The fat mass and obesity-associated (FTO) gene variant rs9939609 predicts long-term incidence of cardiovascular disease and related death independent of the traditional risk factors

Abstract

Objective and methods The impact of the rs9939609 FTO variant on cardiovascular events was investigated in the 19-year follow-up of subjects recruited to the OPERA study.

Results A total of 212 cardiovascular disease (CVD) and 152 coronary heart disease (CHD) events or deaths occurred during follow-up. The logistic regression analysis revealed that among the AA genotype the incidence of CHD (OR 1.905; 95% CI 1.250–2.903, p = 0.001) and CVD (OR 1.849; 1.265–2.702, p = 0.003) events or death was significantly higher when adjusted for age, sex, and study group. After further adjustment with BMI, smoking status, systolic blood pressure, and low-density lipoprotein cholesterol, the higher incidence of CHD and CVD events or death among subjects with the AA genotype remained significant (OR 1.895; p = 0.002 and p = 0.004, respectively). In Cox regression analysis, the AA genotype displayed a higher rate of CVD and CHD death when the model was adjusted for sex, age, and study group (p = 0.006 and p = 0.046). FTO rs9939609 AA genotype improved the C-index of the final predictive model from 0.709 to 0.715. In reclassification analyses, the integrated discrimination index was significant 0.011 (p = 0.010).

Conclusion The AA genotype of FTO rs9939609 seems to be associated with a higher risk of CVD, and this phenomenon seems to be independent of the traditional risk factors for atherosclerosis.

Key messages

• The AA genotype of FTO rs9939609 seems to be associated with a higher risk of cardiovascular disease.

• This phenomenon seems to be independent of the traditional risk factors for atherosclerosis.

Keywords:

• Cardiovascular disease
• FTO
• mortality

Introduction

The fat mass and obesity-associated (FTO) gene is highly conserved between animal species (1). It seems to be expressed ubiquitously with the highest expression in cerebral tissue, especially in the appetite-regulating neurons of the hypothalamus (2).

In animal studies, FTO-/- mice display increased heat production and are resistant to obesity (3), and they have an altered response of white adipose tissue to high-fat diet (4). On the contrary, transgenic animals overexpressing FTO become obese associated with white adipocyte hypertrophy and increased energy intake (5). FTO has also been suggested to play a role in white fat ‘browning’ (6). A quite recent study (7) with Irx3-deficient mice has demonstrated a functional connection between the obesity-associated non-coding sequences within the first intron of FTO partly overlapping with rs9939609 and the iroquois homeobox gene 3 (IRX3). The mice were 30% smaller than their WT controls, with reduced fat mass and increased basal metabolic rate with browning of white adipose tissue. On the other hand, heterozygous deletion of another proximal gene RPGRIP1L resulted also in a mild obesity phenotype (8), indicating a possible complex regulatory interaction between several genes in the same genetic region. In human, FTO mutation has been observed to cause seriously impaired growth and development (9).

FTO SNP rs9939609 has been associated with body weight and adiposity in many studies (10–13), with the A allele being a risk allele. An association with eating behavior (10,12–15) has also been suggested, and this variant has also been demonstrated to regulate ghrelin and modulate responses to eating on brain reward regions (16). Furthermore, this FTO variant has also been associated with resting energy expenditure, plasma leptin level, and thyrotropin (17), as well as with fasting levels of blood glucose, insulin, and plasma leptin (18) in women. An association between the AA genotype of the variant and type 2 diabetes has also been well established in many studies (19–21). Moreover, FTO rs9939609 has been shown to be associated with coronary heart disease (CHD) and cardiovascular disease (CVD) risk independent of BMI in diabetic women (22) and with CVD risk independent of conventional risk factors in men with impaired glucose tolerance or type 2 diabetes (11). The last-mentioned study also reported an association of FTO rs9939609 with high-sensitivity C-reactive protein (hsCRP) level and low high-density lipoprotein (HDL) cholesterol concentration in men (11). Furthermore, it was shown—again in diabetic patients—that rs9939609 was associated with BMI, measures of dyslipidemia, and the risk of myocardial infarction (23). Finally, a meta-analysis of the association of FTO rs9939609 and CVD risk suggested a significant association that was independent of BMI and other conventional CVD risk factors (24).

