Erythrocyte Fatty Acid Composition Is Associated with the Risk of Hypertension in Middle-Aged and Older Women

Abstract

Experimental studies have suggested different effects of various fats on blood pressure. However, epidemiologic evidence of these relations remains limited and inconsistent. We therefore assessed the association of fatty acid (FA) composition in erythrocyte membranes with the risk of hypertension. We selected 516 cases of incident hypertension and 516 matched controls during 12.9 y of follow-up in the Women’s Health Study. Erythrocyte FA was measured in baseline bloods using GC. After controlling matching factors and lifestyle factors, erythrocyte SFA showed a positive association, whereas total cis PUFA, cis (n-3) PUFA, and the ratio of PUFA:SFA (PS ratio) showed an inverse association with the risk of hypertension. The multivariable RR of hypertension across the increasing quartiles of erythrocyte FA subtypes were 1.00, 1.19, 1.44, and 1.76 for total SFA; 1.00, 0.84, 0.88, and 0.56 for total cis PUFA; 1.00, 0.87, 0.66, and 0.65 for cis (n-3) PUFA; and 1.00, 0.99, 0.70, and 0.51 for the PS ratio. After further adjusting for obesity-related metabolic factors, these associations were attenuated and remained significant only for the PS ratio. cis MUFA, cis (n-6) PUFA, and trans unsaturated FA in erythrocyte membranes were not associated with the risk of hypertension. Our study showed that FA composition in erythrocyte membranes is associated with the risk of hypertension in middle-aged and older women. However, after controlling for obesity-related metabolic factors, the associations remained significant only for the PS ratio.

Introduction

Hypertension is the most common chronic disease in the U.S. (1), affecting >74.5 million adults in 2003–2006 (2). Dietary fat is an important modifiable risk factor for hypertension. Animal studies have shown that diets high in saturated fat increase BP⁹ (3, 4), whereas diets enriched with (n-3) polyunsaturated fat protect against BP elevations (5–7). The current dietary recommendations for Americans to prevent and treat hypertension include a low intake of total and saturated fat (8, 9). However, epidemiologic evidence on the association between various subtype and individual FA intake and the risk of developing hypertension remains limited and inconsistent (10, 11).

The extent to which nutritional epidemiology can elucidate the relation between diet and disease outcomes depends on the accuracy and reliability of diet assessment. To provide more objective assessment of dietary fat, FA composition in plasma lipid fractions and erythrocyte membranes has been measured in epidemiologic studies. Two previous studies (12, 13) that investigated the association between plasma FA and risk of hypertension have found that a higher proportion of SFA and a lower proportion of PUFA in baseline plasma lipids were significantly associated with an increased risk of hypertension. To our knowledge, no prior study has examined the FA composition in erythrocyte membranes and risk of hypertension.

We recently found in a large prospective cohort of middle-aged and older U.S. women that baseline intake of SFA, MUFA, and trans FA assessed from FFQ was each positively associated with the risk of incident hypertension during a mean of 12.9 y of follow-up, but only the association for trans FA intake remained significant after adjustment for obesity-related factors (14). To further investigate the relation between FA intake and risk of hypertension, we examined FA composition in erythrocyte membranes in a nested case-control study of hypertension within this prospective cohort. We hypothesized that higher proportions of SFA and trans FA and a lower proportion of PUFA and PS ratio in erythrocyte membranes are associated with an increased risk of hypertension among initially normotensive women.

Participants and Methods

Study population.

The WHS is a randomized, double-blind, placebo-controlled, 2 × 2 factorial trial evaluating the risks and benefits of low-dose aspirin and vitamin E in the primary prevention of CVD and cancer (15, 16). A 3rd component, β-carotene, was initially included in the trial but was terminated after a median treatment of 2.1 y (17). From 1992 to 1995, 39,876 female U.S. health professionals, aged ≥39 y and free from CVD and cancer (except nonmelanoma skin cancer), were randomized into the WHS. Baseline blood samples were collected from 28,345 participants and stored in liquid nitrogen freezers. Written informed consent was obtained from all participants. The trial and ongoing cohort follow-up was approved by the institutional review board of Brigham and Women’s Hospital, Boston, MA.

