Association Between Noninvasive Fibrosis Markers and Postoperative Mortality After Hepatectomy for Hepatocellular Carcinoma

Key Points Question Are the noninvasive fibrosis markers aspartate aminotransferase–platelet ratio index and fibrosis 4 associated with perioperative mortality and overall survival after hepatectomy for hepatocellular carcinoma? Findings In this cohort study of 475 US veterans, aspartate aminotransferase–platelet ratio index and fibrosis 4 were independently associated with increased 30- and 90-day mortality and worse overall survival. They were shown to improve the estimation of postoperative mortality. Meaning This study suggests that the incorporation of aspartate aminotransferase–platelet ratio index and fibrosis 4 in the selection criteria of hepatectomy for hepatocellular carcinoma may be warranted.


Introduction
Hepatectomy and liver transplant are the main curative therapies for hepatocellular carcinoma (HCC). 1 Intention-to-treat studies have demonstrated similar overall survival for these therapies. 2,3 The selection criteria for HCC resection are not well established. 4 The National Comprehensive Cancer Network guidelines for HCC state that patients with preserved liver function, resectable disease, who fit United Network for Organ Sharing criteria, could be considered for resection or transplant, and there remains controversy over which initial strategy is preferable. 1 Hepatocellular carcinoma is currently the fastest rising cause of cancer-related deaths in the United States. 5 Although the incidence of HCC tripled between 1975 and 2005, the 5-year survival rate remains approximately 25%. 6 However, this survival rate could be greater than 70% when curative therapy is applied to early-stage HCC. 2,3,7,8 Hepatic fibrosis is associated with an increased risk of postoperative liver failure and mortality after resection for HCC. 9 The noninvasive fibrosis markers (NIFMs) aspartate aminotransferaseplatelet ratio index (APRI) and fibrosis 4 (Fib4) have been shown to accurately predict the presence of cirrhosis and severe fibrosis when validated with liver biopsy findings. 10,11 Studies conducted in Asian patients with hepatitis B and HCC demonstrated that abnormal levels of NIFMs were associated with worse postoperative outcomes after hepatectomy. [12][13][14] However, to our knowledge, the role of these markers in the US population has not been studied. The aim of this study was to examine the association of APRI and Fib4 with perioperative mortality and overall survival after HCC resection among US veterans. We hypothesized these NIFMs independently associated with perioperative mortality and overall survival after hepatectomy for HCC.

Data Source and Study Cohort
The Veterans Administration (VA) Corporate Data Warehouse database was used. The Corporate Data Warehouse is a robust and comprehensive database that abstracts clinical information directly from the computerized patient record system from VA hospitals. Data were accessed, managed, and analyzed within the VA informatics and computing infrastructure. From January 1, 2000, to December 31, 2012, all patients with the diagnosis of HCC were identified using the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) code 155.0 (liver cell carcinoma) and excluding patients with ICD-9-CM code 155.1 (intrahepatic bile duct carcinoma). 15 We included all patients undergoing hepatectomy using the Current Procedural Terminology codes 47120 (partial lobectomy), 47125 (total left lobectomy), 47130 (total right lobectomy), and 47122 (trisegmentectomy). Data abstraction and management started September 30, 2016, and the analysis was completed on December 30, 2017.
The study protocol was reviewed and approved by the institutional review board of the Southern Arizona VA Health Care System, which also waived informed patient consent because this project was limited to retrospective deidentified data analysis. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline. 16

Patient Demographics and Clinical Factors
Demographic, laboratory, and clinical data were ascertained at the time of resection. Race/ethnicity information was defined by the participants and was available on the CDW database. The preoperative laboratory data were obtained from the closest date to surgery. Liver function was determined using Model for End-Stage Liver Disease (MELD) score (score of 6-9 indicates normal liver function; Ն10 indicates liver dysfunction) and Child-Turcotte-Pugh (CTP) class (A indicates preserved liver function; B, mild to moderate liver dysfunction), as previously described. 17,18 The presence of advanced fibrosis and cirrhosis was determined using the NIFMs Fib4 and APRI, respectively. Based on the studies that validated these NIFMs with liver biopsy and meta-analysis, the cutoff value for Fib4 was 4.0 (>4.0 indicates advanced fibrosis) and for APRI, 1.5 (>1.5 indicates cirrhosis). 10,11,19,20 Standard equations were used: The main outcomes were defined as perioperative mortality and long-term survival; 30-and 90-day mortality were examined. Long-term survival was evaluated with overall survival comparison for patients with at least 1-month follow-up.

Statistical Analysis
A 2-tailed, unpaired t test or Mann-Whitney test was used for univariate comparisons of continuous variables, and Pearson χ 2 or Fisher exact test was used for comparison of categorical variables. A logistic regression model was used to evaluate the association between APRI and Fib4 and perioperative mortality. Overall survival was calculated from the date of surgery to date of death or last follow-up. The Kaplan-Meier method was used to plot the survival curves, and comparisons were performed with the log-rank test. Cox proportional hazards regression analysis was performed to evaluate the association between APRI and Fib4 levels with long-term survival. Previously established predictors of perioperative mortality were included in the multivariate models. To evaluate the predictive contribution of APRI and Fib4, their association with the concordance index of the predictive model composed of the established predictors of outcome (CTP and portal hypertension) were examined as previously described. 21 Complete case analysis was used for missing data.
Statistical comparisons were 2-sided, and P values <.05 were considered significant. The statistical analysis was conducted using the SAS Enterprise Guide, version 7.1 (SAS Institute Inc).

