Introduction

Numbers of patients living with cancer are increasing because of advances in the treatment of malignancies (Jemal et al. 2017). Cancer-related complications are common during the course of the disease (Torres et al. 2016). Accordingly, many of these patients require admission to the intensive care unit (ICU) for the management of underlying pathophysiological disorders such as postoperative status, respiratory failure, and sepsis (Schellongowski et al. 2016).

Organ support methods and sepsis management is the mainstay of treatment for critically ill patients with cancer. However, some malignancy-specific problems including oncological emergencies, organ dysfunction due to expansive or infiltrative cancer, chemotoxicity, radiotoxicity, tumour lysis syndrome, leukostasis, and hemophagocytic lymphohistiocytosis require a specialized therapy (Young and Simmons 2014; Lehmberg et al. 2015; Olcina and Giaccia 2016; Strauss et al. 2017). Thus, a close collaboration between the intensivist and oncologist is mandatory.

Mortality rates of critically ill cancer patients have decreased in the latest decades, because of advances in the management of malignancies and organ failures; however, some studies suggest that mortality of patients with cancer remains higher compared with that in patients without cancer (Taccone et al. 2009). Reasons of ICU admission, nature and number of organ failures, type of malignancy, and therapies before ICU admission may affect outcomes (Kostakou et al. 2014). According to methodological design and included patients, the ICU mortality rate in cancer patients range between 30 and 77% (Anisoglou et al. 2013; Mânica et al. 2013; Ostermann et al. 2013; Yoo et al. 2013; Aygencel et al. 2014). In addition, the ICU mortality rate for critically ill ventilated patients with cancer is greater than 45% (Almeida et al. 2014; Martos-Benítez et al. 2017). The study was aimed to describe the characteristics of cancer patients admitted to an oncological ICU (OICU) and to identify clinical features associated with outcomes.

Patients and methods

Design and setting

This was a prospective cohort study conducted in the OICU of the Institute of Oncology and Radiobiology (IOR) from January 2014 to December 2015. This is a 220-bed, university-affiliated, tertiary care referral centre for cancer patients in Havana, Cuba. The OICU has 12 beds and provides care for about 500 medical and surgical cancer patients per year. The current study was conducted in accordance with the Declaration of Helsinki, and it was approved by the Scientific Council and the Ethics Committee for Scientific Research of the IOR (no. 54/02-11-2013). Written informed consent was obtained from all included patients.

Participants

Over the study period, a total of 992 consecutive cancer patients were included independently of histology, tumour location, clinical stage, or reason for ICU admission. Patients with a length of ICU stay ≤ 1 day were excluded (Fig. 1); they may be divided in two subgroups: (1) patients undergoing intermediate-risk surgery (e.g., head and neck, major gynaecological, major urological) admitted to the ICU for early postoperative surveillance in accordance with institutional protocol (406 patients); and (2) patients who died within 24 h after ICU admission (63 patients). These patients were considered as unsalvageable, independently of any therapeutic effort, because of their severe pathophysiological disturbance; therefore, they can show basic features that distinguish them from general cancer patients admitted to the ICU. Thus, their exclusion reduced the risk of selection bias.

Fig. 1
figure 1

Flow diagram of study participants. ICU intensive care unit

Full size image

For those patients who were admitted more than once to the ICU during the same hospitalization, only the first data on ICU admission were analysed. To be admitted, the oncologist and the ICU physician generally were agreed that patients had a potential chance of recovering from the acute problem. In contrast, patients without any further oncology-specific treatment options were offered end-of-life care on the referring ward and not transferred to the ICU.

Data collection and outcomes

The following demographic and clinical data were obtained at ICU admission: age, sex, Age-Adjusted Charlson Comorbidity Index score with the exclusion of malignancy, primary tumour location, clinical stage of cancer, patient origin (hospital ward or emergency department), type of admission (elective admission or unplanned admission), nature of admission (malignancy-related or non-malignancy-related), Acute Physiology and Chronic Health Evaluation (APACHE) II score, transfusion of blood products within 15 days prior to ICU admission, sepsis, chemotherapy-induced adverse event as reason for admission in ICU, and requirement of invasive mechanical ventilation (MV).

