The Relationship between Military Combat and Cardiovascular Risk: A Systematic Review and Meta-Analysis

Background and Objectives Cardiovascular disease (CVD) is a leading cause of death among military veterans with several reports suggesting a link between combat and related traumatic injury (TI) to an increased CVD risk. The aim of this paper is to conduct a widespread systematic review and meta-analysis of the relationship between military combat ± TI to CVD and its associated risk factors. Methods PubMed, EmbaseProQuest, Cinahl databases and Cochrane Reviews were examined for all published observational studies (any language) reporting on CVD risk and outcomes, following military combat exposure ± TI versus a comparative nonexposed control population. Two investigators independently extracted data. Data quality was rated and rated using the 20-item AXIS Critical Appraisal Tool. The risk of bias (ROB using the ROBANS 6 item tool) and strength of evidence (SOE) were also critically appraised. Results From 4499 citations, 26 studies (14 cross sectional and 12 cohort; 78–100% male) met the inclusion criteria. The follow up period ranged from 1 to 43.6 years with a sample size ranging from 19 to 621901 participants in the combat group. Combat-related TI was associated with a significantly increased risk for CVD (RR 1.80: 95% CI 1.24–2.62; I2 = 59%, p = 0.002) and coronary heart disease (CHD)-related death (risk ratio 1.57: 95% CI 1.35–1.83; I2 = 0%, p = 0.77: p < 0.0001), although the SOE was low. Military combat (without TI) was linked to a marginal, yet significantly lower pooled risk (low SOE) of cardiovascular death in the active combat versus control population (RR 0.90: CI 0.83–0.98; I2 = 47%, p = 0.02). There was insufficient evidence linking combat ± TI to any other cardiovascular outcomes or risk factors. Conclusion There is low SOE to support a link between combat-related TI and both cardiovascular and CHD-related mortality. There is insufficient evidence to support a positive association between military combat ± any other adverse cardiovascular outcomes or risk factors. Data from well conducted prospective cohort studies following combat are needed.


Introduction
In 1979 US Veterans Administration published the results of their review examining the potential causal relationship between traumatic limb amputation and future risk of cardiovascular disease (CVD) [1]. As part of this work a literature review was undertaken to examine the medical literature relating to traumatic amputation and future CVD risk. Among the publications examined were just six studies [2][3][4][5][6][7] that had reported cardiovascular outcomes (including hypertension and cardiovascular death) following traumatic amputation. e results were inconsistent and failed to show a clear relationship.
Owing to the inconsistency of existing published data, coupled with their concern regarding the health implications of a potential link between increased CVD risk, combatrelated amputations and potentially other forms of severe traumatic injury (TI), the Veteran's Administration concluded that more robust data was required. Consequently, the Veteran's Administration and Department of US Defence commissioned a longitudinal study to more robustly investigate the issue. is retrospective cohort study was the first to provide evidence to support a significant link between combatrelated traumatic amputation and a higher risk of future adverse cardiovascular outcomes [8,9]. Unfortunately, this data represented military populations who were injured more than seventy years ago, and the relevance for those injured in current conflicts is open to question. Subsequent to this, only one systematic review and one literature review have emerged. ey were both published approximately 10 years ago, only identified a handful of additional studies and failed to reach a consensus opinion [10,11].
Consequently, and in light of the high tempo and large scale of recent military conflicts, there is a need to re-examine the issue of combat related injury and CVD risk. Recent wars in Iraq and Afghanistan have led to the survival of large numbers of combatants, who have sustained highly complex and severe trauma, which would most likely have proved fatal as little as 20 years ago. Despite this, there has not been a wider examination of the impact of unselected combat on CVD outcome (e.g., cardiovascular death) and its associated risk factors (e.g., hypertension and lipid profiles). e objective of this review was to systematically search and review the literature to determine whether military combat exposure, both with and separately without TI is linked to an increased CVD risk and adverse outcomes.

