Induction Immunosuppression Does Not Worsen Tumor Recurrence After Liver Transplantation for Hepatocellular Carcinoma

Abstract

Background.

Prior studies are inconsistent regarding the impact of antibody induction therapy on outcomes after liver transplantation (LT) for hepatocellular carcinoma (HCC).

Methods.

Adults transplanted with HCC exception priority were identified from February 27, 2002, to March 31, 2019, using the United Network for Organ Sharing database. Time-to-event analyses evaluated the association of antibody induction therapy (none, nondepleting induction [NDI], depleting induction [DI]) with overall post-LT patient survival and HCC recurrence. Separate multivariable models adjusted for tumor characteristics on either last exception or on explant. The interaction of induction and maintenance regimen at LT discharge was investigated.

Results.

Among 22 535 LTs for HCC, 17 688 (78.48%) received no antibody induction, 2984 (13.24%) NDI, and 1863 (8.27%) DI. Minimal differences in patient and tumor characteristics were noted between induction groups, and there was significant center variability in practices. NDI was associated with improved survival, particularly when combined with a calcineurin inhibitor (CNI) and antimetabolite (hazard ratio [HR] 0.73 versus no induction plus 3-drug therapy in the last exception model [P < 0.001]; HR 0.64 in the explant model [P = 0.011]). The combination of DI with CNI alone was also protective (HR 0.43; P = 0.003). Neither NDI nor DI was associated with tumor recurrence (all P > 0.1). However, increased HCC recurrence was observed with no induction plus CNI monotherapy (HR 1.47, P = 0.019; versus no induction plus 3-drug therapy).

Conclusions.

In conclusion, induction immunosuppression was not associated with worse post-LT outcomes in patients transplanted with HCC exception priority. An improvement in survival was possibly observed with NDI.

INTRODUCTION

Hepatocellular carcinoma (HCC) is a leading cause of liver transplantation (LT).¹ Despite careful selection criteria, HCC recurs in approximately 5% to 20% of patients who have confirmed tumor on explant.^2–8 HCC recurrence typically occurs within the first several years after LT and has a poor median survival of about 1 y. Often refractory to the limited treatment options available, recurrent HCC is aggressive with few patients surviving beyond 5 y postdiagnosis.^6–12

Several strategies have been suggested to reduce the risk of post-LT HCC recurrence, including careful pre-LT selection of candidates, adjuvant chemotherapy, and choice of immunosuppression.^11,13 It has long been established that the immune system plays a crucial role in modulating the development and progression of malignancy, including recurrent HCC posttransplant.^14,15 HCC recurrence is believed to be secondary to small populations of residual tumor cells, whose survival could be affected by immunemodulating medications after transplant.^11,13,16,17 Given the accelerated rate of disease progression in LT recipients with HCC recurrence relative to nontransplanted HCC patients, immunosuppression has also been hypothesized to hasten tumor progression.^6–12,18

Immunosuppression drugs may affect the risk of post-LT HCC recurrence through different mechanisms. Mechanistic target of rapamycin inhibitors (mTORi) have unique antiproliferative effects that may reduce HCC recurrence, whereas calcineurin inhibitors (CNIs) may increase this risk.^19–25 Data on corticosteroids, mycophenolate mofetil, and azathioprine are more limited.^11,13,26 To date, few studies have evaluated the impact of induction immunosuppression on the risk of HCC recurrence or overall mortality post-LT.^27,28 In an older study from the 1990s, induction therapy using antilymphocyte antibody (ATG) or anti-CD3 antibody (OKT3) led to increased tumor recurrence.²⁹ Another small single-center retrospective cohort study from Korea in the early 2000s also showed an increased risk of post-LT HCC recurrence in patients receiving basiliximab induction.³⁰ In contrast, a more recent study in 2010 using the Scientific Registry of Transplant Recipients database indicated improved survival with interleukin-2 receptor antagonist (IL2Ra) induction (74% versus 68% at 5 y, P ≤ 0.01) but no difference with ATG (P = 0.44), although tumor explant findings and HCC recurrence were not assessed.³¹ Importantly, each of these studies took place in the era before highly effective direct-acting antivirals (DAAs) for hepatitis C virus (HCV) infection. It is likely that the characteristics and management of the patients included differed substantially from those today, particularly because concerns had been raised previously on the safety of induction immunosuppression in the setting of untreated HCV infection.^32,33

Consensus recommendations on the optimal immunosuppression management of HCC recipients are inconsistent, and specific guidance on induction therapy is lacking. However, the American Association for the Study of Liver Diseases and European Association for the Study of Liver Diseases guidelines do not address induction therapy in this context.^34,35 More recent guidance from the International Liver Transplantation Society and the Società Italiana Trapianti D’Organo highlight the importance of CNI and overall immunosuppression minimization in HCC recipients, although they again do not indicate whether this should be achieved with the support of induction therapy.^36,37 In fact, Società Italiana Trapianti D’Organo advises caution with the use of induction therapy in the setting of HCC. Acknowledging the need for more data, the conclusion from the International Liver Transplantation Society states, “beyond generally minimizing overall immunosuppression, the optimal immunosuppression strategy for minimizing the frequency and severity of recurrence of HCC […] has not been determined.”³⁶

