Evaluation of Absolute Serum α-Fetoprotein Levels in Liver Transplant for Hepatocellular Cancer

Hypothesis  An elevated serum α-fetoprotein (AFP) level before orthotopic liver transplant (OLT) is predictive of mortality after OLT for hepatocellular carcinoma (HCC).

Design  Retrospective analysis of a population-based cohort.

Setting  United Network for Organ Sharing registry (2003-2008).

Patients  We identified 2253 patients who underwent OLT for HCC with available pre-OLT serum AFP values.

Methods  Patients were stratified by AFP levels into low (<20 ng/mL), medium (20-399 ng/mL), or high (≥400 ng/mL) groups. Clinical and pathological characteristics were compared among groups. Survival curves were constructed by the Kaplan-Meier method, and univariate and multivariate Cox-regression analysis was performed.

Results  Of the 2253 patients, 1210 (53.7%), 805 (35.7%), and 238 (10.6%) were in the low, medium, and high AFP groups, respectively. On univariate analysis, the low AFP group demonstrated the best 4-year survival (76%) compared with the medium (65%; P = .001) and high (57%; P < .001) AFP groups. When AFP levels in patients with only stage II HCC underwent assessment, improved survival in the low AFP group was still observed (P < .001). On multivariate analysis, the medium and high AFP groups were associated with higher mortality (hazard ratios, 1.50 [95% confidence interval, 1.19-1.89; P = .001] and 2.11 [1.55-2.88; P < .001], respectively).

Conclusions  Serum AFP level is an independent prognostic predictor of outcome after OLT for HCC. The association between serum AFP value and post-OLT survival warrants further investigation to potentially better allocate donor allografts for HCC.

Hepatocellular carcinoma (HCC) remains a major worldwide health concern resulting in more than 1 million cancer-related deaths annually.^1-3 These outcomes are directly attributable to a generally poor prognosis for patients with HCC, with few achieving 5-year survival.³ Although the incidence of HCC in North America has been relatively low, a steady increase has occurred since the 1980s, while incidence of most other cancers has declined.^4,5 Because of the relative resistance of HCC to systemic chemotherapy, innovative therapies such as transarterial chemoembolization and yttrium 90–labeled radiotherapy have been developed to improve local disease control. Nevertheless, surgical options remain the criterion standard for cure and offer the only prospects for long-term survival.^4,6-8

Recent reports verify the efficacy of surgical therapies in providing long-term survival for patients with HCC.^9,10 However, multicentric hepatic disease, extrahepatic disease, or advanced cirrhosis excludes up to 80% of patients for curative surgical options.¹¹ Because HCC often occurs in the setting of cirrhosis, total hepatectomy with orthotopic liver transplant (OLT) affords the only potential cure for both diseases. However, the critical shortage of available donor organ allografts has limited the widespread use of OLT for the treatment of HCC. Moreover, HCC recurrence after OLT has been problematic and remains a major cause of post-OLT mortality.^5,12-16 Despite the recent trend to expand traditional criteria through downstaging strategies for organ allocation to patients with HCC,^17-19 patient selection is of critical importance.

Further refinement of transplant criteria through evaluation of prognostic factors appears critical to ensure appropriate allocation of scarce organ transplant resources. During the past 50 years, α-fetoprotein (AFP) has been used as a serum tumor marker for HCC, but its role as a potential prognostic factor in OLT for HCC has not been established. A 70-kDa glycoprotein produced by the fetal liver and yolk sac, AFP was first described by Bergstrand and Czar²⁰ in 1956. Prominent interest rose in the 1960s when observational studies correlated increasing levels of serum AFP with germ cell tumors and HCC.²¹ For HCC, the correlation between AFP level and cancer detection has been imperfect, and nearly one-third of patients with HCC have serum AFP levels within the reference range.^22,23 In contrast, significant elevations of AFP may occur with hepatic disease even in the absence of a malignant neoplasm.^23,24 Nonetheless, when radiographic imaging demonstrates a hypervascular hepatic tumor accompanied by cirrhosis and serum AFP levels of 400 ng/mL or more (to convert to micrograms per liter, multiply by 1), these clinical findings are diagnostic for HCC.²⁵

The role or significance of absolute serum AFP levels has not been definitively established for patients undergoing OLT for HCC. Because recurrent HCC has been a major source of mortality among recipients undergoing transplant for HCC,^5,12-16 we hypothesized that an elevated pre-OLT AFP value may be a surrogate for advanced or metastatic disease not detected through routine radiographic imaging modalities. Our objective was to evaluate the pre-OLT serum AFP value as a predictor of patient survival after transplant. We used the United Network of Organ Sharing (UNOS) database, which contains clinical information for all patients in the United States who have undergone OLT for HCC, to assess this potential association.

