Patient Comorbidities Drive High Mortality Rates Associated with Major Liver Resections Irrespective of Hospital Volume

Abstract

Introduction:

Major hepatectomies are utilized to manage primary hepatic malignancies. Reports from high-volume centers (HVCs) with minimal perioperative mortality focus on multiple aspects of perioperative care, although patient-specific factors remain unelucidated. We identified patient factors associated with outcomes and examined whether these contribute to survival differences.

Methods:

We queried the National Cancer Database (2006–2015) for patients with primary liver malignancies managed with major hepatectomy. Facilities were dichotomized by volume (high volume: >15 hepatectomies/year). Perioperative outcomes were compared based on patient demographic and clinical characteristics as well as center volume.

Results:

4263 patients were included with 78.5% receiving care in low-volume centers (LVCs). 90-day postoperative mortality was higher in LVCs vs. HVCs (12% vs. 7.5%; P < .001). Factors associated with undergoing surgery in LVCs included: living in areas with lower income (P = .006) and education (P < .001), having nonprivate insurance (P < .001), residing near the care center (P < .001), and having a comorbidity score (CDS) >1 (P = .014). Patients with CDS ≤ 1 had higher 90-day mortality in LVCs (11.3% vs. 6.6%; P < .001) and had similar outcomes in LVCs and HVCs (15.6% vs. 13.7% P = .6). Patients with CDS > 1 were more likely to receive care in LVCs (16.3% vs. 12.7%; P < .001).

Conclusion:

Reduced perioperative mortality following major hepatectomy in HVCs is driven by optimal management of patients with low CDS. However, nearly 1 in 5 patients who undergo major hepatectomies have a high CDS and approximately 15% of them succumb in the perioperative period irrespective of the treating centers’ experience.

Introduction

The incidence of primary liver malignancies, which include hepatocellular carcinoma (HCC) and intrahepatic cholangiocarcinoma (iCCA), is expected to increase over the next several decades principally as a result of increasing rates of nonalcoholic steatohepatitis.^1,2 Both HCC and iCCA are associated with advanced stage at presentation, portending a dismal prognosis. When possible, surgical resection offers the only hope for long-term cure.^3–7 Resection for such patients commonly involves performing a major hepatic resection, defined as the resection of 4 or more Couinaud liver segments.^8,9 Despite improved rates of morbidity and mortality following major hepatectomy, the operation is not without risk. Specifically, patients undergoing major hepatectomy for primary liver cancers carry particularly high rates of hepatic dysfunction, a major risk factor for post-hepatectomy liver failure and mortality.^10–16

The volume of complex surgical procedures performed by a given center is an established factor affecting short-term postoperative outcomes.^17–19 A major hepatic resection offers a prime example of this association. Reports from specialized hepatopancreatobiliary surgery centers describe particularly good outcomes with perioperative mortality as low as 4%.^11,14 However, patients who receive care in such centers may not be a representative of the general population suffering from primary liver malignancies. Patients receiving care in specialized centers are likely to have the means to travel and good health status, both of which may not reflect the overall population with primary liver malignancies. Outcomes following major hepatectomy in low volume centers (LVCs) and the patient population receiving care in such centers are not well described. Several patient factors have been found to contribute to the likelihood of receiving care in an LVC for other malignancies including distance from high volume centers (HVC) and limited means to travel great distances to receive care.²⁰ In addition, race and socioeconomic factors such as income, education, and insurance status have been shown to be predictive of receiving care in LVCs. These factors are typically associated with patients carrying higher comorbidity indices, a leading cause of worse outcomes following complex operations.^21,22

To date, only single highly experienced center’s retrospective reviews have examined patient outcomes following major hepatectomy. Little is known about the overall short-term outcomes of major hepatectomy or the differences between patients who undergo this procedure in LVCs vs. HVCs. Our aim was to provide a nation-wide examination of patient-specific factors following major hepatectomy for primary liver cancers. Specifically, we sought to identify patient characteristics associated with receiving care in LVCs and to identify risk factors associated with worse perioperative outcomes.

