Differential Effect of Glycemia on the Incidence of Hypertension by Sex: The Epidemiology of Diabetes Complications study

Participants from the Pittsburgh Epidemiology of Diabetes Complications (EDC) study with arterial blood pressure <140/90 at study initiation were selected for study (n = 510). The EDC is a historical cohort study based on incident cases of childhood-onset (prior to their 17th birthday) type 1 diabetes, diagnosed or seen within 1 year of diagnosis (1950–1980) at the Children’s Hospital of Pittsburgh (15). This cohort has been previously shown to be epidemiologically representative of the type 1 diabetes population of Allegheny County, Pennsylvania (16). The first clinical assessment for the EDC study took place between 1986 and 1988, when the mean participant age and diabetes duration were 28 and 19 years, respectively. Subsequently, biennial examinations were conducted for 10 years, with a further examination at 18 years of follow-up. The University of Pittsburgh institutional review board approved the study protocol.

Prior to each clinic visit, participants were sent questionnaires concerning demographic, health care, diabetes self-care, and medical history information. Leisure time physical activity was assessed by the Paffenbarger questionnaire (17), and a previously published algorithm (18) was used to calculate physical activity over the past week and over the past year based on the daily number of city block equivalents walked, the number of flights of stairs climbed, and the frequency and duration of leisure time activity. Three blood pressure measurements were taken by trained and certified personnel with a Hawksley random zero sphygmomanometer, after a 5-min rest in the sitting position according to the Hypertension Detection and Follow-up Program protocol (19). Hypertension was defined as ≥140/90 mmHg (mean of the second and third readings) or use of antihypertensive medications. For the present analyses, participants were only considered “hypertensive” if they were positive on two consecutive examination cycles. Intensive insulin therapy was defined as multiple (three or more) daily insulin injections or continuous subcutaneous insulin infusion in addition to frequent (at least four times daily) glucose testing. Stable glycosylated hemoglobin (HbA₁) was measured by ion-exchange chromatography (Isolab, Akron, OH) for the first 18 months, and the subsequent 10 years by automated high-performance liquid chromatography (Diamat; Bio-Rad, Hercules, CA); the two assays were highly correlated (r = 0.95). For follow-up beyond the 10 years, HbA_1c was measured with the DCA 2000 analyzer (Bayer, Tarrytown, NY). The DCA and Diamat assays were also highly correlated (r = 0.95). Original HbA₁ (1986–1998) and HbA_1c values (1998–2004) were converted to DCCT-aligned standard HbA_1c values using regression formulae derived from duplicate assays (DCCT HbA_1c = [0.83 × Diamat HbA₁] + 0.14 and DCCT HbA_1c = [DCA HbA_1c − 1.13]/0.81). HDL cholesterol was determined enzymatically after precipitation with heparin and manganese chloride, with a modification (20) of the Lipid Research Clinics method (21). Cholesterol and triglycerides were measured enzymatically (22,23). Non-HDL cholesterol was calculated as total minus HDL cholesterol. White blood cell (WBC) count was obtained using a counter S-plus IV and fibrinogen using a biuret colorimetric procedure and a clotting method. Urinary albumin was measured by immunonephelometry (24), and creatinine was assayed by an Ectachem 400 Analyzer (Eastman Kodak Co., Rochester, NY). Glomerular filtration rate was estimated by the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) creatinine equation (25). It should, however, be noted that serum creatinine was not calibrated in this study. All assays were conducted during the cycle that samples were obtained, and thus, prolonged storage would not have affected measurements performed in this study.

Analyses were conducted stratified by sex. Univariate associations were determined using the Student t test for normally distributed continuous variables or Wilcoxon two-sample test for nonnormally distributed continuous variables. The χ² or Fisher exact test, as appropriate, was used for univariate analysis of categorical variables. Cox proportional hazards models with backward elimination were constructed to assess independent predictors of hypertension incidence among traditional risk factors and univariately significant variables. Cox proportional hazards models were also constructed using time-dependent updated means of variables significant in models using baseline characteristics. Survival time was defined as the time in years from study entry to either incident hypertension or censorship during the 18-year follow-up. Nonnormally distributed variables were logarithmically transformed for entry into multivariable models. Statistical analyses were conducted using Statistical Analysis Software (SAS), version 9.2 (SAS Institute, Cary, NC).

