The Palliative Index: Predicting Outcomes of Emergent Surgery in Patients with Cancer

Abstract

Background: The role of emergent palliative surgery in the setting of advanced malignancy remains a subject of controversy.

Objective: The purpose of this study was to identify clinical predictors of outcome in patients with cancer who undergo nonelective abdominal surgery.

Setting/Subjects: Individuals who underwent urgent and emergent abdominal operations between 2006 and 2010 at a tertiary cancer center were identified.

Measurements: Analyses were performed to identify predictors of 30-day morbidity and mortality as well as overall survival (OS). A risk score was derived from predictors of OS.

Results: Of 143 patients, 93 (65%) had active disease (AD; defined as evidence of malignancy at time of surgery). Thirty-day morbidity and mortality were 36.4% and 9.8%, respectively. Independent predictors of 30-day mortality included ASA score >3 (p=0.009) and albumin <2.8 (p=0.040). Median OS was 5.4 months in patients with AD and was not reached in patients without AD (p<0.001). Independent predictors of decreased OS included AD; ASA >3; creatinine >1.3; and a tumor-related indication (i.e., bleeding, obstructing, or perforating tumor). A risk or palliative index (PI) score stratified patients into groups with discreet outcomes.

Conclusions: Although AD did not predict 30-day morbidity, it was the dominant independent predictor of postoperative OS. In cancer patients undergoing emergency abdominal surgery, outcome is anticipated by disease status and other independent predictors of OS.

Introduction

Emergent surgery is associated with considerable morbidity and mortality.^1,2 Recent data from the American College of Surgeons National Surgical Quality Improvement Program (NSQIP) confirmed markedly elevated 30-day mortality and morbidity rates after emergent compared to elective general surgery (5.8% versus 0.8%, p<0.0001 and 19.8% versus 8.8%, p<0.0001, respectively).³ Urgent or emergent surgery is also associated with a substantially increased cost compared to elective surgery.^4,5 The utilization of such resources can often be justified by the potential for lifesaving intervention in patients with a meaningful potential for recovery. In contrast, urgent or emergent surgery may be more difficult to justify in patients with advanced malignancy and a limited life expectancy.

Previous studies have evaluated outcomes after elective and emergent palliative operations⁶ or presentation with specific acute surgical indications (e.g., pneumoperitoneum) in cancer patients.^7,8 While these studies afford insights into the management of surgical disease in patients with advanced malignancy, few studies have specifically explored the broader implications of urgent or emergent operations in this cohort. Acute surgical disease in patients with cancer poses a unique clinical challenge. High patient acuity, the expectation of perioperative morbidity, and poor long-term prognosis often obfuscate treatment priorities. While surgeons at general hospitals encounter this challenging scenario occasionally, it is commonplace at tertiary cancer centers. In this study we analyze the outcomes of urgent and emergent operations at a tertiary cancer center and identify predictors of perioperative morbidity and mortality, and overall survival (OS). We also propose a risk stratification scoring system that incorporates clinical predictors of OS.

Methods

Study cohort

After approval by the institutional review board, all urgent or emergent (nonelective) abdominal operations performed at our institution between September 1, 2006 and August 31, 2010 were identified from the operative case logs. Reoperative cases within 30 days after a prior elective abdominal operation were excluded. Medical records were retrospectively reviewed to identify patient and disease characteristics. Preoperative laboratory studies, underlying cancer diagnosis and status, and indications for emergency intervention were extracted.

Laboratory results of specific interest included leukocytosis (defined as white blood cell count, WBC >11 K/μL); neutropenia (defined as absolute neutrophil count, ANC <1 K/μl); elevated creatinine (defined as serum creatinine >1.3 mg/dL); and hypoalbuminemia (defined as serum albumin <2.8 g/dL). Malignancies were categorized as solid or hematologic as recorded by the treating oncologists. Cancer status was categorized as active disease (AD) versus no active disease (NAD); AD was defined as presence of measurable tumor burden or active hematologic malignancy identified on preoperative or intraoperative evaluation. American Society of Anesthesiologists (ASA) physical status scores were obtained from anesthesia records. Complications of the underlying malignancy that necessitated surgery (i.e., obstruction, perforation, or bleeding from tumor) were termed “tumor-related” indications.

Outcome measures

Perioperative morbidity and mortality rates were calculated by summing adverse events and deaths, respectively, during the first 30 postoperative days and dividing the number of events by the total number of patients. Postoperative complications were classified using the system proposed by Clavien.⁹ In brief, grade 1 complications do not require therapy other than analgesic, antipyretic, antiemetic, or antidiarrheal drugs, or those to treat lower urinary tract infection. Grade 2 complications are defined as potentially life threatening, and require intervention or prolonged hospital stay. Grade 3 complications lead to lasting disability or organ resection, and grade 4 complications are those that lead to mortality. Follow-up duration and overall survival (OS) were calculated from the time of emergent surgery.

