A total of 215 participants who recently experienced a traumatic injury were recruited from an Emergency Department at an urban level 1 trauma center as part of a longitudinal observational study investigating neurobiological and socio-environmental predictors of PTSD [iSTAR study (8,27–29)]. Individuals were considered eligible if they were between the ages of 18 and 65 years old, were English speaking, had no contraindications for magnetic resonance imaging scanning, and had not sustained a head injury more severe than a mild traumatic brain injury. Participants provided written consent and were financially compensated for their time. All study procedures were approved by the local Institutional Review Board at the Medical College of Wisconsin.

Sample characteristics can be found in Table 1. Individual demographics (gender, age, income, and acute PTSD symptoms) were self-reported at the first study visit. Annual household income was assessed on a semicontinuous scale (coded 1–11 in steps of $10,000; 1 = $0–$10,000/year, 11 = greater than $100,000).

Acute PTSD symptoms were measured using the PTSD Checklist Scale for DSM-5 (PCL-5) (30), which consists of 20 self-report items evaluating the severity of PTSD symptoms (31). A total PTSD symptom severity score was created by summing the scores of each item. This sample exhibited pre-dominantly subthreshold PTSD symptoms (mean = 24.91, SD = 17.43) per the proposed clinical cutoff of 30 (32).

Neighborhood socioeconomic disadvantage was measured using the area deprivation index (ADI) (33). Participants’ home addresses were used to derive ADI rankings from a publicly available dataset through the University of Wisconsin School of Medicine and Public Health https://www.neighborhoodatlas.medicine.wisc.edu/ (downloaded February 2020, geocoding completed 3 months after all data were collected). Census block-group rankings were derived from data collected during the National 2014–2018 American Community Survey (33–36). National ADI rankings are factor-based percentile scores representing 17 variables [see (36) for more information on ranking development] and range from 0 (most advantaged) to 100 (most disadvantaged) (36). The current sample lived in largely disadvantaged neighborhoods (mean ADI = 66, SD = 23).

To model the anticipation period, blood oxygen level–dependent signal prior to image onset was modeled for each condition using AFNI’s duration modulation miniblock basis function, because the duration of the anticipation period was variable (1–8 seconds). To model transient response to the images, blood oxygen level–dependent signal at the onset of image presentation was modeled using a generalized linear model and a 14-second tent function with seven tents in AFNI: TENTzero(0,14,8).

Of the recruited participants, 165 participants had anatomical and task scans, and 99 had complete and usable scan data (n = 22 did not complete all runs of the task, n = 44 exceeded motion thresholds). Of the 99 with usable scan data, 6 were unable to be geocoded and 3 were missing other covariate data, yielding a final sample of n = 90.

Two voxelwise linear mixed effects models (3dLMEs) were run to investigate the relationship of ADI and whole-brain activation during 1) the anticipation period (ticking clock) and 2) presentation of task stimuli. The voxelwise statistical threshold was set at p < .001 and corrected for multiple comparisons across the whole brain at p < .05 using Monte Carlo simulations (k > 14) through AFNI’s 3dClustSim. In both models, valence (negative, neutral) and predictability (predictable, unpredictable) were within-subject effects. To remove any acute potential neurobiological consequences of a psychological posttraumatic response (37–39) and to isolate the effects of ADI, we included acute posttraumatic stress symptoms (PCL-5) as a covariate in the model. Additional covariates included age, gender, and income with ADI as the primary variable of interest. The following equation is an illustration of the full model: Valence×predictability×ADI+gender+income+PCL-5+age+(1|participant)

Two additional sets of analyses were run and are reported in the Supplement. The first supplemental analysis demonstrates the effects of ADI without considering acute posttraumatic stress symptoms (i.e., PCL-5 was not included as a covariate). The second is a sensitivity analysis to determine whether the results were driven by individuals living in the most disadvantaged neighborhoods (8,40). We excluded participants with ADI rankings greater than 90 (n = 17) and reran the 3dLME analyses (n = 73).

Six participants reported invalid or unexpected anxiety ratings (i.e., all the same number or in the exact opposite direction from expected); however, exclusion of these participants did not change the pattern of anxiety ratings across the task and are therefore retained in the reported results. Blocks with negative images elicited greater anxiety than blocks with neutral images (main effect of valence: F_1,356 = 97.93, p < .001). Blocks with unpredictable timing did not elicit greater anxiety than blocks with predictable timing (main effect of predictability: F_1,356 = 1.48, p = .19). There was no interaction of valence and predictability (F_1,356 = 0.06, p = .80).