However, although the associations of FTO variant with several variables have been observed, and some of the previous studies have also been performed in impressively large study populations, the major limitations of these studies have often been the reliability of the diagnoses, the lack of sufficient follow-up time, or that only men or women are included in the study. Here, we investigated the association of the FTO SNP rs9939609 with cardiovascular events, related deaths, and total mortality in a 19-year OPERA follow-up study. It is a randomly selected, highly characterized study population of 1,045 northern Finnish subjects, including both sexes.

Materials and methods

Study population

OPERA (Oulu Project Elucidating Risk of Atherosclerosis) is a population-based, epidemiological study designed to address the risk factors and disease end-points of atherosclerotic cardiovascular diseases. This study population and selection criteria have been described previously in detail (25). The subjects were randomly selected, middle-aged drug-treated hypertensives and their age- and sex-matched control subjects who were recruited to the OPERA study between the years 1990 and 1993. The participants were interviewed, examined, and tested in our research laboratory. Mortality and hospital events of a total of 1,045 subjects were followed up until 2009. The study was conducted according to the principles of the Declaration of Helsinki and approved by the Ethics Committee of the Faculty of Medicine, University of Oulu. Written informed consent was obtained from each participant.

Clinical measurements

The blood pressure measurements were conducted according to the recommendations of the American Society of Hypertension. All the blood pressure measurements were undertaken with an automatic oscillometric blood pressure recorder (Dinamap, Critikon Ltd, Ascot, UK). The resting blood pressure was measured three times at 1-min intervals from the right arm after the patient had been seated for at least 5 min. The mean value of the second and the third blood pressure measurement from the sitting patient was used. The lifetime smoking burden was calculated as pack-years (1 pack-year = 20 cigarettes smoked/day in 1 year), and the smoking history was obtained from a questionnaire. BMI was calculated as weight (kg) divided by height squared (m²). Height was measured to the nearest centimeter without shoes and weight to the nearest 0.1 kg with the subject wearing only light underwear without shoes. All of the measurements were performed by the same specially trained nurses.

Laboratory analyses

The very-low-density lipoprotein (VLDL) fraction was separated from plasma by ultracentrifugation at 10,500g for 18 h. The plasma HDL-cholesterol concentration was measured by mixing 0.5 mL of the VLDL-free fraction with 25 mL of 2.8% (wt/vol) heparin and 25 mL of 2M manganese chloride and measuring the cholesterol concentration in the supernatant after centrifugation at 1,000g and 4°C for 30 min. The low-density lipoprotein (LDL) cholesterol concentration was calculated by subtracting the cholesterol concentration in HDL from that in the VLDL-free fraction. The oral glucose tolerance test (OGTT) was performed in the morning after a 12-h fast immediately after fasting blood had been drawn. Fasting glucose concentrations were measured with the glucose dehydrogenase method (Diagnostica, Merck, Darmstadt, Germany) and the serum insulin levels with a two-site immunoenzymometric assay (AIA-PACK IRI, Tosoh Corp., Tokyo, Japan). These laboratory analyses have been previously described in detail (25). HsCRP was analyzed using commercially available ELISA kits (Diagnostic Systems Laboratories, Webster, TX, USA) as described before (26).

Fasting plasma leptin concentration was assessed with a commercial double antibody radioimmunoassay (RIA) (Human Leptin RIA Kit; Linco Research, Inc., St. Charles, MO, USA) and fasting plasma total ghrelin concentrations with a commercial peptide radioimmunoassay kit (Phoenix Pharmaceuticals, Belmont, CA, USA) as described earlier (27). Plasma adiponectin concentrations were measured with an enzyme-linked immunosorbent assay (ELISA) devised in our laboratory, described previously in detail (28).

