SNP-by-fitness and SNP-by-BMI interactions from seven candidate genes and incident hypertension after 20 years of follow-up: the CARDIA Fitness Study


The association of single nucleotide polymorphisms (SNPs) from seven candidate genes, including genotype-by-baseline fitness and genotype-by-baseline body mass index (BMI) interactions, with incident hypertension over 20 years was investigated in 2663 participants (1301 blacks, 1362 whites) of the Coronary Artery Risk Development in Young Adults Study (CARDIA). Baseline cardiorespiratory fitness was determined from duration of a modified Balke treadmill test. A total of 98 SNPs in blacks and 89 SNPs in whites from seven candidate genes were genotyped. Participants that became hypertensive (295 blacks and 146 whites) had significantly higher blood pressure and BMI (both races), and lower fitness (blacks only) at baseline than those who remained normotensive. Markers at the peroxisome proliferative activated receptor gamma coactivator 1α (PPARGC1A) and bradykinin β2 receptor (BDKRB2) genes were nominally associated with greater risk of hypertension, although one marker each at the BDKRB2 and endothelial nitric oxide synthase-3 (NOS3) genes were nominally associated with lower risk. The association of baseline fitness with risk of hypertension was nominally modified by genotype at markers within the angiotensin converting enzyme, angiotensinogen, BDKRB2 and NOS3 genes in blacks and the BDKRB2, endothelin-1 and PPARGC1A genes in whites. BDKRB2 rs4900318 showed nominal interactions with baseline fitness on the risk of hypertension in both races. The association of baseline BMI with risk of hypertension was nominally modified by GNB3 rs2301339 genotype in whites. None of the above associations were statistically significant after correcting for multiple testing. We found that SNPs in these candidate genes did not modify the association between baseline fitness or BMI and risk of hypertension in CARDIA participants.


It is estimated that approximately 65 million adults in the United States have hypertension, a well-known independent risk factor for cardiovascular disease.1 Hypertension is viewed as a complex trait caused by the interaction of multiple susceptibility genes with environmental and behavioural factors. Risk factors for hypertension include poor diet, physical inactivity, low cardiorespiratory fitness (hereafter referred to as fitness), overweight and obesity, smoking and low education level. Prospective epidemiological studies have shown a lower risk of developing hypertension in physically fit compared with unfit individuals.2, 3, 4, 5 The Coronary Artery Risk Development in Young Adults (CARDIA) Study found that during a 15-year follow-up period, participants with low fitness (⩾sixtieth percentile), and the population attributable risk of developing hypertension because of low fitness was 21%.4 Furthermore, overweight and obesity are associated with increased risk of developing hypertension.6, 7, 8, 9 In CARDIA, race- and sex-specific odds ratios (ORs) for incident hypertension over 10 years of follow-up associated with a one-standard deviation (s.d.) increase in body mass index (BMI) ranged from 1.33 to 1.66 (P0.001).6

However, some individuals will become hypertensive despite being physically fit and/or having a normal body weight, whereas others who are unfit and/or overweight may have blood pressure values in the normal range. For example, although fitness level was the strongest predictor of hypertension risk in the HYPGENE Study, 37% of the subjects in the highest fitness decile became hypertensive and one-third of the subjects in the lowest decile remained normotensive during the follow-up period.10 Similarly, although exercise training and weight loss lower blood pressure on average, there are significant inter-individual differences in blood pressure responses to identical interventions.11 The underlying mechanisms for inter-individual variation in the effects of regular physical activity on blood pressure are still poorly understood, but initial blood pressure level and familial factors are involved.11 The HERITAGE Family Study has shown that the heritabilities of endurance training induced changes in hemodynamic phenotypes after completion of a standardized endurance-training program vary between 20 and 40%,12, 13, 14 and some of the classic hypertension candidate genes affect the training responses.15, 16 Thus, identifying the sources of variation in blood pressure is critical for a better understanding of the pathophysiological processes leading to hypertension and the mechanisms by which higher levels of fitness or normal weight may protect against hypertension.

Several cross-sectional studies have found significant single nucleotide polymorphism (SNP)-by-physical activity and SNP-by-BMI interactions on blood pressure phenotypes using the candidate gene approach.17, 18, 19, 20 However, similar data on the contribution of genetic variation to the anti-hypertensive effect of increased fitness levels over time are scarce, mainly because of the lack of suitable data sets to examine such questions. The CARDIA Study is a longitudinal study of black and white young adults that includes measures of symptom-limited exercise test duration at three time points over 20 years of follow-up. Thus, the CARDIA cohort represents an excellent resource to examine hypotheses regarding the genetic basis of hypertension, while taking fitness and BMI levels into account. The purpose of this study was to examine the association of SNPs, as well as SNP-by-baseline fitness and SNP-by-baseline BMI interactions, in seven candidate genes with risk of incident hypertension after 20 years of follow-up in the CARDIA Fitness Study.