In this study, we evaluate whether antibody-based induction immunosuppression is associated with all-cause mortality and HCC recurrence in patients transplanted for HCC in 2 ways: (1) as an intention-to-treat analysis (ie, ignoring the subsequent choice of maintenance immunosuppression) and (2) by modeling the interaction of induction therapy with maintenance immunosuppression at hospital discharge (ie, simulating a per-protocol approach). We also develop separate multivariable models that account for pretransplant tumor burden on imaging (ie, what is known at the time of opting to use induction therapy) and tumor burden on explant (ie, reflective of tumor biology) given historic discrepancies in pre-LT staging and post-LT pathology.³⁸

MATERIALS AND METHODS

Study Population

This was a retrospective cohort study using the United Network for Organ Sharing (UNOS) database (Figure 1). All adults undergoing first LT alone for HCC between February 27, 2002, and March 31, 2019, were included. Given the interest in evaluating tumor characteristics, subjects were required to either have (1) pre-LT HCC exception point request data or (2) HCC explant data in the UNOS database. In the United States, patients with HCC meeting certain criteria (eg, T2 criteria) are eligible for additional Model for End-Stage Liver Disease (MELD) priority points called exception points. Table S1 (SDC, http://links.lww.com/TP/C660) includes a summary of recent US policies for LT allocation in patients transplanted with HCC.^39–41 The majority of subjects (70.41%) in our study met standard exception point criteria for T2 HCC.⁴²

FIGURE 1.

Study cohorts included in analyses and rates of antibody induction use.

Exposures and Covariates

The primary exposure was receipt of antibody-based induction therapy as a 3-level categorical variable: no antibody induction, nondepleting induction (NDI), and depleting induction (DI). The no antibody induction group included patients receiving no induction whatsoever or corticosteroid-only induction. NDI included IL2Ra (daclizumab and basiliximab), and DI included ATG, OKT3, alemtuzumab, and rituximab. As per our previously published methods, if no specific antibody induction agent was reported, patients were assumed to have not received antibody-based induction therapy.^43,44 Maintenance immunosuppression was collected at hospital discharge, 6 mo, and 1 y post-LT. It is important to note that (1) there is no ability to evaluate maintenance immunosuppression longitudinally in the UNOS database and (2) there is significant data missingness after hospital discharge (though maintenance regimen at discharge is complete). Using individual drugs reported in the database, a 5-level variable was developed at each time point that included the following regimens: CNI with antimetabolite (antiM) and steroid, CNI with antiM, CNI with steroid, CNI alone, and mTORi-based. CNI drugs included tacrolimus and cyclosporine, antiM drugs included azathioprine and mycophenolate, and mTORi included sirolimus and everolimus. Subjects were considered to be on an mTORi-based therapy if on any regimen that included an mTORi (eg, mTORi alone or mTORi in combination with another immunosuppression medication).

Demographic characteristics included sex, age, and race and ethnicity. LT year and transplanting center were assessed to evaluate trends in induction use among HCC recipients. Clinical variables at LT included liver disease cause, ascites or encephalopathy severity, native MELD score, and serum creatinine. Tumor characteristics were evaluated at 3 time points: first exception request, last exception request, and explant pathology. Variables from exception request data included the largest tumor size, number of tumors (1, 2, 3, or >3), alpha-fetoprotein (AFP; 0–20, 21–99, 100–499, ≥500 ng/mL), and receipt of HCC therapy. HCC treatment modalities were classified as ablative locoregional therapy (LRT; radiofrequency ablation, thermal ablation, chemical ablation or cryoablation), nonablative LRT (transarterial chemoembolization or microspheres), or other (resection, external beam radiation). Tumor characteristics obtained from explant pathology reports included the largest tumor size, number of tumors (1, 2, 3, or >3 tumors/infiltrative), history of HCC treatment (yes/no), tumor differentiation (well, moderate, or poor differentiation or complete necrosis), vascular invasion (none, microvascular, macrovascular), extrahepatic spread (including lymph node involvement), presence of ≥1 fully viable tumor, and size of the largest fully viable tumor. If explant data were available, patients’ Risk Estimation of Tumor Recurrence After Transplant score was obtained and evaluated in continuous and binary forms (<4 and ≥4).^5,45

Outcomes

This study had 2 primary outcomes: post-LT patient survival and HCC recurrence. HCC recurrence was defined as being present if patients were identified as having recurrence of a pretransplant tumor (ie, malig_pretx_tumor = Y) and the recurrence was that of a nonincidental liver tumor (ie, pre_tum_ty = 16) or post-LT cause of death attributed to metastatic HCC.^45,46 Among those with identified HCC recurrence before death, the earliest date of documented recurrence was used.