Results of the whole-brain voxelwise 3dLME examining activation during the anticipation period are presented in Table 2. After accounting for all covariates (i.e., ADI, gender, income, PCL-5, age), there was a main effect of predictability with greater activation in occipital and superior parietal regions for blocks anticipating predictable compared with unpredictable stimuli. There was a main effect of valence such that the right middle temporal gyrus showed greater activation in anticipation of neutral compared with negative stimuli. Finally, there was a significant predictability × valence interaction. In unpredictable blocks, the right posterior cingulate cortex showed greater activation during anticipation of neutral compared with negative images. In neutral blocks, the right posterior cingulate showed greater activation during unpredictable compared with predictable anticipation, and the left middle occipital gyrus showed greater activation to predictable compared with unpredictable anticipation.

Anticipation period results of specific ADI effects of interest, after accounting for all other covariates, are depicted in Figure 2 and Table 2. There were no significant interactions of ADI with predictability or valence during anticipation. There was a significant three-way interaction of ADI × predictability × valence such that, in neutral blocks, greater neighborhood disadvantage was related to significant activation in the right anterior cingulate cortex (ACC) for predictable compared with unpredictable anticipation. For thoroughness, we note several simple effects of task conditions and ADI. In unpredictable neutral blocks, higher ADI rankings were related to less activation of the left hippocampus. In predictable negative blocks, higher ADI rankings were related to less activation of the bilateral occipital cortices, right primary motor cortex, and left cerebellum. Finally, in predictable neutral blocks, higher ADI rankings were related to greater activation of the right hippocampus.

Results of the whole-brain voxelwise 3dLME examining activation in response to stimulus presentation are detailed in Table 3. After accounting for all covariates, there was a significant main effect of valence such that the occipital and somatosensory cortices showed greater activation in response to negative versus neutral images. Further, in each of the unpredictable and predictable blocks, the occipital and somatosensory cortices showed greater activation in response to negative stimulus presentations compared with neutral presentations. However, there was no significant main effect of predictability.

Stimulus presentation results of specific ADI effects of interest, after accounting for all other covariates, are depicted in Figure 3 and Table 3. There were no significant ADI × valence, ADI × predictability, or ADI × valence × predictability interactions. For thoroughness, we note that there were several simple effects of task conditions related to ADI. In response to unpredictable negative images, greater neighborhood disadvantage was related to less activation of the left temporal cortex. In response to unpredictable neutral images, higher ADI rankings were related to less activation of the left hippocampus. In response to predictable negative images, greater neighborhood disadvantage was related to less activation in the anterior cingulate, prefrontal, primary motor, and temporal cortices. Finally, in response to predictable neutral images, higher ADI rankings were related to less activation in the temporal, posterior cingulate, prefrontal, primary motor, and occipital cortices.

After accounting for all covariates, the observed task activation during anticipation and stimulus presentation periods was consistent with the literature on trauma-exposed populations, with greater activation to negative than to neutral stimuli in the occipital and frontal cortices (41–44). Some studies have found that hypervigilance to threat-related stimuli in PTSD can bias attention resources or require additional effort to disengage (42,45–49). However, others have found this same pattern for those who have experienced trauma broadly, regardless of PTSD diagnosis (41–44). Thus, greater activation to negative stimuli in attention-related cortices, irrespective of PTSD severity, is consistent with previous reports of attentional biases toward negative and trauma-related stimuli in trauma survivors, particularly those without PTSD (43,44).

Although common posttrauma outcomes such as PTSD have frequently been shown to lead to hyperactivity in threat processing (47,48), this body of work has largely overlooked factors related to preexisting functioning and adaptations of threat-processing neural circuitry (4). The inclusion of pretrauma context (i.e., neighborhood disadvantage) and PTSD symptom severity as covariates in this study may help explain why we observed less activation in threat-related neural circuitry compared with the existing literature.

After accounting for individual characteristics (i.e., age, gender, income, and baseline PTSD symptoms), neighborhood disadvantage modulated activation during individual task conditions. Specifically, greater neighborhood disadvantage was related to less activation of the hippocampus in anticipation of unpredictable neutral stimuli, whereas anticipation of predictable neutral stimuli was related to greater hippocampal activation. The hippocampus is essential for regulating fear expression and consolidating fear memories (50). In this context, these results suggest that greater neighborhood disadvantage is related to less sensitivity to unpredictable stimuli, whereas those in more advantaged neighborhoods may respond to these stimuli as if they are more fearful. Along these lines, Harnett et al. (4), showed that negative life experiences lead to blunting of emotional function in brain (e.g., hypoactivity of hippocampus and amygdala during threat conditions) and behavior.

