Recipient and Center Factors Associated With Immunosuppression Practice Beyond the First Year After Liver Transplantation and Impact on Outcomes : Transplantation

Journal Logo

Original Clinical Science—Liver

Recipient and Center Factors Associated With Immunosuppression Practice Beyond the First Year After Liver Transplantation and Impact on Outcomes

Bittermann, Therese MD1,2; Lewis, James D. MD1,2; Goldberg, David S. MD3

Author Information
doi: 10.1097/TP.0000000000004209
  • Free
  • SDC
  • Infographic



Improvements in immunosuppression have made rejection an uncommon cause of allograft failure, lessening its significance as a critical endpoint in clinical and research settings.1,2 Concerns over immunosuppression-related side effects have instead come to the forefront, particularly given the changing demographics of liver transplant (LT) recipients. Although immunological tolerance may be induced in a small subset, this remains largely restricted to research settings.3 Thus, transplant providers must continually balance the benefits of immunosuppression with its inherent risks. The evidence that guides this decision-making is largely derived from small clinical trials with limited generalizability and short follow-up.4 Data on immunosuppression management beyond the first year after LT are particularly sparse, even though 1-y survival now exceeds 90%.5 In LT, immunosuppression management has traditionally focused on achieving calcineurin inhibitor (CNI) monotherapy; however, long-term outcomes with this approach are incompletely understood.6–9

Consensus recommendations on the optimal immunosuppression strategy for LT recipients have only recently become more robust.10 Nevertheless, transplant program preference remains a dominant factor in determining recipients’ immunosuppression early post-LT.11,12 These preferences may have indirectly resulted from participation in industry-funded studies. The increasing availability of large administrative databases offers an opportunity to evaluate the “real-world” effectiveness and safety of immunosuppression regimens in LT. Although the United Network for Organ Sharing (UNOS) routinely collects post-transplant data from US centers, the quality of UNOS immunosuppression data is unknown.

In this study, we investigate the patient and center factors associated with immunosuppression management in the first 5 y after LT among Medicare beneficiaries. Second, we longitudinally evaluate the impact of an immunosuppression regimen on patient and graft survival.


Development of the UNOS-Medicare Study Cohort

All adult LTs performed in the United States between January 1, 2007, and December 31, 2016, were identified in the UNOS database and then subsequently provided to the Centers for Medicare and Medicaid Services for extraction of the linked UNOS-Medicare study cohort. Medicare is federally funded health insurance covering people ≥65 y or younger individuals with disabilities or end-stage renal disease.13 It is the largest single-payer of transplant-related care in the US Medicare covers transplant immunosuppression medications for patients without out-of-pocket prescription coverage if they are enrolled in Medicare Part A (inpatient) at LT and in Part B (outpatient) at the time of the prescription filled.14 Multi-organ and prior transplant (liver or other) recipients were excluded, given inherent differences in immunosuppression practices. Study subjects were required to have ≥1 nonsteroid immunosuppression prescription filled in Medicare within the first 90-d following LT to ensure the cohort included patients using Medicare as their primary source of immunosuppression coverage. Of note, 98.8% of adult first LT alone recipients from 2007 to 2016 had a post-LT length of stay ≤90 d.

Exposures and Outcomes

Variables at LT obtained from UNOS included sex, age, race/ethnicity, liver disease, history of hepatocellular carcinoma (HCC), listing priority for acute liver failure (ALF), diabetes status, native Model for End-stage Liver Disease (MELD) score and individual components, albumin, and Karnofsky Performance Status. Perioperative induction immunosuppression was also obtained from UNOS and categorized as none, lymphocyte nondepleting (IL-2 receptor antagonists), and depleting (anti-thymocyte globulin).11

In the first analysis, immunosuppression was the outcome. Medicare claims for the following oral medications were assessed: (1) CNIs (tacrolimus and cyclosporine), (2) anti-metabolites (antiM; mycophenolic acid derivatives and azathioprine), (3) mechanistic target of rapamycin inhibitors (mTORi; sirolimus and everolimus), and (4) steroids (prednisone and budesonide). Common regimens were then defined: three-drug therapy with CNI+antiM+steroid, two-drug therapy with CNI+antiM or CNI+steroid, CNI monotherapy, and mTORi-based regimens. The remainders were classified as “other,” Immunosuppressive therapy was assessed on the anniversary (± 1 month) of the LT date for the first 5-y post-LT. CNI monotherapy was the primary outcome as this was felt to be the theoretical goal in most LT recipients.15–17

In the second part, the immunosuppression regimen was the exposure, measured as time-updating on a monthly basis. In developing the longitudinal immunosuppression analytic dataset, the last observation carried forward was used to fill gaps in exposure (ie, patients were assumed to remain on the same regimen until a change in regimen was observed). Patient and graft survival were the primary outcomes. Graft loss was defined as retransplantation or death.

Statistical Analysis

To assess its generalizability, the characteristics of the study cohort were compared with all other adults first LT alone recipients nationally during the study period. Chi-square and Kruskal–Wallis tests were used for categorical and continuous variables, respectively. When evaluating immunosuppression at each year post-LT, patients were censored at the end of follow-up, retransplantation, or death. Regimens obtained from Medicare drug claims were first described at each year (± 1 mo) post-LT over the first 5 y and regimen transitions over time were depicted using Sankey diagrams. Center practices in immunosuppression management at 1-, 3-, and 5-y post-LT were evaluated, restricting to those contributing ≥20 recipients to the cohort at each time point to reduce extreme values and increase generalizability.