UNOS coordinates the US organ transplant system by operating the Organ Procurement and Transplantation Network; UNOS collects and maintains data, which have been organized into accessible Standard Transplant Analysis and Research (STAR) files for all organ transplants in the United States since 1986.²⁶ The UNOS transplant information database contains clinical and pathological data on each transplant candidate and recipient.²⁷ After UNOS approval was obtained for this specific investigation, a STAR file containing all clinical data for liver transplants performed in the United States from September 30, 1987, through February 6, 2008, was obtained. The data file included all deceased and living related liver transplant recipients. The STAR files recorded follow-up data at the 6-month and annual posttransplant intervals.

Because 2003 was the first year that UNOS routinely recorded AFP values, a query of the UNOS database identified 2253 of 3076 patients who underwent OLT for HCC from May 9, 2003, through February 6, 2008, and who had verified pre-OLT serum AFP values. In addition to AFP values, the following clinical data were collected for these patients: age, sex, number of HCC lesions, stage, native and transplant Model for End-Stage Liver Disease (MELD) scores,²⁸ liver-directed/regional therapy (LDT), pathological confirmation of HCC, comorbidities including diabetes mellitus and obesity (calculated as body mass index [BMI; the weight in kilograms divided by height in meters squared] of ≥30), and hepatitis B virus (HBV) and hepatitis C virus (HCV) status. The LDTs included transarterial chemoembolization, radiofrequency ablation, and cryoablation. Stage was defined according to the modified TNM staging classification of the American Liver Tumor Study Group.²⁹

The primary outcome measure of interest was overall survival, which was calculated from the date of transplant to the date of death or of last follow-up (March 6, 2009). The main factor of interest was pre-OLT serum AFP value. The entire patient cohort was stratified into 3 groups according to AFP value: low (<20 ng/mL), medium (20-399 ng/mL), and high (≥400 ng/mL). These cutoff points were selected with regard to the reference range of AFP values (ie, 0-20 ng/mL) and the diagnostic threshold for HCC in the absence of a pathological diagnosis (ie, AFP level ≥400 ng/mL).²⁵

Initial MELD scores were reported by UNOS without exception points and were compared among AFP groups as categorical values. The MELD scores at transplant were dichotomized to account for patients receiving UNOS exception points (ie, >20). We compared demographic and clinical factors among AFP groups by χ² test for categorical variables and 1-way analysis of variance for continuous variables. Survival was calculated using the Kaplan-Meier method and compared using the log-rank test. Univariate analysis was performed to identify factors associated with survival, and the Cox proportional hazards regression method was used to examine the association of respective AFP groups with survival and to calculate hazard ratios and 95% confidence intervals. Variables adjusted in the Cox model included age, sex, stage, initial and transplant MELD scores, LDT, HBV and HCV status, diabetes mellitus, BMI, and AFP group. P < .05 was considered statistically significant, and all analyses were performed using SAS statistical software (version 9.1; SAS Institute, Inc, Cary, North Carolina).

The characteristics of the UNOS population of patients with HCC are presented in Table 1. Of the 2253 patients who underwent OLT for HCC during the study period, 461 died by March 6, 2009. Graft failure owing to HCC recurrence or a malignant neoplasm was the cause of mortality in 30.4% of the deceased patients. The mean age at the time of transplant was 56 years, and 1784 of the patients (79.2%) were male. Most patients (2198 [97.6%]) received a whole (ie, cadaveric) liver, and most patients (1512 [67.1%]) had 1 HCC lesion. Nearly all patients had stage I (146 [6.5%]) or stage II (2054 [91.2%]) disease. Of the entire patient cohort, HBV or HCV was detected in 32.8% and 55.3%, respectively. Before transplant, 705 patients (31.3%) of the cohort were classified as obese (ie, BMI ≥30) and 556 (24.7%) were reported to have diabetes mellitus. More than half the entire cohort underwent LDT (1305 patients [58.4%]). At the time of transplant, 97.5% of patients had MELD scores of greater than 20, indicating receipt of UNOS exception points.