Statistical analyses

Continuous variables were described as medians, and categorical variables were described as numbers (percentage). Groups were compared using chi-squared or Fisher's exact tests for categorical variables, and the Wilcoxon rank-sum test for continuous variables. Overall survival (OS) was estimated using the Kaplan-Meier method from the time of emergent surgery. The long-term OS of the AD and NAD cohorts were compared using the log-rank test. Independent predictors of perioperative morbidity and mortality and OS were identified by logistic regression. Cox proportional hazards models were fit with variables associated with a p value of <0.15 on univariate analyses.

Derivation of scoring system

Independent predictors of OS were weighted based on the associated relative risk of death (hazard ratio). The most powerful predictor was given twice the weight (allotted two points) of the other variables (allotted one point each), and a risk scoring system or palliative index (PI) was created. After stratification by score, rates of OS in the subgroups were compared using the log-rank test. A two-sided p<0.05 was considered statistically significant. Statistical analyses were performed by SPSS 19.0 (IBM, Armonk, NY).

Results

Patient characteristics

One hundred and forty-three patients that met inclusion criteria were identified. The median age was 59 years (range 20–90). Sixty-one (43%) patients were female. Ninety-three patients (65%) had AD and among these, 67 (72%) had stage IV disease. A comparison of patients with and without AD (see Table 1) shows similar characteristics with two notable exceptions; a greater percentage of the patients with AD were hypoalbuminemic (26% versus 9%, p=0.002) and had an ASA score >3 (26.9% versus 6%, p=0.0012).

Table 1.

Preoperative Characteristics of Cancer Survivors with and without Active Disease Undergoing Emergent Surgical Intervention

Variable	Total (%)	AD (%)	NAD (%)	p
Number	143	93	50
Median age (range)	59.0 (20–90)	59.0	58.5	0.83
Sex				0.63
Female	61	41 (44%)	20 (40%)
Male	82	52 (56%)	30 (60%)
ASA score				0.0015
1–3	115	68 (73%)	47 (94%)	0.0012
4–5	28	25 (27%)	3 (6%)
WBC (K/μL)
Median	7.30	7.20	7.35	0.53
≤12	114	70 (75%)	44 (88%)	0.06
>12	29	23 (25%)	6 (12%)
ANC (K/μL)
Median	5.40	5.40	5.85	0.57
≤1	5	4 (4%)	1 (2%)	0.45
>1	138	89 (96%)	49 (98%)
Creatinine (mg/dL)
Median	0.8	0.8	0.8	0.71
≤1.3	125	80 (86%)	45 (90%)	0.49
>1.3	18	13 (14%)	5 (10%)
Albumin (g/dL)
Median	3.5	3.2	3.9	<0.0001
≤2.7	27	24 (28%)	3 (6%)	0.002
>2.7	105	63 (72%)	42 (84%)

AD, active disease; ANC, absolute neutrophil count; ASA, American Society of Anesthesiologists; NAD, no active disease; WBC, white blood cell count.

Emergency operations

Indications for operation included peritonitis (e.g., from appendicitis, diverticulitis, cholecystitis, perforated viscous, etc.) or pneumoperitoneum in 52 cases (36%), gastrointestinal tract obstruction in 84 cases (59%), and active bleeding in 7 cases (5%). Indications for surgery were tumor related in 52 cases (56%). Operations included bowel resection in 54 cases, adhesiolysis in 19 cases, intestinal or biliary bypass in 15 cases, intestinal diversion in 15 cases, herniorrhaphy in 14 cases, cholecystectomy in 10 cases, appendectomy in 10 cases, repair of a perforation in 3 cases, drainage in 1 case, diagnostic laparotomy in 1 case, and diagnostic laparoscopy in 1 case. The overwhelming majority of patients (94%) underwent open operations. The remaining 6% of patients underwent laparoscopic procedures.

Perioperative mortality and morbidity

The 30-day mortality rate was 9.8%, 13% (12 patients) in patients with AD and 4% (2 patients) in patients without AD. These rates were not significantly different (p=0.139). Univariate predictors of 30-day mortality included ASA score >3 (p<0.001), leukocytosis (p=0.028), elevated creatinine (p=0.018), and hypoalbuminemia (p=0.001). On multivariate analysis, independent predictors of 30-day mortality were ASA >3 (HR 5.93, p=0.009), elevated creatinine (HR 4.42, p=0.050), and hypoalbuminemia (HR 4.12, p=0.040) (see Table 2).

Table 2.

