Is the Development of Diabetes With Antihypertensive Therapy a Problem?—Pro

Abstract

Some questions about new‐onset diabetes (NOD) must still be completely addressed: 1) its incidence; 2) the possible association between NOD and some classes of antihypertensive drugs; and 3) its prognostic impact. It is well known that diuretics and β blockers can increase plasma glucose and, in available hypertension trials, diuretics and β blockers caused a higher incidence of NOD than new antihypertensive drugs. NOD heralds a high risk of major cardiovascular events, but the absolute difference between old and new drugs was too small to significantly drive the differences in cardiovascular event rates between the two groups of treatment. This evidence suggests a judicious use of drugs more frequently associated with NOD in subjects at high risk of diabetes (impaired fasting glucose, overweight, family history of diabetes, low high‐density lipoprotein cholesterol levels). The lowest effective dose of these drugs should be used, plasma glucose should be checked periodically, and concomitant lifestyle measures to prevent diabetes should be implemented with resolution.

The uncertainty surrounding the clinical impact of new‐onset diabetes (NOD) in treated hypertensive subjects is due to the paucity of data and their controversial interpretation on three main issues: 1) the incidence of NOD; 2) the association between NOD and some classes of antihypertensive drugs, namely diuretics and β blockers; and 3) the prognostic impact of NOD.

It is unquestioned that diabetes mellitus is a condition that confers an excess risk of cardiovascular (CV) disease at any level of blood pressure (BP) and other risk factors. ¹ Based on the results of some intervention trials, ² , ³ several hypertension guidelines ⁴ , ⁵ suggest that antihypertensive treatment in diabetics should be very aggressive, with BP lowered to <130/80 mm Hg, along with concomitant aggressive management of other modifiable risk factors.

The issue of NOD, however, is less obvious and, in part, elusive because most of the available studies on the prognostic impact of diabetes are based on patients with an established diagnosis of diabetes at the beginning of the study. Over the past few years, some intervention trials in patients with hypertension ⁶ , ⁷ , ⁸ , ⁹ , ¹⁰ , ¹¹ , ¹² , ¹³ , ¹⁴ , ¹⁵ showed a consistent difference in the incidence of NOD among the randomized treatment groups, usually with a higher incidence of NOD among the subjects randomized to diuretics and β blockers. Indeed, while the relative differences among the treatment groups were generally quite remarkable (with extremes ranging between 2% and 30% or more), the absolute differences among the groups were smaller, with extremes between 0.6% and 3.5%. Such discrepancy has been discussed in detail by Moser. ¹⁶ The greatest absolute differences were observed in some more recent trials. In the Valsartan Antihypertensive Long‐term Use Evaluation (VALUE), ¹⁵ for example, NOD occurred in 13.1% of patients randomized to valsartan vs. 16.4% of patients randomized to amlodipine, a percentage difference of 20%. In the Antihypertensive and Lipid‐Lowering Treatment to Prevent Heart Attack Trial (ALLHAT), ⁶ NOD occurred in 8.1% of patients taking lisinopril, 11.6% with chlorthalidone, and 9.8% with amlodipine, a 3.5% absolute difference in favor of lisinopril vs. chlorthalidone, and 1.8% in favor of amlodipine vs. chlorthalidone. In the Captopril Prevention Project (CAPPP), ⁸ Swedish Trial in Old Patients with Hypertension‐2 (STOP‐2), ¹⁴ and Second Australian National Blood Pressure (ANBP‐2) ⁷ trials, the absolute excesses in NOD incidence in the diuretic/β‐blocker group compared with the other groups were 1.0%, 0.2%, and 2.1%, respectively.

Not infrequently, NOD was one of the few measures that differed significantly among the randomized treatments in the above trials. Consequently, it received increasing attention, and even achieved the dignity of being a predefined end point in some major intervention trials. ¹⁷ , Figure 1 shows the absolute incidence of NOD, calculated per 100 patient‐years, in the above trials.

Figure 1.

Incidence of new onset diabetes in available hypertension trials. ARB=angiotensin receptor blocker; CCB=calcium channel blocker; ACEI=angiotensin‐converting enzyme inhibitor; Diur=diuretic; beta=β blocker; NORDIL=Nordic Diltiazem; LIFE=Losartan Intervention for Endpoint Reduction in Hypertension; INVEST=International Verapamil‐Trandolapril Study; INSIGHT=International Nifedipine GITS Study‐Intervention as a Goal in Hypertensive Treatment; remainder of trial acronyms are expanded and bibliographic references are given in text or Table

Indeed, most of the current debate on the clinical significance of NOD in treated hypertensive patients arises from a question that offers an apparent paradox: how can NOD be dangerous if most of these trials failed to detect a significant difference in CV outcomes among the treatment groups, despite a significant difference in the incidence of NOD among the groups?

