Insulin Resistance and Risk of Incident Hypertension Among Men

Abstract

The independent association between insulin resistance and the development of hypertension remains in doubt because insulin resistance correlates with other metabolic factors also proposed to be associated with hypertension. The authors examined the association between the insulin sensitivity index and incident hypertension in a prospective nested case‐control study among 1453 men (mean age, 61 years) who participated in the Health Professionals’ Follow‐up Study. The authors computed the insulin sensitivity index for each man in the study based on fasting insulin and triglyceride levels. Logistic regression was performed conditioned on age and adjusted for standard hypertension risk factors as well as renal function, cholesterol, and uric acid. The insulin sensitivity index was 6% lower in the cases compared with the controls (P<.001). The multivariable odds ratio for hypertension comparing the lowest with highest quartile of insulin sensitivity index was 1.09 (0.71–1.65) among the entire sample. However, the association between the insulin sensitivity index and incident hypertension differed significantly by age (P interaction <.001). Among men younger than 60 years, the multivariable odds ratio for the lowest compared with highest quartile was 1.93 (1.01–3.71) but was 0.67 (0.37–1.24) among older men. Insulin resistance is independently associated with incident hypertension among younger men.

Hypertension (HTN) is a leading cause of cardiovascular morbidity and mortality and, in most cases, the etiology is unclear. Insulin resistance may play a role in the pathogenesis of HTN, potentially via adverse effects on the renin‐angiotensin system (RAS), sympathetic nervous system, and vascular muscle tone. ¹ , ² , ³ Welborn and colleagues ⁴ first observed that insulin levels were higher among patients with essential HTN than among normotensive patients. Subsequent cross‐sectional studies examined this relation but results were conflicting. ⁵ , ⁶ , ⁷ , ⁸ In the past decade, many prospective longitudinal studies reported crude positive associations between hyperinsulinemia and incident HTN, but these relations were typically confounded by body mass index (BMI) and baseline blood pressure (BP). ⁹ , ¹⁰ , ¹¹ , ¹² , ¹³ , ¹⁴ , ¹⁵ , ¹⁶ , ¹⁷ , ¹⁸ , ¹⁹ , ²⁰ , ²¹ , ²² , ²³ , ²⁴ , ²⁵ , ²⁶ , ²⁷ Furthermore, insulin resistance correlates with higher uric acid levels, which are themselves associated with HTN, ²⁸ but prior studies did not adjust for uric acid. The independent association between insulin resistance and incident HTN remains, therefore, an area of active debate.

We performed a prospective nested case‐control study of 1453 male participants aged 47 to 81 years in the Health Professionals’ Follow‐up study (HPFS) to examine the independent association between insulin resistance and risk for incident HTN.

Materials and Methods

Study Sample

The HPFS is an ongoing prospective cohort study of 59,529 male health professionals that began in 1986 and has been described in detail elsewhere. ²⁹ Follow‐up of participants was >90% through 2004. In 1993, 18,025 men contributed blood samples that were stored in liquid nitrogen (−130°C). We conducted a nested case‐control study among those men with available blood samples and without prevalent HTN in 1994 (approximately 1 year after blood samples were collected). Only men whose BMI in 1994 was <30 kg/m² and whose blood sample was drawn after fasting for ≥8 hours were considered. The BMI restriction was imposed because BMI is a strong predictor of insulin resistance, and because BMI is a powerful predictor of HTN. We then randomly selected 750 men in whom new HTN was diagnosed from 1994 to 2002 and 750 age‐matched controls by risk‐set sampling. ³⁰ Because risk‐set sampling allows for a control to be selected as a case in a subsequent time period, 713 controls were unique individuals, and 37 controls were later selected as cases and matched to a new control. Risk‐set sampling is commonly used for prospective nested case‐control studies, and the resulting odds ratio (OR) derived from logistic regression directly estimates the relative risk. ³⁰ After exclusion of 10 men with missing laboratory values, the total study sample consisted of 1453 unique individuals and 740 matched case‐control pairs. The institutional review board at Brigham and Women’s Hospital reviewed and approved this study.