Study population

Details of recruitment, study design and methods of the CARDIA Study have been published elsewhere.21 The initial examination included 5115 black and white men and women aged 18–30 years. Participants were recruited to represent proportionate racial, sex, age and education groups from four United States communities: Birmingham, AL; Chicago, IL; Minneapolis, MN; and Oakland, CA. Six sequential examinations have been conducted from the time of initiation of the study in 1985–1986 through year 20 (2005–2006). Retention rates declined from 90 to 72% across examinations.22 All participants provided a written informed consent, and institutional review boards from each field centre and the coordinating centre approved the study annually.

Genotype data were obtained on 4244 individuals who participated in the baseline examination. To be included in analyses of 20-year incident hypertension, participants had to have measurements of systolic and diastolic blood pressures (SBP and DBP, respectively) from both the baseline and year 20 examinations, be free of hypertension and not be taking any blood pressure medication at baseline, and free of other chronic diseases and not pregnant at both examinations. At baseline, we excluded participants that did not have data for treadmill time or BMI and who reported having or were unsure of their status regarding the following health risks: hypertension/blood pressure medication, heart problems, diabetes, cancer and women who were pregnant (total excluded for these reasons: n=967). To minimize potential classification errors, parents’ race (reported by the participants) had to match that of the participant (n=132 excluded). This resulted in 3145 (1490 blacks and 1655 whites) healthy participants at baseline. Exclusions at year 20 included participants reporting heart problems, cancer, diabetes, HIV, women who were pregnant, and those not sure of their hypertension/blood pressure medication status (total excluded for these reasons: n=482). This resulted in a total cohort of 2663 participants (1301 blacks and 1362 whites) being included in analyses of incident hypertension over 20 years of follow-up. There were small baseline differences between participants included in the analyses and those excluded. Participants included in the analyses were younger (24.6 vs 25.5 years old, P<0.0001), had lower BMI (24.0 vs 25.4 kg m−2, P<0.0001), had lower blood pressure (SBP: 109.1 vs 112.8 mm Hg, P<0.0001); (DBP: 67.5 vs 70.8 mm Hg, P<0.0001), and higher treadmill time (10.04 vs 9.31 min., P<0.0001) at baseline compared with those excluded.

Definition of hypertension

Hypertension was defined at baseline as resting SBP 140 mm Hg or resting DBP 90 mm Hg, or positive responses to the questions, ‘Has a doctor or nurse ever said you had high blood pressure?’ or ‘Have you ever taken medication for high blood pressure?’ Hypertension at the year 20 examination was defined similarly except the last question was phrased, ‘Are you (currently) taking medications for high blood pressure?’ The hypertension end point is based on blood pressure cut points used in the Seventh Report of the Joint National Committee on Detection, Evaluation and Treatment of High Blood Pressure.23 Thus, hypertensive participants are defined as those who were normotensive at baseline but developed hypertension over 20 years of follow-up, whereas normotensive participants are those who remained free of hypertension during the 20-year follow-up period. Of included incident hypertension cases (n=441 total), 348 (78.9%) were confirmed on the basis of hypertension diagnosis and medication, 17 (3.9%) had elevated SBP, 38 (8.6%) had elevated DBP and 38 (8.6%) had both elevated SBP and DBP.

Data collection

The Coordinating Centre and the CARDIA Quality Control Commission monitored quality of data collection. The CARDIA Study staff was centrally trained and certified, and followed standardized protocols across study centres and examinations. Participants were asked to fast for at least 12 h and to not smoke or engage in heavy physical activity for at least 2 h before the examination.

Blood pressure

Blood pressure was measured using a Hawksley random zero sphygmomanometer (WA Baum Company, Copaigue, NY, USA) at baseline and the Omron HEM-907XL digital blood pressure monitor (Omron Healthcare, Kyoto, Japan) at year 20. Each participant sat in a quiet room for 5-min before having three blood pressure measurements taken from the right arm at 1-min intervals by trained and certified technicians. The SBP and DBP were recorded as Phases I and V Korotkoff sounds. The average of the second and third measurements was used for analyses.

Fitness assessment

Symptom-limited graded exercise treadmill testing was performed at baseline according to a modified Balke protocol.24 The test consisted of up to nine 2-min stages of progressively increasing difficulty. Stage 1 was at 3.0 mph and 2% grade, stages 2–6 were at 3.4 mph with grade beginning at 6% and increasing by 4% at each stage, stages 7–8 were at 4.2 mph and 22 and 25% grades respectively, and stage 9 was at 5.6 mph and 25% grade. The first six stages could generally be performed by walking. The exercise test consisted primarily of walking to facilitate performance by those unaccustomed to jogging and to allow for easier replication during future follow-up exams.24 Fitness was defined as the duration of the treadmill test in minutes.

Other measurements

Body weight (light clothing) was measured to the nearest 0.23 kg using a beam balance scale. Height without shoes was measured to the nearest 0.5 cm using a vertically mounted ruler and a metal carpenter's square. BMI was calculated as weight (kg) divided by height squared (m2). Each participant's age, race and sex were self-reported during the recruitment phase and verified during the baseline clinic visit. Standardized questionnaires were used to obtain data on sociodemographic risk factors such as years of education (highest number of years of school completed) and smoking status (current, former and never smoker), as well as diagnosis and treatment of hypertension, diabetes and other conditions.