Genotyping

Genotypes for FTO rs9939609 were determined by TaqMan SNP Genotyping Assay (Applied Biosystems, Foster City, CA, USA). The genotyping of FTO rs9939609 was successfully performed for 1037 subjects.

Outcome classification

Information on causes of death and events leading to hospitalization was obtained from the Finnish Causes-of-Death Register and the Hospital Discharge Register. The diagnoses were classified according to the International Classification of Diseases, Eighth Revision (ICD-8) or Ninth Revision (ICD-9) before 1994 and the Tenth Revision (ICD-10) thereafter. CHD was defined as diagnoses I20, I21, I22 (ICD-10) and 410, 4110 (ICD-8/9), coronary artery bypass graft or coronary angioplasty as I20–I25, I46, R96, R98 (ICD-10) and 410–414, 798 (not 7980A) (ICD-8/9) as causes of death. CVD was defined as CHD or stroke that included I61, I63 (not I636), I64 (ICD-10) and 431, 4330A, 4331A, 4339A, 4340A, 4341A, 4349A, 436 (ICD-9) or 431 (excluding 43101, 43191), 433, 434, 436 (ICD-8) according to the FINRISK criteria (29).

Statistical methods

The data were analyzed with IBM SPSS Statistics for Windows, version 19.0. (IBM Corp., Armonk, NY, USA). In order to evaluate the difference in the baseline data or the number of events and deaths during the follow-up between the genotypes the variables were tested with Pearson’s chi-square or ANOVA when appropriate. To normalize the skewed distributions, log-transformed values of BMI, systolic blood pressure (SBP), leptin, adiponectin, CRP, cholesterol levels, as well as of 2-h glucose and fasting insulin concentrations were used in analyses at baseline that required a normal distribution. In the analyses, the variable study group refers to drug-treated hypertensives and their age- and sex-matched control persons. Further analysis was performed with logistic regression to estimate the impact of rs9939609 on the events and death. In the genetic analyses, the subjects with diagnosed myocardial infarction (n = 20) and with stroke diagnosed in the hospital (n = 22) already at baseline were included in the events. Time-to-death analyses were performed with Cox regression, and Kaplan–Meier curves were created. The estimate about the impact of FTO variant on CVD mortality was assessed by using Nagelkerke R² in the logistic regression model. These analyses were carried out with the recessive genetic model. P values less than 0.05 were considered to be statistically significant.

Prognostic significance of the FTO variant in predicting cardiovascular events was assessed by the guidelines recommended for assessing newer risk markers (30).

Results

Baseline data

The allele frequency of the FTO rs9939609 A allele was 40.7% in our study population. The alleles of rs9939609 were in Hardy–Weinberg equilibrium. Allele frequencies and genotype counts are presented in Table I. The only variable differing between the genotypes in the baseline was fasting glucose, with AA homozygotes displaying the highest fasting blood glucose at baseline (p = 0.029) (Table II). The other traditional risk factors did not differ between the genotypes, and measures of obesity, BMI and waist circumference, displayed no statistically significant differences between the genotypes, although a trend toward higher BMI and waist circumference was seen among the AA genotype. Also no differences between the genotypes were observed in the plasma levels of adiponectin, leptin, and ghrelin.

Table I. Allele frequencies, genotype counts, and Hardy–Weinberg equilibrium P value of FTO rs9939609 variant.

Variant	Allele	Frequency	Genotype	Count	HWE p value
rs9939609	A	0.407	AA	183	0.147
			AT	478
	T	0.593	TT	376

HWE = Hardy–Weinberg equilibrium; a non-significant p value (p > 0.05) indicates that the variant does not deviate significantly from HWE.

