A randomized controlled design was used in which participants were randomly assigned to an 8-week wait-listed control (CON) or 8-week adaptive horsemanship (AH) condition.

Nine (30% of targeted enrollment of 30) male veterans between the ages of 18 and 75 who had been deployed and experienced combat completed data collection ( Table 1 ). Potential participants with severe traumatic brain injury, an amputation, a diagnosis or experience of bi-polar disorder, schizophrenia, or substance dependence in the last 3 months or a pacemaker were excluded ( Table 2 ).

An additional nine male non-veterans between the ages of 18 and 75 were enrolled to serve as the other half of the dyad in the social motor synchrony task in which pairs of participants swing a pendulum and the synchrony between pendulums is recorded (see Section 2.5.1 for more details). Synchrony can only be measured in dyads, and the standard for this type of task is to use a healthy individual as the other part of the dyad (37, 38). Non-veterans were excluded if they had a chronic or acute mental or physical health issue or a pacemaker ( Table 2 ). Dyads were pair-matched on the basis of age (non-veteran within ± 5 years of the veteran’s age) and highest level of education completed following the methods of Varlet et al. (38).

Potential participants were recruited from the surrounding area through word of mouth, distribution of flyers through the New Jersey Department of Military and Veteran Affairs, and radio advertisements. Those interested in participating (n = 79 veterans, n = 52 non-veterans; Figures 1 , 2 ) contacted the study coordinator (EMR) to learn more and complete a pre-screening questionnaire over the phone to assess eligibility criteria ( Table 1 ). Non-veterans meeting the eligibility criteria were placed on a waiting list until a veteran with whom they could be matched enrolled in the study.

If veterans met the eligibility criteria in the pre-screening questionnaire, they completed the informed consent process with the study coordinator and the Life Events Checklist (LEC-5) and Clinician Administered PTSD Scale (CAPS-5) interview with a trained graduate student from the Rutgers University School of Applied and Professional Psychology (42–45). Students were overseen by a clinical psychologist. If students were unsure how to score an item given the information provided by the veteran, the issue was discussed with the clinical psychologist and a score agreed upon. Veterans meeting clinical cutoffs for PTSD on the past month version of the CAPS-5 (n= 13; Figure 1 ) were enrolled in the study.

Pre-measures were taken the week before veterans started the 8-week adaptive horsemanship (AH) program or CON period. Post-measures were taken the week following completion of the AH program or CON period. Follow-up measures were taken 2 and 6 months after completion of the AH program. At the pre-measure time point, non-veteran participants arrived 20 minutes before veteran participants to complete the informed consent process and provide demographic information. Only veteran participants completed the follow-up measures. Pre-measure procedures consisted of a social motor synchrony test, blood draw, and completion of the PTSD Checklist (PCL-5) and brief symptom inventory (BSI). These procedures were repeated at the post-measure time point, and veteran participants also completed a questionnaire on medication, treatment, and lifestyle changes. At the follow-up time points, only the PCL-5, BSI, and questionnaire on medication, treatment, and lifestyle changes were completed.

Data were also collected during the adaptive horsemanship lessons from the veterans enrolled in the adaptive horsemanship condition (n=6). Blood samples were drawn in weeks 1, 4, and 8. Surface electromyography (sEMG) and responses on the Human–Animal Interaction Scale were collected in weeks 1–8.

Data were inspected for normality, and those violating the assumptions of normality were analyzed using an alternative distribution in the statistical model. The most appropriate models were selected using inspection of the residual plots and comparison of Akaike Information Criterion (AIC) values. Statistically different means in ANOVAs were separated using Tukey’s method. Significance was set at p < 0.05.

Circular variance values from the test of social motor synchrony and hormone concentrations from pre- and post-measure time points were analyzed with a repeated measures mixed model ANOVA with fixed effects of time point, treatment, and their interaction and a random effect of participant (SAS 9.4, Cary, NC, USA). Time point was considered a repeated measure.

Changes in scores and subscores from the PCL-5 and BSI from pre- to post-measure time point were analyzed using an independent sample t-test with equal variance (SAS 9.4, Cary, NC, USA). Scores from the follow-ups (EAA group only) were analyzed with a repeated measures mixed model ANOVA with a fixed effect of time point and a random effect of participant.

ARV and MF of the sEMG signals were analyzed with a mixed model, repeated measures, lognormal distribution ANOVA with fixed effects of week, time point, side, and their interactions and a random effect of participant. Plasma hormone concentrations during the lessons were analyzed using mixed model, repeated measures ANOVAs with fixed effects of week, time point, and their interactions and a random effect of participant. Week and time point were considered repeated measures. Subscores and total scores from the Human–Animal Interaction Scale were analyzed using a mixed model, repeated measures ANOVA with a fixed effect of week and a random effect of participant.

