Global Cardiovascular Risk and Associated Factors in 2792 French Military and Civilian Aircrew

METHODS

Procedure

This was a single-center, observational, and cross-sectional study based on retrospective analysis of data from subjects included during 6 mo of Toulon AeMC activity. The data were gathered in an anonymous computer database created with Microsoft Excel 2010^© and benefitted from advanced controls in order to limit entry errors. The measured variables corresponded to sociodemographic, biometric, and biological data and some CVRF. They included gender; age; aviation type (civilian or military); aircrew roles; license profile (applicant or trained aircrew); weight; height; body mass index (BMI); waist circumference (WC); systolic blood pressure (SBP); diastolic blood pressure (DBP); smoking; taking lipid-lowering, antihypertensive, or antidiabetic therapy; a medical history of diabetes or obstructive sleep apnea syndrome (OSAS); cardiovascular prevention status (primary or secondary); total cholesterolemia (TC); low density lipoprotein (LDL-c), high density (HDL-c) lipoprotein, and triglycerides (TG); and fasting blood glucose (FBG).

Age was defined as a risk factor from 50 yr old for men and 60 yr old for women. Automated blood pressure measurement was verified by experienced AeMC nurses. In the event of values greater than 140/90 mmHg, a check after 20 min of lying down was systematically carried out. HBP was defined as SBP ≥ 140 mmHg and/or DBP ≥ 90 mmHg or taking antihypertensive therapy. Weight and height were systematically measured and BMI calculated (weight/height² in kg · m⁻²). Obesity was defined as a BMI ≥ 30 kg · m⁻². As tobacco consumption was systematically assessed by a regulatory questionnaire, people could be classified as active smokers (or weaned for less than 3 yr) and non- or ex-smokers. Dyslipidemia was defined on the basis of the latest recommendations from the European Society of Cardiology (ESC 2019).¹⁵ It was thus defined as LDL-c ≥ 1.16 g · L⁻¹ for a low Systematic Coronary Risk Evaluation (SCORE) risk, LDL-c ≥ 1.0 g · L⁻¹ in the event of a moderate SCORE risk, LDL -c ≥ 0.7 g · L⁻¹ for a high SCORE risk, LDL-c ≥ 0.55 g · L⁻¹ for a very high SCORE risk, or, finally, in case of lipid-lowering treatment with statin (as monotherapy or in combination). Prediabetes was defined by World Health Organization (WHO) criteria in case of FBG ≥ 1.10 g · L⁻¹. International Diabetes Federation criteria (IDF 2005) were used to define metabolic syndrome in the presence of a WC ≥ 94 cm (men) or 80 cm (women), associated with at least two factors among: blood pressure ≥ 130/85 mmHg; HDL-c < 0.4 g · L⁻¹ (men) or < 0.5 g · L⁻¹ (women); triglycerides ≥ 1.5 g · L⁻¹; or FBG ≥ 1.0 g · L⁻¹. Finally, people with an established diagnosis of diabetes, OSAS, or being treated with secondary prevention measures after a CVD event were identified by a systematic and documented analysis of their medical file. From these collected data, supplementary variables were derived in order to assess the individual global cardiovascular (CV) risk according to three parameters: the total number of CVRF, the estimated 10-yr CVD risk, and the risk levels categorization using different selected models. The SCORE model offered the advantage of having been developed from European cohorts and of being applied to countries with low incidence rates of CVD, including France. On the other hand, it only estimates a 10-yr cardiovascular mortality risk based on five CVRF: gender, age, SBP, TC, and smoking.³ The Framingham model, as updated by D’Agostino et al. in 2008, is one of the most used over the world. Despite a tendency to overestimation noted in the absence of recalibration for European countries,¹⁶ it makes it possible to assess a global 10-yr risk of cardiovascular morbidity and mortality. This model was, therefore, particularly suited to aviation medicine and the challenge of assessing a risk of incapacitating in-flight events. Retained CVRF are gender, age, SBP, taking antihypertensive medication, HDL-c levels, smoking, and diabetes. Finally, a third older model based on the summation of CVRF was inspired by an approach proposed in 2005 by the French Haute Autorité de Santé in recommendations (now updated) about the management of essential hypertension in France. In comparison with previous prediction models using a mathematical equation of combined factors, this type of approach exposes the pitfall of a global overestimation without capacity to provide a quantified, precise, and reproducible estimated risk. Nevertheless, this third strategy offered the possibility of taking into account both classic CVRF (age, dyslipidemia, hypertension, smoking, diabetes) and other supplementary factors such as obesity, metabolic syndrome, prediabetes, or even OSAS in the stratification of an absolute CV risk. The 10-yr CVD individual probabilities have been calculated using online resources (http://www.cardiorisk.fr). Risk levels were defined as low (SCORE < 1%, Framingham < 10%), moderate (1% ≤ SCORE < 5%, 10% ≤ Framingham < 20%), or high (SCORE ≥ 5%, Framingham ≥ 20%). The very high risk level was reserved in cases of previously documented CVD. For the summation model, a risk matrix was used, classifying diabetic people straight away at a high risk level (Fig. 1). Whatever the model used, primary outcomes were defined by a risk level from moderate to very high and by the presence of at least one CVRF in applicants. Secondary outcomes were represented by all collected CVRF: age, dyslipidemia, HBP, smoking, obesity, (pre)diabetes, metabolic syndrome, and OSAS.