Methods

Study Design

This study was exempt from review from our institutional review board. The National Cancer Database (NCDB) was the source of all analyzed data. The NCDB is a joint program of the Commission on Cancer (CoC) of the American College of Surgeons and the American Cancer Society (ACS) and is a hospital-based registry with data from more than 1500 CoC-accredited hospitals. It includes information about demographics, disease stage, comorbidity, and the first course of treatment for 70% of newly diagnosed cancer cases in the United States. The CoC and ACS have not verified and are not responsible for the analytic or statistical methodology used or for the conclusions drawn from these data. The NCDB was queried to identify all patients with a primary liver malignancy who underwent a major hepatectomy from 2006 to2015. Major hepatectomy was defined as the surgical resection of 4 or more Couinaud liver segments (8, 9). Patients with HCC and iCCA were then identified using the International Classification of Disease for Oncology 3rd edition (ICD-O-3) primary site code for liver and intrahepatic bile ducts (22.0) and the histology codes for HCC (8170 through 8175) and iCCA (8160). Nonoperative cases, resections less than major resections, and resections for other malignancies were excluded from this analysis (Supplemental figure 1).

Patient and Center Characteristics

Standard demographic data including age (<65 vs. ≥65 years), sex, race (white, black, and other), insurance status (private, Medicaid/uninsured, and Medicare), median area of residence income(<$38 000, $38–63 000, and >$63 000), percentage of high school graduates in the area of residence (<21% vs. >21%), type of area of residence (metro/urban vs. rural), and “crowfly” or distance from hospital (within or beyond 50 miles from the center where the patient received surgery^23,24) were recorded. Clinicopathologic data included Charlson-Deyo Comorbidity Score (CDS), disease histology (HCC vs. iCCA), and diagnosis year (Table 1). Patients with CDS = 0 and CDS = 1 were grouped as CDS ≤ 1 and patients with CDS of 2 or more were grouped as CDS > 1 for the purpose of our analyses. Patients younger than 18 years of age or without evidence of malignancy (T0) on final pathology were excluded.

Table 1.

Study Patient Population Characteristics.

Total patients n (%)	4263 (100)
Gender
Male	2858 (67.0)
Female	1405 (33.0)
Age
≥65 years	2285 (53.6)
<65 years	1978 (46.4)
Race
White	3035 (71.2)
Non-Nonwhite	1228 (28.8)
% HS diploma
Low <21%	3350 (78.6)
High ≥21%	913 (21.4)
Income
< $38 000	790 (18.5)
$38–63 000	2124 (49.8)
>$63 000	1349 (31.7)
Insurance
Uninsured/Medicaid	521 (12.2)
Medicare	2102 (49.3)
Private insurance	1640 (38.5)
Crowfly
<50 miles	3261 (76.5)
>50 miles	1002 (23.5)
Patient residence
Metro/Urban	4210 (98.7)
Rural	53 (1.3)
Disease histology
HCC	3694 (86.6)
iCCA	569 (13.4)
Charlson-Deyo index
No comorbidities	2416 (56.7)
1 comorbidity	1189 (27.9)
>1 Comorbidities	658 (15.4)
Center Type
Academic	2671 (62.6)
Community	1592 (37.4)
Center location
Atlantic	1909 (44.8)
Central	1676 (39.3)
Pacific	678 (15.9)
Center volume
Low-volume center	3342 (78.5)
High-volume center	921 (21.5)

Abbreviations: HCC, hepatocellular carcinoma; ICCA, intrahepatic cholangiocarcinoma.

Data on the centers were then documented including center type (academic vs. community), location (east coast, central, and west coast), and the number of hepatic resections performed. This number was used to code and categorize treatment centers as either high or low volume (HVC vs. LVC). Prior to exclusion of nonmajor hepatectomy procedures, treatment centers were individually analyzed to obtain the average annual number of hepatic resections (<4 Couinaud liver segment and wedge resections excluded) performed. Using previously defined cutoffs for volume of hepatic resections, treatment centers were stratified into low volume (fewer than 15 HR per year) vs. high volume (15 or more HR hepatic resections per year).¹¹ We defined HVCs as having performed more than 15 hepatic resections per year for at least 2 years of the study period. The 90-day mortality on all patients was subsequently recorded.