SNP selection

For this study, we included SNPs from seven established blood pressure and/or fitness candidate genes: angiotensin converting enzyme (ACE), angiotensinogen (AGT), bradykinin β2 receptor (BDKRB2), endothelin-1 (EDN1), guanine nucleotide binding protein beta-3 subunit (GNB3), endothelial nitric oxide synthase-3 (NOS3) and peroxisome proliferative activated receptor gamma coactivator-1α (PPARGC1A).25 TagSNPs within these genes with a minor allele frequency (MAF) of greater than 0.05 were derived using the Haploview Program26 ( based on two sources of SNPs: the Caucasian and Yoruban populations from the International HapMap project.27 The algorithm used for SNP selection was Haploview's implementation of the Broad Institute's Tagger software,28 with the R-squared cutoff for Tagger set to 0.8 and the logarithm of the odds threshold to 2. In addition, Tagger was used in aggressive multi-marker mode. All tagSNPs selected by Tagger for the Caucasian population were included in the SNP panel. TagSNPs that were not in blocks, or only tagged themselves in the Yoruban population were not included. Non-synonymous SNPs with a MAF >0.05 were also included. The final SNP set included 101 SNPs in blacks and 90 SNPs in whites with MAF >0.05 and Hardy–Weinberg equilibrium (P0.0008 in the CARDIA cohort. The selected SNPs capture 50–91% of the common genetic variance reported for the Yoruban population and 60–100% for the Caucasian population according to the International HapMap project Phase 2 release 24 data (Supplementary Table S1). One SNP in EDN1 and two SNPs in PPARGC1A in blacks and one SNP each in NOS3 and PPARGC1A and two SNPs in AGT in whites were genotyped that were not reported in HapMap and were not in linkage disequilibrium (LD) with HapMap tagSNPs, and therefore potentially capture part of the remaining genetic variation for these genes in the Yoruban and Caucasian populations.


Genotyping of the polymorphisms was performed using a two-step approach. SNPs were multiplexed in a reaction using the iPLEX MassARRAY genotyping system (Sequenom, San Diego, CA, USA). SNPs that did not perform well on this platform were then genotyped using TaqMan Pre-Validated SNP assays (Applied Biosystems; Foster City, CA, USA). Details for PCR conditions and primer sequences are available on request. Genotyping was successfully performed in 82% of the original SNP set (n=354 SNPs from 17 genes) of the CARDIA Fitness Study. Replicate samples (n=206) were randomly dispersed throughout the genotyping plate set. Only SNPs that had a minimum concordance of 99% were used for further analyses.

Statistical analysis

All statistical analyses were performed with SAS version 9.1 (SAS Institute, Cary, NC, USA). Means and s.d. were computed for all descriptive characteristics by race. Differences in continuous and categorical variables between incident hypertensives and normotensives by race were assessed using t-tests and χ2-tests, respectively. Hardy–Weinberg equilibrium was tested in the included SNPs of each race by comparing observed genotype frequencies with expected frequencies using the ALLELE procedure in SAS. The pairwise LD among the SNPs was assessed using the ldmax program available in the GOLD software package.29 Logistic regression modelling was used to assess the contribution of age, sex, BMI, fitness, DBP, SBP, smoking, follow-up time and education to the risk of incident hypertension by race. All variables were based on baseline values except for follow-up time. Odds ratios and 95% confidence intervals (CIs) were calculated for each variable in the multivariate models.

Logistic regression modelling was used to test the contributions of SNPs, as well as SNP-by-baseline fitness and SNP-by-baseline BMI interactions to the risk of incident hypertension. In the basic models testing the association of SNPs with hypertension, each SNP was entered individually into the model with baseline values of age, sex, fitness, BMI, DBP and SBP included as covariates. BMI was log transformed because of its skewed distribution. The common allele homozygotes served as the reference group in the basic models. Interactions between SNPs and baseline fitness (dichotomous trait: high/low) and SNPs and baseline BMI (dichotomous trait: high/low) were tested by including interaction terms (SNP*fitness group; SNP*BMI group) in the basic logistic regression models. In the interaction models, baseline fitness and BMI values were dichotomized into low- and high- groups based on race- and sex-specific median cut points. Specifically, low- and high-fitness groups based on race- and sex-specific median cut points of baseline treadmill time (blacks: males 11.5 min, females 7.5 min.; whites: males 12.6 min., females 9.3 min), and low- and high-BMI groups based on race- and sex-specific median cut points of baseline BMI (blacks: males 23.5 kg m−2, females 23.7 kg m−2; whites: males 23.6 kg m−2, females 21.8 kg m−2) were used. Baseline values of age, sex, fitness (BMI interaction model), BMI (fitness interaction model), DBP and SBP were included as covariates in the interaction models. The interaction models tested for the main effect of each SNP, as well as the SNP interaction term (SNP*fitness group or SNP*BMI group) in the total sample stratified by race. Post-hoc analyses of nominally significant interactions were performed using race-specific analyses stratified by baseline fitness or BMI group, with no further adjustment for fitness or BMI. The common allele homozygotes in the low-fitness or low-BMI group served as the reference group (OR=1.0) in the interaction models.