View Figure

• Download as Powerpoint
• Download Figure

RESULTS

Of the 2822 aircrew involved in a licensing medical visit during the study period, 2792 were included. The population was predominantly male (86.2%) and military (52.5%), and the mean age was 38.5 yr (SD 12.4). Military airmen were younger than civilians (32.8 yr vs. 44.9 yr, P < 0.001). Almost two-thirds of aircrew were pilots [61.2% (59.4–63.1)]. Almost one in six people (16.0%) applied for a first professional license. Applicants were younger [24.4 yr (23.8–24.9)] than trained crew [41.2 yr (40.7–41.7), P < 0.001] and, more than 8 times out of 10, were applying for a military career [81.2% (77.3–84.7)].

Only 22 subjects had a past history of CVD (coronary or cerebrovascular disease), i.e., 0.8% (0.5–1.2) of the population. They were all men. Their mean age was 51.6 yr (48.5–54.7), with no significant difference from the mean age of people at a moderate risk level defined by SCORE [54.7 yr (54.3–55.1), P = 0.17] or Framingham [53.8 yr (53.2–54.4), P = 0.33]. According to the summation model, this mean age was not significantly different between subjects at a moderate risk level [48.0 yr (47.1–48.9), P = 0.20] or at a high risk level [52.8 yr (51.7–53.9), P = 0.60].

This secondary prevention group comprised 13 military and 9 civilian subjects, including 11 pilots, 8 rear crew, and 3 ATCO. Biometrically, they were on average slightly overweight [mean BMI = 26.1 kg · m⁻² (24.7–27.4)] with a waist circumference of 93.7 cm (89.8–97.6), reflecting excess central adiposity. Their mean blood pressure was 129.4 mmHg for SBP (123.8–134.9) and 76.9 mmHg for DBP (72.4–81.4), while 54.5% (32.2–75.6) were taking antihypertensive therapy. Biologically, their mean lipid levels were 1.66 g · L⁻¹ (1.52–1.81) for TC, 0.90 g · L⁻¹ (0.78–1.03) for LDL-c, 0.52 g · L⁻¹ (0.47–0.58) for HDL-c, and 1.08 g · L⁻¹ (0.90–1.25) for TG. These results were achieved with lipid-lowering treatment in 81.8% (59.7–94.8) of cases. Mean FBG was 0.99 g · L⁻¹ (0.95–1.02), significantly higher than the rest of the study population if based on the 95% confidence intervals. According to WHO criteria, 10 subjects (45.5%) had prediabetes. In terms of risk factors, this sample presented an average of 3.4 risk factors per subject, including five smokers (active or weaned for less than 3 yr), six cases of metabolic syndrome, and one case of OSAS. Unfortunately, the study protocol did not allow for the collection of risk factors prior to the diagnosis of cardiovascular disease in this secondary prevention group. More than two-thirds of the population had at most only one CVRF [69.9% (68.2–71.6)]. In 82.5% of cases, this corresponded to dyslipidemia [55.8% (52.4–59.1)] or smoking [26.7% (23.8–29.8)] (Fig. 2). The mean number of CVRF increased with age (P < 0.001). It was higher in men than in women [1.19 (1.14–1.25) vs. 0.61 (0.53–0.70), P < 0.001] (Table I).

