A Clinician’s Guide to PTSD Biomarkers and Their Potential Future Use

Precision medicine aims to identify individual-level phenomenological characteristics (phenotype), underlying biological correlates (endophenotype), and biological targets implicated in modifying the course of disease. The conceptual framework for biomarker development in PTSD begins with understanding the onset, progression, trajectories of illness, and process of recovery from PTSD. Biomarkers must be specific and linked to clinical phenotypes. Among DSM-5 disorders, PTSD is distinct as a psychological condition defined by exposure to a traumatic event at a particular point (or points) in time. As one of the few psychiatric diagnoses linked to a causal event, PTSD provides unique opportunities to study the time course of biological changes involved in the onset of psychiatric disease. As an exposure-related disorder, therefore, PTSD is a classic example of a Gene × Environment (G×E) interaction that may inform mechanisms of onset in other disorders.

The basis of psychiatric biomarkers is similar to biomarkers for nonpsychiatric disorders in that there are static markers (e.g., genetic testing for Huntington’s disease or BRCA1/BRCA2 for breast cancer) and ones that are quantitative and may change over time (e.g., HgA1c, thyroid-stimulating hormone). In PTSD, biomarkers can be examined over a time course: before and after trauma exposure and during the initial weeks to months after exposure. Similar to previously proposed typologies (12, 13), PTSD biomarkers may therefore include markers as indicated in Box 2.

Given the number of different types of biomarkers that can be investigated, an understanding of how the markers will influence clinical practice is essential to guide development. However, research evidence is not established according to the above types of biomarkers. We therefore summarize the current state of biomarker research with a focus on clinically relevant trends, keeping in mind the prospect of further differentiation of biomarkers according to their projected clinical use.

The use of existing, commercially available measurements, such as a standard laboratory test (e.g., C-reactive protein, electroencephalography [EEG], or functional magnetic resonance imaging) may become feasible if research evidence supports their reliability and validity as markers for PTSD. In contrast, the development of novel chemical, proteomic, or genetic markers involves a longer process, given the additional validation needed for the introduction of a new test, prior to commercialization and clinical implementation.

As is true for other psychiatric disorders, diagnostic biomarkers for PTSD pose a particular difficulty in that there is no biological gold standard. Unlike a tumor, for which discrete tissue samples are available for comparison, PTSD is not readily localized in one brain region and may involve subtle alterations to physiologically normal brain circuits or connectivity. Animal models of PTSD, none of which model the entire disorder, are arguably homogenous by design, to represent and detect basic mechanisms. They have been used to screen for putative biomarkers, but this method of discovery has limitations, given minimal homology with human biology and the heterogeneity present in clinical populations.

The study of postmortem human brains may allow for more robust signal detection, followed by the development of methods to detect corresponding biomarkers in peripheral blood. Appropriate selection of clinical rating instruments and the analysis of dimensional clinical constructs in conjunction with biological findings may further assist in parsing the heterogeneity of PTSD. Combinations of clinical instruments with currently available and new assays in large datasets may be needed to fully characterize the heterogeneity among individuals with a PTSD diagnosis, and machine-learning methods may improve signal detection (14, 15).

Once screening and discovery have identified candidate biomarkers, analytic validation to enhance assay precision and accuracy, as well as to establish thresholds for detection, will be needed (16). Acceptable thresholds for precision, accuracy, and detection should be determined in accordance with the intended clinical use (16). As noted more generally for screening tests, the sensitivity, specificity, positive predictive value, and negative predictive value across a large, diverse population are highly consequential. With an accuracy level highly dependent on the base rate of the disease, laboratory tests can result in costly follow-on evaluations, which themselves may entail risks, additional costs, and inconvenience to the patient. Defining the context of use, including the target population, is therefore important to the biomarker development process, as well as the ways that a biomarker test may enhance the clinician’s practice (17).

