“Hold the Pendulum: Rates of Acute Kidney Injury Are Increased in Patients Who Receive Resuscitation Volumes Less Than Predicted by the Parkland Equation”
Thank you for the opportunity to respond to the recent Letter to the Editor submitted by Drs. Chung et al. We appreciate Dr. Chung and colleagues’ well-thought and excellent comments on our paper “Hold the pendulum: rates of acute kidney injury are increased in patients who receive resuscitation volumes less than predicted by the Parkland equation.”1
Dr. Chung and colleagues expressed concerns about how acute kidney injury (AKI) was defined for the purposes of our study. Specifically, they argue that we did not apply a standardized definition of AKI. We agree that it would have been optimal to utilize the currently accepted consensus definition. However, the data required to apply this definition were not collected as part of the Glue Grant study, from which our data were derived. We believe our definition identified patients with clinically meaningful AKI, rather than transient rises in creatinine that can be observed immediately following burn injury. We are unable to ascertain the potential impact of serum creatinine dilution on the diagnosis of AKI, particularly in patients who received resuscitation volumes in excess of 6 cc/kg/%TBSA. Similarly, the timing of AKI was not collected, and we agree that no conclusions regarding a causal link between resuscitation strategy and AKI can be made without data about their temporal relationship. Finally, Dr. Chung and colleagues posit that patients with chronic renal disease might have been misclassified as AKI. While comorbid renal failure was not measured, the incidence of chronic renal failure in this cohort is likely very low; a prior study from the National Burn Repository estimated an incidence of 0.6% in patients admitted for burn injury.2 Therefore, we feel the absence of data regarding premorbid renal disease is unlikely to be a source of misclassification bias.
As is the case with any retrospective study, there is potential for residual bias to confound the observed association between fluid resuscitation volume and AKI. To address potential confounding related to the baseline patient and injury characteristics that might influence clinician decision-making regarding total fluid received, these variables were included in multivariable modeling. Drs. Chung et al raise the concern of selection bias among patients who do not survive to receive the intervention (ie, 24-hour fluids). However, patients in whom treatment was deemed futile were excluded from the parent study, and among included patients, only 1 died within 24 hours. Therefore, the risk of selection/survival bias is minimal. Drs. Chung et al suggest that patient weight should have been included in multivariable analysis. We utilized the change in estimate approach to build our multivariable model. While patient weight was considered a potential confounder, its inclusion in the model did not change the parameter estimate for AKI by more than 10%, and as such it was not included as a covariate in the final model.
Drs. Chung et al correctly point out that the crude length of stay (LOS) and LOS per percent burn was lower among patients in the restrictive group than in the excessive group. While we considered reporting risk-adjusted estimates of LOS, this is a difficult outcome to model, because LOS is likely to be influenced by center-level practices, varying access to rehabilitation, and other factors. When we attempted to model risk-adjusted LOS for the present study, this unmeasured confounding resulted in overall poor fit of risk-adjustment models (data not published). Thus, we concluded the association between length of stay and resuscitation strategy could not be modeled successfully with the current data. We therefore do not feel the current data are adequate to support a conclusion regarding the association between resuscitation strategy and LOS.