Acute kidney injury in critically ill patients with solid tumours

Abstract

Background

Patients with solid tumours are at risk for acute kidney injury (AKI), however, epidemiological data are limited.

Methods

We conducted a study that included patients with solid tumours admitted to a single-centre intensive care unit (ICU) from January 2011 to December 2015. We analysed factors associated with the occurence of AKI, ICU and Day-90 mortality.

Results

Two-hundred and four patients were included. The incidence of AKI was 59%, chiefly related to sepsis (80%), hypovolaemia (40%) and outflow tract obstruction (17%). Renal replacement therapy was implemented in 12% of the patients, with a hospital mortality of 39%. Independent predictors of AKI were: Simplified Acute Physiological Score II (SAPS II) [odds ratio (OR) 1.05; 95% confidence interval (95% CI) 1.02–1.07; P < 0.001], abdominal or pelvic cancer (OR 2.84; 95% CI 1.35–5.97; P = 0.006), nephrotoxic chemotherapy within the previous 3 months (OR 3.84; 95% CI 1.67–8.84; P = 0.002) and sepsis (OR 2.74; 95% CI 1.30–5.77; P = 0.008). Renal recovery at Day 90 was inversely related to AKI severity. ICU, hospital and Day-90 mortality were 15, 29 and 37%, respectively. Factors independently associated with ICU mortality were: total serum protein (OR per 10 g/L, 0.44; 95% CI 0.23–0.86; P = 0.02) and SAPS II (OR 1.04; 95% CI 1.01–1.07; P = 0.02), while Day-90 mortality was associated with performance status 3–4 (OR 6.59; 95% CI 2.42–18; P < 0.001) and total serum protein (OR 0.60; 95% CI 0.38–0.94; P = 0.02).

Conclusions

AKI in patients with solid tumours was frequent and renal recovery gradually decreased in proportion to AKI severity. However, AKI was not independently associated with a higher short-term mortality.

INTRODUCTION

The burden of acute kidney injury (AKI) in patients with cancer is a growing concern [1]. An increasing number of people are living with cancer owing to the major diagnostic and therapeutic advances of the last few decades [2]. However, these patients may experience renal injuries related to the malignancy itself or its treatment [3]. The occurrence of AKI is likely to complicate many aspects of the patients’ care and adversely affect their prognosis [4, 5]. A major issue is that AKI may compromise the ability to receive optimal anticancer treatment and induce toxicities related to drug overdose [6, 7].

Critically ill patients have undoubtedly the highest risk of AKI, mainly in the context of sepsis, and experience the most severe stage of AKI, which may require renal replacement therapy (RRT) [8]. The management of critically ill cancer patients with AKI is challenging and often requires the close collaboration of intensivists, nephrologists and oncologists. The decision to start RRT in cancer patients often raises ethical issues owing to the very high rate of in-hospital mortality associated with this condition [9, 10]. Thus, recent data on the prognosis and the likelihood of renal recovery are essential to guide clinicians’ decisions. Unfortunately, there is little information on AKI in critically ill patients with solid tumours as studies in the intensive care unit (ICU) setting have mainly investigated patients with haematological malignancies.

The objectives of this study were to describe the clinical features, risk factors and outcomes of AKI in critically ill patients with solid tumours. In addition, we sought to identify factors associated with mortality.

MATERIALS AND METHODS

This study was approved by our institutional review board (CECIC Clermont-Ferrand IRB no.5891; Ref: 2007–16). According to the French law, informed consent was not required in this retrospective observational study in which the collected data were anonymized.

Design and setting

We retrospectively included all adult patients with solid tumours admitted to the medical ICU of Saint-Louis University Hospital for a minimum of 48 h between 1 January 2011 and 31 December 2015.

The Saint-Louis University Hospital is a 650-bed public hospital, with 330 beds dedicated to patients with various conditions associated with immunodeficiency (haematological malignancies, solid cancers and solid organ transplantation). There are four oncology wards: general oncology, lung cancer, gastrointestinal oncology and radiation therapy.

The medical ICU of Saint-Louis University Hospital is a 12-bed unit that admits 750–850 patients per year, of whom about one-third have cancer. Information on the organization of the ICU and criteria for ICU admission has been published elsewhere [11]. Mandatory ICU admission reasons are at least one organ failure and the need for at least one of the following therapies: supplemental nasal oxygen ≥ 3 L/min, use of vasoactive drugs, sustained intravenous fluid administration, invasive or non-invasive mechanical ventilation or RRT. At our institution, intensivists and oncologists are available 24 h a day, 7 days a week and work together to manage all high-risk cancer patients.