Ascertainment of Insulin and Triglycerides

Fasting insulin and triglyceride levels were used as biomarkers of insulin sensitivity and measured using a radioimmunoassay and standard enzymatic methods (coefficient of variations 4.6% and 3.6%, respectively). The insulin sensitivity index (glucose disposal rate [M] corrected for fat‐free mass [MFFM]) was estimated for participants using the following prediction equation that includes fasting insulin and triglyceride levels (triglyceride levels converted to mmol/L):

This estimated MFFM value has been validated ³¹ and accepted as an index of insulin sensitivity (inversely related to insulin resistance). ³²

Ascertainment of HTN

On biennial mailed questionnaires, we asked participants to report whether a clinician had made a new diagnosis of HTN during the preceding 2 years. Self‐reported HTN is highly reliable in HPFS. ³³ Among a subset of men who reported having HTN, 100% had the diagnosis confirmed by medical record review. In addition, self‐reported HTN was highly predictive of subsequent cardiovascular events. ³³ A participant was considered to have prevalent HTN, and thus excluded, if he reported this diagnosis on any questionnaire up to and including the 1994 questionnaire. Therefore, cases only included individuals who first reported HTN on subsequent questionnaires (1996–2002).

Ascertainment of Covariates

Age, BMI (weight in kilograms divided by height in meters squared), smoking status, physical activity, and alcohol intake were ascertained from the 1994 questionnaire. Baseline BP was ascertained from the 1992 questionnaire (not available from 1994), when participants reported their usual BP in categories, and were assigned the median of the chosen category. Questionnaire‐derived information about these covariates has previously been validated, with correlations of 0.97 for weight compared with direct measurement, 0.79 for physical activity compared with physical activity diaries, and 0.90 for alcohol compared with multiple averaged dietary records. ²⁹ , ³⁴ , ³⁵ Family history of HTN was available on the 1990 questionnaire.

In addition to insulin and triglyceride levels, blood samples were also assayed in the same laboratory for creatinine using a modified Jaffe method, uric acid using a uricase oxidization assay, and cholesterol by a standard esterase‐oxidase method. The coefficients of variation for these measurements were 4.0%, 2.7%, and 2.7%, respectively. Estimated glomerular filtration rate (eGFR) was established using the Modification of Diet in Renal Disease (MDRD) equation: ³⁶ .

Statistical Analyses

Because the continuous baseline variables, including the MFFM, were not normally distributed, differences between these variables among cases and controls were compared using the Wilcoxon rank sum test. Differences in categoric variables between cases and controls were compared using the chi‐square test.

To examine the correlations between MFFM and age, BMI, uric acid, eGFR, and cholesterol, we used Spearman partial correlations, in which pair‐wise Spearman correlation coefficients were computed after adjusting for the other variables. For example, the Spearman correlation between BMI and MFFM was adjusted for age, eGFR, and cholesterol. Insulin sensitivity (MFFM) was first examined in quartiles, ranging from the highest MFFM values (most insulin sensitive) to the lowest MFFM values (least insulin sensitive). We additionally examined MFFM as a continuous variable to calculate the OR for HTN per one unit decrease (ie, lower insulin sensitivity).

We analyzed the association between MFFM and HTN using conditional logistic regression conditioning on the matching factor (age). We generated several hierarchical multivariable models: model 1, adjusted for BMI; model 2, adjusted for BMI and also alcohol consumption, physical activity, smoking status, and family history of HTN; model 3, adjusted for variables in model 2 plus baseline systolic BP (SBP); and model 4, adjusted for variables in model 3 plus uric acid, eGFR, and total cholesterol. For all ORs, we calculated 95% confidence intervals (CIs).

The men in our sample were considerably older than the populations studied by others. We therefore investigated whether the association between MFFM and HTN varied by age (<60 years, ≥60 years).

All statistical analyses were performed with SAS version 9.2 (SAS Institute, Inc, Cary, NC).

Results

Baseline Characteristics

The baseline characteristics of the study population by case status are displayed in Table I. The median age was 61 years (interquartile range [IQR], 54 to 68 years). Age was a matching factor, and therefore did not differ between cases and controls. Cases had higher fasting levels of insulin and triglycerides and thus had lower MFFM scores than the control group. The median MFFM scores at baseline were 7.38 units (IQR, 6.24–8.83) among cases and 7.84 units (IQR, 6.46–9.42) among controls (P<.001). Compared with controls, cases had higher baseline BMI, higher SBP, and were more likely to have a family history of HTN. There were no significant differences between cases and controls in physical activity, alcohol consumption, and smoking status.