Mixed-effects logistic regression assessed recipient factors associated with CNI monotherapy at 1-, 3-, and 5-y post-LT. Models were adjusted for sex, age, race/ethnicity, liver disease, history of HCC, creatinine at LT, bilirubin at LT, international normalized ratio at LT, diabetes, induction regimen, UNOS-reported history of acute rejection in the first post-LT year, and era (2012–2016 versus 2007–2011) as fixed effects. The transplant center was included as a random effect. In secondary analyses, factors associated with immunosuppression escalation, defined as transitioning from monotherapy to 2- or 3-drug therapy or from 2-drug to 3-drug therapy, and with mTORi-based treatment were evaluated.

Kaplan–Meier curves evaluated the association of maintenance immunosuppression (time-updating) with unadjusted patient and graft survival. Time zero was set at 1-y post-LT. Deaths and graft losses during the first year were thus excluded, given the goal of evaluating the relationship between long-term maintenance regimens and outcomes and also to allow models to be adjusted for early acute rejection. Given the study duration, the maximum follow-up time was 10 y from LT. Patients not meeting the endpoints of interest were censored at the last follow-up. The association of immunosuppression regimen (time-updating) on patient and graft survival was investigated using Cox proportional hazards models, adjusted for the aforementioned baseline covariates. In secondary analyses, interactions between regimen and the following covariates were explored: sex, race/ethnicity, age, liver disease, history of HCC, pre-LT diabetes, creatinine at LT, induction regimen, and acute rejection at 1 y. These were chosen based on biological plausibility and available evidence. Differences in UNOS cause of death (COD) by regimen were described. As an exploratory analysis, the estimated glomerular filtration rate (eGFR) in the first 3-y post-LT was calculated using creatinine values reported in UNOS and the 4-variable modification of diet in renal disease equation without race correction.18,19 Mixed-effects linear regression models assessed the association of regimen at 1 y and average eGFR between years 1–2 and years 2–3, accounting for clinical covariates including average eGFR between LT to year 1.

All primary analyses used Medicare drug claims exclusively. As a last study objective, the concordance between Medicare and UNOS regimens was described, among patients with nonmissing data in both sources. Unlike prescription claims, UNOS immunosuppression data are reported at discrete time points (approximately annually). UNOS immunosuppression regimen was evaluated over the first 5-y post-LT using a ± 1-mo and ± 3-mo interval from the LT anniversary date. UNOS immunosuppression data missingness was reported accounting for censoring: data were only considered missing if they should have been available (ie, patient alive with native allograft at measurement date).

This study was approved by the Institutional Review Board of the University of Pennsylvania. Analyses used Stata v17 (College Station, TX) and RStudio version 1.2.1335 (Boston, MA).


The linked UNOS-Medicare study cohort included 11 326 subjects, representing 23.1% of all adult first LT alone recipients nationally between January 1, 2007, and December 31, 2016 (Figure S1, SDC, There were 5161 (45.6%) patients alive with their native allograft at 5-y post-LT and 815 (7.2%) patients at 10-y post-LT. The UNOS-Medicare cohort was older (median 61 versus 57 y; P < 0.001) and had higher rates of HCC (30.9% versus 24.8%; P < 0.001) and diabetes (31.8% versus 23.2%; P < 0.001) compared with LT recipients nationally (Table 1).

TABLE 1. - Pre-LT and peri-LT patient characteristics in the baseline UNOS-Medicare cohort and remainder of adult recipients nationally between 2007 and 2016
UNOS-Medicare cohort a (N = 11 326) Remainder of adult LT recipients nationallya, b (N = 37 609) P Value
Male, N (%) 7250 (64.0) 25 582 (68.0) <0.001
Age at LT
 Median (IQR) 61 (54–66) 57 (50–63) <0.001
 Categorical, N (%)
  <50 1365 (12.1) 9770 (26.0) <0.001
  50-64 5464 (48.2) 25 008 (66.5)
  ≥65 4497 (39.7) 2831 (7.5)
Race/ethnicity, N (%) <0.001
 White 8143 (71.9) 26 968 (71.7)
 Black 916 (8.1) 3481 (9.3)
 Hispanic 1663 (14.7) 4868 (12.9)
 Asian 451 (4.0) 1803 (4.8)
 Other 153 (1.4) 489 (1.3)
Diagnosis, N (%) <0.001
 HCV 4760 (42.0) 14 838 (39.5)
 NASH 2350 (20.8) 5778 (15.4)
 Alcohol 1711 (15.1) 6492 (17.3)
 PSC 403 (3.6) 2005 (5.3)
 PBC 409 (3.6) 1165 (3.1)
 AIH 321 (2.8) 1349 (3.6)
 HBV 303 (2.7) 1243 (3.3)
 Other 1069 (9.4) 4739 (12.6)
HCC at LT, N (%) 3503 (30.9) 9309 (24.8) <0.001
Pre-LT diabetes, N (%) 3575 (31.8) 8665 (23.2) <0.001
Acute liver failure at LT, N (%) 97 (0.9) 1398 (3.7) <0.001
Lab MELD at LT, median (IQR) 18 (12–26) 22 (15–33) <0.001
Creatinine at LT (mg/dL), median (IQR) 1.1 (0.8–1.5) 1.2 (0.8–2.2) 0.513
Dialysis pre-LT, N (%) 767 (6.8) 3343 (8.9) <0.001
Albumin at LT (g/dL), median (IQR) 3.0 (2.6–3.5) 3.0 (2.5–3.5) 0.118
Pre-LT location, N (%) <0.001
 Home 8575 (75.7) 25 957 (69.0)
 Inpatient – ward 1783 (15.7) 6945 (18.5)
 Inpatient – ICU 968 (8.6) 4707 (12.5)
Karnofsky Performance Status at LT, <0.001
N (%) 2174 (19.2) 8873 (23.6)
 80%-100% 5281 (46.6) 14 855 (39.5)
 50%-70% 3742 (33.0) 13 334 (35.5)
 10%-30% 129 (1.1) 547 (1.5)
Induction immunosuppression receipt peri-LT c , N (%) 0.307
 None 8098 (71.5) 26 836 (71.4)
 Nondepleting induction 2021 (17.8) 6586 (17.5)
 Depleting induction 1207 (10.7) 4187 (11.1)
Post-LT length of stay (days), median (IQR) 9 (7–15) 13 (5–29) 0.015
aExcludes patients retransplanted and/or who died ≤90 from initial LT.
bSimilar to UNOS-Medicare cohort, this comparison group excludes prior organ (liver or other) and multi-organ transplant recipients.
cSteroids alone peri-LT was considered as not receiving induction therapy. Nondepleting induction included the interleukin-2 receptor antagonists daclizumab and basiliximab. Depleting induction included thymoglobulin, alemtuzumab, and rituximab.
HBV, hepatitis B virus; HCV, hepatitis C virus; IQR, interquartile range; LT, liver transplant; HCC, hepatocellular carcinoma; MELD, Model for End-stage Liver Disease; mTORi, mechanistic target of rapamycin inhibitors; NASH, nonalcoholic steatohepatitis; PBC, primary biliary cholangitis; PSC, primary sclerosing cholangitis; UNOS, United Network for Organ Sharing.

