Faculty Bibliography

BACKGROUND:The Tobacco, Alcohol, Prescription Medication, and Other Substance (TAPS) tool is a screening and brief assessment instrument to identify unhealthy tobacco, alcohol, drug use, and prescription medication use in primary care patients. This secondary analysis compares the TAPS tool to the Alcohol Use Disorders Identification Test-Consumption (AUDIT-C) for alcohol screening. METHODS:Adult primary care patients (1124 female, 874 male) completed the TAPS tool followed by AUDIT-C. Performance of each instrument was evaluated against a reference standard measure, the modified World Mental Health Composite International Diagnostic Interview, to identify problem use and alcohol use disorder (AUD). Area under the curve (AUC) appraised discrimination, and sensitivity and specificity were calculated for Youden optimal score thresholds. RESULTS:For identifying problem use: On the AUDIT-C, AUC was 0.90 (95% Confidence Interval: 0.86-0.92) for females and 0.91 (0.89-0.93) for males. Sensitivity and specificity for females were 0.89 (0.83-0.93) and 0.78 (0.75-0.80), respectively, and for males were 0.84 (0.79-0.88) and 0.82 (0.79-0.85). On the TAPS tool, AUC was 0.82 (0.79-0.86) for females and 0.81 (0.78-0.84) for males. Sensitivity and specificity for females were 0.78 (0.72-0.84) and 0.78 (0.75-0.81), respectively, and for males were 0.76 (0.71-0.81) and 0.76 (0.72-0.79). For AUD: On the AUDIT-C, AUC was 0.90 (0.88-0.93) for both females and males. Sensitivity and specificity for females were 0.83 (0.74-0.90) and 0.83 (0.80-0.85), respectively, while for males, they were 0.81 (0.74-0.87) and 0.84 (0.81-0.87). On the TAPS tool, AUC was 0.84 (0.80-0.89) for females and 0.82 (0.78-0.86) for males. Sensitivity and specificity for females were 0.73 (0.63-0.81) and 0.85 (0.83-0.88), respectively, while for males, they were 0.75 (0.68-0.81) and 0.84 (0.81-0.86). CONCLUSION/CONCLUSIONS:The AUDIT-C performed somewhat better than the TAPS tool for alcohol screening. However, the TAPS tool had an acceptable level of performance for alcohol screening and may be advantageous in practice settings seeking to identify alcohol and other substance use with a single instrument.

Neuroimaging-based subtyping is increasingly used to explain heterogeneity in psychiatric disorders. However, the clinical utility of these subtyping efforts remains unclear, and replication has been challenging. Here we examined how the choice of neuroimaging measures influences the derivation of neuro-subtypes and the consequences for clinical delineation. On a clinically heterogeneous dataset (total n = 566) that included controls (n = 268) and cases (n = 298) of psychiatric conditions, including individuals diagnosed with post-traumatic stress disorder (PTSD), traumatic brain injury (TBI), and comorbidity of both (PTSD&TBI), we identified neuro-subtypes among the cases using either structural, resting-state, or task-based measures. The neuro-subtypes for each modality had high internal validity but did not significantly differ in their clinical and cognitive profiles. We further show that the choice of neuroimaging measures for subtyping substantially impacts the identification of neuro-subtypes, leading to low concordance across subtyping solutions. Similar variability in neuro-subtyping was found in an independent dataset (n = 1642) comprised of major depression disorder (MDD, n = 848) and controls (n = 794). Our results suggest that the highly anticipated relationships between neuro-subtypes and clinical features may be difficult to discover.

We introduce a simple variant of a Purely Random Forest, an Absolute Random Forest (ARF) for clustering. At every node splits of units are determined by a randomly chosen feature and a random threshold drawn from a uniform distribution whose support, the range of the selected feature in the root node, does not change. This enables closed-form estimators of parameters, such as pairwise proximities, to be obtained without having to grow a forest. The probabilistic structure corresponding to an ARF is called a Treeless Absolute Random Forest (TARF). With high probability, the algorithm will split units whose feature vectors are far apart and keep together units whose feature vectors are similar. Thus, the underlying structure of the data drives the growth of the tree. The expected value of pairwise proximities is obtained for three pathway functions. One, a completely common pathway function, is an indicator of whether a pair of units follow the same path from the root to the leaf node. The properties of TARF-based proximity estimators for clustering and classification are compared to other methods in eight real-world data sets and in simulations. Results show substantial performance and computing efficiencies of particular value for large data sets.