Table I.

 Baseline Characteristics of Cases and Control Patients

Characteristic	All	Cases	Controls	P Value
Demographic and physiologic
Age, y	61 (54–68)	61 (54–69)	61 (53–68)
Body mass index, kg/m2	25.0 (23.3–26.8)	25.2 (23.6–27.1)	24.7 (23.1–26.5)	<.001
Physical activity, METs/wk	26.9 (11.7–51.6)	26.5 (11.4–49.5)	27.7 (12.4–53.7)	.34
Systolic blood pressure, mm Hg	120 (120–130)	130 (120–140)	120 (120–130)	<.001
Alcohol, g	6.3 (0.9–17.0)	6.4 (0.9–17.7)	6.1 (0.9–16.1)	.26
Current smoker, %	5.5	6.5	4.5	.09
Past smoker, %	43.3	45.3	41.3	.12
Family history of hypertension, %	39.0	43.1	34.9	.001
Fasting biomarkers
Insulin, IU/mL	6.2 (4.2–9.4)	6.6 (4.5–9.8)	6.0 (4.0–8.9)	<.001
Triglycerides, mg/dL	118 (82–169)	123 (85–176)	113 (80–157)	.004
MFFM	7.6 (6.3–9.2)	7.4 (6.2–8.8)	7.8 (6.5–9.4)	<.001
eGFR, mL/min/1.73 m2	80 (71–90)	80 (71–90)	80 (71–90)	.98
Uric acid, mg/dL	6.0 (5.2–6.8)	6.0 (5.2–6.8)	5.9 (5.2–6.8)	.52
Cholesterol, mg/dL	211 (186–236)	212 (187–238)	208 (186–235)	.15

Abbreviations: eGFR, estimated glomerular filtration rate; METS, metabolic equivalents; MFFM, glucose disposal rate (M) corrected for fat‐free mass. Continuous variables are expressed as median (25th–75th percentile). Categoric variables are expressed as percent. Continuous variables were analyzed using the Wilcoxon rank sum test and categoric variables with the chi‐square test.

No significant differences were observed between cases and controls with respect to eGFR, uric acid, and cholesterol. MFFM was significantly associated with BMI (partial Spearman correlation, r=−0.40; P<.001), uric acid (r=−0.12; P<.001), cholesterol (r=−0.19; P<.001), and eGFR (r=−0.05; P=.04).

Association With HTN

The median MFFM value was 7.6 and 25.3% of participants had an MFFM score ≤6.3, which indicates insulin resistance. Lower MFFM was found to be a significant predictor of HTN in crude but not multivariable adjusted analyses (Table II). Compared with men in the highest MFFM quartile (10.6; range, 9.3–24.5), those in the lowest (5.5; range, 2.9–6.2) had a multivariable OR for incident HTN of 1.09 (0.71–1.65). The multivariable OR for each 1‐unit decrease in MFFM was 1.04 (0.98–1.11).

Table II.

 ORs of Incident Hypertension According to Quartile of Insulin Sensitivity (MFFM)

Parameter	Continuous (Per 1‐Unit Decrease)	Quartile 4 Median (Range) 10.6 (9.3–24.5)	Quartile 3 Median (Range) 8.4 (7.6–9.2)	Quartile 2 Median (Range) 6.9 (6.3–7.5)	Quartile 1 Median (Range) 5.5 (2.9–6.2)
No. of cases	742	162	181	194	205
Age‐matched	1.08 (1.04–1.14)	1.0 (reference)	1.22 (0.92–1.63)	1.42 (1.06–1.89)	1.60 (1.19–2.14)
Multivariate
Model 1	1.05 (1.00–1.11)	1.0 (reference)	1.14 (0.85–1.53)	1.24 (0.92–1.67)	1.32 (0.97–1.81)
Model 2	1.04 (0.99–1.10)	1.0 (reference)	1.13 (0.83–1.54)	1.22 (0.88–1.68)	1.25 (0.89–1.75)
Model 3	1.04 (0.97–1.11)	1.0 (reference)	1.01 (0.70–1.47)	1.23 (0.85–1.78)	1.11 (0.74–1.66)
Model 4	1.04 (0.98–1.11)	1.0 (reference)	1.02 (0.69–1.49)	1.21 (0.83–1.78)	1.09 (0.71–1.65)