Statistical Analyses

Demographic and clinical characteristics at LT were compared across induction groups using chi-square tests for categorical variables and Kruskal-Wallis tests for continuous variables. Descriptive methods were also used to compare tumor characteristics at the time of first exception application, last exception application, and explant pathology. The association of induction type and maintenance immunosuppression regimen was assessed descriptively at discharge, 6 mo, and 1 y post-LT.

We then conducted a series of analyses as shown in Table S2 (SDC, http://links.lww.com/TP/C660). First, multivariable models evaluated antibody induction therapy (3-level: none [reference], NDI, DI) as the primary exposure. In these models, the maintenance regimen was not accounted for, effectively simulating an “intention-to-treat” scenario. First, the relationship between induction immunosuppression and all-cause mortality was investigated using Cox regression. For this part of the analysis, the event was death, with follow-up time defined as the minimum of time from transplant to the earliest of death (from any cause), loss-to-follow-up, or the end of the observation period. Separate multivariable models were developed that adjusted for either tumor features on the last available HCC exception application (ie, all knowledge available pre-LT when the decision to administer induction immunosuppression is made) or tumor pathology features on explant (ie, factors more directly related to tumor biology). Covariates that were obtained from the last exception application included the largest tumor size, tumor number, last AFP reported, number of LRT received pre-LT, and time from first application to LT. Covariates obtained from explant pathology included the largest viable tumor size, number of viable tumors, worst tumor differentiation, vascular invasion, and extrahepatic/lymph node spread.

Second, we investigated the association between induction immunosuppression and HCC recurrence, again using Cox regression. However, for this part of the analysis, we modeled the cause-specific recurrence hazard to accommodate death as a competing risk.⁴⁷ Here, the event of interest was HCC recurrence (including recurrences that were detected at death, as evidenced by the cause of death being HCC), with follow-up time defined as the time between transplant and the earliest of death, HCC recurrence, loss to follow-up, and the end of the observation period. Covariates for each of the HCC recurrence models paralleled those described in the preceding paragraph for all-cause mortality for each of the last exception and explant cohorts.

We then repeated the aforementioned analyses, including an interaction term between induction therapy and maintenance regimen at discharge (Table S2, SDC, http://links.lww.com/TP/C660). This effectively allows for the identification of early immunosuppression regimen combinations that improve or worsen outcomes in HCC patients, simulating what would occur in a per-protocol analysis of a clinical trial. Induction immunosuppression was included as a 3-level covariate and maintenance regimen at hospital discharge as a 5-level covariate. Here, the reference was receipt of no induction plus CNI + antiM + steroid at LT discharge (ie, the most commonly used immunosuppression combination overall). Because of the high degree of missingness in UNOS regarding maintenance immunosuppression at 6 mo and 1 y, as well as the inability to evaluate maintenance therapy longitudinally between these time points, we felt it would be inappropriate to include these covariates in the interaction analyses using the data at hand. Separate multivariable models were again assessed that specified tumor characteristics either on last exception application or on explant, as previously described, for all-cause mortality and HCC recurrence.

All analyses were stratified by LT center to prevent confounding in a flexible manner, given marked differences in induction practices between centers. In addition, all models adjusted for patient clinical characteristics at LT (sex, age, race and ethnicity, primary liver disease cause, ascites severity, encephalopathy grade, native MELD score, and creatinine). We explored LT year as a potential confounder in the models (given temporal trends) and the interaction of induction therapy with HCV (given prior safety concerns and the introduction of DAAs during the study period). Of note, all analyses using explant pathology were inherently smaller, given that collection of explant data by UNOS began on April 12, 2012. Survival curves were generated from a Cox model for overall survival that was very similar to that underlying the primary results presented. This model was stratified by induction type rather than center. The curves pertain to hypothetical LT recipient with continuous covariates set to the median value and categorical covariates set to the most common category.

This study was approved by the Institutional Review Board of the University of Pennsylvania. All analyses were performed using STATA, version 14 (College Station, TX).

RESULTS

The study cohort included 22 535 subjects transplanted with HCC between February 27, 2002, and March 31, 2019. At the time of LT, 17 688 (78.48%) of the subjects received no induction, 2984 (13.24%) received NDI, and 1863 (8.27%) received DI. Among subjects in the starting cohort, 22 059 (97.88%) had at least 1 exception point request pretransplant, and 10 943 (48.56%) had available explant data (since 2012). Maintenance immunosuppression regimens were available for 21 817 (96.8%), 9340 (41.45%), and 13 853 (61.47%) subjects at discharge, 6 mo, and 1 y post-LT, respectively.

Demographic and Clinical Characteristics

There were few meaningful differences in patient characteristics between induction groups, suggesting that these played a limited role in decision-making (Table 1). Patients receiving NDI were more likely to have nonalcoholic steatohepatitis (12.60% versus 9.58% no induction versus 10.41% DI), whereas those receiving DI were more likely to have autoimmune liver disease (3.54% versus 2.74% no induction versus 2.71% NDI; P < 0.001 in pairwise comparisons). Moderate-to-severe ascites were more prevalent in the NDI group (9.54% versus 7.77% no induction versus 5.90% DI; P < 0.001 overall). Although statistically significant, there was no clinical difference in creatinine (median 0.88 for no induction versus 0.90 for NDI versus 0.90 mg/dL for DI; P < 0.001) or native MELD score at LT (median 11 for all cohorts; P < 0.001) by the induction group.

