patients do not remit in response to first- or second-line treatment. To map MDD heterogeneity and markers of treatment response to enable a precision medicine approach, we will acquire several possible predictive markers across several domains, e.g., psychosocial, biochemical, and neuroimaging.

METHODS

All patients are examined before receiving a standardised treatment package for adults aged 18-65 with first-episode depression in six public outpatient clinics in the Capital Region of Denmark. From this population, we will recruit a cohort of 800 patients for whom we will acquire clinical, cognitive, psychometric, and biological data. A subgroup (subcohort I, n = 600) will additionally provide neuroimaging data, i.e., Magnetic Resonance Imaging, and Electroencephalogram, and a subgroup of patients from subcohort I unmedicated at inclusion (subcohort II, n = 60) will also undergo a brain Positron Emission Tomography with the [¹¹C]-UCB-J tracer binding to the presynaptic glycoprotein-SV2A. Subcohort allocation is based on eligibility and willingness to participate. The treatment package typically lasts six months. Depression severity is assessed with the Quick Inventory of Depressive Symptomatology (QIDS) at baseline, and 6, 12 and 18 months after treatment initiation. The primary outcome is remission (QIDS ≤ 5) and clinical improvement (≥ 50% reduction in QIDS) after 6 months. Secondary endpoints include remission at 12 and 18 months and %-change in QIDS, 10-item Symptom Checklist, 5-item WHO Well-Being Index, and modified Disability Scale from baseline through follow-up. We also assess psychotherapy and medication side-effects. We will use machine learning to determine a combination of characteristics that best predict treatment outcomes and statistical models to investigate the association between individual measures and clinical outcomes. We will assess associations between patient characteristics, treatment choices, and clinical outcomes using path analysis, enabling us to estimate the effect of treatment choices and timing on the clinical outcome.

DISCUSSION

The BrainDrugs-Depression study is a real-world deep-phenotyping clinical cohort study of first-episode MDD patients.

TRIAL REGISTRATION

Registered at clinicaltrials.gov November 15th, 2022 (NCT05616559).

The Precision in Psychiatry (PIP) study: Testing an internet-based methodology for accelerating research in treatment prediction and personalisation

BACKGROUND

Evidence-based treatments for depression exist but not all patients benefit from them. Efforts to develop predictive models that can assist clinicians in allocating treatments are ongoing, but there are major issues with acquiring the volume and breadth of data needed to train these models. We examined the feasibility, tolerability, patient characteristics, and data quality of a novel protocol for internet-based treatment research in psychiatry that may help advance this field.

METHODS

A fully internet-based protocol was used to gather repeated observational data from patient cohorts receiving internet-based cognitive behavioural therapy (iCBT) (N = 600) or antidepressant medication treatment (N = 110). At baseline, participants provided > 600 data points of self-report data, spanning socio-demographics, lifestyle, physical health, clinical and other psychological variables and completed 4 cognitive tests. They were followed weekly and completed another detailed clinical and cognitive assessment at week 4. In this paper, we describe our study design, the demographic and clinical characteristics of participants, their treatment adherence, study retention and compliance, the quality of the data gathered, and qualitative feedback from patients on study design and implementation.

RESULTS

Participant retention was 92% at week 3 and 84% for the final assessment. The relatively short study duration of 4 weeks was sufficient to reveal early treatment effects; there were significant reductions in 11 transdiagnostic psychiatric symptoms assessed, with the largest improvement seen for depression. Most participants (66%) reported being distracted at some point during the study, 11% failed 1 or more attention checks and 3% consumed an intoxicating substance. Data quality was nonetheless high, with near perfect 4-week test retest reliability for self-reported height (ICC = 0.97).

CONCLUSIONS

An internet-based methodology can be used efficiently to gather large amounts of detailed patient data during iCBT and antidepressant treatment. Recruitment was rapid, retention was relatively high and data quality was good. This paper provides a template methodology for future internet-based treatment studies, showing that such an approach facilitates data collection at a scale required for machine learning and other data-intensive methods that hope to deliver algorithmic tools that can aid clinical decision-making in psychiatry.

Predicting treatment outcome in depression: an introduction into current concepts and challenges

Improving response and remission rates in major depressive disorder (MDD) remains an important challenge. Matching patients to the treatment they will most likely respond to should be the ultimate goal. Even though numerous studies have investigated patient-specific indicators of treatment efficacy, no (bio)markers or empirical tests for use in clinical practice have resulted as of now. Therefore, clinical decisions regarding the treatment of MDD still have to be made on the basis of questionnaire- or interview-based assessments and general guidelines without the support of a (laboratory) test. We conducted a narrative review of current approaches to characterize and predict outcome to pharmacological treatments in MDD. We particularly focused on findings from newer computational studies using machine learning and on the resulting implementation into clinical decision support systems. The main issues seem to rest upon the unavailability of robust predictive variables and the lacking application of empirical findings and predictive models in clinical practice. We outline several challenges that need to be tackled on different stages of the translational process, from current concepts and definitions to generalizable prediction models and their successful implementation into digital support systems. By bridging the addressed gaps in translational psychiatric research, advances in data quantity and new technologies may enable the next steps toward precision psychiatry.

Evidence for machine learning guided early prediction of acute outcomes in the treatment of depressed children and adolescents with antidepressants

BACKGROUND

The treatment of depression in children and adolescents is a substantial public health challenge. This study examined artificial intelligence tools for the prediction of early outcomes in depressed children and adolescents treated with fluoxetine, duloxetine, or placebo.

METHODS

The study samples included training datasets (N = 271) from patients with major depressive disorder (MDD) treated with fluoxetine and testing datasets from patients with MDD treated with duloxetine (N = 255) or placebo (N = 265). Treatment trajectories were generated using probabilistic graphical models (PGMs). Unsupervised machine learning identified specific depressive symptom profiles and related thresholds of improvement during acute treatment.

