Association of ideal cardiovascular health with cardiovascular events and risk advancement periods in a Mediterranean population-based cohort

Participants that agreed to enroll in the cohort were contacted from one of the 25 participating primary healthcare centers. Trained nurses collected via face-to-face structured interviews participants’ self-reported socio-demographic and lifestyle information, including smoking habit, physical activity, adherence to the Mediterranean diet (MedDiet), individual and family medical history, and current medication use. The pharmacist and physician team members reviewed the participants’ medication for each condition and resolved with nurses any potential misclassification or conflict. Adherence to the MedDiet was measured using a previously validated 14-item questionnaire [24, 25]. Physical activity levels were assessed using the validated Spanish version of the Minnesota Leisure Time Physical Activity Questionnaire [26, 27]. Metabolic equivalent hours per week (METs-h/week) were calculated based on time spent and intensity required for each performed activity according to the international compendium of physical activity [28]. Anthropometric measures were conducted in duplicate by registered nurses using standardized techniques. Blood pressure was measured in triplicate using a validated semiautomatic oscillometer (OMRON® M4-1). Finally, laboratory test analyses were conducted to determine biological parameters, including fasting blood glucose, lipid profile, and high-sensitivity C-reactive protein level, among others [22, 23].

Our ICVH score was based on 7 equally weighted health metrics as proposed by the AHA [6]. We assigned 1 point for each of the following health metrics that participants met at baseline: never-smoker or quitting > 12 months ago, ≥ 9 points of adherence to MedDiet, ≥ 150 min/week of moderate physical activity or ≥ 75 min/week of vigorous intensity or an equivalent combination, BMI < 25 kg/m², untreated total cholesterol < 200 mg/dl, untreated blood pressure < 120/80 mmHg, and untreated fasting glucose < 100 mg/dl. Physical activity at “goal” was defined based on the World Health Organization recommendations [29], which is considered to be the equivalent of ≥ 500 METs-min/week based on the Compendium of Physical Activities (using a value of 3.3 METs, the level associated with a brisk [3 miles per hour] walking pace) [30]. This ≥ 500 METs-min/week cut-off was also used in previous publications to define physical activity guidelines [19, 31]. The MedDiet was selected as dietary pattern that ‘promotes cardiovascular health’ due to the large, robust, and consistent evidence of its cardiovascular beneficial effects [32, 33]. The ≥ 9 cut-point score was selected based on the definition of good adherence provided by previous MedDiet studies [34, 35]. Thus, the total ICVH score ranged from 0 (worst) to 7 (best) points. Participants were categorized into 4 groups according to the number of achieved metrics (0–2, 3, 4, and 5–7) to ensure an adequate distribution of the sample in each score group. The upper and lower extremes of the ICVH score range were collapsed due to the low proportion of participants with these numbers of metrics.

The primary endpoint of the study was a composite of myocardial infarction, stroke, or death from cardiovascular causes  (all major cardiovascular events). The secondary endpoints were the individual components of the main composite endpoint and an expanded composite of major cardiovascular events that additionally included other ischemic heart diseases, other cerebrovascular diseases, and peripheral arterial disease. Definitions of study endpoints are described in Additional file 1: Fig. S1. For the ascertainment of cardiovascular events, several sources of information were used: primary electronic health record database, hospital discharge database, population based myocardial infarction registry, and regional and national mortality registries (managed by the National Statistical Institute, Madrid, Spain), which provided information on date and cause of death. Nearly all the 650,000 inhabitants of the community of Navarre are covered by the primary care electronic health record database, which has resulted to be a valid source for epidemiological surveillance in previous validation studies of disease diagnoses [36, 37]. To further identify potential cardiovascular events requiring hospitalization, we additionally reviewed the Assisted Morbidity Registry of Navarre, which involved the Minimum Basic Data Set with information of all public and private hospitals in the Navarre Health System. Record linkage additionally included the Regional Registry of Myocardial Infarction of Navarre and the National Statistics Institute of Spain. The combination of 4 sources of information and the utilization of primary electronic health records assured the study to have a strict follow-up of vital status of participants (< 1% losses). Medical records were reviewed by members of the study team. A validation process was carried out to confirm and classify the cardiovascular events identified in the different information sources. Detailed information of the outcome assessment can be found somewhere else [23].