TABLE 1.

Demographics and patient characteristics

	No induction (N = 17 688)	Nondepleting induction (N = 2984)	Depleting induction (N = 1863)	P
Male, n (%)	13 602 (76.90)	2305 (77.25)	1440 (77.29)	0.866
Age at LT (y), median (IQR)	59 (54–64)	60 (55–64)	59 (55–64)	0.064
Race and ethnicity, n (%)				0.035
White	11 878 (67.15)	1957 (65.58)	1243 (66.72)
African American	1613 (9.12)	278 (9.32)	160 (8.59)
Hispanic	2516 (14.22)	467 (15.65)	257 (13.79)
Asian	1454 (8.22)	228 (7.64)	179 (9.61)
Other	227 (1.28)	54 (1.81)	24 (1.29)
Diagnosis, n (%)				<0.001
HCV	10 323 (58.36)	1675 (56.13)	1115 (59.85)
NASH/cryptogenic	1695 (9.58)	376 (12.60)	194 (10.41)
Alcohol	1677 (9.48)	292 (9.79)	153 (8.21)
HBV	1060 (5.99)	149 (4.99)	95 (5.10)
Autoimmune	484 (2.74)	81 (2.71)	66 (3.54)
Other	2,449 (13.85)	411 (13.77)	240 (12.88)
Encephalopathy at LT, n (%)				<0.001
None	10 634 (60.37)	1917 (64.59)	1133 (61.28)
Grade 1–2	6747 (38.30)	1003 (33.79)	680 (36.78)
Grade 3–4	233 (1.32)	48 (1.62)	36 (1.95)
Ascites at LT, n (%)				<0.001
None	8268 (46.93)	1445 (48.69)	891 (47.19)
Mild	7979 (45.29)	1240 (41.78)	849 (45.92)
Moderate-severe	1369 (7.77)	283 (9.54)	109 (5.90)
Creatinine at LT (mg/dL), median (IQR)	0.88 (0.70–1.03)	0.90 (0.73–1.10)	0.90 (0.71–1.09)	<0.001
Native MELD score at LT, median (IQR)	11 (8–15)	11 (9–15)	11 (8–14)	<0.001

IQR, interquartile range; HBV, hepatitis B virus; HCV, hepatitis C virus; LT, liver transplant; MELD, Model for End-Stage Liver Disease; NASH, nonalcoholic steatohepatitis.

Tumor Characteristics Reported by Exception Application and on Explant

Differences in tumor characteristics were small across induction groups, suggesting that HCC burden was not influential for induction immunosuppression decision-making (Table 2). Waiting time from first exception to transplant was greatest in the NDI cohort (5.06 mo) relative to the no induction and DI cohorts (4.63 and 3.35 mo, respectively; P < 0.001). This trend also held true for the time from last exception to transplant (1.41 mo for NDI versus 1.35 mo for no induction versus 1.25 mo for DI; P = 0.012), although differences were small.

TABLE 2.

