The authors previously reported that addition of valsartan ameliorated the negative metabolic effects of hydrochlorothiazide in obese hypertensive patients through an enhanced postprandial insulin response. In this secondary analysis, the authors tested whether this enhanced insulin response to valsartan/hydrochlorothiazide was influenced by serum potassium levels, which were reduced to a lesser extent, when compared with amlodipine/hydrochlorothiazide. Results showed that the early insulin response with valsartan plus hydrochlorothiazide occurred regardless of serum potassium levels. Heightened insulin response was, however, not significantly different when patients with normal potassium (>3.9 mEq/L) at baseline and low potassium (≤3.9 mEq/L) at the end of the study were compared with the amlodipine/hydrochlorothiazide group. Despite the influence of serum potassium on insulin secretory response to a glucose challenge, the addition of valsartan maintained normoglycemia in patients given hydrochlorothiazide. Thus, the metabolic response to hydrochlorothiazide was improved with addition of valsartan through an enhanced insulin response that was not greatly affected by changes in potassium levels.
Several large, randomized, controlled trials with diuretics as first-line agents or as add-on therapy have reported increases in the incidence rates of new-onset diabetes. For example, the incidence of new-onset diabetes in the Antihypertensive and Lipid-Lowering Treatment to Prevent Heart Attack Trial (ALLHAT) was 11.6% in the group randomized to chlorthalidone vs 9.8% and 8.1% in the amlodipine and lisinopril groups, respectively.1 Similarly, in the Anglo-Scandinavian Cardiac Outcomes Trial–Blood Pressure Lowering Arm (ASCOT-BPLA), new-onset diabetes was 30% lower in the amlodipine-based therapy group (amlodipine plus perindopril as needed) than in the atenolol-based therapy group (atenolol plus bendroflumethiazide and potassium [K+] as needed).2 The perceived mechanism by which diuretics precipitate diabetes is not well established, however, but it is thought to be related to the reduction in serum K+. Studies with the addition of a renin-angiotensin system (RAS) inhibitor to diuretic therapy have reported a lower risk of new-onset diabetes and it has largely been attributed to lessening of the diuretic-induced reduction in serum K+ levels.3,4 In the Valsartan Antihypertensive Long-Term Use Evaluation (VALUE) study in high-risk hypertensive patients (N=15,245), the rate of new-onset diabetes was 23% higher among those treated with amlodipine (±hydrochlorothiazide [HCTZ] and other antihypertensive agents) compared with valsartan (±HCTZ and other antihypertensive agents), and it was thought that RAS inhibition helped minimize the negative metabolic effect of HCTZ.5 Not all evidence, however, supports the hypothesis of the role for RAS inhibition for affecting the onset of diabetes with diuretics.6
Traditionally, it is recognized that the reduction in serum K+ levels (range 0.2–0.6 mmol/L) with diuretic therapy increases the risk for the development of diabetes as hypokalemia can affect insulin secretion and/or peripheral insulin resistance.4,7,8 Diuretic-induced reductions in serum K+ levels can be mitigated with the combined use of a RAS inhibitor; however, previous studies have not shown consistent correlations between changes in serum K+ levels and changes in fasting and postprandial glucose and insulin levels.7,9,10
Previously, in the Valsartan and Hydrochlorothiazide in Hypertensive Abdominally Obese Patients (VITAE) study, combination therapy with valsartan/HCTZ reduced blood pressure (BP) to a similar extent as that observed with amlodipine/HCTZ treatment in centrally obese, nondiabetic hypertensive patients.11 Individuals treated with valsartan/HCTZ had a reduced hyperglycemic response to a glucose challenge relative to patients treated with amlodipine/HCTZ, which was attributed to a heightened early insulin response in valsartan recipients. We conducted a post-hoc analysis of the VITAE study to identify whether the mitigating effect of valsartan on the diuretic-induced hyperglycemic response and the heightened insulin response associated with valsartan were influenced by study serum K+ levels.
We conducted a post hoc analysis of the 16-week, randomized, double-blind, forced-titration VITAE study, which was conducted in nondiabetic patients with hypertension and abdominal obesity. Methods for this study have been previously described in detail11 and are briefly summarized here. The study protocol was approved by the ethics committee or institutional review board at each center, and the study was conducted according to the ethical principles of the Declaration of Helsinki. All patients provided written informed consent.