Immunosuppression Regimens in the First 5-y Post-LT

Half (51.9%) of patients received CNI monotherapy at 1-y post-LT, which increased to 68.6% at 5 y (Figure 1A). After CNI monotherapy and CNI+antiM, mTORi-based regimens were the third most commonly used in all years. Although statistically significant, differences in maintenance regimen at 1 y by induction regimen were small (P < 0.001; Figure 1B). Rates of individual drugs in the first 5-y post-LT are shown in Figure S2, SDC,

Immunosuppression regimens used in the first 5-y post-LT among patients in the UNOS-Medicare cohort. CNI, calcineurin inhibitor; mTORi, mechanistic target of rapamycin inhibitors; UNOS, United Network for Organ Sharing.

Patients on CNI monotherapy at 1 y largely kept this regimen at 3- and 5-y post-LT (87.6% and 85.9%, respectively; Figure 2). By 5 y, 55.7% of CNI+antiM and 44.2% of CNI+antiM+steroid at 1 y were down to CNI monotherapy. In contrast, immunosuppression regimen escalation was infrequent (6.0% between post-LT years 1 and 3). Half of the patients on CNI+antiM (50.4%) or on an mTORi-based regimen (55.1%) at 1 y kept the same regimen between 3-y and 5-y post-LT, whereas one-third (35.4%) of CNI+antiM+steroid patients did so.

Sankey diagram demonstrating changes in immunosuppression regimen in the first 5-y post-LT. Excludes patients with missing data; patients are censored at end of follow-up, retransplantation or death. CNI, calcineurin inhibitor; mTORi, mechanistic target of rapamycin inhibitors.

Minimal differences in regimen transitions were noted between patients with liver disease from viral hepatitis and those from alcohol or nonalcoholic steatohepatitis (NASH; Figures S3A-B, SDC, Transplants for autoimmune liver disease received multidrug therapy, including long-term steroid-based regimens, at higher rates throughout the 5-y time-window (Figure S3C, SDC, In HCC patients, mTORi-based regimens waned over time (13.7% at 1 y, 11.2% at 3 y, and 8.8% at 5 y; Figure S3D, SDC, ALF recipients received multidrug regimens at higher rates throughout the first 5-y post-LT (Figure S4, SDC,

Center Variability in Immunosuppression Practices

Center immunosuppression practices differed, though this was most pronounced early post-LT (Figure 3A–C). Center median CNI monotherapy use was 51.9% (interquartile range [IQR], 37.6%–63.5%) at year 1, 68.0% (IQR, 54.0%–74.7%) at year 3, and 69.6% (IQR, 57.7%–78.3%) at year 5 post-LT (Figure S5A-C, SDC, Center CNI+antiM use was ≥20% at 25% (N = 23/92) of centers at 3 y and 19.5% (N = 24/82) of centers at 5 y (Figure 3B-C). Center preference for mTORi-based regimens was variable, ranging from 0% to 38.8% at 1 y and 0% to 29.6% at 5 y (Table 2). There were largely no differences in center practices by center volume, with the exception of CNI+antiM+steroid at 1 y being more common among centers ≤75th percentile for average yearly volume (4.8% versus 2.9% for centers >75th percentile, P = 0.038; Table S1, SDC,

TABLE 2. - Immunosuppression practice variability among centers at years 1, 3, and 5 post-LT
Percent of recipients on regimen by center
Year 1(N = 98 centers) Year 3(N = 92 centers) Year 5(N = 82 centers)
CNI monotherapy
 Median 51.8 68.0 69.6
 IQR 37.6–63.5 54.0–74.7 57.7–78.3
 Range 20.0–79.7 24.5–100.0 15.4–95.2
CNI + antiM
 Median 21.9 13.5 12.3
 IQR 13.6–29.3 8.2–20.2 8.0–18.8
 Range 2.7–52.9 0–47.3 0–39.3
CNI + steroid
 Median 5.6 4.2 3.6
 IQR 3.2–8.1 2.5–6.5 2.1–6.5
 Range 0–35.0 0–19.6 0–15.9
CNI + antiM + steroid
 Median 4.3 2.2 2.5
 IQR 2.5–7.7 0–4.1 0–4.6
 Range 0–31.7 0–18.9 0–20.5
 Median 8.8 7.1 6.2
 IQR 5.1–13.2 3.5–10.8 2.3–9.1
 Range 0–38.8 0–38.5 0–29.6
Restricted to centers contributing ≥20 recipients to the United Network for Organ Sharing-Medicare cohort at each time point to avoid extreme values.
CNI, calcineurin inhibitor; IQR, interquartile range; LT, liver transplant; mTORi, mechanistic target of rapamycin inhibitors.