RESULTS

Variation in six depressive symptoms (difficulty having fun, social withdrawal, excessive fatigue, irritability, low self-esteem, and depressed feelings) assessed with the Children's Depression Rating Scale-Revised at 4-6 weeks predicted treatment outcomes with fluoxetine at 10-12 weeks with an average accuracy of 73% in the training dataset. The same six symptoms predicted 10-12 week outcomes at 4-6 weeks in (a) duloxetine testing datasets with an average accuracy of 76% and (b) placebo-treated patients with accuracies of 67%. In placebo-treated patients, the accuracies of predicting response and remission were similar to antidepressants. Accuracies for predicting nonresponse to placebo treatment were significantly lower than antidepressants.

CONCLUSIONS

PGMs provided clinically meaningful predictions in samples of depressed children and adolescents treated with fluoxetine or duloxetine. Future work should augment PGMs with biological data for refined predictions to guide the selection of pharmacological and psychotherapeutic treatment in children and adolescents with depression.

Creating sparser prediction models of treatment outcome in depression: a proof-of-concept study using simultaneous feature selection and hyperparameter tuning

BACKGROUND

Predicting treatment outcome in major depressive disorder (MDD) remains an essential challenge for precision psychiatry. Clinical prediction models (CPMs) based on supervised machine learning have been a promising approach for this endeavor. However, only few CPMs have focused on model sparsity even though sparser models might facilitate the translation into clinical practice and lower the expenses of their application.

METHODS

In this study, we developed a predictive modeling pipeline that combines hyperparameter tuning and recursive feature elimination in a nested cross-validation framework. We applied this pipeline to a real-world clinical data set on MDD treatment response and to a second simulated data set using three different classification algorithms. Performance was evaluated by permutation testing and comparison to a reference pipeline without nested feature selection.

RESULTS

Across all models, the proposed pipeline led to sparser CPMs compared to the reference pipeline. Except for one comparison, the proposed pipeline resulted in equally or more accurate predictions. For MDD treatment response, balanced accuracy scores ranged between 61 and 71% when models were applied to hold-out validation data.

CONCLUSIONS

The resulting models might be particularly interesting for clinical applications as they could reduce expenses for clinical institutions and stress for patients.

Biomarkers as predictors of treatment response to tricyclic antidepressants in major depressive disorder: A systematic review

Tricyclic antidepressants (TCAs) are frequently prescribed in case of non-response to first-line antidepressants in Major Depressive Disorder (MDD). Treatment of MDD often entails a trial-and-error process of finding a suitable antidepressant and its appropriate dose. Nowadays, a shift is seen towards a more personalized treatment strategy in MDD to increase treatment efficacy. One of these strategies involves the use of biomarkers for the prediction of antidepressant treatment response. We aimed to summarize biomarkers for prediction of TCA specific (i.e. per agent, not for the TCA as a drug class) treatment response in unipolar nonpsychotic MDD. We performed a systematic search in PubMed and MEDLINE. After full-text screening, 36 papers were included. Seven genetic biomarkers were identified for nortriptyline treatment response. For desipramine, we identified two biomarkers; one genetic and one nongenetic. Three nongenetic biomarkers were identified for imipramine. None of these biomarkers were replicated. Quality assessment demonstrated that biomarker studies vary in endpoint definitions and frequently lack power calculations. None of the biomarkers can be confirmed as a predictor for TCA treatment response. Despite the necessity for TCA treatment optimization, biomarker studies reporting drug-specific results for TCAs are limited and adequate replication studies are lacking. Moreover, biomarker studies generally use small sample sizes. To move forward, larger cohorts, pooled data or biomarkers combined with other clinical characteristics should be used to improve predictive power.

Predicting the Treatment Outcomes of Antidepressants Using a Deep Neural Network of Deep Learning in Drug-Naïve Major Depressive Patients

Predicting the treatment response to antidepressants by pretreatment features would be useful, as up to 70–90% of patients with major depressive disorder (MDD) do not respond to treatment as expected. Therefore, we aim to establish a deep neural network (DNN) model of deep learning to predict the treatment outcomes of antidepressants in drug-naïve and first-diagnosis MDD patients during severe depressive stage using different domains of signature profiles of clinical features, peripheral biochemistry, psychosocial factors, and genetic polymorphisms. The multilayer feedforward neural network containing two hidden layers was applied to build models with tenfold cross-validation. The areas under the curve (AUC) of the receiver operating characteristic curves were used to evaluate the performance of the models. The results demonstrated that the AUCs of the model ranged between 0.7 and 0.8 using a combination of different domains of categorical variables. Moreover, models using the extracted variables demonstrated better performance, and the best performing model was characterized by an AUC of 0.825, using the levels of cortisol and oxytocin, scales of social support and quality of life, and polymorphisms of the OXTR gene. A complex interactions model developed through DNN could be useful at the clinical level for predicting the individualized outcomes of antidepressants.

Polygenic risk scores: An overview from bench to bedside for personalised medicine

Since the first polygenic risk score (PRS) in 2007, research in this area has progressed significantly. The increasing number of SNPs that have been identified by large scale GWAS analyses has fuelled the development of a myriad of PRSs for a wide variety of diseases and, more recently, to PRSs that potentially identify differential response to specific drugs. PRSs constitute a composite genomic biomarker and potential applications for PRSs in clinical practice encompass risk prediction and disease screening, early diagnosis, prognostication, and drug stratification to improve efficacy or reduce adverse drug reactions. Nevertheless, to our knowledge, no PRSs have yet been adopted into routine clinical practice. Beyond the technical considerations of PRS development, the major challenges that face PRSs include demonstrating clinical utility and circumnavigating the implementation of novel genomic technologies at scale into stretched healthcare systems. In this review, we discuss progress in developing disease susceptibility PRSs across multiple medical specialties, development of pharmacogenomic PRSs, and future directions for the field.

