Novel association of the obesity risk-allele near Fas Apoptotic Inhibitory Molecule 2 (FAIM2) gene with heart rate and study of its effects on myocardial infarction in diabetic participants of the PREDIMED trial

We analyzed 7,161 participants (3,049 men and 4,112 women) in the PREDIMED (PREvención con DIeta MEDiterránea) trial from whom the FAIM2-rs7138803 polymorphism was successfully determined. The PREDIMED study is a multi-center, randomized, controlled clinical trial (controlled-trials.com number, ISRCTN35739639) aimed at assessing the effects of the MedDiet on the primary prevention of cardiovascular disease [36]. The completion date of this study was December 2010 and the total number of randomized subjects was 7,447. The 7,161 participants included in the present study did not differ in the main characteristics from those of the total cohort. Additional file 1: Figure S1 shows the CONSORT flowchart of the trial. Details of the PREDIMED trial including sample size calculations and interim analysis have been fully described elsewhere [36‐38]. Briefly, from October 2003 physicians in Primary Care Centers selected high cardiovascular risk subjects. Eligible subjects were community-dwelling people (55–80 years of age for men; 60–80 years of age for women) who fulfilled at least one of two criteria: type 2 diabetes [39] or three or more cardiovascular risk factors as previously detailed [36‐38]. Exclusion criteria included a personal history of cardiovascular disease, any severe chronic illness, and drug or alcohol addiction [36]. Participants were randomly assigned to three interventions: MedDiet with extra-virgin olive oil, MedDiet with mixed nuts and control group (low-fat diet). Participants assigned to both MedDiet groups received intensive training to follow the MedDiet and allocations of either extra virgin olive oil (1 litre/week, for all the family) or mixed nuts (30 g/d) at no cost [36, 37]. Participants assigned to the control diet received recommendations on reducing the intake of all types of fat. A detailed description of the nutritional interventions has been provided elsewhere [36]. The Institutional Review Board/Ethics Committee of each participating center approved the study protocol. All participants provided written informed consent. Participants were followed up for a median of 4.8 years (interquartile range, 2.8 to 5.8 years). Data were analyzed cross-sectionally (at baseline) as well as longitudinally (pooling together the MedDiet intervention groups versus the control group). By December 2010, a total of 502 participants (7.0%) had been lost to follow-up for 2 or more years. Dropout rates were higher in the control group (11%) than in the MedDiet groups (5%) (Additional file 1: Figure S1). As compared with participants who remained in the trial, those who dropped out were younger (by 1.4 years), had a higher BMI and a lower score for the adherence to the MedDiet as previously detailed elsewhere [36]. The effect of the FAIM2-rs7138803 polymorphism on heart rate was also examined longitudinally analyzing all subjects having heart rate data at baseline, at 1-year, at 3-years and at 5-years (n = 2,310). The reduction in the number of cases in the longitudinal analysis was mainly due no to patients being lost during follow up, but to the fact that patients were not all recruited in the same year at the beginning of the study, but progressively over the following years (from 2003 to 2009) as previously described [36]. Therefore, individuals who entered the study in 2003 completed the 5-year follow-up period in the year 2010, which is when follow-up ended, whereas those who entered the study in later years could not complete the 5-year follow-up period and so were not included in the longitudinal analyses of the study on heart rate. The Institutional Review Board/Ethics Committee of each participating center approved the study protocol. All participants provided written informed consent.

The baseline examination included an assessment of standard cardiovascular risk factors, medication use, socio-demographic factors and lifestyle variables, as previously detailed [36, 38]. Food consumption was determined by a validated semi-quantitative food frequency questionnaire [40]. Adherence to MedDiet at baseline was assessed by a validated 14-item questionnaire [41]. The final score ranged from 0 to 14. The greater the score obtained from the questionnaire, the greater the adherence to the MedDiet. Physical activity was estimated by the Minnesota Leisure-Time Physical Activity Questionnaire, as previously described [36].

Weight and height were measured with calibrated scales and a wall-mounted stadiometer, respectively. BMI was calculated as weight in kilograms divided by the square of height in meters.

At baseline and yearly thereafter, trained personnel measured blood pressure and heart rate in triplicate using a validated semiautomatic oscillometer (Omron HEM-705CP, Hoofddorp, the Netherlands) with a 5-minute interval between each measurement with the patients seated and at rest in a peaceful setting. The means of these measurements were calculated.

