Complete necrosis of HCC improves recurrence-free and overall survival after transplantation independent of baseline risk.
Oct 23, 2026DOI: 10.1097/hc9.0000000000001074Journal ArticleHepatology Communications
Written by: Zachary T BermanBeau B ToskichChristopher D MaloneKristopher P CroomeRobert J LewandowskiRiad SalemMaria B Majella DoyleKyril L ColeNeil Mehta
INTRODUCTION
Liver transplantation offers the best chance of cure for patients with HCC, but most patients will receive locoregional therapy to increase access to transplant via either downstaging or bridging therapy. Historically, transarterial chemoembolization (TACE) was the most widely used locoregional therapy modality, which has recently been surpassed by Y90 transarterial radioembolization (TARE).1Complete pathologic necrosis (CPN), defined as the absence of any viable tumor on explant histology, has been previously identified as a marker of favorable tumor biology and treatment response. DiNorcia et al2 analyzed 3439 patients from the US Multicenter HCC Transplant Consortium and found that the 23% of patients who achieved a complete pathologic response had significantly lower 5-year recurrence (5.2% vs. 16.4%) and higher overall survival (75% vs. 68%) when compared with those who did not, a finding consistent with an earlier 501-patient single-center series reported by Agopian et al.3 These associations have since shaped how many centers approach pretransplant tumor therapy. Current AASLD practice guidance recommends locoregional therapy for candidates expected to wait more than 6 months for transplant, to reduce the risk of tumor progression beyond transplant criteria or to downstage tumors initially beyond criteria.4 However, whether CPN-transplant outcome associations reflect a true effect or selection bias, such as those with less advanced liver disease or favorable disease biology who could tolerate more aggressive locoregional therapy, has been questioned.5In addition, whether bridging therapy confers benefit when CPN is not achieved remains debated. Prior direct comparisons have found more recurrence among patients treated without a complete pathologic response than among untreated patients, in both within-Milan and beyond-Milan cohorts, raising the possibility of an aggravating role of locoregional therapy in tumors with unfavorable biology.6,7The previously mentioned multicenter studies establishing the prognostic importance of CPN were conducted before TARE became a common bridging modality. Reported CPN rates differ substantially among modalities, with TARE and percutaneous ablation generally outperforming TACE.8 CPN rates after external beam radiation therapy (EBRT) have been variable across series.9–11 Given these nonuniform findings, some of the observed differences may reflect patient selection or tumor biology rather than a true treatment modality effect.We therefore evaluated whether CPN was independently associated with posttransplant overall survival and recurrence after adjustment for hepatic reserve and tumor burden in the largest available US national transplant database. Secondary aims were to examine the association between bridging modality and CPN and to compare outcomes among patients who achieved CPN, patients treated without CPN, and patients who received no bridging therapy. In addition, we sought to determine whether this association persisted in subgroups of patients who had low and high risk of waitlist dropout.
METHODS
We conducted a retrospective cohort study using the Organ Procurement and Transplantation Network (OPTN) Standard Transplant Analysis and Research (STAR) files, which include data for nearly all US liver transplant candidates and recipients. This study was determined to be nonhuman subjects research by the University of California San Diego Institutional Review Board since no identifiable patient information was ever available to the investigators. The cohort comprised candidates listed with an HCC MELD exception with at least one locoregional therapy recorded on an exception application (TACE, TARE, percutaneous or chemical ablation, cryoablation, EBRT, or a combination), who underwent deceased-donor liver transplantation and had an explant pathology report linked to a recorded hepatic function assessment at listing (Figure 1). Patients under age 18 at listing were excluded. Candidates with no locoregional therapy recorded on any application were excluded from the main analytic cohort.
Cohort flow from all bridged candidates with HCC in the OPTN STAR registry to the primary analytic cohort. The comparator group used in the 3-arm analysis is shown at right, after screening for evidence of unrecorded pretransplant treatment, and is mutually exclusive of the treated cohort. Abbreviations: CPN, complete pathologic necrosis; OPTN, Organ Procurement and Transplantation Network; STAR, Standard Transplant Analysis and Research.
The main cohort (n=16,955) included transplants performed between 2012 and 2026. The parent population of candidates with HCC who underwent bridge therapy and deceased-donor transplantation (n=27,982, Figure 1) spans 2002 through 2026. The cohort was restricted to transplants performed between 2012 and 2026, since the hepatic-severity fields and transplant pathology reports needed were not consistently recorded before that year (Figure 1).
