A Blood Test Might One Day Mass-Screen Military Personnel for PTSD

[T]he U.S. government is on the hunt for effective ways to screen large populations for PTSD.

In a process of elimination based on how consistently certain blood-based markers could be linked to PTSD, the researchers sieved out almost all of the candidate factors. They finally ended with a group of 27 chemical signatures that together with heart rate offered the best accuracy for identifying PTSD. Some but not all of the markers had previously been linked to the condition and include measures related to insulin levels and blood clotting.

The test panel detected a person as having a positive PTSD result 85 percent of the time and properly identified that a person did not have the condition 77 percent of the time.

The screening test in real-world use would return a probability value, he says, showing how likely a person is to have PTSD. In cases of high probability, the military could refer the person for appropriate interventions that it has in place.

The plan, Marmar says, is to perform the test in bigger populations of male veterans, then to include female veterans, and then turn to mixed general populations before seeking FDA approval. “There will be attempts to validate these same markers in civilian contexts, such as for disaster victims, sexual assault survivors, or industrial accident survivors,”

Here is a link to the actual journal article. Molecular Psychiatry Article Published: 10 September 2019 Multi-omic biomarker identification and validation for diagnosing warzone-related post-traumatic stress disorder Kelsey R. Dean, Rasha Hammamieh, […]Charles Marmar https://www.nature.com/articles/s41380-019-0496-z?utm_source=feedburner&utm_medium=feed&utm_campaign=Feed%3A+mp%2Frss%2Fcurrent+%28Molecular+Psychiatry+-+Issue%29 Article Open Access Published: 10 September 2019 Multi-omic biomarker identification and validation for diagnosing warzone-related post-traumatic stress disorder Kelsey R. Dean, Rasha Hammamieh, […]Charles Marmar Molecular Psychiatry (2019) | Download Citation Abstract Post-traumatic stress disorder (PTSD) impacts many veterans and active duty soldiers, but diagnosis can be problematic due to biases in self-disclosure of symptoms, stigma within military populations, and limitations identifying those at risk. Prior studies suggest that PTSD may be a systemic illness, affecting not just the brain, but the entire body. Therefore, disease signals likely span multiple biological domains, including genes, proteins, cells, tissues, and organism-level physiological changes. Identification of these signals could aid in diagnostics, treatment decision-making, and risk evaluation. In the search for PTSD diagnostic biomarkers, we ascertained over one million molecular, cellular, physiological, and clinical features from three cohorts of male veterans. In a discovery cohort of 83 warzone-related PTSD cases and 82 warzone-exposed controls, we identified a set of 343 candidate biomarkers. These candidate biomarkers were selected from an integrated approach using (1) data-driven methods, including Support Vector Machine with Recursive Feature Elimination and other standard or published methodologies, and (2) hypothesis-driven approaches, using previous genetic studies for polygenic risk, or other PTSD-related literature. After reassessment of ~30% of these participants, we refined this set of markers from 343 to 28, based on their performance and ability to track changes in phenotype over time. The final diagnostic panel of 28 features was validated in an independent cohort (26 cases, 26 controls) with good performance (AUC = 0.80, 81% accuracy, 85% sensitivity, and 77% specificity). The identification and validation of this diverse diagnostic panel represents a powerful and novel approach to improve accuracy and reduce bias in diagnosing combat-related PTSD. Introduction Materials and methods Results Participant recruitment and multi-omic data generation A set of three cohorts totaling 281 samples from male combat veterans from OEF/OIF conflicts were recruited as part of a larger study designed to identify biomarkers for PTSD diagnosis using a combination of clinical, genetic, endocrine, multi-omic, and imaging information (Fig. 1). Participants were recruited in three cohorts: discovery, recall, and validation (Fig. 2a and Table 1). The discovery cohort (cohort 1) consisted of 83 PTSD and 82 trauma-exposed control participants who met the inclusion and exclusion criteria (described in Materials and Methods and Supplementary Material). All participants completed clinical interviews and blood draws. After assessment of data quality, 77 PTSD and 74 trauma-exposed control samples were available with all completed blood marker assays. This discovery cohort was used to generate an initial pool of candidate biomarkers. Participants from the discovery cohort were invited back for clinical re-evaluation and a blood draw approximately three years after their initial evaluation. This cohort of recalled subjects (recall cohort, cohort 2), included 55 participants from the initial discovery cohort. Some of these participants showed PTSD symptom and status changes based on clinical assessment (Fig. 2b). In addition, some participants no longer met the original inclusion/exclusion criteria for the study; these participants had symptoms intermediate between the PTSD and control groups, in some cases