Malignancy-related admission was defined as a cause directly related with tumoural tissue [e.g., primary surgical intervention (tumoural resection or tumoural occlusion/perforation), tumoural infection/haemorrhage, tumour lysis syndrome, paraneoplastic syndrome, and infiltrative or metastatic organ dysfunction]. Otherwise, the nature of admission was categorized as non-malignancy-related (e.g., infection/haemorrhage of non-tumoural tissue, surgical reintervention due to any complication of primary surgery, and adverse event to oncospecific therapy).

The primary outcome of interest was hospital mortality. Secondary outcomes were ICU mortality, length of ICU stay, ICU re-admission, and length of hospitalization.

Statistical analysis

Categorical variables are showed as count with percentage and numerical variables as median with 25th–75th interquartile range (IQR). Difference between groups was performed using Pearson’s Chi-square test (2) for categorical variables; because of lack of normality, the Mann–Whitney U test was used for numerical variables.

Multivariate logistic regression analysis was used to identify risk factors associated with hospital mortality. Because of lack of normality, APACHE II score was transformed by natural logarithm (Ln). Then, Ln(APACHE II) and Age-Adjusted Charlson Comorbidity Index were included as confounders, and a model for each clinical characteristic was built. Only those explanatory variables yielding p values ≤ 0.25 in the univariate analysis and obvious clinical implications were considered to enter in logistic regression models. Parsimony of the models was guaranteed. The Hosmer–Lemeshow test was used to check the goodness of fit. Discrimination capability was evaluated by determination of the area under the receiver-operating characteristic (ROC) curve. Results were summarized as odds ratio (OR) and respective 95% confidence interval (CI).

Statistical test with a two tailed p value ≤ 0.05 was considered as significant. Data were analysed using IBM® SPSS® Statistics 23.0 (IBM, Chicago, IL, USA).

Results

Characteristics of study population

A total of 522 patients were analysed (Fig. 1). The median age was 60.5 years (IQR 51.0–69.0 years). The median in the Age-Adjusted Charlson Comorbidity Index score was 3.0 points (IQR 2.0–5.0 points). The most common primary tumours were localized in gastrointestinal tract (35.2%), thorax (25.3%), and gynaecological tract (10.0%). Advanced cancer was observed in 53.3% of patients (stage III 29.7%; stage IV 23.6%). Four hundred and seventy-one patients (90.2%) arrived to ICU from hospital ward. Unplanned ICU admission accounted for 25.3% of patients (Table 1). The most common reasons for unplanned admission were acute respiratory failure (36.7%), infections (32.1%), cardiovascular disorders (19.3%), neurological disorders (6.4%), and emergency surgery (4.4%).

Table 1 Characteristics of patients and univariate analysis of factors associated with hospital mortality
Full size table

The median APACHE II score was 12.0 points (IQR 10.0–14.0 points), and the estimated probability of death was 11.7% (IQR 8.5–18.5%). Invasive ventilatory support was required for 15.5% of patients; the median ventilation time was 9.0 days (IQR 6.0–15.0 days). Eighty-one patients (15.5%) required vasopressors during their stay in ICU. The main characteristics of study population are depicted in Table 1.

Outcomes of cancer patients admitted to intensive care unit

The overall ICU mortality rate and overall hospital mortality rate was 10.2 and 13.0%, respectively. The median length of ICU and hospital stay was 3.0 and 9.0 days, respectively. Twenty-three patients were re-admitted to ICU during the same hospitalization. ICU mortality (p < 0.0001), length of ICU stay (p = 0.001), and hospital mortality (p < 0.0001) were higher for patients with unplanned admission than those for electively admitted patients. Table 2 summarizes the outcomes of studied participants.