Inclusion criteria

All adult patients (≥18 years old) were identified from the ICU database using codes for malignant neoplasms (C00-C97) from the International Classification of Diseases, Tenth Revision system. All the medical records of the patients were thereafter reviewed by two investigators (D.K. and E.C.) to confirm the diagnosis of solid tumours. Patients with haematological malignancies were excluded from the study as were those with complete remission of the solid tumours for >5 years before ICU admission.

Data collection

All data were obtained from medical records and patient charts. Baseline patient characteristics were collected, including demographics, comorbidities and the best performance status (PS) within 3 months before ICU admission. Data regarding the type of cancer, cancer status and anticancer treatments were abstracted from the medical charts. Chemotherapy was classified nephrotoxic if the prescription included at least one drug for which episodes of nephrotoxicity were reported in the literature (in a similar way to the criteria used by Launay-Vacher et al. [7]). The variables recorded regarding ICU admission and treatments were: data relative to clinical presentation, reason for ICU admission, sepsis at ICU admission time of admission and discharge from ICU, diagnosis, therapies implemented and outcome data (vital status at ICU and hospital discharge, as well as 90 days after ICU admission). Disease severity at ICU admission was assessed using the Simplified Acute Physiology Score II (SAPS II) at Day 1 [12]. The decisions to withdraw or withhold life-sustaining therapies or other treatments during the ICU stay were abstracted from the medical charts.

Definitions

Definition and staging of AKI were reported according to the 2012 Kidney Disease: Improving Global Outcomes (KDIGO) criteria [13]. Decisions regarding the initiation, discontinuation and modalities of RRT were made by senior nephrologists (D.K., M.V., L.Z. and E.C.) based on the 2012 KDIGO guidelines [14]. The serum creatinine value used as the baseline was the lowest value obtained within 3 months before ICU admission. For patients without any measure of baseline renal function, and with no past history of renal disease, the baseline creatinine was back-calculated using the Modification of Diet in Renal Disease formula (four-variable equation) and assuming an estimated glomerular filtration rate of 75 mL/min/1.73 m² before ICU admission [15]. Chronic kidney disease (CKD) was defined according to the KDIGO definition, for diagnosis and staging [16].

The clinical, laboratory and imaging data of each patient with AKI were reviewed by four intensivists who are also nephrologists (D.K., M.V., L.Z. and E.C.) in order to reach a consensus regarding the cause of AKI using the following definitions. Septic AKI was defined as the simultaneous presence of sepsis and AKI using consensus definitions for both [17]. Hypoperfusion was defined using a combination of criteria: clinical evidence of hypovolaemia (poor fluid intake, diarrhoea, vomiting, oliguria, tachycardia and hypotension), urine analysis findings (e.g. urine sodium <10 mmol/L, urine sodium/potassium ratio <1, fractional excretion of sodium <1%, urinary osmolality >500 mOsm/L), a positive response to intravenous fluid administration (reversal of clinical and laboratory signs of volume depletion) and the return of serum creatinine level to baseline within 24—72 h. Urinary tract obstruction was defined as AKI associated with dilatation of the collecting system in one or both kidneys, assessed by imaging. Tumour lysis syndrome (TLS) was defined according to the Cairo and Bishop criteria [18]. Exposure to nephrotoxic agents (at least one of the following agents: aminoglycosides, vancomycin, angiotensin-converting enzyme inhibitors, angiotensin II type 1 receptor blockers, diuretics, intravenous radiocontrast agents, non-steroidal anti-inflammatory drugs) within 3 months prior to ICU admission and during ICU stay was extracted from the medical charts.

Recovery of renal function was assessed using the serum creatinine value 90 days after ICU admission. Based on the study conducted by Pannu et al. [19], complete recovery was defined as a return to the baseline serum creatinine or a serum creatinine increase of no more than 25% that of the baseline serum creatinine. Partial recovery was defined as dialysis-independence and a serum creatinine increase over 25% that of the baseline serum creatinine. Non-recovery was defined as dialysis dependency 90 days after ICU admission.

The primary aim of the study was to report the clinical features of AKI and to identify independent factors associated with the occurrence of AKI in critically ill patients with solid tumours. The secondary objective was to describe the independent factors associated with mortality both at ICU discharge and 90 days after ICU admission.

Statistical analysis

Data for categorical variables are summarized as frequencies and percentages; quantitative variables are presented with medians, 25th and 75th percentiles [interquartile range (IQR)].