The goal of this review is to discuss three main points: first, NOD is associated with an excess risk of CV events and therefore its occurrence should be prevented; second, diuretics and β blockers increase the risk of NOD when compared with other antihypertensive drugs; and third, despite being a harmful condition, NOD is unlikely to drive the difference in CV event rates among treatment groups in available mega‐trials.

EXCESS CV RISK ASSOCIATED WITH NOD

In a longitudinal study from Sweden, ¹⁸ subjects with established diabetes at entry had a more than two‐fold higher risk of coronary heart disease than those without diabetes, but the 48% excess risk associated with NOD was not statistically significant. In another study from Sweden, the risk of myocardial infarction after age 60 was predicted by the rise in blood glucose between the ages of 50 and 60 years in a group of hypertensive men treated mostly with diuretics and β blockers, but not in a group of subjects not treated for hypertension. ¹⁹ This study has been criticized because BP and other risk markers consistently differed between the two groups, thus exerting a confounding effect on results. In a study by Alderman et al., ²⁰ a history of diabetes doubled the risk of CV disease, but the excess risk associated with NOD did not achieve statistical significance.

In a study from our group, ²¹ 795 initially untreated hypertensive subjects were examined before and during treatment and then followed for up to 16 years. At entry, 6.5% of subjects had a prior diagnosis of diabetes and 5.8% of the remaining subjects developed NOD, with an incidence of 1.9%/yr. The incidence of NOD noted in our study was not far from the 1.6%/yr incidence noted in major hypertension trials. We defined diabetes as a fasting plasma glucose of ≥7.0 mmol/L (126 mg/dL) on repeated occasions, or current antidiabetic therapy, while impaired fasting glucose (IFG) was defined as a fasting plasma glucose level of 6.1–6.9 mmol/L (110–125 mg/dL). ²² At entry, IFG was present in 42% of subjects with future diabetes, vs. 7% of those in whom diabetes did not develop (p<0.001). In the subjects with NOD, the crude rate of CV events after diagnosis of NOD was 3.9/100 patient years. Notably, the incidence of events did not differ between the subjects with NOD and those with diabetes at entry (Figure 2). We speculated that the long duration of follow‐up in our study (up to 16 years) allowed a full disclosure of the clinical complications of new diabetes. ²¹

Figure 2.

Cardiovascular events in treated hypertensive subjects without diabetes (A), new‐onset diabetes (B), and previously known diabetes (C). Adapted from Hypertension. 2004;43:963–969. ²¹

DIURETICS, β BLOCKERS, AND NOD

In our study, ²¹ antihypertensive drugs administered during follow‐up included a diuretic in 53.5% of patients with NOD vs. 30.4% of those without NOD (p=0.002). More than 90% of diuretic users received hydrochlorothiazide or chlorthalidone in a dose range of only 12.5–25.0 mg/d, and a minority were given furosemide or potassium‐sparing diuretics. The participants in whom NOD developed during follow‐up, however, also received calcium channel blockers (CCBs) (p<0.05) and angiotensin‐converting enzyme inhibitors (ACEIs) (p<0.05) more frequently than those in whom NOD did not occur. This was probably due to the higher BP levels at entry in the former than in the latter group. In a multivariate logistic regression analysis, however, only glucose concentration at the baseline visit (p<0.0001) and exposure to diuretics at the follow‐up visit (p=0.004) were independent predictors of NOD (Figure 3).

Figure 3.

Probability of new‐onset diabetes in relation to baseline glucose concentration and exposure to diuretics in treated hypertensive subjects. Adapted from Hypertension. 2004;43:963–969. ²¹

It is well known that diuretics can increase plasma glucose and that glucose tolerance usually improves when diuretics are discontinued. ²³ , ²⁴ , ²⁵ In addition, β blockers may increase plasma glucose. In the Atherosclerosis Risk in Communities (ARIC) ²⁶ study, β blockers were associated with a 25%–28% higher risk of diabetes mellitus compared with the absence of antihypertensive drugs, and such difference was statistically significant.