The primary endpoint was the occurrence of the first major cardiovascular events and comprised myocardial infarction, stroke or cardiovascular death [36]. We used four sources of information to identify end-points: 1) repeated contacts with participants; 2) family physicians; 3) yearly review of medical records; and 4) consultation of the National Death Index. The end-point adjudication committee, whose members were blind to treatment allocation, examined all medical records related to end-points. Only end-points confirmed by the adjudication committee that occurred between October 1, 2003, and December 1, 2010 were included in the analyses. The criteria for adjudicating primary outcomes are detailed elsewhere [36].

At baseline, blood samples were obtained from each participant after an overnight fast and were frozen at -80°C. Fasting glucose, total cholesterol, triglycerides, HDL-C and LDL-C were measured using standard methods as previously described [37].

Data were analyzed both at baseline and longitudinally in the intervention trial. Chi-square tests were used to test differences in percentages. T and ANOVA tests were applied to compare crude means of continuous variables among genotypes at baseline. Multivariable adjustments for continuous variables were carried out by linear regression analysis and regression coefficients (B) were estimated. Models were sequentially adjusted for age, sex, center, type 2 diabetes, total energy intake, adherence to MedDiet, alcohol, tobacco, physical activity, hypertension, and medications (antihypertensive, lipid-lowering and hypoglycemic drugs) as indicated. Antihypertensive medication included angiotensin-converting enzymes (ACE) inhibitors (48% of the population took these), diuretics (21% of the population) and other antihypertensive drugs (beta blockers, calcium channel blockers, etc.) regardless of the dosage, as previously described [36]. Three dummy variables (taking or not taking the corresponding drug regardless of the dosage) were considered for the antihypertensive medication including ACE inhibitors, diuretics and other antihypertensive drugs. In addition a dummy variable (taking or not taking) for lipid-lowering drugs and two dummy variables (one for insulin and other for oral antidiabetic agents) were included in the adjustment for medications. Hypertension was defined as systolic blood pressure > =140 mm Hg or diastolic blood pressure > =90 mmHg or under antihypertensive medication. The FAIM2-rs7138803 polymorphism was first tested as additive for continuous traits analyzed. However, when the differences between heterozygous and homozygous subjects for the variant A-allele were very small, a recessive model was selected (for heart rate). A dichotomous variable for the pre-randomization adherence to MedDiet was created using the sample means as cut-off values (9 points). Obesity was defined as BMI > = 30 kg/m². Logistic regression methods were also used to estimate the contribution of the FAIM2-rs7138803 polymorphism to predict obesity. In addition, BMI was categorized into four categories (BMI < 25 kg/m²; 25–30 kg/m²; 30–35 kg/m² and > =35 kg/m²). Heart rate in beats per minute (bmp) was also categorized into five categories (<=60, 60–70, 70–80, 80–90 and > =90 bmp) for categorical analyses. The interaction between the FAIM2-rs7138803 polymorphism and adherence to the MedDiet in determining BMI, blood pressure and resting heart rate at baseline was tested in multivariable regression models (linear or logistic for categorical) including the corresponding main effects and interaction terms. Subgroup analyses stratified by type 2 diabetes status were carried out according to our “a priori” hypothesis and aim. The effect of the FAIM2-rs7138803 polymorphism on heart rate in the whole population and stratified by type 2 diabetes status was also examined longitudinally analyzing all subjects having heart rate data at baseline, at 1-year, at 3-years and at 5-years (n = 2,310) in a multivariable model of repeated measures with interactions terms after adjustment for the dietary intervention and the indicated covariates. Dietary intervention was considered as a dichotomous variable pooling together the MedDiet intervention groups versus the control group after having checked the homogeneity of the effect in the two MedDiet groups). In addition, we analyzed the interaction term between the FAIM2-rs7138803 and dietary intervention on changes in heart rate at 5-years by linear regression analysis with multivariable adjustment. Finally, we examined the association between the FAIM 2-rs7138803 polymorphism and cardiovascular events (total cardiovascular events, myocardial infarction or stroke) in the whole population and in type 2 diabetic subjects by Cox regression models with the length of follow-up as the primary time variable and adjustment for the dietary intervention and the other covariates as indicated. The exposure time was calculated as the time between randomization and the date of a major cardiovascular event, the date when completing the last interview, December 1st 2010, or the date at death, whichever came first. Hazard ratios (HR) with 95% confidence intervals (CI) for the FAIM2-rs7138803 polymorphism (recessive model) for cardiovascular diseases were calculated in the population as a whole. In addition, we carried out the stratified analysis by type-2 diabetes status. Analyses were based on the intention-to-treat principle. In multivariable model 1 we adjusted for sex, age, center and dietary intervention. Multivariable model 2 included additional adjustments for BMI, total energy intake, alcohol, smoking, type-2 diabetes, medications and physical activity. Additional adjustments for baseline hypertension were also undertaken. The proportionality of hazards in determining the main outcomes in the PREDIMED study was tested with the use of time-varying covariates as previously described [36]. Here we graphically checked the assumption of proportional hazards for the genetic factor (FAIM2) by looking at the Log-Log plot of survival. We obtained parallel lines for the curves of genotypes in all the models fitted. Finally, Kaplan-Meier survival curves were plotted to estimate the probability of remaining free of myocardial infarction during follow-up. Statistical analyses were performed with the SPSS package, version 18.0 (SPSS, Chicago, IL). All tests were two-tailed and the nominal significance level was set at a P-value <0.05. In addition, the Bonferroni correction was applied to compensate for multiple comparisons. Our approach consisted of using an adjusted alpha level equal to the nominal alpha level (0.05), divided by the number of new hypothesis tested. With the aim of achieving a balance between type I and type II errors, we have chosen to undertake the Bonferroni correction considering the three new groups of variables that we were investigating (heart rate, blood pressure and cardiovascular diseases). Anthropometric measures were not included in the correction because in this case we were not dealing with a new hypothesis of association. Thus, the adjusted alpha was 0.05/3 = 0.017. The significance of each original (uncorrected) test was assessed at this level. Thus, a P-value <0.017 was considered to be statistically significant taking into account the correction for multiple comparisons.