Post-traumatic stress disorder (PTSD) is a mental disorder diagnosed by clinical interviews, self-report measures and neuropsychological testing. Traumatic brain injury (TBI) can have neuropsychiatric symptoms similar to PTSD. Diagnosing PTSD and TBI is challenging and more so for providers lacking specialized training facing time pressures in primary care and other general medical settings. Diagnosis relies heavily on patient self-report and patients frequently under-report or over-report their symptoms due to stigma or seeking compensation. We aimed to create objective diagnostic screening tests utilizing Clinical Laboratory Improvement Amendments (CLIA) blood tests available in most clinical settings. CLIA blood test results were ascertained in 475 male veterans with and without PTSD and TBI following warzone exposure in Iraq or Afghanistan. Using random forest (RF) methods, four classification models were derived to predict PTSD and TBI status. CLIA features were selected utilizing a stepwise forward variable selection RF procedure. The AUC, accuracy, sensitivity, and specificity were 0.730, 0.706, 0.659, and 0.715, respectively for differentiating PTSD and healthy controls (HC), 0.704, 0.677, 0.671, and 0.681 for TBI vs. HC, 0.739, 0.742, 0.635, and 0.766 for PTSD comorbid with TBI vs HC, and 0.726, 0.723, 0.636, and 0.747 for PTSD vs. TBI. Comorbid alcohol abuse, major depressive disorder, and BMI are not confounders in these RF models. Markers of glucose metabolism and inflammation are among the most significant CLIA features in our models. Routine CLIA blood tests have the potential for discriminating PTSD and TBI cases from healthy controls and from each other. These findings hold promise for the development of accessible and low-cost biomarker tests as screening measures for PTSD and TBI in primary care and specialty settings.

Importance/UNASSIGNED:Although classic psychedelic medications have shown promise in the treatment of alcohol use disorder (AUD), the efficacy of psilocybin remains unknown. Objective/UNASSIGNED:To evaluate whether 2 administrations of high-dose psilocybin improve the percentage of heavy drinking days in patients with AUD undergoing psychotherapy relative to outcomes observed with active placebo medication and psychotherapy. Design, Setting, and Participants/UNASSIGNED:In this double-blind randomized clinical trial, participants were offered 12 weeks of manualized psychotherapy and were randomly assigned to receive psilocybin vs diphenhydramine during 2 day-long medication sessions at weeks 4 and 8. Outcomes were assessed over the 32-week double-blind period following the first dose of study medication. The study was conducted at 2 academic centers in the US. Participants were recruited from the community between March 12, 2014, and March 19, 2020. Adults aged 25 to 65 years with a DSM-IV diagnosis of alcohol dependence and at least 4 heavy drinking days during the 30 days prior to screening were included. Exclusion criteria included major psychiatric and drug use disorders, hallucinogen use, medical conditions that contraindicated the study medications, use of exclusionary medications, and current treatment for AUD. Interventions/UNASSIGNED:Study medications were psilocybin, 25 mg/70 kg, vs diphenhydramine, 50 mg (first session), and psilocybin, 25-40 mg/70 kg, vs diphenhydramine, 50-100 mg (second session). Psychotherapy included motivational enhancement therapy and cognitive behavioral therapy. Main Outcomes and Measures/UNASSIGNED:The primary outcome was percentage of heavy drinking days, assessed using a timeline followback interview, contrasted between groups over the 32-week period following the first administration of study medication using multivariate repeated-measures analysis of variance. Results/UNASSIGNED:A total of 95 participants (mean [SD] age, 46 [12] years; 42 [44.2%] female) were randomized (49 to psilocybin and 46 to diphenhydramine). One participant (1.1%) was American Indian/Alaska Native, 5 (5.3%) were Black, 16 (16.8%) were Hispanic, and 75 (78.9%) were non-Hispanic White. Of the 95 randomized participants, 93 received at least 1 dose of study medication and were included in the primary outcome analysis. Percentage of heavy drinking days during the 32-week double-blind period was 9.7% for the psilocybin group and 23.6% for the diphenhydramine group, a mean difference of 13.9%; (95% CI, 3.0-24.7; F1,86 = 6.43; P = .01). Mean daily alcohol consumption (number of standard drinks per day) was also lower in the psilocybin group. There were no serious adverse events among participants who received psilocybin. Conclusions and Relevance/UNASSIGNED:Psilocybin administered in combination with psychotherapy produced robust decreases in percentage of heavy drinking days over and above those produced by active placebo and psychotherapy. These results provide support for further study of psilocybin-assisted treatment for AUD. Trial Registration/UNASSIGNED:ClinicalTrials.gov Identifier: NCT02061293.

The quality of a cluster analysis of unlabeled units depends on the quality of the between units dissimilarity measures. Data dependent dissimilarity is more objective than data independent geometric measures such as Euclidean distance. As suggested by Breiman, many data driven approaches are based on decision tree ensembles, such as a random forest (RF), that produce a proximity matrix that can easily be transformed into a dissimilarity matrix. A RF can be obtained using labels that distinguish units with real data from units with synthetic data. The resulting dissimilarity matrix is input to a clustering program and units are assigned labels corresponding to cluster membership. We introduce a General Iterative Cluster (GIC) algorithm that improves the proximity matrix and clusters of the base RF. The cluster labels are used to grow a new RF yielding an updated proximity matrix which is entered into the clustering program. The process is repeated until convergence. The same procedure can be used with many base procedures such as the Extremely Randomized Tree ensemble. We evaluate the performance of the GIC algorithm using benchmark and simulated data sets. The properties measured by the Silhouette Score are substantially superior to the base clustering algorithm. The GIC package has been released in R: https://cran.r-project.org/web/packages/GIC/index.html.