HCC tumor characteristics

	No induction (N = 17 688)	Nondepleting induction (N = 2984)	Depleting induction (N = 1863)	P
First exception
Largest tumor size (cm)	2.3 (1.8–3.0)	2.2 (0–3.0)	2.2 (0–2.9)	<0.001
No. of tumors				0.070
1	13 078 (73.94)	2244 (75.20)	1334 (71.60)
2	3134 (17.72)	503 (16.86)	376 (20.18)
3	1041 (5.89)	172 (5.76)	125 (6.71)
>3	42 (0.24)	6 (0.21)	4 (0.22)
Time from first exception to transplant (mo)	4.63 (1.55–8.68)	5.06 (1.74–9.24)	3.35 (1.02–7.79)	<0.001
AFP (ng/mL), n (%)				0.002
0–20	10 736 (66.34)	1929 (69.24)	1199 (66.35)
21–99	3176 (19.63)	544 (19.53)	361 (19.98
100–499	1576 (9.74)	227 (8.15)	184 (10.18)
≥500	695 (4.29)	86 (3.09)	63 (3.49)
Last exception
Largest tumor size (cm)	2.0 (0–2.8)	1.9 (0–2.7)	2.0 (0–2.7)	<0.001
No. of tumors				0.040
1	13 466 (76.13)	2334 (78.22)	1394 (74.83)
2	2791 (15.78)	423 (14.18)	329 (17.66)
3	992 (5.61)	164 (5.50)	112 (6.01)
>3	46 (0.27)	4 (0.14)	4 (0.22)
Time from last exception to transplant (mo)	1.35 (0.69–2.20)	1.41 (0.69–2.24)	1.25 (0.59–2.24)	0.012
AFP (ng/mL), n (%)				<0.001
0–20	11 297 (69.35)	2044 (73.00)	1231 (68.05)
21–99	3025 (18.57)	471 (16.82)	359 (19.85)
100–499	1376 (8.45)	214 (7.64)	167 (9.23)
≥500	591 (3.63)	71 (2.54)	52 (2.87)
No. of HCC treatments				<0.001
0	4790 (27.70)	706 (24.14)	560 (31.45)
1	8055 (46.58)	1368 (46.77)	873 (47.47)
2	3191 (18.45)	598 (20.44)	299 (16.26)
3	847 (4.90)	178 (6.09)	68 (3.70)
>3	410 (2.37)	75 (2.56)	39 (2.12)
Explant
Largest tumor size	2.5 (1.8–3.8)	2.7 (1.3–4)	2.6 (1.8–3.5)	0.002
No. of tumors				0.008
1	3892 (49.86)	829 (53.38)	420 (46.46)
2	1857 (23.79)	359 (23.12)	222 (24.56)
3	947 (12.13)	157 (10.11)	115 (12.72)
>3 or infiltrative	1110 (14.22)	208 (13.39)	147 (16.26)
Received HCC treatment	7851 (95.24)	1601 (95.18)	865 (90.77)	<0.001
Worst tumor differentiation				0.050
Complete necrosis	1681 (21.54)	353 (22.73)	188 (20.80)
Poor	620 (7.94)	100 (6.22)	68 (7.52)
Moderate	3701 (47.42)	755 (48.62)	469 (51.88)
Well	1803 (23.10)	345 (22.22)	179 (19.80)
Vascular invasion				<0.001
Macrovascular	157 (2.01)	29 (1.87)	14 (1.55)
Microvascular	1145 (14.67)	157 (10.12)	92 (10.18)
None	6504 (83.32)	1366 (88.02)	798 (88.27)
Extrahepatic spread (including LN involvement)	172 (2.20)	36 (2.32)	46 (5.09)	<0.001
Macrovascular or extrahepatic involvement	313 (4.01)	58 (3.74)	58 (6.42)	0.002
At least 1 viable tumor	5687 (73.31)	1116 (72.42)	695 (76.97)	0.036
No. of viable tumors				0.367
0	2070 (26.69)	425 (27.58)	208 (23.03)
1	3039 (39.18)	615 (39.91)	362 (40.09)
2	1316 (16.97)	248 (16.09)	167 (18.49)
3	664 (8.56)	125 (8.11)	79 (8.75)
>3	668 (8.61)	128 (8.31)	87 (9.63)
Largest viable tumor size (ie, fully nonnecrotic)	1.8 (0–3.0)	1.9 (0–3.1)	1.8 (0–3.0)	0.103
RETREAT score	3 (1–5)	3 (1–5)	3 (1–5)	0.003
RETREAT score ≥4	1373 (18.44)	220 (14.71)	136 (15.37)	<0.001

AFP, alpha fetoprotein; HCC, hepatocellular carcinoma; RETREAT, Risk Estimation of Tumor Recurrence After Transplant.

Candidates receiving DI were more likely to have macrovascular or extrahepatic involvement (6.42% versus 4.01% for no induction versus 3.74% for NDI; P = 0.002) and ≥1 viable tumor (76.97% versus 73.31% for no induction versus 72.42% for NDI; P = 0.036) on explant. Conversely, candidates who received no induction were more likely to have microvascular invasion (14.67% compared with 10.12% for NDI and 10.18% for DI; P < 0.001). A greater proportion of patients in the no induction cohort (18.44%) had an elevated Risk Estimation of Tumor Recurrence After Transplant score ≥4 compared with the NDI and DI cohorts (14.71% and 15.37%, respectively; P < 0.001).

Temporal and Geographic Trends

Antibody induction in HCC recipients increased between 2002 and 2010 but subsequently decreased (Figure 2A). For example, NDI use increased from 11.01% in 2002 to a peak of 17.97% in 2014 and was more recently 12.50% as of 2019. Similarly, 1.89% of HCC recipients were treated with DI in 2002, peaking at 10.59% in 2015 and decreasing to 6.77% in 2019. Wide center variability in induction practices was observed in the setting of HCC (Figure 2B). In 29 centers (23.02%), more than half of HCC recipients received either NDI or DI. The median center rate of HCC recurrence was 4.6% (interquartile range, 2.1%–7.0%; Figure S1A, SDC, http://links.lww.com/TP/C660). No correlation was found between center HCC recurrence and induction practices (r_s = −0.08, P = 0.35; Figure S1B, SDC, http://links.lww.com/TP/C660).

FIGURE 2.

Temporal (A) and geographic (B) trends in antibody induction use among patients transplanted with hepatocellular carcinoma exception points. Note: In panel B, each vertical bar represents 1 transplant center.