We observed a statistically significant association between the FAIM2-rs7138803 polymorphism and BMI (B_adjusted = 0.15: 95% CI: 0.02-0.27 kg/m² per variant-allele; P = 0.023) in the additive model after adjustment for sex, age, centre, type 2 diabetes, hypertension and medications (Table 2). We also found additive effects in the association of this polymorphism with obesity (OR_adjusted: 1.08; 95% CI: 1.01-1.17 per variant-allele; P: 0.011). Likewise, we obtained a statistically significant association with diastolic blood pressure (DBP) in the additive model adjusted for covariates that remained significant after BMI adjustment (P = 0.015). Moreover, we observed a novel association between this polymorphism and resting heart rate which was greater in the recessive model, and that remained statistically significant after multivariable adjustment including BMI (P = 0.012). Additional adjustment of the model for DBP did not attenuate the statistical significance of this association (P = 0.010). All these association were statistically significant not only at the nominal P-value (P < 0.05) but also after correction for multiple comparisons taking the level of P < 0.017 into account.

In the stratified analysis by type-2 diabetes status (48% diabetic patients), we observed that the association of the polymorphism with heart rate was greater and statistically significant (P = 0.010) in type-2 diabetic subjects (B_adjusted = 1.44: 95% CI: 0.23-2.56 bmp for AA versus carriers of the G-allele). No significant association was found in non-diabetics (B_adjusted: 0.62: 95% CI: -0.47-1.69 bmp; P = 0.258, in non-diabetics for AA versus G-carriers). However, in terms of BMI, we found no heterogeneity of the effects of the polymorphism by diabetes status, obtaining similar associations in type 2 diabetic subjects as in non-diabetics (not shown).

We also analyzed (Additional file 1: Table S2) the association of the FAIM2-rs7138803 polymorphism with different categories of BMI and observed that the association increased with BMI categories (P-for trend: 0.008). Categorizing heart rate (Table 3) into five categories (<=60, 60–70, 70–80, 80–90 and > =90 bpm), we also observed that the association with the polymorphism was greater as the heart rate category increased (P-for linear trend: 0.002). Thus, prevalence of AA homozygotes was 9.9% in subjects having a heart rate < =60 bpm, whereas its prevalence was 15.0% among subjects having a heart rate > =90 bpm (OR_adjusted: 1.66, 95% CI: 1.13-2.43; P = 0.009 for AA homozygotes versus G-allele carriers) in the multivariable adjusted model including BMI.