The change in total scores from the PCL-5 and scores from the cluster of cognition and mood alteration symptoms (Cluster D) tended to differ between AH and CON groups with decreases in the AH group and increases in the CON group (p = 0.0989 and 0.0889) ( Table 3 ). The change in global severity scores and somatization and psychoticism subscores from the BSI differed significantly between groups with scores decreasing in the AH group and increasing in the CON group (p ≤ 0.0488) ( Table 3 ). The change in depression, hostility, and paranoid ideation subscores from the BSI tended to differ between groups with scores decreasing in the AH group and increasing in the CON group (p ≤ 0.0721).

Obsessive–compulsive, anxiety, and paranoid ideation subscores from the BSI were significantly lower (p = 0.0414, 0.0271, and 0.0456) in veterans in the AH group at 2 months post-intervention as compared to the pre-time point ( Table 4 ).

There were no significant differences (p ≥ 0.2151) in circular variance across time points or groups ( Figure 3 ).

There were no significant differences (p ≥ 0.1432) in cortisol, epinephrine, norepinephrine, or oxytocin concentrations across pre- and post-measure time points or groups ( Table 5 ).

During AH lessons, there were no significant differences (p ≥ 0.2748) in cortisol, norepinephrine, or oxytocin concentrations across time points within lessons or weeks ( Table 6 ). Epinephrine concentrations trended toward a statistically significant difference (p = 0.0744) in weeks 1 and 4 with concentrations being lower in week 4 ( Table 6 ).

There was a significant main effect of side (p < 0.0001) and side by week interaction (p = 0.0209) on ARV in the masseter ( Table 7 ). ARVs were higher on the left than on the right side (p < 0.0001). ARVs were higher on the right side in week 1 than on the left side in weeks 4, 5, and 8 (p ≤ 0.0421). ARV on the left side in week 1 was higher than on the right side (p = 0.0023). There were no significant differences in MF of the masseter ( Table 8 ).

There were significant main effects of week (p < 0.0001) and side (p = 0.0025) and side by week interaction (p = 0.0071) on ARV in the sternocleidomastoid ( Table 7 ). ARV in week 1 was lower than weeks 3, 4, and 8 (p ≤ 0.0067). ARV in week 5 was lower than weeks 3, 4, and 8 (p ≤ 0.0236). ARV in week 7 was lower than weeks 3 and 8 (p = 0.0130 and 0.007). ARV was lower on the right side than on the left (p = 0.0025). In week 1, ARVs were higher on the left side than on the right (p = 0.0111). ARV on the left side in week 5 was higher in weeks 3, 6, and 8 on the left side and week 8 on the right side (p ≤ 0.0436). ARVs were higher on the left side in week 3 than the right side in weeks 1, 6 and 7 (p ≤ 0.0152). ARV on the right in week 1 was lower than on the left side in weeks 4, 6, and 8 and on the right side in weeks 3, 4, and 8 (p ≤ 0.0034). ARV on the left side in week 8 was higher than on the right side in weeks 6 and 7 (p = 0.0454 and 0.0449). There were no statistically significant differences in MF of the sternocleidomastoid ( Table 8 ).

There was a significant main effect of week (p < 0.0001) on ARV in the cervical trapezius ( Table 7 ). ARVs in week 8 were lower than in weeks 1, 3, 4, and 5 (p ≤ 0.0188). ARV in week 3 were higher than in weeks 2 and 6 (p = 0.0089 and 0.0008). ARVs in week 5 were higher than week 6 (p = 0.0170). There were no statistically significant effects on MF in the cervical trapezius ( Table 8 ).

Human subscores on the Human–Animal Interaction Scale were higher in weeks 7 (p = 0.0151) and 8 (p = 0.0013) as compared to that in week 1. Animal subscores were higher in weeks 3 (p = 0.0362), 4 (p = 0.0006), and 7 (p = 0.0059) as compared to that in week 1. Total scores were higher (p ≤ 0.0437) starting in week 3 and remained higher through week 8 as compared to week 1 scores ( Figure 4 ).