Machine learning in the prediction of depression treatment outcomes: a systematic review and meta-analysis

BACKGROUND

Multiple treatments are effective for major depressive disorder (MDD), but the outcomes of each treatment vary broadly among individuals. Accurate prediction of outcomes is needed to help select a treatment that is likely to work for a given person. We aim to examine the performance of machine learning methods in delivering replicable predictions of treatment outcomes.

METHODS

Of 7732 non-duplicate records identified through literature search, we retained 59 eligible reports and extracted data on sample, treatment, predictors, machine learning method, and treatment outcome prediction. A minimum sample size of 100 and an adequate validation method were used to identify adequate-quality studies. The effects of study features on prediction accuracy were tested with mixed-effects models. Fifty-four of the studies provided accuracy estimates or other estimates that allowed calculation of balanced accuracy of predicting outcomes of treatment.

RESULTS

Eight adequate-quality studies reported a mean accuracy of 0.63 [95% confidence interval (CI) 0.56-0.71], which was significantly lower than a mean accuracy of 0.75 (95% CI 0.72-0.78) in the other 46 studies. Among the adequate-quality studies, accuracies were higher when predicting treatment resistance (0.69) and lower when predicting remission (0.60) or response (0.56). The choice of machine learning method, feature selection, and the ratio of features to individuals were not associated with reported accuracy.

CONCLUSIONS

The negative relationship between study quality and prediction accuracy, combined with a lack of independent replication, invites caution when evaluating the potential of machine learning applications for personalizing the treatment of depression.

Effectiveness of common antidepressants: a post market release study

BACKGROUND

This study summarizes the experiences of patients, who have multiple comorbidities, with 15 mono-treated antidepressants.

METHODS

This is a retrospective, observational, matched case control study. The cohort was organized using claims data available through OptumLabs for depressed patients treated with antidepressants between January 1, 2001 and December 31, 2018. The cohort included patients from all states within United States of America. The analysis focused on 3,678,082 patients with major depression who had 10,221,145 antidepressant treatments. Using the robust, and large predictors of remission, and propensity to prescribe an antidepressant, the study created 16,770 subgroups of patients. The study reports the remission rate for the antidepressants within the subgroups. The overall impact of antidepressant on remission was calculated as the common odds ratio across the strata.

FINDINGS

The study accurately modelled clinicians' prescription patterns (cross-validated Area under the Receiver Operating Curve, AROC, of 82.0%, varied from 77% to 90%) and patients' remission (cross-validated AROC of 72.0%, varied from 69.5% to 78%). In different strata, contrary to published randomized studies, remission rates differed significantly and antidepressants were not equally effective. For example, in age and gender subgroups, the best antidepressant had an average remission rate of 50.78%, 1.5 times higher than the average antidepressant (30.30% remission rate) and 20 times higher than the worst antidepressant. The Breslow-Day chi-square test for homogeneity showed that across strata a homogenous common odds-ratio did not exist (alpha<0.0001). Therefore, the choice of the optimal antidepressant depended on the strata defined by the patient's medical history.

INTERPRETATION

Study findings may not be appropriate for specific patients. To help clinicians assess the transferability of study findings to specific patient, the web site http://hi.gmu.edu/ad assesses the patient's medical history, finds similar cases in our data, and recommends an antidepressant based on the experience of remission in our data. Patients can share this site's recommendations with their clinicians, who can then assess the appropriateness of the recommendations.

FUNDING

This project was funded by the Robert Wood Johnson foundation grant #76786. The development of related web site was supported by grant 247-02-20 from Virginia's Commonwealth Health Research Board.

Multilevel Survival Modeling with Structured Penalties for Disease Prediction from Imaging Genetics data

This paper introduces a framework for disease prediction from multimodal genetic and imaging data. We propose a multilevel survival model which allows predicting the time of occurrence of a future disease state in patients initially exhibiting mild symptoms. This new multilevel setting allows modeling the interactions between genetic and imaging variables. This is in contrast with classical additive models which treat all modalities in the same manner and can result in undesirable elimination of specific modalities when their contributions are unbalanced. Moreover, the use of a survival model allows overcoming the limitations of previous approaches based on classification which consider a fixed time frame. Furthermore, we introduce specific penalties taking into account the structure of the different types of data, such as a group lasso penalty over the genetic modality and a L2-penalty over the imaging modality. Finally, we propose a fast optimization algorithm, based on a proximal gradient method. The approach was applied to the prediction of Alzheimer's disease (AD) among patients with mild cognitive impairment (MCI) based on genetic (single nucleotide polymorphisms - SNP) and imaging (anatomical MRI measures) data from the ADNI database. The experiments demonstrate the effectiveness of the method for predicting the time of conversion to AD. It revealed how genetic variants and brain imaging alterations interact in the prediction of future disease status. The approach is generic and could potentially be useful for the prediction of other diseases.

The promise of machine learning in predicting treatment outcomes in psychiatry

For many years, psychiatrists have tried to understand factors involved in response to medications or psychotherapies, in order to personalize their treatment choices. There is now a broad and growing interest in the idea that we can develop models to personalize treatment decisions using new statistical approaches from the field of machine learning and applying them to larger volumes of data. In this pursuit, there has been a paradigm shift away from experimental studies to confirm or refute specific hypotheses towards a focus on the overall explanatory power of a predictive model when tested on new, unseen datasets. In this paper, we review key studies using machine learning to predict treatment outcomes in psychiatry, ranging from medications and psychotherapies to digital interventions and neurobiological treatments. Next, we focus on some new sources of data that are being used for the development of predictive models based on machine learning, such as electronic health records, smartphone and social media data, and on the potential utility of data from genetics, electrophysiology, neuroimaging and cognitive testing. Finally, we discuss how far the field has come towards implementing prediction tools in real-world clinical practice. Relatively few retrospective studies to-date include appropriate external validation procedures, and there are even fewer prospective studies testing the clinical feasibility and effectiveness of predictive models. Applications of machine learning in psychiatry face some of the same ethical challenges posed by these techniques in other areas of medicine or computer science, which we discuss here. In short, machine learning is a nascent but important approach to improve the effectiveness of mental health care, and several prospective clinical studies suggest that it may be working already.