OBJECTIVE:To further the precision medicine goal of tailoring medical treatment to individual patient characteristics by providing a method of analysis of the effect of test treatment, T, compared to a reference treatment, R, in participants in a RCT who are likely responders to T. METHODS:Likely responders to T are individuals whose expected response at baseline exceeds a prespecified minimum. A prognostic score, the expected response predicted as a function of baseline covariates, is obtained at trial completion. It is a balancing score that can be used to match likely responders randomized to T with those randomized to R; the result is comparable treatment groups that have a common covariance distribution. Treatments are compared based on observed outcomes in this enriched sample. The approach is illustrated in a RCT comparing two treatments for opioid use disorder. RESULTS:A standard statistical analysis of the opioid use disorder RCT found no treatment difference in the total sample. However, a subset of likely responders to T were identified and in this group, T was statistically superior to R. CONCLUSION/CONCLUSIONS:The causal treatment effect of T relative to R among likely responders may be more important than the effect in the whole target population. The prognostic score function provides quantitative information to support patient specific treatment decisions regarding T furthering the goal of precision medicine.

Active-duty Army personnel can be exposed to traumatic warzone events and are at increased risk for developing post-traumatic stress disorder (PTSD) compared with the general population. PTSD is associated with high individual and societal costs, but identification of predictive markers to determine deployment readiness and risk mitigation strategies is not well understood. This prospective longitudinal naturalistic cohort study-the Fort Campbell Cohort study-examined the value of using a large multidimensional dataset collected from soldiers prior to deployment to Afghanistan for predicting post-deployment PTSD status. The dataset consisted of polygenic, epigenetic, metabolomic, endocrine, inflammatory and routine clinical lab markers, computerized neurocognitive testing, and symptom self-reports. The analysis was computed on active-duty Army personnel (N = 473) of the 101st Airborne at Fort Campbell, Kentucky. Machine-learning models predicted provisional PTSD diagnosis 90-180 days post deployment (random forest: AUC = 0.78, 95% CI = 0.67-0.89, sensitivity = 0.78, specificity = 0.71; SVM: AUC = 0.88, 95% CI = 0.78-0.98, sensitivity = 0.89, specificity = 0.79) and longitudinal PTSD symptom trajectories identified with latent growth mixture modeling (random forest: AUC = 0.85, 95% CI = 0.75-0.96, sensitivity = 0.88, specificity = 0.69; SVM: AUC = 0.87, 95% CI = 0.79-0.96, sensitivity = 0.80, specificity = 0.85). Among the highest-ranked predictive features were pre-deployment sleep quality, anxiety, depression, sustained attention, and cognitive flexibility. Blood-based biomarkers including metabolites, epigenomic, immune, inflammatory, and liver function markers complemented the most important predictors. The clinical prediction of post-deployment symptom trajectories and provisional PTSD diagnosis based on pre-deployment data achieved high discriminatory power. The predictive models may be used to determine deployment readiness and to determine novel pre-deployment interventions to mitigate the risk for deployment-related PTSD.

We sought to find clinical subtypes of posttraumatic stress disorder (PTSD) in veterans 6-10 years post-trauma exposure based on current symptom assessments and to examine whether blood biomarkers could differentiate them. Samples were males deployed to Iraq and Afghanistan studied by the PTSD Systems Biology Consortium: a discovery sample of 74 PTSD cases and 71 healthy controls (HC), and a validation sample of 26 PTSD cases and 36 HC. A machine learning method, random forests (RF), in conjunction with a clustering method, partitioning around medoids, were used to identify subtypes derived from 16 self-report and clinician assessment scales, including the clinician-administered PTSD scale for DSM-IV (CAPS). Two subtypes were identified, designated S1 and S2, differing on mean current CAPS total scores: S2 = 75.6 (sd 14.6) and S1 = 54.3 (sd 6.6). S2 had greater symptom severity scores than both S1 and HC on all scale items. The mean first principal component score derived from clinical summary scales was three times higher in S2 than in S1. Distinct RFs were grown to classify S1 and S2 vs. HCs and vs. each other on multi-omic blood markers feature classes of current medical comorbidities, neurocognitive functioning, demographics, pre-military trauma, and psychiatric history. Among these classes, in each RF intergroup comparison of S1, S2, and HC, multi-omic biomarkers yielded the highest AUC-ROCs (0.819-0.922); other classes added little to further discrimination of the subtypes. Among the top five biomarkers in each of these RFs were methylation, micro RNA, and lactate markers, suggesting their biological role in symptom severity.