Table II. Measurements of the risk factors for atherosclerosis and obesity-related peptide hormones at baseline in relation to FTO rs9939609 genotypes.

	AA n = 183	AT n = 478	TT n = 376	P value
Males/females (n)	97/86	234/244	181/195	0.541^a
Hypertensive/control (n)	93/90	241/237	184/192	0.882^a
Age (years)	51 (6)	51 (6)	52 (6)	0.237^b
BMI (kg/m^2)	28.3 (5.1)	27.6 (4.5)	27.5 (4.5)	0.200^b
Waist circumference (cm)	92.0 (13.5)	90.3 (13.2)	90.3 (13.2)	0.349^b
SBP (mmHg)	149 (22)	148 (22)	148 (23)	0.919^b
Plasma triglycerides (mmol/L)	1.7 (1.3)	1.5 (0.9)	1.6 (1.0)	0.393^b
Total cholesterol (mmol/L)	5.8 (1.2)	5.7 (1.0)	5.7 (1.1)	0.567^b
LDL-C (mmol/L)	3.5 (1.0)	3.5 (0.9)	3.6 (1.0)	0.958^b
HDL-C (mmol/L)	1.3 (0.4)	1.3 (0.4)	1.4 (0.4)	0.943^b
Smoking (pack-year)	12.7 (14.6)	11.7 (13.5)	12.1 (14.4)	0.187^b
Fasting glucose (mmol/L)^c	5.0 (1.8)	4.6 (1.1)	4.8 (1.7)	0.029^b
Fasting glucose (mmol/L)^d	4.4 (0.5)	4.4 (0.5)	4.4 (0.5)	0.882^b
Fasting insulin (mU/L)^c	13.5 (10.1)	14.0 (12.8)	13.2 (9.0)	0.302^b
Fasting insulin (mU/L)^d	12.5 (9.8)	13.2 (11.7)	12.3 (7.4)	0.178^b
OGTT-2h glucose (mmol/L)^c	6.6 (3.7)	5.9 (2.6)	5.9 (3.0)	0.035^b
OGTT-2h glucose (mmol/L)^d	5.5 (1.6)	5.4 (1.5)	5.3 (1.4)	0.456^b
CRP (mg/L)	4.1 (6.9)	4.0 (7.7)	3.6 (7.5)	0.598^b
Leptin (ng/mL)	10.9 (8.7)	10.7 (8.0)	10.4 (8.2)	0.378^b
Adiponectin (mg/L)	15.2 (7.0)	15.9 (6.8)	16.2 (6.9)	0.544^b
Ghrelin (pg/mL)	645 (258)	675 (246)	671 (233)	0.459^b
MI at baseline, n (%)	5 (2.7)	8 (1.7)	7 (1.9)	0.671^a
Stroke at baseline, n (%)	3 (1.6)	8 (1.7)	11 (2.9)	0.399^a
Diabetes at baseline, n (%)	30 (16.4)	33 (6.9)	29 (7.7)	0.000^a

Values are means and SD.

^a,bP values are obtained with Pearson’s chi-square^a or ANCOVA^b adjusted for sex, age, and BMI when appropriate. Significant p-values are bolded.

^cAll subjects.

^dDiabetic subjects excluded.

BMI = body mass index; CRP = C-reactive protein; HDL-C = high-density lipoprotein cholesterol; LDL-C = low-density lipoprotein cholesterol; MI = myocardial infarction; OGTT = oral glucose tolerance test; SBP = systolic blood pressure.

There were also 20 subjects with a reliably diagnosed myocardial infarction (MI), 22 subjects with stroke diagnosed in the hospital, and 92 subjects with diabetes already at baseline. These other baseline events were evenly distributed between the genotypes, but a history of diabetes varied statistically significantly, with the prevalence of diabetes being 16.4% among the AA genotype and only 6.9% and 7.7%, respectively, among the AT and TT genotypes (p < 0.001) (Table II).