Genome-wide analysis suggests the importance of vascular processes and neuroinflammation in late-life antidepressant response

Antidepressant outcomes in older adults with depression is poor, possibly because of comorbidities such as cerebrovascular disease. Therefore, we leveraged multiple genome-wide approaches to understand the genetic architecture of antidepressant response. Our sample included 307 older adults (≥60 years) with current major depression, treated with venlafaxine extended-release for 12 weeks. A standard genome-wide association study (GWAS) was conducted for post-treatment remission status, followed by in silico biological characterization of associated genes, as well as polygenic risk scoring for depression, neurodegenerative and cerebrovascular disease. The top-associated variants for remission status and percentage symptom improvement were PIEZO1 rs12597726 (OR = 0.33 [0.21, 0.51], p = 1.42 × 10^-6) and intergenic rs6916777 (Beta = 14.03 [8.47, 19.59], p = 1.25 × 10^-6), respectively. Pathway analysis revealed significant contributions from genes involved in the ubiquitin-proteasome system, which regulates intracellular protein degradation with has implications for inflammation, as well as atherosclerotic cardiovascular disease (n = 25 of 190 genes, p = 8.03 × 10^-6, FDR-corrected p = 0.01). Given the polygenicity of complex outcomes such as antidepressant response, we also explored 11 polygenic risk scores associated with risk for Alzheimer's disease and stroke. Of the 11 scores, risk for cardioembolic stroke was the second-best predictor of non-remission, after being male (Accuracy = 0.70 [0.59, 0.79], Sensitivity = 0.72, Specificity = 0.67; p = 2.45 × 10^-4). Although our findings did not reach genome-wide significance, they point to previously-implicated mechanisms and provide support for the roles of vascular and inflammatory pathways in LLD. Overall, significant enrichment of genes involved in protein degradation pathways that may be impaired, as well as the predictive capacity of risk for cardioembolic stroke, support a link between late-life depression remission and risk for vascular dysfunction.

Inclusion of genetic variants in an ensemble of gradient boosting decision trees does not improve the prediction of citalopram treatment response

Identifying in advance who is unlikely to respond to a specific antidepressant treatment is crucial to precision medicine efforts. The current work leverages genome-wide genetic variation and machine learning to predict response to the antidepressant citalopram using data from the Sequenced Treatment Alternatives to Relieve Depression (STAR*D) trial (n = 1257 with both valid genomic and outcome data). A confirmatory approach selected 11 SNPs previously reported to predict response to escitalopram in a sample different from the current study. A novel exploratory approach selected SNPs from across the genome using nested cross-validation with elastic net logistic regression with a predominantly lasso penalty (alpha = 0.99). SNPs from each approach were combined with baseline clinical predictors and treatment response outcomes were predicted using a stacked ensemble of gradient boosting decision trees. Using pre-treatment clinical and symptom predictors only, out-of-fold prediction of a novel treatment response definition based on STAR*D treatment guidelines was acceptable, AUC = .659, 95% CI [0.629, 0.689]. The inclusion of SNPs using confirmatory or exploratory selection methods did not improve the out-of-fold prediction of treatment response (AUCs were .662, 95% CI [0.632, 0.692] and .655, 95% CI [0.625, 0.685], respectively). A similar pattern of results were observed for the secondary outcomes of the presence or absence of distressing side effects regardless of treatment response and achieving remission or satisfactory partial response, assuming medication tolerance. In the current study, incorporating SNP variation into prognostic models did not enhance the prediction of citalopram response in the STAR*D sample.

Could Polygenic Risk Scores Be Useful in Psychiatry?: A Review

IMPORTANCE

Polygenic risk scores (PRS) are predictors of the genetic susceptibility to diseases, calculated for individuals as weighted counts of thousands of risk variants in which the risk variants and their weights have been identified in genome-wide association studies. Polygenic risk scores show promise in aiding clinical decision-making in many areas of medical practice. This review evaluates the potential use of PRS in psychiatry.

OBSERVATIONS

On their own, PRS will never be able to establish or definitively predict a diagnosis of common complex conditions (eg, mental health disorders), because genetic factors only contribute part of the risk and PRS will only ever capture part of the genetic contribution. Combining PRS with other risk factors has potential to improve outcome prediction and aid clinical decision-making (eg, determining follow-up options for individuals seeking help who are at clinical risk of future illness). Prognostication of adverse physical health outcomes or response to treatment in clinical populations are of great interest for psychiatric practice, but data from larger samples are needed to develop and evaluate PRS.

CONCLUSIONS AND RELEVANCE

Polygenic risk scores will contribute to risk assessment in clinical psychiatry as it evolves to combine information from molecular, clinical, and lifestyle metrics. The genome-wide genotype data needed to calculate PRS are inexpensive to generate and could become available to psychiatrists as a by-product of practices in other medical specialties. The utility of PRS in clinical psychiatry, as well as ethical issues associated with their use, should be evaluated in the context of realistic expectations of what PRS can and cannot deliver. Clinical psychiatry has lagged behind other fields of health care in its use of new technologies and routine clinical data for research. Now is the time to catch up.

Using a simulation centre to evaluate preliminary acceptability and impact of an artificial intelligence-powered clinical decision support system for depression treatment on the physician-patient interaction

BACKGROUND

Recently, artificial intelligence-powered devices have been put forward as potentially powerful tools for the improvement of mental healthcare. An important question is how these devices impact the physician-patient interaction.