Associations between covariates and AKI, ICU mortality and Day-90 mortality were examined using logistic regression models, reporting estimated odds ratios (OR) and their 95% confidence intervals (95% CIs). Log-linearity of continuous explanatory variables was checked using comparison of linear modelling to restricted cubic splines modelling.

Factors that were significant at the 10% level in univariate analysis were candidate for multivariate adjusted analysis of the endpoints. Final adjusted were selected using a stepwise backward procedure on P-value (threshold 0.05).

All tests were two-sided and P-value <0.05 was considered as indicating significant association. Analyses were performed using R statistical platform, version 3.0.2 (htpps://cran.r-projet.org).

RESULTS

Study population

During the study period, 2184 patients were admitted to the ICU, of which 204 (9.3%) had solid tumours. Patient characteristics are reported in Table 1. The median age was 64 (IQR 53–70) years old and 129 (78%) had a good general condition (PS 0–2). The median time between cancer diagnosis and ICU admission was 10.5 (2.8–34.8) months, with 24 (12%) patients newly diagnosed during the ICU stay. The five most common types of cancer were breast (n = 44, 21%), lung (n = 34, 16%), prostate (n = 28, 13%), colorectal (n = 18, 9%) and bladder (n = 14, 7%) cancers. Metastases were reported in 128 (63%) patients. Roughly two-thirds of the patients had received anticancer treatments within the 90 days before ICU admission, including chemotherapy in 94 (46%) patients. The chemotherapy prescriptions included had at least one drug with potential nephrotoxicity in 62 (66%) patients. The nephrotoxic molecules most often prescribed were cisplatin (n = 27, 29%), carboplatin (n = 17, 18%) and oxaliplatin (n = 8, 9%) (Supplementary data, Table S1). Twenty-eight (14%) patients had previous CKD (Stage 3a, n = 18; Stage 3 b, n = 5; Stage 4, n = 5 according to the KDIGO criteria).

Clinical features of AKI

The median baseline serum creatinine of the 204 patients was 70 (51–90) µmol/L. According to the KDIGO criteria, AKI was diagnosed in 113 (59%) of the 193 patients without missing data on renal function, of whom 43 (22%) were classified in the Stage 1 category, 20 (10%) in the Stage 2 and 50 (26%) in the Stage 3 (Table 2). Patients with AKI Stage 3 had the highest ICU mortality rate (24%). The main causes of AKI were sepsis (n = 90, 80%), hypovolaemia (n = 45, 40%) and urinary tract obstruction (n = 19, 17%) while TLS was reported in only 4 (3.5%) patients (Supplementary data, Table S2). RRT was implemented in 24 (12%) patients, with a median time-to-RRT of 1 (1–1) day and for a median duration of 5 (2–14) days. The most common reasons for RRT initiation were anuria (74%), severe metabolic acidosis (48%), hyperkalaemia (26%) and acute pulmonary oedema (4%). The modalities of RRT were intermittent haemodialysis (n = 11, 46%), continuous haemofiltration (n = 6, 25%) or alternative use of both techniques (n = 7, 29%). At hospital discharge, among the 24 RRT patients, 39% were deceased, 13% were still undergoing RTT treatment and 48% were RRT-free with a median serum creatinine of 134 (81–205) µmol/L.

The likelihood of complete renal recovery decreased with AKI severity (Figure 1). This effect was remarkable from Stage 1, where 65% of the patients fully recovered, whilst AKI Stage 3 was associated with the poorest renal prognosis (only 28% had a complete recovery of renal function).

Table 3 reports the assessment of various factors for developing AKI. In the univariate analysis, baseline serum creatinine, nephrotoxic chemotherapy administered within the last 90 days, nephrotoxic agent exposure within the last 90 days, SAPS II at Day 1 and sepsis at ICU admission were associated with a higher risk for AKI. In the multivariate model, significantly higher risks for AKI persisted with solid tumour located in the abdomen or the pelvis (OR 2.84; 95% CI 1.35–5.97; P = 0.006), nephrotoxic chemotherapy administered within the last 90 days (OR 3.84; 95% CI 1.67–8.84; P = 0.002), SAPS II at Day 1 (OR per point, 1.05; 95% CI 1.02–1.07; P < 0.001) and sepsis (OR 2.74; 95% CI 1.30–5.77; P = 0.008).