Outpatients 40 years and older with central obesity (waist circumference >40 inches in men [>35 inches in Asian American men], >35 inches in women [>31 inches in Asian American women]) and stage 1 or 2 hypertension (mean sitting systolic BP ≥150 mm Hg and <180 mm Hg, mean sitting diastolic BP <110 mm Hg) were eligible for participation. Exclusion criteria included receipt of >3 antihypertensive medications at the time of screening, inability to discontinue all prior antihypertensive medications safely during the washout period, and weight loss of >10 pounds during the screening/washout period. Patients with type 1 or 2 diabetes, a fasting plasma glucose level ≥126 mg/dL, or a serum K+ level <3.5 mEq/L or >5.5 mEq/L were also excluded. All serum K+ values were measured using a central laboratory (CRL. medinet, Lenexa, KS).
Briefly, after a washout period of up to 4 weeks, eligible patients were randomized to receive either valsartan/HCTZ 160/12.5 mg or HCTZ 12.5 mg alone.11 After 4 weeks taking this treatment, patients were force-titrated to valsartan/HCTZ 320/25 mg and HCTZ 25 mg, respectively. Patients in the valsartan/HCTZ group remained at the same dose (320/25 mg) through the end of the 16-week study. At weeks 8 and 12, patients in the HCTZ group received add-on amlodipine 5 mg and 10 mg, respectively, to minimize BP differences between the 2 treatment groups at study completion. All study medication was administered once daily.
Metabolic and Inflammatory Marker Assessments
An oral glucose tolerance test (OGTT) was performed at baseline and at the end of the study (EOS; week 16) after an overnight fast to assess fasting and postprandial plasma glucose and serum insulin levels, as previously described.11 In addition, the insulinogenic index,12 a measure of early insulin response (30 minutes) derived from OGTT data, was determined at baseline and EOS.
All analyses were performed after stratification of patients into 3 groups according to the following baseline/EOS serum K+ levels: (1) baseline and EOS K+ >3.9 mEq/L, (2) baseline K+ >3.9 mEq/L, EOS K+ ≤3.9 mEq/L, and (3) baseline K+ ≤3.9 mEq/L, EOS K+ >3.9 or ≤3.9 mEq/L. The 3.9-mEq/L cut-off was chosen because it represented the baseline median serum K+ level for the study; this concentration of serum K+ is consistent with previous findings in nondiabetic, abdominally obese, hypertensive patients.9 Within each stratified subgroup, demographic and baseline characteristics were compared between the 2 treatment groups using a 2-sample t test for continuous variables and chi-square test for categorical variables. For metabolic parameters, within-treatment changes from baseline to EOS were analyzed using a paired t test, and between-treatment differences were analyzed using an analysis of covariance model. Estimated correlations between change from baseline in serum K+ level and change from baseline in levels of fasting plasma glucose, postprandial glucose, fasting serum insulin, postprandial insulin, and insulinogenic index were evaluated using the Spearman correlation coefficients.
Patient disposition and baseline demographics for the overall study population (N=412) were previously reported.11 Briefly, the intent-to-treat population comprised 401 patients (197 valsartan/HCTZ, 204 amlodipine/HCTZ); 348 patients (170 valsartan/HCTZ, 178 amlodipine/HCTZ) completed the study. Major findings included statistically significant increases in fasting and 2-hour glucose levels (P<.05) with amlodipine/HCTZ (but not with valsartan/HCTZ) and substantially greater increases in postprandial insulin levels with valsartan/HCTZ than with amlodipine/HCTZ (P=.001). In both treatments, serum K+ levels were similar at baseline in the valsartan/HCTZ (4.24 mEq/L) and amlodipine/HCTZ (4.30 mEq/L) group and were reduced significantly at EOS (P<.05), but reductions in serum K+ levels were smaller in the valsartan/HCTZ than in the amlodipine/HCTZ group (−0.25 mEq/L vs −0.47 mEq/L, respectively; P<.001).
In this post hoc analysis, patients were stratified by baseline/EOS serum K+ levels. Demographic and baseline characteristics for both treatment groups, stratified by baseline and EOS serum K+ levels, are shown in the Table. A total of 385 patients had data available for analysis. At baseline, the median serum K+ level was 3.9 mEq/L. Among patients with serum K+ levels >3.9 mEq/L at baseline (n=143 valsartan/HCTZ; n=150 amlodipine/HCTZ), a greater percentage randomized to amlodipine/HCTZ (58% [87 of 150]) vs valsartan/HCTZ (44% [63 of 143]) had reductions in serum K+ levels to ≤3.9 mEq/L at EOS. Overall, mean age was approximately 56 years, and most patients were female (65%), Caucasian (52%), and had the cardiometabolic syndrome (70%); 30% were prediabetic (fasting plasma glucose ≥100 mg/dL and <126 mg/dL). Patients who were diabetic at the screening visit were excluded from the study as per the study entry criteria; however, some patients (n=22) were included in the study who were prediabetic prior to randomization, but had fasting plasma glucose values in the diabetic range at the randomization visit.