Patients with IFG or other conditions associated with a risk of NOD may be expected to develop NOD more frequently when exposed to diuretics and/or β blockers, particularly when relatively high doses of these drugs are used. In the Prospective Cardiovascular Munster (PROCAM) ²⁷ study, independent predictors of NOD were IFG, overweight, family history of diabetes, hypertension, and low high‐density lipoprotein (HDL) cholesterol levels. Of note, IFG itself is a condition of increased CV risk in hypertensive subjects. ²⁸

NOD AND OUTCOME IN HYPERTENSION TRIALS

The duration of follow‐up in individual trials might exert some impact on the incidence of NOD. For example, a short duration of follow‐up could limit the detection of the diabetogenic effect of one drug class over another, while a long duration of follow‐up might increase the likelihood of crossover between treatments, with potential dilution of differences. Indeed, trials with a long duration of follow‐up like CAPPP, the Study on Cognition and Prognosis in the Elderly (SCOPE), ¹³ and STOP‐2 (duration of 6.1, 5, and 6 years, respectively) showed a small absolute difference in NOD between the groups, ranging from 0.1 to 1.0%.

In a meta‐analysis, Opie and Schall ²⁹ estimated the number of patients needed to treat for 4 years with ACEIs, CCBs, or angiotensin receptor blockers (“new” antihypertensives) rather than diuretics and β blockers (“old” antihypertensives) to prevent one case of new diabetes. According to this analysis, one case of NOD would be prevented by treating 60–70 patients for about 4 years with new rather than old antihypertensive drugs. The logical next step would be to estimate how many events specifically attributable to NOD would be prevented by treating patients with new rather than old antihypertensive drugs. In the Progetto Ipertensione Umbria Monitoraggio Ambulatoriale (PIUMA) study, ²¹ the rate of major CV events was 3.90/100 patient‐years in the subset with NOD vs. 0.97/100 patient‐years in the subset without NOD. Therefore, we tried to combine the number needed to treat calculated by Opie and Schall (100 cases of new diabetes prevented for every 6000–7000 patients treated with new rather than old drugs for 4 years) with the PIUMA prognostic estimate. About 15.6 major CV events associated with NOD would be prevented by treating 6000–7000 patients with new rather than old antihypertensive drugs for about 4 years (i.e., one event prevented for every 385–449 patients treated for 4 years). Of particular note, this estimate assumes a constant rate of CV events beginning just after the diagnosis of NOD—and therefore should be applied with caution.

Our estimate suggested that NOD was unlikely to significantly drive the differences in CV event rates between the groups treated with new vs. old antihypertensive drugs in available hypertension trials. Four conclusions should be considered: 1) overall NOD incidence is 1.5–2/100 patient‐years; 2) one case of NOD occurs in every 50–60 initially nondiabetic patients treated with new rather than old antihypertensive drugs for about 4 years; 3) the event rate is 3.5–4/100 patient‐years after the diagnosis of NOD; and 4) one CV event related to NOD is prevented for every 385–449 patients treated with new rather than old antihypertensive drugs for 4 years. These numbers suggest that the absolute number of NOD cases and, most importantly, CV events associated with NOD, in the groups exposed to old rather to new antihypertensive drugs, has been too small in trials lasting from 3 to 6 years to exert a consistent impact on results. The subjects at increased risk of NOD due to obesity or IFG, however, might receive a greater benefit from new rather than old antihypertensive drugs both in terms of prevention of new diabetes and, possibly, related prognostic implications.

CONCLUSIONS

In available hypertension trials, diuretics and β blockers caused more cases of NOD when compared with new antihypertensive drugs, but the absolute difference was too small to cause an important disparity in CV outcome between the groups over the follow‐up period of these studies, usually ranging between 3.5 and 6 years. From a clinical standpoint, we suggest that the risk of NOD in treated hypertensive subjects should not be dismissed on the basis of the above considerations from hypertension trials. In fact, regardless of the drugs potentially inducing its development, NOD heralds a high risk of major CV events, which approximates 4%/yr. These considerations should prompt a judicious use of drugs more frequently associated with NOD, namely diuretics and β blockers, in subjects at high risk of diabetes (IFG, overweight, family history of diabetes, low HDL cholesterol levels). The lowest effective dose should be used, plasma glucose should be checked periodically, and concomitant lifestyle measures to prevent diabetes should be implemented with resolution.

Acknowledgment and disclosure: We gratefully thank Miss Francesca Saveri for secretarial assistance. This work has been supported by the Associazione Umbra Cuore e Ipertensione, Perugia, Italy.