Greater neighborhood disadvantage was related to less activation of the ACC during anticipation of unpredictable versus predictable neutral stimuli. We note that this interaction emerged only during anticipation, and not during presentation, of neutral stimuli, suggesting a precise effect of pretrauma context on stimulus anticipation in the ACC. Further analysis of stimulus presentation effects showed that for predictable negative and neutral stimuli, higher ADI rankings were related to less activation of the ACC and posterior cingulate cortex, respectively. The ACC is critical for prediction of stimuli and execution of decision making in uncertainty (47,51). The posterior cingulate cortex is a critical node of the default mode network shown to be vulnerable to stress wherein reduced activity impairs top-down emotion regulation (52). This pattern of cingulate cortex activation is broadly consistent with previous work in trauma-exposed individuals. Less activation of the cingulate cortex in response to trauma-related stimuli is commonly reported in trauma populations with and without PTSD (45,53). Activation of the ACC has been shown to relate to PTSD symptom severity, although the direction of effect is widely debated in the literature and may depend on the comparison group and subregion of the ACC used in analysis (46,49,52).

Finally, in response to predictable neutral stimuli, higher ADI rankings were also associated with less activation of prefrontal, premotor, and occipital cortices. Higher ADI was also related to less activation of occipital cortices during the anticipation of and response to predictable negative stimuli. The direction of our findings is not unprecedented. In a study with veterans, less activation to conditioned unpredictable threat cues compared with predictable threat cues in trauma-exposed control subjects was reported in the amygdala, hippocampus, insula, and bilateral superior and middle temporal gyri, implying hypersensitivity to unpredictability (54). However, Dretsch et al. (54) did not account for prior life trauma, and thus inclusion of pretrauma context in this study may explain why our results showed less activation to unpredictable stimuli only in the occipital cortices and not in the amygdala, hippocampus, or insula. Moreover, Dretsch et al. (54) used a threatening auditory stimulus, whereas this study used visual stimuli and thus the superior and middle occipital gyri were recruited. Still, the majority of work in this realm has not considered pretrauma contexts, such as measures of neighborhood socioeconomic disadvantage, or even individual measures of socioeconomic circumstances (e.g., income).

Together, the results of this study may be understood as an adaptive response to unpredictable and/or negative stimuli, stemming from chronic exposure to socioeconomic-based uncertainties. We use adaptive here without a negative or positive connotation, rather to reflect an ecologically driven short-term change in neural circuitry suited to a given environment. The chronic unpredictability and stress of a disadvantaged context may sensitize emotion regulation and uncertainty processing circuitry, yielding less activation of these regions (4,55) as opposed to potentially harmful increases in activity and vigilance to uncertainty observed in anxiety disorders (56). We hypothesize that the chronic stress associated with greater neighborhood disadvantage may have altered reactivity to predictable stimuli given the endurance needed for maintaining and monitoring predictable and/or negative events in the environment. In other words, living with chronic stress makes living with unpredictability predictable, which impacts vigilance systems. We further suggest that uncertainty neural circuitry is not as engaged in laboratory tasks of unpredictability (51) given how frequent uncertainties occur in the lives of individuals residing in more disadvantaged environments (11,13–15). However, future investigations should specifically probe the relationship between uncertainty neural circuitry and socioeconomic-based uncertainties across different task contexts (e.g., reward, threat, safety).

Several considerations temper the generalizability of our findings and the ability to infer how these alterations in neurobiology may be affecting the participants’ everyday lives. First, most inferences were drawn in response to observed simple effects because only one robust interaction with neighborhood socioeconomic disadvantage emerged from the full model. Given that the sample had just experienced a physical traumatic injury, although not worse than a mild traumatic brain injury, the recent exposure to trauma may impact threat-related processing beyond what can be captured by covarying for acute posttraumatic symptoms in the PCL-5. Unfortunately, there was excessive motion in the imaging data, which led to the exclusion of 44 participants. The nature of the task (i.e., presentation of negative stimuli and mixed design) resulted in a subset of participants (n = 22) who were either unable or uncomfortable with completing all necessary trials (all four runs). The exclusion of these participants was also unfortunate because it greatly limited our ability to examine other factors that may interact with socioeconomic-based uncertainty. For example, negative life experiences have been shown to help explain racial differences in threat processing (4).

Racial identity and neighborhood disadvantage are independent, although not mutually exclusive (57); both constructs underscore means by which chronic environmental stress can impact neurobiology and trauma outcomes. In the United States, members of minoritized racial and ethnic groups are disproportionately exposed to neighborhood disadvantage. Racialized socioeconomic inequities are considered drivers of racial and ethnic health disparities, and alterations to neurobiology may be just one mechanism linking these inequities to health outcomes. Different aspects of identity, such as socioeconomic circumstances, race, ethnicity, and gender identity, form complex connections that shape how an individual responds in different contexts (58,59). We call for future neuroscience research, in trauma populations and otherwise, to consider using an intersectional approach to examine the independent effects and interactions between sociodemographic variables (60). Finally, this study did not evaluate the duration of exposure to current socioeconomic circumstances, including neighborhood disadvantage and income. Childhood socioeconomic position impacts neurodevelopment and influences adulthood socioeconomic position. Therefore, future work should consider evaluating how stability of income and ADI interacts with neural processing of uncertainty.