Search Strategy.
We conducted a systematic review and meta-analysis according to a pre-defined protocol and in accordance to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [12]. e protocol of this review was prospectively registered at PROSPERO. Four electronic databases were used: PubMed, Embase, Cumulative Index to Nursing and Allied Health Literature (CINAHL) and ProQuest. Cochrane Reviews was also searched to identify any previous systematic reviews. A systematic search was undertaken for articles published between the 1 st of January 1980 and 22 nd December 2018, in any language. Two reviewers (NDV and CJB) worked in conjunction with a Medical Librarian to create a search algorithm, which used Medical Subject Headings (MeSH) terms, where available. e search was conducted in adherence to the PICO (Population, Intervention/exposure, Comparison/control and Outcome) tool [13]. e population search terms examined were ("Military", "veterans", "combat", "servicemen", "Iraq", "Afghanistan", "Army", "armed services", "marines" or "infantry"). e Intervention search terms used were ("traumatic", "trauma-related", "amputation", "amputees", "traumatic injury", "wounded", "wounding", "combat", "warfare", or "battlefield").

Study Selection.
Only observational studies that evaluated the impact of combat exposure ± TI on future cardiovascular outcomes were included. Individual case reports, conference abstracts, animal studies, in vivo/in vitro studies and those involving children, were excluded. Studies relating to starvation, cold injury or famine were excluded. Studies that examined selected groups of combatants with traumatic brain injury, spinal cord injury and post-traumatic stress disorder (PTSD) were also excluded.
All selected studies needed to include a population of currently serving, or ex-military and predominantly (>75%) male servicemen (veterans) who had been exposed to combat operations. A Comparator or Control group of nonexposed controls was required with a period of follow up from exposure to outcome of at least one year (see selection algorithm Supplement Table 1).
Two reviewers (CJB and NDV) examined all of the screened records independently to determine potential for inclusion. For each study a preliminary grading of "include, exclude or unclear" was made. Study eligibility was assessed on the basis of the article title, followed by examination of the abstract. A er the preliminary screening process, full text versions of articles deemed "included" and "unclear" were scrutinised further. Eligible studies were identified based on the inclusion criteria. Any disagreements between reviewers were resolved by a detailed discussion in order to come to a consensus. e PRISMA flowchart for the selection of included studies is shown in Figure 1.

Data Collection and Abstraction.
Following selection, the data for each study was extracted using a pre-designed data extraction form, which included author, year of publication, military conflict and population studied, number of participants, type of study, sex, duration of follow up, study outcomes and findings (Table 1).

Quality Assessment.
e AXIS Critical Appraisal tool was used to critically appraise the quality of each of the included studies [14] (Supplement Table 2). e AXIS tool consists of 20 questions relating to study conduct. Studies with a total score of >15 were deemed to be high quality, those of 10-15 of moderate quality, whilst those scoring <10 were deemed poor quality. e Risk of Bias (ROB) of all included studies was examined using the Risk of Bias Assessment Tool for Nonrandomized Studies (RoBANS), which consists of six questions (Supplement Table 3) [15]. Using the total scores for each study the ROB was graded as low (scores of 0), moderate (1)(2) or high (>2).

Study Outcomes.
e study outcomes were cardiovascular death, CHD-related death, CHD and myocardial infarction, arterial hypertension, atrial fibrillation, stroke, heart failure, aortic aneurysms, peripheral vascular disease, diabetes mellitus, metabolic syndrome, carotid intima medial thickness and measures of arterial stiffness (including augmentation index and pulse wave velocity).

Statistical Analysis.
Due to the heterogeneity of the included studies (cross sectional and cohort designs, mode of data presentation, etc.) a pooled analysis was inappropriate for the majority of the reported data. Accordingly, a narrative synthesis was undertaken for most included studies. A metaanalysis was undertaken separately for the binary outcome of CVD and CHD-related death, as these outcomes were only reported from cohort studies. Data was pooled using a random-effects model and the Cochran-Mantel-Haenszel Estimate for the generation of weighted risk ratios and their 95% confidence intervals.
Analysis of continuous data was undertaken using GraphPad Prism® (version 6.07) with results presented as the mean ± standard deviation (SD). Meta-analyses were conducted using RevMan (Review Manager) so ware (Version 5.3). Heterogeneity was evaluated using forest plots and the 2 statistic; 2 values of 25%, 50%, and 75% were considered evidence of low, moderate, and high heterogeneity, respectively [16]. e overall strength of evidence (SOE) for the included studies was assessed using the five domains of consistency, precision, reporting bias, study limitations and directness, as supported by the Cochrane Collaboration Tool [17] . e overall SOE was rated by a single investigator (CJB) with a final rating of high, moderate, low, or insufficient as previously described [18].