Statistical Analysis

Descriptive statistics were calculated for all variables of interest. Categorical measures were summarized using frequencies with percentages. Comparisons of categorical variables were performed using the chi-squared or Fisher’s exact test. P values < .05 were considered statistically significant. Poisson regression with robust error variance models were constructed to identify factors associated with risk of 90-day mortality. The risk ratio for 90-day mortality in an LVC compared to an HVC was also calculated. Clinically significant confounding variables and/or those with P values of < .2 on bivariable analysis were included in the model for adjustment. Measures of association were presented as risk ratios (RRs) or percent increase/decrease in risk with 95% confidence intervals (CIs). In addition, patients were stratified by CDS (scores ≤1 vs. > 1), and subgroup analysis was performed evaluating associations between 90-day mortality and volume center status using logistic regression. Statistical analyses were conducted using R v3.5.0 (R foundation for statistical computing, Vienna, Austria).

Results

Outcomes According to Surgical Volume

4263 patients met our inclusion criteria, including 3346 (78.5%) who received care at LVCs and 917 (21.5%) who received care at HVCs. There were 17 HVCs and 590 LVCs identified (Supplemental figure 1). The mean number of hepatectomies performed per year in HVCs was 29.6 (15.3–58) vs. 2.1 (.08–14.8) for LVC. The 90-day mortality among those who underwent major hepatectomy in LVCs was 12.0% (n = 401) compared to 7.5% (n = 69) in HVCs (P < .001). The RR for 90-day mortality in an LVC when compared to an HVC was 1.60 (CI 1.25–2.05; P < .001).

Characteristics of Patients Treated in Low-Volume Centers

Bivariable analysis of socioeconomic and demographic factors revealed that race, income, education, and insurance status were different among those who underwent a major hepatic resection at an LVC compared to an HVC (Table 2). On multivariable analysis, patients with Medicare insurance were 8.9% (P < .001) more likely to receive care at an LVC. Those who lived in areas with a median income of $62 999 or less had a higher likelihood of receiving care in an LVC (<$38 000, P < .001 and $38 000–62 999, P = .012; respectively). We also noted that patients living in areas with fewer than 21% high school graduates were 8.6% (P < .001) more likely to receive care at an LVC (Figure 1).

Table 2.

Comparison of Demographic And Disease Characteristics of Patients Having Undergone A Major Hepatic Resection in (LVC) Compared to (HVC).

	Patients in LVC (%)	Patients in HVC (%)	P-value
Total patients n (%)	3346 (78.5)	917 (21.5)
Gender
Male	2223 (66.4)	635 (69.2)	.109
Female	1123 (33.6)	282 (30.8)
Age
≥65 years	1817 (54.3)	468 (51.0)	.085
<65 years	1529 (45.7)	449 (49.0)
Race
White	2418 (72.3)	617 (67.3)	<.001a
Black	465 (13.9)	97 (10.6)
Asian and other	463 (13.8)	203 (22.1)
Education (% HS diploma)
≥21%	698 (20.9)	215 (23.4)	.008a
<21%	2648 (79.1)	702 (76.6)
Income
< $38 000	649 (19.4)	141 (15.4)	.006a
$38 000-$62 999	1654 (49.4)	470 (51.3)
> $63 000	1043 (31.2)	306 (33.4)
Insurance
Uninsured/medicaid	412 (12.3)	109 (11.9)	<.001a
Private	1230 (36.8)	410 (44.7)
Medicare	1704 (50.9)	398 (43.4)
Crowfly
<50 miles	2707 (80.9)	554 (60.4)	<.001a
>50 miles	639 (19.1)	363 (39.6)
Patient residence
Metro/urban	3303 (98.7)	907 (98.9)	.712
Rural	43 (1.3)	10 (1.1)
Charlson-deyo index
No comorbidities	1859 (55.6)	557 (60.7)	.007a
1 comorbidity	945 (28.2)	244 (26.6)
>1 comorbidities	542 (16.2)	116 (12.6)
Disease histology
HCC	2895 (86.5)	799 (87.1)	.63
iCCA	451 (13.5)	118 (12.9)

Abbrevations: LVC, low-volume centers; HVC, high-volume centers; HCC, hepatocellular carcinoma; iCCA, intrahepatic cholangiocarcinoma.