When the sexes were analyzed separately the baseline results did not differ from the original analyses in which the sexes were included in the same analyses. The only significant finding was that baseline diabetes was statistically significantly more prevalent among the AA genotype in women (AA 16.3%, AT 5.3%, and TT 7.2%; p = 0.005). A similar trend was also seen in men, but it failed to reach statistical significance. The measures of obesity, BMI and waist circumference, displayed no statistically significant differences between the genotypes either in men or women separately.

The association of rs9939609 with CVD and CHD events and mortality

Out of 1037 subjects genotyped for rs9939609, 152 coronary heart disease (CHD) and 212 cardiovascular disease (CVD) events or deaths, including 58 CHD and 69 CVD deaths, occurred during the follow-up period, and a total of 166 subjects out of 1037 subjects died. No differences were seen in the total mortality rates or cancer incidence between the genotypes, whereas AA genotypes displayed significantly higher incidence for both CVD and CHD events and death (p = 0.007 and p = 0.009) as well as CVD and CHD mortality (p = 0.009 and p = 0.032) (Table III). There were no interaction effects of FTO genotype and statin use on CVD outcome (p = 0.100).

Table III. The all-cause mortality and major outcomes in relation to FTO rs9939609.

	AA n = 183	AT n = 478	TT n = 376	p value
All deaths	35 (19.1%)	64 (13.4%)	67 (17.8%)	0.096
CVD event or death	53 (29.0%)	88 (18.4%)	71 (18.9%)	0.007
CVD deaths	21 (11.5%)	23 (4.8%)	25 (6.6%)	0.009
CHD event or death	40 (21.9%)	61 (12.8%)	51 (13.6%)	0.009
CHD deaths	16 (8.7%)	18 (3.8%)	24 (6.4%)	0.032
Cancer event or death	17 (9.3%)	60 (12.6%)	53 (14.1%)	0.273
Cancer death	7 (3.8%)	15 (3.1%)	22 (5.9%)	0.142

Values are numbers of the subjects with event or death (% within the genotype). p values are achieved from Pearson’s chi-square test. Significant p-values are bolded.

CHD = major coronary heart disease; CVD = cardiovascular disease.

Since we observed an increased mortality among the AA genotype, we performed further genetic analyses with a recessive model by comparing the AA genotype against the AT and TT genotypes. The logistic regression analyses of cardiovascular event and death (Table IV) revealed that the AA group indeed had a higher rate of CVD and CHD events or death when adjusted with age, sex, and study group (p = 0.001 and p = 0.003, respectively). The effect of AA genotype on both CVD and CHD event or death remained significant after further adjustment for BMI, smoking status, systolic blood pressure, and LDL-cholesterol (p = 0.002 and p = 0.004). The association of rs9939609 on CVD and CHD events and mortality remained significant when adjusted for baseline diabetes (data not shown).

Table IV. The logistic regression analysis of the total mortality and cardiovascular outcomes in relation to FTO rs9939609.

	Model 1			Model 2
	OR	95% CI	p value	OR	95% CI	p value
Total mortality	1.323	0.863–2.026	0.199	1.273	0.815–1.987	0.289
CVD event or death	1.849	1.265–2.702	0.001	1.844	1.246–2.729	0.002
CHD event or death	1.905	1.250–2.903	0.003	1.895	1.222–2.938	0.004

The analysis is performed as comparing AA genotype to AT and TT genotypes. Model 1 adjusted for sex, age, and study group; model 2 adjusted for sex, age, study group, SBP, BMI, LDL, and smoking. Significant p-values are bolded.