Results
e initial search retrieved 4499 potentially relevant studies, from which 345 duplicates were removed immediately. Following the preliminary title and abstract review, 73 full-text articles were screened for eligibility. Forty eight full-text articles failed to fully meet the systemic review inclusion criteria and were excluded. One additional study, which met our selection criteria, was identified from the reference lists ( Figure 1) and included. Hence, 26 studies were included in this review. Agreement between the two reviewers on the selection of fulltext articles was moderate Cohens Cohen's 0.70 (86.3% agreement).

Study Characteristics.
All of the included studies were observational, 14 being cross sectional and 12 being cohort studies (Table 1). e follow up period ranged from 1 to 43.6 years. e sample size of the combat exposed population ranged from 19 to 621901 participants. One study was published in French. Fourteen studies related to US military war veterans (Table 1)  International Journal of Vascular Medicine 8 푝 = 0.77: 푝 < 0.0001) (Figure 3). In the only cohort study of combat versus no combat (without significant TI) Schlenger (moderate ROB) did not observe a significant difference in CHD-related (3.02% vs. 2.33%) deaths among US Vietnam combat versus noncombat veterans [22].

Myocardial Infarction and CHD.
e risk of CHD or myocardial infarction was reported in ten studies (Tables  1 and 2). Four studies (two cohort and two cross sectional) reported outcome data following TI; two reported an increased risk [9,24], whilst two were neutral [19,25]. In the first of the two studies reporting an increased risk, Yekutiel and colleagues observed a higher risk of CHD among 53 lower limb amputees wounded from 1948 to 1973 versus 159 agematched healthy controls [26]. In the second study, Stewart and colleagues reported a significantly higher risk of CHD risk among combatants surviving very serious TI versus the age-matched general population [27]. Furthermore, each fivepoint increase in the Injury Severity Score was linked to a 13% increase in the adjusted risk of CHD (Hazard ratio 1.13; 95% CI, 1.03-1.25; 푝 = 0.01) [27]. ere were six studies (one cohort and five cross sectional) that reported on CHD risk following combat exposure without TI. An increased risk with combat was observed in three studies (two low and one moderate ROB) [28][29][30], with no significant difference noted in another three (one low, one moderate and one high ROB) [31][32][33]. (Tables 1  and 2), but these were predominantly cross sectional studies and possessed a significant ROB. In the only cohort study of combatants following TI (moderate ROB) there was no reported difference in stroke rates between veterans with and without proximal amputation (three versus two strokes, respectively) [19]. Among the population with combat exposure without TI five studies were identified; three of these studies reported an increased risk with combat versus non-combat controls (all cross sectional; one with low ROB, one moderate, one high) [28,29,31]. Two of the five studies reported no difference in risk (one cross sectional and one cohort; one low ROB and one moderate) [32,33].

Aortic Aneurysm.
Two cross sectional studies reported the risk of aortic aneurysms following TI. In one study, Vollmar et al. observed a higher prevalence of infrarenal aortic aneurysms (5.8% vs 1.1%) among veterans (푛 = 329) with above knee amputations, compared to deployed veterans without (푛 = 702) (high ROB) [34]. In the second study, there was no difference (4.4% vs. 4.0%) in ultrasound detected aortic aneurysms among veterans with lower limb amputation versus a control population of similar age (high ROB) [25]. (Tables 1 and 2). ere were ten studies (eight cross sectional and two cohort) relating to TI. e risk of hypertension, was reported to be increased in five studies (one moderate and four high ROB) [27,[35][36][37][38], with another five studies reporting no influence of TI (one moderate and four high ROB) in systolic hypertension [19,25,26,34,39]. ere were nine studies of combatants without TI (three cohort and six Twelve studies comprised of participants with combat-related TI, whilst 14 studies included combat veterans without a significant burden of TI (predominantly noninjured). e age ranges at the time of study enrolment ranged from 18 to 89 years. e study populations were predominantly male (range 78-100%). e majority of participants were, where stated, Caucasian (62.4-100%), with the vast majority (≥74% of stated) being of nonofficer rank at the time of combat (±TI).