AIMS

Aifred is an artificial intelligence-powered clinical decision support system (CDSS) for the treatment of major depression. Here, we explore the use of a simulation centre environment in evaluating the usability of Aifred, particularly its impact on the physician-patient interaction.

METHOD

Twenty psychiatry and family medicine attending staff and residents were recruited to complete a 2.5-h study at a clinical interaction simulation centre with standardised patients. Each physician had the option of using the CDSS to inform their treatment choice in three 10-min clinical scenarios with standardised patients portraying mild, moderate and severe episodes of major depression. Feasibility and acceptability data were collected through self-report questionnaires, scenario observations, interviews and standardised patient feedback.

RESULTS

All 20 participants completed the study. Initial results indicate that the tool was acceptable to clinicians and feasible for use during clinical encounters. Clinicians indicated a willingness to use the tool in real clinical practice, a significant degree of trust in the system's predictions to assist with treatment selection, and reported that the tool helped increase patient understanding of and trust in treatment. The simulation environment allowed for the evaluation of the tool's impact on the physician-patient interaction.

CONCLUSIONS

The simulation centre allowed for direct observations of clinician use and impact of the tool on the clinician-patient interaction before clinical studies. It may therefore offer a useful and important environment in the early testing of new technological tools. The present results will inform further tool development and clinician training materials.

Combining machine learning algorithms for prediction of antidepressant treatment response

OBJECTIVES

Predictors for unfavorable treatment outcome in major depressive disorder (MDD) applicable for treatment selection are still lacking. The database of a longitudinal multicenter study on 1079 acutely depressed patients, performed by the German research network on depression (GRND), allows supervised and unsupervised learning to further elucidate the interplay of clinical and psycho-sociodemographic variables and their predictive impact on treatment outcome phenotypes.

EXPERIMENTAL PROCEDURES

Treatment response was defined by a change of HAM-D 17-item baseline score ≥50% and remission by the established threshold of ≤7, respectively, after up to eight weeks of inpatient treatment. After hierarchical symptom clustering and stratification by treatment subtypes (serotonin reuptake inhibitors, tricyclic antidepressants, antipsychotic, and lithium augmentation), prediction models for different outcome phenotypes were computed with random forest in a cross-center validation design. In total, 88 predictors were implemented.

RESULTS

Clustering revealed four distinct HAM-D subscores related to emotional, anxious, sleep, and appetite symptoms, respectively. After feature selection, classification models reached moderate to high accuracies up to 0.85. Highest accuracies were observed for the SSRI and TCA subgroups and for sleep and appetite symptoms, while anxious symptoms showed poor predictability.

CONCLUSION

Our results support a decisive role for machine learning in the management of antidepressant treatment. Treatment- and symptom-specific algorithms may increase accuracies by reducing heterogeneity. Especially, predictors related to duration of illness, baseline depression severity, anxiety and somatic symptoms, and personality traits moderate treatment success. However, prospectives application of machine learning models will be necessary to prove their value for the clinic.

Comparison Between Expression Microarrays and RNA-Sequencing Using UKBEC Dataset Identified a trans-eQTL Associated with MPZ Gene in Substantia Nigra

In recent years, the advantages of RNA-sequencing (RNA-Seq) have made it the platform of choice for measuring gene expression over traditional microarrays. However, RNA-Seq comes with bioinformatical challenges and higher computational costs. Therefore, this study set out to assess whether the increased depth of transcriptomic information facilitated by RNA-Seq is worth the increased computation over microarrays, specifically at three levels: absolute expression levels, differentially expressed genes identification, and expression QTL (eQTL) mapping in regions of the human brain. Using the United Kingdom Brain Expression Consortium (UKBEC) dataset, there is high agreement of gene expression levels measured by microarrays and RNA-seq when quantifying absolute expression levels and when identifying differentially expressed genes. These findings suggest that depending on the aims of a study, the relative ease of working with microarray data may outweigh the computational time and costs of RNA-Seq pipelines. On the other, there was low agreement when mapping eQTLs. However, a number of eQTLs associated with genes that play important roles in the brain were found in both platforms. For example, a trans-eQTL was mapped that is associated with the MPZ gene in the substantia nigra. These eQTLs that we have highlighted are extremely promising candidates that merit further investigation.

Genetics and major depressive disorder: clinical implications for disease risk, prognosis and treatment

In the post-genomic era, genetics has led to limited clinical applications in the diagnosis and treatment of major depressive disorder (MDD). Variants in genes coding for cytochrome enzymes are included in guidelines for assisting in antidepressant choice and dosing, but there are no recommendations involving genes responsible for antidepressant pharmacodynamics and no consensus applications for guiding diagnosis or prognosis. However, genetics has contributed to a better understanding of MDD pathogenesis and the mechanisms of antidepressant action, also thanks to recent methodological innovations that overcome the challenges posed by the polygenic architecture of these traits. Polygenic risk scores can be used to estimate the risk of disease at the individual level, which may have clinical relevance in cases with extremely high scores (e.g. top 1%). Genetic studies have also shed light on a wide genetic overlap between MDD and other psychiatric disorders. The relationships between genes/pathways associated with MDD and known drug targets are a promising tool for drug repurposing and identification of new pharmacological targets. Increase in power thanks to larger samples and methods integrating genetic data with gene expression, the integration of common variants and rare variants, are expected to advance our knowledge and assist in personalized psychiatry.

'Precision' or 'personalized' psychiatry: different terms - same content?

