This is the first study to investigate the relationship between CRTI and predicted 10-year CVD risk. Overall, there was no significant difference in absolute or relative predicted CVD risk among a similar sized cohort of injured versus uninjured combat veterans frequency matched by age, sex, rank and deployment. However, the relative CVD risk, but not absolute QRISK®3 scores, were significantly higher in the amputee versus non-amputee injured and uninjured groups. After adjustment, CRTI and worsening injury severity (higher NISS and previous limb amputation) and the mechanism of blast injury were independently associated with increased QRISK®3 scores and predicted 10-year CVD risk.

There are several important factors that shaped our decision to use the QRISK®3 calculator in this study. The QRISK®3 has been validated for use in UK adults which is relevant to our UK military population [13, 14]. Secondly, QRISK®3 uses dynamically and annually updated health information to reflect changes in population characteristics and to enhance its predictive precision [12, 13]. Thirdly, it includes a broad range of ethnicities known to affect CVD risk and incorporates a far wider age range than the vast majority of other cardiovascular risk calculators [15]. Finally, QRISK®3 includes a greater number of ‘modifiable’ risk factors (eg BMI, blood pressure, smoking, mental health lipids and steroid use) than other available risk calculators [12]. Despite these enhancements a number of important modifiable risk factors, known to influence CVD risk, such as diet and exercise, are not included in QRISK®3. One reason for this might be the fact that accurate reporting of dietary and exercise data is known to be challenging, subjective and highly prone to recall bias lessening their reliability and potential use in CVD risk models [20, 21].

It is unfortunate that we were not able to include post code data in this study. This is perhaps not unsurprising given the population of combat veterans examined. Military servicemen tend to be a highly mobile population where variable postings within the UK and abroad are common. The transient colocation of addresses due to similar military accommodation was expected and borne out by the multiple duplicate post codes identified. Consequently, we could not include social deprivation information (using post codes) in our QRISK®3 data entry. However, we were able to undertake an indirect examination of the influence of socioeconomic status, using military rank. We found that lower rank status (at injury/deployment) was independently associated with increased QRISK®3. Military rank has been used as a proxy measure of socioeconomic [22]. Lower socioeconomic status is a well-established risk factor for future CVD and this concept was supported by our data [23, 24].

We did not observe a significant difference in absolute QRISK®3 scores between the amputees and severely injured on cross-sectional analysis. This may be due to the significantly lower age of these groups and might explain why a higher NISS and amputee status were independently associated with increased QRISK®3 scores after adjustment for confounding factors including sampling age. The independent association between blast injury (versus other injury mechanisms) and QRISK scores is interesting. This may relate to the fact that blast injuries were associated with the most severe injuries. The relationship between injury mechanism and health outcomes is a highly complex process as it well recognized that blast leads to a number of associated injuries including burns. There is ongoing work, with ADVANCE, investigating the relationship between detailed injury mechanisms/types and both physical activity and quality of life.

We found that that the relative CVD risk was > 1.0 for both the uninjured and injured groups in our study; in fact they were on average > 1.6 and hence well above the population expected risk [13]. The relative risk is essentially a means of interpreting our population’s risk versus on the expected UK population predicted CVD risk for persons of similar age, sex and ethnic group without known cardiovascular risk factors [13]. This could be interpreted as suggesting that military personnel are at higher CVD risk than the expected population risk. However, this cannot be confidently concluded from our data and would require a comparative examination of CVD risk among military and matched non-military adults of similar age, sex etc.

It is reassuring to note that only 15 participants (1.3%) of entire cohort has a QRISK®3 score of > 10%. It is generally recommended that a 10-year predicted CVD risk above 10% highlights high-risk individuals who should be specifically targeted for aggressive primary prevention strategies including statin treatment. Scrutiny of the CVD risk factors in our ADVANCE cohort versus the QRISK®3’s own derivation cohort used as part of its validation does, in part, explain their differences. For example the average BMI (27.8 kg/m2) and systolic blood pressures (128.8 ± 11.2 mmHg) were higher in our ADVANCE cohort compared with the UK population expected values [13]. The prevalence of ex-smokers (30.2% vs 15.4%), severe mental illness (6.4% vs 4.8%) and erectile dysfunction (8.8% vs 5.1%) were also higher in our ADVANCE population versus the > 3.5 million men included in the QRISK3 derivation cohort respectively [13]. It is interesting that this was observed despite the fact that the ADVANCE cohort was on average > 8 years younger (34.1 vs 42.6 years) than that of the QRISK®3 derivation cohort [13]. Whilst indirect comparisons are difficult these differences could be explained in part by more robust attainment of risk factors in the ADVANCE cohort where all participants underwent a detailed research visit to collect the QRISK3 variables. Also the specific definitions used for severe mental illness and erectile dysfunction are not identical in ADVANCE and QRISK®3 derivation cohort, with the later relying predominantly on available GP/medical records and available blood results rather than that obtained from a single study visit.