Table TABLE. Demographic and Baseline Characteristics Stratified by Serum K+ Levels at Baseline and EOS
Serum K+ Level and Drug Treatment
Baseline: >3.9 mEq/L EOS: >3.9 mEq/L
Baseline: >3.9 mEq/L EOS: ≤3.9 mEq/L
Baseline: ≤3.9 mEq/L EOS: >3.9 or ≤3.9 mEq/L
Abbreviations: Aml, amlodipine; eGFR, estimated glomerular filtration rate; EOS, end of study; HbA1c, glycated hemoglobin; HCTZ, hydrochlorothiazide; K+, potassium; SD, standard deviation; Val, valsartan. aFasting plasma glucose <100 mg/dL at baseline. bFasting plasma glucose ≥100 mg/dL and <126 mg/dL at baseline. cFasting plasma glucose ≥126 mg/dL at baseline. dP<.05 for between-treatment comparison.
Age, y, mean (SD)
≥65 y, No. (%)
Diabetes status, No. (%)
eGFR, mL/min/1.73 m2, mean (SD)
HbA1c, %, mean (SD)
Baseline serum K+, mEq/L
EOS serum K+, mEq/L
The fasting plasma glucose levels were elevated in the amlodipine/HCTZ group at EOS compared with the valsartan/HCTZ group in all serum K+ subgroups. Statistical significance was reached in the amlodipine/HCTZ group (baseline K+ >3.9 mEq/L, EOS K+ ≤3.9 mEq/L subgroup) in which a mean (standard error [SE]) increase of 4.4 (1.9) mg/dL was observed (P=.023). In this subgroup, the least-square mean (LSM) difference in change from baseline to EOS between the 2 treatment groups was 5.2 mg/dL (P=.030). A similar pattern, albeit not achieving statistical significance, was observed in patients with serum K+ levels >3.9 mEq/L at baseline and EOS; mean (SE) fasting plasma glucose increased by 3.3 (2.2) mg/dL in the amlodipine/HCTZ group (P=.14) and declined by 0.9 (1.3) mg/dL in the valsartan/HCTZ group (P=.50).
Across the serum K+ stratification subgroups, patients taking amlodipine/HCTZ showed greater increases from baseline to EOS in postprandial plasma glucose at most time points after OGTT challenge than patients taking valsartan/HCTZ (Figure 1A, Figure 1B, Figure 1C). Treatment with amlodipine/HCTZ was associated with significant (P<.05) increases from baseline to EOS in postprandial plasma glucose (assessed by area under the curve [AUC] from 0 to 120 minutes following glucose administration) in all 3 serum K+ stratification subgroups (Figure 2A). The elevations in mean AUC values ranged from 1065 mg/dL×minutes (baseline and EOS K+ >3.9 mEq/L subgroup) to 2087 mg/dL×minutes (baseline K+ ≤3.9 mEq/L, EOS K+ >3.9, or ≤3.9 mEq/L subgroup). However, treatment with valsartan/HCTZ was associated with no significant change from baseline in the postprandial glucose response in 2 of the serum K+ stratification subgroups (baseline and EOS K+ >3.9 mEq/L; baseline K+ >3.9 mEq/L, EOS K+ ≤3.9 mEq/L); in both of these subgroups, between-treatment differences were significant (P<.05). The relative between-treatment difference in postprandial plasma glucose AUC0–120 min was similar across all 3 serum K+ stratification subgroups. The correlation of change in serum K+ level to change in postprandial glucose level or change in AUC glucose was ≤0.1 and nonsignificant for both treatment groups.
In each serum K+ stratification subgroup, fasting serum insulin levels increased slightly from baseline to EOS in both treatment groups (0.6–7.2 μU/mL in the valsartan/HCTZ group, 2.6–4.8 μU/mL in the amlodipine/HCTZ group). The only significant increase was seen with amlodipine/HCTZ treatment (baseline and EOS K+ >3.9 mEq/L subgroup) in which a mean (SE) increase of 4.8 (2.2) μU/mL was observed (P=.033). There were no between-treatment differences that were statistically significant.