Maintenance Immunosuppression

Two-thirds of the patients (64.74%) in the no induction group and half of those (52.24%) in the NDI group were discharged on triple therapy after LT (Figure 3; Table S3, SDC, http://links.lww.com/TP/C660). The DI group was primarily discharged on either CNI + antiM (48.42%) or CNI monotherapy (25.74%; P < 0.001 in overall intragroup pairwise comparisons). mTORi therapy at discharge was twice as frequent in the NDI group compared with either the no induction or DI cohort (P < 0.001 for pairwise comparisons). CNI monotherapy was the dominant regimen at 6 mo and 1 y in the DI group (51.66% and 50.04%, respectively). Overall, mTORi use was 5.63% and 10.40% at 6 mo and 1 y, respectively. The use of mTORi was most common in the NDI cohort (12.93% overall in the first year versus 5.28% for the no induction and 4.43% for the DI groups; P < 0.001). Of mTORi-based therapies (Table S4, SDC, http://links.lww.com/TP/C660), mTORi + CNI + steroids (47.20%) was the most common regimen at discharge.

FIGURE 3.

Maintenance immunosuppression at liver transplant discharge by induction group among patients transplanted with hepatocellular carcinoma exception points. Note: P < 0.001 each for discharge, 6 mo, and 1 y; data are missing for 13 195 patients (58.55%) at the 6-mo follow-up and 8682 patients (38.53%) at the 1-y follow-up. antiM, antimetabolite; CNI, calcineurin inhibitor; mTORi, mechanistic target of rapamycin inhibitor.

All-Cause Post-LT Mortality

Among the 22 059 patients with ≥1 HCC exception application, the median follow-up time was 3.95 y (interquartile range, 1.28–7.19), during which 6366 (28.86%) deaths were recorded. Covariate-adjusted 5-y survival was estimated at 81% for NDI, 79% for DI, and 78% for no induction (Figure 4). NDI was associated with a reduction in all-cause mortality in the last exception cohort (adjusted hazard ratio [HR] 0.88, P = 0.044) but not in the explant cohort (P = 0.500, Table 3; full multivariable model output in Tables S5 and S6, SDC, http://links.lww.com/TP/C660). DI was not associated with all-cause mortality in either the exception or explant cohort (all P > 0.1). There was no difference in the point estimates obtained when adjusted for LT year, and no interaction with HCV was found, suggesting the introduction of DAA therapy for HCV had no significant effect on study outcomes.

FIGURE 4.

Overall survival probability by induction type. DI, depleting induction; NDI, nondepleting induction.

TABLE 3.

Association of induction immunosuppression on all-cause mortality and HCC recurrence accounting for tumor exception characteristics or explant features

Cohort	Discharge induction regimen	All-cause mortality		HCC recurrence
		HR (95% CI)	P	HR (95% CI)	P
Last exception cohorta,b (N = 22 033)	None	Reference	-	Reference	-
	NDI	0.88 (0.80–0.97)	0.01	0.92 (0.76–1.12)	0.43
	DI	0.98 (0.86–1.12)	0.79	1.05 (0.80–1.38)	0.74
Explant cohorta,c (N = 10 915)	None	Reference	-	Reference	-
	NDI	0.88 (0.71–1.09)	0.25	0.80 (0.55–1.18)	0.27
	DI	0.95 (0.73–1.25)	0.72	0.92 (0.54–1.60)	0.78

^aAll analyses adjusted for demographic and clinical factors: sex, age at LT, race and ethnicity, primary liver disease cause, presence of ascites or encephalopathy at LT, native MELD score at LT, and serum creatinine at LT; models additionally stratified by center.

^bMultivariable model 1 adjusted for characteristics at last exception: induction immunosuppression, sex, age at LT, race and ethnicity, primary liver disease etiology, ascites severity, encephalopathy grade, native MELD score at LT, creatinine at LT, largest tumor size, tumor number, last alpha-fetoprotein reported, the number of locoregional therapies received pre-LT, and time from first application to LT.

^cMultivariable model 2 adjusted for characteristics on explant: induction immunosuppression, sex, age at LT, race and ethnicity, primary liver disease cause, ascites severity, encephalopathy grade, native MELD score at LT, creatinine at LT, largest viable tumor size, number of viable tumors, worst tumor differentiation, vascular invasion, and extrahepatic/lymph node spread.

CI, confidence interval; DI, depleting induction; HCC, hepatocellular carcinoma; HR, hazard ratio; LT, liver transplant; MELD, Model for End-Stage Liver Disease; NDI, nondepleting induction.

Point estimates in bold indicate statistically significant results with p <0.05.

Compared with the reference of no induction plus CNI + antiM + steroid at discharge, a steroid-free approach with CNI + antiM with or without NDI was associated with significant reduction in all-cause mortality: adjusted HR 0.73 (P < 0.001) and HR 0.88 (P = 0.031), respectively, in the last exception cohort (Table 4). This result persisted in the explant cohort for the group receiving NDI with CNI + antiM (adjusted HR 0.64, P = 0.011). In addition, the combination of DI with CNI alone was also protective in the explant cohort (adjusted HR 0.43, P = 0.003). Using the approach of DI with mTORi-based maintenance at discharge was associated with increased all-cause mortality, although this was only observed in the last exception cohort (adjusted HR 1.55, P = 0.026).

TABLE 4.