Center variation in immunosuppression management at 1-y (A), 3-y (B), and 5-y (C) post-LT. Each vertical bar includes one transplant center; figures are restricted to centers contributing ≥20 recipients to the cohort at each time point; centers are ranked by increasing rate of CNI monotherapy from left to right. CNI, calcineurin inhibitor; LT, liver transplant; mTORi, mechanistic target of rapamycin inhibitors.

Factors Associated with Immunosuppressive Therapy in the First 5-y Post-LT

At 1- and 3-y post-LT, Black race, history of alcohol-associated liver disease, NASH or autoimmune liver disease, and serum creatinine at LT were all independently associated with a lower likelihood of CNI monotherapy (Table 3). HCC recipients were less likely to receive CNI monotherapy at 1 y (OR, 0.80; 95% CI, 0.72-0.89; P < 0.001), but this effect waned over time (eg, P = 0.046 at 3 y and P = 0.524 at 5 y). Early acute rejection not only reduced the likelihood of CNI monotherapy at 1 y (aOR, 0.76; 95% CI, 0.66-0.87, P < 0.001) but also at 3 y (OR, 0.82; 95% CI, 0.69-0.98; P = 0.027).

TABLE 3. - Multivariable model evaluating predictors of CNI monotherapy in the first 5 y post-LT
Year 1N = 10 219 Year 3N = 7289 Year 5 a N = 4725
aOR(95% CI) P Value aOR(95% CI) P Value aOR(95% CI) P Value
Female 0.96(0.88-1.05) 0.378 1.02(0.91-1.15) 0.694 0.84(0.72-0.97) 0.037
Age at LT, per 5 y 1.07(1.05-1.10) <0.001 1.05(1.01-1.08) 0.004 1.01(0.97-1.05) 0.721
Race/ethnicity 0.140 0.019 0.153
 White Reference Reference Reference
 Black 0.83(0.71-0.98) 0.78(0.64-0.95) 0.78(0.60-1.01)
 Hispanic 0.94(0.82-1.07) 1.09(0.93-1.28) 0.82(0.67-1.00)
 Asian 0.95(0.76-1.18) 0.92(0.70-1.20) 0.90(0.65-1.25)
 Other 1.23(0.85-1.78) 1.63(0.97-2.74) 1.13(0.61-2.10)
Liver disease pathogenesis <0.001 <0.001 <0.001
 Viral Reference Reference Reference
 ALD 0.87(0.77-0.99) 0.74(0.63-0.88) 0.84(0.68-1.04)
 NASH 0.86(0.76-0.97) 0.72(0.61-0.84) 0.91(0.75-1.11)
 AI disease b 0.59(0.50-0.68) 0.44(0.37-0.54) 0.47(0.38-0.59)
 Other 0.77(0.66-0.90) 0.71(0.59-0.85) 0.75(0.60-0.95)
Pre-LT HCC 0.80(0.72-0.89) <0.001 0.87(0.76-0.99) 0.046 1.06(0.89-1.26) 0.524
Bilirubin at LT, per 1 mg/dL 1.00(0.99-1.00) 0.157 1.00(0.99-1.01) 0.832 1.00 (0.99-1.01) 0.491
Creatinine at LT, per 1 mg/dL 0.90(0.85-0.94) <0.001 0.86(0.81-0.92) <0.001 0.88(0.81-0.95) 0.002
INR, per 1 unit 0.93(0.87-0.99) 0.018 1.05(0.96-1.14) 0.220 1.03(0.95-1.18) 0.275
Pre-LT diabetes 1.00(0.91-1.10) 0.980 0.99(0.88-1.12) 0.873 1.03(0.88-1.20) 0.702
Induction 0.387 0.258 0.150
 None Reference Reference Reference
 NDI 0.99(0.86-1.13) 1.01(0.85-1.19) 1.24(1.00-1.53)
 DI 1.14(0.94-1.39) 1.23(0.96-1.58) 1.06(0.78-1.45)
Rejection in first year 0.76(0.66-0.87) <0.001 0.82(0.69-0.98) 0.027 0.93(0.74-1.16) 0.519
Era <0.001 <0.001 --- ---
 2007-2011 Reference Reference
 2012-2016 1.35(1.23-1.46) 1.98(1.77-2.20)
aYear 5 model not adjusted for era as all transplants occurred between 2007 and 2011 by default.
bIncludes primary biliary cholangitis, autoimmune hepatitis and primary sclerosing cholangitis.
AI, autoimmune; ALD, alcohol-associated liver disease; aOR, adjusted odds ratio; CI, confidence interval; DI, depleting; HCC, hepatocellular carcinoma; INR, international normalized ratio; LT, liver transplant; NASH, nonalcoholic steatohepatitis; NDI, non-depleting.