CPN was defined as the absence of an identifiable or viable tumor on explant. Tumor burden was taken from the original presenting tumor recorded on the exception application. Milan status was derived from the individual original lesion sizes. Child-Pugh class and albumin-bilirubin grade at listing were calculated from albumin, bilirubin, international normalized ratio, ascites grade, and encephalopathy grade recorded on the initial listing application. The MELD laboratory score was taken directly from the registry. Child-Pugh class, albumin-bilirubin grade, and MELD score served as the measures of hepatic reserve adjusted for in the multivariable analyses below. Era was categorized as pre-2014, 2014–2021, and post-2021, corresponding approximately to an early period, a TACE-predominant period, and a subsequent period in which TARE and thermal ablation together accounted for most bridging applications nationally.
The OPTN HCC exception application records treatment only as one of a fixed set of locoregional and surgical options and contains no field for systemic therapy. Systemic therapy given before or after locoregional therapy, therefore, could not be identified, and patients who received systemic therapy in addition to locoregional therapy were classified by their recorded locoregional modality. The exclusion of patients treated with multiple modalities from the lesion-level model, therefore, applied only to combinations of recorded locoregional modalities.
To determine whether locoregional therapy without CPN was associated with survival or recurrence relative to no bridging therapy, we identified candidates with HCC transplanted without recorded locoregional treatment. Candidates with an approved HCC exception, a recorded index tumor, and no locoregional therapy recorded on any exception application numbered 7537. Because the exception application captures only treatments entered by the listing center, we screened this group against every other treatment signal available in the registry and excluded any candidate with plausible evidence of unrecorded therapy. This excluded additional patients: the explant form (but not exception point form) recording pretransplant treatment for HCC (795 patients), necrosis coded on any explant tumor (63), a tumor recorded as smaller than its original presenting size (70), a tumor that shrank across serial exception applications (624), or an application submitted indicating “0 cm” of HCC, which implied an already treated tumor by the time of listing (118). The last 3 criteria rest on the assumption that untreated tumors do not shrink. These criteria removed 1670 candidates (22.2%) and yielded the comparator group of 5867 recipients with no plausible evidence of pretransplant treatment that was used in the primary 3-arm analysis. Because bridging therapy has become more common over time, these recipients were concentrated in an earlier transplant era (median transplant year 2008; 84% transplanted in or before 2015). All analyses were therefore adjusted for era as above. Additional sensitivity analyses using the no listed locoregional therapy comparator and restricting all arms to transplants performed in 2012 or later are reported in Supplemental Tables S7 and S8, Supplemental Digital Content 1, https://links.lww.com/HC9/C509.
A post hoc risk-stratified subgroup analysis compared 2 cohorts at opposite ends of the baseline risk of waitlist dropout spectrum: a very-low-risk cohort (solitary tumor 2–3 cm and alpha-fetoprotein [AFP] <100 ng/mL) and a high-risk cohort (AFP ≥100 ng/mL or tumor burden beyond Milan criteria). These cohorts were modeled on previously described phenotypes associated with low and high risk of waitlist dropout.12–14
Overall survival and recurrence-free survival were compared by CPN status using Kaplan-Meier estimates and log-rank tests. Time zero was defined as the date of liver transplantation. Follow-up was censored at each patient’s last reported status update in the STAR files (as of July 7, 2026). Cumulative incidence of recurrence was estimated with the Aalen-Johansen method, treating death without recurrence as a competing event. Multivariable Cox proportional-hazards models were used for overall and recurrence-free survival, and cause-specific Cox models were used for recurrence with death censored. All multivariable models were additionally adjusted for transplant era.
The 3-group comparison of CPN achieved, treated without CPN, and the comparator with no evidence of pretransplant treatment used a log-rank test for overall survival and Gray test for the cumulative incidence of recurrence, together with a 3-level categorical Cox model with the comparator group as reference. The association between bridging modality and CPN was modeled with multivariable logistic regression adjusted for tumor burden at original presentation, tumor number, AFP, and transplant era. Patients who received more than one bridging modality were classified as TACE+Ablation if those were the only 2 modalities used. Patients who received other combinations of multiple modalities were excluded from the lesion-level model, since an exposure involving multiple treatment modalities could not be attributed to a specific modality. A lesion-level analysis used generalized estimating equations clustered by patients. Because per-lesion size was not separately recorded for every lesion, the lesion-level model was adjusted for the index (largest recorded) tumor burden at original presentation as a patient-level proxy for overall tumor burden, together with tumor number, AFP, and transplant era, matching the patient-level model.
To assess whether partial tumor destruction was associated with outcome, explant necrosis was also classified using the registry’s three-level per-tumor field (none, incomplete, or complete), assigning each patient the least favorable category. Multivariable Cox models for overall survival, recurrence, and recurrence-free survival were fit on the patients with complete data for every model covariate (n=16,677 of 16,955). Missing data were not imputed.