Using a prospective, nested, case-control study design, we randomly selected 516 incident hypertension cases who had provided sufficient baseline blood samples. Hypertension at baseline was defined by meeting any of the following 4 criteria: a physician’s diagnosis of hypertension, systolic BP ≥ 140 mm Hg, diastolic BP ≥ 90 mm Hg, or use of antihypertensive treatment. Incident hypertension was identified as women who had no hypertension at baseline but reported newly developed hypertension during follow-up on the basis of a new physician’s diagnosis of hypertension, elevated systolic BP ≥ 140 mm Hg or diastolic BP ≥ 90 mm Hg, or newly initiated antihypertensive treatment. Individuals who developed CVD, for which the management may affect BP levels, after baseline but before the development of hypertension were censored on the date of CVD diagnosis. On the baseline and follow-up questionnaires, participants reported systolic BP in 1 of 9 ordinal categories from <110 to ≥180 mm Hg and diastolic BP in 1 of 7 ordinal categories from <65 to ≥105 mm Hg. In similar cohorts of health professionals, self-reported BP highly correlated with measured systolic (r = 0.72) and diastolic (r = 0.60) BP (18) and the validity of self-reported hypertension was high (11, 19).

For each selected case of hypertension, one control participant was randomly chosen from women who provided baseline blood samples and remained free of hypertension until the case was identified. Each case and the respective control was matched on age (±1 y) and follow-up time (±3 mo). When we selected the 516 case-control pairs, we first included 388 case-control pairs previously selected, of whom 345 pairs were also matched on race/ethnicity (344 pairs of white and 1 pair of African American participants). We then added 128 case-control pairs by oversampling African American (64 pairs) and Asian/Pacific Islander (64 pairs) participants to increase the ethnic diversity of the current study.

Blood assays of erythrocyte FA composition.

Baseline blood samples of all cases and controls were handled identically throughout sample collection, long-term storage, sample retrieval, and assays. All investigators and laboratory personnel were unaware of the participants’ case-control status. The FA profile was measured at the Department of Laboratory Medicine and Pathology, University of Minnesota, using the method by Cao et al. (20). FA in erythrocyte membranes were extracted with a mixture of chloroform and methanol (2:1, v:v), dissolved in heptane, and injected onto a capillary Varian CP7420 100-m column with a Hewlett Packard 5890 gas chromatograph equipped with a HP6890A autosampler. The gas chromatograph was configured for a single capillary column with a flame ionization detector and interfaced with HP chemstation software. Adequate separation of FAME was obtained over a 50-min period with an initial temperature of 190°C followed by the subsequent temperature gradually increased to 240°C. FA from 12:0 through 24:1(n-9) were separated, identified, and expressed as percent of total FA. FA subtypes, including SFA, MUFA, PUFA, and trans FA, were each calculated as the sum of the respective individual FA. The PS ratio and ratio of (n-6):(n-3) PUFA were also calculated. The CV on 51 blind triplicates from 17 individual samples were 2.1% for SFA, 3.0% for MUFA, 3.3% for PUFA, and 3.6% for trans FA.

Other baseline covariates.

On the baseline questionnaire, women provided self-reports of age, weight and height, smoking status, alcohol use, recreational exercise, menopausal status, and postmenopausal hormone use. History of diabetes was defined by a physician’s diagnosis. History of hypercholesterolemia was defined as having a physician’s diagnosis, self-reported cholesterol level ≥ 240 mg/dL (6.2 mmol/L), or current treatment for high cholesterol. BMI (in kg/m²) was calculated from self-reported height and weight. Diet was assessed from a 131-item, validated, semiquantitative FFQ (21). Nutrient intakes including FA were computed according to food composition tables from the USDA (22). Values for the trans isomer content of foods were estimated based on the analyses by Enig et al. (23) and Slover et al. (24), which determined the FA component of selected food items using capillary chromatography. In addition, previously completed blood assays, including total cholesterol, LDL cholesterol, HDL cholesterol (Roche Diagnostics), and hs-CRP (Denka Seiken) (25) were also available for analyses.

Data analyses.