^aPatient and disease characteristics found to be statistically significantly different between patient receiving care in LVC and HVC on bivariable analysis. These variables were subsequently included and incorporated into a multivariable analysis.

Figure 1.

Forest plot displaying the percent likelihood (risk ratio) of undergoing a major hepatic resection in a low-volume center based on patient characteristics listed on the left.

Distance travelled to receive care was also significantly different among those who received care in an LVC compared to those receiving care in an HVC. On multivariable analysis, patients living farther (>50 miles) from the center where they underwent surgery were 26.5% (P < .001) less likely to receive care in an LVC (Figure 1).

Bivariable analysis revealed that patients receiving care in LVCs tended to have higher CDS. On multivariable analysis, patients with CDS >1 were 5.3% (P = .014) more likely to receive care in an LVC when compared to patients with CDS ≤ 1.

Influence of Volume Status on Outcomes Among Patients with High Comorbidity

Among patients with CDS ≤ 1, 90-day mortality was 11.3% in LVCs compared to 6.6% in HVCs (P < .001). Interestingly, among patients with CDS > 1, 90-day mortality was 15.6% among patients who underwent major hepatectomy in LVCs compared to 13.6% in HVCs (P = .600) suggesting that perioperative mortality is not affected by center volume status among high-risk patients. While the majority of patients in both LVCs and HVCs had CDS ≤ 1 (83.7% and 87.3%, respectively), LVCs operated on a significantly greater proportion of patients with CDS > 1 than HVCs (16.3% vs. 12.7% (P < .001)) (Figure 2).

Figure 2.

Comparison of mortality rates among patients with low vs. high CDS. Left columns compare overall mortality rates regardless of CDS. Middle columns compare patients with CDS ≤1. Right columns compare patients with CDS >1. (***:P-value < .001, ns: nonsignificant). CDS, comorbidity score.

Discussion

The impact of center volume on perioperative outcomes has been emphasized in the management of several malignancies including primary liver malignancies.²⁵ The degree of a center’s experience can influence sufficient preoperative screening, surgical complications, and the rigor of postoperative care.²⁶ Patient-related socioeconomic and disease characteristics have also been implicated in predicting where a patient receives oncologic surgical care ^27–29. Yet, for such a high-risk procedure, little is known regarding patient factors associated with the worse perioperative outcomes following major hepatic resection performed in LVCs. We used a national database to evaluate perioperative outcomes of major hepatectomy for primary liver malignancies, and as suspected, found that 90-day mortality was considerably higher in LVCs. We also noted that patients living farther from HVCs, belonging to more disadvantaged socioeconomic groups, and those with high CDS were more likely to receive care in LVCs. As high burden of comorbidities is a well-recognized factor associated with worse perioperative outcomes, we further examined the relationship between 90-day mortality, center volume, and CDS. We found that perioperative mortality was considerably higher in LVCs among patients with low CDS. Interestingly, 90-day mortality did not differ by treatment center volume among patients with high CDS. We also noted that LVCs operated on a higher proportion of patients with HVCs. These findings underscore the importance of careful patient selection for major hepatic resection given the risk that this procedure poses to patients and the high level of care required postoperatively and the particularly high risk of this operation to those with high comorbidity burdens.

Studies examining determinants of outcomes following pancreatic resections have concluded that a “defined minimum hospital experience” was necessary to minimize perioperative deaths following pancreatic resections and that large volume tertiary centers performed pancreaticoduodenectomies with improved outcomes.^17,18 This was also queried in 2 single HVC studies reporting lower in-hospital mortality following hepatectomy at one HVC than that of all other centers in the state, and the second proposing that center experience and surgeon expertise allowed for optimal surgical and peroperative care among 400 patients treated for HCC.^11,14 Though these single HVC studies provide valuable insight into such centers’ major hepatectomy outcomes, they do not provide generalizable conclusions. A recent multicenter retrospective study demonstrated improved outcomes among patient receiving resection for iCCA in HVCs.²⁵ Nevertheless, none of these studies characterize the patient population treated in LVCs and the possible factors associated with worse outcomes in these centers. To our knowledge, our study is the first to provide national multicenter data on perioperative outcomes and patient characteristics for those who undergo major hepatic resections in LVCs.

