Evaluation of Patients With Hepatocellular Carcinomas That Do Not Produce α-Fetoprotein.

To compare characteristics and outcomes among LT recipients with radiographically apparent HCC lesions with AFP-producing tumors or with tumors that do not produce AFP (hereafter non-AFP–producing tumors), and to identify factors influencing recurrence in LT recipients with non-AFP–producing tumors. is negligible for patients within the Milan criteria. Stratifying patients by AFP status in addition to radiological criteria may improve the selection process for and the prioritization of transplant candidates.

A liver transplant (LT) for hepatocellular carcinoma (HCC) has unequivocally been established as the gold standard therapy for select patients with underlying liver dysfunction, largely owing to the reproducibly excellent results when transplantation is limited to tumors meeting the widely accepted Milan criteria. [1][2][3] The prioritization of LT recipients with HCC by use of the Model for End-Stage Liver Disease (MELD) allocation system [4][5][6] has resulted in LTs for HCCs now accounting for nearly 25% of all transplants in the United States. 7 Despite continuous refinements in the HCC MELD exception policy, LT recipients with HCC continue to be overprioritized compared with patients with liver failure who do not have cancer, with many studies revealing higher transplant rates despite lower risks of waitlist dropout and inferior survival. [8][9][10] The recently implemented regional Share MELD 35 policy, 11 intended to increase access to regional allografts for the sickest LT recipients, has had the occasional unintentional consequence of diverting lifesaving organs from patients with liver failure to LT recipients with HCC whose MELD scores have been allowed to mature to 35 or greater. This has led to the latest modification of the HCC allocation policy by the Organ Procurement and Transplantation Network, with a requirement of a mandatory 6-month waiting period prior to awarding of exception points and a capping of the MELD score at 34 for LT recipients with HCC with preserved underlying liver function. 12 The contemporary practice of awarding equal prioritization for all LT recipients with HCC who qualify for MELD exception points, without taking into account their variable risks of waitlist dropout or posttransplant recurrence, aggravates the current problem. Multiple studies have demonstrated that radiologic size criteria alone do not reliably capture the biological behavior of tumors. 13,14 Important serum biomarkers, explant pathologic factors, and response to pretransplant locoregional therapy 15-20 have all been shown to significantly improve the prognostic ability of radiologic criteria alone, but none have formally been incorporated in the MELD prioritization scheme.
Serum α-fetoprotein (AFP) is a well-established prognostic marker of increased tumor virulence in HCC 15,21-27 and has been shown to be associated with worse tumor phenotype and aggressiveness. While multiple studies have shown inferior outcomes following an LT for HCC at various prognostic cutoffs, data regarding the characteristics and outcomes of patients with tumors that do not produce AFP (hereafter referred to as non-AFP-producing tumors) have been scarce. The specific aims of our study were to evaluate the proportion of LT recipients whose tumors do not produce AFP (≤10 ng/mL [to convert to micrograms per liter, multiply by 1.0]), compare their outcomes with those of patients with AFPproducing tumors, and identify factors that specifically affect posttransplant recurrence in these patients in order to allow for better stratification of patients for LT.

