Concerns About the Hold the Pendulum Paper
We read with great interest the recent article by Mason et al1 entitled “Hold the pendulum: rates of acute kidney injury are increased in patients who receive resuscitation volumes less than predicted by the Parkland equation”. We commend the authors for their efforts to improve burn care outcomes, and for providing a counterbalance to the movement away from over resuscitation (“fluid creep”). However, we have several serious concerns about how this study was conducted and the resulting conclusions. First, the definition utilized for acute kidney injury (AKI) in the article is flawed. Second, use of resuscitation volumes as an independent variable is problematic. Third, there are problems with the statistical analysis. Finally, the title overstates the results of the study and thus is potentially misleading.
The AKI definition used in this study differs importantly from the currently accepted consensus definition.2 Mason et al defined AKI as an absolute creatinine >1.5 mg/dL at any point during hospitalization. This could have resulted in misclassification of a patient with chronic kidney disease (CKD), or of one with a higher muscle mass. Since patients were not randomly assigned to a resuscitation volume in this study, it is possible that clinicians recognized CKD and opted for a less aggressive resuscitation. Additionally, the timing of AKI in relation to fluid administration was not taken into account. If a patient had AKI on admission and then received lower fluid volumes, one cannot conclude that the lower fluid volumes caused AKI. Furthermore, burn patients can and do develop AKI during extended hospitalizations that are unrelated to their resuscitations. Finally, the effect of creatinine dilution was not considered. Increased total body water during burn resuscitation results in an increased volume of distribution of creatinine. Thus, dilution of serum creatinine may have led to an under-diagnosis of AKI in the higher-volume groups. If the net fluid balance of these patients was known, methods could have been applied to take this into account.3
In this article, 24-hour resuscitation volumes are treated as an independent variable—as if the patients were randomized to 1 of 3 different fluid regimens (restrictive, standard, or excessive). On the contrary, in this multicenter observational study, each patient was resuscitated based on a calculated starting volume, which was then adjusted according to site-specific protocols using various end-points to include urine output. Hence, the 24-hour resuscitation volume is actually an outcome, and is dependent on a host of independent variables such as burn size, presence of inhalation injury, and premorbid conditions. This is an example of indication bias, a type of selection bias, described as one of the “seven deadly sins” of trauma outcomes research by del Junco et al.4 The patients were classified by the results of the intervention they received, that is, fluids. Patients who may have died before receiving the complete intervention would have been selected out. The presence of this type of bias often leads to false conclusions.
Further, weight differs significantly among groups. Patients in the restrictive group are significantly heavier than the other groups. This represents a significant confounder and should have been controlled for in the final model. AKI was not different among groups on univariate analysis (Table 2); this finding only became manifest on subsequent multivariate analysis. Indeed, the authors gloss over a remarkable finding, which is present on univariate analysis: the restrictive group had a lower length of stay (per percent burn) than the excessive group by almost 50% (Table 2).
The conclusion of the article reflects an observed association between AKI and those patients who ended up in the restrictive group.