Dr. Rashdan cites the results of a recent trial in which probucol was used to limit restenosis after angioplasty [1]. This report was published on 7 August 1997, well after our manuscript had been accepted and typeset for publication. We did not include the abstract form of this study because the number of probucol recipients reported in the abstract was small (n = 91) and the reduction in restenosis (35.2% in the placebo group and 22.6% in the probucol group; P = 0.055) was of borderline significance [2]. Experience with restenosis trials has repeatedly shown that drugs, thought to be beneficial in small studies, often have no effect when examined in large trials. In this case, however, the full-length paper reported on 180 probucol recipients and showed significant results. The value of an abstract is sometimes hard to judge. Data presented in some abstracts are excellent and result in major publications; others, however, do not survive the peer review process or are substantially different when published in full-length form. For this reason, some journals do not allow the citation of abstracts and others only allow the citation of abstracts less than 2 years old.

The letter from Drs. McCullough and O'Neill requires two comments. First, these authors express concern about what constitutes an appropriate literature search for a review article. At the time we wrote our paper, their study of contrast nephropathy after coronary interventions had been published twice in abstract form. Cardiology abstracts are not listed in any of the major computerized databases, including MEDLINE. The only way to search abstracts is by manually reviewing abstract booklets or finding them in the reference list of a published paper. In the abstracts for the 1996 meeting of the American College of Cardiology, the study by McCullough and colleagues is presented in a section titled "Age, Lipids, Clots and Dysrhythmias," an odd place for an abstract related to contrast nephropathy [3]. Moreover, the abstract is not indexed under the key words angioplasty, angiography, or contrast media but is listed only under renal function. This abstract was not included because, honestly, we just missed it. Had we not, we would have included it because the retrospective analysis is the only information we are aware of specifically related to contrast nephropathy after coronary interventions.

Second, what are the important points of this abstract, and how do they differ from our paper? McCullough and O'Neill's observations reaffirm almost everything we said in our paper. Contrast nephropathy rarely requires dialysis (0.8% of cases in McCullough and colleagues' data set) and is independently related to baseline creatinine level, diabetes, and the amount of contrast used. Our results differ on the long-term outlook for patients. McCullough and O'Neill state that the in-hospital mortality rate for patients requiring dialysis is 36%, with a 2-year survival rate of only 19%. These sobering statistics differ from our own experience and published data on contrast nephropathy occurring in situations other than coronary interventions. This difference may exist because of other comorbid conditions, such as acute myocardial infarction, but McCullough and colleagues' abstract provides no such data that might allow for risk adjustments. This lack of information highlights some of the difficulties encountered in the evaluation of abstracts. McCullough and associates' perspective is important, and we are sorry we failed to include it. Readers should note that the full-length report of McCullough and colleagues' work was recently published [4].

Dr. Marinella lists several drugs that contribute to hyperkalemia, especially in the setting of contrast nephropathy. All of these drugs are important, and we thank him. Both Dr. Marinella and Dr. Nzerue comment that we neglected to mention cholesterol embolization. In our experience, this is a serious but uncommon event. As Dr. Nzerue notes, there are differences between contrast nephropathy and cholesterol atheroembolism.