Methods
We retrospectively reviewed a prospectively maintained transplant database and identified all adult patients (18 years of age or older) who underwent an LT for HCC and who received a pretransplant diagnosis of HCC with available data on serum AFP level at UCLA from 1989 to 2013. The UCLA institutional review board approved the study. Informed consent was waived because the data were deidentified.
Recipients were classified as having an AFP-producing tumor (serum AFP level of >10 ng/mL at any time prior to LT irrespective of tumor treatment) or a non-AFP-producing tumor (serum AFP level of ≤10 ng/mL at all times prior to LT irrespective of tumor treatment) based on pretransplant AFP levels. The primary objective was to evaluate the proportion of radiographically apparent HCC lesions that were non-AFPproducing tumors and the effect of AFP status on posttransplant outcomes. Only recipients with explant pathology confirming viable HCC or patients with a pretransplant radiographic HCC diagnosis and directed locoregional tumor treatment with explant pathology confirming complete tumor necrosis were included. Variables for analysis included recipient demographics (age, sex, primary end-stage liver disease diagnosis, body mass index, diabetes, and hypertension), laboratory results (physiologic MELD score, total cholesterol level, AFP level, and neutrophil to lymphocyte ratio), radiologic characteristics (number of lesions, maximum tumor diameter, cumulative tumor diameter, and proportion within radiologic Milan criteria), treatment-specific factors (number and type of locoregional therapy), and pathologic characteristics (number of lesions, maximum tumor diameter, cumulative tumor diameter, extent of pathologic necrosis, tumor grade/differentiation, 28 presence of microvascular and macrovascular invasion, and American Joint Committee on Cancer T stage 29 ). The selection of locoregional therapy modality is based on the size, number, and location of lesions, without consideration for serum AFP.
Posttransplant immunosuppression included our standard 3-drug regimen with methylprednisolone sodium succinate, mycophenolate mofetil, and a calcineurin inhibitor. 30 Occasionally, LT recipients were initiated on a mammalian target of rapamycin inhibitor with withdrawal of calcineurin inhibitors if the explant pathology had poor prognostic features. Tumor surveillance included intravenous contrastenhanced axial abdominal imaging (computed tomography or magnetic resonance imaging), noncontrast computed tomography of the chest, and a serum AFP sample obtained every 3 months prior to the LT and every 6 months following the LT.

Key Points
Question What are the radiologic and pathologic features and posttransplant outcomes of hepatocellular carcinomas (HCCs) that do not produce α-fetoprotein (AFP)?
Findings In this single-center retrospective analysis of 665 liver transplant recipients, the patients with tumors that did not produce AFP had radiographic characteristics similar to those of patients with AFP-producing tumors but significantly more favorable pathologic features and posttransplant outcomes.
Meaning Stratifying patients with HCC by AFP status in addition to radiologic criteria may improve transplant candidate selection and prioritization.
Copyright 2016 American Medical Association. All rights reserved.
Comparisons were made among LT recipients with AFPproducing tumors and LT recipients with non-AFP-producing tumors, with subanalyses stratified by both AFP status and radiologic size criteria. Continuous variables were compared using the Wilcoxon rank sum test and reported as median values with interquartile ranges (IQRs). Categorical and ordinal variables were compared using the χ 2 test/Fisher exact test or the Cochran-Armitage test for trend, respectively, and summarized as percentages. Patient survival curves were computed using Kaplan-Meier methods and were compared using log-rank tests. Cumulative incidence curves for recurrence were computed in each specified group while taking into account the competing risk of non-HCC-related mortality, and were compared using the Fine-Gray test. Cox regression analysis was used to identify factors associated with recurrence for both patients with AFPproducing tumors and patients with non-AFP-producing tumors, controlling for the competing risk of non-HCC-related mortality. An interaction P value was reported to indicate whether the hazard ratios (HRs) for any covariates differentially affected recurrence between patients with AFPproducing tumors and patients with non-AFP-producing tumors. P < .05 was considered statistically significant.

Results
Of 665 patients with radiographically apparent HCC undergoing an LT during the study period, 457 (68.7%) had AFPproducing tumors, and 208 (31.3%) had non-AFP-producing tumors. The overall median follow-up time was 27.3 months (IQR, 8.8-67.8 months), with a median follow-up of 43.4 months (IQR, 12.1-78.5 months) for recipients without recurrence or mortality at last follow-up.