A separate multivariable logistic regression evaluated predictors of CPN achievement among treated patients, adjusting for Child-Pugh class, albumin-bilirubin grade, age, sex, MELD score, diabetes, PVT history, HCV status, index tumor size, tumor number, AFP, macrovascular invasion, and waitlist time, to test whether hepatic reserve was independently associated with CPN after accounting for tumor burden and other clinical factors (Supplemental Table S5, https://links.lww.com/HC9/C509). All tests were 2-sided, and p<0.05 was considered significant. Analyses were performed in R (R Foundation for Statistical Computing) and Python v3.11 (Python Software Foundation).
ChatGPT (OpenAI) was used to assist with grammatical, stylistic, and formatting edits during manuscript preparation, and all suggested changes were reviewed and approved by the authors, who take full responsibility for the final manuscript.
RESULTS
Of 43,294 candidates listed with an HCC MELD exception who had at least one locoregional therapy recorded, 27,982 underwent deceased-donor liver transplantation, and 16,955 patients were ultimately evaluable. In total, 5132 (30.3%) achieved CPN, and 11,823 (69.7%) did not (Figure 1). Baseline characteristics by CPN status are shown in Table 1. The median follow-up was 5.0 years.
Baseline characteristics of the analytic cohort, by complete pathologic necrosis status
Note: n=16,955.
Continuous variables are summarized as median (IQR) and compared with the Wilcoxon rank-sum test; categorical variables are summarized as number (%) and compared with the χ2 test. Percentages are of nonmissing observations for each variable.
A small number of patients are missing individual covariates, so category counts within a variable may not sum to the column total.
Overall survival differed among patients who achieved CPN (n=5106), patients treated without CPN (n=11,792), and the comparator with no evidence of pretransplant treatment (p<0.001; Figure 2C). Survival was superior in the CPN group to that in the treated-without-CPN group (HR: 0.68, 95% CI: 0.63–0.72). Five-year overall survival was 84.9% with CPN, 76.9% with treatment without CPN, and 71.6% in the comparator group, and 5-year cumulative incidence of recurrence was 2.9%, 8.5%, and 7.0%, respectively (Figure. 2D). In a 3-level categorical Cox model, CPN was associated with better survival (HR: 0.64, 95% CI: 0.59–0.70; p<0.001) and lower recurrence (HR: 0.47, 95% CI: 0.38–0.58; p<0.001), whereas treatment without CPN was not associated with a survival advantage (HR: 0.95, 95% CI: 0.89–1.01; p=0.12) and was associated with higher recurrence (HR: 1.44, 95% CI: 1.25–1.66; p<0.001). One hundred thirty-four of 5106 patients (2.6%) who achieved CPN recurred, compared with 962 of 11,792 (8.2%) treated without CPN. Patients with incomplete necrosis (n=6171) and patients with no necrosis (n=5047) had similar overall survival. Relative to complete necrosis (n=5655; the necrosis category was indeterminate in 82 patients), the aHRs were 1.37 (95% CI: 1.27–1.47) for incomplete necrosis and 1.39 (95% CI: 1.28–1.50) for no necrosis (Figures 2A, B). For recurrence, the aHRs were 2.60 (95% CI: 2.17–3.10) for incomplete necrosis and 2.49 (95% CI: 2.07–2.99) for no necrosis (both p<0.001).
(A) Kaplan-Meier overall survival and (B) cumulative incidence of recurrence with death as a competing risk, stratified by patient-level worst necrosis category (none, incomplete, or complete). (C) Overall survival and (D) cumulative incidence of recurrence for the 3-arm comparison of patients who achieved CPN, patients treated without CPN, and the comparator group with no evidence of pretransplant treatment. Abbreviation: CPN, complete pathologic necrosis.
Figure 3A shows the full overall survival model, and Figure 3B shows the full cause-specific recurrence model. Full recurrence and recurrence-free survival models are shown in Supplemental Tables S1 and S2, https://links.lww.com/HC9/C509. Explant evidence of macrovascular invasion was the strongest predictor of worse outcomes (HR: 2.11 for overall survival and 3.54 for recurrence), and complete pathologic necrosis was the strongest predictor of improved survival and reduced recurrence.
