XUE et al.1 suggest adjusting for baseline serum creatinine or glomerular filtration rate in the primary analysis. The mean ± SD baseline serum creatinine concentrations for patients who had postoperative acute kidney injury (AKI) stage 0 was 0.81 (0.18); for AKI stage I it was 0.85 (0.21); and for AKI stage II/III it was 0.71 (0.18) mg/dl. When results are adjusted for baseline serum creatinine in addition to the variables in Table 1 of our report,2 the estimating odds ratio [95% CI] assessing the association between postoperative AKI stage 1 and long-term renal dysfunction is 2.4 [1.9, 2.9], which is virtually identical to the reported odds ratio of 2.4 [2.0, 3.0]. We therefore conclude that our published results were not biased by baseline creatinine. We further assessed the interaction between baseline creatinine and postoperative AKI stage on long-term outcome and found it to be convincingly negative (P = 0.35).
Xue et al. note that AKI severity and duration are important—and we agree. Severity is characterized by AKI stage and was fully reported. We do not have details about the immediate postoperative duration of AKI, but we agree that pronged acute injury is likely to be a harbinger of poor prognosis including development of chronic kidney disease. Duration of acute AKI might help parse which patients are most likely to develop persistent injury. However, it would not change the reported fraction of patients at each immediate postoperative AKI stage who developed sustained or worsened renal function 1 to 2 yr later.
Xue et al. also suggest that our estimated associations between postoperative AKI stage and long-term renal outcomes and mortality might be biased because we did not adjust for postoperative factors possibly associated with or contributing to immediate or long-term renal injury. By definition, confounding variables must be linked to both exposure (immediate AKI) and outcome (long-term AKI). Confounding variables must therefore occur before exposure.
As in many observational analyses, the timing of potential confounding factors and exposure is unclear in our patients. Postoperative cardiac dysfunction, sepsis, and use of nephrotoxic drugs might well contribute to immediate AKI. And we do not doubt that postoperative pulmonary and neurologic complications, heart failure, and new-onset arrhythmia are precipitating factors for short- and long-term mortality after noncardiac surgery, although they seem unlikely to influence the relationship between short-term and long-term AKI. On the other hand, new-onset proteinuria is probably consequent to renal injury and thus a potential mediator of persistent injury. Adjusting for proteinuria might therefore incorrectly diminish the true association of interest.3 In general, when an exposure such as postoperative AKI is studied, it is best to only consider potential confounding variables which were measurable before the exposure is documented, that is either before or during surgery.
Competing Interests
The authors declare no competing interests.