Results of the search
The primary stage of searches yielded 858 citations. Additionally, 44 studies also identified through gray literature. After removing duplicates, 657 studies selected to investigate the titles and abstracts, and 108 articles remained for the full-text review. Of these, 46 studies fulfilled the inclusion criteria. Finally, 30 articles, involving 1879170 individuals, were extracted for the meta-analysis. The reasons for exclusion of 645 papers were irrelevant title (n = 567), studies with inappropriate population (n = 12), not reporting prevalence of either diabetes or prediabetes (n = 11), reporting incidence of either diabetes or prediabetes (n = 2), review or meta-analysis study (n = 6), non-English papers (n = 10), and editorial, conference abstracts, or poster papers (n = 7). The flowchart of the included studies is shown in Fig. 1.
Description of studies
Among 30 included studies (26 cross-sectional and four cohort studies), 11 studies conducted in India [7, 8, 17-25], four in the USA [26-29], two in Iran [30, 31], two in Saudi Arabia [32, 33], one in Brazil [34], one in Sweden [35], one in Canada [36], one in Poland [37], one in Zambia [38], one in Jordan [39], one in Guinea-Bissau [40], one in Thailand [41], one in Brunei [42], one in Sri Lanka [43], and one in Ethiopia [6], with report of diabetes (n = 30) and prediabetes (n = 9) prevalence. Studies participants consisted of military personnel, soldiers, and conscripts, as well as police personnel and officer. The participants’ age ranged from 17 to 60 years old. Fasting blood glucose (FBG) was used as the main diagnosis method of diabetes and prediabetes in most studies. The basic characteristics of the included studies are shown in Table 1.
Quality assessment
Based on the JBI checklist, included studies categorized as satisfactory (n = 21), good (n = 9), and low (n = 0) quality. The results of the quality assessment are summarized in Table 1 and Supplementary file 1.
Heterogeneity
Considerable inter-study heterogeneity was observed based on the Cochran test and the I2 index. The results of heterogeneity and the models of analysis are expressed in Table 2.
Results of the meta-regression
The Results of meta-regression demonstrated an association between publication year of study and the prevalence of diabetes among military personnel (Reg Coef = 0.062, p = 0.069). However, this was not significantly associated with HDI score (Reg Coef = -8.11, p = 0.441), sample size (Reg Coef = 0.001, p = 0.935), and mean age of the participants (Reg Coef = 0.024, p = 0.879).
On the other hand, the obtained results showed an association between the mean age of the participants and the prevalence of diabetes among police officers (Reg Coef = 0.064, p = 0.089). Such a meaningful finding was not observed for HDI score (Reg Coef = 2.21, p = 0.182), sample size (Reg Coef = -0.001, p = 0.593), and publication year of study (Reg Coef = -0.001, p = 0.919).
Results of the meta-analysis
Prevalence of diabetes
The analysis of pooled data indicated the overall prevalence of diabetes among the military personnel was 9.15% (95% confidence intervals [CI]: 5.35-12.96; I2 = 99.8%, p <0.001). The results of subgroup analyses showed the countries with medium HDI (36.40%; CI: 23.68-49.12) and female gender (23.80%; CI: 22.90-24.70) had the highest prevalence of diabetes among the military personnel. The prevalence of diabetes among the police officers was 9.58% (CI: 5.98-13.18; I2 = 98.6%, p <0.001). The highest prevalence of diabetes among the police officers was observed in countries with medium HDI (13.17%; CI: 5.11-21.24, I2 = 98.4 %, p = 0.001) and male gender (11.99%; CI: 8.45-15.52, I2 = 98.9%, p <0.001) (Supplementary file 3). The results of the subgroup analysis stratified by gender are presented in Supplementary file 4. Based on the geographical regions, the highest prevalence of diabetes was related to the South Asia countries among both military personnel (36.40%; CI: 23.68-49.12) and police officers (14.20%; CI: 4.80-23.60) (Supplementary file 5). The results of the meta-analysis of diabetes prevalence stratified by a different variable are demonstrated in Table 2.
Prevalence of prediabetes
The overall prevalence of prediabetes among the military personnel and police officers were 7.32% (CI: 4.22-10.42, I2 = 44.3%, p = 0.166) and 6.39% (CI: 4.54-8.23, I2 = 81.5%, p = 0.005), respectively (Supplementary file 6). The highest prevalence of prediabetes among the military personnel was observed in countries with medium HDI (6.50%; CI: 5.80-7.20, I2 = 0.0 %, p = 0.724). The results of the meta-analysis of prediabetes prevalence among armed forces personnel are demonstrated in Table 2.
Sensitivity analysis
The findings were reanalyzed by deleting each study separately. The results demonstrated the studies with small sample size, and the outlier studies had no significant effect on the final finding.
Publication bias
Results of Egger’s tests demonstrated a lack of publication bias (p >0.10).