Abbreviations: MFFM, glucose disposal rate (M) corrected for fat‐free mass; ORs, odds ratios. Results are expressed as ORs (95% confidence interval). Model 1, adjusted for body mass index (BMI). Model 2, adjusted for BMI, alcohol, physical activity, smoking status, and family history of hypertension. Model 3, adjusted for BMI, alcohol, physical activity, smoking status, family history of hypertension, and systolic blood pressure (SBP). Model 4, adjusted for BMI, alcohol, physical activity, smoking status, family history of hypertension, SBP, and for the following biomarkers: uric acid, estimated glomerular filtration rate and total cholesterol.

Age‐Stratified Models

Because the pathophysiology of HTN occurring in younger vs older individuals may differ, we analyzed the association between MFFM and incident HTN after stratifying participants according to age (<60 years vs ≥60 years). As shown in Table III, younger men consumed more alcohol, had lower baseline SBP, were less likely to have smoked previously, and were more likely to report a family history of HTN. The only biomarker that differed significantly between younger and older men was eGFR, which was anticipated because age is included in the equation.

Table III.

 Baseline Characteristics Stratified by Age

Characteristic	Age <60 y	Age ≥60 y	P Value
Demographic and physiologic
Age, y	53 (50–56)	67 (63–72)
Body mass index, kg/m2	25.1 (23.6–26.8)	25.0 (23.1–26.8)	.14
Physical activity, METs/wk	26.8 (12.5–50.8)	27.4 (11.6–54.3)	.40
Systolic blood pressure, mm Hg	120 (120–130)	130 (120–140)	<.001
Alcohol, g	7.3 (1.1–17.4)	5.1 (0.0–16.4)	.02
Current smoker, %	6.1	5.1	.41
Past smoker, %	36.3	48.8	<.001
Family history of hypertension, %	46.5	33.1	<.001
Fasting biomarkers
Insulin, IU/mL	6.2 (4.2–9.2)	6.2 (4.3–14.0))	<.51
Triglycerides, mg/dL	115 (81–174)	119 (82–167)	.94
MFFM	7.6 (6.4–9.3)	7.6 (6.3–9.1)	.67
eGFR, mL/min/1.73 m2	82 (74–92)	78 (69–88)	<.001
Uric acid, mg/dL	5.9 (5.2–6.8)	6.0 (5.2–6.8)	.61
Cholesterol, mg/dL	209 (186–234)	211 (187–238)	.25

Abbreviations: eGFR, estimated glomerular filtration rate; METs, metabolic equivalents; MFFM, glucose disposal rate (M) corrected for fat‐free mass. Continuous variables are expressed as median (25th–75th percentile). Categoric variables are expressed as percent. Continuous variables were analyzed using the Wilcoxon rank sum test and categoric variables with the chi‐square test.

The multivariable OR for incident HTN for each 1‐unit decrease in MFFM varied according to age group (P<.001 for interaction) (Table IV). For each unit lower MFFM score, the OR for HTN was 1.15 (95% CI, 1.04–1.28) among men younger than 60 years and was 0.97 (95% CI, 0.88–1.07) among men 60 years and older. Similarly, the OR for the lowest compared with highest quartile of insulin sensitivity was 1.93 (95% CI, 1.01–3.71) among younger men and 0.67 (95% CI, 0.37–1.24) among older men (Table IV).

Table IV.