According to a recent glossary from the U.S. Food and Drug Administration and the National Institutes of Health, a biomarker is a “defined characteristic that is measured as an indicator of normal biological processes, pathogenic processes, or responses to an exposure or intervention, including therapeutic interventions. Molecular, histologic, radiographic, or physiological characteristics are types of biomarkers. A biomarker is not an assessment of how an individual feels, functions, or survives” (13). The term indicator suggests that a biomarker does not necessarily index a mechanistic pathway or serve as a target for intervention. A biomarker may be a characteristic that is correlated with disease or produced as the biological result of a disease process and thereby serve as a marker without a demonstrated causal relationship (18).

One notable example of the complexity in the clinical implementation of biomarkers is the medical community’s experience with prostate-specific antigen (PSA). Beginning in the 1980s, PSA was used as a standard screening measure for prostate cancer among older men. In the early 2000s, several large studies to determine the efficacy of this screening process (19) showed that universal screening resulted in substantial adverse outcomes from follow-up biopsies, whereas the number of prostate cancer deaths prevented was less than 1 in 1,000 men screened (20). Although PSA is no longer recommended for general screening among older men, it remains a useful tool in situations such as monitoring tumor recurrence for individuals previously treated for prostate cancer. This example from general medicine illustrates the point that when one is thinking through the clinical use of biomarkers, it is important to carefully consider the context of use and secondary effects of broad-based testing in contrast to tailored use within specified populations or known disease processes.

PTSD biomarkers thus far have been used exclusively in the research setting. As described above, PTSD has a specific event as its starting point, with the opportunity to examine evolution over time and distinguish trajectories of recovery, a process that is optimally studied with longitudinal data with repeated measures. A promising example of longitudinal design is a currently ongoing study funded by the National Institute of Mental Health involving 5,000 trauma-exposed participants recruited in an emergency room setting. This study will use biological markers for phenotyping and conduct surveys and continuous real-time physiological monitoring over eight weeks, with a 52-week total follow-up period (21). In addition, the research use of biomarkers over the course of interventional studies is well recognized, although clinical trials have generally recruited on the basis of a PTSD diagnosis, using biomarkers to define moderators and subgroups of responders. The collection of blood and saliva samples to be stored for future biomarker studies has become a common practice in clinical trials.

These research uses of biomarkers for PTSD set the stage for their future clinical use, although their implementation in clinical practice raises many difficult questions. The potential development of a test that will assist in establishing a PTSD categorical diagnosis has been considered. However, the clinician immediately faces an issue regarding the manner in which an assay can be used to enhance clinical judgment. Does a negative result necessarily lead to the conclusion that the individual does not have PTSD? A biomarker test might assist in screening for possible PTSD in a primary care setting for subsequent referral to specialty consultation. Alternatively, a biomarker assay in the specialty setting might be used to help establish or rule out the diagnosis in clinical presentations that are ambiguous, where the clinician might have concerns if the patient has a low level of awareness of symptoms and their affect on functioning.

The cost-benefit issues for different types of PTSD biomarkers have been discussed in detail elsewhere (12). Guidelines are needed to establish standards so that biomarkers are not used to supplant the clinician’s judgment or decrease the care with which clinical assessment is performed. In the future, there may well be biomarker-positive PTSD cases as well as biomarker-negative cases among individuals who nonetheless require treatment. The potential use of biological markers in legal and administrative contexts has yet to be discussed in detail and resolved.

For instance, the ethical use of biomarkers for decisions regarding recruitment of military service members, fitness for duty determinations for police officers, or examination of malingering in forensic settings requires a clear understanding and consensus regarding the benefits and limitations of particular tests. The level of accuracy needed to rely on biomarkers for decision making in these contexts would likely be very high. Using biomarkers to deny disability benefits or to disqualify individuals from military service would be ethically problematic from the perspectives of autonomy and potential harm as well as fairness to the overall population concerned.

Biological markers identified in earlier investigations include heart rate, electrodermal response (skin conductance), cortisol levels, and glucocorticoid receptor (GR) density in peripheral blood cells (22). In the past decade, biomarker research has focused further on these, along with genetic and epigenetic markers, mitochondrial genes, mRNA, and micro RNA. A large number of chemical, protein, and genetic markers have been examined among individuals with PTSD (see (22–24 and Box 2).