It could be argued given the high relative risk of our ADVANCE cohort that they are at genuinely higher CVD risk than that of the average UK population. This might seem surprising given the typical perception of military servicemen as fit young adult who are required to maintain relatively high standards of basic fitness. Contrary to public perception, it has been previously reported that UK servicemen may be at higher CVD risk than that of the age-matched UK population [25]. The explanation for this is complex and includes selection bias as well as cultural practices specific to military life (eg increased smoking [particularly on deployment] and alcohol consumption) and the greater representation of lower socioeconomic status linked to greater cardiovascular risk [25,26,27]. It is encouraging that several recent policy documents have helped to address this potential inequality [28].

The independent association between worsening injury severity and QRISK®3 scores is interesting. This complements previously published data from ADVANCE in which it was shown that CRTI and worsening injury severity was independently associated with increased arterial augmentation index (a measure of reduced arterial compliance) and metabolic syndrome [4]. Moreover, there is data suggesting that higher NISS and in particular a NISS ≥ 25 is associated with worsening all-cause mortality following traumatic injury [19]. However, this data relates to acute rather than historical traumatic injury and not selective military populations as in our study. Nevertheless we also found that the highest quartile of NISS (> 22) was independently associated with the highest QRISK®3 scores. Whether this relationship between worsening injury severity and a previous traumatic limb amputation will eventually translate into genuine adverse clinical outcomes, remains unknown and this determination as well as the potential mechanisms is the key tenet of the ADVANCE study.

One of the Achilles heals of risk prediction is the concept of prediction itself which is not an exact science and relies on complex mathematical equations to generate a 10-year CVD risk prediction using a variety of known cardiovascular risk factors [29]. It relies on accurate data entry to optimise its risk precision which has been shown to be stronger for population-related analysis as in this study than that for an individual. Hence, we cannot discount the possibility that we have underestimated or even overestimated our population’s true CVD risk. However QRISK®3 has been robustly validated and its precision has been graded as ‘excellent’ with an overall discriminatory c-statistic of > 0.84 which is highest in adults < 65 years (c-statistic ≥ 0.86) as in the ADVANCE cohort. We feel that by using contemporaneous and robustly collected data during research study visits strengthens our findings. In addition we had very little missing data [13, 14].

There are a number of additional limitations that need to be mentioned. Our sample size is relatively small and its power calculation based on ≥ 20 year long term follow up [16]. We have only included men in this study due to the very small number of female injuries during Operation HERRICK. The findings of this study focus solely on the baseline data obtained from the ADVANCE cohort, providing merely a snapshot of the cohorts CVD ‘predicted’ risk at singular point in time. Since the study is of a prospective longitudinal nature, CVD risk profiles will change over time. At present there is no justification for widespread primary CVD prevention strategies that specifically target the injured. However, the minority of participants with a QRISK®3 > 10% should be highlighted for closer primary prevention management [15]. We did not to control for multiplicity of comparisons, as our hypotheses were highly focused and all p values were reported alongside confidence intervals for interpretability. Decreasing the type I error could potentially increase the risk of a type II error. Finally, although our regression analyses were adjusted for age, time from injury/deployment, rank and ethnicity there is the possibility of unmeasured or residual confounding which may have resulted in biased estimates.

In conclusion, CRTI and its worsening severity, depicted by increasing NISS and a history of traumatic amputation, and blast mechanism of injury is independently associated with an increased 10-year predicted CVD risk score using the QRISK®3 scoring system. Long term follow up of this cohort is required to determine whether this increased estimated risk will translate into genuine adverse cardiovascular events or need for targeted primary prevention strategies for the injured.