Statistical analyses were performed using SAS software version 9.1 (SAS Institute). All statistical tests were 2-sided and P < 0.05 was considered significant. We first assessed the correlation of FA composition in erythrocyte membranes with dietary FA and known hypertension risk factors. We then compared cases and controls on major hypertension risk factors, using paired t tests for means and McNemar’s tests for proportions. The composition of FA in erythrocyte membranes was also compared between cases and controls as well as between whites and non-whites. We divided each subtype and individual FA into quartiles based on its distribution in 516 controls and performed conditional logistic regression to calculate RR and 95% CI for incident hypertension according to quartiles of FA, with the lowest quartile serving as the reference. The values presented in the text and tables are RR (95% CI). Initial models only controlled for matching factors including age and race. Multivariable models then adjusted for lifestyle and dietary factors, including smoking, exercise, alcohol use, menopausal status, postmenopausal hormone use, and total energy intake (multivariable model 1). Because dietary fat influences FA in erythrocyte membranes as well as in adipose tissues, obesity and related abnormalities in lipid and glucose metabolism could be potential mediating factors in the causal pathway between FA and hypertension. We therefore further examined the impact of adjustment for BMI and history of diabetes and history of hypercholesterolemia on the RR (multivariable model 2). Linear trends across increasing quartiles were tested using the median value of each quartile as an ordinal variable. In secondary analyses, we stratified all analyses by baseline age [<55 vs. ≥55 y, cutpoint selected based on previous findings in the WHS cohort (14)], race (whites vs. non-whites), BMI (<25 vs. ≥25 kg/m²), and systolic/diastolic BP (<120/80 vs. ≥120/80 mm Hg), with the interactions tested using Wald chi-square tests. We also repeated the analyses for each individual FA separately.

Results

Among 516 control women, the correlations between erythrocyte FA and dietary FA were moderately strong for cis (n-3) PUFA (Spearman r = 0.26; P < 0.0001) and trans FA (r = 0.39; P < 0.0001) but weak and nonsignificant for total SFA (r = 0.01; P = 0.76), total cis MUFA (r = −0.008; P = 0.87), total cis PUFA (r = 0.02; P = 0.63), and cis (n-6) PUFA (r = 0.05; P = 0.27). Erythrocyte total SFA was positively correlated with BMI (r = 0.10; P = 0.03) and hs-CRP concentrations (r = 0.17; P < 0.0001), whereas total cis PUFA (r = −0.18; P < 0.0001 and r = −0.13; P = 0.002, respectively) and the PS ratio (r = −0.16; P = 0.0004 and r = −0.18; P < 0.0001, respectively) were inversely correlated with BMI and hs-CRP.

Compared with controls, incident hypertension cases had higher baseline BMI, higher baseline levels of hs-CRP, lower levels of HDL cholesterol, and were more likely to have a history of diabetes and hypercholesterolemia (Table 1). Smoking status, alcohol use, exercise, and blood concentrations of total and LDL cholesterol did not significantly differ between cases and controls. As expected, the baseline systolic and diastolic BP was higher in hypertension cases than in controls. Comparing erythrocyte FA between cases and controls, the proportions of total cis PUFA, cis (n-3) PUFA, 22:5(n-3), and 22:6(n-3) were significantly lower in cases, the proportions of 16:0, cis 16:1 and 20:3(n-6) were significantly higher in cases, whereas total SFA, total cis MUFA, and total and individual trans FA did not differ between cases and respective controls (Table 2). When erythrocyte FA were compared between white and nonwhite women, the proportions of total cis MUFA and total trans FA were significantly higher, whereas the proportion of total cis (n-3) PUFA was significantly lower in whites than in nonwhites, regardless of case-control status. The proportion of total cis (n-6) PUFA was higher in white cases than in nonwhite cases but similar in white and nonwhite controls.

TABLE 1.

