Differing Effects of Antihypertensive Drugs on the Incidence of Diabetes Mellitus Among Patients With Hypertensive Kidney Disease FREE

The incidence and prevalence of type 2 diabetes mellitus (DM) and its complications are higher among African Americans, native Americans, and Latinos than among white Americans.^1- 5 In addition, hypertension and insulin resistance frequently coexist, often as part of the metabolic syndrome.^6- 10 Recent hypertension trials have reported a lower incidence of DM among participants treated with an angiotensin-converting enzyme inhibitor (ACEI) or an angiotensin II receptor blocker compared with other classes of antihypertensive medications.^11- 19 Only 1 trial enrolled more than 20% African Americans, and the association of the drug interventions and the risk of DM was not stratified by race.¹⁷ Therefore, it is unknown whether the lower associated risk of DM with ACEI treatment also occurs in African Americans. Furthermore, only 1 trial had more than 2 randomized drug groups¹⁷; hence, it is difficult to assess whether the net effect of ACEI treatment results from a benefit of the ACEI or from an adverse effect of the comparison therapy.

The African American Study of Kidney Disease and Hypertension (AASK) was a multicenter trial of nondiabetic African American adults with chronic kidney disease due to hypertension.²⁰ Participants were randomized to 1 of 3 classes of initial antihypertensive medications: an ACEI, a β-blocker, or a calcium channel blocker. Therefore, the trial is well suited to examine the association of these agents with the risk of DM and impaired fasting glucose (IFG) in a racial group at high risk for DM. The primary objectives of this article are to document the association of the AASK randomized therapies on the serum glucose level and on DM and IFG in this African American population and to explore potential risk factors for development of DM and IFG.

Before initiating recruitment, each participating center received approval from their local institutional review board for human research. All participants provided informed consent. The AASK study design has been previously described.^20- 21 Briefly, the AASK was a multicenter, randomized, double-blind study designed to examine the effects of 2 levels of blood pressure control and 3 classes of antihypertensive medications on the rate of progression of hypertensive nephrosclerosis. Enrollment criteria included African Americans, aged 18 to 70 years, who had hypertension (ie, a seated diastolic blood pressure >95 mm Hg). All participants had reduced renal function, defined as a prerandomization iodine I 125–iothalamate glomerular filtration rate of 20 to 65 mL per minute per 1.73 m². Individuals with a history of type 1 or type 2 DM were excluded, as were potential participants who met the criteria for DM using the 1979 criteria of the National Diabetes Data Group²² (ie, a fasting serum glucose [FSG] concentration ≥140 mg/dL [≥7.8 mmol/L] on 2 occasions or a random serum glucose concentration >200 mg/dL [>11.1 mmol/L]) during the screening phase of the study. Participants were randomized to 1 of 2 blood pressure goals (a mean arterial pressure of 102-107 mm Hg [termed the usual goal] or a mean arterial pressure of ≤92 mm Hg [termed the low goal]) and to 1 of 3 classes of initial antihypertensive medications (the β-blocker metoprolol, the angiotensin-converting enzyme inhibitor ramipril, or the calcium channel blocker amlodipine). Randomization was performed with equal allocation to the 2 blood pressure goals, and participants were randomized in a 2:2:1 ratio to the metoprolol-, ramipril-, and amlodipine-treated groups, respectively.²⁰ If the blood pressure goal was not achieved using the study drug, additional unmasked drugs were added as previously described.²¹ The amlodipine intervention was terminated approximately 1 year before the end of the trial (September 2000) on the recommendation of the Data Safety and Monitoring Board²¹; however, the participants in the amlodipine-treated group maintained their assigned blood pressure goals. The primary outcome of the study was the rate of decline (slope) of the glomerular filtration rate in milliliters per minute per 1.73 m² per year.

One thousand ninety-four individuals were followed up for a planned 3.0 to 6.4 years. Fasting serum cholesterol and glucose concentrations were measured by the AASK central laboratory at baseline and annually for the duration of the study. Participants were weighed at every study visit (ie, ≥6 times per year). At each encounter, participants were queried regarding changes in medications and health status. At baseline and at each annual visit, participants were queried by an AASK physician about their mean weekly amount of exercise. All data were reported to the AASK data coordinating center.