Comparison of Patient and Tumor Characteristics by AFP Status
The characteristics of the LT recipients, the donors, and the procedure are shown in the eTable in the Supplement. Patients with non-AFP-producing tumors were more likely than patients with AFP-producing tumors to have diabetes (35.0% There were no significant differences in age, sex, MELD score, total cholesterol level, the presence of hypertension or obesity, or any donor and procedure characteristics between the 2 groups of patients.
Radiologic, treatment, and pathologic characteristics are shown in Table 1. Patients with non-AFP-producing tumors had radiologic characteristics that were similar to those of patients with AFP-producing tumor, but they had significantly more favorable pathologic features, including fewer numbers of tumors (

HCC Recurrence and Posttransplant Survival
Of the 665 patients who underwent an LT, 103 (15.5%) experienced HCC recurrence, with a median time to recurrence of 15.9 months. Patients with non-AFP-producing tumors had significantly superior recurrence-free survival at 1, 3, and 5 years (88%, 74%, and 67% vs 76%, 59%, and 51%, respectively) (P = .002; Figure 1A), a lower cumulative incidence of recurrence at 5 years (8.8% vs 22%) (P < .001; Figure 1B), and a longer median time to recurrence for those who had developed recurrence (30.9 vs 13.4 months; P = .01) than did patients with AFP-producing tumors. When stratified by radiologic Milan criteria at presentation, in addition to AFP status, 1-, 3-, and 5-year recurrence-free survival was better for patients with non-AFPproducing tumors within the Milan criteria (89%, 77%, and 71%, respectively) and was worse for patients with AFPproducing tumors outside the Milan criteria (69%, 42%, and 40%, respectively) (P < .001; Figure 2A). Patients with AFPproducing tumors within the Milan criteria (78%, 64%, and 55%, respectively) and patients with non-AFP-producing tumors outside the Milan criteria (83%, 59%, and 47%, respectively) had intermediate but acceptable survival at 1, 3, and 5 years ( Figure 2A). The 5-year cumulative incidence of recurrence, while controlling for the competing risk of non-HCC mortality, was highest for patients with AFP-producing tumors outside the Milan criteria (42%) and lowest for patients with non-AFP-producing tumors within the Milan criteria (6%) (overall P < .001), with AFP status being able to further discriminate posttransplant recurrence for both groups of patients within the Milan criteria (6% of patients with non-AFPproducing tumors vs 15% of patients with AFP-producing tumors; P = .004) and patients outside the Milan criteria (19% of patients with non-AFP-producing tumors vs 42% of patients with AFP-producing tumors; P = .02) ( Figure 2B).
Analysis of LT recipients initially outside the Milan criteria revealed a trend toward improved 5-year survival (53% vs 39%; P = .35; Figure 3A) and significantly reduced recurrence (0% vs 43%; P = .002; Figure 3B) for patients with non-AFPproducing tumors that were downstaged compared with patients with non-AFP-producing tumors that were unable to be downstaged by pretransplant locoregional therapy. Similarly, for patients with AFP-producing tumors initially outside the Milan criteria, there was significantly superior 5-year recurrence-free survival (58% vs 23%; P = .01; Figure 3A) and lower Bivariate predictors of HCC recurrence for both patients with AFP-producing tumors and patients with non-AFPproducing tumors are shown in Table 2. For patients with non-AFP-producing tumors, the important radiologic factors in- While the HRs for any given predictor are slightly different between the patients with AFP-producing tumors and the patients with non-AFP-producing tumors, none of these covariates had a statistically significant differential effect on posttransplant recurrence between the 2 groups of patients, as evidenced by all the reported interaction P values greater than .05. Of the 208 patients with non-AFP-producing tumors, 16 (7.7%) experienced a posttransplant recurrence; of these 16 patients, 9 (56.3%) had AFP levels of greater than 10 ng/mL, with a median postrecurrence AFP level of 15.8 ng/mL (IQR, 4.0-123.0 ng/mL). Conversely, although the majority (n = 55) of patients with AFP-producing tumors who experienced a posttransplant recurrence demonstrated elevated serum AFP levels, 14 patients (25.5%) experienced a proven recurrence with an AFP level of less than 10 ng/mL. Median survival following recurrence was significantly superior for patients with non-AFP-producing recurrent tumors compared with patients with AFP-producing recurrent tumors (70.6 vs 12.4 months; P = .002).