Table 2 Outcomes of patients with cancer admitted to the intensive care unit
Full size table

In the univariate analysis, Age-Adjusted Charlson Comorbidity Index (p = 0.001), clinical stage of cancer (stage I 0.0% vs. stage II 20.8% vs. stage III 20.8% vs. stage IV 58.5%; p < 0.0001), patients origin (hospital ward 75.5% vs. emergency department 24.5%; p < 0.0001), type of admission (elective 28.3% vs. unplanned 71,7%; p < 0.0001), nature of admission (malignancy-related 49.1% vs. non-malignancy-related 50.9%; p < 0.0001), sepsis (73.6 vs. 26.45%; p < 0.0001), chemotherapy-induced adverse event (79.2 vs. 20.8%; p < 0.0001), invasive mechanical ventilation (66.0 vs. 34.0%; p < 0.0001), and APACHE II score (p = 0.001) were related with ICU mortality.

With regard to hospital mortality, Table 1 shows that higher Age-Adjusted Charlson Comorbidity Index score, primary tumour location, higher clinical stage of cancer, patients admitted from emergency department, unplanned admission, non-malignancy-related admission, sepsis, chemotherapy-induced adverse event, invasive mechanical ventilation, and higher APACHE II score were related to a higher hospital mortality rate in univariate analysis.

In multivariate logistic regression analysis, after adjust for confounders, stage IV of cancer, patients admitted from emergency department, unplanned admission, non-malignancy-related admission, sepsis, chemotherapy-induced adverse event, and invasive mechanical ventilation were independently associated with increased risk of hospital mortality (Table 3).

Table 3 Results of multivariate logistic regression analysis of factors associated with hospital mortality in cancer patients admitted to the intensive care unit
Full size table

Discussion

This prospective study showed an overall ICU and hospital mortality rate of 10.2 and 13.0%, respectively. Mortality rates were higher in the previous studies (Anisoglou et al. 2013; Mânica et al. 2013; Ostermann et al. 2013; Yoo et al. 2013; Aygencel et al. 2014). Differences in the composition of the studied participants could explain these differences; for example, elective postoperative patients prevailed in our study. In addition, MV and vasopressors were only required for 15.5% of patients. In this regard, characteristics of cancer patients admitted to ICU vary around the world. Results of studies analyzing a wide cohort of patients are similar to those found by us, with a VM and vasopressors rate of 18.0–28.4 and 10.8–23.3%, respectively (Soares et al. 2010; Xing et al. 2015; Fisher et al. 2016; Shrime et al. 2016). However, ICU and hospital mortality rate was 28.8 and 35.6%, respectively, when unplanned patients were analysed. These mortality rates are comparable to those reported in international cohorts (Bos et al. 2012; Aygencel et al. 2014; Fisher et al. 2016).

The major findings of this study were that stage IV of cancer, patients admitted from the emergency department, unplanned admission, non-malignancy-related admission, sepsis, chemotherapy-induced adverse event, and invasive mechanical ventilation were independent clinical risk factors for hospital mortality in multivariate analysis.

Advanced cancer has been associated with higher mortality in the previous studies (Fisher et al. 2016; Xia and Wang 2016). Such relationship could be explained by two ways: (1) because of infiltrative/ metastatic organ dysfunction; (2) increasing the risk of other mortality-related factors such as acute respiratory failure (ARF) and sepsis. In fact, both severe ARF requiring MV and sepsis were factors independently associated with mortality in our study. For ventilated cancer patients, ICU mortality rates and hospital mortality rates have been reported greater than 50% (Almeida et al. 2014; Martos-Benítez et al. 2017) and 65% (Azevedo et al. 2014; Martos-Benítez et al. 2017), respectively. Barotrauma, oxygen toxicity, hemodynamic compromise, ventilator-induced lung injury, ventilator-associated pneumonia, as well as local and systemic effects of tumour explain higher mortality associated with MV (Slutsky and Ranieri 2013; Blot et al. 2014; Serpa et al. 2014; Spieth et al. 2014). On the other hand, sepsis is a major risk factor for mortality among critically ill cancer patients (Schellongowski et al. 2016). Recent data suggest that active disease, haematological malignancies, compromised performance status, presence of 3–4 systemic inflammatory response syndrome criteria, the number of acute organ dysfunctions, and polymicrobial infections are associated with increased mortality in these subjects (Rosolem et al. 2012; Torres et al. 2015).