Due to the increased lifetime prevalence and personal, social, and economic burden of mental disorders, psychiatry is in need of a significant change in several aspects of its clinical and research approaches. Over the last few decades, the development of personalized / precision medicine in psychiatry focusing on tailored therapies that fit each patient's unique individual, physiological, and genetic profile has not achieved the same results as those obtained in other branches, such as oncology. The long-awaited revolution has not yet surfaced. There are various explanations for this including imprecise diagnostic criteria, incomplete understanding of the molecular pathology involved, absence of available clinical tools and, finally, the characteristics of the patient. Since then, the co-existence of the two terms has sparked a great deal of discussion around the definition and differentiation between the two types of psychiatry, as they often seem similar or even superimposable. Generally, the two terminologies are used indiscriminately, alternatively, and / or separately, within the same scientific works. In this paper, an overview is provided on the overlap between the application and meaning of the terms 'precision psychiatry' and 'personalized psychiatry'.

The State of Our Understanding of the Pathophysiology and Optimal Treatment of Depression: Glass Half Full or Half Empty?

Major depressive disorder is a remarkably common and often severe psychiatric disorder associated with high levels of morbidity and mortality. Patients with major depression are prone to several comorbid psychiatric conditions, including posttraumatic stress disorder, anxiety disorders, obsessive-compulsive disorder, and substance use disorders, and medical conditions, including cardiovascular disease, diabetes, stroke, cancer, which, coupled with the risk of suicide, result in a shortened life expectancy. The goal of this review is to provide an overview of our current understanding of major depression, from pathophysiology to treatment. In spite of decades of research, relatively little is known about its pathogenesis, other than that risk is largely defined by a combination of ill-defined genetic and environmental factors. Although we know that female sex, a history of childhood maltreatment, and family history as well as more recent stressors are risk factors, precisely how these environmental influences interact with genetic vulnerability remains obscure. In recent years, considerable advances have been made in beginning to understand the genetic substrates that underlie disease vulnerability, and the interaction of genes, early-life adversity, and the epigenome in influencing gene expression is now being intensively studied. The role of inflammation and other immune system dysfunction in the pathogenesis of major depression is also being intensively investigated. Brain imaging studies have provided a firmer understanding of the circuitry involved in major depression, providing potential new therapeutic targets. Despite a broad armamentarium for major depression, including antidepressants, evidence-based psychotherapies, nonpharmacological somatic treatments, and a host of augmentation strategies, a sizable percentage of patients remain nonresponsive or poorly responsive to available treatments. Investigational agents with novel mechanisms of action are under active study. Personalized medicine in psychiatry provides the hope of escape from the current standard trial-and-error approach to treatment, moving to a more refined method that augurs a new era for patients and clinicians alike.

Predicting second-generation antidepressant effectiveness in treating sadness using demographic and clinical information: A machine learning approach

INTRODUCTION

Current guidelines for choosing antidepressant medications involve a trial-and-error process. Most patients try multiple antidepressants before finding an effective antidepressant. This study uses demographic and clinical information to create models predicting effectiveness of different antidepressants in treating sadness in a nationally representative sample of US adults.

METHODS

A secondary analysis of the Collaborative Psychiatric Epidemiology Survey (CPES) was performed. Participants with or without a mental health diagnosis who reported sadness as a symptom, and were taking fluoxetine (n=156), sertraline (n=224), citalopram (n=91), paroxetine (n=156), venlafaxine (n=69), bupropion (n=92), or trazadone (n=26) within the past year were included. Two sets of principal component analyses (PCAs) and logistic regressions were performed: one determined associations between symptom clusters and antidepressant effectiveness for sadness, and the other created models to predict effectiveness. Both PCAs controlled for psychiatric and medical diagnoses, substance use, psychiatric medications, alternative treatments, and demographics.

RESULTS

Anxiety was associated with ineffectiveness of fluoxetine in treating sadness. Low mood scores were associated with ineffectiveness of paroxetine and venlafaxine, and fatigue was associated with ineffectiveness of sertraline. The models for predicting drug effectiveness had a mean accuracy of 83% and internal validity of 72%.

LIMITATIONS

CPES data were collected from 2001-2003, so newer drugs were not included. Effectiveness was for sadness, so results are not directly comparable to studies using overall depressive symptom reductions as outcomes.

CONCLUSION

Since fewer than 50% of patients currently respond to their first antidepressant, this model could provide modest improvement to choosing starting antidepressants in treating sadness.

Personalized Psychiatry and Depression: The Role of Sociodemographic and Clinical Variables

Despite several pharmacological options, the clinical outcomes of major depressive disorder (MDD) are often unsatisfactory. Personalized psychiatry attempts to tailor therapeutic interventions according to each patient's unique profile and characteristics. This approach can be a crucial strategy in improving pharmacological outcomes in MDD and overcoming trial-and-error treatment choices. In this narrative review, we evaluate whether sociodemographic (i.e., gender, age, race/ethnicity, and socioeconomic status) and clinical [i.e., body mass index (BMI), severity of depressive symptoms, and symptom profiles] variables that are easily assessable in clinical practice may help clinicians to optimize the selection of antidepressant treatment for each patient with MDD at the early stages of the disorder. We found that several variables were associated with poorer outcomes for all antidepressants. However, only preliminary associations were found between some clinical variables (i.e., BMI, anhedonia, and MDD with melancholic/atypical features) and possible benefits with some specific antidepressants. Finally, in clinical practice, the assessment of sociodemographic and clinical variables considered in our review can be valuable for early identification of depressed individuals at high risk for poor responses to antidepressants, but there are not enough data on which to ground any reliable selection of specific antidepressant class or compounds. Recent advances in computational resources, such as machine learning techniques, which are able to integrate multiple potential predictors, such as individual/ clinical variables, biomarkers, and genetic factors, may offer future reliable tools to guide personalized antidepressant choice for each patient with MDD.