Metabolic Syndrome.
Only one cross sectional study reported metabolic syndrome as a specific outcome. Etjahed and colleagues observed a 2-fold higher risk of metabolic syndrome (Defined according to the ATP III Criteria) [45] among 235 veterans with bilateral traumatic lower limb amputation versus controls from the general population (high ROB) [46].

Blood Lipid Levels.
Comparative lipid levels and risk of hyperlipidaemia were reported in 11 studies. Among TI veterans two studies (cross sectional) reported an increased lipid profile compared with controls (high ROB) [37,46]. ere were six (two cohort and four cross sectional) studies that all reported no difference in lipid levels or risk of hyperlipidaemia among combatants with TI versus controls of TI (3 moderate and three high ROB) [19, 24-26, 34, 39]. ere were three studies of uninjured combat veterans (one cohort and two cross sectional); one study [40] reported a higher risk cross sectional), four of which found risk of hypertension to be higher following combat (two low, one moderate and one high ROB) [28,29,40,41]; four studies found no effect of combat (two low, one moderate and one high ROB) [31][32][33]42], whilst one study found combat was associated with a lower risk (moderate ROB) [30].
risk. e SOE supporting of a link between combat-related TI and an increased risk of cardiovascular death and CHD-related death is low. ere is insufficient SOE linking combat-related TI (mainly lower limb amputations) to adverse cardiovascular outcome or risk factors. ere is also insufficient evidence that combat exposure, in the absence of significant traumatic injury, is associated with an increase in adverse cardiovascular outcomes or cardiovascular risk.

Discussion
is is the first systematic review to examine the effects of combat exposure and TI on CVD-related mortality, as well as a wide range of cardiovascular risk factors. ere is low SOE to support an association between combat-related TI and an increased risk of cardiovascular death and CHD-related mortality. ere is insufficient evidence to support an association between combat-related TI and increased cardiovascular risk. Furthermore, there is (moderate ROB) and two studies found no differences in lipids levels of combatants versus controls (one moderate and one high ROB) [31,44].

Other Cardiovascular Risk Factors.
Only one study examined carotid intimal thickness (CIMT), a known surrogate for subclinical atherosclerosis, among noninjured combat and noncombat veterans versus their civilian population [33]. CIMT was found to be higher among combat veterans (802.4 ± 182.2 m) compared with noncombat veterans (757.7 ± 164.1 m) even a er adjustment for potential confounders (including age and race). However, there was no significant difference in carotid plaque burden a er adjusting for confounders. ere were no identified studies that examined the comparative measures of arterial stiffness or atrial fibrillation with traumatic injury or military combat. cardiovascular risk following traumatic amputation. ese included increased insulin resistance, psychological stress, high risk behaviour among exposed subjects and the effects of abnormal blood flow proximal to the amputation. Other risk factors remain largely speculative and further research with a need to examine the other types of combat related TI. ese two previous reviews highlight the need for further, more contemporaneous data from studies of participants in more recent military conflicts. ere is also a need to more robustly examine clinical cardiovascular endpoints, such as cardiovascular death, and recognised cardiovascular risk factors. If traumatic amputation truly leads to an increased risk of CVD, then it would be expected that there would be a relatively higher burden and prevalence of established cardiovascular risk factors. Our current analysis did not find this to be the case. Based on the assimilated data within our meta-analysis, there is insufficient evidence, at present, to support a link between severe combat related TI and the insufficient evidence to support a link between combat exposure without trauma and adverse cardiovascular risk or outcomes.