Descriptive statistics were used to summarize the distribution of the baseline characteristics of the study population across groups of the ICVH.

We used Nelson-Aalen estimators, adjusted for potential confounders with inverse probability weighting [38], to describe the incidence of the study’s endpoints over follow-up time across the 4 groups, and the differences between the groups were tested using the log-rank test.

Multivariate Cox proportional hazards models were used to calculate the hazard ratios (HRs) and 95% confidence intervals (CIs) of risk of developing cardiovascular events during follow-up. We considered the lowest score group (0–2 metrics) as the reference category. Time on study was the interval between the date of completion of the baseline interview and the date of diagnosis of cardiovascular events, or the date of death, whichever occurred first and the last recorded follow-up for non-cases. Follow-up of participants without cardiovascular events exceeding the end of follow-up date (31 December 2017) was censored at this date. After crude models, we fitted age and sex-adjusted models, and multivariable models additionally adjusted for higher level of attained education (primary or less, secondary, and college/university), and occupation (executives/managers, clerical workers, and manual workers).

Tests of linear trend across successive categories of the ICVH score were calculated by assigning the median value to each category treating the resulting variables as continuous. The potential non-linear relationship between ICVH metrics and the primary endpoint was graphically examined using fully adjusted restricted cubic splines with three knots. A p-value for non-linearity was calculated by the likelihood ratio test [39].

RAPs for the study endpoints were calculated according to Brenner et al. [13]. In our study, RAPs represent the equivalent risk associated with ICVH metric in terms of years of chronological age. Thus, RAPs represent the baseline age difference at which exposed subjects (participants with > 2 metrics) reach the same risk of disease (endpoint events) as unexposed subjects (participants with ≤ 2 metrics). Point estimates of RAPs and 95% CIs were derived from multivariate Cox proportional hazard models adjusted for the same potential confounders. We verified the proportionality of hazard assumption graphically and conducting Schoenfeld residuals to correctly interpret RAP estimates [14, 40].

The association of each individual health metric with the risk of developing major cardiovascular events was additionally calculated. Cox models were fitted for each health metric of the score adjusting the models for the effect of the other metrics and the same confounders as previous models. The reference category for each independent analysis was not meeting the criterion (0 points).

Sensitivity analyses were also conducted using additional adjustments: alcohol consumption (never [0 g/day]; light [< 5 g/day]; moderate [women: 5–15 g/day and men 5–30 g/day]; heavy [women: > 15 g/day and men > 30 g/day]), and serum concentration of high-sensitivity C-reactive protein (continuous). We additionally explored the association between individual health metrics and the primary endpoint of the study by removing from the models the adjustments of the effect of the other metrics.

This study has several limitations. First, ICVH metric information was assessed at baseline, and it may be possible that participants might have changed their metrics over follow-up. Second, despite our statistical models were adjusted for potential confounders, the absence of confounding cannot be completely ruled out. Third, the number of study endpoints in our cohort was limited, particularly secondary endpoints, which may have attenuated the statistical power of our results; nevertheless, previous findings of the RIVANA cohort on CVD have been reliable [23]. And fourth, it may be possible that certain lifestyle behaviors such as physical activity, diet, or smoking may be greatly rooted in a country’s culture; yet, it seems plausible the observed findings can be generalized to other populations. Despite these limitations, the strengths of our study are reflected in the population-based nature of the cohort, the rigorous method of collection of cardiovascular events supported by public resources and validated electronic health record database [36, 37], the long follow-up, the high retention proportion, and the inclusion of sensitivity analyses.