Participants were grouped by minor allele carrier status for SNPs with MAF <10% showing nominally significant (P<0.05) associations with incident hypertension. Minor allele carrier status was also used when the cell size for a genotype-by-hypertension or genotype-by-hypertension-by fitness or BMI group was less than 10. As multiple SNPs were used in the association analyses, we applied a multiple-testing correction proposed by Nyholt.30 Briefly, the method uses spectral decomposition of matrices of pairwise LDs (r) to estimate variance of eigenvalues. The effective number of independent SNPs (Meff) at each gene can be calculated based on the ratio of observed eigenvalue variance (λobs) and its maximum (M): Meff=1+(M−1) (1−(Var λobs/M)). This procedure was performed separately for each gene locus and the effective number of SNPs from each locus was then added over all loci to derive the total number of effective SNPs. The total number of effective SNPs can then be used to adjust the standard α-level (for example, 5%) using a standard Bonferroni correction. As LD between the SNPs differed between black and white participants, the total effective number of SNPs was also different (98 in blacks and 64 in whites). Thus, in our study the corrected threshold for statistical significance was set to P<0.0005 in black participants and P<0.0008 in white participants. Post-hoc power analyses were performed using QUANTO version 1.2.4 (


Table 1 shows the baseline characteristics of participants by race and 20-year incident hypertension status. In black participants, hypertensives (n=295) were significantly older, had higher BMI and resting blood pressures (both SBP and DBP) and were less fit at baseline than normotensives. In whites, a higher proportion of hypertensives were men, whereas a higher proportion of normotensives were women. Furthermore, in white participants, hypertensives (n=146) were significantly heavier and had higher BMI and resting blood pressures (both SBP and DBP) at baseline compared with normotensives. Overall, a higher proportion of subjects with low fitness at baseline were hypertensive (blacks: 26% of low-fit group, whites: 12%) compared with subjects that were in the high-fitness group at baseline (blacks: 20% of high-fit group, whites: 9%), with the difference being significant in blacks only (P=0.01). Similarly, a higher proportion of hypertensives were in the high-BMI compared with low-BMI group at baseline in both races (blacks: 27% of high-BMI group compared with 18% of low-BMI group, P<0.0001; whites: 14% of high-BMI group compared with 8% of low-BMI group, P=0.001).

Table 1 Baseline (1985–86) characteristics, by race and year 20 (2005–06) incident hypertension status, of healthy participants without hypertension: the CARDIA Fitness Study

Results of the multivariate logistic regression models for the risk of incident hypertension by race are shown in Table 2. In both races, baseline SBP showed the strongest association with the risk of hypertension, followed by baseline DBP. Greater baseline fitness level was associated with lower risk of hypertension in both races, whereas no other variables were significantly associated with incident hypertension. The allele frequencies, Hardy–Weinberg equilibrium and the pairwise LD among all included SNPs are summarized by race in Supplementary Tables S2 and S3. The results for the association of SNPs with incident hypertension and SNP-by-baseline fitness and SNP-by-baseline BMI interactions for all SNPs are summarized by race in Supplementary Tables S4 and S5.

Table 2 Multivariate odds ratios (ORs) and 95% confidence intervals (CIs) by race for predictors of 20-year incident hypertension: the CARDIA Fitness Study

Association of SNPs with incident hypertension

The nominally significant associations of SNPs with incident hypertension by race are presented in Table 3. SNPs from four genes (AGT, ACE, EDN1 and GNB3) were not associated with incident hypertension in either race. The minor alleles of three PPARGC1A SNPs in blacks and two PPARGC1A and three BDKRB2 SNPs in whites were nominally associated (P<0.05) with greater risk of incident hypertension (Table 3). In contrast, the minor alleles at the BDKRB2 rs8016905 locus in blacks and the NOS3 rs1808593 locus in whites were nominally associated with a lower risk of incident hypertension. None of these associations were statistically significant when taking into account the race-specific multiple-testing corrected thresholds for statistical significance.

Table 3 Nominally significant main effects of SNPs on risk of 20-year incident hypertension, by race: the CARDIA Fitness Study

SNP-by-fitness interactions and incident hypertension

There was no evidence of SNP-by-fitness interactions on the risk of incident hypertension in either race for SNPs in the GNB3 gene locus. The BDKRB2 rs4900318 marker showed nominally significant interactions with baseline fitness level on the risk of incident hypertension in both black and white participants (Table 4), although the interaction patterns and minor alleles differed by race (Figure 1). In blacks, participants who were highly-fit at baseline and homozygous for the common allele (A/A) of the BDKRB2 rs4900318 marker had a 44% lower risk of incident hypertension, while no association was observed in low-fit participants (Figure 1a). In whites, minor allele homozygotes (A/A) of the BDKRB2 rs4900318 marker had an over threefold greater risk of incident hypertension only in participants that were high-fit at baseline (Figure 1b).

Table 4 Nominally significant results for SNP-by-baseline fitness interactions on risk of incident hypertension, by race: the CARDIA Fitness Study
Figure 1

BDKRB2 rs4900318-by-fitness interactions on the risk of incident hypertension in black (a) and white (b) participants of the CARDIA Study. Genotype distribution by hypertension and fitness group is given below the figure.