Patients with cancer are frequently encountered in the emergency department due to clinical conditions requiring ICU admission such as infection, acute respiratory failure, need of emergency surgery, haemorrhage, and complications of chemotherapy (Young and Simmons 2014; Cornillon et al. 2016; Lash et al. 2017). Recently, García-Gigorro et al. found that ICU complications such as acute renal failure and acute respiratory distress syndrome were the variables with the greatest impact on hospital mortality among patients admitted to ICU from the emergency department (García-Gigorro et al. 2017).

In cancer patients with unplanned admission to ICU, the hospital mortality rate ranges between 30 and 55% (Bos et al. 2012; Aygencel et al. 2014; Fisher et al. 2016). In addition, mortality rate is almost twice higher in medical cancer patients (40.6%) than in medical patients without cancer (23.7%) as recently reported Bos et al. (Bos et al. 2012). The presence of metastasis, APACHE II score, a Glasgow Coma Scale < 7 on admission to ICU, sepsis/septic shock, and vasopressor requirement during ICU stay are the main mortality-explicative variables (Aygencel et al. 2014; Fisher et al. 2016).

The nature of admission has not been analysed in the previous studies, so it is an important contribution of this study with significant clinical implication because of their impact on prediction of mortality. We considered malignancy-related admission when tumoural tissue was directly implicated as cause of ICU admission. Otherwise, the nature of admission was defined as non-malignancy-related. This definition should be used in future studies to be compared with our results.

A recent systematic review suggests that adverse drug events requiring ICU admission are frequents (Jolivot et al. 2014). Oncospecific-drug therapy is associated with several adverse events leading clinical decline, impairment of quality of life, and death (Davis 2015). In fact, disorders such as neutropenia, mucositis, pneumonitis, and encephalopathy are commonly related with drugs toxicity and mortality in cancer patients admitted to ICU (Legrand et al. 2012; Yoo et al. 2013; Iacovelli et al. 2014).

Length of ICU stay and time of hospitalization was lower than those reported in specific subgroups of critically ill cancer patients (e.g., cancer patients with sepsis, lung cancer patients, and cancer patients with acute respiratory failure) (Mânica et al. 2013; Ferreira et al. 2015; Torres et al. 2015). However, length of stay was comparable with those described in studies analyzing a wide cohort of critically ill patients with cancer (Soares et al. 2010; Bos et al. 2012; Xing et al. 2015; Fisher et al. 2016), including a large prospective study carried out in several ICU of Europe (Taccone et al. 2009).

The present study has several shortcomings. First, our study was conducted at a specialized ICU of single institution confined to oncological patients. Therefore, the result of this study may not be generalized to other general medical centres. Second, there were no patients who had leukaemia, and haematological patients represented a small number of participants. Finally, we have evaluated a relatively large number of patients for a single centre study; however, it could be considered as a limitation.

Conclusions

Clinical predictors of hospital mortality for cancer patients admitted to the ICU are advanced cancer, patients admitted from the emergency department, unplanned admission, non-malignancy-related admission, sepsis, chemotherapy-induced adverse event, and invasive mechanical ventilation. The decision to admit critically ill patients with cancer to the ICU should be based on the probability of surviving the acute illness and not on malignance by itself. Consequently, a general reluctance to admit these patients to the ICU is not justified. Palliative care is a justified alternative for patients with severe pathophysiological disturbances whom ICU admission is considered as futile; but, critically ill patients with cancer should always receive a specialized care. Prospective studies examining the impact of these and others clinical prognostic factors on outcomes are warranted.