View Figure

• Download as Powerpoint
• Download Figure

Table I. Distribution of Parameters for Global Cardiovascular Risk According to Sociodemographic Data.

View Fullscreen

After adjusting for age and gender, this number was found to be identical between civilians and military personnel (P = 0.45). It was significantly unfavorable for two professional categories. That observed among controllers was higher than that of pilots [+0.39 factors (0.20–0.57), P < 0.001], cabin crew [+0.40 factors (0.14–0.66), P < 0.001], and paratroopers[+0.60 factors (0.16–1.04), P < 0.01], while that of rear crew was higher than that of pilots [+0.22 factors (0.07–0.38), P < 0.001] and paratroopers [+0.44 factors (0.01–0.87), P < 0.05] (Fig. 3A).

View Figure

• Download as Powerpoint
• Download Figure

In applicants, this mean number of CVRF was significantly higher than that documented among trained crew [+0.40 (0.29–0.51), P < 0.001]. In the presence of at least one risk factor, 51.7% (44.0–59.4) of the applicants were dyslipidemic and 48.3% (40.6–46.0) were smokers. The 10-yr global risk of CVD was estimated at 0.58% (0.54–0.62) for SCORE and 5.16% (4.93–5.38) according to Framingham. This risk was thus defined as low for the entire population studied. The two models had an excellent correlation between them [Pearson coefficient calculated at 0.89 (0.88–0.89), P < 0.001]. These values were higher in men than in women: respectively, 0.65% (0.61–0.69) vs. 0.14% (0.11–0.17) (P < 0.001) for SCORE, 63% (5.38–5.89) vs. 2.22% (1.96–2.48) (P < 0.001) for Framingham. Both scores increased with age (P < 0.001) (Table I). After adjusting for age and gender, the estimates were identical between civilians and military personnel (P = 0.35 for SCORE and P = 0.43 for Framingham). The comparison of the global risk between the professional categories was most often to the benefit of the cabin crew. Their risk according to SCORE was indeed significantly lower than that of pilots [−0.41% (0.25–0.57), P < 0.001], controllers [−0.31% (0.01–0.60), P < 0.05], navigators [−0.34% (0.06–0.63), P < 0.01], and rear crew [−0.38% (0.20–0, 56), P < 0.001] (Fig. 3B). Their risk according to Framingham was also lower than that of pilots [−1.34% (0.41–2.27), P < 0.001], controllers [−1.72% (0.67–2.78), P < 0.001], and rear crew [−1.50% (0.45–2.54), P < 0.001] (Fig. 3C).

In addition, applicants presented a global risk significantly higher than that documented among trained aircrew: +0.45% (0.37–0.53) for SCORE (P < 0.001) and +1.72% (1.27–2.17) for Framingham (P < 0.001). Those with at least one CVRF had a Framingham risk 0.42% (0.21–0.63) higher than that of the other applicants (P < 0.001). In contrast, the SCORE risk was not significantly different between these two groups (P = 0.49).

A global risk level from moderate to very high was shown in one in five subjects according to the SCORE model [20.1% (18.6–21.6)], one in six subjects according to Framingham [16.3% (14.9–17.7)], and nearly one in three subjects according to the summation model [30.1% (28.4–31.9)]. Regardless of the model used, these proportions were higher in men than in women (P < 0.001) and increased with age (P < 0.001) (Table I).