A polygenic predictor of treatment-resistant depression using whole exome sequencing and genome-wide genotyping

Treatment-resistant depression (TRD) occurs in ~30% of patients with major depressive disorder (MDD) but the genetics of TRD was previously poorly investigated. Whole exome sequencing and genome-wide genotyping were available in 1209 MDD patients after quality control. Antidepressant response was compared to non-response to one treatment and non-response to two or more treatments (TRD). Differences in the risk of carrying damaging variants were tested. A score expressing the burden of variants in genes and pathways was calculated weighting each variant for its functional (Eigen) score and frequency. Gene-based and pathway-based scores were used to develop predictive models of TRD and non-response using gradient boosting in 70% of the sample (training) which were tested in the remaining 30% (testing), evaluating also the addition of clinical predictors. Independent replication was tested in STAR*D and GENDEP using exome array-based data. TRD and non-responders did not show higher risk to carry damaging variants compared to responders. Genes/pathways associated with TRD included those modulating cell survival and proliferation, neurodegeneration, and immune response. Genetic models showed significant prediction of TRD vs. response and they were improved by the addition of clinical predictors, but they were not significantly better than clinical predictors alone. Replication results were driven by clinical factors, except for a model developed in subjects treated with serotonergic antidepressants, which showed a clear improvement in prediction at the extremes of the genetic score distribution in STAR*D. These results suggested relevant biological mechanisms implicated in TRD and a new methodological approach to the prediction of TRD.

Using structural MRI to identify bipolar disorders - 13 site machine learning study in 3020 individuals from the ENIGMA Bipolar Disorders Working Group

Bipolar disorders (BDs) are among the leading causes of morbidity and disability. Objective biological markers, such as those based on brain imaging, could aid in clinical management of BD. Machine learning (ML) brings neuroimaging analyses to individual subject level and may potentially allow for their diagnostic use. However, fair and optimal application of ML requires large, multi-site datasets. We applied ML (support vector machines) to MRI data (regional cortical thickness, surface area, subcortical volumes) from 853 BD and 2167 control participants from 13 cohorts in the ENIGMA consortium. We attempted to differentiate BD from control participants, investigated different data handling strategies and studied the neuroimaging/clinical features most important for classification. Individual site accuracies ranged from 45.23% to 81.07%. Aggregate subject-level analyses yielded the highest accuracy (65.23%, 95% CI = 63.47-67.00, ROC-AUC = 71.49%, 95% CI = 69.39-73.59), followed by leave-one-site-out cross-validation (accuracy = 58.67%, 95% CI = 56.70-60.63). Meta-analysis of individual site accuracies did not provide above chance results. There was substantial agreement between the regions that contributed to identification of BD participants in the best performing site and in the aggregate dataset (Cohen's Kappa = 0.83, 95% CI = 0.829-0.831). Treatment with anticonvulsants and age were associated with greater odds of correct classification. Although short of the 80% clinically relevant accuracy threshold, the results are promising and provide a fair and realistic estimate of classification performance, which can be achieved in a large, ecologically valid, multi-site sample of BD participants based on regional neurostructural measures. Furthermore, the significant classification in different samples was based on plausible and similar neuroanatomical features. Future multi-site studies should move towards sharing of raw/voxelwise neuroimaging data.

Multi-omics signature of brain amyloid deposition in asymptomatic individuals at-risk for Alzheimer's disease: The INSIGHT-preAD study

BACKGROUND

One of the biggest challenge in Alzheimer's disease (AD) is to identify pathways and markers of disease prediction easily accessible, for prevention and treatment. Here we analysed blood samples from the INveStIGation of AlzHeimer's predicTors (INSIGHT-preAD) cohort of elderly asymptomatic individuals with and without brain amyloid load.

METHODS

We performed blood RNAseq, and plasma metabolomics and lipidomics using liquid chromatography-mass spectrometry on 48 individuals amyloid positive and 48 amyloid negative (SUVr cut-off of 0·7918). The three data sets were analysed separately using differential gene expression based on negative binomial distribution, non-parametric (Wilcoxon) and parametric (correlation-adjusted Student't) tests. Data integration was conducted using sparse partial least squares-discriminant and principal component analyses. Bootstrap-selected top-ten features from the three data sets were tested for their discriminant power using Receiver Operating Characteristic curve. Longitudinal metabolomic analysis was carried out on a subset of 22 subjects.

FINDINGS

Univariate analyses identified three medium chain fatty acids, 4-nitrophenol and a set of 64 transcripts enriched for inflammation and fatty acid metabolism differentially quantified in amyloid positive and negative subjects. Importantly, the amounts of the three medium chain fatty acids were correlated over time in a subset of 22 subjects (p < 0·05). Multi-omics integrative analyses showed that metabolites efficiently discriminated between subjects according to their amyloid status while lipids did not and transcripts showed trends. Finally, the ten top metabolites and transcripts represented the most discriminant omics features with 99·4% chance prediction for amyloid positivity.

INTERPRETATION

This study suggests a potential blood omics signature for prediction of amyloid positivity in asymptomatic at-risk subjects, allowing for a less invasive, more accessible, and less expensive risk assessment of AD as compared to PET studies or lumbar puncture. FUND: Institut Hospitalo-Universitaire and Institut du Cerveau et de la Moelle Epiniere (IHU-A-ICM), French Ministry of Research, Fondation Alzheimer, Pfizer, and Avid.