One SNP each at the AGT, NOS3 and ACE genes and three SNPs at the BDKRB2 gene showed nominally significant interactions with baseline fitness level on the risk of incident hypertension in black participants (Table 4). High-fit participants carrying the minor allele of the ACE rs4303 marker had a 55% lower risk of incident hypertension, whereas no association was observed in participants that were low-fit at baseline (Supplementary Figure S1). Minor allele homozygotes at the NOS3 rs3918188 marker had an almost twofold greater risk of incident hypertension only in participants low-fit at baseline, whereas no association was observed in high-fit participants (Supplementary Figure S2). Within the BDKRB2 gene, minor allele homozygotes of the rs945039 marker had a 64% lower risk of incident hypertension only in high-fit participants, whereas minor allele homozygotes of the rs4905474 marker had twofold greater risk of incident hypertension only in high-fit participants (Supplementary Figure S3).


Two SNPs each at the BDKRB2 and PPARGC1A genes and the EDN1 rs2071943 marker showed nominally significant interactions with baseline fitness level on the risk of incident hypertension in white participants (Table 4). Minor allele carriers of the PPARGC1A rs2932965 marker had a 55% lower risk of incident hypertension only in high-fit participants, whereas common allele homozygotes of the PPARGC1A rs6838600 marker had a 45% lower risk of incident hypertension only in high-fit participants (Supplementary Figure S4). Additional figures showing nominal SNP-by-fitness interactions can be found in Supplementary Figures S5–S7. None of the SNP-by-fitness interaction terms in either race were statistically significant when taking into account the race-specific multiple-testing corrected thresholds for statistical significance.

SNP-by-BMI interactions and incident hypertension

The SNP-by-BMI interaction term was nominally significant for the GNB3 rs2301339 marker in white participants (P=0.006). In white participants, minor allele carriers of the GNB3 rs2301339 marker showed a 64% lower risk of incident hypertension compared with common allele homozygotes in the low-baseline BMI group, whereas genotype was not associated with incident hypertension risk among high-baseline BMI subjects (Supplementary Figure S8). However, none of the SNP-by-BMI interaction terms in either race were statistically significant when taking into account the race-specific multiple-testing corrected thresholds for statistical significance.


We found no statistically significant evidence of the main effects of almost 100 SNPs from seven blood pressure candidate genes on incident hypertension, although taking into account baseline values of age, sex, BMI, fitness and blood pressures (both DBP and SBP). Furthermore, the panel of SNPs tested did not modify the association between the risk of incident hypertension over 20 years and baseline fitness or BMI. The candidate gene approach of this study has been previously used in numerous studies of hypertension with inconsistent results.32, 33 A possible explanation for the lack of success of this approach may be related to an incorrect a priori assumption that predisposition to hypertension is driven by classical blood pressure regulating genes such as AGT or EDN1.34 As such, genome-wide association studies (GWASs) have become the most common approach to identify susceptibility genes for common diseases, as GWASs are unbiased by prior assumptions about the DNA alterations responsible for phenotypic variation. The results from recent GWASs for other complex diseases suggest that truly causative, common, consistently associated variants may not have an obvious physiological relationship to the common disease.

The evidence from GWASs for the existence of genetic susceptibility variants for hypertension remains also weak and inconsistent, and the early published GWASs for hypertension did not even find any genetic variant significantly associated with hypertension at the genome-wide level.35, 36 Furthermore, none of the SNPs previously identified in candidate gene studies showed evidence of an association with hypertension in these early GWASs. These results may be caused by several factors. It is possible that hypertension may have fewer common risk alleles with large effect sizes compared with other common diseases and that the true hypertension susceptibility variants may not be represented on the chips used for GWAS genotyping.34 Furthermore, hypertension may be more susceptible than other complex phenotypes to misclassification bias because of the presence of latent hypertension in the control population.37 Phenotyping error along with population stratification, insufficient statistical power, lack of consideration of LD in the human genome and imprecise selection of genetic markers are also possible explanations for the failure of the candidate gene approach.34

A better understanding of the interactions between environmental and genetic factors constitutes a key issue in the understanding of the pathogenesis of hypertension.38 The evidence suggests that most of the susceptibility genes for common diseases, such as hypertension, do not have a strong primary aetiological role in predisposition to disease, but rather act as response modifiers to exogenous environmental factors.38 To be sure, the context dependency of the genetic risk is evident in previous studies of hypertension. A genetic marker may have a modest effect on risk in individuals who maintain a low environmental risk, but a major effect in a high-risk environment.39 The ability to detect gene–environment interactions is highly dependent not only on the magnitude of effect and sample size, but also on the precision with which the outcome and environmental exposures are measured.40, 41 The detection of SNP-by-fitness and SNP-by-BMI interactions in this study was aided by the objective assessment of both baseline fitness and BMI, which should help keep our potential phenotypic error low. The CARDIA Study uses a standardized data collection strategy, which supported the prospective verification of hypertension cases and controls over 20 years of follow-up. In addition, we used a strict classification of hypertension along with stringent inclusion criteria, including minimizing potential racial admixture bias and controlling for baseline values of possible effect modifiers, to eliminate participants with prevalent disease and include only those participants that were apparently healthy and non-hypertensive at baseline.