After adjusting for age and gender, these proportions were identical between military and civilian personnel, whatever the model used. After post hoc comparisons between professional categories, the proportions for the SCORE risk were identical. For the Framingham risk, the only significant difference concerned the controllers, with a higher proportion than that observed among the pilots (P < 0.05). The summation model found this same difference to the detriment of the controllers (P < 0.001) as well as the rear crew (P < 0.05). On the other hand, only this last model documented a higher proportion of risk level from moderate to very high in applicants than among trained aircrew {OR = 3.75 (2.45–5.71), P < 0.001]. Regardless of the model used, the proportion of risk observed in applicants with at least one risk factor was not different from that of the other applicants (P = 0.99).

The CVRF distribution according to sociodemographic data and global risk parameters is presented in Table II. The results of multivariate analyses by logistic regression models are presented in Table III.

Table II. Distribution of Cardiovascular Risk Factors Based on Sociodemographic Data and Global Cardiovascular Risk Parameters.

View Fullscreen

Table III. Associated Factors with the Different CVRF (Multivariate Analyses by Logistic Regression Models).

View Fullscreen

The prevalence of dyslipidemia was 44.6% (42.8–46.5), with an overall tendency to increase with age (P < 0.001). Men were more affected than women (47.3% vs. 27.7%, P < 0.001). Dyslipidemic people had a mean number of CVRF of 2.06 (1.99–2.12) vs. 0.35 (0.32–0.39) if there was no lipid metabolism disorder (P < 0.001) (Table II). In multivariate analysis, controllers [OR = 1.56 (1.12–2.18), P < 0.01] and rear crew [OR = 1.44 (1.09–1.89), P < 0.01] duties were positively and significantly associated with the risk of dyslipidemia compared to pilots (reference), while military status was protective [OR = 0.76 (0.61–0.96), P < 0.05]. Among other risk factors, hypertension [OR = 1.72 (1.27–2.33), P < 0.001] and the presence of OSAS and/or metabolic syndrome [OR = 2, 02 (1.40–2.94), P < 0.001] were also significantly associated with the risk of dyslipidemia (Table III).

The prevalence of HBP was 12.4% (11.2–13.7) [13.8% (12.4–15.2) in men and 3.9% (2.2–6.3) in women, P < 0.001]. It increased significantly with age, from 4.9% in 18–34 yr olds to 63.2% in 65–74 yr olds (P < 0.001). The prevalence in men was higher than women for all age groups. People with HBP presented on average 2.96 (2.81–3.10) CVRF against 0.85 (0.82–0.89) in normotensive subjects (P < 0.001) with, in almost a third of cases, associated OSAS and/or metabolic syndrome [31.7% (26.8–36.9)] (Table II). In multivariate analysis, regardless of age and gender, ATCO duties [OR = 1.91 (1.22–2.96), P < 0.01] were positively and significantly associated with the risk of HBP vs. pilots (reference). Among the other risk factors, dyslipidemia [OR = 1.74 (1.29–2.37), P < 0.001], obesity [OR = 2.37 (1.54–3.60), P < 0.001], and the presence of OSAS and/or metabolic syndrome [OR = 3.12 (2.24–4.34), P < 0.001] were also associated with the risk of HBP (Table III).

Tobacco consumption, current or weaned for less than 3 yr, affected 17.4% (16.1–18.9) of the population, without significant differences between age groups. Women smoked more than men (22.5% vs. 16.6%, P < 0.01). Smokers had a mean number of risk factors of 1.95 (1.85–2.06) compared to 0.94 (0.89–0.98) in non- or ex- smokers (P < 0.001). Almost one in two smokers was also dyslipidemic [46.4% (41.9–50.9)] (Table II). In multivariate analysis, gender was no longer a factor significantly associated with tobacco consumption. On the other hand, all professional categories, with the exception of navigators, were significantly associated with a higher probability of exposure to tobacco than that of pilots (reference): rear crew [OR = 2.32 (1.73–3.11), P < 0.001], controllers [OR = 2.81 (1.98–3.96), P < 0.001], cabin crew [OR = 3.39 (2.24–5.13), P < 0.001], paratroopers [OR = 2.09 (0.99–4.09), P < 0.05], and Other [OR = 4.85 (2.07–10.9), P < 0.001]. Military status was associated with a lower likelihood of smoking [OR = 0.70 (0.53–0.92), P < 0.05] while the presence of OSAS and/or metabolic syndrome was significantly associated with smoking [OR = 1.45 (1.00–2.08), P < 0.05] (Table III).