Multivariable Cox proportional-hazards model for (A) overall survival and (B) cause-specific recurrence, using tumor size and Milan status at original presentation (n=16,677). Macrovascular invasion in this model was seen on explant pathology, not pretransplant imaging. Models refit without explant macrovascular invasion are shown in Supplemental Table S6, https://links.lww.com/HC9/C509. Abbreviations: AFP, alpha-fetoprotein; ALBI, albumin-bilirubin;
Complete pathologic necrosis was associated with improved overall survival and lower HCC recurrence in both the very-low-risk cohort (n=7468, CPN achieved in 34.7%) and the high-risk cohort (n=2801, CPN achieved in 23.2%) (Figure 4, Supplemental Table S3, https://links.lww.com/HC9/C509). In the very-low-risk group, 5-year overall survival was 85.0% among patients who achieved CPN versus 79.4% among those who did not, and 5-year recurrence was 1.9% versus 6.5% (adjusted HR for CPN 0.80, 95% CI: 0.72–0.89 for overall survival; HR: 0.31, 95% CI: 0.23–0.43 for recurrence; both p<0.001). In the high risk of dropout cohort, 5-year overall survival was 86.2% versus 72.3%, and 5-year recurrence was 5.9% versus 12.6% (adjusted HR: 0.67, 95% CI: 0.56–0.80 for overall survival; HR: 0.43, 95% CI: 0.30–0.62 for recurrence; both p<0.001).
(A) Kaplan-Meier overall survival and (B) cumulative incidence of recurrence with death as a competing risk, stratified by complete pathologic necrosis (CPN) status within two post hoc risk-stratified cohorts re-derived from tumor size and Milan status at original presentation: a very-low-risk cohort (solitary tumor 2–3 cm, AFP <100 ng/mL; n=7468) and a high-risk cohort (AFP ≥100 ng/mL or tumor burden beyond Milan criteria; n=2801). Abbreviation: CPN, complete pathologic necrosis.
At the patient level, unadjusted CPN rates in the full analytic cohort ranged from 23.1% with TACE, 30.8% with TACE plus ablation, and 31.9% with EBRT, to 38.5% with percutaneous ablation and 41.9% with TARE. After adjustment for tumor burden at original presentation, tumor number, AFP, and transplant era, TARE (OR: 2.19, 95% CI: 1.97–2.44), ablation alone (OR: 1.84, 95% CI: 1.68–2.03), TACE plus ablation (OR: 1.46, 95% CI: 1.32–1.62), EBRT (OR: 1.38, 95% CI: 1.04–1.83), and other or combination therapy (OR: 1.29, 95% CI: 1.15–1.46) were each associated with higher odds of CPN than TACE (all p<0.001 except EBRT, p=0.026) (Figure 5A). The lesion-level analysis comprised 23,208 lesions in 12,487 patients treated with a single modality (Figure 5B). After adjustment, percutaneous ablation and TARE were both associated with higher odds of lesion-level CPN than TACE (OR: 2.07, 95% CI: 1.91–2.25 and OR: 2.08, 95% CI: 1.89–2.30; both p<0.001), whereas EBRT was not (OR: 1.08, 95% CI: 0.83–1.41; p=0.55). Full models underlying Figure 5, including covariate coefficients and the number of patients or lesions included and excluded by modality, are provided in Supplemental Table S4A (patient level) and Supplemental Table S4B (lesion level), https://links.lww.com/HC9/C509.
Bridging modality and achievement of CPN, with TACE as reference. (A) Patient level, in which the CPN requires complete necrosis of every tumor. (B) Lesion level, in which each treated lesion contributes separately. Combination therapy groups appear only in panel A because a lesion treated with more than one modality cannot be attributed to a single method in this dataset. CPN, complete pathologic necrosis; EBRT, external beam radiation therapy; TACE, transarterial chemoembolization; TARE, transarterial radioembolization.
DISCUSSION
In this national cohort of 16,955 recipients of transplant, the association between CPN and posttransplant outcomes persisted after adjustment for hepatic reserve, tumor burden, AFP, wait time, and transplant era. This further reinforces findings from the 3439-patient cohort reported by DiNorcia et al2 and the 501-patient series reported by Agopian et al3 to a substantially larger and more contemporary population that represents current locoregional therapy practice. The results are in contrast to the smaller propensity-matched analysis by Grąt et al,5 in which complete pathologic response was no longer independently associated with recurrence or survival after adjustment.
Bridging modality was found to be independently associated with CPN. TARE and thermal ablation were associated with approximately twice the odds of CPN compared with TACE, which is consistent with the previously reported data by Sarwar et al.8 Fakhoury et al,15 in a national cohort of 8367 candidates bridged with TACE or TARE, found that TARE use increased from 22.5% to 51.7% between 2016 and 2024 and was independently associated with lower waitlist dropout than TACE, consistent with the higher rate of CPN from TARE observed in this study. TACE plus ablation’s intermediate odds of CPN may also partly reflect bias from patient selection, as treating physicians may favor combination therapy for anatomically challenging tumors (eg, near critical structures) not captured in this database. Given changes in treatment techniques, the median follow-up ranged from 6.8 years after TACE to 2.1 years after TARE; therefore, treatment modality remains closely confounded with the transplant era. To evaluate the effect of individual modalities on posttransplant survival and recurrence will require future study once these cohorts have matured.