Glucocorticoid regulation in PTSD has been well characterized, and many potential markers pertain to glucocorticoid pathways (25). The hypothalamic-pituitary-adrenal (HPA) axis is a complex system that controls stress response. Although basal serum cortisol levels have demonstrated mixed results as a predictor or marker of PTSD, decreased cortisol reactivity to an experimental stressor was found to predict the development of PTSD (22). PTSD symptom severity has been associated with high pre-exposure GR number in peripheral blood mononuclear cells (25–27) and T-cell sensitivity to glucocorticoids for PTSD participants without depression (28). Data also show that individuals with PTSD had elevation of cortisol (29) and corticotropin-releasing hormone (CRH) concentrations (30) in cerebrospinal fluid when compared with healthy controls. In contrast, CRH concentrations in cerebrospinal fluid seem to decrease after exposure to a trauma-related stimulus among individuals with PTSD (31). Other potential biomarkers include genetic variants of the CRH receptor (32) and lower methylation of a promotor region for the GR gene (33).

A number of biomarkers related to additional pathways involved in regulating the HPA axis have been proposed. FK506 binding protein 51 (FKBP5) is a heat-shock protein that binds to and regulates GR activity (34), with variants that are implicated in a number of psychiatric disorders, including PTSD (35). Among the numerous proteins that bind to the immunosuppressant drug FK506, functional polymorphisms in the gene-encoding FKBP5 have been associated with risk among adults of developing stress-related psychiatric disorders (34, 36), as well as with PTSD symptom severity (37). Carriers of one risk allele had GR supersensitivity and PTSD, which suggests that functional variants in FKBP5 are associated with biologically distinct subtypes of PTSD (38). The FKBP5 promoter region contains a glucocorticoid response element whose demethylation is associated with childhood exposure to trauma and may be linked to stress-dependent changes in gene transcription (39). The demethylation of the FKBP5 gene is associated with long-term dysregulation of the stress response and global immune cell functioning (39). Low FKBP5 mRNA among predeployment soldiers predicted more intense PTSD symptoms (27).

Also related to the HPA axis stress response in PTSD is the pituitary adenylate cyclase-activating polypeptide (PACAP). PACAP and its receptors have been identified in both autonomic stress and HPA pathways (40). Higher blood levels of PACAP have been associated with a PTSD diagnosis, symptom severity, and potentiated startle among female but not male participants (41). A further genetic study revealed that a polymorphism in the PACAP receptor gene, found within an estrogen response element, was associated with PTSD diagnosis among women and that differential methylation of this gene correlated with PTSD symptoms (41). Further supporting the role of stress-response dysregulation, expression of serum and glucocorticoid regulated kinase 1 was down-regulated in the prefrontal cortex (PFC) of postmortem samples from PTSD participants (42).

p11, a calcium-binding protein associated with PTSD as well as depression and regulated by glucocorticoids, has been found to be overexpressed in the postmortem brain of individuals with PTSD (43). In contrast, it is decreased in other psychiatric disorders, such as depression (44), which suggests that p11 may serve as specific biomarker for PTSD. The overexpression in PTSD is thought to be regulated by glucocorticoids associated with stress, and it results in a significant increase of GR activation. Activated GR is then transferred into the nucleus, where it interacts with glucocorticoid response elements in p11. Concomitantly, mRNA expression is decreased in PTSD, whereas it is elevated in major depressive disorder, bipolar disorder, and schizophrenia, which suggests that p11 may serve as a biomarker specific to trauma-related disorders (45).

Brain-derived neurotrophic factor (BDNF) is implicated in dendritic spine density, synaptic plasticity, and memory formation (46, 47) as well as in a number of psychiatric disorders (46, 48). Chronic exposure to stress is associated with elevated BDNF in the basolateral amygdala and decreased levels in the hippocampus (47). BDNF is modulated by mineralocorticoids and glucocorticoids, both implicated in traumatic stress reactions (47).