Discussion
Although an LT has been established as the ideal treatment for select patients with HCC, the scarcity of donor organs and the continued overprioritization of LT recipients with HCC by use of the MELD allocation system 9,10 obligate a change in the current stratification scheme to include markers of tumor biology beyond radiologic size criteria that better discriminate a patient's need for and predicted outcome following an LT. Serum AFP, an important HCC tumor biomarker, has indisputably been shown to be associated with increased risks of waitlist dropout 23,31,32 and posttransplant recurrence. 15,16,21,22,24-27,33 However, data on patients with HCCs that do not produce AFP are scarce. We report the largest single-center study evaluating LT recipients with non-AFP-producing tumors, who comprised nearly one-third of the 665 LT recipients with HCC. We evaluated the effect of AFP status on important pathological characteristics that drive tumor biology and on posttransplant cancer outcomes, and we identified important factors predicting HCC recurrence in LT recipients with non-AFP-producing tumors.
Numerous prior studies 1,15,26 have clearly established pathological tumor differentiation and microvascular invasion as 2 of the most important variables influencing posttransplant cancer outcomes; however, these variables cannot reliably be ascertained prior to an LT. 34,35 In the present study, despite no significant differences in any radiologic characteristics between non-AFP-producing tumors and AFP-producing tumors, we found that non-AFP-producing tumors were significantly less likely to be poorly differentiated or to show the presence of microvascular or macrovascular invasion. Our findings are consistent with the few prior studies demonstrating this association of AFP status with pathologic differentiation and vascular invasion. In a multicenter French study reported by Duvoux et al, 22   In the present study, the patients with non-AFPproducing tumors had significantly improved disease-free survival and lower recurrence rates compared with the patients with AFP-producing tumors. Even for patients who did experience posttransplant recurrence, the median time to recurrence was significantly longer for the patients with non-AFPproducing tumors than for patients with AFP-producing tumors  Patients with non-AFP-producing tumors within Milan criteria Patients with AFP-producing tumors within Milan criteria Patients with non-AFP-producing tumors outside Milan criteria Patients with AFP-producing tumors outside Milan criteria P =.001 P =.24 P =.02 P =.004 The curves compare patients with AFP-producing tumors with patients with tumors that do not produce AFP (non-AFP-producing tumors); AFP status was used to further discriminate survival among liver transplant (LT) recipients within or outside the Milan criteria (A), and HCC recurrence in patients within or outside the Milan criteria (B).
(30.9 vs 13.4 months; P = .01), which supports the idea that non-AFP-producing tumors are less virulent. Interestingly, the AFP status of the recurrence itself also affected postrecurrence survival, with superior survival for patients with non-AFPproducing recurrent tumors compared with patients with AFPproducing recurrent tumors, irrespective of the AFP status prior to the LT.
Perhaps most importantly, AFP status was able to further discriminate outcomes among patients who had tumors within or outside the radiologic Milan criteria, indicating that size criteria alone fail to capture the tumor heterogeneity in these patients. For example, patients with non-AFP-producing tumors within the Milan criteria had an excellent 5-year recurrence-free survival of 71%, with only a 6% adjusted risk of recurrence, compared with a 55% recurrence-free survival and a 15% recurrence rate at 5 years for patients with AFPproducing tumors within the Milan criteria ( Figure 2). This stratification by AFP status appeared even more important in the subset of patients with tumors outside the Milan criteria, allowing for identification of a group of patients with non-AFP-producing tumors outside of the Milan criteria who demonstrated an acceptable 5-year recurrence rate of 19% compared with an unacceptably high 5-year recurrence rate of 42% for patients with AFP-producing tumors outside the Milan criteria ( Figure 2B). Similar to prior reports, 15,37-39 the ability to downstage tumors to meet the Milan criteria was an important requirement to achieve acceptable cancer outcomes in the subset of  The curves compare patients with AFP-producing tumors with patients with tumors that do not produce AFP (non-AFP-producing tumors), for a subset of liver transplant (LT) recipients initially outside the Milan criteria. The ability to successfully downstage recipients to the Milan criteria was important for recurrence-free survival outside the Milan criteria (A) and the incidence of HCC recurrence (B).