This analysis was designed to examine 2 main outcome variables, DM and a composite outcome of IFG or DM (IFG/DM). For this analysis, DM is defined as the first occurrence of at least 1 follow-up FSG level of at least 126 mg/dL (≥7.0 mmol/L) or a clinical diagnosis of DM by the local center. Impaired fasting glucose is defined as an FSG level of 100 to 125 mg/dL (5.6-6.9 mmol/L),²³ and IFG/DM is defined as the first occurrence of at least 1 follow-up FSG level of at least 100 mg/dL (≥5.6 mmol/L) or a clinical diagnosis of DM. Although the amlodipine-treated arm was terminated in September 2000, data for this report were retained to the end of the study (September 30, 2001).

Event rates for DM and IFG/DM were calculated as the total number of events divided by the total patient-years of follow-up. Rates were expressed per patient-year.

The effects of the randomized treatment interventions on the DM and the IFG/DM outcomes were evaluated by Cox regression analysis, stratified by the Clinical Center, controlling for prerandomization FSG level, body mass index (BMI, calculated as weight in kilograms divided by the square of height in meters), and 5 prespecified baseline factors for renal disease progression (age, sex, history of heart disease, mean arterial pressure, and urinary protein-creatinine ratio). Analyses of DM were restricted to participants with a prerandomization FSG level of 125 mg/dL or less (≤6.9 mmol/L), and analyses of IFG/DM were restricted to participants with a prerandomization FSG level less than 100 mg/dL (<5.6 mmol/L). Obesity was defined as a BMI greater than 30, and overweight as a BMI between 25 and 30.²⁴

Sensitivity analyses were performed using time-dependent Cox regression models to evaluate the effects of the randomized groups on the risk of DM and IFG/DM after adjusting for the use of thiazide diuretics during the study (defined as whether thiazide diuretics were being used at the most recent follow-up visit). We also tested whether the effects of the randomized group interventions differed among participants using thiazide diuretics at the most recent follow-up visit. Similar sensitivity analyses were performed for the use of any type of diuretic during the study.

The mean FSG concentration at baseline and at year 1 and the mean of years 1 to 3 (expressed as actual mean ± SD) were summarized by randomized group. A mixed-effects model was used to compare the adjusted mean FSG concentration at year 1 and the mean of years 1 to 3 between randomized groups, controlling for baseline FSG level, BMI, and 5 prespecified baseline factors.

Separate Cox regression models were used to relate DM and IFG/DM individually to each of the following 11 baseline factors, controlling for randomized treatment group: sex, FSG level, BMI, body weight, age at randomization, hours of exercise per week, urinary protein-creatinine ratio, mean arterial pressure, prerandomization glomerular filtration rate, and high-density lipoprotein and non–high-density lipoprotein cholesterol. A multiple Cox regression model was used to jointly relate DM and IFG/DM to each of these baseline factors, except for body weight (because of the collinearity of body weight with BMI).

Follow-up time for computation of event rates and for Cox regression models was censored at death, the occurrence of end-stage renal disease, or loss to follow-up (9 participants). All results were considered statistically significant at 2-sided P<.05 without adjustment for multiple comparisons. The expression “relative risk” (RR) is used to refer to hazard ratios in Cox regression analyses due to its greater familiarity.

Baseline FSG levels were missing for 19 participants, and 58 individuals had a prerandomization FSG level of at least 126 mg/dL (≥7.0 mmol/L). Two hundred forty-one participants had a prerandomization FSG level between 100 and 125 mg/dL (5.6-6.9 mmol/L). Therefore, data from 1017 of 1094 randomized participants were available for analysis of the incidence of DM, and data from 776 participants were available for analysis of the IFG/DM composite outcome. Prerandomization demographic and clinical features of the study participants are summarized in Table 1. Sixty-one percent of the study participants were men. More than 75% of the AASK participants were overweight or obese. On average, participants reported less than 1.5 hours of exercise per week.