ICU management and outcome analysis

The most common reasons for ICU admission were acute respiratory failure (n = 79, 39%), shock (n = 60, 29%) and coma (n = 20, 10%). The median SAPS II at Day 1 was 39 (29–52). Over two-thirds of the patients were admitted for sepsis, with microbiological documentation in 46% of the cases. During the ICU stay, mechanical ventilation was needed in 64 (31%) patients, vasopressors in 61 (30%) and RRT in 24 (12%). Median ICU and hospital lengths of stay were 4 (3–6) days and 15 (9–24) days, respectively. Decisions to withhold or withdraw life-sustaining therapies were taken in 40 (20%) patients. These latter had a Day-90 mortality of 95%. ICU hospital and Day-90 mortalities of the whole cohort were 15% (n = 31), 27% (n = 55) and 37% (n = 66), respectively (Supplementary data, Figure S1).

Table 4 shows the factors associated with ICU mortality. In the univariate analysis, the following three variables had higher OR for ICU mortality: lower total serum protein at Day 1, SAPS II at Day 1 and AKI Stage 3. Only total serum protein (OR per 10 g/L, 0.44; 95% CI 0.23–0.86; P = 0.02) and SAPS II (OR 1.04; 95% CI 1.01–1.07; P = 0.02) remained independently associated with ICU mortality in the multivariate model.

Factors associated with mortality 90 days after ICU admission are indicated in Table 5. In the univariate analysis, patients with PS 3–4, metastases, high SAPS II, low total serum protein and newly diagnosed cancer had higher mortality. Two factors were identified as independent predictors of Day-90 mortality in the multivariate model: PS 3–4 (OR 6.59; 95% CI 2.42–18.0; P < 0.001) and total serum protein (OR per 10 g/L, 0.60; 95% CI 0.38–0.94; P = 0.03).

DISCUSSION

We analysed renal outcomes among a population of critically ill patients with solid tumours over a 5-year period and found that 6 in every 10 patients developed AKI, predominantly moderate-to-severe stages. Although sepsis and hypovolaemia were the main causes of AKI, obstruction of the outflow tract was reported in nearly a fifth of the cases. The likelihood of developing AKI relied not only on the severity of the patient condition but also on the type of cancer and chemotherapy administered. Even though patients with AKI Stage 3 had the highest short-term mortality, AKI was not identified as an independent factor associated with mortality. Furthermore, two-thirds of the patients treated by RRT were alive at hospital discharge. However, another important finding was that renal recovery after AKI was highly variable and inversely correlated to the severity of AKI.

The management of cancer patients in the ICU is a growing concern [20, 21]. While major therapeutic advances of the past decade have undoubtedly resulted in extended survival [22], the repeated administration of anticancer treatments increases the risk of developing meaningful complications. For instance, Lapi et al. have reported that the use of androgen deprivation therapy in patients with advanced prostate cancer was associated with a significantly increased risk of AKI [23]. Moreover, in a recent study conducted in the UK, 5.2% of the 118 541 patients newly diagnosed for cancer experienced a critical illness resulting in unplanned ICU admission within 2 years of cancer diagnosis [24]. In our study, almost 1 in 10 ICU patients had solid tumours, which is in line with other studies conducted in Europe and Brazil where cancer patients accounted for 10 to 21.5% of the ICU population [25–27]. AKI was rarely the main reason for ICU admission (7% of the cases) and yet 59% of the critically ill cancer patients had AKI, mainly Stage 3 according the KDIGO criteria (44% of the patients with AKI). This result is similar to that of the Acute Kidney Injury-Epidemiologic Prospective Investigation (AKI-EPI) study, where the incidence of AKI in a non-selected ICU population was 57.3%, among which 52.3% were classified Stage 3 [28]. Although previous studies have identified cancer as an independent risk factor of AKI in the ICU setting [29, 30], the patients included in these studies had mainly haematological malignancies rather than solid tumours. Conversely, in the AKI-EPI study, cancer patients did not have a higher occurrence of AKI compared with patients without cancer [28]. Thus, in the ICU setting, the rate of AKI in patients with solid tumours does not seem to be higher than that reported in patients without cancer.

