The differences in post-liver transplant outcomes of patients with autoimmune hepatitis who present with overlapping autoimmune liver diseases

Patients with autoimmune hepatitis (AIH) may co-present with features of primary biliary cholangitis (PBC) or primary sclerosing cholangitis (PSC). Using a national transplant registry, the outcomes of patients with these autoimmune liver conditions were compared. The UNOS-STAR registry was used to select a study population of AIH, PSC, and PBC liver transplant (LT) patients. Living and multi-organ transplant cases were excluded. Using the UNOS-registered diagnoses, the study population was subdivided into those with nonoverlapping autoimmune liver diseases and those with overlapping forms (e.g., AIH–PBC). Outcomes were compared, using endpoints such as all-cause mortality, graft failure, and organ-system specific causes of death. The main analysis featured 2048 entries, with 1927 entries having nonoverlapping AIH, 52 entries having PSC overlap, and 69 entries having PBC overlap. Patients with PBC overlap were more likely to have graft failure (adjusted hazard ratio [aHR] 3.46 95% CI 1.70–7.05), mortality secondary to respiratory causes (aHR 3.57 95% CI 1.23–10.43), and mortality secondary to recurrent disease (aHR 9.53 95% CI 1.85–49.09). Case incidence rates reflected these findings, expressed in events per 1000 person-years. For patients with PBC overlap and nonoverlapping AIH cases, respectively. Graft failure: 28.87 events vs. 9.42 events, mortality secondary to respiratory causes: 12.83 deaths vs. 3.77 deaths, mortality secondary to recurrent disease: 6.42 deaths vs. 1.26 deaths. Those with AIH–PSC overlap experienced a higher risk of death from graft infection (aHR 10.43 95% CI 1.08–100.37; case-incidence rate: 3.89 vs. 0.31 mortalities per 1000 person-years). Supplementary analysis showed similar findings, in which overlapping autoimmune conditions were associated with higher adverse outcome rates. Patients with AIH-PBC overlap have higher risk of mortality due to recurrent liver disease and respiratory causes, and patients with AIH-PSC overlap have higher risk of mortality due to graft infection. While further prospective studies are needed to clarify the underlying mechanisms related to these findings, our study characterizes the prognostic implications of AIH overlap on post-LT mortality and graft failure risks.


Introduction
Autoimmune hepatitis (AIH) is a chronic, inflammatory, liver pathology characterized by a loss of tolerance to hepatocyte-specific autoantigens [1]. AIH manifestations can range from an asymptomatic condition to end-stage liver disease [2], with fulminant forms carrying a significant mortality risk [3]. The phenotypic expression of AIH can also vary at the tissue and cellular levels [3]. AIH overlaps histologically with PBC and PSC at a rate of 10-20% and 2-8% [4], and these subtle differences in inflammation physiology may translate to discrepant clinical outcomes. Evidence for this theory has been demonstrated in several studies [5][6][7]. Indeed, AIH-PSC overlap has been linked to the highest morbidity and mortality risk [5].
While overlapping autoimmune liver conditions (e.g., AIH and PBC, PSC and PBC, AIH and PSC) are thought to complicate outcomes in non-transplant hepatology, less is known about the implications of overlap in the setting of liver transplant (LT). This knowledge gap persists despite the fact that certain instances of fulminant and chronic AIH progress to hepatic failure [8], thereby indicating LT as a necessary, life-saving intervention. It is possible that the differences in physiology among overlapping and non-overlapping autoimmune liver diseases may confer discrepant post-LT prognoses, via the mechanisms of disease recurrence or nonhepatic manifestations of the underlying disorder.
To evaluate the risk of carrying an overlapping form of autoimmune liver disease, we used a national transplant registry to investigate outcomes post-LT, with special attention to the AIH LT population. Patients were stratified via their diagnostic history of overlapping autoimmune liver disease. Furthermore, we controlled for confounding via multiple iterations of Cox regressions analyses, to estimate the corresponding risks for all-cause mortality and system-specific causes of death.

