Drs. Buller and Perazella state that studies of acute renal failure associated with NSAIDs should be limited to persons at risk for this condition. We agree with this view, but hospitalized patients ill enough to receive parenteral analgesics are at risk for acute renal failure, which in some cases is blamed on NSAIDs [1]. Furthermore, a substantial minority have conditions (such as volume depletion or hypotension) in which renal function depends on vasodilator prostaglandins; this may place them at higher risk for acute renal failure from such cyclooxygenase inhibitors as ketorolac.

Drs. Buller and Perazella questioned whether our study had sufficient power to detect an effect of ketorolac among patients who are exposed to diuretics or have a history of renal disease, cirrhosis, or heart failure. We, too, were concerned about a potentially enhanced toxicity from ketorolac in this subgroup. We therefore explored whether there was a statistically significant interaction between membership in this subgroup and the occurrence of ketorolac-induced acute renal failure, but we found no such interaction. Thus, ketorolac was not differentially toxic in this subgroup, and we did not present a subanalysis for these 3891 patients (2111 in the opioid group and 1780 in the ketorolac group). However, to respond to Drs. Buller and Perazella's specific inquiry, we have calculated the hazard ratio in this subgroup for the comparison of ketorolac with opioids (hazard ratio, 0.86 [95% CI, 0.63 to 1.17]); this confirms our previous finding of no enhanced risk in this subgroup.

We are puzzled by Drs. Buller and Perazella's statement that multivariate adjustment cannot account for baseline differences between our study groups. This is exactly the goal of multivariate analysis (to control for the potential confounding influence of measured baseline characteristics) [2]. It is conceivable that unmeasured differences in our study groups could have led to bias, but we believe that multivariate adjustment for previous renal disease and many other measured characteristics makes such a potential bias very unlikely.

Drs. Buller and Perazella note that we observed a rate of acute renal failure that was lower than that cited in Hou and colleagues' report of hospital-acquired acute renal failure [3]. There are several reasons for these differences in the rates of acute renal failure, each related to differences between Hou and colleagues' objective and our objective. Our study was designed to evaluate the comparative risk for acute renal failure in patients receiving parenteral analgesics. First, the definition of acute renal failure that we used was more stringent. For example, we required an elevation of serum creatinine level of 44.2 µmol/L if the baseline level was less than 132.6 µmol/L, whereas Hou and colleagues used an elevation of 44.2 µmol/L for baseline levels as high as 168 µmol/L. Second, we looked for acute renal failure only during the period of analgesic administration and for 3 days thereafter rather than at any time during the hospital stay. Finally, we examined only hospitalized patients who received parenteral analgesics, not all hospitalized patients.

We emphasize that caution should be used when parenteral ketorolac is administered for a prolonged period.