 Baseline characteristics of 516 women who developed incident hypertension (cases) and 516 women who remained free of hypertension (controls) in a nested case-control study within the Women's Health Study (WHS)¹²

Characteristics	Cases (n = 516)	Controls (n = 516)	P3
Age,4y	54.0 ± 6.3	54.0 ± 6.3	—
BMI, kg/m2	26.5 ± 4.8	24.5 ± 4.0	<0.0001
Race,4%			—
White	70.9	72.3
Black	14.5	13.7
Asian	13.0	13.5
Others	1.6	0.6
Smoking, %			0.87
Current	11.6	11.4
Past	30.8	33.1
Never	57.6	55.4
Alcohol, %			0.42
Rarely/never	46.5	42.4
10–45 g ethanol/mo	15.1	13.8
10–90 g ethanol/wk	30.8	34.5
≥13 g ethanol/d	7.6	9.3
Exercise, %			0.31
Rarely/never	37.2	32.6
<1 time/wk	17.8	17.6
1–3 times/wk	33.3	38.8
≥ 4 times/wk	11.6	11.1
Postmenopausal, %	54.2	52.4	0.81
Current postmenopausal hormone use, %	42.9	43.7	0.29
History of diabetes, %	2.5	0.9	0.03
History of hypercholesterolemia, %	28.1	22.1	0.03
Plasma analytes
hs-CRP,5mg/L	2.2 (0.9; 4.4)	1.3 (0.5; 3.3)	<0.0001
Total cholesterol,6mg/dL	209.1 ± 40.0	207.8 ± 36.4	0.59
LDL-cholesterol,6mg/dL	123.2 ± 33.1	120.8 ± 30.7	0.23
HDL-cholesterol,6mg/dL	52.3 ± 13.1	55.9 ± 15.3	<0.0001
Systolic BP, %			<0.0001
<110 mm Hg	10.0	25.4
110–119 mm Hg	29.1	45.1
120–129 mm Hg	41.5	23.9
130–139 mm Hg	19.3	5.6
Diastolic BP, %			<0.0001
<65 mm Hg	4.3	15.2
65–74 mm Hg	31.8	48.7
75–84 mm Hg	47.7	32.6
85–89 mm Hg	16.1	3.5

¹Values are mean ± SD or median (IQR) for continuous variables and percentage for categorical variables.

²BP, blood pressure; hs-CRP, high-sensitivity C-reactive protein.

³-values were derived from paired t test for means and McNemar's test for proportions.

⁴Matching variable.

⁵To convert hs-CRP concentrations from mg/L to nmol/L, multiply by 9.524.

⁶To convert blood cholesterol concentrations from mg/dL to mmol/L, divide by 38.67.

TABLE 2.

 Erythrocyte fatty acid (FA) composition in baseline blood samples among hypertension cases and controls in a nested case-control study within the Women's Health Study (WHS)¹