The interaction between the variant and sex on CVD events was significant (p < 0.001). Therefore, we also performed the analyses of CVD event or death as well as CVD mortality separately for men and women. In men, the incidence of CVD event or death was higher among the AA genotype (38.1% compared to AT 23.1% and TT 24.3%; p = 0.013). The CVD mortality was also more common among the AA genotype (16.5% compared to AT 6.8% and TT 11.6%; p = 0.025) in men. Also the logistic regression analysis showed an association between rs9939609 and CVD event or death in men when the model was adjusted for age and study group (OR 2.150, 95% CI 1.334–3.463; p = 0.002), and it remained significant even when adjusted for age, study group, SBP, BMI, LDL, and smoking (OR 2.209, 95% CI 1.345–3.630; p = 0.002).

A similar trend as in men was seen also in women for CVD event or death (AA 15.1%, AT 11.1%, and TT 10.3%) as well as for CVD mortality (AA 5.8%, AT 2.9%, and TT 2.1%), but as the incidence of events was much lower than in men these failed to reach statistical significance. The logistic regression analysis showed no association between rs9939609 and CVD event and death in women as the model was adjusted for age and study group (OR 1.404, 95% CI 0.710–2.777; p = 0.329) or when adjusted for age, study group, SBP, BMI, LDL, and smoking (OR 1.364, 95% CI 0.672–2.766; p = 0.390).

Modeling of the predictability of rs9939609 on cardiovascular mortality

The Kaplan–Meier curves are shown in Figure 1. They demonstrate the association of FTO rs9939609 genotype on CVD (p = 0.006) and CHD mortality (p = 0.048), with the AA genotype displaying lower survival. Mortality and major causes of death were analyzed also with Cox regression which takes the follow-up time into account (Table V). When the model was adjusted for sex, age, and study group, the AA genotype displayed a higher rate of CVD and CHD death (p = 0.006 and p = 0.046). When baseline diabetes was included in the model, the results did not change. We performed also a further modeling for total CVD mortality (Table VI). Again, even when the best model was adjusted for sex, age, SBP, smoking, and diabetes status at baseline, rs9939609 displayed an association with the CVD mortality, suggesting that the variant exerted an independent effect on CVD mortality by AA genotype, displaying approximately 1.7-fold risk for CVD mortality compared to other FTO genotypes. A similar association between CHD mortality and AA genotype (p < 0.01) was also observed after adjustments (data not shown).

Figure 1. The Kaplan–Meier curves of total, CVD, CHD, and cancer mortality. The graph includes both sexes together. The survival differs statistically significantly with respect to both CVD and CHD mortality.

Table V. The Cox regression analysis of the total mortality and major causes of death in relation to FTO rs9939609.

	AA (n = 183)	AT + TT (n = 854)	HR	95% CI	p value
All deaths	35 (19.1%)	131 (15.3%)	1.268	0.873–1.843	0.213
Cause of death
CVD	21 (11.5%)	48 (5.6%)	2.067	1.235–3.459	0.006
CHD	16 (8.7%)	42 (4.9%)	1.801	1.010–3.212	0.046
Cancer	7 (3.8%)	37 (4.3%)	0.895	0.398–2.011	0.788

The model is adjusted for sex, age, and study group. CVD deaths include CHD deaths. Significant p-values are bolded.

Table VI. The Cox regression modeling of factors predicting CVD mortality.

	Crude	Model 1	Model 2	Model 3	Model 4	Model 5
rs9939609	2.031 (1.216–3.392)**	2.004 (1.199–3.350)**	1.954 (1.167–3.270)*	2.039 (1.217–3.416)**	2.066 (1.232–3.465)**	1.715 (1.006–2.922)*
Sex	3.250 (1.848–5.714)**	3.483 (1.975–6.141)**	3.494 (1.979–6.169)**	3.053 (1.722–5.412)**	1.990 (1.103–3.589)*	2.073 (1.133–3.792)*
Age	1.054 (1.012–1.098)*	1.062 (1.020–1.110)**	1.063 (1.019–1.109)**	1.050 (1.006–1.097)*	1.036 (0.991–1.083)	1.023 (0.978–1.071)
Study group	1.618 (0.998–2.624)
BMI	1.070 (1.025–1.118)**		1.067 (1.018–1.118)**	1.035 (0.984–1.089)
SBP	1.025 (1.015–1.036)**			1.019 (1.007–1.030)**	1.022 (1.012–1.032)**	1.018 (1.008–1.029)**
Smoking	1.058 (1.041–1.076)**				1.047 (1.030–1.065)**	1.051 (1.033–1.069)**
LDL-C	1.250 (0.976–1.601)
Diabetes	4.120 (2.403–7.063)**					3.229 (1.809–5.764)**