In a study of traffic accident survivors evaluated at six months, BDNF levels in plasma and the Val66Met genotype were not associated with PTSD, but plasma BDNF was correlated with more lifetime trauma exposures (49). In another study, PTSD diagnosis was associated with significantly lower plasma BDNF (50), although BDNF may be elevated early in the course of illness, decreasing with time (51). A single nucleotide polymorphism, Val66Met, has been implicated in susceptibility to a number of neuropsychiatric disorders, including PTSD (46). The frequency of at least one Met allele was twofold higher among individuals with probable PTSD than among controls, and plasma BDNF was also significantly higher among those with probable PTSD (52). Furthermore, the frequency of the BDNF Val66met polymorphism was significantly higher among participants with PTSD who also had an exaggerated startle reflex. In contrast, the frequency of the Val/Val genotype was higher among participants without PTSD and those with a low startle response, compared with those with PTSD and with a high startle response.

Data also suggest a G×E interaction such that at a lower level of exposure to lifetime stressors, the Met/Met genotype was associated with an elevated risk of hyperarousal, whereas Val allele carriers had a lower rate of hyperarousal. Among individuals who reported fewer lifetime stressors, the Met/Met genotype was more highly positively associated with the PTSD Checklist total score than were Val carriers. This suggests that at a lower level of stressful events, Met/Met genotype individuals had higher levels of PTSD symptoms, and Val carriers had lower levels of PTSD symptoms (52). This result contrasts with other studies of genetic association that showed a lack of relationship between the BDNF Val66Met allele and a PTSD diagnosis (49, 53, 54), possibly as a result of methodological differences. Although with further research BDNF may hold promise as a biomarker for trauma- and stressor-related disorders, BDNF is also altered in phobia and panic disorder (46) and other central nervous system disorders, including Huntington’s disease, Alzheimer’s disease, schizophrenia, depression, and substance abuse (48). BDNF involvement in a wide array of disorders suggests that its role in PTSD potentially involves common pathways or mechanisms.

Epigenetic regulation of genes through DNA methylation is a promising area of research for detection of PTSD-related alterations (55). Studies have predominantly used models in animals and human peripheral blood samples. Consistent with the impetus for a PTSD brain bank, Zannas et al. (56) reinforced that, given the neuroanatomical specificity of epigenetic modifications, human postmortem studies in brain tissue are crucial for improved detection.

In addition to the above, a number of other candidate biomarkers have been explored, encompassing a wide range of neuroendocrine, genetic, and inflammatory markers that has been well summarized elsewhere (22). These include 5-HTTLPR polymorphisms in the promoter region of the serotonin transporter gene (SLC6A4) (57). These studies suggest that among European Americans there is an interactive effect of childhood exposure and SLC6A4 gene alleles on risk for PTSD.

In studies of serum in participants with PTSD and controls among U.S. military service members recently deployed to Afghanistan or Iraq, LINE-1 (long interspersed element, a DNA sequence that makes up 17% of the human genome) was hypermethylated among controls and hypomethylated postdeployment among participants with PTSD when compared with predeployment measurements (58). Alu, another repetitive DNA sequence, was also hypermethylated predeployment among participants who eventually were diagnosed with PTSD, when compared with controls (58). Thus, the methylation status of LINE-1 and Alu DNA may serve as a marker of exposure and susceptibility for service members for whom PTSD might occur, which suggests possible resilience or vulnerability factors. The SLC6A3 (solute carrier family 6 [neurotransmitter transporter, dopamine], member 3) locus has also been associated with increased risk of PTSD (59, 60). Other biomarkers that have been explored include inflammatory biomarkers, such as C-reactive protein (61, 62); variants of the gene-encoding FAAH (63, 64); the endocannabinoids anandamide and 2-AG (65, 66); arginine as a marker of nitric oxide synthesis capacity (67); telomere length (68–70); and the interaction of attachment style and oxytocin gene polymorphisms (71).

Researchers have also worked to identify biomarkers by examining shared genetic characteristics across several disorders, through studies recruiting participants inclusive of PTSD, major depressive disorder, and anxiety disorders (72). Transdiagnostic investigation of G×E relationships may identify common biological underpinnings and allow for differentiation (73). This approach has also been used to examine core constructs within study samples that include nonclinical and subthreshold as well as clinical groups.