More than a third of the population [38.3% (36.5–40.2)] was overweight or obese. This involved 41.5% (39.5–43.5) of men and 18.7% (14.9–22.9) of women (significant difference according to sex: P < 0.001). The prevalence of obesity was 4.8% (4.0–5.7), higher in men than in women [5.2% (4.3–6.1) vs. 2.6% (1.2–4.7), P < 0.05] (Table II). Almost 9 in 10 obese subjects [86.6% (79.6–91.8)] presented with so-called moderate obesity (30 ≤ BMI < 35 kg · m⁻²).

In multivariate analysis, gender was not an independent risk factor for obesity (P = 0.69). On the other hand, ATCO duties were significantly associated [OR = 2.08 (1.11–3.83), P < 0.05] with the probability of obesity greater than that of pilots (reference) while dyslipidemia [OR = 1.65 (1.02–2.72), P < 0.05], HBP [OR = 2.32 (1.51–3.55), P < 0.001], and OSAS and/or metabolic syndrome [OR = 6.02 (3.91–9.27), P < 0.001] were significantly associated with obesity (Table III).

In total, 3.0% (2.4–3.8) of the population had diabetes or prediabetes. These patients had a mean number of risk factors of 3.72 (3.33–4.10) vs. 1.03 (0.99–1.08) (P < 0.001) if there were no glucose metabolism disorders. Dyslipidemia was associated in 70.6% (59.7–80.0) of cases (Table II).

Due to the small number of diabetic people, diabetes and prediabetes criteria were combined to perform multivariate analyses. Thus, ATCO duties were, in comparison with pilots (reference), an independent risk factor for diabetes or prediabetes [OR = 2.34 (1.09–4.91), P < 0.05]; likewise with the presence of OSAS and/or metabolic syndrome [OR = 7.57 (4.43–13.05), P < 0.001] (Table III).

The prevalence of metabolic syndrome was 9.1% (8.1–10.3). It was significantly higher in men than in women [10.1% (8.9–11.4) vs. 3.1% (1.6–5.4), P < 0.001], with a tendency to age-related increase (P < 0.001). Only 10 cases of OSAS, diagnosed and treated, were collected [0.4% (0.2–0.7)], with no difference related to gender (P > 0.99). After post hoc comparisons between age groups, no difference was retained as significant (P > 0.20). Patients with OSAS and/or metabolic syndrome had a mean number of risk factors of 3.39 (3.22–3.55) vs. 0.88 (0.85–0.92) for patients without (P < 0.001). Almost 8 times out of 10 [79.2% (73.7–83.9)], dyslipidemia was associated (Table II) with higher risk. The small number of patients with OSAS led to grouping with metabolic syndrome to perform multivariate analyses on the documented basis of a very frequent pathophysiological association.⁸ Risk factors sharing a common field of definition with metabolic syndrome were thus significantly associated with it: dyslipidemia [2.08 (1.44–3.04), P < 0.001], HBP [OR = 3.14 (2.25–4.37), P < 0.001], obesity [OR = 6.11 (3.95–9.44), P < 0.001], and diabetes or prediabetes [OR = 8.10 (4.65–14, 15), P < 0.001] (Table III).

DISCUSSION

The distribution by age groups shows a gradual attrition of aircrews. The majority of the population studied was military. They are significantly younger than civilians and have shorter careers in aviation, at the end of which retraining in the civil aviation sector is not systematic. At the same time, the oldest crews are those whose state of health remains compatible with the pursuit of aeronautical activities. This implies a better medical (in general) and cardiovascular (in particular) condition than other people of the same age who have interrupted their aeronautical career by choice (retirement) or for medical reasons (unfitness, with a high statistical representation of cardiovascular disease, as mentioned above¹⁷).

The prevalence of CVD in our sample was nearly 10 times lower than that of the French general population (0.8% vs. 7.6%).²⁵ The estimated probabilities of cardiovascular morbidity and mortality were generally low and, according to the validated models, our results seemed both consistent and better than in the general population, even after adjusting for the targeted age groups (40–65 yr for SCORE and 30–74 yr for Framingham).

