This study focused on the use of cause-specific hazard regression and Fine-Gray models versus competing risk as death due to noncardiac causes to investigate the effect of risk factors related to the incidence of coronary artery disease in overweight and obese populations of the city of Yazd, aged 2074 years, according to cohort data. In the present study, the 10-year cumulative crude incidence of coronary artery disease was 10.6% vs 6.8%.
The percentage of the competing cause as noncardiac deaths. Additionally, in line with previous studies, older age(33), higher systolic blood pressure(33), higher uric acid, male sex, diabetes(33), and hypertriglyceridemia increased the likelihood of developing coronary artery disease, and those with higher education (inversely) were less likely to develop CAD.
In the present study, the cause-specific and Fine-Gray models performed similarly in the 10-year prediction of coronary artery disease, in line with the study of Wilber et al.(34). According to the findings, the cause-specific hazard and subdistribution hazard of coronary artery disease associated with a per ten-year increase in age, systolic blood pressure, and uric acid had a very slight difference and similarly the cause-specific hazard and subdistribution hazard of coronary artery disease associated in during ten-year in diabetics, hypertriglyceridemia and higher education were similar.
However, Wilber's reported that 18% and 8% of people were classified at high risk of coronary artery disease according to the Cox and Fine-Gray models, respectively(34). Puddu et al. also showed that the use of the CIF model based on the Fine-Gray model was more appropriate and could avoid overestimations of the Kaplan–Meier method based on methods such as the Cox model(35). The study of Mackenzie reported that Cox proportional hazard models were not a good predictor of individual risk (36).
According to our findings, a per ten-year increase in age was associated with a 3.3% increase in the cause-specific hazard rate of coronary artery disease and a similar 3.2% increase in the risk of coronary artery disease (relative incidence) for at-risk individuals (Table 3). Age is an irreversible risk factor that plays a vital role in mortalities due to cardiovascular disease, remaining as a predictor variable in multivariate models even after adjusting for other confounders(37–39). In the present study, most cases of coronary artery disease occurred in the age of 65-74 years old.
Sex is an important risk factor for CVD in aging adults (37). In our study, men developed 52 (13.5%), and women developed 39 (7.9%) new cases of CAD. Consistent with the results of this study, Tsiampalis et al. reported that women were less likely to die from CHD, but Govender et al. reported that women were 1.96 times more likely than men to have a recurrence of cardiovascular disease(40).
Our findings showed that a per10-year increase in systolic blood pressure was associated with a 25.4% increase in the cause-specific hazard rate of coronary artery disease and, similarly, a 26% increase in the relative incidence of at-risk coronary artery disease. Consistent with our results, Puddu et al., using the Cox model (HR: 1.402, 95% CI: 1.3931.410, P <0.001) and Fine-Gray (SHR: 1.152, 95% CI: 1.1451.159, P <0.001), showed that increased systolic blood pressure was associated with a higher risk of CHD death(35). Ramezanian et al. also showed in the multivariate Cox model that a one-unit increase in systolic and diastolic blood pressure was associated with a 1% increase in the risk of CAD. Using a competing risk model, Dianatkhah et al. reported that hypertension was associated with a higher risk of cardiovascular disease(33).
In this study, diabetic patients were more likely to develop CAD. The cause-specific hazard and subdistribution hazard of coronary artery disease in high-risk diabetic patients were 2.5% and 2.5% higher than those in nondiabetic patients, respectively, and 19.9% of diabetic patients at the beginning of the study (n=176) developed CAD over ten years. Diabetic patients have a worse prognosis of CAD, so the risk of cardiac death is 2 to 4 times higher in these patients than in others(41). The increase in obesity epidemic in the world has led to the prevalence of CAD and diabetes(42). In many countries, the prevalence of diabetes in patients with coronary artery disease is reported to be 50% (43). Ramazanian et al. showed in the multivariate Cox model that each unit increase in fasting blood sugar increased the risk of CAD to 10% and 13% (44) in women and men, respectively. Tsiampalis et al. showed that increasing each unit in fasting blood sugar increased the risk of CHS death by 0.3%, and women were less likely to die from CHD(45). The cause-specific hazard and subdistribution hazard of CAD were approximately twice as high in at-risk people with hypertriglyceridemia as those with normal triglycerides. Additionally, a per 10-year increase in uric acid was associated with a 21% increase in the cause-specific hazard rate and subdistribution hazard of CAD in at-risk individuals. Longitudinal epidemiological studies have shown that serum uric acid is a predictor for the incidence of cardiovascular diseases and cardiac deaths(46). A study showed an association between elevated uric acid levels and overweight and obesity(47). Additionally, our findings showed that the risk of coronary artery disease decreased with increasing level of education, taking into account the competing event (death due to noncardiac causes) (Table 3).
Education is one of the socioeconomic factors affecting the risk of cardiovascular diseases and is completed by adulthood(48, 49). In a cohort study on 9226 patients without a history of cardiovascular disease, Dégano et al. reported that the risk of cardiovascular disease was lower in people with a university degree than in those with a lower or primary education (HR = 0.51, 95% CI: 0.300.85)(49). It seems that preventive strategies should be planned and made available to individuals according to their level of education(49). For example, some people may have learning disabilities, increasing lifestyle-related risk for CVD in these individuals(48). It also seems that people with higher education will have a healthier lifestyle despite economic changes, and the risk of cardiovascular disease will be reduced in these people(48). The findings of this study are consistent with other studies.
The at-risk group in the Kaplan–Meier model does not include individuals who experience a competing event before the event in question(18). The Cox model considered competing risk as censorship when examining the effect of variables on a particular cause(18). One of the challenges of competing risks in medical research on survival analysis is the one-to-one mismatch between the hazard function (rate) and the cumulative incidence function(18). In the Cox model, there is a one-to-one correspondence between the hazard function and the cumulative incidence function; therefore, the hazard ratio is equivalent to the rate in epidemiology and reflects the risk in population studies(18). The cumulative incidence function in the cause-specific model is affected by both the cause-specific function of the desired event and the competing event(34). The results of the cause-specific model for a particular variable should not be reported as an increase or decrease in the incidence of the disease because the effect of the variables on both the cause-specific hazard function and the competing event must be determined to examine the effect of variables on the incidence of the disease(34, 50).
The subdistribution function still keeps those who experienced the competing risk before the desired event in the at-risk set(11, 50). For example, if the competing risk is death due to cardiac causes, deceased heart patients are no longer at risk for the disease; therefore, the subdistribution function is not considered a standard epidemiological rate as a hazard ratio(51, 52). The hazard ratio in the Fine-Gray model is interpreted as a relative change in the desired event rate in people who have not experienced the desired event and the competing cause by a certain time(50). and directly determines the effect of variables on disease incidence(14, 18). In general, it is suggested to use the cause-specific hazard function for studies of disease etiology(11, 53). and the subdistributed hazard function for prognostic, predictive, and risk score classification in survival studies(11, 14, 54).