Efficacy and Safety of Immunosuppression Withdrawal in Pediatric Liver Transplant Recipients: Moving Toward Personalized Management

Abstract

Background and aims: Tolerance is transplantation's holy grail, as it denotes allograft health without immunosuppression and its toxicities. Our aim was to determine, among stable long-term pediatric liver transplant recipients, the efficacy and safety of immunosuppression withdrawal to identify operational tolerance.

Approach and results: We conducted a multicenter, single-arm trial of immunosuppression withdrawal over 36-48 weeks. Liver tests were monitored biweekly (year 1), monthly (year 2), and bimonthly (years 3-4). For-cause biopsies were done at investigators' discretion but mandated when alanine aminotransferase or gamma glutamyltransferase exceeded 100 U/L. All subjects underwent final liver biopsy at trial end. The primary efficacy endpoint was operational tolerance, defined by strict biochemical and histological criteria 1 year after stopping immunosuppression. Among 88 subjects (median age 11 years; 39 boys; 57 deceased donor grafts), 33 (37.5%; 95% confidence interval [CI] 27.4%, 48.5%) were operationally tolerant, 16 were nontolerant by histology (met biochemical but failed histological criteria), and 39 were nontolerant by rejection. Rejection, predicted by subtle liver inflammation in trial entry biopsies, typically (n = 32) occurred at ≤32% of the trial-entry immunosuppression dose and was treated with corticosteroids (n = 32) and/or tacrolimus (n = 38) with resolution (liver tests within 1.5 times the baseline) for all but 1 subject. No death, graft loss, or chronic, severe, or refractory rejection occurred. Neither fibrosis stage nor the expression level of a rejection gene set increased over 4 years for either tolerant or nontolerant subjects.

Conclusions: Immunosuppression withdrawal showed that 37.5% of selected pediatric liver-transplant recipients were operationally tolerant. Allograft histology did not deteriorate for either tolerant or nontolerant subjects. The timing and reversibility of failed withdrawal justifies future trials exploring the efficacy, safety, and potential benefits of immunosuppression minimization.

Trial registration: ClinicalTrials.gov NCT01638559.

FIGURE 1:. iWITH enrollment diagram

Beginning with 1,178 potentially eligible recipients, 1,090 patients were sequentially excluded, resulting in 88 subjects who were fully eligible and initiated ISW.

FIGURE 2:. iWITH primary endpoint: the outcome of immunosuppression withdrawal

Eighty-eight subjects initiated immunosuppression withdrawal in 7 steps according to a protocol-specified algorithm provided in Figure S2. A. Thirty-three subjects met biochemical and histological criteria for operational tolerance. Fifty-five subjects were non-tolerant: 39 failed secondary to rejection; 16 failed secondary to histological findings although they met biochemical criteria. Three subjects did not complete trial participation, secondary to being lost to follow-up (operationally tolerant subject, after tolerance adjudication), withdrawal of consent and refusal to travel for the end-of-trial (year 4) biopsy (both subjects who rejected). B. Among the 55 non-tolerant subjects, 35 rejected prior to tolerance adjudication. The remaining 20 subjects were determined to be non-tolerant based on adjudication biopsy findings: 16 subjects failed to meet histologic criteria of operational tolerance (Table S1) and 4 subjects met biopsy criteria for rejection despite stable relative to baseline serum ALT and GGT levels. We classified 39 subjects as nontolerant by rejection, 35 before the tolerance adjudication biopsy and 4 at the time of the adjudication biopsy.

FIGURE 3:. Data regarding 33 tolerant subjects

A. Trial entry, peak, and end-of-trial ALT and GGT values [mean (interquartile range; IQR)] for operationally tolerant subjects; ALT and GGT levels over time are shown in Figure S3A. B. Change in key features of the final (year 4) compared to the baseline (year 0) biopsy for operationally tolerant subjects; changes in additional biopsy features are presented in Figure S5A. Each row represents a single subject. In both figures, subjects are presented in the same order, sorted by change in portal inflammation and then subject identification number. To calculate change over time, absolute scores at year 0 were subtracted from scores at year 4 for the following parameters: portal inflammation, portal, sinusoidal, and perivenular fibrosis and the LAFSc. All score scales ranged from 0 to 3 except the LAFSc scale which ranged from 0 to 9 (25). Pink indicates progression while green indicates regression; increasing intensity of either pink or green indicates larger magnitude of change. Gray indicates missing data; one operationally tolerant subject was lost to follow-up after 3 years of trial participation. C. The TCMR probability is plotted for protocol-driven liver biopsies collected at 3 timepoints (yr 0: trial entry; yr 2: tolerance adjudication; yr 4: end-of-trial). The transcriptional probability of rejection was calculated based on the expression levels of genes in the TCMR signature. The dotted line corresponds to the optimal probability threshold to identify biopsies diagnostic of acute rejection. P-values correspond to an unpaired Mann-Whitney test.