At 5-y post-LT, factors associated with reduced CNI monotherapy use included female sex (OR, 0.84; 95% CI, 0.72-0.97; P = 0.037), autoimmune liver disease (OR, 0.47; 95% CI, 0.38-0.59; P < 0.001 versus viral hepatitis), and creatinine at LT (OR, 0.88 per 1 mg/dL increase; 95% CI, 0.81-0.95; P = 0.002). Induction immunosuppression was not associated with CNI monotherapy at any time point (Table 3). Recipients between 2012 and 2016 were more likely to receive CNI monotherapy at 1 y (aOR, 1.35; 95% CI, 1.23-1.46; P < 0.001) and even more so at 3 y (aOR, 1.98; 95% CI, 1.77-2.20; P < 0.001) than LTs between 2007 and 2011.

In secondary analyses, Black race was associated with a 43% greater likelihood of immunosuppression escalation between years 1 and 3 (aOR, 1.43; versus White, 95% CI, 1.03-1.99; P = 0.031; Table S2, SDC Worse renal dysfunction at LT was also an important factor (aOR, of 1.12 per 1 mg/dL increase; 95% CI, 1.01-1.26; P = 0.039). Factors independently associated with mTORi-based treatment at 1-y post-LT included increasing age (aOR, 1.06 per 5-y increase, 95% CI, 1.02-1.11; P = 0.006), HCC (aOR, 1.51; 95% CI, 1.28-1.79; P < 0.001), increasing creatinine at LT (aOR 1.21 per 1 mg/dL increase, 95% CI: 1.12-1.32; P < 0.001) and history of acute rejection in the first year post-LT (aOR 1.30, 95% CI: 1.04-1.63; P = 0.020; Table S3

Association of Immunosuppression Regimen With Patient and Graft Survival

Maintenance immunosuppression was associated with unadjusted patient and graft survival (P < 0.001 for both; Figure S6, SDC, Differences in primary COD according to regimen at 1 y (P = 0.002), 3 y (P = 0.155), and 5 y (P = 0.755) post-LT were small (Figure S7, SDC, In covariate-adjusted models, CNI+antiM was associated with improved patient survival: hazard ratio (HR), 0.59 (95% CI, 0.51-0.69) versus CNI alone (P < 0.001; Table 4). Worse survival was observed with both CNI+steroid and mTORi-based regimens: covariate-adjusted HR, 1.76 (95% CI, 1.49-2.09) and 1.43 (95% CI, 1.25-1.64). The addition of steroids to CNI+antiM attenuated the improvement in patient survival obtained with CNI+antiM (adjusted HR, 1.03; 95% CI, 0.78-1.35). Similar trends were observed for graft survival, with CNI+antiM achieving the most favorable outcome: covariate-adjusted HR, 0.62 versus CNI alone (95% CI, 0.53-0.71; P < 0.001; Table 4).

TABLE 4. - Association of immunosuppression regimen with patient and graft survival beyond year 1 post-LTa
Patient survival, covariate-adjusted b HR (95% CI)
(N = 10 409) P Value Graft survival, covariate-adjustedHR (95% CI)(N = 10 409) P Value
CNI alone Reference <0.001 Reference <0.001
CNI + antiM 0.59 (0.51-0.69) 0.62 (0.53-0.71)
CNI + steroid 1.76 (1.49-2.09) 1.81 (1.54-2.14)
CNI + antiM + steroid 1.11 (0.85-1.47) 1.32 (1.03-1.69)
mTORi-based 1.43 (1.25-1.64) 1.46 (1.27-1.66)
In all years and for all regimens the most frequent diagnosis was hepatitis C virus.
aTime zero set at y from LT date, maximum follow-up 10-y post-LT.
bModel adjusted for sex, race/ethnicity, age, liver disease pathogenesis, history of hepatocellular carcinoma, pre-LT diabetes, creatinine at LT, bilirubin at LT, international normalized ratio at LT, induction regimen, acute rejection during year 1 post-LT, transplant era.
CI, confidence interval; CNI, calcineurin inhibitor; HR, hazard ratio; LT, liver transplant.

The interaction of regimen and age was statistically significant in the graft survival analysis and nonignorable in the patient survival analysis (P = 0.010 and P = 0.109, respectively). In both models, the survival benefit of CNI+antiM strengthened with increasing age and was greatest for recipients aged ≥65 y (eg, HR, 0.52; 95% CI, 0.49-0.67; P < 0.001 versus CNI monotherapy for patient survival; Table S4, SDC, In contrast, the deleterious effect of the CNI+steroid regimen on outcomes was worst in recipients aged <50 y (eg, HR, 2.17; 95% CI, 1.45-3.23; P < 0.001 versus CNI monotherapy for patient survival). These stratified models accounted for LT indication, and the distribution of diagnosis by age group is shown in Table S5, SDC, None of the other interactions studied were statistically significant (all P > 0.2). For example, there was no interaction found between regimen and diagnosis (P = 0.293 for patient survival and P = 0.411 for graft survival), indicating that our findings (Table 4) applied irrespective of patients’ underlying liver disease.

Differences in eGFR in the first 3 y post-LT according to the regimen at 1 y were not clinically meaningful, though UNOS creatinine data were limited (Tables S6 and S7, SDC, In adjusted analyses, there were no association between regimen at 1 y and adjusted eGFR between years 1–2 (P = 0.186) and years 2–3 (P = 0.175).