High rates of potential participant attrition occurred in this study with 84% of potential participants being lost or excluded from the study between initial contact and completion of informed consent. While higher than the attrition rate reported by Fisher et al. (12) in their study of equine-assisted psychotherapy for veterans with PTSD (66%), it would seem that a high rate of attrition is to be expected in these types of studies, particularly if attrition rate is calculated from initial contact. In the present study, this high attrition rate contributes to the low number of participants completing data collection, which is a severe limitation and makes this a pilot study. High attrition rates and low therapy session attendance are a common problem plaguing research involving human participants (5, 56). Attrition rates in the current study were further complicated by the need for pair-matched control participants. Researchers planning to work with human participants, particularly this population of veterans with PTSD, need to account for these high attrition rates in their study design and recruitment efforts. While some participants were lost after completing the informed consent process, no participants were lost after starting the AH intervention. This is a promising finding given the high rates of dropout associated with traditional therapies for veterans with PTSD and has been corroborated by other researchers (5, 13).

The high attrition rate observed in the beginning stages of the enrollment process may have been compounded by using the CAPS-5 to determine PTSD diagnosis and eligibility for enrollment in the study. Anecdotally, some potential participants had previously received a PTSD diagnosis but failed to meet the criteria for a past-month diagnosis of PTSD using the CAPS-5 at the time of potential enrollment in the study. This discrepancy could be due to a multitude of factors including a diminishment of symptoms since the previous diagnosis, differences in diagnostic criteria, and underreporting of symptoms during the enrollment process. In the future, researchers may want to consider using the lifetime CAPS-5 instead of the past month CAPS-5 or accepting external proof of past PTSD diagnosis to help combat high attrition. Low enrollment in this study was the primary challenge in conducting the study.

The decrease in depression, hostility, and paranoid ideation as measured on the BSI and the tendency for cognition and mood altering symptoms as measured on the PCL-5, and somatization, and psychoticism as measured on the BSI, and overall scores on both instruments to decrease are in line with outcomes reported in other studies, which also reported decreases in PTSD symptoms following various forms of equine-assisted services (8–13, 21). Variation (SE) in scores for PTSD symptoms in the current study was numerically higher than that reported by Malinowski et al. (8) and closer to that reported in the other studies. This is likely due to the time frame of the studies as the study of Malinowski et al. (8) occurred over 5 days and the other studies occurred over periods of weeks as did this study. With the exception of Malinowski et al. (8), these studies only reported overall scores on the PCL-5 or PTSD Checklist—Military Version (PCL-M), although all reported significant decreases in PCL-5 or PCL-M scores. Malinowski et al. (8) reported significant decreases in overall scores on both the PCL-5 and BSI. Hyperarousal symptoms were the only category of symptoms on the PCL-5, which showed significant reductions, while all symptom subcategories except for phobic anxiety and interpersonal sensitivity on the BSI showed significant decreases in the study of Malinowski et al. (8). The disparity in symptom clusters or subcategories showing change across studies is worth investigating further. Different types of equine-assisted services may be more or less effective in addressing particular clusters of symptoms, as the two studies in question employed different approaches—adaptive horsemanship lessons vs. psychotherapy integrating horses. While the changes seen in the current study are insufficient to guide treatment decisions, they do support continued investigation of equine-assisted services as a treatment for PTSD.

Studies by Arnon et al. (13), Fisher et al. (12), Lanning et al. (22, 26), and Marchand et al. (24) have reported data from follow-up times points ranging from 30 days to 3 months after the intervention. Except for Arnon et al. (13), these studies reported persistence of symptom reduction at the follow-up time point. In the current study, obsessive–compulsive, anxiety, and paranoid ideation subscores on the BSI were lower at the 2-month follow-up than at the pre-intervention time point. The preliminary results obtained in the current study and results from other studies reporting data from follow-up time points lead us to suggest that future studies should include measurements at follow-up time points to more closely track how symptoms change during the period following cessation of the activity or therapy. This is valuable information for providers and clinicians to guide their decisions about intervention duration and cessation.

A circular variance value of 1 represents perfect synchrony, and a value of 0 represents no synchrony. Overall, circular variance was lower than values reported in other studies (37, 38). The reason for this discrepancy is unclear and may have contributed to the lack of differences observed at pre- and post-measurement time points. It should be noted that similarly low values were also obtained from a group of veterans without PTSD (57). While researchers may wish to explore this measure further in future studies, the difficulty of locating pair-matched individuals for the dyadic task limits the feasibility of this measure.

All values for cortisol, epinephrine, and norepinephrine concentrations during all measurement time points fell within or only slightly above normal reference ranges for these hormones (58). Plasma oxytocin concentrations were similar to those reported by authors using the same quantification methods; however, there is debate over the accuracy of these quantification methods (59). Dilution recommendations provided by the manufacturer of the ELISA kit were followed. The lack of change in basal plasma hormone concentrations (cortisol, epinephrine, norepinephrine, and oxytocin) is likely related to the fact that these hormone concentrations were within normal ranges at the beginning of the study and did not differ significantly from the healthy participants and could be confounded by the small sample size (57).

