Costs and Outcomes of Acute Kidney Injury in Critically Ill Patients with Cancer
Acute kidney injury (AKI) is a common complication in critically ill patients with cancer. The RIFLE criteria define three levels of AKI based on the percent increase in serum creatinine (Scr) from baseline: risk (≥50%), injury (≥100%), and failure (≥200% or requiring dialysis). The utility of the RIFLE criteria in critically ill patients with cancer is not known.
b Included if patient received therapy at any time from ICU admission to date of maximum creatinine.
The absolute number of patients developing AKI or requiring dialysis by hospital service is depicted in Figure 1. The incidence of AKI was higher among patients on a medical vs. a surgical service (21% vs. 6.6%). Patients with hematologic malignancies (leukemia, lymphoma, and myeloma) had the highest incidence of AKI and need for dialysis (28% and 9.3%, respectively). Among patients on a medical service, the odds for developing AKI or requiring dialysis were increased 1.9-fold and 5.4-fold, respectively, for patients with an underlying hematologic malignancy.
Figure 1.
Number of Patients with AKI or Needing Dialysis by Hospital Service
AKI, defined as a minimum 50% increase in serum creatinine from baseline, occurred in 301 patients (12.6%), of whom 56 (2.3%) required dialysis. By further defining AKI by the RIFLE criteria, we classified 6%, 3%, and 4% of patients into the RIFLE risk, injury, and failure categories, respectively. The median elevations in creatinine from baseline were 0.6, 1.1, and 2 mg/dL, respectively. The median time to maximum creatinine was two days for all patients with AKI. There was a stepwise decrease in estimated survival associated with each RIFLE category (Figure 2). Among patients in the RIFLE failure group, the estimated survival was similar between those who required dialysis and those who did not (P = 0.99, log-rank). Although survival for patients requiring dialysis was dismal overall, it was significantly worse for patients with underlying hematological malignancy vs. solid tumor (3% vs. 20%, respectively).
The results of the logistic regression model for predictors of death at 60 days after ICU admission is presented in Table 2. Race and gender were not significant on univariate or multivariate analyses. Although significant on univariate analysis, hematologic malignancy, prior hematopoietic cell transplant (HCT), baseline comorbidities (hypertension, diabetes, heart failure, liver disease), and sepsis were also eliminated during model reduction. After adjusting for the remaining covariates, the RIFLE risk, injury, and failure categories remained significantly associated with 60-day mortality with odds ratios of 2.3, 3.0, and 14, respectively.
| VARIABLE | UNIVARIATE | MULTIVARIATE | |||
|---|---|---|---|---|---|
| OR | P | OR | 95% CI | P | |
| Age ≥55 years | 1.2 | 0.08 | 1.5 | 1.1–1.9 | 0.007 |
| Male vs. female | 0.997 | 0.98 | |||
| Ethnicity | |||||
| Black vs. white | 2.0 | <0.001 | |||
| Hispanic vs. white | 1.1 | 0.39 | |||
| Other vs. white | 0.8 | 0.46 | |||
| Hypertension | 1.3 | 0.02 | |||
| Diabetes | 1.6 | <0.001 | |||
| Heart failure | 2.5 | <0.001 | |||
| Chronic liver disease | 1.8 | 0.02 | |||
| RIFLE category | |||||
| Risk vs. no AKI | 4.1 | <0.001 | 2.3 | 1.5–3.6 | <0.001 |
| Injury vs. no AKI | 8.1 | <0.001 | 3.0 | 1.6–5.8 | 0.001 |
| Failure vs. no AKI | 35 | <0.001 | 14.3 | 7.2–29.0 | <0.001 |
| Amphotericin | 10.9 | <0.001 | 1.9 | 1.1–3.3 | 0.03 |
| Vasopressors | 6.3 | <0.001 | 2.0 | 1.4–2.6 | <0.001 |
| Mechanical ventilation | 2.1 | <0.001 | 1.9 | 1.4–2.5 | <0.001 |
| IV diuretics | 3.8 | <0.001 | 1.4 | 1.1–1.9 | 0.015 |
| Sepsis | 5.7 | <0.001 | |||
| Medical vs. surgical service | 9.9 | <0.001 | 2.2 | 1.5–3.1 | <0.001 |
| Liquid vs. solid tumor | 5.5 | <0.001 | |||
| Prior HCT | |||||
| Autologous | 1.7 | 0.23 | |||
| Allogeneic | 6.0 | <0.001 | |||
| Advanced vs. locoregional stage (SEER) | 4.4 | <0.001 | 2.1 | 1.6–2.6 | <0.001 |
| ER admission | 11.3 | <0.001 | 5.3 | 3.7–7.6 | <0.001 |
| Pre-ICU length of stay | 1.06 | <0.001 | 1.02 | 1.0–1.03 | 0.02 |
Likelihood ratio x2(12) = 818 (P < 0.001), positive predictive value 72%, negative predictive value 88%; area under the receiver operating curve = 0.88, Hosmer-Lemeshow x2(8) = 6.8 (P = 0.56).
OR, odds ratio; AKI, acute kidney injury; HCT, hematopoietic cell transplant; ER, emergency room; ICU, intensive care unit.
To further assess the relationship between serum creatinine and mortality, a separate logistic regression was performed using “percent rise in creatinine” as a continuous predictor variable (Table 3). Need for dialysis was also included as an independent variable. Aside from “percent rise in creatinine” and dialysis, model reduction yielded the same covariates as in the initial model. Dialysis had the largest effect on the odds of 60-day mortality (odds ratio = 6.2). After adjusting for dialysis, “percent rise in creatinine” remained significantly associated with 60-day mortality. For example, a 10% rise in creatinine increased the odds of mortality by 8%. The predictive capabilities of both logistic regression models were similar.