meeting criteria for subthreshold PTSD. The 55 recall participants included 15 PTSD, 11 subthreshold PTSD, and 29 control participants. The third cohort, an independent group of 26 PTSD and 26 control participants, became the validation cohort (cohort 3), used for validating the final set of PTSD biomarkers. Fig. 1 figure1 Overview of PTSD biomarker identification approach—details of cohort recruitment, and biomarker identification, down-selection, and validation Full size image Fig. 2 figure2 Overview of molecular datasets and cohort symptom severity. a Flow diagram for participant recruitment and enrollment. Participant eligibility was determined through a phone pre-screen and a baseline diagnostic clinical interview. Eligible participants completed fasting blood draws for multi-omic molecular assays. Participants in the initial discovery cohort were invited to return for follow-up in the recall cohort. Some participants returned with symptom changes, including “subthreshold” PTSD symptoms (below original study inclusion criteria). b Trajectory of PTSD symptoms in recalled participants. CAPS total for current symptoms at baseline (T0) and follow-up (T1) for each participant are connected. Participants who remained in the PTSD + group at both time points are shown in red. Participants who remained in the PTSD- group are shown blue. Participants with PTSD status changes are shown in gray, including participants who became “subthreshold” PTSD cases. c Distribution of molecular data types at three stages of biomarker identification: full exploratory dataset (All Data), reduced set of 343 potential biomarkers (candidate set) and the final panel of 28 biomarker (final set). Methylation and GWAS data represents 99% of initial data screen due to high-throughput arrays. Other molecular data types are well represented in the second and final stages of biomarker identification and selection Full size image Table 1 Summary of cohort demographics and clinical symptoms Full size table PTSD cohorts and multi-omic datasets To identify a minimally invasive PTSD diagnostic panel, blood-based multi-omics and other analytes were assayed for each individual (and during both visits for recalled participants), including DNA methylation, proteomics, metabolomics, miRNAs, small molecules, endocrine markers, and routine clinical lab panels. Additionally, physiological measures were recorded and nonlinear marker combinations were computed. Using a strategy described in the next sections, a robust and diverse 28-member biomarker panel for diagnosing PTSD was identified from this pool of more than one million markers (Fig. 2c). Three-stage biomarker identification and down-selection from exploratory set of multi-omic data We used a “wisdom of crowds” approach to identify candidate PTSD biomarkers from the large set of measured blood analytes. Utilizing domain area expertize of multiple researchers, as well as multiple algorithms and methodologies, collective intelligence has the potential to identify successful candidate biomarkers from a large dataset, particularly when knowledge is limited. Collective intelligence and “wisdom of crowds” approaches are often used in financial modeling and predictions [24], have been evaluated in medical decision-making [25], and are the motivation for ensemble classification methods, which have been shown to outperform individual classifiers [26]. From a diverse set of data-driven, hypothesis-driven, hybrid, and other approaches (Table S3), we identified a set of candidate diagnostic panels, totaling 343 unique potential biomarkers (Step 2 from Fig. 1 and Table S4). These approaches included COMBINER [27], polygenic risk [28, 29], as well as traditional Support Vector Machine with Recursive Feature Elimination (SVM-RFE), random forest, and other classification algorithms, and feature selection approaches, including p-value, q-value, and fold-change filtering. Details of these algorithms are listed in the Supplementary Material. To filter and refine the pool of candidate biomarkers, we used data from recalled participants (recall cohort, cohort 2). Many of these returning participants experienced symptom changes over the 3.3 ± 0.9 years (mean ± sd) between the initial and follow-up evaluation. CAPS totals for recalled participants at both time points are shown in Fig. 2b. The panel was refined using the recall cohort along with a two-stage down-selection approach to select the final set of PTSD biomarkers (Steps 4–5 from Fig. 1). The two-stage down-selection process is based on the following methodology. In the first stage, poor performing candidate biomarkers were removed one-by-one based on the largest average AUC of the remaining biomarker set (Step 4, Fig. 1). The trajectory of AUC scores in the recall cohort is shown in Supplementary Fig. 1A, showing the average AUC at each step of the one-by-one elimination. The biomarker set with the largest average AUC prior to the final performance decline was selected, resulting in 77 remaining biomarkers. To further reduce the number of features in the panel, we implemented a second stage of down-selection, based on random forest variable importance (Fig. 1, Step 5). Using the recall cohort, the remaining 77 biomarkers were sorted based on random forest variable importance (Supplementary Fig. 1B). We retained biomarkers with importance >30% of the maximum importance score for the final biomarker panel (n = 