1. Introduction
Not only are military personnel exposed to stressful events in civilian life, they must also cope with high-level stressors when deployed in overseas theatres. Private and professional stressors are repeatedly combined, leading some personnel to experience physiological [
1], and psychological impairments [
2]. These pathologies can be moderate in terms of frequency or severity [
3,
4,
5,
6], or be more disabling [
2,
7,
8,
9,
10].
Regardless of the medical outcome, which is a function of the individual, the relationship between exposure to stressors and impaired health deserves discussion. The understanding of stress biology has largely evolved over the past decades [
11]. The biological framework allows to broaden the well-known three stages of the general adaptation syndrome model by taking into account the biological cumulative impact of stressor exposure on health outcomes [
11,
12,
13]. In this context, the leading mechanism that is thought to underlie the relation is repeated, stress-induced allostasis [
11]. This concept, initially called heterostasis [
12], refers to the high-cost mode of functioning of an organism under stress, which is very different to the usual, economical state of homeostasis [
13,
14,
15]. Chronic strain and life events increase the allostatic load. The latter reflects the functional and structural cost of stress [
14,
15], and is an indicator of the essential protective and adaptive effects of the physiological mediators that maintain homeostasis, or their cumulative impact on daily life if they are mismanaged or overused [
16,
17]. The wide-ranging impacts of allostasis have resulted in its integration into studies of physiological regulation in response to psychosocial and socioeconomic stressors, notably with respect to how adjustments are made to minimize the latter’s impact [
18]. Several biological mediators of allostatic load have been identified [
19]. Among them, cortisol, oxidative stress, and brain-derived neurotrophic factor (BDNF) are three, key independent factors [
19,
20].
The physiological response to a challenge is shaped by the concomitant stress response and its biological sources. The stress response can be described in terms of both allostatic load and tissue tropism. It is mainly seen in cortisol levels, as cortisol is the main hormone controlled by the hypothalamic–pituitary–adreno–cortical (HPA) axis, and is a biomarker of both the stress response and allostasis [
21,
22]. Nocturnal urinary excretion of cortisol reflects the basal tone of the HPA axis [
23], and provides information on the quality of HPA inhibitory feedback. Conversely, blood cortisol concentration, measured in a challenging environment, is an indicator of stress reactivity [
24].
The accumulation of free radicals is another marker of allostatic load [
21,
25]. Free radicals can be indirectly detected by a large panel of biomarkers, notably 8-iso-prostaglandin F2α (8-iso-PGF2α) [
26], as the latter increases in chronic stress [
27] and depression [
28]. Neurotrophic factor production also plays a role in allostasis by protecting neurons [
20,
29]. In particular, BDNF increases with moderate stress [
30], but decreases with high-level stress [
31]. Therefore, it is possible to define the adaptative physiological response to an environmental challenge based on nocturnal urinary cortisol excretion, morning blood cortisol concentration, 8-iso-PGF2α, and BDNF.
Although allostasis has been applied in a number of biomedical contexts, few studies have attempted to use allostasis mediators to connect biomedical and ecological data. Studies have focused on pathologies such as post-traumatic stress disorder (PTSD), which may occur after exposure to a highly stressful event that induces an intense reaction, and anxiety and depression that may occur after repeated exposure to non-traumatic life events [
32,
33]. However, these studies do not address the consequences of exposure to a stressor. In practice, the majority of people who are exposed to stressors in their private or professional life only present infra-clinical psychological suffering, and no psychopathology. However, biological scars may still be detected in people who have been exposed to stressors, despite their lack of clinical symptoms. Identifying this specific state may be an important way to protect soldiers from further psychopathologies. Therefore, we focus on biological pathways for stress become more primed and prepared for future stress, in turn leading to one’s resting allostasis geared toward higher maladaptive patterns of reactivity [
34]. We hypothesize that the four markers of allostatic load noted above can be used to characterize certain stress-related psychiatric conditions, and identify abnormal patterns in a healthy population.
