Faculty Bibliography

OBJECTIVE: Prospectively identifying individuals at heightened risk for depression can alleviate the disease burden of distal physical and mental health consequences after cancer onset. Our objective was to identify heterogeneous trajectories of adjustment in cancer patients, using treatment-type as a predictor. METHODS: Participants were followed for 6 years within the Health and Retirement Study (HRS), a prospective population-based cohort study. The sample consisted of 1,294 middle-aged participants who were assessed once before and 3 time points after their report of an initial cancer diagnosis. In addition to self-reported depressive symptoms, subjects indicated receipt of surgical, radiological, or chemical interventions as part of their usual oncological care. RESULTS: Four symptom trajectories were identified with Latent Growth Mixture Modeling: an increasing depression (10.5%), chronic depression (8.0%), depressed-improved (7.8%), and stable-low depression (73.7%). A conditional model using participants with available predictor data (n = 545) showed individuals in the emerging depression class were significantly more likely to have received chemo/medication therapy when compared with the remitting depression, stable-low, and chronic depression classes. Participants in the chronic and depressed-improved classes generally had worse baseline health, and the depressed-improved were also younger in age. CONCLUSION: Patients who exhibited increasing depressive symptoms had a greater probability of receiving chemo/medication therapy than any other adjustment trajectory group, although the majority of chemotherapy patients did not exhibit depressive symptom changes. These data underscore the diversity of ways that patients adjust to cancer, and suggest cancer treatment, baseline health, and age may influence long-term patterns of psychological adjustment. (PsycINFO Database Record

Background Traumatic injuries affect millions of patients each year, and resulting post-traumatic stress disorder (PTSD) significantly contributes to subsequent impairment. Aims To map the distinctive long-term trajectories of PTSD responses over 6 years by using latent growth mixture modelling. Method Randomly selected injury patients (n = 1084) admitted to four hospitals around Australia were assessed in hospital, and at 3, 12, 24 and 72 months. Lifetime psychiatric history and current PTSD severity and funxctioning were assessed. Results Five trajectories of PTSD response were noted across the 6 years: (a) chronic (4%), (b) recovery (6%), (c) worsening/recovery (8%), (d) worsening (10%) and (e) resilient (73%). A poorer trajectory was predicted by female gender, recent life stressors, presence of mild traumatic brain injury and admission to intensive care unit. Conclusions These findings demonstrate the long-term PTSD effects that can occur following traumatic injury. The different trajectories highlight that monitoring a subset of patients over time is probably a more accurate means of identifying PTSD rather than relying on factors that can be assessed during hospital admission.

IMPORTANCE: Psychiatric diagnoses are currently distinguished based on sets of specific symptoms. However, genetic and clinical analyses find similarities across a wide variety of diagnoses, suggesting that a common neurobiological substrate may exist across mental illness. OBJECTIVE: To conduct a meta-analysis of structural neuroimaging studies across multiple psychiatric diagnoses, followed by parallel analyses of 3 large-scale healthy participant data sets to help interpret structural findings in the meta-analysis. DATA SOURCES: PubMed was searched to identify voxel-based morphometry studies through July 2012 comparing psychiatric patients to healthy control individuals for the meta-analysis. The 3 parallel healthy participant data sets included resting-state functional magnetic resonance imaging, a database of activation foci across thousands of neuroimaging experiments, and a data set with structural imaging and cognitive task performance data. DATA EXTRACTION AND SYNTHESIS: Studies were included in the meta-analysis if they reported voxel-based morphometry differences between patients with an Axis I diagnosis and control individuals in stereotactic coordinates across the whole brain, did not present predominantly in childhood, and had at least 10 studies contributing to that diagnosis (or across closely related diagnoses). The meta-analysis was conducted on peak voxel coordinates using an activation likelihood estimation approach. MAIN OUTCOMES AND MEASURES: We tested for areas of common gray matter volume increase or decrease across Axis I diagnoses, as well as areas differing between diagnoses. Follow-up analyses on other healthy participant data sets tested connectivity related to regions arising from the meta-analysis and the relationship of gray matter volume to cognition. RESULTS: Based on the voxel-based morphometry meta-analysis of 193 studies comprising 15 892 individuals across 6 diverse diagnostic groups (schizophrenia, bipolar disorder, depression, addiction, obsessive-compulsive disorder, and anxiety), we found that gray matter loss converged across diagnoses in 3 regions: the dorsal anterior cingulate, right insula, and left insula. By contrast, there were few diagnosis-specific effects, distinguishing only schizophrenia and depression from other diagnoses. In the parallel follow-up analyses of the 3 independent healthy participant data sets, we found that the common gray matter loss regions formed a tightly interconnected network during tasks and at resting and that lower gray matter in this network was associated with poor executive functioning. CONCLUSIONS AND REVELANCE: We identified a concordance across psychiatric diagnoses in terms of integrity of an anterior insula/dorsal anterior cingulate-based network, which may relate to executive function deficits observed across diagnoses. This concordance provides an organizing model that emphasizes the importance of shared neural substrates across psychopathology, despite likely diverse etiologies, which is currently not an explicit component of psychiatric nosology.