The main causes of AKI in our cohort were sepsis and hypovolaemia, which have been extensively reported as the most common aetiologies in the ICU setting [10, 28, 30]. However, it is worth noting that over a quarter of patients had a cause of AKI related to the cancer (outflow tract obstruction, hypercalcaemia and TLS). Indeed, 17% of the patients had evidence of outflow tract obstruction, while this aetiology is usually reported in <5% of patients with haematological malignancies [4, 31] and 1.4–2.6% of ICU patients [28, 30]. Early identification of these aetiologies is crucial as they may benefit from specific preventive and therapeutic strategies, such as drainage procedures to relieve obstruction, bisphosphonate to treat hypercalcaemia [32], and urate oxidase for TLS [33]. Furthermore, we found that AKI not only developed in more severely ill patients at ICU admission, but also when the solid tumours were located in the abdomen or the pelvis and when nephrotoxic chemotherapy had been administered. The significance of cancer location might have several explanations, such as outflow tract compression by a bulky tumour, retroperitoneal fibrosis, kidney infiltration by malignancy, or hypovolaemia related to peritoneal carcinosis or bowel obstruction. The nephrotoxicity of some chemotherapeutic agents has been well characterized in previous studies [7, 34]. In addition to their immediate toxic effect on the renal parenchyma, these agents might reduce the renal functional reserve [35] when repeatedly administered, and thus make the kidneys more sensitive to a further source of injury, mainly sepsis in the ICU setting. The growing development and use of targeted therapies, which are known to have renal toxicity [35–37], might amplify this issue in the future.

The decision to commence RRT in cancer patients with severe AKI is a matter of debate owing to its reputation for a poor prognosis [37, 38]. Indeed, hospital mortality rates ranging from 70 to 100% have been reported [9, 10], raising doubts about the usefulness of RRT in this population. This was not our experience. In our study, AKI was not an independent predictor of mortality and even though RRT patients had the highest in-hospital mortality (39%), this figure is similar to that reported in patients without cancer [39, 40]. It can be hypothesized that cancer patients have benefited from the recent advances in supportive care [41–43] and the upward trend in survival of RRT patients [44]. While we acknowledge that our admission policy may have influenced our results [11], we advocate that RRT access should not be denied to cancer patients in good general condition, regardless of the cancer status.

Our final finding is data on renal recovery 90 days after AKI in patients with solid tumours. The complete recovery of renal function decreased gradually from 65% in patients with AKI Stage 1, to 38% in AKI Stage 2 and 28% in those with AKI Stage 3. Non-recovery is known to be associated with a significant long-term morbidity and mortality [45, 46]. These findings strongly support the recommendation for a nephrology follow-up after AKI in patients with solid tumours [47].

Our study had several limitations. First, it was conducted at a single institution. Thus, our admission policy and patient recruitment patterns may have influenced our findings. However, the large number of cancer patients treated in the numerous oncology wards at our hospital and the standardized policy of ICU admission suggest that our results may also apply to other settings. Secondly, kidney biopsies were not performed due to the meaningful level of complications related to the procedure in the ICU setting [48, 49]. Thus, we may have underestimated some difficult to diagnose causes of AKI. Thirdly, as various types and status (newly diagnosed, in remission, under treatment) of solid tumours were included in the study, we were unable to assess the impact of AKI on the ability to receive further anticancer treatments and to achieve complete remission. Fourthly, some laboratory data known to be associated with AKI (proteinuria, uric acid) were not available and thus, not analysed. Finally, we did not include patients without cancer in this study and hence, we did not compare the characteristics and outcomes between patients with and without solid tumours. The strengths of our study include the large and homogeneous patient population included over a short and recent period (5 years) during which no changes in treatment practices occurred. Also, our unit has extensive experience in managing critically ill cancer patients [4, 31, 50].

In conclusion, AKI was frequent and severe in critically ill patients with solid tumours. Acute illness severity, abdominal or pelvic cancer and nephrotoxic chemotherapy were predictors of AKI. Mortality was strongly linked to patients’ acute illness severity and general condition but not to AKI or cancer characteristics. Renal recovery was highly variable and adversely affected by AKI severity. Our findings pave the way for the development of a bundle of diagnostic and therapeutic measures dedicated to AKI in patients with solid tumours. This might help clinicians in quickly identifying the reversible causes of AKI (e.g. hypovolaemia, urinary tract obstruction, hypercalcaemia, TLS) and thus, implementing appropriate treatments. Further studies are required to assess whether such strategy will translate into better outcomes.

ACKNOWLEDGEMENTS

The authors thank Victoria Mears for helping with English editing.

AUTHORS’ CONTRIBUTIONS

E.C. was the principal investigator and takes primary responsibility for the article. D.K., L.B., L.K., L.Z., M.V., L.T., B.S. and E.A. recruited the patients. L.B. participated in the statistical analysis. E.C. and D.K. co-ordinated the research. E.C. and D.K. wrote the article.

CONFLICT OF INTEREST STATEMENT

D.K., L.B., L.K., L.Z., M.V., L.T., B.S. and E.S. declare that they have no competing interests. E.A. declares the following statements: board member for Gilead, lectures for Alexion, MSD, Astellas, research grants from MSD, Fisher & Paykel, Pfizer (2012).

REFERENCES