During 18 years of follow-up, 39.2% (n = 200) of individuals developed incident hypertension, for an incidence rate of 31.2 per 1,000 person-years. Incidence was slightly lower in women (35.4%) than in men (43.2%, P = 0.07). Table 1 presents characteristics of male and female participants at study entry by incident hypertension. In both sexes, compared with participants whose arterial blood pressure remained within a normal range, those who subsequently developed hypertension were more likely to be older and have elevated baseline blood pressure, non-HDL cholesterol, albumin excretion rate (AER), and WBC count. In addition, male participants who developed hypertension were more likely to have a longer duration of type 1 diabetes, larger waist-to-hip ratio (WHR), and elevated HbA_1c and fibrinogen levels at study entry. Among women, those who subsequently developed hypertension had lower HDL cholesterol concentrations at study entry compared with women who maintained normal arterial blood pressure. No participant received angiotensin-converting enzyme inhibitors/angiotensin receptor blockers at study baseline.

To evaluate whether a dissimilar distribution of HbA_1c between male and female participants at study entry was responsible for the observed discrepancy in its association with hypertension incidence by sex, we assessed incidence by HbA_1c tertiles. As shown in Fig. 1A, the proportion of incident cases was similar by sex within the first HbA_1c tertile but appeared to be slightly increased among men compared with women within the second and third tertiles. Moreover, although hypertension incidence appeared to increase linearly with increasing HbA_1c tertile among men (P value for trend = 0.0008), a similar increase in risk was not apparent among women (P value for trend = 0.31). Figure 1B and C depicts the diabetes duration–adjusted 18-year survival free of hypertension for men and women, respectively, by tertiles of HbA_1c at study entry. These graphs clearly show a strong association between HbA_1c and hypertension incidence among men but a much weaker, nonsignificant relationship among women. Thus, compared with men with HbA_1c <8%, the estimated relative hazard of developing hypertension was 2.42 (95% CI 1.43–4.10) times higher among those with HbA_1c between 8 and 9.2% and four (2.35–6.80) times higher among men whose levels were >9.2%. Among women, however, even levels >9.2% were only associated with a borderline significantly increased hazard compared with HbA_1c levels <8% (P = 0.07).

The presence of effect modification of HbA_1c by sex was confirmed with the results of a significant interaction term (P = 0.009) in a Cox proportional hazards model that included variables for sex, HbA_1c, and the interaction term. Thus, separate multivariable models (with backward elimination) were constructed for male and female participants. Since only 4 men and 21 women followed an intensive insulin therapy protocol at study entry, this variable was not included in multivariable models of baseline predictors. Among men (Table 2), elevated HbA_1c, systolic blood pressure (SBP), and AER predicted hypertension development. In women, elevated SBP and AER were also predictors, along with non-HDL cholesterol. HbA_1c, however, was not an independent predictor in women.

To better understand the modification of the effect of HbA_1c by sex, we also performed separate Cox proportional hazards models by HbA_1c tertile (Table 3). Interestingly, these analyses revealed the presence of an over twofold significantly increased risk for hypertension incidence in women compared with men among those in the lowest HbA_1c tertile (hazard ratio 2.24 [95% CI 1.17–4.27]). No difference in risk by sex was noted in the second tertile of HbA_1c (1.30 [0.77–2.18]), whereas a significantly lower risk for women compared with men was observed in the third HbA_1c tertile (0.53 [0.32–0.86]). These findings could not be attributed to sex differences in the distribution of HbA_1c within each tertile (the P value for a difference by sex was 0.81, 0.60, and 0.47 for the first, second, and third tertile, respectively).

A significant interaction between sex and HbA_1c (P < 0.0001) was also noted when analyses were repeated using time-dependent updated means of baseline independent predictors of hypertension incidence. Thus, although a significant effect of HbA_1c was seen even among female participants (hazard ratio 1.21 [95% CI 1.00–1.46]), the effect of glycemic control on hypertension development was stronger in men (1.59 [1.29–1.97]) (data not shown). As intensive insulin therapy became more prevalent past study entry, this variable was allowed for in time-dependent models; however, intensive insulin therapy was not associated with incident hypertension in either sex. Although surprising, this finding is likely attributed to the fact that >60% of incident hypertension cases had developed by the sixth examination cycle (1996–1998), whereas intensive insulin therapy appears to have been largely adopted after that time period in this population.

In this cohort of individuals with long-standing type 1 diabetes, the effect of hyperglycemia, as measured by greater HbA_1c levels at study entry, on new-onset hypertension was much stronger in men compared with women. The very small number of men and women following intensive insulin therapy at study entry (4 and 21, respectively) hindered assessment of such practice in multivariable models. When included in time-dependent analyses, intensive insulin therapy did not predict hypertension in either sex, potentially because such therapy was largely adopted in the later examination cycles, by which time hypertension would have developed in a large proportion of participants.

Similar findings of a more prominent role of hyperglycemia in the incidence of hypertension in male compared with female participants were also noted when mean HbA_1c levels throughout follow-up were considered. Moreover, the observed protective effect of insulin dose per kilogram body weight against hypertension incidence was also restricted to men in these analyses. The weaker relationship between glycemia and hypertension in women did not appear to be explained by a different distribution of HbA_1c by sex.