	Cases			Controls
Erythrocyte FA (% of total)	All (n = 516)	White (n = 361)	Non-white (n = 148)	All (n = 516)	White (n = 370)	Non-white (n = 142)	P2
	% of total FA
Total SFA	39.2 ± 3.2	39.1 ± 3.1	39.5 ± 3.6	38.9 ± 2.9	38.9 ± 2.9	38.9 ± 2.9	0.10
16:0	23.5 ± 2.4	23.5 ± 2.4	23.6 ± 2.5	23.2 ± 2.1	23.3 ± 2.2	23.1 ± 2.1	0.05
18:0	11.8 ± 1.4	11.8 ± 1.4	11.9 ± 1.3	11.8 ± 1.4	11.8 ± 1.5	11.9 ± 1.1	0.92
Total cis MUFA	17.5 ± 1.9	17.8 ± 1.8	17.0 ± 1.8*	17.5 ± 1.7	17.8 ± 1.6	16.8 ± 2.0*	0.58
cis 16:1	0.44 ± 0.2	0.50 ± 0.2	0.31 ± 0.2*	0.40 ± 0.2	0.44 ± 0.2	0.30 ± 0.2*	0.0009
cis 18:1	14.9 ± 1.6	15.0 ± 1.6	14.5 ± 1.7*	14.8 ± 1.4	15.0 ± 1.3	14.5 ± 1.6*	0.86
Total cis PUFA	32.7 ± 3.9	32.8 ± 3.5	32.4 ± 4.8	33.1 ± 3.5	33.0 ± 3.3	33.5 ± 4.1	0.03
cis (n-3) PUFA	5.91 ± 1.6	5.70 ± 1.3	6.44 ± 2.1*	6.11 ± 1.5	5.88 ± 1.3	6.71 ± 1.7*	0.02
18:3(n-3)	0.16 ± 0.05	0.16 ± 0.05	0.16 ± 0.05	0.16 ± 0.05	0.16 ± 0.06	0.17 ± 0.05	0.94
20:5(n-3), EPA	0.53 ± 0.3	0.49 ± 0.2	0.63 ± 0.4*	0.53 ± 0.2	0.49 ± 0.2	0.65 ± 0.3*	0.72
22:5(n-3)	1.80 ± 0.4	1.84 ± 0.4	1.69 ± 0.5*	1.85 ± 0.4	1.87 ± 0.4	1.79 ± 0.4*	0.03
22:6(n-3), DHA	3.42 ± 1.2	3.21 ± 1.0	3.96 ± 1.5*	3.57 ± 1.1	3.36 ± 1.0	4.10 ± 1.3*	0.03
cis (n-6) PUFA	26.8 ± 3.1	27.1 ± 2.8	26.0 ± 3.7*	27.0 ± 2.8	27.1 ± 2.7	26.8 ± 3.2	0.12
18:2(n-6)	12.2 ± 1.4	12.3 ± 1.3	12.1 ± 1.7	12.4 ± 1.5	12.4 ± 1.4	12.4 ± 1.5	0.09
20:3(n-6)	1.57 ± 0.4	1.65 ± 0.4	1.38 ± 0.3*	1.50 ± 0.4	1.55 ± 0.4	1.37 ± 0.3*	0.003
20:4(n-6)	12.6 ± 2.5	12.7 ± 2.2	12.2 ± 3.0	12.8 ± 2.3	12.8 ± 2.2	12.7 ± 2.5	0.15
Total trans FA	1.97 ± 0.6	2.03 ± 0.5	1.81 ± 0.6*	2.02 ± 0.6	2.09 ± 0.6	1.86 ± 0.5*	0.11
trans 16:1	0.05 ± 0.03	0.05 ± 0.03	0.05 ± 0.03	0.05 ± 0.03	0.06 ± 0.02	0.05 ± 0.03*	0.78
trans 18:1	1.70 ± 0.5	1.76 ± 0.5	1.56 ± 0.5*	1.75 ± 0.5	1.81 ± 0.5	1.61 ± 0.5*	0.10
trans 18:2	0.21 ± 0.06	0.21 ± 0.06	0.19 ± 0.05*	0.22 ± 0.06	0.22 ± 0.07	0.20 ± 0.05*	0.52

¹Values are mean ± SD. *Different from white, < 0.05.

²-values were derived from comparison between cases and matched controls using paired t test.

In the initial models that only controlled for age and race and in the multivariable model 1 that additionally adjusted for lifestyle factors and dietary factors, the risk of hypertension increased across the increasing quartiles of total SFA and decreased across quartiles of total cis PUFA, cis (n-3) PUFA, and the PS ratio. After further adjustment for obesity-related metabolic factors in the multivariable model 2, these associations were attenuated and remained significant only for the PS ratio (Table 3). The ratio of (n-6):(n-3) PUFA showed a trend of positive association with risk of hypertension in the initial model and the multivariable model 1 but not in the multivariable model 2. cis MUFA, cis (n-6) PUFA, and trans FA were not associated with risk of hypertension in any model.

TABLE 3.

 RR and 95% CI of hypertension according to baseline erythrocyte fatty acid (FA) composition in a nested case-control study within the Women's Health Study (WHS)