Overall all-cause post-liver transplant mortality by induction and maintenance immunosuppression regimens at discharge

		Maintenance regimen at discharge
		CNI + antiM + steroid	CNI + antiM	CNI + steroid	CNI alone	mTORi-based
Last exception cohorta,b (N = 22 033)	None	Reference	0.88 (0.78–0.99)	1.04 (0.96–1.13)	1.05 (0.89–1.23)	1.10 (0.93–1.31)
	NDI	0.94 (0.82–1.06)	0.73 (0.61–0.87)	1.10 (0.86–1.41)	0.88 (0.62–1.24)	0.86 (0.65–1.14)
	DI	1.05 (0.84–1.31)	0.91 (0.75–1.11)	1.01 (0.69–1.49)	0.81 (0.63–1.04)	1.55 (1.06–2.27)
Explant cohorta,c (N = 10 915)	None	Reference	0.79 (0.62–1.01)	0.80 (0.63–1.03)	0.82 (0.53–1.28)	0.96 (0.65–1.41)
	NDI	0.90 (0.71–1.15)	0.64 (0.45–0.90)	1.09 (0.59–2.01)	0.94 (0.37–2.38)	1.12 (0.47–2.69)
	DI	1.13 (0.74–1.73)	0.85 (0.58–1.25)	1.29 (0.50–3.33)	0.43 (0.25–0.75)	0.86 (0.36–2.04)

^aAll analyses adjusted for demographic and clinical factors: sex, age at LT, race and ethnicity, primary liver disease etiology, presence of ascites or encephalopathy at LT, native MELD score at LT, and serum creatinine at LT; models additionally stratified by center.

^bMultivariable model 1 adjusted for characteristics at last exception: induction immunosuppression, sex, age at LT, race and ethnicity, primary liver disease etiology, ascites severity, encephalopathy grade, native MELD score at LT, creatinine at LT, largest tumor size, tumor number, last alpha-fetoprotein reported, the number of locoregional therapies received pre-LT, and time from first application to LT.

^cMultivariable model 2 adjusted for characteristics on explant: induction immunosuppression, sex, age at LT, race and ethnicity, primary liver disease etiology, ascites severity, encephalopathy grade, native MELD score at LT, creatinine at LT, largest viable tumor size, number of viable tumors, worst tumor differentiation, vascular invasion, and extrahepatic/lymph node spread.

antiM, antimetabolite; CNI, calcineurin inhibitor; DI, depleting induction; HCC, hepatocellular carcinoma; LT, liver transplant; MELD, Model for End-Stage Liver Disease; mTORi, mammalian target of rapamycin inhibitor; NDI, nondepleting induction.

Point estimates in bold indicate statistically significant results with p <0.05.

HCC Recurrence

During follow-up, 1487 of 22 059 patients (6.74%) of the cohort were identified as having HCC recurrence. In the last exception cohort, no induction with CNI alone at discharge was associated with a significantly higher risk of HCC recurrence (HR 1.47 versus no induction with CNI + antiM + steroid, P = 0.019; Table 5; Table S5, SDC, http://links.lww.com/TP/C660), but this was not observed in the explant cohort (P = 0.566). There was possibly increased HCC recurrence in the last exception cohort in patients not receiving induction who were discharged on CNI + steroid (HR 1.18, P = 0.071; CNI + antiM + steroid reference) and mTORi (HR 1.37, P = 0.085). There were no associations between either NDI or DI and HCC recurrence in the explant cohort (all P > 0.1; Tables 3 and 5; and Table S6, SDC, http://links.lww.com/TP/C660).

TABLE 5.

Hepatocellular carcinoma recurrence by induction and maintenance immunosuppression regimens at discharge

		Maintenance regimen at discharge
		CNI + antiM + steroid	CNI + antiM	CNI + steroid	CNI alone	mTORi-based
Last exception cohorta,b (N = 22 033)	None	Reference	1.04 (0.82–1.31)	1.18 (0.99–1.42)	1.47 (1.07–2.04)	1.37 (0.96–1.95)
	NDI	0.90 (0.69–1.17)	0.89 (0.57–1.30)	1.15 (0.66–2.01)	1.54 (0.82–2.91)	1.51 (0.87–2.60)
	DI	1.05 (0.64–1.72)	1.38 (0.92–2.06)	1.58 (0.72–3.46)	1.14 (0.64–2.03)	1.04 (0.62–1.73)
Explant cohorta,c (N = 10 915)	None	Reference	1.26 (0.82–1.92)	1.02 (0.62–1.67)	1.32 (0.51–3.41)	0.60 (0.29–1.23)
	NDI	0.88 (0.56–1.39)	0.85 (0.47–1.54)	0.46 (0.10–2.06)	–	1.91 (0.35–10.26)
	DI	0.82 (0.33–2.04)	1.05 (0.47–2.36)	3.81 (0.73–19.95)	1.13 (0.29–4.42)	0.45 (0.06–3.56)

^aAll analyses adjusted for demographic and clinical factors: sex, age at LT, race and ethnicity, primary liver disease cause, presence of ascites or encephalopathy at LT, native MELD score at LT, and serum creatinine at LT; models additionally stratified by center.

