Final outcome results of the multicenter isradipine diuretic atherosclerosis study (MIDAS)

Authors: Borhani N, Mercuri M, Borhani P, Buckalew V, Canossa-Terris M, et al.
Source: JAMA. 276:785-91. September 11, 1996.
Institutions: UCD; University of Nevada; Bowman-Gray; Miami Heart Research Institute; Wayne State; Preventive Cardiology Inc; Scientific Application Co. Inc.
Financial support: Sandoz Pharmaceuticals.



Some studies in animals have suggested an anti-atherogenic effect of calcium channel blockers. This study was designed to test the hypothesis in humans. Hypertensive patients were randomized to treatment with hydrochlorothiazide or isradipine, a dihydropyridine calcium antagonist (DynaCirc ®, Sandoz). B-mode ultrasound measurement of the carotid artery intimal-medial thickness was used as an index of atherosclerosis progression.




In an accompanying editorial, Aram Chobanian, MD (Boston University School of Medicine) notes that several design problems may have contributed to the failure of MIDAS to detect any effect of isradipine on the rate of progression of carotid atherosclerosis. The sample size was based on the rate of progression of IMT in a different population (patients with hypercholesterolemia, who were excluded in this trial); enalapril was added to a quarter of patients in both groups; systolic BP was better controlled in the HCTZ group; problems with ultrasound interpretation required a post-hoc correction.

As for the increase in vascular events, in particular angina, Dr. Chobanian notes that a similar effect has been suggested in other studies which have looked at the use of calcium channel blockers for the treatment of coronary disease. Some evidence points to an adverse effect of short-acting dihydropyridines only, some to an effect of all calcium channel blockers. Much of this data is case-control and thus inconclusive, and MIDAS was not designed to look at these issues specifically. Nevertheless, the results do contribute to the evidence that caution should be exercised when prescribing calcium channel blockers, that the long-acting and non-dihydropyridines should be considered preferentially and that first-line therapy for hypertension should probably include other types of agents.


A digression on the topic of endpoints

This study illustrates the concepts of primary and secondary endpoints (and other measurements). When a study result is significant with a p-value of 0.05, there is a 5% probability that it could be due to chance alone. The more results that are obtained, the higher the likelihood that at least one of them will be due to chance. For example, with 14 comparisons, the probability is better than 50% that one of them will attain a significance of p<0.05 due to chance alone. One way to counteract this problem is to demand a higher level of significance (such as p<0.01) whenever multiple comparisons are made. The problem with this approach is that it reduces the likelihood that a true difference will be uncovered, unless the sample size is increased. Thus, the number of parameters that can be investigated by a study is limited by the sample size. Careful consideration must be given to deciding on the study's endpoints.

The primary endpoint(s) is the main hypothesis the study is designed to look at. In general, the sample size will be determined such that the study will have adequate power to confirm or exclude this primary endpoint (to a given statistical confidence level). In the paper reviewed here, the primary endpoint was the rate of increase in IMT over three years. The sample size was determined such that there would be a 90% probability of uncovering a 30 to 40% decrease in this rate (at a p-value of 0.05).

Secondary endpoints are other aspects that are being investigated. Since multiple secondary endpoints are often specified, a higher level of significance is sometimes required. For this reason, and because secondary endpoints frequently involve subgroup analyses, a positive result remains significant but a negative one does not exclude a true difference (i.e. the study will usually not have adequate power to rule out a true difference as far as the secondary endpoints are concerned).

Finally, any other data that is gathered and compared which is neither a primary nor a secondary endpoint must be viewed with much caution. The risk of coming up with "statistically significant" results due to chance alone increases with the amount of data gathered. Anything that turns up with post-hoc analysis should be viewed as tentative and subject to confirmation by other studies.

In the paper presented here, the increase in vascular events was neither a primary, nor a secondary endpoint; as a result, it needs to be investigated further by other studies. However, this result did not occur in a vacuum -- previous case-control investigations have suggested a similar problem with some calcium channel blockers. This, in turn, reduces the probability that we are dealing with merely a statistical fluke and lends credence to its validity as a problem. As should be apparent from the above considerations, the possibilities of using statistics to back up a pro- or contra- calcium channel blocker position are practically endless. What is needed are more studies (some in progress) and prudent, common sense.

September 15, 1996

Reader comments

Date: Wed, 18 Sep 1996
From: "Gilbert W. Gleim" <>

I would like to point out a few things about the MIDAS study.

1. The group assigned to isradipine had about 5 times more angina at baseline, according to table 1 of the article. That was significant at P=.10 (for those readers who don't know the meaning of a Z score), a strong trend. Under "major vascular events" at the end of the study, angina was the only event found to be statistically higher in the isradipine group. It was approximately 3.5 times higher at this point. Obviously, close inspection of table 4 reveals that the only reason "major vascular events" were greater in the isradipine group is due to inclusion of angina as a "major vascular event".

2. The study was not powered to find the difference in rate of progression of IMT observed. The study was designed around a rate of progression of .15 to .22 mm/year in the diuretic group. In fact, the diuretic group did not attain that degree of change over the course of 3 years. This means that a larger sample size would have been necessary. It should reiterated that the isradipine group had less total progression than the hydrochlorothiazide group over the course of three years, .121 v .149 respectively, P=.02. It seems to have worked.

I would urge all of your readers to view the authors' conclusions of this study with a great deal of skepticism.

Gilbert W. Gleim, Ph.D.
Director of Research
NISMAT, Lenox Hill Hospital
130 East 77th Street
NY, NY 10021
212-434-2687 (fax)

Submit a comment about this article



References related to this article from the NLM's PubMed database.

Return to Journal Club home page

Site Meter