As described in previous studies, our analysis revealed that receiving care in LVCs was associated with a significantly higher perioperative mortality following oncologic resections.^20–22 Our study also identified, on a national level, several patient factors associated with undergoing major hepatectomy in LVCs. Patient socioeconomic and geographic factors were found to be associated with postoperative outcomes and with receipt of care in LVCs. Distance from specialized cancer centers has been reported to correlate negatively with receipt of adequate cancer care including chemoradiation for several cancers.^23,24 For patients on liver transplant waitlists or those requiring liver resection, increased mortality was noted among patients who lived farther from a tertiary referral center.^27,30 Consistent with these studies, we noted that patients who received care in LVCs were less likely to travel a far distance to undergo surgery. Our findings indicate that patients unable or unwilling to travel far distances are more likely to receive care in centers associated with a higher perioperative mortality.³⁰ While we are unable to obtain information regarding specific patient decisions, distance to the nearest center, insurance status, means to travel, and access to information on specialized care, we were able to ascertain that the distance needed to travel affects where patients obtain treatment. Examination of patient socioeconomic factors revealed that carrying nonprivate insurance, residing in areas with lower mean income, and lower rates of high school graduates were all associated with receiving care in LVCs. These factors are often interlinked and, when combined, may present a significant barrier to receiving care in specialized centers.

In addition, patients from lower socioeconomic backgrounds commonly present with higher comorbidity burdens.^28,29 In accordance with these findings, our study demonstrated that a higher proportion of patients receiving care in LVCs had a CDS >1. Our findings corroborate those of previous studies concluding that cancer patients with lower socioeconomic status carry a higher burden of comorbidities and an increased mortality.^28,29 Similar to previous studies examining drivers of increased perioperative mortality following hepatic resection, we found that a higher comorbidity profile was associated with increased 90-day mortality.^16,31 Perhaps our most interesting finding was that among patients with high-comorbidity scores, center volume no longer significantly impacted perioperative mortality. This provides clear evidence that such patients are at very high risk even when receiving care in highly specialized HPB centers. Our findings suggest that HVCs may perform more judicious preoperative patient selection resulting in a smaller proportion of high-risk patients undergoing resection in these centers. This is likely to contribute to the disparate outcomes between LVCs and HVCs. Given the risks posed by major hepatic surgery, experienced centers may recommend other interventions or medical management for patients with serious medical conditions, thereby reducing rates of perioperative mortality. Some of these patients may chose to seek care elsewhere and undergo surgical intervention in centers with less stringent restriction or in LVCs that rarely perform such procedures, putting these patients at undue risk. The lower rate of patients with high CDS combined with the improved outcomes among patients with low CDS suggests that optimal outcomes for such a surgical procedure can only be ensured by high levels of experience. The expertise present in HVCs likely results in the ability to minimize complications, optimize postoperative rescue attempts, and to perform very judicious selection of adequate surgical candidates. These attributes are difficult to achieve by centers lacking expertise in this field.

We acknowledge that this study is not without limitations. As a national database that reports data from multiple centers, errors in the reporting and collection of data may occur. In addition, due to the structure of the reported variables, exact information on patient health status and demographic information such as individual levels of education, income, and insurance status cannot be ascertained. The retrospective nature and inability to obtain information on postoperative complications or cause of death limit the scope of our analysis. Reliable information on baseline liver function and cirrhosis could not be ascertained which was limiting given the high risk for posthepatectomy liver failure in cirrhotic patients. Though we could not identify individual centers, HVCs were all academic centers and were predominantly located on the east coast. Limited access to such centers by large segments of the US population may explain some of the observed trends.

A comprehensive and careful preoperative evaluation, particularly of comorbid conditions, may help to identify patients at increased risk of perioperative death following major hepatic resection. While some of these patients may benefit from medical optimization and referral to specialized HVCs, our findings indicate that many of these patients may simply be at very high risk to undergo such an operation. A judicious case by case analysis of highrisk patients is essential prior to performing major hepatic resections. Further studies aimed at understanding what patient characteristics and LVC attributes contribute to increased mortality in this setting are warranted, particularly given that most major hepatic resections are being performed at LVCs.