The mean and median follow-up times for participants in the DM analyses were 3.8 years and 3.9 years, respectively, and for the IFG/DM analyses they were 3.2 years and 3.2 years, respectively. Antihypertensive medication use among study participants during follow-up is summarized in Table 2. Most patients needed more than 1 antihypertensive medication to achieve their assigned blood pressure goal. During the course of the study, 147 (14.5%) of 1017 participants developed DM, at a mean rate of 3.8% per patient-year (Table 3). Of these, 123 diagnoses were based on at least 1 FSG level of at least 126 mg/dL (≥7.0 mmol/L) in the AASK central laboratory. An additional 24 DM events were identified by the local centers. Most of the DM events occurred during the second and third years of follow-up. Impaired fasting glucose or DM developed in 333 (42.9%) of 776 participants with a prerandomization FSG level less than 100 mg/dL (<5.6 mmol/L), an event rate of 13.5% per patient-year. Most cases (326/333) were based on an FSG concentration of 100 to 125 mg/dL (5.6-6.9 mmol/L). Seven cases were based on a diagnosis of DM by the local centers.

The RR of DM was significantly lower in the group randomized to ramipril treatment compared with metoprolol (RR, 0.53; 95% confidence interval [CI], 0.36-0.78; P = .001) and amlodipine (RR, 0.49; 95% CI, 0.31-0.79; P = .003) treatments (Figure 1A and Table 4). Participants randomized to ramipril treatment had a lower risk for the IFG/DM outcome compared with those randomized to metoprolol treatment (RR, 0.64; 95% CI, 0.50-0.82; P<.001) and a trend toward a lower risk compared with those randomized to amlodipine treatment (RR, 0.76; 95% CI, 0.56-1.05; P = .09) (Figure 2A and Table 4). There were no significant differences between the amlodipine- and metoprolol-treated groups for DM or IFG/DM outcomes.

There was no significant difference in the incidence of DM between the low and usual blood pressure goal groups (Figure 1B and Table 4). However, for the IFG/DM outcome, the low blood pressure goal group had a significantly higher risk compared with the usual blood pressure group (RR, 1.36; 95% CI, 1.09-1.70; P = .007) (Figure 2B and Table 4).

Because the proportion of follow-up visits at which thiazide diuretics were being used was higher in the low blood pressure group than in the usual blood pressure group (14% vs 18%), the use of thiazide diuretics was a potential confounding variable in the DM and IFG/DM outcomes. Repeating the analyses in Table 4 after controlling for the use of thiazides at the most recent follow-up visit did not change the randomized group comparison results for either outcome (data not shown). The effects of the randomized drug and blood pressure interventions did not differ between participants using thiazide diuretics and those not using thiazide diuretics at the most recent follow-up visit. Similar results were observed when controlling for the use of any diuretic during follow-up visits.

At baseline, there were no significant prerandomization differences in mean FSG concentrations among the randomized groups (Table 5). Compared with the metoprolol-treated group, the mean adjusted FSG level in the ramipril-treated group was 6.2 mg/dL (0.3 mmol/L) (95% CI, 2.6-9.9 mg/dL [0.1-0.5 mmol/L]; P<.001) lower at year 1 and 3.4 mg/dL (0.19 mmol/L) (95% CI, 1.0-5.8 mg/dL [0.06-0.32 mmol/L]; P = .007) lower when averaged over years 1 through 3. Compared with the amlodipine-treated group, the mean FSG level in the ramipril-treated group was 5.1 mg/dL (0.28 mmol/L) (95% CI, 1.9-8.2 mg/dL [0.11-0.46 mmol/L]; P = .002) lower at year 1 and 3.5 mg/dL (0.19 mmol/L) (95% CI, −0.2 to 7.2 mg/dL [−0.01 to 0.40 mmol/L); P = .06) lower when averaged over years 1 through 3. No significant differences were found between the changes in FSG levels in the amlodipine- and metoprolol-treated groups or in the low and usual blood pressure groups.

