Human granulocytic ehrlichiosis presenting as acute abdomen in an adult.

Second, we would like to underscore one of the main messages of our article [2], which is that the statistical methods used for estimating the cost impact of an adverse event have a major effect on the results. Costs attributed to adverse events, such as nephrotoxicity, should include only those costs incurred after occurrence of the adverse event. Observing an association between high hospital costs and nephrotoxicity is not a reliable indicator that nephrotoxicity is the direct cause of the higher costs, because many confounding factors probably exist that are common causes of higher cost and nephrotoxicity, including factors that are unmeasured. It is not safe to assume that this confounding is removed simply by building a multivariable regression model, particularly when criteria used for variable inclusion are based solely on statistical significance. The recent study [4] cited by Prendergast and Tong [1] found that nephrotoxicity was one of the factors associated with increased costs in a multivariable regression model, when all costs incurred after study entry, both before and after the adverse event, were grouped together. Another finding from this study worth noting is that the agents that were compared in the clinical trial (amphotericin B lipid complex and liposomal amphotericin B) exhibited similar efficacy but dramatically different incidences of adverse events (e.g., the incidences of nephrotoxicity were 42% and 14%–15%, respectively). However, total poststudy entry costs (excluding the cost of the study drug) across treatment groups were similar, a finding parallel to a study by Cagnoni et al. [5]. Thus, unconfounded, intent-to-treat analyses do not support the contention that nephrotoxicity has a large causal effect on hospital costs.