Preoperative Clinical Factors Associated with 30-Day Postoperative Mortality

	30-day mortality		Univariate	Multivariate
Variable	No	Yes	P	HR	95% CI	p
ASA score
4–5	19	9	<0.001	5.93	1.57–22.41	0.009
1–3	110	5
WBC (K/μL)
>11	23	6	0.038	1.67	0.41–6.86	0.475
<11	54	60
ANC (K/μL)
<1	3	2	0.660
>1	69	69
Creatinine (mg/dL)
>1.3	13	5	0.018	4.42	1.00–19.53	0.050
<1.3	116	9
Albumin (g/dL)
<2.8	21	8	0.001	4.12	1.07–15.87	0.040
>2.8	108	6
AD	82	12	0.139	1.63	0.30–8.96	0.572
NAD	47	2

AD, active disease; ANC, absolute neutrophil count; ASA, American Society of Anesthesiologists; NAD, no active disease; WBC, white blood cell count.

The 30-day complication rate was 36.4%, 39.8% in patients with AD and 30% in patients with NAD (p=0.556). The rate of Clavien grade 4 complications was 14% in patients with and without AD. None of the studied clinical variables was significantly associated with 30-day morbidity.

Overall survival

The median follow-up of the entire cohort was 9.2 (range, 0–50) months, 24.9 months in surviving patients. Median OS was 14.3 months (95% CI, 7.5–21.1). Median OS was 5.8 (95% CI, 3.8–7.8) months among those with AD, but was not reached in patients without AD (p<0.001; see Figure 1). Independent predictors of OS included ASA >3 (HR 2.16, 95% CI, 1.27–3.66, p<0.004), elevated creatinine (HR 2.60, 95% CI, 1.41–4.79, p=0.002), tumor-related indication for surgery (HR 1.76, 95% CI, 1.54–2.96, p<0.031), and AD (HR 4.12, 95% CI, 1.94–8.75, p<0.001) (see Table 3).

FIG. 1.

Overall survival in patients with and without active disease.

Table 3.

Univariate and Multivariate Predictors of Postoperative Overall Survival and Derivation of Preoperative Risk Score for Postoperative Prognosis

	Univariate	Multivariate			Risk score
Clinical factors	p	HR	95% CI	p	Points assigned
Tumor-related indication	<0.001	1.764	1.54–2.96	0.031	1
ASA >3	<0.001	2.155	1.27–3.66	0.004	1
Creatinine >1.3	0.006	2.596	1.41–4.79	0.002	1
AD	<0.001	4.123	1.94–8.75	<0.001	2
WBC >12 K/μL	0.005	1.252	0.73–2.16	0.419
Albumin <2.8 g/dL	0.004	1.594	0.95–2.67	0.076
Stage IV	<0.001	0.958	0.57–1.61	0.870

AD, active disease; ASA, American Society of Anesthesiologists; HR, hazard ratio; WBC, white blood cell count.

Stratification of patients using a palliative index

All patients were assigned a risk score with one point allotted for ASA >3, elevated creatinine, and tumor-related indication, each. Two points were allotted for AD. Scores of 0, 1–2, 3, and 4–5 were associated with significant differences in median survival (p<0.001; see Figure 2). Patients with a score of 1–2 had a median survival of 14.50 (95% CI, 4.04–24.97) months. Patients with a score of 3 had a median survival of 5.20 (95% CI, 3.24–7.16) months. Patients with a score of 4–5 had a median survival of 1.1 (95% CI, 0.12–2.08) months. This translated into 12-month actuarial OS of 92%, 56%, 33%, and 18%, in the four groups, respectively (p<0.001; see Table 4).

FIG. 2.

Overall survival stratified by risk score.

Table 4.

Postoperative Overall Survival Stratified by Preoperative Risk Score

	Actuarial overall survival
Score	6 month	12 month	24 month
0	97%	92%	86%
1–2	82%	56%	32%
3	46%	33%	22%
4–5	25%	18%	0%

Discussion

In this study, AD was not an independent predictor of early (<30 day) morbidity after nonelective surgery, but was a powerful predictor of poor OS. A scoring system incorporating disease status with other independent predictors of poor OS (high ASA score, elevated creatinine, and tumor-related indication for surgery) had prognostic value. These findings suggest that palliative surgery can be offered to patients with malignancy and acute abdominal surgical disease with a substantial but acceptable risk of early morbidity and mortality comparable to that associated with urgent or emergent surgery in patients without AD. The prognosis in patients with cancer who require such interventions, however, is poor overall and a subgroup of patients, identifiable on the basis of clinical parameters, have an especially grave prognosis.