A goal of this approach is the identification of intermediate phenotypes or subtypes. Doing so may assist in the detection of significant loci in genomewide association studies (GWAS). GWAS studies for PTSD have proposed several loci, including rs406001 on 7p12 (74); PRTFDC1, a tumor suppressor gene highly expressed in the central nervous system (10); and ANKRD55 (75). In the largest PTSD GWAS thus far, with a sample size of N=20,730, no significant SNPs were identified, and results from prior studies were not replicated (76). The authors posited that much larger sample sizes are needed, along with the inclusion of more participants with no trauma exposure and improved phenotyping to decrease heterogeneity (76).

Researchers have investigated comorbidities and symptom clusters to further inform biological markers for PTSD. Examination of key clinical dimensions within PTSD—for example, specific symptoms, such as sleep disturbance (77)—has shown promise in further parsing the heterogeneity of PTSD populations. The most notable finding emerging from this approach has been the characterization of neurobiological findings related to PTSD with dissociative symptoms (78), which has resulted in the establishment of a dissociative subtype of PTSD in the DSM-5 (1, 79). Further extension of this set of findings resulted in a theoretical framework to distinguish PTSD symptoms associated with an altered state of consciousness from symptoms that are characteristic of a state of normal awareness (e.g., flashbacks vs. intrusive memories) (80, 81).

Clinical and physiological biomarkers, ones that are not blood-based bioassays, involve a broad set of approaches that includes several different types of measurement. Although the research use of psychophysiological measurements, such as skin conductance and heart rate variability (82–88) and EEG (89, 90), in PTSD is longstanding, the use of such measures to characterize clinical subpopulations of PTSD for treatment selection is a promising area of current research. Exaggerated startle response is among the DSM-5 criteria for PTSD, which thus makes this symptom of particular relevance to study that uses psychophysiological methods, such as pupil response (91) or electromyography measurement of response in potentiated startle (92). Impaired inhibition of fear-potentiated startle has been studied among participants with PTSD, with the prospect for use of eye-blink reactivity as a differentiator of subtypes.

Neuroimaging data from functional magnetic resonance imaging and positron emission tomography studies have identified a number of brain regions that are likely involved in the pathophysiology of PTSD. These regions include the ventromedial PFC and anterior cingulate cortex, amygdala, hippocampus, and insula (22, 93). Although research has shown that these areas are affected among individuals with PTSD, the interplay among these regions may be more critical to the understanding and treatment of PTSD than are specific regional markers (94).

Increased amygdala activation in response to trauma-related stimuli supports the role of the amygdala in PTSD (22, 93). Studies of trauma-related stimuli have used script-driven imagery, aversive sounds, and backward-masked faces. Emotional content unrelated to trauma has also been examined in relation to increased amygdala activation among participants with PTSD compared with mentally healthy controls (93). Patients with PTSD also demonstrate increased activation of the amygdala at rest and in response to neutral stimuli, as well as greater activation when presented with fearful faces compared with neutral faces (93). Studies of the hippocampus in PTSD have shown mixed results, generally demonstrating possible decreased activation and decreased hippocampal volume among people with PTSD (22, 93).

The connectivity of regions involved in PTSD may be vital to understanding this complex disorder. The PFC, involved in executive function, serves a role in inhibition of the amygdala, a key locus for affective responses to trauma (22). The hippocampus, in parallel, further regulates the emotion-related activation of the overall system through memory and learning. In the context of a traumatic stimulus, the capacity of the PFC to provide control through inhibition is exceeded by increased activity in the amygdala and insula (93). Current efforts to combine functional neuroimaging with genomic data for enhanced phenotyping (94) and to use repeated measurements with machine learning to predict the time course of symptoms and recovery from PTSD (95) are examples of approaches that show promise for further clarifying neurobiological characteristics of PTSD.

The identification of phenotypic features of PTSD that are highly correlated with validated biomarker findings may result in clinically feasible subtyping and treatment selection. In the future, biomarkers could be used as part of an initial clinical evaluation to determine whether the patient is more likely to respond to medication or psychotherapy or to decide whether the best choice is a particular class of medications or one among several available psychotherapies. As discussed below, a number of studies have used biomarkers for identification of responder subgroups in treatment studies.