^bMultivariable model 1 adjusted for characteristics at last exception: induction immunosuppression, sex, age at LT, race and ethnicity, primary liver disease cause, ascites severity, encephalopathy grade, native MELD score at LT, creatinine at LT, largest tumor size, tumor number, last alpha-fetoprotein reported, the number of locoregional therapies received pre-LT, and time from first application to LT.

^cMultivariable model 2 adjusted for characteristics on explant: induction immunosuppression, sex, age at LT, race and ethnicity, primary liver disease etiology, ascites severity, encephalopathy grade, native MELD score at LT, creatinine at LT, largest viable tumor size, number of viable tumors, worst tumor differentiation, vascular invasion, and extrahepatic/lymph node spread.

antiM, antimetabolite; CNI, calcineurin inhibitor; DI, depleting induction; LT, liver transplant; MELD, Model for End-Stage Liver Disease; mTORi, mammalian target of rapamycin inhibitor; NDI, nondepleting induction.

Point estimates in bold indicate statistically significant results with p <0.05.

DISCUSSION

Using national data, we evaluate the effects of induction immunosuppression on post-LT outcomes in patients transplanted with HCC. Our study demonstrates that receipt of induction immunosuppression is not associated with an increased risk of adverse outcomes. In fact, NDI may be associated with reduced all-cause mortality, particularly when combined with a steroid-free CNI + antiM regimen at hospital discharge. This observation was seen not only when adjusting for tumor characteristics on last exception but also when accounting for tumor burden on explant pathology. The combination of DI with CNI monotherapy was also associated with improved survival. In contrast, the combination of no antibody induction with CNI monotherapy at discharge may lead to an increased risk of HCC recurrence, which could highlight the potential benefit of CNI dose reduction. Other combinations of induction and maintenance regimen were not definitively associated with either an increased or a decreased risk of HCC recurrence, although certain cell sizes were small. This study contributes important new evidence on the optimal individualized approach to early immunosuppression management after LT for HCC.

Equally important, induction practices were markedly heterogeneous and not clearly driven by either patient or tumor characteristics. Although there were small differences in demographics, clinical characteristics, and degree of tumor burden observed between treatment groups, the choice of induction therapy was primarily influenced by a patient’s transplant center with widely variable practices nationally. There are no published clinical protocols or consensus/expert recommendations guiding the choice of induction immunosuppression in HCC recipients. This and the limited prior studies in this area likely underlie the significant heterogeneity in practice. These findings also highlight the need for further research to elucidate the potential mechanisms by which induction therapy leads to improved outcomes in specific clinical situations, which will better identify those candidates who may benefit the most from treatment.

Our finding of potentially improved post-LT survival with NDI has also been observed in the non-HCC population.^48,49 Such benefits are hypothesized to result from reducing CNI or steroid exposure, allowing for decreased incidence and severity of related drug adverse effects, such as renal toxicity or impaired glucose metabolism, respectively. It is possible that these largely remain the chief reasons for the improved mortality seen in our study of HCC patients, too, given that we did not find a significant association with HCC recurrence. It should also be noted that the NDI group had the highest rates of mTORi use, which may have also played a role. Although we did not observe a clear beneficial effect of combining NDI with mTORi-based regimens in our multivariable models, there was loss of precision because of small cell sizes. Moreover, we were limited to evaluating the maintenance regimen at discharge (which infrequently uses mTORi), given increased missingness at later time points.

There were several limitations of our study. Explant pathology data have only been available in the UNOS database since 2012; thus, analyses accounting for tumor biology had smaller sample sizes. Although we used a common approach to identify patients with HCC recurrence, we suspect that events were underestimated by our study (6.74%).^45,46 This potential misclassification was felt nondifferential with respect to the type of induction received and, therefore, unlikely to yield significant bias; however, it may have affected analytic power. Other studies evaluating HCC recurrence using UNOS have also suffered from this limitation.^5,45,46,50 Reassuringly, our multivariable analyses redemonstrated known associations between several established patient and tumor characteristics (eg, AFP, vascular invasion, size, etc.) and the outcomes studied, supporting the reliability of our data.⁵¹ Given data missingness beyond LT discharge in UNOS, we could not reliably explore the relationship between induction type and maintenance immunosuppression regimen at later time points. Because of data limitations, we also could not confirm that the survival benefit of DI combined with CNI monotherapy occurred in the setting of CNI dose reduction. These issues should be further explored using alternate data sources.

In conclusion, among patients transplanted for HCC, induction immunosuppression is not associated with worse post-LT outcomes, and improved survival may be achieved by using induction therapy in combination with certain maintenance regimens at LT discharge. We demonstrate highly variable practices in the use of induction immunosuppression that are not driven by patient or tumor characteristics, highlighting a need for clearer guidance. As the next steps, we recommend that future research studies evaluate these associations further using additional granular and longitudinal immunosuppression regimen data. These studies will be crucial to help provide the evidence to support consensus guidance on the optimal immunosuppression practices for patients transplanted with HCC.