FIGURE 4:. Data regarding 39 non-tolerant by rejection subjects

A. Timing of rejection episodes. The time of rejection for each subject diagnosed with rejection is represented by a bar. Bar segments represent ISW steps (Figure S2); segment length represents step duration. Time of rejection diagnosis is marked by a circle for biopsy-proven acute rejection (n=37), based on central pathology assessment according to Banff criteria (22) or by a star for clinical rejection (n=2), defined by the trial protocol (Supplementary Appendix) as elevated liver tests treated with increased or re-initiation of immunosuppression but without biopsy confirmation. Rejection occurred during withdrawal for 33 subjects and after stopping immunosuppression for 6 subjects. Of these 6, 4 subjects with tolerance (Table S1) and 4 subjects met biopsy criteria for rejection despite stable relative to baseline liver tests were diagnosed with biopsy-proven acute rejection based on the tolerance adjudication biopsy and are noted with an asterisk. B. Trial entry, peak, and end-of-trial ALT and GGT values [mean (IQR)] for non-tolerant by rejection subjects; ALT and GGT levels over time are shown in Figure S3B. C. Time to resolution of rejection for those with elevated liver tests are shown (n=35); 4 subjects with stable relative to baseline ALT and GGT values but biopsy-proven acute rejection at the tolerance adjudication biopsy are excluded. Two definitions for resolution are presented: i) ALT and GGT values ≤1.5X baseline as defined in the trial protocol (black); one unresolved episode is censored (O) at the end of the trial; ii) ALT and GGT values <50 units per milliliter (gray). D. Immunosuppression exposure over the 4-year trial is shown for those who were non-tolerant by rejection and on tacrolimus (n=38); one subject that converted to azathioprine monotherapy was excluded. Expected exposure (X axis) was calculated assuming that the subject was maintained on the dose at trial entry and plotted against actual exposure (Y axis). Pink circles (n=19) identify subjects with higher actual than expected exposure while green circles (n=19) identify subjects with lower actual than expected exposures. Color intensity increases with larger differences between actual and expected exposures. E. Change in key features of the final (year 4) compared to the baseline (year 0) biopsy for non-tolerant by rejection subjects; changes in additional biopsy features are presented in Figure S5B. Each row represents a single subject. In both figures, subjects are presented in the same order, sorted by change in portal inflammation and then subject identification number. To calculate change over time, absolute scores at year 0 were subtracted from scores at year 4 for the following parameters: portal inflammation, portal, sinusoidal, and perivenular fibrosis and the LAFSc. All score scales ranged from 0 to 3 except the LAFSc scale which ranged from 0 to 9 (25). Pink indicates progression while green indicates regression; increasing intensity of either pink or green indicates larger magnitude of change. Gray indicates missing data. Two subjects did not complete trial participation one withdrew assent/consent after 3 years; the other refused to travel for the end of trial biopsy. F. The TCMR probability is plotted for liver biopsies collected at 3 timepoints (yr 0: trial entry; rej: time of rejection diagnosis; yr 4: end-of-trial). The transcriptional probability of rejection was calculated based on the expression levels of genes in the TCMR signature. The dotted line corresponds to the optimal probability threshold to identify biopsies diagnostic of acute rejection. P-values correspond to an unpaired Mann-Whitney test.