Results of the European Group for the Study of Resistant Depression (GSRD) - basis for further research and clinical practice

Objectives: The overview outlines two decades of research from the European Group for the Study of Resistant Depression (GSRD) that fundamentally impacted evidence-based algorithms for diagnostics and psychopharmacotherapy of treatment-resistant depression (TRD). Methods: The GSRD staging model characterising response, non-response and resistance to antidepressant (AD) treatment was applied to 2762 patients in eight European countries. Results: In case of non-response, dose escalation and switching between different AD classes did not show superiority over continuation of original AD treatment. Predictors for TRD were symptom severity, duration of the current major depressive episode (MDE), suicidality, psychotic and melancholic features, comorbid anxiety and personality disorders, add-on treatment, non-response to the first AD, adverse effects, high occupational level, recurrent disease course, previous hospitalisations, positive family history of MDD, early age of onset and novel associations of single nucleoid polymorphisms (SNPs) within the PPP3CC, ST8SIA2, CHL1, GAP43 and ITGB3 genes and gene pathways associated with neuroplasticity, intracellular signalling and chromatin silencing. A prediction model reaching accuracy of above 0.7 highlighted symptom severity, suicidality, comorbid anxiety and lifetime MDEs as the most informative predictors for TRD. Applying machine-learning algorithms, a signature of three SNPs of the BDNF, PPP3CC and HTR2A genes and lacking melancholia predicted treatment response. Conclusions: The GSRD findings offer a unique and balanced perspective on TRD representing foundation for further research elaborating on specific clinical and genetic hypotheses and treatment strategies within appropriate study-designs, especially interaction-based models and randomized controlled trials.

Pharmacogenomics-Driven Prediction of Antidepressant Treatment Outcomes: A Machine-Learning Approach With Multi-trial Replication

We set out to determine whether machine learning–based algorithms that included functionally validated pharmacogenomic biomarkers joined with clinical measures could predict selective serotonin reuptake inhibitor (SSRI) remission/response in patients with major depressive disorder (MDD). We studied 1,030 white outpatients with MDD treated with citalopram/escitalopram in the Mayo Clinic Pharmacogenomics Research Network Antidepressant Medication Pharmacogenomic Study (PGRN‐AMPS; n = 398), Sequenced Treatment Alternatives to Relieve Depression (STAR*D; n = 467), and International SSRI Pharmacogenomics Consortium (ISPC; n = 165) trials. A genomewide association study for PGRN‐AMPS plasma metabolites associated with SSRI response (serotonin) and baseline MDD severity (kynurenine) identified single nucleotide polymorphisms (SNPs) in DEFB1,ERICH3,AHR, and TSPAN5 that we tested as predictors. Supervised machine‐learning methods trained using SNPs and total baseline depression scores predicted remission and response at 8 weeks with area under the receiver operating curve (AUC) > 0.7 (P < 0.04) in PGRN‐AMPS patients, with comparable prediction accuracies > 69% (P ≤ 0.07) in STAR*D and ISPC. These results demonstrate that machine learning can achieve accurate and, importantly, replicable prediction of SSRI therapy response using total baseline depression severity combined with pharmacogenomic biomarkers.

Big data in pharmacogenomics: current applications, perspectives and pitfalls

The efficiency of new generation sequencing methods and the reduction of their cost has led pharmacogenomics to gradually supplant pharmacogenetics, leading to new applications in personalized medicine along with new perspectives in drug design or identification of drug response factors. The amount of data generated in genomics fits the definition of big data, and need a specific bioinformatics processing following standard steps: data collection, processing, analysis and interpretation. Pitfalls of pharmacogenomics studies are directly related to these steps. This review aims to describe these steps from a pharmacogenomic point of view, focusing on bioinformatics aspects.

Na⁺, K⁺-ATPase Signaling and Bipolar Disorder

Bipolar disorder (BD) is a severe and common chronic mental illness characterized by recurrent mood swings between depression and mania. The biological basis of the disease is poorly understood and its treatment is unsatisfactory. Although in past decades the "monoamine hypothesis" has dominated our understanding of both the pathophysiology of depressive disorders and the action of pharmacological treatments, recent studies focus on the involvement of additional neurotransmitters/neuromodulators systems and cellular processes in BD. Here, evidence for the participation of Na⁺, K⁺-ATPase and its endogenous regulators, the endogenous cardiac steroids (ECS), in the etiology of BD is reviewed. Proof for the involvement of brain Na⁺, K⁺-ATPase and ECS in behavior is summarized and it is hypothesized that ECS-Na⁺, K⁺-ATPase-induced activation of intracellular signaling participates in the mechanisms underlying BD. We propose that the activation of ERK, AKT, and NFκB, resulting from ECS-Na⁺, K⁺-ATPase interaction, modifies neuronal activity and neurotransmission which, in turn, participate in the regulation of behavior and BD. These observations suggest Na⁺, K⁺-ATPase-mediated signaling is a potential target for drug development for the treatment of BD.

Effects of personality on the association between paroxetine plasma concentration and response

BACKGROUND

We studied the differences between groups that were divided according to personality characteristics with respect to the relationship between drug concentration and symptom improvement.

METHODS

A total of 120 patients with major depressive disorder were treated with paroxetine for 6 weeks, and 89 patients completed the protocol. The Montgomery-Åsberg Depression Rating Scale (MADRS) was used to evaluate the patients. Patients' paroxetine plasma concentrations at week 6 were measured. Their personalities were evaluated by the Temperament and Character Inventory (TCI) at the first visit. We divided the patients into two groups according to the median of each TCI dimension. We compared the responder rate between "high" and "low" groups in each TCI dimension and analyzed Pearson's correlation coefficients of paroxetine plasma concentration and MADRS-improvement rate.

RESULTS

A total of 62 patients completed the TCI. Low-novelty-seeking, high-harm-avoidance, low-reward-dependence, and low-self-directedness groups exhibited significant negative correlations between paroxetine plasma concentration and MADRS improvement. Among the groups with combined personality traits, the high-harm-avoidance and low-self-directedness groups showed a markedly significant negative correlation.

CONCLUSION

Patients with depression exhibiting specific personality traits, especially those with high harm-avoidance and low self-directedness scores, exhibited a significant negative association between paroxetine plasma concentration and MADRS-improvement rate. Therefore, a lower dose might be suitable for patients with specific personality traits.