For more than 15 years, nonsteroidal antiinflammatory drugs (NSAIDs) have been associated with acute renal failure as a potential side effect. However, this phenomenon occurs only when renal blood flow significantly depends on renal vasodilator prostaglandin production [1]. In "normal" persons without risk factors for NSAID-related acute renal failure, only one clearcut report has described an NSAID overdose causing acute renal failure [2]. Thus, the relative incidence of NSAID-induced acute renal failure in any population depends on the prevalence of known risk conditions (such as chronic renal failure, cirrhosis, and heart failure) in that population and the severity of those conditions. The incidence of NSAID-induced acute renal failure in 1 million persons without risk conditions would probably be almost zero.

Feldman and colleagues [3] retrospectively compared the rate of renal failure in hospitalized patients who received ketorolac with the rate in those treated with opioids. Patients were matched only by hospital, admitting service, and date of initiation of therapy. The authors concluded that "when used for 5 days or less ... ketorolac does not increase the risk for acute renal failure." They also note that "ketorolac may [our emphasis] be associated with an elevated risk for acute renal failure" when used for more than 5 days. We believe that these conclusions cannot be safely drawn from the data presented. First, the authors report that the incidence of a history of chronic renal failure, papillary necrosis, and the nephrotic syndrome was statistically significantly higher in the opioid group (before they received opioids) than in the ketorolac group. This finding suggests that the opioid group had a higher risk for acute renal failure while hospitalized, possibly offsetting a relatively higher rate of ketorolac-induced acute renal failure. Moreover, this finding suggests a likely bias that is present in any such retrospective study; we hope that the prescribers of these medications knew that patients with risk factors for NSAID-induced acute renal failure should not receive ketorolac and preferentially treated such patients with opioids instead. No retrospective multivariate adjustment can control for this sort of bias.

Although more than 10 000 courses of analgesic therapy were analyzed, the study may not have been powerful enough to detect ketorolac-induced acute renal failure. We are not told how many patients in each group had underlying renal insufficiency, cirrhosis, the nephrotic syndrome, or heart failure or used diuretics, nor do we have any index of the severity of such conditions. Despite multivariate adjustment for these risk conditions, if only a handful of patients in the ketorolac group had major risk conditions or if the opioid group consisted of many more such persons with risk factors for acute renal failure, then the comparison between groups would probably not be able to detect ketorolac-induced acute renal failure.

Finally, the authors note that the overall incidence of acute renal failure was low (1.1%) in the population analyzed (as compared with the 4.9% incidence reported by Hou and colleagues [4]). What was the overall incidence of acute renal failure in all patients hospitalized during the study? Because the basis for the selection of patients was the use of parenteral ketorolac, the investigators may have preselected a group at very low risk for NSAID-induced acute renal failure. This preselection would have occurred because physicians who prescribe ketorolac would avoid prescribing the drug to patients at risk for NSAID-induced acute renal failure.

Because of data from prospective studies, other retrospective analyses, case reports on NSAID-induced acute renal failure (even occurring early in the course of ketorolac therapy [5]), and our concerns, we urge all physicians to use caution when using any NSAID in patients at risk for NSAID-induced acute renal failure.