For each 5-kg increment in body weight, there was an 8% higher risk of DM (P<.001) and a 5% higher risk of IFG/DM (P<.001) (Table 6). Obesity was associated with a significant 2.61-fold (95% CI, 1.53-4.47; P<.001) higher risk of DM and a 1.84-fold (95% CI, 1.35-2.52; P<.001) higher risk of IFG/DM compared with those with a BMI lower than 25. The significant association between obesity and DM and IFG/DM persisted on multiple regression analysis (Table 7). When BMI as a continuous variable was entered in the multivariable models, the RR of DM was 1.09 (95% CI, 1.03-1.15; P = .002), and the RR of IFG/DM was 1.05 (95% CI, 1.01-1.09; P = .005) for every 2-kg/m² increment in BMI.

In multivariable analysis there was a 35% (95% CI, 2%-57%) lower risk of DM among women compared with men, an association that was not present on univariate analysis. However, there was no association between sex and the risk of IFG/DM (Tables 5 and 6).

These analyses document that FSG levels increased over time in African Americans who have hypertensive kidney disease and that these patients were at high risk of developing DM and IFG/DM. Treatment with ramipril-based antihypertensive therapy was associated with a smaller rise in FSG levels compared with amlodipine-based or metoprolol-based treatment. Overall, 14.5% of participants developed DM, and 42.9% of participants developed IFG/DM. Compared with individuals randomized to metoprolol-based and amlodipine-based treatments, participants randomized to ramipril-based treatment had significantly lower FSG levels at year 1, by a mean difference of 5 to 6 mg/dL (0.28-0.33 mmol/L). This translated to a risk of DM that was approximately 50% lower. Ramipril-based therapy was also associated with a lower incidence of IFG/DM compared with metoprolol-based treatment. There was no significant difference between the effects of metoprolol and amlodipine treatments on the risk of DM and IFG/DM. As expected, DM and IFG/DM were also associated with weight gain and obesity.

The increased risk of IFG/DM in the group randomized to the low blood pressure goal was unexpected and was not observed for DM. It persisted when the analyses were repeated after adjusting for the specific use of thiazide diuretics and for the overall use of diuretics. Confounding by multiple variables (eg, sex, body weight, baseline FSG level, serum potassium level, and concomitant medications) made adjustment for diuretic dosage impractical in this study. Additional studies to determine the cause of this finding are needed.

The strengths of this study are the number of treatment arms, the racial composition of the participants, and the focus on patients with chronic kidney disease. However, the study has limitations. Because the AASK was not designed to study DM as a main outcome, data are unavailable for several variables related to DM (eg, waist-hip ratios, baseline insulin levels, and family history of DM). Second, self-reported weekly hours of exercise is an imprecise measure of physical activity, which may explain the lack of association of physical activity with the risk of DM and IFG/DM. Third, data pertaining to changes in mean FSG levels should be interpreted with caution. Because FSG was measured annually and participants with elevated FSG levels may have received interim DM treatment, the observed change in FSG concentration over time may be underestimated.

Investigators in recent trials have reported favorable effects of angiotensin-converting enzyme inhibitor or angiotensin II receptor blocker treatment on glucose and DM.^{11- 13,15- 19,25} It is difficult to compare the incidence rates of DM across studies because there was no uniformity in the diagnostic criteria and because of the diversity of diseases and diabetic risk profiles of the study populations. The AASK is distinctive because it compares multiple classes of drugs. The results of this analysis are the first to demonstrate the superiority of an angiotensin-converting enzyme inhibitor over a metabolically neutral agent such as a calcium channel blocker, suggesting that the metabolic effects of blockade of the renin-angiotensin system are more than simple neutralization of the adverse effects of β-blockers or thiazide diuretics.

The racial composition of the AASK participants is also unique. The primary results of the AASK²⁰ and those of the present analysis underscore the beneficial effects of angiotensin-converting enzyme inhibitors among African Americans with renal disease and demonstrate that these drugs have additional benefits that are independent of their blood pressure effects.