Database
The UNOS (United Network for Organ Sharing) STAR Standard Transplant Analysis and Research) database is a registry that describes the organ transplant system in the United States. UNOS-STAR data consists of preoperative recipient and donor characteristics, periprocedural data, and longitudinal outcomes of graft recipients. The database is publicly available to biomedical researchers via data use agreements, and patient confidentiality is protected because all information is de-identified. While UNOS is sponsored by the federal government, the contents of this paper were not influenced by any regulatory authorities. All findings and interpretations are solely the authors' viewpoints.

Patient selection
From 2005 to 2019, there were 99,987 LT cases tracked by UNOS STAR. This sum was refined via the following exclusions: entries lost to follow-up (n = 3445), entries with a second LT operation (n = 4310), entries with a history of LT (n = 5437), entries with erroneous datapoints (e.g., creatinine < 0, n = 5), entries with transplant age under 18 years (n = 6872), entries with a non-heart beating donor (n = 4427), entries with a living donor (n = 3081), entries with a partial liver (n = 1012), entries undergoing multi-organ transplants (e.g., simultaneous liver-kidney transplant) (n = 6462), and entries without a diagnosis of AIH, PBC, or PSC (n = 58,282). To qualify as having a diagnosis of AIH, PSC, or PBC, cases must have had either a primary or secondary diagnosis of these conditions at one of two timepoints-the time of transplant listing or the time of transplant. Cases were further reviewed to determine if they carried a primary or secondary diagnosis of PSC or PBC, using the same two timepoints of either transplant listing or time of transplant. This procedure resulted in the main and supplementary analysis cohorts, depicted in Fig. 1. The total number of patients included in the study was 6654 as indicated in the patient selection process. Patients with overlapping autoimmune conditions were used in both main and supplementary analyses.

Study variables
Overlap presence was the study exposure for each analysis. Multiple schemes were produced to assess outcomes between these classifications, further described below. Baseline characteristics of interest included patient demographics, gender mismatch variables, comorbidities, liver complications and status (e.g., ascites, MELD score), immunosuppressant medications, laboratory markers, critical care and life support utilization, donor demographics, donor comorbidities, and donor laboratory markers. Compared liver disease comorbidities in the recipient consisted of diabetes, hepatitis B virus, hepatitis C virus, acute liver failure, ulcerative colitis, and Crohn's Disease. The primary study endpoints were all-cause mortality and graft failure. Secondary mortality endpoints were systems-specific causes of death, which included respiratory etiologies, graft infection etiologies, recurrent disease etiologies, and infectious etiologies. The death due to respiratory causes endpoint was a composite of respiratory failure, acute respiratory distress syndrome, pulmonary edema or insufficiency, and opportunistic infection etiologies, whereas death due to graft infection referred to situations in which mortality resulted from graft-centered microbe-driven process. Examples of pathogens that could precipitate this mortality etiology include cytomegalovirus, Epstein-Barr virus, and Herpes Simplex virus. Infectious cause of death was a composite of viral, bacterial, and fungal-induced mortality from any organ system (e.g., pneumonia, bacteremia, meningitis, etc.).