 Odds of Incident Hypertension According to MFFM, Stratified by Age

Parameter	Continuous (Per 1‐Unit Decrease)	Quartile of MFFM (Unit)
		Quartile 4 10.6 (9.3–24.5)	Quartile 3 8.4 (7.6–9.2)	Quartile 2 6.9 (6.3–7.5)	Quartile 1 5.5 (2.9–6.2)
Age <60 y
No. of cases	324	72	73	89	90
Age‐matched	1.15 (1.06–1.24)	1.0 (reference)	1.35 (0.87–2.09)	1.68 (1.08–2.62)	1.92 (1.21–3.06)
Model 1	1.12 (1.04–1.22)	1.0 (reference)	1.28 (0.82–1.98)	1.52 (0.97–2.39)	1.67 (1.03–2.70)
Model 2	1.14 (1.05–1.24)	1.0 (reference)	1.34 (0.82–2.17)	1.62 (0.98–2.67)	1.92 (1.12–3.29)
Model 3	1.15 (1.04–1.28)	1.0 (reference)	1.51 (0.82–2.75)	1.89 (1.04–3.45)	1.97 (1.05–3.71)
Model 4	1.15 (1.04–1.28)	1.0 (reference)	1.45 (0.78–2.67)	1.91 (1.01–3.59)	1.93 (1.01–3.71)
Age >60 y
No. of cases	418	90	108	105	115
Age‐matched	1.03 (0.97–1.10)	1.0 (reference)	1.07 (0.72–1.59)	1.20 (0.81–1.78)	1.35 (0.90–2.02)
Model 1	0.99 (0.93–1.07)	1.0 (reference)	0.97 (0.66–1.48)	1.04 (0.68–1.57)	1.09 (0.70–1.69)
Model 2	0.97 (0.90–1.04)	1.0 (reference)	0.91 (0.59–1.39)	1.00 (0.65–1.53)	0.89 (0.56–1.41)
Model 3	0.95 (0.87–1.04)	1.0 (reference)	0.67 (0.40–1.14)	0.91 (0.55–1.50)	0.68 (0.38–1.20)
Model 4	0.97 (0.88–1.07)	1.0 (reference)	0.67 (0.39–1.16)	0.96 (0.56–1.64)	0.67 (0.37–1.24)

Abbreviation: MFFM, glucose disposal rate (M) corrected for fat‐free mass. Results are expressed as odds ratios (95% confidence interval). Model 1, adjusted for body mass index (BMI). Model 2, adjusted for BMI, alcohol, physical activity, smoking status, and family history of hypertension. Model 3, adjusted for BMI, alcohol, physical activity, smoking status, family history of hypertension, and systolic blood pressure (SBP). Model 4, adjusted for BMI, alcohol, physical activity, smoking status, family history of hypertension, SBP, and for the following biomarkers: uric acid, estimated glomerular filtration rate, and total cholesterol.

To determine whether fasting insulin levels alone (without conversion to the insulin sensitivity index) were also independently associated with incident HTN in young men, we analyzed fully adjusted models that included insulin and triglyceride levels separately. Neither fasting insulin as a continuous variable (OR, 1.01; 95% CI, 0.97–1.05) nor the highest compared with the lowest quartile (OR, 1.48; 95% CI, 0.76–2.89) was associated with incident HTN among men younger than 60 years.

Discussion

We found a robust and independent inverse association between insulin sensitivity and the risk of HTN among younger, but not older, men. To our knowledge, this is the first prospective study of this association to control not only for standard HTN risk factors, but also for baseline BP and other metabolic factors such as renal function, cholesterol, and uric acid.

Results from previous examinations of the association between insulin and HTN are conflicting. An early meta‐analysis of cross‐sectional studies conducted from 1983 to 1991 concluded that fasting serum insulin levels were significantly correlated with both systolic and diastolic BP, supporting a role for hyperinsulinemia in the development of HTN. ³⁷ More recently, a handful of prospective studies reported a positive relation between hyperinsulinemia and risk of incident HTN, but most of the studies were limited in size and not fully adjusted. ⁹ , ¹⁰ , ¹¹ , ¹² , ¹³ , ¹⁴ , ¹⁵ , ¹⁶ , ¹⁷ , ¹⁸ , ¹⁹ , ²⁰ , ²¹ , ²² , ²³ , ²⁴ , ²⁵ , ²⁶ , ²⁷ For example, only one large‐scale analysis demonstrated that fasting insulin level was associated with the risk of HTN after multivariable adjustment for age, obesity, baseline BP, and other factors. ³⁸

In the present study, we chose the insulin sensitivity index (MFFM score) rather than fasting insulin levels to assess insulin resistance because fasting insulin levels may be confounded by variations in insulin secretion. ³⁹ Furthermore, we corrected not only for standard HTN risk factors, but also for baseline BP as well as renal function, cholesterol, and uric acid. These latter biomarkers were important to consider given their association with HTN in prior studies, ⁴⁰ , ⁴¹ , ⁴² as well as their correlation with insulin sensitivity.