In a 2014 systematic review of structural and functional changes related to treatment effects, pharmacotherapy response was associated with improvement in structural abnormalities, whereas psychotherapy response was associated with decreased amygdala activation and increased activation of dorsolateral prefrontal regions (96). This suggests the possibility of detecting other biological predictors and correlates of response for specific treatments (see Box 2). Neuroimaging, however, is unlikely to be a preferred modality for biological measurement in the treatment setting, given feasibility and cost.

Fewer studies to date have examined biological markers of response to drug treatment for PTSD. In a small, open-label study of escitalopram, BDNF levels did not change significantly over the course of treatment. However, mean BDNF levels were correlated with the slope of symptom change—that is, lower mean BDNF was associated with a larger decrease in PTSD symptoms (97). In a secondary analysis of data from a randomized placebo-controlled trial of prazosin for PTSD, higher standing systolic blood pressure at baseline was associated with greater reductions in total Clinician-Administered PTSD Scale score (98), which suggests that blood pressure may be a biomarker for response to prazosin.

Research has more extensively focused on biomarkers of response to psychotherapy for PTSD. In a systematic review, neuroimaging studies generally showed that elevated activity in the dorsal anterior cingulate cortex was associated with treatment response, whereas elevated activity in the amygdala and insula at baseline were associated with lack of response. Successful therapy was associated with decreased activity in the amygdala and increased activation of the dorsal anterior cingulate cortex and hippocampus (99).

In a systematic review of baseline (pretreatment) biomarkers and psychotherapy outcomes in studies meeting Preferred Reporting Items for Systematic Reviews and Meta-Analyses criteria published prior to May 2016, brain regions and circuits implicated in processing trauma-related information and memory were predictive of response (100). In addition, markers related to the HPA axis, glucocorticoid sensitivity, and metabolism were predictive of response to psychotherapy. DNA methylation and genes related to glucocorticoids such as FKBP1 (101) and GR sensitivity, as well as elements in the serotonin system, were predictive of response (100). Increased baseline heart rate was also associated with response to psychotherapy. In a more recent report using psychophysiological measures and salivary cortisol at baseline prior to virtual-reality-based exposure therapy with and without d-cycloserine, increased reactivity to startle and lower cortisol reactivity to virtual reality were predictive of response (102).

Biomarkers—whether readily available ones, such as standing systolic blood pressure (as described above in one study of prazosin), or new tests that require a complex development process, such as a genetic assay—have promise for utility in the context of treatment selection and monitoring of response. Although biomarkers have been used to identify characteristics correlated with treatment response in clinical trials, the next step in furthering this approach involves the prospective use of a biomarker as an inclusion or exclusion criterion to define an enhanced study population with an increased likelihood of benefit. Doing so would serve several purposes to advance our understanding of PTSD therapeutics: Using a biomarker to select participants more likely to be well matched to the treatment might increase effect sizes, thereby allowing for improved signal detection. In addition, such biomarker use might improve treatment matching and early effectiveness when treatment is recommended for a particular individual.

In the context of advances in basic and translational research on biological characteristics of PTSD, it is challenging to extract relevant findings for practical application. The identification of reliable and accurate PTSD biomarkers will be a substantial breakthrough for improving clinical care. Ongoing awareness of advances in subtyping, clinical indicators of prognosis, and treatment matching can assist clinicians by providing context for individualized judgments.

Given that individual treatment needs are determined through a collaborative process between patient and clinician, the ultimate clinical utility of biomarkers as a part of precision medicine will enhance rather than supplant existing decision-making processes. The utility of biomarkers will be limited if the additional information from a biomarker does not exceed the clinician’s interview and any subtyping or treatment selection that can be established on the basis of phenomenology. Biomarkers have utility in research for identifying pathophysiological mechanisms and groups of individuals within the categorical diagnosis of PTSD. Attaining sufficient reliability and validity for use in individual treatment decisions is the hope of the near future.