A comparison between the short-term outcomes reported here and those in studies focusing on both elective and emergent surgery in patients with advanced malignancy provides one context for the findings of the current study. Miner and colleagues reported on 1022 palliative procedures at Memorial Sloan-Kettering Cancer Center and reported 30-day perioperative morbidity and mortality of 29% and 11%, respectively.⁶ Tseng and colleagues evaluated 30-day morbidity and mortality after operation in patients with stage IV cancer.¹⁰ Unadjusted 30-day morbidity and mortality rates in that study were 28% and 9%, respectively. Accounting for the selection of only urgent and emergent cases, a paucity of small procedures, and extent of comorbidity, the 30-day morbidity and mortality rates in the current study (40% and 14%, respectively) are comparable.

The role that careful patient selection played in these reported outcomes should not be underestimated. The current study included only those patients who underwent surgery rather than patients evaluated for surgery; many others declined, or were not offered operation. Some insight into the selection process is provided in a study reported by Badgwell and colleagues of 123 patients with cancer and pneumoperitoneum from our institution.⁷ Seventy-two patients were managed operatively with a 30-day mortality of 15% and complication rate of 46%, rates very close to those in the present series. Forty-two percent of the total cohort was managed nonoperatively. While this cohort is discrete from that in the present study (i.e., the study intervals do not overlap), this rate of operative intervention reflects the practice patterns that influenced outcomes in the current study as well.

Despite acceptable short-term outcomes, and it bears emphasis that many of the patients who underwent operation would have succumbed to their acute illness without operation, the prognosis for patients with cancer and acute intraabdominal disease requiring operation is poor, overall. Indeed, the OS observed in patients with AD affirms that most urgent operations in these patients should be viewed as palliative. The likelihood of short-term morbidity and mortality is, therefore, important when considering surgical intervention. In our series, high ASA score, elevated creatinine, and hypoalbuminemia predicted higher risk of perioperative mortality. Prediction of perioperative risk was also the focus of a recent study of morbidity and mortality after surgery in patients with stage IV cancer. In that study, Tseng and colleagues identified older age, impaired respiratory function, ascites, hypoalbuminemia, elevated creatinine, and leukocytosis as predictors of mortality and morbidity.¹⁰

While valuable, studies that describe short-term outcome only do not address the question that is frequently of equal or greater concern to surgeon and patient—what is the likelihood of long-term survival? Indeed, few previous studies have addressed overall survival after urgent and emergent surgery in patients with malignancy. A recent study of patients on phase I clinical trials highlighted the dismal outcomes associated with ICU admission in patients with advanced cancer. Interestingly, patients admitted to the ICU after surgical intervention did markedly better than those admitted for other reasons.¹¹ The present study goes further in describing surgical outcomes in patients with advanced malignancy and similarly suggests that surgical intervention is justified in selected patients. Likewise, poor overall survival can and should be anticipated in a subset of patients.

The experience described in the present study is unusual in that it includes patients with active cancer diagnoses as well as cancer survivors. The latter serve as a control group; the operative indications in this group were, by definition, benign. Patients with PI scores of 0 and 1, therefore, more closely resemble urgent and emergent surgical patients routinely treated at general hospitals, with the important caveat that these patients have all been affected by malignancy. Patients with scores of 1 and 2 were grouped together on account of their similar observed outcomes, suggesting that in the absence of an elevated creatinine, high ASA score, or tumor-related surgical indication, patients with AD fare as well as patients without AD. In contrast, the presence of additional adverse clinical features shortens expected survival substantially. Indeed, patients with PI scores of 4 and 5 have a grave prognosis and are less likely to derive meaningful benefit from aggressive intervention.

This study has several limitations, including a relatively small and heterogeneous patient cohort. The former may have served to minimize the observed difference in short-term outcome between patients with and without AD. Notwithstanding, this study does reflect a breadth and depth of experience somewhat unique to a high volume, tertiary cancer center and highlights the frequent tensions between patients' long-term cancer care goals and immediate palliative needs. Importantly, the proposed risk stratification score should not be taken as a tool to exclude patients from emergent surgical intervention (e.g., some patients with high scores but meaningful cancer treatment options may be reasonable surgical candidates). Rather, these data provide a basis for patient counseling.

This study does not address many important implications of emergent surgery in the setting of malignancy, particularly its impact on quality of life and its cost. Indeed, as issues of cost and health care priorities assume a dominant place in the public discourse, decision making in these cases will become even more complex. Further studies should address the cost and value (not only in length but in quality of life) associated with these interventions. This study does affirm that the approach to emergent surgical disease in the setting of cancer, particularly advanced cancer, must reflect patients' priorities and be grounded in realistic expectations. These data provide a basis for these expectations.

Acknowledgments

This work was supported by the University of Texas Cancer Center Core Support Grant (P30CA016672). The authors wish to acknowledge Ryan Yan Ling for his contributions to this project.

Author Disclosure Statement

No competing financial interests exist.