Values are HR (95% CI). Study group (hypertensive/control), BMI (body mass index), SBP (systolic blood pressure, mmHg), smoking (pack-year), LDL-C (low-density lipoprotein cholesterol, mmol/L), diabetes diagnosed at baseline.

*P < 0.05, **P < 0.01.

As we wanted to assess the impact of FTO variant on model predictability, we performed also a logistic regression analysis with the variables of the best Cox regression model (rs9939609, sex, SBP, smoking, and baseline diabetes status). The effect of FTO variant was estimated using the difference of Nagelkerke R² in the model with rs9939609 included (Nagelkerke R² = 0.214) or absent (Nagelkerke R²= 0.204). Estimated with this method, FTO rs9939609 alone seems to explain approximately 1% of the CVD mortality in our model.

To assess the discrimination of the risk markers, the C-index and Nagelkerke’s binary R² were calculated for the final model which included FTO rs9939609 variant, sex, age, SBP, smoking, and baseline diabetes status. For age and systolic blood pressure, the optimal cut-off value was defined from receiver-operating characteristic analysis as the maximum sum of sensitivity and specificity above the median with sensitivity at least 20% using cardiovascular death as the end-point. Thereafter, C-index and Nagelkerke’s binary R² were calculated after removal of one variable at the time. The integrated discrimination index and net reclassification index (risk levels: 0% – <6%, 6% – <20%, and ≥20%) were also calculated for the final model. When FTO rs9939609 variant was removed from the model a modest impairment in C-index was observed (from 0.715 to 0.709) (Table VII). In reclassification analyses, integrated discrimination index was significant 0.011 (p = 0.010), whereas net reclassification index was not (0.015; p = 0.52) (Table VIII).

Table VII. Discrimination of the risk models in prediction of cardiovascular event.

	Cardiovascular event
	C-index	Binary R^2
Final model	0.715	0.151
Removed variable:
Sex = male	0.676	0.103
Age ≥ 54 years	0.701	0.124
Systolic blood pressure (≥141 mmHg)	0.705	0.136
Current smoker = yes	0.705	0.139
Baseline diabetes status = yes	0.709	0.138
FTO rs9939609 variant	0.709	0.139

Table VIII. Reclassification of cases and non-cases according to cardiovascular event or death with and without FTO rs9939609 variant in the model where variables of best regression model (sex, age, systolic blood pressure, smoking, and baseline diabetes status) were included.

	Predicted risk with FTO variant
Predicted risk without FTO variant	0 – <6%	6 – <20%	≥20%
Cases (CVD event or death, n = 195)
0 – <6%	1	0	0
6 – <20%	0	70	13
≥20%	0	4	107
Non-cases (n = 839)
0 – <6%	92	9	0
6 – <20%	0	446	44
≥20%	0	27	221

Integrated discrimination index = 0.010 (p = 0.011); net reclassification index = 0.015 (p = 0.52).

Discussion and conclusion

The aim of our study was to investigate the impact of FTO rs9939609 variant on cardiovascular events and related death as well as on total mortality in the 19-year follow-up of the OPERA study. The most significant finding of this study is that the FTO rs9939609 predicts CVD with the AA genotype displaying a higher risk of cardiovascular events and death. Furthermore, this phenomenon seems to be independent of the traditional risk factors for atherosclerosis and also independent of the baseline diabetes diagnosis.