However, there were several limitations of this study. The number of incident hypertension cases in both races was somewhat limited, which resulted in a modest power to detect associations between hypertension and common SNPs, especially in white participants. Under an additive model, the genotype effect sizes detectable at 80% power on incident hypertension had ORs ranging from 1.51 to 1.88 in blacks and 1.70 to 2.14 in whites with MAF ranging from 0.10 to 0.50 using the multiple-testing corrected thresholds (P=0.0005 in blacks and P=0.0008 in whites). These ORs are considerably higher than would be realistically expected (OR=1.1–1.3) for common SNPs and diseases based on recent GWAS reports. For example, the number of cases that would be needed to detect OR=1.2 with 80% power (a priori) is at least five times greater than in our study.

All of the nominally significant SNPs found in this study, except for the ACE rs4303 marker, were located in non-coding regions and the mechanism of their association with incident hypertension and/or interaction with fitness or BMI is unknown. Additionally, the sequence variants nominally associated with incident hypertension risk were different in blacks and whites. There are several possibilities for the observed difference. A partially different panel of markers was used in each race, as blacks and whites differed in MAF values. As exhibited in this study, specific SNPs may have differing contributions and may not capture the same degree of information in each ethnic group. Differences in baseline values of variables, such as blood pressure, BMI and fitness, may have altered the physiological pathways leading to 20-year incident hypertension in each ethnic group. Significantly fewer white participants became hypertensive over 20 years of follow-up compared with black participants. Furthermore, the association of baseline fitness and incident hypertension was not as strong in whites as it was in blacks. As we stratified baseline fitness and BMI levels into dichotomous variables using median cut points, we may have missed potentially significant interactions that may be uncovered only at the high or low ends of the fitness and BMI distributions.


We found that the tested SNPs in selected candidate genes did not modify the association between the risk of hypertension over 20 years with baseline fitness or baseline BMI in CARDIA participants. Despite these negative results, there is a need for additional genetic association studies of hypertension that incorporate behavioural and physiological traits to better understand whether there are interactions between DNA sequence variation and lifestyle factors affecting blood pressure. However, even if such genetic and environmental interactions were shown in cohort studies, they would need to be tested in clinical trials before being used to revise treatment guidelines.42, 43 Evidence from clinical trials could help identify individuals whose blood pressures are favourably responsive to increased cardiorespiratory fitness. In ‘responsive’ individuals, support might be offered to achieve a higher level of fitness. Whereas in ‘non-responsive’ individuals, increased physical activity levels should nonetheless be promoted given the other health benefits of exercise,44 but other strategies for blood pressure reduction would have to be used to ensure optimal risk reduction.


  1. 1

    Fields LE, Burt VL, Cutler JA, Hughes J, Roccella EJ, Sorlie P . The burden of adult hypertension in the United States 1999 to 2000: a rising tide. Hypertension 2004; 44: 398–404.

    CAS  Article  Google Scholar 

  2. 2

    Barlow CE, LaMonte MJ, FitzGerald SJ, Kampert JB, Perrin JL, Blair SN . Cardiorespiratory fitness is an independent predictor of hypertension incidence among initially normotensive healthy women. Am J Epidemiol 2006; 163: 142–150.

    Article  Google Scholar 

  3. 3

    Blair SN, Goodyear NN, Gibbons LW, Cooper KH . Physical fitness and incidence of hypertension in healthy normotensive men and women. JAMA 1984; 252: 487–490.

    CAS  Article  Google Scholar 

  4. 4

    Carnethon MR, Gidding SS, Nehgme R, Sidney S, Jacobs Jr DR, Liu K . Cardiorespiratory fitness in young adulthood and the development of cardiovascular disease risk factors. JAMA 2003; 290: 3092–3100.

    CAS  Article  Google Scholar 

  5. 5

    Sawada S, Tanaka H, Funakoshi M, Shindo M, Kono S, Ishiko T . Five year prospective study on blood pressure and maximal oxygen uptake. Clin Exp Pharmacol Physiol 1993; 20: 483–487.

    CAS  Article  Google Scholar 

  6. 6

    Dyer AR, Liu K, Walsh M, Kiefe C, Jacobs Jr DR, Bild DE . Ten-year incidence of elevated blood pressure and its predictors: the CARDIA study. Coronary artery risk development in (young) adults. J Hum Hypertens 1999; 13: 13–21.

    CAS  Article  Google Scholar 

  7. 7

    Gelber RP, Gaziano JM, Manson JE, Buring JE, Sesso HD . A prospective study of body mass index and the risk of developing hypertension in men. Am J Hypertens 2007; 20: 370–377.

    Article  Google Scholar 

  8. 8

    Hu G, Barengo NC, Tuomilehto J, Lakka TA, Nissinen A, Jousilahti P . Relationship of physical activity and body mass index to the risk of hypertension: a prospective study in Finland. Hypertension 2004; 43: 25–30.

    CAS  Article  Google Scholar 

  9. 9

    Wilson PW, D’Agostino RB, Sullivan L, Parise H, Kannel WB . Overweight and obesity as determinants of cardiovascular risk: the Framingham experience. Arch Intern Med 2002; 162: 1867–1872.