Across the serum K+ stratification subgroups, patients taking valsartan/HCTZ showed greater increases from baseline to EOS in postprandial serum insulin as early as 30 minutes after OGTT challenge than patients taking amlodipine/HCTZ (Figure 1D, Figure 1E, Figure 1F). Significant (P<.05) increases from baseline to EOS in postprandial serum insulin AUC0–120 min were noted in each treatment group and serum K+ stratification subgroup (Figure 2B). The insulin response to a glucose challenge, however, was consistently higher with valsartan/HCTZ treatment than with amlodipine/HCTZ treatment. The difference achieved significance in the subgroup of patients with baseline and EOS K+ >3.9 mEq/L (P=.016), but the magnitude of the insulin response to valsartan/HCTZ was attenuated in the subgroup of patients with baseline K+ >3.9 mEq/L and EOS K+ ≤3.9 mEq/L such that the between-treatment difference was not significant (P=.20).
The early insulin response (30 minutes) to valsartan/HCTZ was confirmed by increases from baseline to EOS in insulinogenic index across all 3 serum K+ stratification subgroups. The response at 30 minutes was heightened in the subgroup of patients with baseline K+ >3.9 mEq/L and EOS K+ ≤3.9 mEq/L (mean [SE] increase of 22.8 [9.5] with valsartan/HCTZ [P=.020] vs 1.0 [2.0] with amlodipine/HCTZ [P=.64]; P=.012 between treatments). Correlation analysis of change from baseline in serum K+ level and change from baseline in postprandial insulin level and insulinogenic index showed that all Spearman correlation coefficients were ≤0.1 and nonsignificant.
The results of VITAE demonstrated the beneficial metabolic effect of the addition of an angiotensin receptor blocker to diuretic therapy.11 Specifically, the addition of valsartan to HCTZ overcame HCTZ-induced increases in fasting and postprandial plasma glucose levels in obese hypertensive patients. Since a significantly greater decrease in serum K+ levels was observed in the HCTZ plus amlodipine group in the primary analysis, we performed this secondary analysis to examine relationships between serum K+ and insulin secretion and glucose levels to explore the possibility that valsartan mitigated the negative metabolic effects of HCTZ through effects on K+.
The primary finding of this post hoc analysis was that valsartan mitigated the thiazide effect on fasting and postprandial glucose levels through an enhanced insulin response that was not greatly influenced by serum K+ levels. Specifically, fasting and postprandial plasma glucose levels were increased in the amlodipine/HCTZ group but glucose levels were maintained similar to baseline levels in the valsartan/HCTZ group, irrespective of serum K+ levels. In patients with normal serum K+ levels at baseline and EOS, there was a heightened postprandial insulin response (AUC) in the valsartan/HCTZ group. The magnitude of the postprandial insulin response (AUC) in the valsartan/HCTZ group was attenuated in patients with normal serum K+ levels at baseline and low serum K+ levels EOS, but this subgroup still experienced a heightened early postprandial insulin response relative to the amlodipine/HCTZ group. Despite the influence of serum K+ levels on the insulin secretory response to a glucose challenge, the addition of valsartan still maintained normoglycemia in this population. Additionally, changes in serum K+ levels were not correlated to changes in glucose and insulin. The lack of correlation does not exclude the role of serum K+ (due to sample size), but it may only contribute as an additional factor.
Previous studies have reported that diuretic therapy affects glucose metabolism through reductions in serum K+ levels. For example, Weinberger and colleagues13 provided early evidence of the influence of RAS blockade on diuretic-associated metabolic effects in hypertension and identified a potential role of serum K+ in the mechanism of this effect. In their comparison of 207 hypertensive patients randomized to either captopril or HCTZ or the combination, significant increases in glucose (7.2%; P<.05) and decreases in serum K+ levels (15.2%; P<.05) were reported in patients randomized to HCTZ alone but not in patients randomized to captopril alone. The addition of captopril to HCTZ blunted these effects, with the combination-therapy group demonstrating higher mean serum K+ levels relative to the HCTZ-monotherapy group (P<.05) and lower glucose levels (P value not provided). This finding for combined RAS blockade with diuretic therapy has been a consistent finding but the mechanism through which it exerts its effect(s) has been elusive.3,14 Primarily, the relationship between the metabolic effect of diuretics and reduction in serum K+ levels has not been consistent. The hypothesis that low K+ levels could impair insulin secretion and subsequently increase plasma glucose remains incompletely understood and does not adequately explain the findings in this study.7,10 Diuretic-induced hypokalemia can also affect peripheral insulin sensitivity by depleting muscle K+ levels.15 Future studies need to assess whole-body K+ levels along with plasma K+ levels to better evaluate the impact of hypokalemia on glucose metabolism and to investigate other possible mitigating factors separate from plasma K+.