Our results show that rs9939609 is also associated with CVD and CHD risk independent of traditional risk factors, and also associated with diabetes or measures of diabetes. The subjects with AA genotype are at increased risk for cardiovascular disease. Our results are consistent with the previous findings of studies suggesting that rs9939609 is associated with cardiovascular disease in diabetic men (11) and in diabetic women (22) as well as with myocardial infarction in patients with type 2 diabetes (23). However, in our study there was no interaction effect of FTO genotype and statin use on CVD outcome as was demonstrated previously (23).

Also an association of rs9939609 with type 2 diabetes diagnoses as well as with some of the blood glucose measurements was observed in this study. The subjects with AA genotype displayed higher blood glucose and elevated incidence of type 2 diabetes at baseline. This is well in line with the previous studies as the association between this FTO variant and type 2 diabetes has been shown in large-scale studies (19–21), and also these identified the A allele as a risk allele.

As our modeling (in Table VI) suggests that FTO rs9939609 predicts cardiovascular mortality independent of baseline diabetes diagnosis, we wanted to confirm this further by replacing the diabetes diagnosis with baseline plasma fasting glucose in model 5. The results remained similar, suggesting that the rs9939609 does actually predict the cardiovascular mortality independently of the measures of diabetes.

The major strength of the present study is that the diagnoses are based on reliably defined disease criteria as they are diagnosed in the hospital. Also the Finnish registers are known to be trustworthy (31). Another major strength of this study is the relatively long follow-up time of 19 years.

When the results were analyzed separately in men and women, the only statistically significant phenomenon seen at the baseline was that rs9939609 associated with diabetes diagnosis in women, with the AA genotype having more diabetes than other genotypes. As the CVD events occurring during the follow-up time were analyzed in both sexes separately, the effect of risk genotype AA was statistically significant only in men. The trend was seen also in women, but it failed to reach statistical significance as the number of events among women was too low. With a larger number of female study subjects the results would have been more reliable concerning the analysis of the sexes separately.

Although rs9939609 has been associated with body weight and adiposity in many studies (10–13), no association to BMI and waist circumference was observed in this study. Previous studies have also demonstrated an association with plasma leptin level (16,17), ghrelin (18), as well as with hsCRP (11) and blood lipids (11,23). However, none of these associations were seen in our study, possibly due to the older age of the study subjects who were middle-aged. It has been shown that FTO SNPs affect body weight with the highest impact in young adulthood, while the effect weakens later in adulthood (32,33). For the same reason also the other associations, such as with IRX3, could be difficult to show.

One possible mechanism through which FTO might affect the cardiovascular risk could involve its suggested amino acid sensing and regulatory effect on mammalian target of rapamycin (mTOR) signaling which is known to regulate autophagy (34). If cells are lacking FTO, their mTORC1 pathway is decreased due to impaired amino acid sensing, and the autophagy increases. Autophagy is a degradation mechanism in a cell that protects from cell death by maintaining energy levels. It is possible that intronic FTO risk alleles affect the expression level of FTO. Moreover, it can affect also nutrient preference, not only the amount of food, as it is highly expressed in brains, at the hypothalamic region (35).

In conclusion, these results suggest that the AA genotype of FTO rs9939609 is associated with a higher risk of cardiovascular events and death and that this is independent of the traditional risk factors for atherosclerosis; the association of AA genotype with cardiovascular disease remained significant after adjustment for baseline diabetes, which suggests that it is not dependent on the diabetogenic effect of the variant.

Acknowledgements

We would like to thank Heidi Häikiö, Saija Kortetjärvi, Liisa Mannermaa, Leena Ukkola, and Kirsi Kvist-Mäkelä for expert technical assistance as well as Elina Malo, MSc, and Pauliina Pisto, MD, for help with the data. Funded by Finnish Foundation for Cardiovascular Research.

Declaration of interest

The authors report no conflicts of interest.