    Article  Google Scholar 

  10. 10

    Rankinen T, Church TS, Rice T, Bouchard C, Blair SN . Cardiorespiratory fitness, BMI and risk of hypertension: the HYPGENE study. Med Sci Sports Exerc 2007; 39: 1687–1692.

    Article  Google Scholar 

  11. 11

    Bouchard C, Rankinen T . Individual differences in response to regular physical activity. Med Sci Sports Exerc 2001; 33: S446–S451.

    CAS  Article  Google Scholar 

  12. 12

    An P, Perusse L, Rankinen T, Borecki IB, Gagnon J, Leon AS et al. Familial aggregation of exercise heart rate and blood pressure in response to 20 weeks of endurance training: the HERITAGE family study. Int J Sports Med 2003; 24: 57–62.

    CAS  Article  Google Scholar 

  13. 13

    An P, Rice T, Gagnon J, Leon AS, Skinner JS, Bouchard C et al. Familial aggregation of stroke volume and cardiac output during submaximal exercise: the HERITAGE Family Study. Int J Sports Med 2000; 21: 566–572.

    CAS  Article  Google Scholar 

  14. 14

    Rice T, An P, Gagnon J, Leon AS, Skinner JS, Wilmore JH et al. Heritability of HR and BP response to exercise training in the HERITAGE family Study. Med Sci Sports Exerc 2002; 34: 972–979.

    Article  Google Scholar 

  15. 15

    Rankinen T, Gagnon J, Perusse L, Chagnon YC, Rice T, Leon AS et al. AGT M235T and ACE ID polymorphisms and exercise blood pressure in the HERITAGE Family Study. Am J Physiol Heart Circ Physiol 2000; 279: H368–H374.

    CAS  Article  Google Scholar 

  16. 16

    Rankinen T, Rice T, Perusse L, Chagnon YC, Gagnon J, Leon AS et al. NOS3 Glu298Asp genotype and blood pressure response to endurance training: the HERITAGE family study. Hypertension 2000; 36: 885–889.

    CAS  Article  Google Scholar 

  17. 17

    Brito EC, Vimaleswaran KS, Brage S, Andersen LB, Sardinha LB, Wareham NJ et al. PPARGC1A sequence variation and cardiovascular risk-factor levels: a study of the main genetic effects and gene × environment interactions in children from the European Youth Heart Study. Diabetologia 2009; 52: 609–613.

    CAS  Article  Google Scholar 

  18. 18

    Grove ML, Morrison A, Folsom AR, Boerwinkle E, Hoelscher DM, Bray MS . Gene–environment interaction and the GNB3 gene in the atherosclerosis risk in communities study. Int J Obes (Lond) 2007; 31: 919–926.

    CAS  Article  Google Scholar 

  19. 19

    Kimura T, Yokoyama T, Matsumura Y, Yoshiike N, Date C, Muramatsu M et al. NOS3 genotype-dependent correlation between blood pressure and physical activity. Hypertension 2003; 41: 355–360.

    CAS  Article  Google Scholar 

  20. 20

    Vimaleswaran KS, Franks PW, Barroso I, Brage S, Ekelund U, Wareham NJ et al. Habitual energy expenditure modifies the association between NOS3 gene polymorphisms and blood pressure. Am J Hypertens 2008; 21: 297–302.

    CAS  Article  Google Scholar 

  21. 21

    Friedman GD, Cutter GR, Donahue RP, Hughes GH, Hulley SB, Jacobs Jr DR et al. CARDIA: study design, recruitment, and some characteristics of the examined subjects. J Clin Epidemiol 1988; 41: 1105–1116.

    CAS  Article  Google Scholar 

  22. 22

    Carnethon MR, Sternfeld B, Schreiner PJ, Jacobs Jr DR, Lewis CE, Liu K et al. Association of 20-year changes in cardiorespiratory fitness with incident type 2 diabetes: the coronary artery risk development in young adults (CARDIA) fitness study. Diabetes Care 2009; 32: 1284–1288.

    Article  Google Scholar 

  23. 23

    Chobanian AV, Bakris GL, Black HR, Cushman WC, Green LA, Izzo Jr JL et al. The Seventh Report of the Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure: the JNC 7 report. Jama 2003; 289: 2560–2572.

    CAS  Article  Google Scholar 

  24. 24

    Sidney S, Haskell WL, Crow R, Sternfeld B, Oberman A, Armstrong MA et al. Symptom-limited graded treadmill exercise testing in young adults in the CARDIA study. Med Sci Sports Exerc 1992; 24: 177–183.

    CAS  Article  Google Scholar 

  25. 25

    Bray MS, Hagberg JM, Perusse L, Rankinen T, Roth SM, Wolfarth B et al. The human gene map for performance and health-related fitness phenotypes: the 2006-2007 update. Med Sci Sports Exerc 2009; 41: 35–73.

    Article  Google Scholar 

  26. 26

    Barrett JC, Fry B, Maller J, Daly MJ . Haploview: analysis and visualization of LD and haplotype maps. Bioinformatics 2005; 21: 263–265.