Additional evidence suggestive of a role for factors other than serum K+ was provided by the 12-week, randomized, double-blind, 3-way crossover, placebo-controlled mechanisms for the Diabetes Preventing Effect of Candesartan (MEDICA) study (N=22).9 In this study, fasting serum K+ was lower following diuretic (HCTZ) vs candesartan therapy (3.6 mEq/L vs 3.9 mEq/L, respectively), and peripheral insulin sensitivity, as assessed by an euglycemic hyperinsulinemic clamp, was impaired following HCTZ administration (∼20% vs candesartan and placebo); however, fasting glucose and insulin response to an oral glucose challenge did not vary significantly among treatments and no correlation between reduction in serum K+ level and differences in insulin sensitivity was detected.
Previous valsartan studies have also demonstrated the mitigating effects of RAS blockade on HCTZ-induced metabolic dysregulation.5,16,17 The results of the VALUE trial suggest that mechanistic factors other than serum K+ contribute to the significant decrease in incidence of new-onset diabetes vs amlodipine.5,16 Despite reductions in the serum K+ level in the amlodipine (−0.2 mmol/L) vs the valsartan-based treatment groups, the lack of hypokalemia associated with add-on HCTZ suggests that another mechanism may well explain the between-treatment differences in new-onset diabetes.5,16 Our results are similar in that patients taking amlodipine/HCTZ had higher fasting glucose with lower serum K+ levels, but, even after stratification by serum K+, the negative metabolic effects of HCTZ in the amlodipine/HCTZ group were similar and the enhanced postprandial insulin response of valsartan was preserved across all K+ subgroups. This novel finding with combined RAS inhibition and diuretic therapy has not been reported before and the results support previous studies that have suggested diuretic therapy also affects peripheral insulin sensitivity and that RAS inhibition can improve pancreatic function.11 A previous study reported an improved early-phase insulin response (ie, increased insulinogenic index after 30 minutes in response to a glucose challenge) after 12 weeks of candesartan treatment in hypertensive patients with impaired glucose tolerance.18 Preclinical studies have suggested that inhibition of the RAS can improve pancreatic islet blood flow and islet morphology, resulting in improvements in first- or early-phase insulin secretion.19,20
We recognize that the current analysis has a number of limitations that may have influenced the results. First, the original study was of relatively short duration (16 weeks) and was not powered for comparison based on stratification into K+ subgroups. Data from adequately powered long-term studies are needed to confirm these findings. Second, the study enrolled obese hypertensive patients, thus results may not be the same for prediabetic patients with impaired glucose tolerance and/or impaired fasting glucose patients. Finally, amlodipine was assumed to be metabolically neutral.
The addition of valsartan to HCTZ was associated with improved fasting and postprandial glucose levels achieved primarily through increased insulin response. When the role of serum K+ was examined, we saw no relation between K+ and insulin/glucose.
Acknowledgments and disclosures: The authors wish to express their appreciation to Susan Ritter of Novartis Pharmaceuticals Corporation for expert assistance in project management. ClinicalTrials.gov identifier: NCT00439738. The study was supported by Novartis Pharmaceuticals Corporation. The authors wish to thank Mary A. Tom, PharmD, and Michael S. McNamara, MS, of Oxford PharmaGenesis, Inc, Newtown, Pennsylvania, for editorial and writing assistance. The editorial and writing assistance was funded by Novartis Pharmaceuticals Corporation. Dr Deedwania has served as a consultant and speaker for Novartis Pharmaceuticals Corporation, Forest Pharmaceuticals, GlaxoSmithKline, and Pfizer, and has received research support from Novartis Pharmaceuticals Corporation and AstraZeneca Pharmaceuticals. Drs Zappe, Purkayastha, and Samuel are employees of Novartis Pharmaceuticals Corporation. Dr Egan has served as a consultant and speaker for Novartis Pharmaceuticals Corporation, Pfizer, and GlaxoSmithKline, and has received research support from Novartis Pharmaceuticals Corporation and AstraZeneca Pharmaceuticals. Dr Sowers has National Institutes of Health and Veteran’s Administration funding and has served as a consultant for Novartis Pharmaceuticals Corporation and Forest Pharmaceuticals; research funding grants were provided to University of Missouri by Novartis Pharmaceuticals Corporation and Forest Pharmaceuticals.