The main objective of our study is, therefore, to determine the allostatic load in a fit-for-duty cohort of soldiers preparing for deployment. In particular, we consider the groups’ psychometric and psychopathological status (whether they present a psychopathology or not), with two objectives. First, we aim to evaluate the allostatic load of soldiers diagnosed as suffering from PTSD, anxiety, or depression based on nocturnal urinary cortisol, morning blood cortisol, 8-iso-PGF2α excretion, and BDNF concentrations. We hypothesize that among mission-ready soldiers, those with high scores of psychological suffering suffer from a higher allostatic load than those with low scores. Second, we aim to determine the allostatic correlates of the psychological profiles of soldiers who do not suffer from PTSD, anxiety, or depression. Our hypothesis is that there are different biological profiles of allostasis characterizing different psychological profiles.
2. Materials and Methods
2.1. Population
The study was conducted in a population of 405 soldiers in the French army who were scheduled for a six-month deployment in Afghanistan in the spring of 2011. Inclusion criteria were having volunteered to participate in the study, being aged between 18 and 50, and being medically fit for military deployment. There were no exclusion criteria. Recruitment took place during pre-deployment training. The study was approved by both the French Armies’ Health Service Ethics Committee, and the French Health Authority (under number 2010-A01232-37). In compliance with the Helsinki Convention that controls and regulates experiments on humans, informed consent was obtained from all participants.
2.2. Protocol
We adopted a cross-sectional ecologic design. The objectives of our investigation were presented by military health authorities during a briefing that was carried out approximately one month before deployment. Participants were asked to collect their urine between 22:00 and 06:00, and to report the following morning for blood collection and psychological assessment. They were asked to not practice sport, drink coffee, or smoke in the two hours preceding blood collection. Blood was collected between 08:30 and 10:30 to control for circadian variation, and after ten minutes spent relaxing. Participants also completed a set of paper-and-pencil standardized assessments that captured sociodemographic data, and details of psychological and pathological functioning. These assessments took approximately one hour to complete.
2.3. Biological Variables
The volume of urine samples was measured, and 2 mL extracts were collected and stored at −80 °C until analysis. Urinary cortisol (U-CORT) concentrations were measured using enzyme-linked immunosorbent assay kits (IBL International GMBH, Hamburg, Germany). Urinary 8-iso-PGF2α (U-PGF) concentrations were measured using enzyme-linked immunosorbent assay kits (Eurobio, DRG, Heidelberg, Germany). Urinary excretion was calculated according to diuresis and creatinine excretion rates.
Blood samples were clotted and centrifuged, while plasma and serum were sampled into 1.5 mL aliquots that were stored at −80 °C until analysis. Plasma cortisol (B-CORT) concentrations were analyzed using enzyme-linked immunosorbent assay kits (IBL international GMBH; Hamburg, Germany). Serum BDNF (B-BDNF) concentrations were determined at a dilution of 1:10 with a commercial BDNF assay (Promega Corporation, Madison, WI, USA) in 96-well plates (Corning Costar® EIA plate, New York, NY, USA). All tests were run in duplicate and according to the manufacturer’s instructions.
2.4. Psychological Variables
2.4.1. Sociodemographic Evaluation
Sociodemographic variables included age, gender, marital status, tobacco use, experience (measured as time served), previous overseas deployments (if any), and, if so, the number of deployments.
2.4.2. Psychopathological Evaluation
Although all members of the cohort had been declared healthy following a medical examination, they completed the Hospital Anxiety-Depression Scale (HAD), and the Posttraumatic Stress Disorder Check List (PCL) to evaluate their adaptation to the environment. The HAD consists of two subscales that aim to detect anxiety (HAD-A) and depression (HAD-D) in general, non-psychiatric medical outpatients [
35,
36]. A cut-off of ≥11 was chosen in order to prioritize specificity (0.92) over sensitivity (0.56) [
37]. Internal consistency was acceptable (Cronbach alpha between 0.67 and 0.68 for HAD-D and HAD-A, respectively). The PCL was used to detect PTSD based on DSM-IV-TR criteria [
38]. A cut-off of ≥44 was chosen to optimize sensitivity (0.864) and specificity (0.944) [
39]. However, this value may overestimate the prevalence of PTSD [
40]. Internal consistency was good (Cronbach alpha: 0.94).