Our finding that lower levels of insulin sensitivity strongly predicted the development of HTN in younger participants is consistent with data from previous studies. In the San Luis Valley Diabetes Study, Shetterly and colleagues ¹⁵ examined the predictors of HTN and found that among younger individuals (46–60 years), impaired glucose tolerance increased the risk of HTN by 1.67‐fold compared with older individuals (61–74 years). Arnlov and colleagues ⁴³ analyzed the relation between insulin resistance and BP increase among Framingham Study participants using models stratified by age, BMI, and baseline BP. The authors observed that insulin resistance was associated with BP increases principally in younger (<51 years) and normotensive participants. Taken together, our findings, as well as the findings by others, support a role for insulin resistance in the development of HTN among younger but not older individuals.

Several mechanisms have been proposed to explain the association between insulin resistance and incident HTN. First, hyperinsulinemia directly increases sodium reabsorption in renal tubules and may therefore lead to sodium and fluid retention. ⁴⁴ , ⁴⁵ , ⁴⁶ , ⁴⁷ Second, hyperinsulinemia activates the sympathetic nervous system. ² Euglycemic clamp studies demonstrated that hyperinsulinemia increases catecholamine levels and elevates SBP, a phenomenon that was independent of blood glucose levels. ¹ , ⁴⁸ Interestingly, the effect of hyperinsulinemia on the sympathetic drive may vary by age. Hausberg and colleagues ⁴⁹ observed that, compared with the increases in norepinephrine and heart rate after insulin infusion in normotensive young individuals, the sympathetic response to insulin infusion in normotensive elderly individuals was substantially blunted. These data provide a potential mechanistic explanation for the age‐dependent variability in the association. Third, insulin may increase the activity of the RAS. Perfusion of rat kidneys with insulin, but not with other hormones tested, stimulated renin release in one study. ⁵⁰ In a second study, stimulation of cultured human adipocytes with insulin increased production of both angiotensinogen and angiotensin II in a dose‐dependent manner. ⁵¹ The activity of the renal tissue RAS is also positively associated with circulating levels of insulin. ⁵² Other proposed mechanisms linking insulin resistance with HTN include increases in smooth muscle proliferation ⁵³ , ⁵⁴ and endothelial dysfunction. ⁵⁵

In contrast, HTN does not develop in everyone with insulin resistance and hyperinsulinemia, ⁵⁶ and acute insulin infusion may produce vasodilatation and no increase in BP. ⁵⁷ Therefore, a mechanism whereby fasting hyperinsulinemia alone explains the association between insulin resistance and HTN may be overly simplistic. Future studies may shed more light on the potential mechanism of how insulin resistance impacts BP.

Strengths and Limitations

Strengths of the present study include its prospective matched nested case‐control design, long duration of follow‐up, high‐quality information on relevant confounders, multivariable adjustment for other circulating factors such as renal function and uric acid, and the use of the insulin sensitivity index rather than fasting insulin levels as a marker of insulin resistance. Our study also has limitations. First, we relied on self‐reported HTN and BP; however, all participants were trained health professionals, and HTN reporting was previously shown to be highly accurate in this population. ³³ Second, controls may have been unaware of existing HTN, thereby not self‐reporting the diagnosis and leading to misclassification. But since we only chose controls who had clinician examinations during the follow‐up period, this possibility is greatly reduced. Moreover, this type of misclassification, if it existed, would drive ORs toward 1.0 (null), leading us to underestimate the true association. Third, we purposefully restricted our sample to men with BMI values <30 kg/m². Although this limits the generalizability of our findings to nonobese men, others have suggested that the association between insulin resistance and HTN may be stronger in leaner individuals. ¹⁵ , ²¹ Finally, this was not a randomized controlled trial and, as with all observational research, the possibility exists for residual confounding from measured and unmeasured factors.

Conclusions

Lower degrees of insulin sensitivity strongly and independently predict incident HTN among younger men. The seemingly age‐dependent association between insulin resistance and HTN may be important because HTN that occurs in younger individuals tends to be phenotypically distinct from HTN that occurs in older individuals ⁵⁸ and may also have important public health implications should measures of insulin resistance ever be clinically used as risk factors. Future studies should address whether interventions to enhance insulin sensitivity would lower the risk of HTN among younger men.