Strength of Evidence.
A PubMed search identified one systematic review and one literature review relating to cardiovascular risk following traumatic amputation in the last 30 years [10,11]. Robbins et al. [11] undertook a systematic review of combat and non-combat related amputations on the outcomes of CVD (four studies) [9,19,36,47]) and cardiometabolic risk (two studies) [43,48], as well as examining joint and phantom limb pain. Unlike our current review, their injured cohort included both civilian and military participants. e most recent single study included in their systematic review was published 17 years ago and a pooled analysis of objective clinical outcomes (CVD and CHD-related death) was not undertaken. Naschitz & Lenger [10] undertook a literature review that was also published 10 years ago and the most recent study included was published >30 years ago [39]. ey suggested a number of potential aetiological factors that may be implicated in an increased is a major concern, we failed to identify any unpublished studies (on reviewing the grey literature) that would support this concern. One considerable source of potential bias relates to the large heterogeneity between differing military conflicts in terms of obvious differences in the type, intensity and duration of combat exposure (Vietnam vs. Iraq/Afghanistan Wars). Finally, this review included a large number of studies consisting of variable control groups and relating to historical conflicts that occurred more than 40 years ago, which raises some concern about the reliability of diagnoses and outcome reporting. e limitations we have identified within the existing literature highlight the need for prospective cohort studies of combat veterans who were engaged in recent armed conflicts.
ese future studies should be designed in such a way that combat veterans with TI are compared with an age matched control population, without known cardiovascular disease, that were deployed to the same operations and at a similar time are followed up and reviewed to examine their relative cardiovascular risk profiles, as well as their psychological health. Addressing the evidence deficits identified above is the focus of the ongoing ADVANCE study, which is in the final phases of its baseline recruitment [56].
In conclusion, there is insufficient data to either support or refute an association between combat or combat related TI and either CVD or an increased burden of cardiovascular risk factors. ere is a weak SOE in support of a link between severe combat related TI and CVD and CHD-related death.
ere is a need for further data from well conducted prospective cohort studies following recent combat operations.
Data Availability e data used to support the findings of this study are included within the article.

Conflicts of Interest
e authors declare that there are no conflicts of interest. Table 1: supplement study selection from search criteria. Table  2: supplement axis quality appraisal tool for included studies. Table 3: supplement ROBANS risk of bias assessment for included studies. (Supplementary Materials) development of diabetes, metabolic syndrome, hyperlipidaemia, hypertension or increased arterial stiffness. Whilst our pooled analysis did demonstrate a significant association between combat related TI and both CVD and CHD related death, these data are drawn from only two studies, with a moderate degree of bias, for each of these outcomes. Given these facts, coupled with the moderate heterogeneity of the studies and their findings, the strength of evidence to support a link remains low. e decision to additionally examine the published data relating to military combat in the absence of significant trauma on cardiovascular risk and outcomes was important. is was done in order to better understand the additive risk associated with combat related injury. Pooled data identified a marginal, yet significantly lower relative risk of cardiovascular death among the combat versus control groups. is is likely to be explained by the fact that combatants were likely fitter and younger than that of the comparator population of noncombat veterans and civilians. ere have been several recent publications suggesting a potential link between combat, in the absence of TI, and adverse cardiovascular outcomes [30][31][32][33]41] and we were keen to more robustly assimilate the current evidence.

Supplementary Materials
We deliberately excluded previous studies that examined cardiovascular outcomes among selected military populations including those with PTSD, isolated traumatic brain and spinal injuries. is was undertaken to reduce potential bias. PTSD can be triggered by a wide variety of adverse life events including military combat [49]. It has been consistently linked to an increased cardiovascular risk compared that of 'non-exposed' individuals [49][50][51][52][53]. is relationship was further endorsed in a very recent meta-analysis of Iraq and Afghanistan war veterans in which a number of plausible mechanism were explored [54]. Traumatic brain and spinal injury have also been linked to adverse cardiovascular risk [55]. e studies included in our systematic review would have likely contained some individuals with these injuries as well as PTSD. However this would have represented a far lower proportion of cases than that of a selected study of these conditions. For example, whilst the burden of PTSD is influenced by the population studied it generally affects about 20% of combat veterans [49]. By exploring the wider context of combat and traumatic injury (hence multiple injury types) beyond that of lower limb amputation, which tended to be the main focus of previous studies we would be able to appreciate the cumulative effects of these exposures including the potentially positive (e.g., related to improved fitness for deployments etc.,) and negative (e.g., PTSD). is wider population inclusion is critical, given the wide range and complexity of traumatic injuries following recent armed conflicts.
is review has a number of important limitations that need to be acknowledged. ere were a large number of cross sectional studies and the majority of cohort studies were retrospective. e majority of included studies had a significant ROB with variable, and in several cases no, adjustment for important confounders. Many of the studies did not report effect estimates with confidence intervals and there was inconsistent reporting of the participant demographics. Although publication bias and selective outcome reporting