Consistent with prior work, in this study, locoregional therapy without CPN was not associated with a survival advantage over the comparator with no evidence of pretransplant treatment, but was associated with higher recurrence. Agopian et al6 found the same, with treatment without a complete pathologic response conferring no survival benefit and more recurrence than no treatment (HR: 1.32; p=0.044).6 Our recurrence estimate (HR: 1.44, 95% CI: 1.25–1.66) is also of similar magnitude to the association reported by Kardashian et al.7 Patients with incomplete and with absent necrosis had similarly worse survival than those with complete necrosis (adjusted HR: 1.37 and 1.39), suggesting that any residual viable tumor is associated with increased risk.
However, an important aim of any bridging and downstaging therapy should be safety prior to transplant, as a patient who becomes ineligible for transplant due to toxicity from local therapy is worse off than the small risk of recurrence if transplanted without CPN. Mehta et al,16 in the multicenter MERITS-LT consortium, prospectively followed 209 patients undergoing downstaging and found no differences in waitlist dropout between TACE and TARE.16 As stated before, Fakhoury and colleagues reported that TARE was associated with a lower hazard of waitlist dropout than TACE (HR: 0.83), as did a separate analysis of 5677 candidates for liver transplant in the United Network for Organ Sharing (HR: 0.78).15,17 Percutaneous ablation has fewer comparative data, though a 10-year intention-to-treat analysis of radiofrequency ablation reported a waitlist dropout rate of 7.4%, similar to the bridging and downstaging studies above.18 Overall, using modalities that have higher rates of CPN does not appear to confer additional risk of transplant waitlist dropout, although therapy-related toxicity should still be considered when selecting a treatment modality.
Interpretation of the presented results should also account for the broad, multiera study window. For thermal ablation, modern stereotactic multiprobe procedures with intraprocedural image fusion achieved complete pathologic necrosis in 94.8% of patients.19 For radiation segmentectomy, Montazeri and colleagues reported an increase from 49% to 76% after dosimetry-guided treatment intensification, and the COBRAS study reported rates of 91% when optimized dose thresholds were met (importantly, doses above 440 Gy).20,21 A contemporary single-center tumor-level comparison using these improved thresholds reported CPN rates of 29.0% after TACE and 83.3% after TARE.22 Stereotactic body radiation therapy results are more heterogeneous and vary widely by technique and follow-up interval.9–11 These differences may also reflect variation in patient selection, treatment intent, and technique.
This study has several limitations. Explant pathology was available only for patients who reached transplantation, so the analysis does not capture waitlist dropout or other intention-to-treat outcomes, which can affect the interpretation of the benefit of bridging therapy.23 Treatment was not randomly assigned, and residual confounding by unmeasured tumor biology, imaging modalities, treatment intensity, retreatment, institutional experience, or physician selection remains possible. In addition, the registry records necrosis as none, incomplete, or complete rather than as a percentage, precluding analysis of thresholds such as 50% or 90%, though single-center data suggest “extensive” (>50%) necrosis may confer outcomes similar to complete necrosis, a distinction this registry’s 3-level field cannot capture.24 Because recurrence estimation in OPTN relies on voluntary center reporting through the posttransplant malignancy follow-up form, absolute recurrence estimates likely understate true recurrence, though this limitation should affect the CPN and comparison groups similarly. The registry’s EBRT and TARE flags also do not capture dosimetry or distinguish stereotactic body radiotherapy from conventional fractionated regimens or palliative TARE from the ablative “radiation segmentectomy” technique. The OPTN exception application has no field for systemic therapy, so systemic agents given before or after locoregional therapy could not be identified or adjusted for. This is a growing limitation as immunotherapy-based regimens are being used more frequently before transplant.
CONCLUSIONS
Among 16,955 recipients of liver transplant, complete pathologic necrosis after bridging locoregional therapy was associated with improved overall survival and lower recurrence after adjustment for hepatic reserve and tumor burden. TARE and percutaneous ablation were associated with approximately twice the odds of CPN compared with TACE. Treatment without CPN was not associated with better survival than no locoregional therapy and was associated with more recurrence. These observed improvements in survival and recurrence outcomes support complete pathologic necrosis as a therapeutic goal of bridging and downstaging locoregional therapy for HCC.