The median AFP value was 17 (range, 1-71 941) ng/mL for the entire cohort. After stratifying the entire cohort by AFP value into low (1210 patients [53.7%]), medium (805 [35.7%]), and high (238 [10.6%]) groups, we compared demographic and clinicopathological factors among the AFP groups (Table 2). When we compared the low, medium, and high AFP groups, the high AFP group had a greater percentage of female patients (18.8%, 20.7%, and 31.5%, respectively; P = .001). There was a greater percentage of deceased patients with increased AFP values (16.0%, 24.5%, and 29.8%, respectively; P < .001), with a corresponding higher rate of HCC recurrence/malignant neoplasm as the cause of mortality (29.2%, 36.7%, and 43.8%, respectively; P = .03). A higher rate of HCV was also observed in patients with medium and high AFP values (56.8%, 71.3%, and 67.3%, respectively; P < .001). A lower percentage of patients with diabetes mellitus and obesity (BMI ≥30) was observed with higher AFP values (diabetes, 28.0%, 22.2%, and 19.1%, respectively; P = .001, and BMI ≥30, 33.9%, 29.7%, and 25.3%, respectively; P = .01). Finally, there was no significant difference between AFP groups with regard to age, stage, LDT, or HBV status (all P values, >.05).

The overall median survival for the entire cohort was 66.6 months (5.6 years) with 1- and 4-year survival rates of 90% and 70%, respectively. When survival was assessed according to AFP group, the medium and high AFP groups had significantly worse overall survival compared with the low AFP group (P < .001) (Figure 1). Survival curves for patients with only stage II disease stratified by AFP level demonstrated similar outcomes, showing that patients with low AFP values had significantly longer survival than patients with middle or high AFP values (4-year survival, 76%, 65%, and 57%, respectively; P < .001) (Figure 2). Survival curves for each AFP group from the entire cohort were also compared with the survival of patients who underwent OLT owing to nonmalignant disease during the same study period (Figure 3). The survival curve for nonmalignant disease was similar to the curve for patients with low AFP values (1 year, 87% vs 92%; 2 year, 83% vs 87%; 3 year, 78% vs 82%; and 4 year, 76% vs 76%; respectively).

Univariate and multivariate analyses for the overall cohort are presented in Table 3. After univariate analysis, older age at transplant, positive HCV status, and higher AFP values (ie, medium and high AFP groups) were associated with higher mortality. After adjustment for clinical factors, medium and high AFP groups had higher mortality compared with the low AFP group (hazard ratios, 1.50 [95% confidence interval, 1.19-1.89; P = .001] and 2.11 [1.55-2.88; P < .001], respectively).

A 1983 National Institutes of Health consensus development conference endorsed OLT as a viable therapeutic option.³⁰ Since then, access to OLT has rapidly expanded, along with the indications for its application. A learning curve for OLT was apparent, with early results for HCC showing relatively poor 5-year survival rates ranging from 15% to 36%.^31-33 With refined selection criteria, surgical techniques, and postoperative care, improved outcomes have been achieved. In a landmark article establishing HCC lesion number and size criteria for OLT (ie, the Milan criteria), Mazzaferro et al⁸ reported impressive short- and long-term survival for OLT in patients with HCC. However, recent UNOS data show a disparity in long-term outcomes between patients undergoing transplant for HCC and nonmalignant disease.^13,26,34

Despite an ever-increasing demand for a critically short supply of donor organs, the HCC criteria for OLT have been expanded. Yao et al¹⁹ reported that patients with HCC who have undergone OLT outside standard criteria have outcomes similar to those of patients who meet the Milan criteria. However, a recent multi-institutional study³⁵ demonstrated that long-term outcomes outside the Milan criteria resulted in suboptimal outcomes as a whole compared with patients who underwent transplant for HCC within the Milan criteria. This expansion of criteria has also occurred despite a 13% increase in 2007 in the number of patients who died while waiting for an allograft and, despite tumor recurrence remaining a serious problem, with recurrence rates of up to 57%.^12-16,36 Considering the successful treatment of end-stage liver disease with OLT in patients who have no concerns for tumor recurrence, the criteria for patients to receive a donor allograft for HCC should be optimized. In deference to the true goals of appropriately allocating scarce resources and improving outcomes, relevant clinical details regarding patients with HCC must be continuously reassessed.

