January 19, 2007 (Chicago, IL) - Angiotensin receptor blockers (ARBs) appear to have the lowest risk of any antihypertensive of precipitating diabetes, according to a new network meta-analysis [ 1].
The analysis, published in the January 20, 2007 issue of the Lancet, included data from 143 153 patients in 22 clinical trials and found that ARBs had less of an association with new-onset diabetes than even ACE inhibitors, which came in second in terms of lowest risk. Both ARBs and ACE inhibitors were protective against diabetes compared with placebo. Third were the calcium antagonists, which showed similar effects to placebo, and beta blockers and diuretics came fourth and fifth, respectively, with increased risks compared with placebo.
What is a network meta-analysis?
A network meta-analysis is a relatively new concept that allows comparison of various agents across a class, by including all relevant clinical trials and incorporating both direct comparisons of one drug with another and indirect comparisons of two drugs that may not have been directly compared. This indirect comparison is particularly relevant for ACE inhibitors vs ARBs, as these two classes have not been compared directly in a large-scale head-to-head study.
In an interview with heartwire , lead author Dr William Elliott (Rush University Medical Center, Chicago, IL) used a football analogy to explain how the network meta-analysis worked. "If two football teams were playing each other for the first time in a particular competition, you might be able to get an idea how they might do from looking how each team had performed against a third team in previous matches. If the Chicago Bears had beaten the third team by a long way but the New Orleans Saints had only drawn with that third team, we might say that the Chicago Bears seemed better than the New Orleans Saints. This is an indirect comparison. The same can be done with drugs in clinical trials, to try to tease out which one may be better than another one in a certain aspect. By including many different clinical trials with various antihypertensives in one network meta-analysis, all the various direct and indirect comparisons can be made, and all the information is put together to then rank the various drugs in terms of the one particular property you are looking for."
Elliott further noted that as long as all the trials included are randomized, the randomization process should allow any biases to be evened out. He added that network meta-analyses worked best when similar trials are included, and to judge how well the various trials are matched, the analysis is given an "incoherence value." In general, like a p value, the more zeros there are in an incoherence value the better matched are the trials and the more reliable is the result of the network meta-analysis. "The more zeros in the incoherence value, the better the data fit together. This network meta-analysis has an incoherence value of 0.000017. That is four zeros after the decimal point--a very low number, which suggests a very reliable result. Three zeros would be quite good, but two zeros would be a bit more doubtful."
In the paper, Elliot and his colleague Dr Peter Meyer (Rush University Medical Center) note that four independent meta-analyses have shown that direct inhibitors of the renin-angiotensin system reduce the risk of incident diabetes, but no comparison was attempted in these studies between ACE inhibitors and ARBs.
They used the network meta-analysis technique to estimate the relative odds of developing diabetes during long-term treatment with an initial class of antihypertensive drug, on the basis of the reported numbers of participants with or at risk of incident diabetes in randomized clinical trials. They included 48 randomized groups of 22 clinical trials with 143 153 participants who did not have diabetes at randomization.
Results--expressed as the odds ratio of developing diabetes with a given antihypertensive drug class or placebo, with the diuretic drug class used as the reference standard--are shown in the table.
Results of network meta-analysis comparing risk of new-onset diabetes with various antihypertensive agents
| Agent |
Odds ratio of diabetes |
95% CI |
p |
| ARBs |
0.57 |
0.46-0.72 |
<0.0001 |
| ACE inhibitors |
0.67 |
0.56-0.80 |
<0.0001 |
| Calcium blockers |
0.75 |
0.62-0.90 |
0.002 |
| Placebo |
0.77 |
0.63-0.94 |
0.009 |
| Beta blockers |
0.90 |
0.75-1.09 |
0.30 |
| Diuretics (reference) |
1.00 |
— |
— |
Elliott and Meyer say that the network meta-analysis technique overcomes the significant heterogeneity in traditional meta-analyses for several drug classes vs all other antihypertensive agents, because it can attribute risk across all classes of initial antihypertensive drugs, rather than being restricted to comparisons of one class vs all other classes. "The findings of this network meta-analysis are robust, in terms of both the low estimate of incoherence within the model itself and in many sensitivity analyses," they add. They note that the results are consistent with those of previous meta-analyses but go beyond them, providing "a useful and complete picture of the propensity of antihypertensive drugs to be associated with incident diabetes."
Elliott commented to heartwire that as the results of this network meta-analysis are broadly in line with what would have been expected, with ARBs and ACE inhibitors showing the best protective effects against diabetes and beta blockers and diuretics the worst diabetic profile, this should enhance confidence in the technique. "Maybe people out there will look to this paper as a reason to use network meta-analyses in other fields," he said.
Elliot added that he cannot explain why ARBs may be more protective against diabetes than ACE inhibitors, but there are now some large-scale trials ongoing that are comparing these two classes, and "when these are completed we will know."
Messerli impressed ?
Commenting on this paper for heartwire , Dr Franz Messerli (St Luke's-Roosevelt Hospital, New York) said the network meta-analysis was "thorough and comprehensive and corroborates and extends previous work." He notes that given that there are about 20 million patients on thiazide diuretics and an almost equal number on beta blockers, the increased risk of diabetes with these drugs translates into 250 000 cases of new-onset diabetes, accounting for about a quarter of the new cases of diabetes occurring every year in the US. "The prevalence of obesity, metabolic syndrome, and diabetes have doubled in the US within a decade. Hypertensive patients are at a higher risk of becoming diabetic than are matched normotensive subjects. How long are we to prescribe diuretics and beta blockers in uncomplicated hypertension? How many more hypertensive patients will have to become diabetic just because guidelines advise to use diuretics and beta blockers as first-line therapy?" he adds.
Elliott W J and Meyer P M. Incident diabetes in clinical trials of antihypertensive drugs: a network meta-analysis. Lancet 2007; 369: 201–207.
The complete contents of Heartwire , a professional news service of WebMD, can be found at www.theheart.org, a Web site for cardiovascular healthcare professionals.
Heartwire from Medscape © 2007 Medscape
Cite this: Sue Hughes. ARBs May Protect Against Diabetes Better Than ACE Inhibitors - Medscape - Jan 19, 2007.
Comments