We did not observe any statistically significant gene-diet interaction in the multivariable adjusted models (by sex, age, centre, type 2 diabetes, medications, hypertension, smoking, drinking, physical activity and total energy intake) between adherence to the MedDiet (two categories based on the population mean of 9 points) at baseline and the FAIM2-rs7138803 polymorphism in determining BMI (P-interaction: 0.716) (Additional file 1: Figure S2A), waist circumference (P-interaction: 0.658), obesity (P-interaction = 0.679), DBP (P-interaction: 0.346) or heart rate (P-interaction: 0.371) (Additional file 1: Figure S2B).

Although in this study, prevalence of diagnosed hypertension at baseline was very high (82.8%), we analyzed whether the association of the polymorphism with DBP (Additional file 1: Figure S3A) and heart rate (Additional file 1: Figure S3B) differed by hypertension status. We obtained no statistically significant interactions and observed homogeneous results across the strata. Statistically significant differences for the polymorphism were detected in subjects with hypertension because this group had a higher sample size.

Given that that heart rate was the parameter in which we found the most novel association, we wanted to study whether the association between the FAIM2-rs7138803 polymorphism and heart rate was maintained over time and how the intervention with MedDiet (MedDiet groups vs control group after having observed no heterogeneity in the MedDiet groups) influenced that association. We used longitudinal data from 5-year follow-up analyzing data for all subjects having heart rate measured at baseline, at 1-year, at 3-years and at 5-years (n = 2,310) in a model for repeated measures. Prevalence of the FAIM2-rs7138803 genotypes in this group was 41.3% GG, 44.8% GA and 13.9% AA, very close to the population as a whole. The mean age was the same (67.1 ± 6.0 years) as in the population as a whole. There were slight differences in other demographic or clinical variables. Thus, in this group prevalence of women was 56.4%, prevalence of type 2 diabetes was 45.1%, mean BMI was 29.7 ± 3.6 kg/m², mean SBP was 84.4 ± 10.8 mmHg and mean heart rate was 70.5 ± 10.8 bpm.

In the analysis stratified by diabetes status (n = 1,076 subjects with type-2 diabetes and n = 1,234 non-diabetics), we found that the longitudinal association between the FAIM2-rs7138803 polymorphism and heart rate was higher and statistically significant in subjects with type-2 diabetes (P = 0.002 for the polymorphism in the longitudinal analysis of the multivariable adjusted model) (Figure 1). This longitudinal association remained statistically significant after correction for multiple comparisons. No statistically significant association for the polymorphism in the follow-up was detected in non-diabetics (P = 0.135). Even in type 2 diabetic subjects we did not detect a significant interaction between the polymorphism and the intervention with MedDiet (P-interaction = 0.587, multivariable adjusted) in the follow-up.

After a median follow-up of 4.8 years (interquartile range, 2.8 to 5.8 years), 268 major cardiovascular events occurred among the 7,161 participants analyzed (30,360 person-years of observation). Of these, 135 were strokes, 100 myocardial infarctions and the others were cardiovascular deaths. Firstly, we longitudinally analyzed the association of the polymorphism with the incidence of cardiovascular events and adjusted the model for dietary intervention (MedDiets and control group). Additional file 1: Table S3 of the Supporting Online Material shows the HRs for total cardiovascular events, stroke and myocardial infarction. In the whole population, we did not observe a significant association between the FAIM2-rs7138803 polymorphism (in the recessive model) and myocardial infarction either in the basic model or in the model adjusted for BMI (HR_adjusted: 1.57, 95% CI: 0.96-2.62). In the analysis stratified by type 2 diabetes, we observed a nominally significant association in type 2 diabetes patients (HR_adjusted:1.86; 95% CI:1.03-3.37; P = 0.041 in AA versus G-carriers). However, this association did not remain statistically significant following correction for multiple comparisons. Figure 2 shows cumulative myocardial infarction free-survival by the FAIM2-rs7138803 polymorphism in the whole population (A) and in type-2 diabetic subjects (B).