	Quartiles of FA
Erythrocyte FA subtypes	1st	2nd	3rd	4th	P-trend1
Total SFA
Median, % of total FA	36.5	37.8	38.9	41.2
Cases/controls, n/n	109/129	120/129	135/129	152/129
Initial model2	1.00 (reference)	1.14 (0.78–1.67)	1.33 (0.89–1.97)	1.53 (1.03–2.29)	0.03
Multivariable model 13	1.00 (reference)	1.19 (0.79–1.78)	1.44 (0.94–2.20)	1.76 (1.14–2.73)	0.009
Multivariable model 24	1.00 (reference)	1.16 (0.75–1.79)	1.37 (0.86–2.18)	1.53 (0.94–2.47)	0.08
Total cis MUFA
Median, % of total FA	15.8	16.9	17.8	19.1
Cases/controls, n/n	121/129	120/129	152/129	123/129
Initial model2	1.00 (reference)	1.09 (0.74–1.61)	1.32 (0.89–1.94)	1.11 (0.75–1.64)	0.50
Multivariable model 13	1.00 (reference)	0.97 (0.64–1.48)	1.26 (0.83–1.91)	1.02 (0.66–1.56)	0.67
Multivariable model 24	1.00 (reference)	0.88 (0.56–1.40)	1.18 (0.75–1.85)	0.92 (0.58–1.46)	0.98
Total cis PUFA
Median, % of total FA	30.6	33.3	34.4	35.9
Cases/controls, n/n	149/129	131/ 129	138/129	98/129
Initial model2	1.00 (reference)	0.90 (0.62–1.31)	0.91 (0.64–1.31)	0.58 (0.39–0.87)	0.03
Multivariable model 13	1.00 (reference)	0.84 (0.56–1.25)	0.88 (0.60–1.29)	0.56 (0.36–0.87)	0.03
Multivariable model 24	1.00 (reference)	0.90 (0.58–1.39)	0.88 (0.58–1.34)	0.66 (0.41–1.05)	0.11
cis (n-3) PUFA
Median, % of total FA	4.7	5.7	6.4	7.8
Cases/controls, n/n	150/129	144/ 129	112/129	110/129
Initial model2	1.00 (reference)	0.95 (0.67–1.36)	0.74 (0.51–1.06)	0.69 (0.47–1.02)	0.03
Multivariable model 13	1.00 (reference)	0.87 (0.59–1.28)	0.66 (0.44–0.99)	0.65 (0.42–1.00)	0.03
Multivariable model 24	1.00 (reference)	0.84 (0.55–1.27)	0.67 (0.43–1.04)	0.75 (0.47–1.20)	0.19
cis (n-6) PUFA
Median, % of total FA	24.8	27.0	28.2	29.3
Cases/controls, n/n	135/129	131/ 129	140/129	110/129
Initial model2	1.00 (reference)	0.96 (0.67–1.39)	1.05 (0.72–1.52)	0.78 (0.53–1.14)	0.35
Multivariable model 13	1.00 (reference)	1.00 (0.67–1.48)	1.07 (0.71–1.62)	0.77 (0.51–1.16)	0.35
Multivariable model 24	1.00 (reference)	0.94 (0.61–1.45)	1.13 (0.72–1.78)	0.81 (0.52–1.26)	0.54
Total trans FA
Median, % of total FA	1.3	1.8	2.2	2.7
Cases/controls, n/n	140/129	152/ 129	106/129	118/129
Initial model2	1.00 (reference)	1.04 (0.73–1.49)	0.72 (0.50–1.05)	0.81 (0.57–1.17)	0.12
Multivariable model 13	1.00 (reference)	1.06 (0.72–1.54)	0.71 (0.47–1.07)	0.76 (0.51–1.14)	0.09
Multivariable model 24	1.00 (reference)	1.12 (0.74–1.68)	0.85 (0.54–1.32)	0.96 (0.62–1.48)	0.66
PS ratio5
Median	0.77	0.87	0.91	0.96
Cases/controls, n/n	151/129	143/ 129	123/129	99/129
Initial model2	1.00 (reference)	0.99 (0.69–1.43)	0.77 (0.53–1.11)	0.57 (0.38–0.86)	0.01
Multivariable model 13	1.00 (reference)	0.99 (0.67–1.47)	0.70 (0.47–1.04)	0.51 (0.33–0.80)	0.004
Multivariable model 24	1.00 (reference)	0.98 (0.64–1.51)	0.77 (0.49–1.18)	0.60 (0.37–0.97)	0.04
(n-6):(n-3) PUFA ratio
Median	3.4	4.2	4.9	5.8
Cases/controls, n/n	123/129	99/129	137/129	157/129
Initial model2	1.00 (reference)	0.82 (0.56–1.21)	1.17 (0.82–1.69)	1.37 (0.93–2.01)	0.04
Multivariable model 13	1.00 (reference)	0.75 (0.50–1.14)	1.20 (0.81–1.79)	1.42 (0.92–2.18)	0.03
Multivariable model 24	1.00 (reference)	0.70 (0.45–1.10)	1.02 (0.67–1.57)	1.26 (0.79–2.00)	0.17

¹Linear trend across increasing quartiles was tested using the median value of each quartile as an ordinal variable.