Assessing the Validity of UNOS Follow-up Immunosuppression Data

Data missingness for Medicare drug claims was low, ranging 2.7%–3.1% during the first 5 y. In contrast, missingness in UNOS data was 29.6% at 1-y post-LT and reached 79.8% at 5-y post-LT (Figure 4A). UNOS and Medicare regimen concordance were overall low, though improved over time from 56.4% at 1-y to 70.4% at 5-y post-LT (Figure 4B). The most frequent discordance was evidence of CNI monotherapy according to Medicare drug claims versus report of multidrug therapy in UNOS (most commonly CNI+antiM). Discordance in multidrug regimens was also frequent, accounting for 22.6%–28.3% of discordances.

A, Frequency of patients with missing immunosuppression data in UNOS database during the first 5-y post-LT in the UNOS-Medicare study cohort. B, Frequency of discordant UNOS-Medicare regimens and their underlying reasons. Reasons for discordance (and percent reported) are among the subset of patients with discordant regimens as shown in (B), which is reflected in the smaller subgroup sizes. CNI, calcineurin inhibitor; LT, liver transplant; UNOS, United Network for Organ Sharing.


Immunosuppression is the cornerstone of post-transplant management, yet there is surprisingly little evidence on optimal practices beyond the first post-LT year. In a recent meta-analysis, only 2.3% of immunosuppression-related clinical trials studied an intervention occurring ≥1-y post-transplant.4 Moreover, Organ Procurement and Transplantation Network (OPTN) annual reports focus solely on practices at LT hospital discharge and 1-y post-transplant.5,20–23 Yet the vast majority of LT recipients survive the first year and outcomes have continued to improve over time.5 This is the first study to explore this critical aspect of post-transplant care using a novel linked database of national transplant registry data and Medicare drug claims. Our study highlights the heterogeneity in treatment practices among transplant centers, particularly in the first post-LT year. Moreover, several key recipient factors such as race/ethnicity and pathogenesis of liver disease influence treatment choice. Lastly, we found that the combination of CNI+antiM was associated with improved long-term patient and graft survival over CNI monotherapy among Medicare beneficiaries, an effect that was further strengthened by increasing age. These findings build upon the existing evidence supporting this strategy in other specific situations (eg, renal dysfunction).10 As the transplant community aims to individualize the approach to immunosuppression after LT, our study provides critical foundational work to understand the current practices and their outcomes in a large contemporary cohort.

From a methods perspective, this is the first study to use up to 10 y of monthly updating, longitudinal data to explore the benefits and harms of maintenance immunosuppression practices in a large recipient population. We specifically focus on long-term outcomes, given the ongoing unmet need in the existing literature.2 Aiming for immunosuppression minimization, CNI monotherapy has traditionally been the primary objective in LT recipients.6–8,24 However, our results suggest that long-term combination therapy with CNI+antiM may offer key benefits, particularly among older recipients, that should be further validated prospectively. Though we did not find a clear signal in our exploratory analysis using the limited UNOS creatinine data available, we hypothesize that improvement in renal function from CNI minimization is one potential mechanism.10 In addition, lower CNI exposure has also been shown to reduce post-transplant malignancy risk, which may explain why CNI monotherapy was less commonly employed in recipients with HCC in our cohort.25,26 Given that the dominant reason for graft loss in our cohort was patient death, these reasons likely explain the improvement in graft survival found. Interestingly, similar long-term benefits of using CNI+antiM at LT hospital discharge were also observed in a prior study that did not account for regimen changes over time.27 Future efforts will warrant more comprehensive laboratory data to definitively determine the mechanisms by which combination therapy leads to improved outcomes. In addition, our finding of the decreased patient and graft survival with CNI+steroid immunosuppression, particularly in younger patients, should also be further investigated to elucidate possible causal pathways (eg, recurrence of autoimmune disease and toxicity from long-term corticosteroids).

In our study, Black recipients were less likely to receive CNI monotherapy at 1- and 3-y post-LT. We hypothesize that this may result from their actual or perceived risk of progressive renal dysfunction, which may specifically benefit from CNI minimization through combination therapy, though this should be further studied.10,28 Differences in the pharmacokinetics and immunosuppressive effects of key transplant drugs in African Americans could have also contributed to these findings.29–32 In our longitudinal analysis of patient and graft survival, we did not identify an interaction between regimen and race, suggesting that the benefit of CNI+antiM was uniform among subgroups. Regarding socioeconomic factors, because Medicare frequently covers LT immunosuppression without a supplemental out-of-pocket plan, it is possible that the benefits of multidrug immunosuppression we observed are not generalizable to patients with financial barriers to medication adherence.33,34 This is particularly relevant to the cost difference between antiM drugs, such that the benefits afforded by azathioprine versus mycophenolic acid may not be equal.35

UNOS immunosuppression data are the same as those published in annual OPTN reports, which therefore reflect their limitations.5,20–23 We show that UNOS data are insufficient alone for pharmacoepidemiology studies in transplant recipients. However, UNOS data contain details not included in claims data (eg, laboratory parameters). As such, the linkage of UNOS and other data sources (eg, other registries and healthcare claims data) creates a uniquely detailed resource for pharmacoepidemiology research.12,36 Importantly, we observed similar trends in the use of individual drugs at 1-y post-LT in our linked cohort as described in OPTN reports.20,22 In determining the optimal immunosuppression strategy post-LT, data from multiple types of studies should be taken into consideration (eg, clinical trials, meta-analyses and large observational cohorts). As highlighted in the case of induction immunosuppression in LT, for example, each highlight different perspectives and limitations.11,37,38 The inclusion of UNOS data also permits a broader evaluation of differences in center practices, which naturally also facilitates investigation into the nonpatient factors that drive immunosuppression decision-making. For example, future studies should explore the role of provider type, subspecialty, and experience on regimen choices and outcomes.

