Study design
The PERU MIGRANT Study is a prospective population-based cohort [4], that was established to assess the risk factors of cardiovascular disease in three population groups in Peru: urban dwellers, who were born and currently live in Pampas de San Juan de Miraflores, a periurban area in Lima; rural inhabitants from San José de Secce and Chacas in Ayacucho (highlands settings); and rural-to-urban migrants, inhabitants born in Ayacucho (rural) who migrated to Pampas de San Juan de Miraflores (urban).
Participants
Participants were recruited from 2007 to 2008 using a random sampling technique stratified by sex and age (30–39, 40–49, 50–59 and 60 + years) in each population group using household census data [4]. For all study groups, men and women ≥ 30 years of age and habitual residents (≥ 6 months) of the study area were considered eligible. Pregnant women and people with mental disorders that prevented them from giving informed consent were excluded.
Following a random selection process, participants were invited to participate. They provided consent to participate in the study, completed a questionnaire, and had their anthropometric measurements assessed [4]. Survival data were retrieved through the National Registry of Identification and Civil Status (RENIEC) information to ascertain vital status and date of death (or censoring) when relevant.
Variables
Outcome
All-cause mortality at 10 years of follow-up was the outcome of interest. After the RENIEC database was searched, participants were classified as alive or dead. Those participants who were not available in that database (i.e. their national identity number was incorrectly recorded), were considered censored at the last time of contact (i.e. first or second in-person follow-up) [5, 6].
Independent variables
Aggregation and pair-wise combinations of cardiovascular risk factors were the independent variables. Risk factors included hypertension, type 2 diabetes mellitus, hypercholesterolemia, and overweight/obesity. These chronic cardiovascular risk factors were studied at baseline following standard procedures for all participants [4]. Hypertension was defined as systolic blood pressure (SBP) ≥ 140 mm Hg, diastolic blood pressure (DBP) ≥ 90 mm Hg or self-reported diagnosis with the use of anti-hypertensive medications [7]. Type 2 diabetes mellitus was defined based on fasting blood glucose ≥ 126 mg/dl or self-reported diagnosis with the use of anti-diabetic medications [8]. Overweight/obesity was defined as a body mass index (BMI) ≥ 25 kg/m2 [9]. Hypercholesterolemia was defined as total cholesterol ≥ 200 mg/dl [10], and was assessed from blood samples drawn after an 8-h fasting period.
We examined mortality rates with different cardiovascular risk profiles. In the first part, cardiovascular risk factors were aggregated and then split into one, two, and three or more cardiovascular risk factors. On the other hand, the second part comprised six groups of analysis with selected cardiovascular risk factors (e.g., group 1: hypertension with type 2 diabetes mellitus; group 2: hypertension with hypercholesterolemia, etc.) Each group was evaluated in four categories: disease-free individuals (reference), people with only one disease (e.g., only hypertension and only type 2 diabetes mellitus), and the pair-wise combination (e.g., hypertension with type 2 diabetes mellitus). The rationale of using this approach was to capture the individual effect of each chronic condition in addition with the pair-wise combination using the same model, and in this way, avoiding misclassification bias.
Covariables
Other variables were included in the analysis as potential confounders. Sociodemographic variables included age (30–39, 40–49, 50–59, and 60 + years), sex, migrant status (rural, migrant or urban), education level (< 7 years of education vs. ≥ 7 years of education), and socioeconomic status (low, medium, or high), measured using a wealth index based on household income, assets and household facilities. Lifestyle variables included tobacco use (having smoked at least one cigarette per day compared to never users in the 6 months prior to the interview, and classified as yes or no), alcohol consumption (self-reported consumption of ≥ 6 beers or its equivalent in alcohol with other beverages on the same occasion at least once a month, and classified into low or high), HDL cholesterol (> HDL-c 40 mg/dl in men and > 50 mg/dl in women) [10], and high waist circumference (> 80 cm in women and > 90 cm in men) [11].
Statistical analysis
The statistical analysis was conducted in STATA 14 for Windows (STATA Corp, College Station, TX, USA). The characteristics of the study population were tabulated according to the population group at baseline, and Chi-squared tests were used to compare categorical variables.
For the bivariate analysis, the log rank test was used to evaluate the association between sociodemographic variables, lifestyles, each cardiovascular risk factor, and aggregated cardiovascular risk factors and all-cause mortality. The Cox proportional hazard model was used to estimate the effect of the aggregation of cardiovascular risk factors on all-cause mortality. Crude and adjusted models were computed. Adjusted models included sociodemographic and lifestyle variables.
The association between the pair-wise combinations of cardiovascular risk factors and all-cause mortality were evaluated separately using crude and adjusted Cox regression models. Adjusted models included sociodemographic and lifestyles variables, as well as the cardiovascular risk factors not included in the pair-wise combination. Cox regression assumptions of independent observations, and independent censorship of survival and proportional risks were assessed. The latter assumption was evaluated using the global test of Schoenfeld residuals. Given the number of confounding variables, the variance inflation factor was also utilized to evaluate the presence of collinearity.