28). The dynamics and distribution of these 28 biomarkers in the discovery and recall cohorts is shown in Supplementary Figs. 2 and 3. Validation of a robust, multi-omic PTSD biomarker panel After the two-stage feature reduction strategy, the final biomarker set consisted of 28 features, including methylation, metabolomics, miRNA, protein, and other data types. A random forest model trained on the combined cohorts 1 and 2 predicted PTSD status in an independent validation set (cohort 3) with an area under the ROC curve (AUC) of 0.80 (95% CI 0.66–0.93, Fig. 3a). Using the point closest to (0,1) on the ROC curve (shown in Fig. 3a), the model was validated with an accuracy of 81%, sensitivity of 85%, and specificity of 77%. The PTSD participants in the validation cohort had CAPS scores ranging from 47–114. We found that predicted PTSD scores from the random forest model for these cases were correlated with total CAPS (r = 0.59, p = 0.001), indicating the current biomarker model predicts not only disease status, but potentially PTSD symptom severity of cases (Fig. 3b). In addition, predicted PTSD scores were moderately correlated with DSM-IV re-experiencing, avoidance, and hyperarousal symptoms (r = 0.44–0.53, Supplementary Fig. 4), suggesting that the identified molecular markers are not specific to a single symptom cluster, but to overall symptoms. Fig. 3 figure3 Validation of biomarker panels. a ROC curve for identified biomarker panel (28 markers), illustrating good performance in an independent validation dataset (26 cases, 26 controls). Shaded region indicates 95% confidence interval, determined by 2000 bootstrapping iterations. Operating point closest to (0,1) on ROC curve used for calculating sensitivity, specificity, and accuracy. b Predicted probability of PTSD based on trained random forest model using a biomarker panel of 28 features. In PTSD participants, predicted PTSD probability is correlated with PTSD symptom severity, measured by CAPS (r = 0.59, p < 0.01). c Random forest variable importance of the final 28 biomarkers. Variable importance was determined using biomarker model training data (cohorts 1 and 2). The top 10 biomarkers, based on random forest variable importance, contain multiple data types, including methylation markers (cg01208318, cg20578780, and cg15687973), physiological features (heart rate), miRNAs (miR-133a-1-3p, miR-192-5p, and miR-9-1-5p), clinical lab measurements (insulin and mean platelet volume), and metabolites (gammaglutamyltyrosine). d Correlation between PTSD biomarkers. Pearson correlation coefficients were computed in the combined set of all three cohorts. The final set of identified biomarkers show small clusters of moderately correlated features, primarily grouped by molecular data type (proteins, miRNAs, and methylation markers). e Biomarker panel performance evaluation during panel refinement, across molecular data types, and in nonlinear features. The validation AUC improves after biomarker down-selection and model refinement. The final biomarker panel validates with greater AUC over the initial biomarker candidate pool (343 markers, AUC = 0.74), and stage one refined panel (77 markers, AUC = 0.75). The final multi-omic panel also outperforms each individual molecular data type. Performance metrics for nonlinear feature combinations, Global Arginine Bioavailability Ratio (GABR) and lactate/citrate. Both nonlinear combinations outperform their individual components in AUC (0.60 vs. 0.51 and 0.55 vs. 0.52 in GABR and lactate/citrate, respectively). Error bars indicate 95% confidence interval, determined by 2000 bootstrapping iterations. f Validation performance by ethnicity, and in the presence of major depressive disorder (MDD). Validation performance in Hispanic participants was higher than other ethnicities (non-Hispanic White, non-Hispanic Black, non-Hispanic Asian). PTSD cases with comorbid MDD (n = 9) are easily distinguishable from all combat-exposed controls (n = 26), with AUC = 0.92, while PTSD cases without comorbid MDD (n = 17) are only moderately distinguishable from controls (n = 26), with AUC = 0.73 Full size image Overall, the set of identified PTSD biomarkers contains many molecular data types (DNA methylation, miRNAs, proteins, metabolites, and others), with signals primarily including under-expressed proteins and miRNAs, and signatures of both DNA hyper- and hypomethylation. Of the 28 markers comprising the final panel, 16 markers had consistent fold-change directions in all three cohorts (Table 2). Five of the final 28 markers were retained during panel refinement even though the fold-change direction was inconsistent between the discovery and recall cohorts, indicating that these features may contain relevant PTSD signal that is not purely measured by group differences in mean. A post hoc analysis of the biomarker panel performance without these inconsistent features resulted in decreased validation performance (AUC = 0.74 and 0.71 when using only markers with consistent fold-change directions across the discovery and recall cohorts (23 markers), and all three cohorts (16 markers), respectively). Table 2 Overview of biomarker signals in each of the three cohorts Full size table Using random forest variable importance, the top 10 biomarkers from the final 28-marker panel included five of the six molecular data types: DNA methylation, physiological, miRNAs, clinical lab measures, and metabolites (Fig. 