Statistical methods and analytic design
Baseline characteristics were compared using mean-based and nominal-based statistics. Student's T-tests and Mann-Whitney U tests were used for parametric and non-parametric continuous factors, respectively, whereas Chi-square and Fisher's Exact tests were used for categorical data. Iterative Cox regression models were created for the AIHoverlapping versus non-overlapping evaluations, using the primary or secondary outcomes of interest as the endpoint. The regression models included the following covariates: model 1-adjusted for age, BMI, ethnicity, and gender; model 2-adjusted with additional inclusion of comorbidities of diabetes, hepatitis, alcoholic liver disease, and hepatocellular carcinoma; model 3-additional inclusion of ascites or encephalopathy presence, laboratory data, and MELD score; model 4-additional inclusion of LT donor demographics. Standard 95% two-tail confidence intervals with an alpha level of 0.05 were used to indicate statistical significance. For the all-cause mortality and graft failure endpoints, hazard-event analyses were conducted to determine the log-rank statistics, using the prespecified strata to evaluate the comparative outcomes. Forest plots showing adjusted hazard ratios (aHR) for the AIH overlap exposure as well as covariates were constructed, assessing their relationship with the all-cause mortality and graft failure endpoints.
Multiple analytic designs were created to characterize outcomes of patients with autoimmune-type liver diseases. The main cohort compared patients with AIH-PSC to those without PSC, and AIH-PBC patients to those without PBC. Additional analyses were as follows: assessment of baseline characteristics between AIH-PSC versus AIH-PBC overlapping phenotypes (Supplementary Table 1 Supplementary Tables 13 through  15). For all supplementary analyses, the same statistical protocol was followed, which included baseline characteristic comparison, all-cause mortality and graft failure primary outcome sequential Cox regression, and system-specific secondary outcome sequential Cox regression.
Variables were plotted to assess missing values patterns, and random forest plot generation was used to compensate for missing data, reduce potential biases, and improve power [9]. All statistics were conducted with RStudio version 1.2.5042, R code version 3.6.3.

Follow-up and baseline characteristics of main analysis
Median time periods of follow-up were as follows: for AIH-PSC overlap-3.89 years (25- were not significantly different between cohorts. PSC-AIH overlap patients were more likely to carry a diagnosis of ulcerative colitis (32.7% vs. 0.0%, p < 0.001) and Crohn's Disease (11.5% vs. 0.0%, p < 0.001), but PBC-AIH overlap patients were not characterized by higher rates of inflammatory bowel disease. Neither overlapping cohorts saw higher rates of diabetes compared to their reference groups. Further baseline comparisons are pictured in Tables 1, 2.

Analyses after excluding hepatitis B and C
AIH patients with overlapping liver disease were again compared; however, entries with other liver disease etiologies (hepatitis B, C, alcoholic cirrhosis or non-alcoholic steatohepatitis) were excluded. Findings were similar to those reported in the main analysis. The presence of PSC-overlap was not associated with a difference in all-cause mortality hazard (aHR 1.08 95% CI 0.59-1.99) nor graft failure hazard (aHR 1.68 95% CI 0.60-4.72). PBC-overlap trended toward an increase in all-cause mortality risk, but this result did not reach significance (aHR 1.52 95% CI 0.99-2.33). However, graft failure risks were elevated in this patient group (aHR 3.20 95% CI 1.56-6.56; case-incidence rate: 29.73 vs. 9.50 failures per 1000 person-year). Regarding secondary   Tables 5 through 12 further describe these data, and assessments of baseline data are also included.

Analyses stratified by tacrolimus use
The main analysis was repeated, albeit with the study cohort restricted to patients having a tacrolimus regimen.