    CAS  Article  Google Scholar 

  27. 27

    International HapMap Consortium. The International HapMap Project. Nature 2003; 426: 789–796.

    Article  Google Scholar 

  28. 28

    de Bakker PI, Yelensky R, Pe’er I, Gabriel SB, Daly MJ, Altshuler D . Efficiency and power in genetic association studies. Nat Genet 2005; 37: 1217–1223.

    CAS  Article  Google Scholar 

  29. 29

    Abecasis GR, Cookson WOC . GOLD--graphical overview of linkage disequilibrium. Bioinformatics 2000; 16: 182–183.

    CAS  Article  Google Scholar 

  30. 30

    Nyholt DR . A simple correction for multiple testing for single-nucleotide polymorphisms in linkage disequilibrium with each other. Am J Hum Genet 2004; 74: 765–769.

    CAS  Article  Google Scholar 

  31. 31

    Gauderman WJ, Morrison JM . QUANTO 1.1: A computer program for power and sample size calculations for genetic-epidemiology studies. 2006; Available from:

  32. 32

    Agarwal A, Williams GH, Fisher ND . Genetics of human hypertension. Trends Endocrinol Metab 2005; 16: 127–133.

    CAS  Article  Google Scholar 

  33. 33

    Marteau JB, Zaiou M, Siest G, Visvikis-Siest S . Genetic determinants of blood pressure regulation. J Hypertens 2005; 23: 2127–2143.

    CAS  Article  Google Scholar 

  34. 34

    Wellcome Trust Case Control Consortium. Genome-wide association study of 14 000 cases of seven common diseases and 3000 shared controls. Nature 2007; 447: 661–678.

    Article  Google Scholar 

  35. 35

    Charchar F, Zimmerli L, Tomaszewski M . The pressure of finding human hypertension genes: new tools, old dilemmas. J Hum Hypertens 2008; 22: 821–828.

    CAS  Article  Google Scholar 

  36. 36

    Levy D, Larson MG, Benjamin EJ, Newton-Cheh C, Wang TJ, Hwang SJ et al. Framingham Heart Study 100K project: genome-wide associations for blood pressure and arterial stiffness. BMC Med Genet 2007; 8: S3.

    Article  Google Scholar 

  37. 37

    McCarthy MI, Abecasis GR, Cardon LR, Goldstein DB, Little J, Ioannidis JP et al. Genome-wide association studies for complex traits: consensus, uncertainty and challenges. Nat Rev Genet 2008; 9: 356–369.

    CAS  Article  Google Scholar 

  38. 38

    Tiret L . Gene-environment interaction: a central concept in multifactorial diseases. Proc Nutr Soc 2002; 61: 457–463.

    Article  Google Scholar 

  39. 39

    Talmud PJ . How to identify gene-environment interactions in a multifactorial disease: CHD as an example. Proc Nutr Soc 2004; 63: 5–10.

    CAS  Article  Google Scholar 

  40. 40

    Luan JA, Wong MY, Day NE, Wareham NJ . Sample size determination for studies of gene-environment interaction. Int J Epidemiol 2001; 30: 1035–1040.

    CAS  Article  Google Scholar 

  41. 41

    Wong MY, Day NE, Luan JA, Chan KP, Wareham NJ . The detection of gene-environment interaction for continuous traits: should we deal with measurement error by bigger studies or better measurement? Int J Epidemiol 2003; 32: 51–57.

    CAS  Article  Google Scholar 

  42. 42

    Goodman SN, Berlin JA . The use of predicted confidence intervals when planning experiments and the misuse of power when interpreting results. Ann Intern Med 1994; 121: 200–206.

    CAS  Article  Google Scholar 

  43. 43

    Franks PW . Identifying genes for primary hypertension: methodological limitations and gene-environment interactions. J Hum Hypertens 2009; 23: 227–237.

    CAS  Article  Google Scholar 

  44. 44

    Haskell WL, Lee IM, Pate RR, Powell KE, Blair SN, Franklin BA . Physical Activity and Public Health: updated recommendation for adults from the American College of Sports Medicine the American Heart Association. Circulation 2007; 116: 1081–1093.

    Article  Google Scholar 

Download references


The Coronary Artery Risk Development in Young Adults study is supported by the National Heart, Lung,and Blood Institute (grants N01-HC-48047, N01-HC-48048, N01-HC-48049, N01-HC-48050 and N01-HC-95095). The CARDIA Fitness study is supported by R01 HL 078972.

Author information



Corresponding author

Correspondence to T Rankinen.

Ethics declarations

Competing interests

The authors declare no conflict of interest.

Additional information

Supplementary Information accompanies the paper on the Journal of Human Hypertension website

Supplementary information

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Sarzynski, M., Rankinen, T., Sternfeld, B. et al. SNP-by-fitness and SNP-by-BMI interactions from seven candidate genes and incident hypertension after 20 years of follow-up: the CARDIA Fitness Study. J Hum Hypertens 25, 509–518 (2011).

Download citation


  • cardiorespiratory fitness
  • gene–environment interaction
  • blood pressure
  • genotype

Further reading


Quick links