²Initial model controlled only for matching factors including age and race.

³Multivariable model 1 controlled for matching factors including age and race and additionally adjusted for total energy intake (continuous), smoking status (current, past, never), alcohol use (rarely/never, 10–45 g ethanol/mo, 10–90 g ethanol/wk, ≥13 g ethanol/d), exercise (rarely/never, <1, 1–3, ≥4 times/wk), menopause status (premenopausal, postmenopausal, unknown), and postmenopausal hormone use (current, past, never).

⁴Multivariable model 2 assessed the impact of additional adjustment for potential mediating factors including BMI (continuous), history of diabetes (yes, no), and history of hypercholesterolemia (yes, no).

⁵ PS ratio, ratio of PUFA:SFA.

In the secondary analyses, we found relatively stronger associations between erythrocyte FA and risk of hypertension among women aged <55 y than among women aged ≥55 y (Supplemental Table 1). Among women younger than 55 y, the multivariable model 2-adjusted RR of hypertension in the highest compared with the lowest quartile were 2.50 for SFA, 0.47 for total cis PUFA, 0.59 for cis (n-3) PUFA, 0.44 for cis (n-6) PUFA, and 0.35 for the PS ratio. In contrast, among women aged 55 y or older, none of the major erythrocyte FA subtypes, except cis (n-6) PUFA, had significant associations with risk of hypertension. Nevertheless, the tests for interaction with age reached significance only for total cis MUFA (P = 0.02), cis (n-6) PUFA (P = 0.03), and the PS ratio (P = 0.05). When FA were examined individually, the patterns of association for major individual FA were largely similar with their respective FA subtypes (Supplemental Table 2). Generally, there were no different associations by race/ethnicity, baseline BMI, and baseline BP (Supplemental Table 3).

Discussion

In this prospective, nested, case-control study among middle-aged and older women, we found that the proportion of SFA in erythrocyte membranes was positively associated with risk of hypertension and the proportion of total cis PUFA and the PS ratio were inversely associated with risk of hypertension. After adjustment for BMI, history of diabetes, and hypercholesterolemia, these associations were attenuated and remained significant only for the PS ratio, which likely reflect that obesity-related metabolic disorders partially mediate the causal pathway between FA and development of hypertension.

Experimental studies found that feeding rats with SFA impairs endothelial function (3), enhances sympathetic nervous system activity (4), and increases BP. In contrast, consumption of long-chain (n-3) PUFA changes the plasma phospholipid composition and cell membrane fluidity, increases the production of vasodilators, reduces cardiac adrenergic activity, and lowers BP (26, 27). Despite the in vivo and in vitro study findings, epidemiologic evidence on dietary FA in relation to risk of hypertension remains scarce. Two large-scale prospective cohort studies, the Nurses’ Health Study (10) and the Health Professionals Follow-up Study (11), reported no association between baseline intake of SFA, MUFA, PUFA, or trans FA assessed from FFQ and incident hypertension during a 4-y period of follow-up. However, our recent analyses in the WHS showed that dietary SFA, MUFA, and trans FA assessed from a similar FFQ were each positively associated with the risk of hypertension over a 12.9-y follow-up and the associations were partially explained by obesity-related factors (14).

FA in biospecimens has been proposed to provide a more accurate assessment of fat intake than self-report instruments. Adipose tissue is considered the best component to reflect an individual’s long-term fat intake (28–30), whereas plasma lipid fractions (30, 31) and erythrocyte membranes (30), which reflect short- to medium-term fat intake (over days and months, respectively), are more widely used in epidemiological studies because of their accessibility and low cost. To our knowledge, 2 studies have previously assessed the association of FA in plasma with the risk of hypertension (12, 13); our study is the first to examine erythrocyte FA and subsequent risk of hypertension. Consistent with the studies of plasma FA, we found a positive association for the proportion of SFA and an inverse association for the proportion of PUFA and the PS ratio in erythrocyte membranes with risk of hypertension. Because our earlier analyses in the entire WHS cohort found a positive association for trans fats intake assessed from the FFQ (14), lack of association between erythrocyte trans FA and hypertension risk in this nested case-control sample was unexpected. The reason for this discrepancy is unclear but may be due to the selective incorporation of trans FA into the macromolecular components within cell membranes. Trans FA differ from the naturally occurring cis counterpart in the manner they are incorporated into TG and phospholipids and in the specificity of their cholesterol esters to cholesterol esterases (32). Another possible explanation could be that the FFQ aims to assess long-term dietary FA intake, whereas erythrocyte membrane FA composition reflects a short- to medium-term exposure. In a cross-sectional study of erythrocyte FA and metabolic syndrome (33), trans FA in erythrocyte membranes were also uncorrelated with BP but were inversely correlated with HDL cholesterol, suggesting that in the link between trans FA and cardiovascular health, the adverse effect on blood lipid might be stronger than on BP.