BACKGROUND: Predicting Posttraumatic Stress Disorder (PTSD) is a pre-requisite for targeted prevention. Current research has identified group-level risk-indicators, many of which (e.g., head trauma, receiving opiates) concern but a subset of survivors. Identifying interchangeable sets of risk indicators may increase the efficiency of early risk assessment. The study goal is to use supervised machine learning (ML) to uncover interchangeable, maximally predictive combinations of early risk indicators. METHODS: Data variables (features) reflecting event characteristics, emergency department (ED) records and early symptoms were collected in 957 trauma survivors within ten days of ED admission, and used to predict PTSD symptom trajectories during the following fifteen months. A Target Information Equivalence Algorithm (TIE*) identified all minimal sets of features (Markov Boundaries; MBs) that maximized the prediction of a non-remitting PTSD symptom trajectory when integrated in a support vector machine (SVM). The predictive accuracy of each set of predictors was evaluated in a repeated 10-fold cross-validation and expressed as average area under the Receiver Operating Characteristics curve (AUC) for all validation trials. RESULTS: The average number of MBs per cross validation was 800. MBs' mean AUC was 0.75 (95% range: 0.67-0.80). The average number of features per MB was 18 (range: 12-32) with 13 features present in over 75% of the sets. CONCLUSIONS: Our findings support the hypothesized existence of multiple and interchangeable sets of risk indicators that equally and exhaustively predict non-remitting PTSD. ML's ability to increase prediction versatility is a promising step towards developing algorithmic, knowledge-based, personalized prediction of post-traumatic psychopathology.

BACKGROUND: Scientific research into mental health outcomes following trauma is undergoing a revolution as scientists refocus their efforts to identify underlying dimensions of health and psychopathology. This effort is in stark contrast to the previous focus which was to characterize individuals based on Diagnostic and Statistical Manual of Mental Disorders (DSM) diagnostic status (Insel et al., 2010). A significant unresolved issue underlying this shift is how to characterize clinically relevant populations without reliance on the categorical definitions provided by the DSM. Classifying individuals based on their pattern of stress adaptation over time holds significant promise for capturing inherent inter-individual heterogeneity as responses including chronicity, recovery, delayed onset, and resilience can only be determined longitudinally (Galatzer-Levy & Bryant, 2013) and then characterizing these patterns for future research (Depaoli, Van de Schoot, Van Loey, & Sijbrandij, 2015). Such an approach allows for the identification of phenominologically similar patterns of response to diverse extreme environmental stressors (Bonanno, Kennedy, Galatzer-Levy, Lude, & Elfstom, 2012; Galatzer-Levy & Bonanno, 2012; Galatzer-Levy, Brown, et al., 2013; Galatzer-Levy, Burton, & Bonanno, 2012) including translational animal models of stress adaptation (Galatzer-Levy, Bonanno, Bush, & LeDoux, 2013; Galatzer-Levy, Moscarello, et al., 2014). The empirical identification of heterogeneous stress response patterns can increase the identification of mechanisms (Galatzer-Levy, Steenkamp, et al., 2014), consequences (Galatzer-Levy & Bonanno, 2014), treatment effects (Galatzer-Levy, Ankri, et al., 2013), and prediction (Galatzer-Levy, Karstoft, Statnikov, & Shalev, 2014) of individual differences in response to trauma. METHOD: METHODological and theoretical considerations for the application of Latent Growth Mixture Modeling (LGMM) and allied methods such as Latent Class Growth Analysis (LCGA) for the identification of heterogeneous populations defined by their pattern of change over time will be presented (Van De Schoot, 2015). Common pitfalls including non-identification, over identification, and issues related to model specification will be discussed as well as the benefits of applying such methods along with the theoretical grounding of such approaches. CONCLUSIONS: LGMM and allied methods have significant potential for improving the science of stress pathology as well as our understanding of healthy adaptation (resilience).