Impaired fasting glucose and insulin resistance are well-recognized risk factors for DM^25- 27 and cardiovascular events.^28- 29 Among the AASK cohort, we found a high incidence of IFG. While 29.1% of the AASK participants had evidence of IFG or frank DM at study enrollment, 59.5% had IFG or DM by the end of the study. Possible reasons for the prevalence of IFG include the high prevalence of obesity and overweight (75% of the study population), physical inactivity, hypertension, and chronic kidney disease.^{7- 10,28- 31} Excess weight, particularly obesity, was associated with a substantial risk of DM and IFG/DM. Hence, the association of a lower incidence of DM and IFG/DM with ramipril-based therapy in this population does not diminish the necessity for interventions such as exercise, weight loss, and the possible use of thiazolidinediones, which have been shown in other studies^32- 33 to reduce the incidence of DM.

The long-term clinical significance of our observations is uncertain. It is unclear whether the risk associated with the modest increases in glucose levels observed in this study is clinically equivalent to the risk associated with an adverse lifestyle (eg, weight gain and inactivity). Although the results of a study³⁴ with up to 16-years' follow-up suggested an increase in cardiovascular disease rates among patients with new-onset DM, the metabolic consequences of the use of diuretics and β-blockers in clinical trials such as the Antihypertensive and Lipid-Lowering Treatment to Prevent Heart Attack Trial,³⁵ Valsartan Antihypertensive Long-term Use Evaluation,¹⁹ and Swedish Trial in Old Patients With Hypertension-2 study,³⁶ with up to 8-years' follow-up, have not been associated with an increase in cardiovascular or cerebrovascular events. In the prospective meta-analysis of the Blood Pressure Lowering Treatment Trialists' Collaboration,³⁷ there is no indication of any difference in cardiovascular disease or coronary heart disease between subjects receiving ramipril and those receiving diuretics or metoprolol among more than 46 000 participants with more than 6000 cardiovascular disease events and more than 1800 coronary heart disease events. In the Systolic Hypertension in the Elderly Program,³⁸ subjects who developed DM in the placebo group increased their cardiovascular disease mortality hazard ratio to 1.56 (95% CI, 1.12-2.18), while subjects who developed DM and were receiving chlorthalidone or atenolol had no increase in their cardiovascular disease mortality hazard ratio (hazard ratio, 1.04; 95% CI, 0.75-1.16) after 14.3 years of follow-up.

In summary, ramipril-based therapy was associated with a significantly lower incidence of DM and IFG/DM among a group of subjects at high risk for these disorders compared with amlodipine-based or metoprolol-based therapy. Further studies are indicated to determine the mechanisms and long-term consequences of the effect associated with this angiotensin-converting enzyme inhibitor.

Correspondence: Denyse Thornley-Brown, MD, Division of Nephrology, University of Alabama at Birmingham, PB 234, 1530 Third Ave S, Birmingham, AL 35294-0007 (dtb@uab.edu).

Accepted for Publication: October 14, 2005.

Financial Disclosure: Dr Wright has received research grants, honoraria, and consultation fees from the following pharmaceutical companies: AstraZeneca, Aventis Pharmaceuticals, Bayer, Bristol-Myers Squibb Company, Eli Lilly and Company, Merck & Co Inc, Novartis Pharma AG, Pfizer Inc, Phoenix Pharmaceuticals Inc, GlaxoSmithKline, and Solvay/Unimed.

Funding/Support: This study was supported under a cooperative agreement from the National Institute of Diabetes and Digestive and Kidney Diseases, Bethesda; by institutional general clinical research center grants M01 RR00080, 5M01 RR00071, M01 RR00032, P20 RR11145, M01 RR00827, and 2P20 RR11104 from the National Institutes of Health; by the National Center on Minority Health and Health Disparities, Bethesda; and by Pfizer Inc, AstraZeneca, and King Pharmaceuticals Inc.

Role of the Sponsor: Medications used in this study were donated by Pfizer Inc, AstraZeneca, and King Pharmaceuticals Inc.

Previous Presentation: This study was presented at The American Society of Nephrology Renal Week 2003; November 15, 2003; San Diego, Calif.

Acknowledgment: We thank the AASK participants for their time and commitment to the study.

Group Information: A listing of the African American Study of Kidney Disease and Hypertension Study Group participating centers and investigators was published in JAMA (2002;288:2421).