Although AFP values have diagnostic utility for HCC, a role for AFP values in determining eligibility or prioritization for transplant among patients with HCC has not been established. Herein, we demonstrate that survival after OLT is directly associated with the pre-OLT serum AFP value, and that higher AFP values correlate with increased mortality secondary to HCC recurrence/malignant neoplasm (Table 2). Patients with very high AFP values (ie, ≥1000 ng/mL [n = 125] and ≥5000 ng/mL [n = 37]) had median survivals of 4.1 and 3.8 years, respectively (4-year survival, 55% and 42%, respectively) (data not shown). We also stratified AFP values according to guidelines set by the American Association for the Study of Liver Diseases.³⁷ In accord with these guidelines, when an AFP value of greater than 200 ng/mL was the cutoff for the high AFP group, the survival advantage for patients with low AFP values persisted compared with patients with middle or high AFP values (data not shown).

We used the UNOS registry to determine the association of AFP values with OLT for HCC. As a comprehensive registry of all liver transplants performed in the United States, this registry has a sufficient patient cohort to provide the statistical power to determine any potential association between AFP and OLT. However, our results should be interpreted in light of the following limitations. Although the UNOS registry lists all OLTs performed for any reason in the United States, we excluded approximately 800 patients because of the absence of preoperative AFP values for patients with HCC. In addition, AFP test dates were not available from the UNOS database. However, the period between AFP testing and OLT in our cohort was likely relatively short given that the median time from listing to transplant was 3.0 months; and 28.5% of patients underwent transplant in less than 1 month. Finally, we were unable to control for specific pathological features (ie, tumor differentiation and microvascular invasion) that have been associated with post-OLT tumor recurrence.¹²

Other groups have examined the potential role of AFP value for OLT in small cohort studies.^38-41 In these single-institution studies, patients with low AFP values compared with those with high values had better outcomes after OLT. Furthermore, they reported that absolute AFP values or increases thereof were associated with higher HCC recurrence. In addition, a separate examination of UNOS data proposed a model, including a MELD score of at least 20 and an AFP level of at least 455 ng/mL, that may predict poor posttransplant survival.⁴² Overall, these studies suggest that high AFP values, which may reflect proportionately more advanced disease than that observed in patients with lower AFP values, correlate with poor OLT outcomes.

Despite its shortcomings, OLT provides the prospect for cure in select patients with HCC. Currently, patients undergo transplant evaluation involving analysis of hepatic function, determination of the degree or extent of disease, and assessment of medical comorbidities. Although these criteria have resulted in successful and safe OLT for thousands of patients, there is a subset of HCC patients who have suboptimal outcomes after OLT. Herein we demonstrate that patients with AFP values above the reference range have a survival disadvantage. Because UNOS reevaluates the criteria for OLT in the United States, evaluation of AFP values should be considered for inclusion in transplant criteria for HCC.

Correspondence: Joseph Kim, MD, Comprehensive Cancer Center, City of Hope, 1500 E Duarte Rd, Duarte, CA 91010 (jokim@coh.org).

Accepted for Publication: October 5, 2009.

Author Contributions: Dr Kim had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Mailey, Artinyan, Nissen, Colquhoun, and Kim. Acquisition of data: Khalili, Denitz, Sanchez-Luege, Sun, Bhatia, Nissen, Colquhoun, and Kim. Analysis and interpretation of data: Khalili, Denitz, Sun, Bhatia, Nissen, Colquhoun, and Kim. Drafting of the manuscript: Mailey, Artinyan, Khalili, Denitz, Sanchez-Luege, and Kim. Critical revision of the manuscript for important intellectual content: Mailey, Sun, Bhatia, Nissen, Colquhoun, and Kim. Statistical analysis: Sun and Bhatia. Administrative, technical, and material support: Mailey, Artinyan, Khalili, Denitz, Sun, Bhatia, Nissen, Colquhoun, and Kim.

Financial Disclosure: None reported.

Previous Presentation: This paper was presented at the 20th Annual Scientific Meeting of the Southern California Chapter of the American College of Surgeons; January 16, 2009; Santa Barbara, California.