In France, a survey about the evolution of cardiovascular risk and coronary mortality in the urban community of Lille found a SCORE risk of 2.2% in men and 0.7% in women among the 1636 ELISABET study participants over the period 2011–2013.² In the same age groups, the results obtained among aircrew were 1.24% (1.18–1.30) for men and 0.30% (0.23–0.37) for women. On a French national level, the study of 50,856 volunteers from the CONSTANCES cohort reported a global median SCORE risk of 0.9% (0.3–2.1) [1.7% (0.8–3.2) in men and 0.5% (0.2–1.2) in women]. Based on the 95% confidence intervals, our results were not significantly different: the median was 0.95% (0.90–1.00) [1.05% (1.0–1.10) in men and 0.25% (0.20–0.28) in women]. On the other hand, the distribution of risk levels according to gender seemed to be favorable to this population.¹⁸ In parallel, the ELISABET study reported a Framingham score of 14–15% in men and 6% in women, compared with 9.8% and 3.7%, respectively, in aircrew. From 2007–2012, a study carried out in the “Hauts-de-Seine” department on 6504 adults in primary prevention without treatment made it possible to estimate this same risk at 11.7% in men and 5.9% in women.¹³ For identical age groups, our results were 9.8% and 5.7%, respectively.

Internationally, a favorable trend for aircrew was also revealed in the light of comparisons of the Framingham risk with the general populations of southern Europe¹ and North America.⁷ Our results, when limited to civilian pilots only, found an overall Framingham score identical to that of British airline pilots (8.29% vs. 8.41%).¹¹ The aircrew population, who is presumed to be in good health through medical selection and regular monitoring, would therefore be less at risk from a cardiovascular perspective than the general population. The role of social determinants in cardiovascular risk has thus been demonstrated at the level of European populations.^21,22 However, after multivariate analyses based on logistic regression models, no significant association between cardiovascular risk levels and aviation professional duties could be retained (Table IV).

Table IV. Associated Factors Associated with the Primary Outcomes (Multivariate Analyses).

View Fullscreen

One of this study objectives was to find a possible profile of interest in applicants with at least one CVRF in order to better characterize the factors of CV risk emergence. In this subgroup, the results highlighted systolic blood pressure and LDL-c as the only clinical and biological parameters significantly associated with the CV risk, in parallel with a high prevalence of dyslipidemia and smoking as concrete CVRF in this early risk identification. On the basis of these criteria, the Framingham score allowed documentation of a significantly higher global CV risk in the presence of at least one CVRF among applicants. Finally, at this number of applicants, the highlighting of an absolute risk greater than that of trained aircrew (after adjustment for age and gender) did not find an obvious explanation. The hypothesis of an influential contribution of the professional category “Others” within applicants could be put forward, the latter bringing together military people at an already advanced stage of their career, with more than a third of them being smokers (38.5%), dyslipidemic (38.5%), or affected by prediabetes (30.8%). Another hypothesis would be to consider the benefit of the selection criteria and close medical supervision on the CV risk control in favor of people already enrolled in professional aircrew categories.

In the current state of our knowledge, this study relates to one of the largest cohorts devoted to CVRF and the global 10-yr CV risk assessment in aircrew, without equivalent data in France except for a few declarative collections by questionnaires. Our activity, within one of the only two French military AeMC, allowed the recruitment of the largest possible panel of aircrew categories in France, including both military and civilian, but also private components issued from industrial and commercial sectors.

The comparisons between aircrew categories constituted a strong point of this survey, with reference to the previously cited publications. Several pitfalls are, however, to be highlighted, including its monocentric nature, which limited its representativeness. Some analyses could lack power due to limited numbers for the oldest age groups or for the female population, very weakly represented among the majority of the professional roles studied. Anthropometric data were collected on the day of the licensing medical visit by trained health professionals. On the other hand, biological data underlines a methodological limitation. In fact, with the exception of applicants for whom blood exams were processed by the biology laboratory at Sainte-Anne Military Hospital, the majority of biochemical analyses could not be performed on the day of the visit. The data, therefore, came from city laboratories, with mean delay of 1.39 yr (1.32–1.46) at the date of the test. In addition, the cases of diabetes and OSAS, very poorly represented in our study, were selected on the basis of a previously documented diagnosis in the medical file: this was a potential source of underestimation. The treatment of missing data based on multiple imputations by chained equations allowed us, despite sometimes high proportions, to exploit all the available data without affecting the statistical analyses power.