APRI and Fib4 Association With Perioperative Mortality and Survival
Logistic regression was performed to examine the association between APRI and Fib4 and the 30and 90-day mortality. Compared with patients with APRI values 1.5 or lower, multivariable analysis revealed that APRI greater than 1.5 was associated with worse 30-day (OR, 6 Table 2).

JAMA Network Open | Surgery
Kaplan-Meier survival analysis showed that patients with APRI levels greater than 1.5 had worse overall survival than patients with APRI levels of 1.5 or lower (mean survival time, 3.6 vs 5.4 years; log-rank test P < .001) ( Figure 1A). The median survival for patients with APRI greater than 1.5 and APRI 1.5 or less were 2.2 and 4.3 years, respectively. When patients were stratified by Fib4 index, the overall survival between patients with Fib4 greater than 4.0 was significantly worse than patients with Fib4 of 4.0 or lower (mean survival, 4.1 vs 5.3 years; log-rank test P = .01) ( Figure 1B). The median survival for patients with Fib4 greater than 4.0 and 4.0 or lower were 2.7 and 4.3 years,  A, Patients stratified by aspartate aminotransferaseplatelet ratio index. High risk was considered an aspartate aminotransferase-platelet ratio index value greater than 1.5 (cirrhosis); low risk, aspartate aminotransferase-platelet ratio index 1.5 or lower (logrank test P < .001). B, Patients stratified by fibrosis 4. High risk was considered a fibrosis 4 value greater than 4.0 (advanced fibrosis); low risk, fibrosis 4 value of 4.0 or lower (log-rank test P = .01).

Change in the Concordance Index and APRI and Fib4
Concordance index is similar to the area under the receiver operating characteristic (ROC) curve and measures the probability that, given 2 random patients, the one with the worse outcome is predicted to have a worse outcome. 22 To evaluate whether APRI and Fib4 could more accurately predict the postoperative mortality and overall survival after hepatectomy for HCC, these NIFMS were added to the model containing the established predictors: CTP class and portal hypertension. Table 3  In addition, APRI improved the overall survival prediction by 0.18 (P < .001). Table 3 also demonstrates the association between Fib4 and the prediction accuracy of perioperative mortality and overall survival. The improvement of the predictive accuracy of 30-and 90-day mortality by adding Fib4 to the model was 0.13 (P = .004) and 0.07 (P = .01), respectively. The long-term survival prediction was improved by 0.17 (P = .003). Figure 2 illustrates the ROC curves and their respective areas under the curve, demonstrating the association between APRI and Fib4 with the predictive accuracy of 30-and 90-day mortality in patients after hepatectomy for HCC.

Discussion
To our knowledge, this is the largest Western series to examine the role of NIFMs in patients undergoing hepatectomy for HCC. This study demonstrates the association of APRI and Fib4 with the perioperative mortality and overall survival after hepatectomy for HCC. Using cutoff values described and validated by previous studies, both APRI and Fib4 were independently associated with increased 30-and 90-day mortality. 10,11,20,23 In addition, APRI was associated with worse overall survival, after adjusting for CTP class and portal hypertension. Child-Turcotte-Pugh classification and portal hypertension are the most validated and established predictors of outcome after liver resection, and are broadly used for selection of patients for liver resection for HCC. 24 The perioperative mortality  21 The decrease in the concordance index associated with the addition of APRI or Fib4 to the regression models was statistically significant for the prediction of 30-and 90-day mortality and overall survival. These findings suggest that both APRI and Fib4 contribute to the prediction of short-and long-term survival outcomes after resection of HCC, beyond what is achieved by CTP and portal hypertension.
This study supports the findings of other authors that APRI is independently associated with poor overall survival after liver resection for HCC. 12   Noninvasive fibrosis markers, such as APRI and Fib4, are attractive markers because they are easily accessible and determined with routine preoperative laboratory tests. Another unique characteristic of APRI and Fib4 is that they include the platelet count, which is useful because a low platelet count has been shown to be associated with increased perioperative morbidity and mortality. 29,30 In addition, APRI and Fib4 are NIFMs that were validated prospectively with liver biopsy. 10 reporting an area under the ROC curve of 0.756 for severe fibrosis. Therefore, APRI and Fib4 are good predictors of cirrhosis and fibrosis, respectively. Liver fibrosis and cirrhosis have been described as markers of impaired hepatic regeneration and increased risk of postoperative liver failure after hepatectomy. 9,31 These findings, associated with the results of the present study, indicate that APRI and Fib4 are potential predictors of postoperative mortality after hepatectomy for HCC. Limitations This study has several limitations, which are inherent to its observational retrospective design. Some unmeasured patient characteristics could have changed the results, in particular, the lack of information on preoperative comorbidities. However, the large number of patients of this cohort and the analysis accounting for the validated predictors of surgical outcomes helped to minimize possible selection bias. The VA CDW database constitutes predominately white, male patients. Therefore, the results of this study may not be generalizable to the overall population. Information on HCC recurrence and specific cause of death were not available. In addition, information on postoperative complications was not ascertained. Therefore, it is unknown whether perioperative mortality was related to liver failure or other causes.

Conclusions
This study suggests that APRI greater than 1.5 and Fib4 greater than 4.0 were independently associated with increased 30-and 90-day mortality after hepatectomy for HCC. In addition, APRI was independently associated with worse overall survival. The contribution of APRI and Fib4 to improve the ability of established markers in predicting perioperative mortality and overall survival was supported by valid methods. These findings suggest that incorporating APRI and Fib4 in the selection process for hepatectomy for HCC as new predictors of mortality may be warranted.