2.4.3. Psychological Evaluation
Stress reactivity was assessed using four questionnaires. Perceived stress was evaluated using the validated French version [
41,
42,
43] of the Perceived Stress Scale (PSS) [
41]. The PSS is a self-report measure of the degree to which the respondent has perceived stressful situations in his/her life in the past month. Alexithymia was assessed using the validated French version [
44] of the Toronto Alexithymia scale (TAS) [
45,
46], where a score below 44 indicates no alexithymia [
47]. Trait anxiety was measured using the validated French version [
46] of the Spielberger Trait Anxiety Scale (STAI-T) [
48]; scores over 42 are considered to be high. State anxiety was evaluated using the validated French version [
49] of the Spielberger State Anxiety Scale (STAI-S) [
48]; scores over 35 are considered high. For these assessments, internal consistency was good (Cronbach alpha between 0.78–0.89).
Mental health was evaluated using four measures. The Burnout Measure Short version (BMS) evaluates the level of burnout [
50,
51]. Developed for use with all occupational groups, it is well-suited to a military population. No cut-off is defined in the French version [
51]. The Positive and Negative Affect Scale (PANAS) [
52]) is a good index of distress. Scores above 33.3 indicate positive affect (PA), and scores above 17.4 suggest negative affect (NA) [
53,
54]. Finally, participants completed the validated French version [
53] of the 28-item General Health Questionnaire (GHQ28) usually used in healthy populations [
55,
56]; a score over 22 indicates psychological distress [
57]. For these assessments, internal consistency was good (Cronbach alpha between 0.74–0.9).
Questionnaires were excluded from further analysis if more than two items were not completed. If only one item was not completed, its value was considered to be the mean of the other items.
2.5. Statistical Analysis
All statistical analyses were performed using Statistica software (Stastsoft France, Maison Alfort, France, version 7.1). Clustering was carried out using SPSS software (SPSS INC, Chicago, IL, USA, version 24.0).
The relation between the four biological variables (U-CORT, U-PGF, B-BDNF, and B-CORT) was analyzed using a factorial analysis with normalized varimax rotation. Two factors were above the eigenvalue threshold of one, and explained 56.8% of the variance (F1: 29.4%; F2: 27.4%). F1 combined U-CORT (weight = 0.7128) and B-CORT (weight = −0.7461), while F2 combined U-PGF (weight = −0.7020) and B-BDNF (weight = 0.6513).
The population was divided into two groups according to scores recorded for the three psychopathological assessments (the PCL, the HAD-D, and the HAD-A). The aim was to separate low-scoring (LS) soldiers who reported no psychopathological suffering (no scores below a cut-off) from high-scoring subjects (HS) with at least one score above a cut-off. Participants who scored above the cut-off on the HAD-D, the HAD-A, or the PCL were termed HAD-D+, HAD-A+, or PCLs+, respectively.
For the LS group, hierarchical tree clustering was applied, based on the four biological markers. Ward’s method was used for aggregation, and the Euclidean distance for distance calculation, after
z-score normalization [
58]. A four-cluster solution (C1, C2, C3, and C4) was selected as the best compromise between precision and discrimination in the context of four co-evolving variables, and comparisons were carried out between them. The C1 subgroup was considered as the Reference group, based on the normality of all considered variables. A factorial Analysis of Variance (ANOVA) was used for between-group comparisons followed, if necessary, by post hoc Bonferroni tests. Correlations were based on regression methods, and only results where
R2 > 0.10 were considered.
For the HS group, comparisons were carried out for each pathology, and with the LS group. The exception was PCLs+, which was also compared to the PCLs+ + HAD-A+ subgroup.
Results are expressed as mean ± standard error of the mean. Statistical significance was set at p < 0.05. Where the group was small, t < 0.10 was considered as evidence of a trend.