FIGURE 5:. Data regarding 16 non-tolerant by histology subjects

A. Changes in the specific histological features utilized to adjudicate operational tolerance (Table S1); each row represents a single subject. The upper 8 rows represent subjects who were kept off immunosuppression; the lower 8 rows represent subjects who were restarted on immunosuppression as a result of the tolerance adjudication biopsy. Two subjects, identified by carets, were re-initiated on immunosuppression prior to the end of the trial. To calculate change over time, absolute scores at year 0 were subtracted from those at year 2 for the following: 3 parameters of inflammation (portal, interface, and perivenular), 2 parameters of fibrosis (Ishak and perivenular), bile duct damage, and isolated arteriopathy. All score scales ranged from 0 to 3 except Ishak fibrosis stage which ranged from 0 to 6 (22). Pink indicates progression and green indicates regression; increasing intensity indicates larger magnitude of change. Gray indicates missing data. All except one subject failed the primary endpoint due to new onset necro-inflammatory-type interface activity with or without other disqualifying features, such as increase in fibrosis stage of 2 or new onset isolated arteriopathy. B. Trial entry, peak, and end-of-trial ALT and GGT values [mean (IQR)] for non-tolerant by histology subjects; ALT and GGT levels over time are shown in Figure S3B. C. Immunosuppression exposure over the 4-year trial for those who were non-tolerant by histology and on tacrolimus (n=13); 3 subjects on cyclosporine were excluded. Expected exposure (X axis) was calculated assuming that the subject was maintained on the dose at trial entry and plotted against actual exposure (Y axis). The pink circle identifies the single subject with higher actual than expected exposure. The remaining subjects (n=12) with lower actual than expected exposures are identified by green symbols: green circles (n=5) identify subjects who resumed tacrolimus and green stars (n=7) identify subjects who remained off immunosuppression after the tolerance adjudication biopsy (year 2). Color intensity increases with larger differences between actual and expected exposures. D. Change in key features of the final (year 4) compared to the baseline (year 0) biopsy for non-tolerant by histology subjects; changes in additional biopsy features are presented in Figure S5C. Each row represents a single subject. The upper 8 rows represent subjects who were kept off immunosuppression, while the lower 8 rows represent subjects who were restarted on immunosuppression as a result of the biopsy. In both figures, subjects in the groups of 8 are presented in the same order, sorted by change in portal inflammation and then subject identification number. Two subjects, identified by carets, were re-initiated on immunosuppression prior to the end of the trial (Supplementary Methods). To calculate change over time, absolute scores at year 0 were subtracted from scores at year 4 for the following parameters: portal inflammation, portal, sinusoidal, and perivenular fibrosis and the LAFSc. All score scales ranged from 0 to 3 except the LAFSc scale which ranged from 0 to 9(25). Pink indicates progression and green indicates regression; increasing intensity indicates larger magnitude of change. Gray indicates missing data. One subject did not undergo the end of trial biopsy secondary to for-cause within the preceding 6 months. E. The TCMR probability for protocol-driven liver biopsies collected at 3 timepoints (yr 0: trial entry; yr 2: tolerance adjudication; yr 4: end-of-trial). The transcriptional probability of rejection was calculated based on the expression levels of genes in the TCMR signature. The dotted line corresponds to the optimal probability threshold to identify biopsies diagnostic of acute rejection. P-values correspond to an unpaired Mann-Whitney test.

FIGURE 6:. Factors associated with operational tolerance

A. Multiplex immunohistochemistry parameters of the eligibility biopsy separate tolerant from non-tolerant subjects. Shown is a 3-dimensional scatter plot of tolerant (green circles; n=18) and non-tolerant (red squares; n=35) subjects according to the number of CD8+ cells per mm² (T effector cells; X axis), lobular CD45+/MHCII+ pairs per mm² (leukocyte/antigen-presenting cell pairs; Y axis), and MAC387+ cells per mm² (infiltrating macrophages; Z axis) in the eligibility biopsy. The inner cube identifies thresholds that, simultaneously, maximizes the number of tolerant subjects (17 of 18; 94%) and minimizes the number of non-tolerant subjects (12 of 35; 34%). Subjects within the inner cube are closed symbols; those outside are open symbols. Plots only show subjects for which values of all 3 parameters were available. B. Eligibility biopsies with comparable portal and lobular inflammation grade but different immunohistochemical inflammatory loads. Hematoxylin and eosin sections are shown in the top row while corresponding immunostained sections [CD34 (green) /CD45 (teal) /MHCII (red)] are shown in the bottom row. The left column is the eligibility biopsy from an operationally tolerant subject while the right column is from a non-tolerant subject. Using a scale from 0 to 3, both biopsies were graded as 0 for both portal and lobular inflammation. Algorithmically detected pairings of leukocytes (CD45+) and antigen-presenting cells (MHCII+), shown in high magnification in the inset, are highlighted in yellow circles in the immunostained sections. The number of pairings was 7.6 per mm² for the tolerant (lower left) and 15.3 per mm² for the non-tolerant (lower right) subject. C. Class II DSA presence during ISW (year 0 to 1) is shown in 3 heatmaps. A minimum mean fluorescence intensity (MFI) threshold of 1,000 was used to identify a positive class II DSA. Heat maps show the maximum MFI for class II DSA with a range from 1,000 to 20,000; white indicates missing data. Immunosuppression was reduced stepwise according to a protocol-specified algorithm (Figure S2). Class II DSA was determined at baseline (year 0), weeks 12, 24, 36, and year 1 as long as subjects continued to withdraw immunosuppression. After diagnosis of rejection, subjects were not tested for class II DSA until the year 1 visit. Hence, non-tolerant subjects have a high frequency of missing data, particularly at the week 24 and 36 timepoints. Subjects are divided into those who did not show any DSA during year 1 (n=22), those who have detectible DSA at trial entry (n=44), and those who develop DSA as immunosuppression is reduced (n=20); 2 subjects with missing data at trial entry were excluded. Subjects within each group were ordered first by tolerance status, then timepoint, and finally MFI. Univariable logistic regression models were used to explore class II DSA status during ISW for association with operational tolerance. The accompanying table shows ORs and 95% CIs.