Kalisch and colleagues identify several routes to a better understanding of mechanisms underlying resilience and highlight the need to integrate findings from neuroscience and animal learning. We argue that appreciating methodological complexity and integrating neurobiological perspectives will advance the science of resilience and ultimately help improve the lives of those exposed to stress and adversity.

There is broad interest in predicting the clinical course of mental disorders from early, multimodal clinical and biological information. Current computational models, however, constitute a significant barrier to realizing this goal. The early identification of trauma survivors at risk of post-traumatic stress disorder (PTSD) is plausible given the disorder's salient onset and the abundance of putative biological and clinical risk indicators. This work evaluates the ability of Machine Learning (ML) forecasting approaches to identify and integrate a panel of unique predictive characteristics and determine their accuracy in forecasting non-remitting PTSD from information collected within10 days of a traumatic event. Data on event characteristics, emergency department observations, and early symptoms were collected in 957 trauma survivors, followed for fifteen months. An ML feature selection algorithm identified a set of predictors that rendered all others redundant. Support Vector Machines (SVMs) as well as other ML classification algorithms were used to evaluate the forecasting accuracy of i) ML selected features, ii) all available features without selection, and iii) Acute Stress Disorder (ASD) symptoms alone. SVM also compared the prediction of a) PTSD diagnostic status at 15 months to b) posterior probability of membership in an empirically derived non-remitting PTSD symptom trajectory. Results are expressed as mean Area Under Receiver Operating Characteristics Curve (AUC). The feature selection algorithm identified 16 predictors, present in >/=95% cross-validation trials. The accuracy of predicting non-remitting PTSD from that set (AUC = .77) did not differ from predicting from all available information (AUC = .78). Predicting from ASD symptoms was not better then chance (AUC = .60). The prediction of PTSD status was less accurate than that of membership in a non-remitting trajectory (AUC = .71). ML methods may fill a critical gap in forecasting PTSD. The ability to identify and integrate unique risk indicators makes this a promising approach for developing algorithms that infer probabilistic risk of chronic posttraumatic stress psychopathology based on complex sources of biological, psychological, and social information.

The course of depression in relation to myocardial infarction (MI), commonly known as heart attack, and the consequences for mortality are not well characterized. Further, optimism may predict both the effects of MI on depression as well as mortality secondary to MI. In the current study, we utilized a large population-based prospective sample of older adults (N=2,147) to identify heterogeneous trajectories of depression from 6 years prior to their first-reported MI to 4 years after. Findings indicated that individuals were at significantly increased risk for mortality when depression emerged after their first-reported MI, compared with resilient individuals who had no significant post-MI elevation in depression symptomatology. Individuals with chronic depression and those demonstrating pre-event depression followed by recovery after MI were not at increased risk. Further, optimism, measured before MI, prospectively differentiated all depressed individuals from participants who were resilient.

Heterogeneity in glucocorticoid response to experimental stress conditions has shown to differentiate individuals with healthy from maladaptive real-life stress responses in a number of distinct domains. However, it is not known if this heterogeneity influences the risk for developing stress related disorders or if it is a biological consequence of the stress response itself. Determining if glucocorticoid response to stress induction prospectively predicts psychological vulnerability to significant real life stressors can adjudicate this issue. To test this relationship, salivary cortisol as well as catecholamine responses to a laboratory stressor during academy training were examined as predictors of empirically identified distress trajectories through the subsequent 4 years of active duty among urban police officers routinely exposed to potentially traumatic events and routine life stressors (N = 234). During training, officers were exposed to a video vignette of police officers exposed to real-life trauma. Changes in salivary 3-methoxy-4-hydroxyphenylglycol (MHPG) and cortisol in response to this video challenge were examined as predictors of trajectory membership while controlling for age, gender, and baseline neuroendocrine levels. Officers who followed trajectories of resilience and recovery over 4 years mounted significant increases in cortisol in response to the experimental stressor, while those following a trajectory of chronic increasing distress had no significant cortisol change in response to the challenge. MHPG responses were not associated with distress trajectories. Cortisol response prospectively differentiated trajectories of distress response suggesting that a blunted cortisol response to a laboratory stressor is a risk factor for later vulnerability to distress following significant life stressors.