In our study, the associations of erythrocyte SFA and PUFA with risk of hypertension were substantially attenuated after adjustment for BMI. This finding supports a mechanistic role of obesity and related abnormalities, including lipid and glucose metabolism disorders and systemic inflammation, in the association between dietary fats and hypertension risk. Based on our earlier finding in the entire WHS cohort (14), we also stratified current analysis by baseline age. Although not all tests for interaction reached significance, the earlier study and the current study consistently showed stronger and clearer associations of dietary FA or its biomarkers with hypertension risk in younger than in older women. It has been speculated that the pathogenesis of hypertension may be different in younger than in older populations (34). It is also possible that individuals remaining free of major chronic diseases in their old age have biochemical, physiological, and/or genetic profiles less vulnerable to adverse environmental factors (35). Different baseline risk may provide another explanation for the interaction from a statistical perspective. Specifically, a fixed absolute effect by a given risk factor will correspond to a smaller relative effect when the baseline risk of disease is higher (such as in an elderly population).

FA in the human body is mainly provided from diet. Meanwhile, FA can also be endogenously synthesized. In humans, because the capacity of PUFA elongation is very limited, long-chain PUFA in human tissues are obtained primarily from diet. Trans isomers of FA cannot be converted in humans, so the only source of trans FA in tissues is via the diet. In our study, long-chain (n-3) PUFA and trans FA in erythrocyte membranes showed a moderately strong correlation with the corresponding dietary intake. In contrast, erythrocyte SFA and MUFA, which can be synthesized, were only weakly correlated with the corresponding dietary intake (all P > 0.05). The processes of FA absorption, transport, metabolism, and utilization can all influence the FA composition in specific tissues, thus reducing the utility of tissue FA as an indicator of dietary intake but enhancing its role as a marker of available FA at the cellular level (28). The associations observed for erythrocyte SFA and total PUFA, although not ideally reflecting dietary FA, demonstrate an important role of FA that have been incorporated into cell membranes in the development of hypertension.

Several limitations of the current study deserve comment. First, hypertension was identified based on self-reported information and subject to misclassification. However, the accuracy of self-reported hypertension in health professionals has been demonstrated in previous validation studies using clinical measurements (18) and medical record review (11, 19). Second, we have only a single baseline measurement of erythrocyte FA and thus cannot take into account any change of erythrocyte FA during follow-up. Furthermore, the FA composition of other tissues and FFA concentrations in blood were not available for our study. As a result, possible random misclassification in exposure may lead to an underestimation of true associations. Third, despite comprehensive adjustment for multiple lifestyle, dietary, and clinical factors, residual confounding cannot be ruled out in our study as in other observational studies. Fourth, because the WHS participants are U.S. health professionals and are predominantly white women, the findings from this study may not apply to other socioeconomic or ethnic populations. However, we oversampled nonwhite women and did not find significantly different results by ethnicities, which argues against a limited generalizability of our findings.

In conclusion, our study found a positive association of erythrocyte SFA and an inverse association of erythrocyte cis PUFA and the PS ratio with risk of hypertension among middle-aged and older women. After adjustment for potential mediating factors including obesity and related metabolic disorders, these associations were attenuated and only the association for the PS ratio remained significant. MUFA and trans FA in erythrocyte membranes were not associated with risk of hypertension. Future studies need to identify both dietary and nondietary factors that affect the incorporation of FA in different tissues and their role in the development of hypertension and other cardiometabolic diseases.