Some CVRF, although unanimously recognized by the scientific community, have not been taken into account, such as heredity, sedentary lifestyle linked to a lack of physical activities, an unsuitable diet, or even excessive alcohol consumption.²⁴ This methodological choice made it possible to limit an information bias linked to strictly declarative data, in particular in this medical context concluded by a fitness to fly decision and for which a certain degree of relative omission should be considered. Note that despite its strictly declarative nature, smoking habits were nevertheless included. This systematic assessment in our daily practice has indeed encouraged us to utilize it so as not to overlook a major factor included in predictive risk models. Therefore, the main useful parameters for estimating the global 10-yr CVD risk were collected.

The comparisons with the general population were based on estimated prevalences by direct standardization and weighting results by the gender and age group distributions of reference populations. These comparisons are both an advantage and one of the main limitations of our study, illustrating an important “Healthy Worker Effect” as selection bias.¹⁹ Indeed, morbidity and mortality rates are commonly lower among aviation and military workers than in the general population, on the intuitive basis that their health conditions must be compatible with standards of recruitment and retention in professional roles. The aircrew population is particularly sensitive to this bias, imposed by the rigor of selection and medical supervision standards. Comparisons with the general population are a clear illustration of this, but, in the particular case of primary CVD prevention, they nevertheless demonstrate the performance of aviation medicine in the control of determinants likely to compromise professional suitability.

In current practice, our results suggest paying particular attention to cardiovascular prevention in people at “moderate” risk; the occurrence of CVD for such a level of risk could prompt strategic thinking around preventive medicine. A large international cohort revealed that the majority of coronary artery disease, especially in Western Europe, occurred in the presence of a single “conventional” risk factor (smoking, dyslipidemia, diabetes, or hypertension).¹⁴ Age appeared to be the most important sociodemographic factor associated with the risk models tested (Table IV) and our results showed, despite a very low number of people in secondary prevention, consistent results through a similar age profile between “moderate” and “very high” risk levels. Prevention therefore deserves to be promoted for people with a “moderate” global risk level. It could be organized around screening and reinforced control of CVRF, in particular dyslipidemia and arterial hypertension, in accordance with the issues already identified in the general population^5,20 and those predicted by recent recommendations.¹⁵ Smoking should be the subject of a priority prevention axis among applicants.

Our results also underline the need to carry out future comparisons through larger scale studies between aircrew categories to better document the possible translation of various operational and workplace constraints. Despite the lack of significant association with the tested predictive models, the profile of air traffic controllers, highly involved in aviation safety and exposed to sedentary and shift work, nevertheless seems to call for particular vigilance. Finally, models for assessing a global CVD risk can benefit from recalibrations in order to guarantee a relevant use for the benefit of populations with variable morbidity and mortality and maintain their performance by integrating the epidemiological trends of the CVRF.²³ The value of a prospective study designed to evaluate these models with a view to a specific adjustment to certain professional categories involved in safety functions, in particular aeronautics, could be discussed. As an illustration, a New Zealand case-control study conducted with airline pilots did not demonstrate the expected 5-yr predictive performance of the Framingham model.²⁷

This original study was conducted for the benefit of a professional population whose medical approach is inseparable from the essential concepts of risk control and aviation safety. Beyond the translation of favorable cardiovascular health, crediting the medical selection standards and close supervision, areas for improvement have been identified, inviting the development of prevention strategies around the “moderate” risk level, but also some conventional risk factors, which already are a source of issues in the general population, as reported above. The challenges and limits represented by global risk modeling will also have to change in response to individualized decisions based on additional risk factors currently not taken into account, such as metabolic syndrome, prediabetes, or OSAS, whose predictive involvement will deserve to be specified in the future.