Discussion
In our study, those with AIH-PBC were at a higher risk of graft failure than those with nonoverlapping or uncomplicated AIH. Similar findings persisted across multiple subanalyses, including tests in which PSC or PBC populations were used as reference, hepatitis B and C patients were excluded, and patients were stratified by tacrolimus use.
Upon evaluating system-specific causes of death, AIH-PSC patients were associated with graft infection-related mortality, whereas AIH-PBC patients were at a higher hazard for respiratory and recurrent liver disease-induced death. The prognostic implications of overlap syndrome in AIH patients who undergo LT have been seldom investigated. Some single-center studies have shown statistically equivalent survival between uncomplicated AIH and overlapping   AIH; however, they have also found a greater incidence of disease recurrence among the overlapping populations [11,12]. These findings apparently contrast with a number of non-transplant studies that have associated comorbid autoimmune liver disorders with worse outcomes in advanced-stage AIH patients [5][6][7]13]. In the pre-LT setting, a possible explanation for this difference may be related to discrepant response to standard steroid therapy among overlapping patients [14]. For example, it was reported that a combined therapy targeting both PBC and AIH was more likely to attain disease remission than single therapy [14,15]. Given the relevance of post-LT immunosuppression, these types of differences in regimen response suggest that post-LT prognosis may be altered by the presence of AIH-overlap. Prior studies have suggested heightened rates of disease recurrence following LT among patients with autoimmune liver diseases (AIH, PBC, and PSC) [16][17][18][19][20][21][22], with recurrence rates of 16-43%, 10-50%, and 20%, respectively. AIH-PBC patients experienced escalated risk of mortality from original disease recurrence, suggesting that some concomitant autoimmune disorders could cause liver disease in the new liver graft. Mechanisms driving this observation are poorly defined, but it is probable that they can be attributed to the variable response of AIH-PBC overlap to steroid therapy [23,24]. Furthermore, the histological distortions of the liver cytoarchitecture may be increased by the copresence of AIH-and PBC-related disease activity, which may augment the deleterious effects of AIH-induced graft necroinflammation and disease recurrence [25].
Similarly, PSC is thought to recur at an increased rate post-LT, resulting in biliary strictures and graft malfunction [17,18]. PSC patients are at a higher risk of steroid-resistant graft rejection [26,27], which can debilitate the viability of the graft and result in graft failure. Combined, these features can exacerbate the rates of graft damage among AIH-PSC overlapping patients. With steroid-resistant PSC, a stronger immunosuppressant regimen may be needed to achieve a therapeutic threshold, but this approach could introduce a potentially higher likelihood of infectious complications. With disease recurrence, the stenotic trees in the graft biliary tract may serve as a nidus for infection in already immunosuppressed AIH-PSC patients. AIH-PSC patients were also characterized by higher rates of ulcerative colitis and Crohn's Disease. It is probable that the coexisting inflammatory bowel disease among AIH-PSC patients may serve as entry points for infectious pathogens, resulting in graft infections and similar complications [28].
Currently, it is unclear why AIH-PBC patients experience higher rates of respiratory-induced death following LT. A possible mechanism could be contingent upon the higher immunosuppressant dose needed to prevent rejection in the AIH-PBC patient. Increased immunosuppression can increase risk for infection, and this could manifest through complications such as pneumonia and pneumonia-related acute respiratory distress syndrome [29]. Nonetheless, further validating studies are needed to determine the underlying cause.

Implications and limitations
Clinically, the care of post-LT AIH patients with PSC or PBC-overlap should be characterized by an increased clinical awareness of the adverse outcomes observed in this study. For AIH-PSC patients, graft function should be routinely monitored per the standard of care, but with greater clinical caution of potential graft infection. Symptoms of infection or changes in clinical status should warrant a workup that is comprehensive and includes liver graft and other systems including the respiratory system. PBC-overlap patients should undergo routine liver function checks during the immediate to subacute post-LT period to monitor PBCdisease activity (i.e., alkaline phosphatase).
In the future, prospective studies that focus on the post-LT effects of histological overlap between AIH and autoimmune liver disorders should be conducted. Given the limitations  of the UNOS database, we used the registrant diagnoses to indicate the clinical overlap of AIH and autoimmune liver conditions. Histological review of samples was not possible. While this method precludes the disease-specific histological features from defining the different subtypes of AIH, we ensured that the alternative liver diagnoses (AIH with either PSC or PBC) were established prior to comparing the prognostic risks. Nonetheless, the clinical diagnosis of AIH overlap cannot ascertain the underlying histological characteristics and hence requires further investigation. In addition, the UNOS database does not include medication regimen data for recipients prior to transplant. Therefore, differences in therapies could not be controlled despite the iterative Cox regression procedure.

Conclusion
The presence of diagnostic overlap as presented by the concurrence of PBC or PSC among patients with AIH affects the post-liver transplant outcomes; for AIH-PBC overlap, there is a higher risk of death due to recurrent liver disease and respiratory causes, whereas for AIH-PSC overlap there is a higher risk of death due to graft infection. While further prospective studies are needed to elucidate the underlying features of pre-LT recipient histology that affect the post-LT outcomes, the current observations characterize the prognostic implications of the AIH-diagnostic overlap on post-LT mortality and graft failure risks.