In the current analyses, hypertension incidence was slightly higher in men than in women. Male sex per se was also associated with increased hypertension risk in the DCCT/EDIC study of individuals with type 1 diabetes (11), as well as in the biracial cohort of the Coronary Artery Risk Development in (Young) Adults (CARDIA) study (26) and the Framingham Heart Study (27), among many others in the general population. Although not all reports concur that risk is greater among men, blood pressure has generally been reported to be higher in men than in women within the general population (28), with women exhibiting a lower risk for hypertension especially in the years prior to menopause, whereas risk becomes comparable between age-matched men and postmenopausal women (29). This sex dimorphism and the observation of an important role of menopause have led to the hypothesis that sex hormones may act as modulators of vascular function and the pathogenesis of hypertension (30). Interestingly, a previous report also suggested that sex hormone–binding globulin and total testosterone are higher in male, but not female, children and young adults with diabetes compared with nondiabetic siblings, a finding apparently related to the absence of endogenous insulin (31). Whether such differences may account for the differential association between HbA_1c and hypertension incidence by sex could, unfortunately, not be evaluated, as hormone data are not available for EDC study participants. Moreover, as the associations reported in this manuscript were derived from a population where the majority of women are premenopausal, any conclusions made should be restricted to younger and middle-aged adults, as deviations from these relationships may be noted with longer follow-up, when more women would have reached menopause.

It has been suggested that the presence of hyperglycemia contributes to the excess risk of hypertension in individuals with diabetes by promoting vascular stiffness (10). Indeed, considerable evidence links hyperglycemia with increased flux through the polyol pathway and the reduction of glucose to sorbitol, increased formation of advanced glycosylation end products, and, importantly, generation of reactive oxygen species, which lead to vascular damage (10). Interestingly, our results suggest that the incidence of new-onset hypertension is higher in women compared with men among participants at better glycemic control but higher in men compared with women among those at worse glycemic control. The reason why glycemic stress may differentially affect a person’s risk for developing hypertension based on their sex is currently unclear. As HbA_1c has improved over follow-up, it is possible that the impact of male predominance for hypertension at high HbA_1c has diminished in time, in a similar manner to what we have reported for renal disease (13) where the elimination of the male excess in advanced renal disease in the more recently diagnosed cohort may be potentially linked to improved glycemic control. A large body of literature has provided evidence that in addition to structural differences, developmental/environmental stressors may provoke a distinctive physiological response by sex (32). Thus, as here, although only a weak difference in hypertension incidence exists by sex, the pathogenesis of blood pressure elevation is likely to differ between men and women. This is consistent with our earlier observation of similar coronary artery disease incidences between men and women despite differences in risk factors by sex (12) and our recent report that HDL cholesterol shows a different relationship to coronary artery disease in women (U shaped) compared with men (inverse linear) in the EDC study, potentially explaining some of the loss of female protection against heart disease seen in type 1 diabetes (14).

Despite previously reported associations between BMI and hypertension incidence in the general population (26,27) as well as in type 1 diabetes (11), measures of body fatness (WHR and, in separate models, BMI) were not selected in the final prediction models in this study. However, AER, as measured at baseline and also as a time-dependent updated mean over the follow-up period, was significantly associated with increased risk for hypertension in both sexes. The categorization of study participants into normo- or microalbuminuric at the baseline assessment was also associated with increased risk of hypertension in both men (almost fourfold increased risk) and women (twofold increased risk). AER was strongly associated with the incidence of hypertension also in the report from the DCCT/EDIC study (11). A direct association between plasma lipid concentrations and hypertension incidence was also observed among women, although similarly strong associations were not seen among men.

In conclusion, the results of this study suggest that although hyperglycemia contributes to the development of hypertension, this effect is much stronger in men compared with women with type 1 diabetes. The reasons for this difference are not clear but merit further investigation as sex differences appear to be common in the natural history of type 1 diabetes complications, including coronary artery disease and kidney disease, both of which are closely related to hypertension.

This research was supported by National Institutes of Health grants DK-34818 and DK-082900.

No potential conflicts of interest relevant to this article were reported.

T.C. researched and analyzed data and wrote the manuscript. T.J.O. researched data, contributed to the discussion, and reviewed and edited the manuscript. All authors have read and approved the final version of the manuscript. T.C. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Preliminary analyses of this study were presented at the Scientific Sessions of the American Heart Association, Orlando, Florida, 14–18 November 2009.

The authors are indebted to the participants of the Pittsburgh EDC study who have tirelessly volunteered their time for more than 20 years.