3c). These data types contribute primarily uncorrelated signals, with only small clusters of moderate to highly correlated biomarkers from three data types: proteins, miRNAs, and DNA methylation (Fig. 3d). Through the biomarker identification and down-selection process, two intermediate biomarker sets were identified, consisting of 343 and 77 candidate biomarkers. Trained random forest models on these biomarker sets validated with slightly lower AUCs than the final biomarker panel (AUCs of 0.74, 0.75, and 0.80 in the 343, 77, and 28 biomarker panels; Fig. 3e). The consistent validation AUC indicates robust signal in these sets of candidate biomarkers, without loss of signal during down-selection from 343 to 28 features. The final panel of 28 markers consisted of six different data types: routine clinical lab markers, metabolites, DNA methylation marks, miRNAs, proteins, and physiological measurements. The combined panel out-performed all six panels composed of each individual data type (Fig. 3e), demonstrating the power of combining different types of markers in a diverse biomarker panel, capable of capturing the complexities of PTSD. Two biomarker features included in our final panel are computed, nonlinear metrics: Global Arginine Bioavailability Ratio (GABR, defined as arginine/[ornithine + citrulline]) and lactate/citrate. These computed ratios outperform their combined individual components in predictive performance, indicating biologically-driven nonlinear features may enhance low signals (Fig. 3e). In addition, these ratios begin to alleviate single-sample normalization issues that need to be addressed for clinical use of a biomarker panel. Evaluation of clinical and demographic factors The cohorts recruited for this study are diverse in terms of ethnicity, educational background, clinical symptoms, overall health, and comorbid diseases and conditions. The heterogeneity of the participants included in these three cohorts, including race, age, and clinical comorbidities, as well as PTSD severity are shown in Table 1. To evaluate the performance of this biomarker panel in the context of participant demographics and other clinical factors, we computed biomarker performance in stratified subsets of the validation cohort. While biomarker performance was highest in Hispanic participants (AUC = 0.95), we observed no statistically significant differences in AUC across ethnicities (Fig. 3f). Multiple studies have examined the increased prevalence and greater symptom severity of PTSD in Hispanic populations [47, 48], which may correspond to stronger biological signals, leading to the differences in AUC. In the validation cohort, 35% of PTSD cases also met the criteria for major depressive disorder (MDD). Using the identified biomarker panel and model, these PTSD + /MDD + cases could be distinguished from all controls with an AUC of 0.92, while the PTSD + /MDD – could only be distinguished from controls with an AUC of 0.73 (Fig. 3f). Similarly, predicted PTSD scores were more strongly correlated with PTSD symptom severity in PTSD + /MDD + participants than in PTSD + /MDD– participants, with r = 0.64 and r = 0.37, respectively (Supplementary Fig. 5). This decrease in prediction accuracy and correlation with PTSD symptoms in the absence of comorbid MDD indicates a potential overlap of biological signals for MDD and PTSD that should be explored further. Discussion This study presents the identification and validation of a biomarker panel for the diagnosis of combat-related PTSD. The panel consists of 28 features that perform well in identifying PTSD cases from combat-exposed controls in a male, veteran population (81% accuracy). Some of the biomarkers have been linked to PTSD previously, including elevated heart rate [36] and decreased level of coagulation factors [10], and other included markers have been linked to MDD, anxiety, and other comorbid conditions, including platelet volume [43, 44], insulin resistance [41, 49], alterations in the SHANK2 gene [30], and PDE9A expression [31] (Table 2). In particular, the circulating miRNAs selected in the panel reflect the diverse pathology and comorbidities present in PTSD populations, including connections to metabolic diseases and cardiovascular conditions. The miR-133-3p, a member of myomiRs that are highly abundant in muscle, including cardiac muscle, has been implicated in cardiomyocyte differentiation and proliferation [50]. The circulating miR-133-3p level has been linked to various cardiovascular disorders, including myocardial infarction, heart failure, and cardiac fibrosis [51, 52]. The miR-9-5p is enriched in brain [40] and known as a regulator for neurogenesis. It is also involved in heart development and heart hypertrophy [53]. The miR-192 is highly abundant in the liver and circulating miR-192-5p levels have been associated with various liver conditions as well metabolic diseases such as obesity and diabetes [37, 38]. The circulating miR-192 level has also been used as a biomarker for ischemic heart failure [54]. *************************************** Comorbidity had to be a major problem in addition the side effects of current treatments.

Disclaimer: Opinions posted on Free Republic are those of the individual posters and do not necessarily represent the opinion of Free Republic or its management. All materials posted herein are protected by copyright law and the exemption for fair use of copyrighted works.