This study had several limitations. The UNOS-Medicare cohort included a greater proportion of older recipients, which also led to greater rates of pre-LT NASH, HCC, and diabetes. Given that metabolic risk factors, renal dysfunction, and malignancy risk may influence immunosuppression decision-making, the management practices we identified may not mirror those of all recipients. For example, our finding of improved post-LT outcomes using CNI+antiM versus CNI monotherapy may not be generalizable to younger patients at large, particularly since Medicare eligibility <65 y implies disability before transplant. Similarly, although we found worse outcomes with a CNI+steroid approach in young patients, only a small subset of these had autoimmune liver disease. Thus, these observations should be validated prospectively and/or using alternate pharmacy claims data.

We were also not able to investigate drug dose or trough level with these data. Thus, it was not possible to confirm whether the addition of antiM to CNI occurred in the setting of CNI minimization versus increased immunosuppression. For example, our observation of increased multidrug immunosuppression among ALF recipients may reflect their increased likelihood of underlying immune-mediated liver disease.39 Moreover, similar to the development of tolerance, the immunologic effects of multidrug therapy are likely also heterogeneous.3 We attempted to evaluate the relevance of post-LT renal function to our results; however, UNOS-reported laboratory parameters post-LT were too limited to draw definitive conclusions. We also explored differences in COD by regimen, but these efforts were again limited by incomplete data. Reliably establishing the underlying COD requires a systematic approach to avoid underestimating the contribution of key post-transplant complications.40,41 These aspects should be further investigated prospectively and/or using alternate data sources to better understand the mechanisms by which immunosuppression regimens may alter post-LT outcomes.

As another limitation, brief gaps in drug fills were observed, which is why we allowed flexibility in measurement. However, this may have led to recently discontinued therapies being captured as actively prescribed, though it is anticipated that this bias would be nondifferential by regimen. Given the study’s duration, a 5-y regimen could only be assessed in LTs performed between 2006 and 2011. Immunosuppression practices may have since changed, particularly given the recent availability of effective direct-acting antiviral treatment for hepatitis C virus and the subsequent increase in LTs for alcohol-associated liver disease and NASH. Lastly, when considering center practice differences, it should be acknowledged that the study population included only a subset of recipients at each center.

This first analysis of “real world” immunosuppression practices among Medicare beneficiaries addresses an important evidence gap on the optimal long-term management after LT. Our results indicate that combination therapy with CNI+antiM may be a preferred strategy in a broad group of recipients. We also highlight the heterogeneity in practices both among recipients and centers, emphasizing the ongoing need for further studies and greater guidance on the ideal approach to minimizing immunosuppression in a manner that individualizes recipients’ risks and benefits. We additionally demonstrate the value of data linkages to augment the capture of longitudinal changes in management and outcomes beyond the early postoperative period. These findings help create a road map for future lines of investigation in this critical aspect of post-LT care.