Oxytocin, cortisol, and norepinephrine plasma concentrations remained relatively stable across session timepoints and weeks in the veterans. As this is the first study to have measured these hormones in this population during sessions, comparison among studies is impossible. The social bonding at play during humans’ interactions with dogs that leads to increases in oxytocin concentrations may not be present in human–horse interactions, although further work is needed to confirm this hypothesis (60, 61). The response of the hypothalamic–pituitary–adrenal (HPA) axis also appeared to be unaffected by veterans’ interactions with horses, as no changes in cortisol concentrations were found. Responses of the autonomic nervous system are not as straightforward, as no changes in norepinephrine concentrations were found, but epinephrine tended to decrease between weeks 1 and 4. High concentrations of epinephrine in week 1 and the subsequent decrease lead us to suggest that veterans experienced a shift toward sympathetic activation in the autonomic nervous system during week 1 and that in later weeks, the balance shifted toward parasympathetic activation. Given that anxiety and heightened arousal, especially in unfamiliar situations, is a component of PTSD, it would seem reasonable to expect veterans to experience these types of responses in entering an unfamiliar environment with novel expectations (62, 63). Given the small sample size, these are preliminary results and should only be used in guiding the development of future research studies.

Previous research has failed to find links between salivary and plasma catecholamines, making salivary measurement of these hormones scientifically unsound (64, 65). Thus, if researchers wish to continue pursuing this line of investigation, blood sampling will be needed. The sampling procedures (catheter) used in this study resulted in no adverse events, and no participants voiced concern over the procedures. The samples were quickly drawn with the veteran seated during the procedure. While further spacing between sample time points would be recommended; in this population, the procedure was feasible to implement. The fact that all participants were adults and a small sample was taken at each time point contributed to the feasibility of the blood draws. Additionally, a registered nurse with extensive experience in conducting blood draws for research studies was employed for these procedures and contributed to the success of the blood draws.

Changes in ARV in the sternocleidomastoid and cervical trapezius muscles are likely an artifact of the veterans’ movements during the lessons, as EMG traces for these muscles were visually observed to change with the veteran’s movement and activity during live monitoring of data collection. These same patterns were not visually apparent in the EMG trace of the masseter. Results should still be interpreted with caution as recordings occurred during anisometric, dynamic movements rather than the isometric, static movements usually used for robust interpretation of results (66). ARV values in the current study were lower than those generally reported in the literature (67, 68). It should be borne in mind, however, that most research using EMG measures report results from targeted, specific activities or movements, whereas the current study included data from periods of relatively low activity when low muscle activation would be expected. Measuring sEMG as a potential marker of stress via muscle tension in an applied context seems unlikely to be useful given the results of this study.

This study is the first known report of using the Human–Animal Interaction Scale with horses, as it was developed with other animal species (40). The increases in overall scores starting in week 3 were expected as Wharton et al. (11) also reported increases in client–therapist and client–horse relationships following psychotherapy incorporating horses. The Human–Animal Bond Scale used by Wharton et al. (11) had not been validated, however. The increase in scores in the present study indicates more positive human–horse interactions, as higher scores on the Human–Animal Interaction Scale are indicative of more positive behaviors, fewer negative behaviors, or both (40). The changes in horse subscores and total scores did not appear to be related to measurable behavioral changes in the horses, as no differences in horse behavior were found (41). The behavioral measures used in this study did not account for affiliative behaviors, though, which would be included in the Human–Animal Interaction Scale. Future studies should look at behavior and scores on instruments like the Human–Animal Interaction Scale, as there may also be differences in self-report and researcher reported measures as perception can become a confounding variable. In making decisions about which instrument to use, researchers should carefully consider whether the scale incorporates human behavior, animal behavior, or both. It may be useful to develop a scale specific to horse–human interactions, as behavior does vary across species.

The preliminary results of this pilot study support the need for further research addressing the outcomes of equine-assisted services for veterans with PTSD. The results from this small pilot study contribute to the literature that reports positive outcomes in terms of symptom severity following equine-assisted services. It also demonstrates the feasibility of collecting physiological data, such as hormone concentrations, during lessons without adverse effects for the participants. Recruitment was the greatest challenge in this study, and researchers planning to pursue this line of research should carefully plan to mitigate this challenge.