Hepatocellular Carcinomas That Do Not Produce α-Fetoprotein
Original One of the main objectives of the present study was to identify important predictors of HCC recurrence in patients with non-AFP-producing tumors and to compare these predictors with those in patients with AFP-producing tumors. As expected, some of the most important predictors for HCC recurrence in patients with non-AFP-producing tumors included radiologic size, number of tumors, adherence to size criteria, pathologic size, and presence of vascular invasion. These same factors were equally important in predicting HCC recurrence in patients with AFP-producing tumors, with no evidence that any covariate was differentially significant between the 2 groups (P > .05 for interaction).
As important as it is to have identified factors predicting posttransplant recurrence, differential prioritization of LT recipients with HCC also requires the determination of the factors that affect waitlist dropout. While we have not yet examined the effect of AFP status on waitlist dropout at our center, fortuitously, several studies have shown that the factors affecting waitlist dropout due to tumor progression similarly influence posttransplant outcomes. 40,41 In a recent study, 23 more than 40% of patients with an initial AFP level of higher than 1000 ng/mL experienced waitlist dropout due to tumor progression; not surprisingly, patients with an AFP level of higher than 1000 ng/mL undergoing an LT had a 5-year disease-free survival of 52.7%, significantly inferior to the 80.3% survival observed among patients with an AFP level of lower than 1000 ng/mL. The determination of the "optimal" AFP cutoff remains to be determined. However, along with previously reported experiences, our findings support the concept that prioritization of LT recipients by AFP status in addition to radiologic size criteria will align the goals of minimizing waitlist dropout and optimizing posttransplant cancer outcomes.
In addition to the inherent flaws in a retrospective analysis, several limitations to the present study warrant discussion. At an AFP threshold of 10 ng/mL, it is likely that some patients were categorized as having AFP-producing tumors, even though an underlying hepatitis, and not their tumor, per se, was the source of an elevated AFP level of higher than 10 ng/mL. Furthermore, the longer median waitlist times in our region 42 may have resulted in some patients with initially non-AFP-producing tumors beginning to produce AFP and, hence, to be categorized as AFP-producing tumors, which may limit the generalizability of our findings to other centers. Despite this, it is important to emphasize that we have not proposed an AFP threshold of 10 ng/mL as a prognostic cutoff, but rather we have identified important radiologic and pathologic characteristics of tumors that do not produce AFP even over a prolonged period of time. Ultimately, a change in the HCC allocation policy will require a prospective assessment of dynamic changes in both the serum AFP and radiologic parameters to pretransplant locoregional therapy to minimize the risk of waitlist dropout while preserving posttransplant oncologic outcomes.

Conclusions
In conclusion, despite significant improvements in posttransplant outcomes for LT recipients with HCC with the institution of the Milan criteria, there is now overwhelming evidence that radiologic size criteria alone do not optimally stratify patients with HCC for LT. Serum AFP is an important biomarker of tumor behavior that reliably indicates a more aggressive tumor phenotype, and we show that a higher AFP expression is associated with a significantly greater frequency of poorly differentiated tumors and microvascular and macrovascular invasion. The inclusion of AFP status to the existing radiologic Milan criteria allows for improved discrimination of tumor biology for LT recipients within or outside the Milan criteria. We propose that the inclusion of serum AFP status to the current radiologic size criteria will improve transplant candidate selection and prioritization.