1. Neuberger J, Adams DH. What is the significance of acute liver allograft rejection? J Hepatol. 1998;29:143–150.
2. Lerut JP, Gondolesi GE. Immunosuppression in liver and intestinal transplantation. Best Pract Res Clin Gastroenterol. 2021;54-55:101767.
3. Starzl TE, Murase N, Abu-Elmagd K, et al. Tolerogenic immunosuppression for organ transplantation. Lancet. 2003;361:1502–1510.
4. Rodríguez-Perálvarez M, Guerrero-Misas M, Thorburn D, et al. Maintenance immunosuppression for adults undergoing liver transplantation: a network meta-analysis. Cochrane Database Syst Rev. 2017;3:CD011639.
5. Kwong AJ, Kim WR, Lake JR, et al. OPTN/SRTR 2019 annual data report: liver. Am J Transplant. 2021;21:208–315.
6. Lerut J, Mathys J, Verbaandert C, et al. Tacrolimus monotherapy in liver transplantation: one-year results of a prospective, randomized, double-blind, placebo-controlled study. Ann Surg. 2008;248:956–967.
7. Moench C, Barreiros AP, Schuchmann M, et al. Tacrolimus monotherapy without steroids after liver transplantation–a prospective randomized double-blinded placebo-controlled trial. Am J Transplant. 2007;7:1616–1623.
8. Lan X, Liu MG, Chen HX, et al. Efficacy of immunosuppression monotherapy after liver transplantation: a meta-analysis. World J Gastroenterol. 2014;20:12330–12340.
9. Weiler N, Thrun I, Hoppe-Lotichius M, et al. Early steroid-free immunosuppression with FK506 after liver transplantation: long-term results of a prospectively randomized double-blinded trial. Transplantation. 2010;90:1562–1566.
10. Charlton M, Levitsky J, Aqel B, et al. International liver transplantation society consensus statement on immunosuppression in liver transplant recipients. Transplantation. 2018;102:727–743.
11. Bittermann T, Hubbard RA, Lewis JD, et al. The use of induction therapy in liver transplantation is highly variable and is associated with posttransplant outcomes. Am J Transplant. 2019;19:3319–3327.
12. Nazzal M, Lentine KL, Naik AS, et al. Center-driven and clinically driven variation in US liver transplant maintenance immunosuppression therapy: a national practice patterns analysis. Transplant Direct. 2018;4:e364.
13. Centers for Medicare & Medicaid Services. What’s Medicare? Available at Accessed April 26, 2022.
14. Centers for Medicare & Medicaid Services. Local Coverage Article: Immunosuppressive Drugs - Policy Article (A52474). Available at Accessed September 1, 2021.
15. Jain AB, Kashyap R, Rakela J, et al. Primary adult liver transplantation under tacrolimus: more than 90 months actual follow-up survival and adverse events. Liver Transpl Surg. 1999;5:144–150.
16. Margarit C, Rimola A, Gonzalez-Pinto I, et al. Efficacy and safety of oral low-dose tacrolimus treatment in liver transplantation. Transpl Int. 1998;11:S260–S266.
17. Raimondo ML, Burroughs AK. Single-agent immunosuppression after liver transplantation: what is possible? Drugs. 2002;62:1587–1597.
18. Levey AS, Coresh J, Greene T, et al.; Chronic Kidney Disease Epidemiology Collaboration. Using standardized serum creatinine values in the modification of diet in renal disease study equation for estimating glomerular filtration rate. Ann Intern Med. 2006;145:247–254.
19. Yap E, Prysyazhnyuk Y, Ouyang J, et al. The implication of dropping race from the MDRD equation to estimate GFR in an African American-only cohort. Int J Nephrol. 2021;2021:1880499.
20. Kim WR, Lake JR, Smith JM, et al. OPTN/SRTR 2016 annual data report: liver. Am J Transplant. 2018;18:172–253.
21. Kim WR, Lake JR, Smith JM, et al. OPTN/SRTR 2017 annual data report: liver. Am J Transplant. 2019;19:184–283.
22. Kim WR, Lake JR, Smith JM, et al. OPTN/SRTR 2015 annual data report: liver. Am J Transplant. 2017;17:174–251.
23. Kwong A, Kim WR, Lake JR, et al. OPTN/SRTR 2018 annual data report: liver. Am J Transplant. 2020;20:193–299.
24. Lerut JP, Pinheiro RS, Lai Q, et al. Is minimal, [almost] steroid-free immunosuppression a safe approach in adult liver transplantation? Long-term outcome of a prospective, double blind, placebo-controlled, randomized, investigator-driven study. Ann Surg. 2014;260:886–891; discussion 891.
25. Rodríguez-Perálvarez M, Colmenero J, González A, et al. Cumulative exposure to tacrolimus and incidence of cancer after liver transplantation. Am J Transplant. 2022;22:1671–1682.
26. Carenco C, Assenat E, Faure S, et al. Tacrolimus and the risk of solid cancers after liver transplant: a dose effect relationship. Am J Transplant. 2015;15:678–686.
27. Wiesner RH, Shorr JS, Steffen BJ, et al. Mycophenolate mofetil combination therapy improves long-term outcomes after liver transplantation in patients with and without hepatitis C. Liver Transpl. 2005;11:750–759.
28. Alvarado M, Schaubel DE, Reddy KR, et al. Black race is associated with higher rates of early-onset end-stage renal disease and increased mortality following liver transplantation. Liver Transpl. 2021;27:1154–1164.
29. Nagashima N, Watanabe T, Nakamura M, et al. Decreased effect of immunosuppression on immunocompetence in African–Americans after kidney and liver transplantation. Clin Transplant. 2001;15:111–115.
30. Dirks NL, Huth B, Yates CR, et al. Pharmacokinetics of immunosuppressants: a perspective on ethnic differences. Int J Clin Pharmacol Ther. 2004;42:701–718.
31. Jacobson PA, Oetting WS, Brearley AM, et al.; DeKAF Investigators. Novel polymorphisms associated with tacrolimus trough concentrations: results from a multicenter kidney transplant consortium. Transplantation. 2011;91:300–308.
32. Taber DJ, Gebregziabher MG, Srinivas TR, et al. African-American race modifies the influence of tacrolimus concentrations on acute rejection and toxicity in kidney transplant recipients. Pharmacotherapy. 2015;35:569–577.
33. Serper M, Reese PP, Patzer RR, et al. The prevalence, risk factors, and outcomes of medication trade-offs in kidney and liver transplant recipients: a pilot study. Transpl Int. 2018;31:870–879.
34. Moayed MS, Khatiban M, Nassiri Toosi M, et al. Barriers to adherence to medical care programs in liver transplant recipients: a qualitative study. Int J Organ Transplant Med. 2019;10:115–126.
35. Germani G, Pleguezuelo M, Villamil F, et al. Azathioprine in liver transplantation: a reevaluation of its use and a comparison with mycophenolate mofetil. Am J Transplant. 2009;9:1725–1731.
36. Mahmud N, Goldberg DS, Bittermann T. Best practices in large database clinical epidemiology research in hepatology: barriers and opportunities. Liver Transpl. 2022;28:113–122.
37. Penninga L, Wettergren A, Wilson CH, et al. Antibody induction versus placebo, no induction, or another type of antibody induction for liver transplant recipients. Cochrane Database Syst Rev. 2014;2014:Cd010253.
38. Penninga L, Wettergren A, Wilson CH, et al. Antibody induction versus corticosteroid induction for liver transplant recipients. Cochrane Database Syst Rev. 2014;31:Cd010252.
39. Germani G, Theocharidou E, Adam R, et al. Liver transplantation for acute liver failure in Europe: outcomes over 20 years from the ELTR database. J Hepatol. 2012;57:288–296.
40. Navez J, Iesari S, Kourta D, et al. The real incidence of biliary tract complications after adult liver transplantation: the role of the prospective routine use of cholangiography during post-transplant follow-up. Transpl Int. 2021;34:245–258.
41. Wareham NE, Da Cunha-Bang C, Borges ÁH, et al. Classification of death causes after transplantation (CLASS): evaluation of methodology and initial results. Medicine (Baltimore). 2018;97:e11564.

Supplemental Digital Content

Copyright © 2022 Wolters Kluwer Health, Inc. All rights reserved.