Long-term Outcomes Following Kidney and Liver Transplant in Recipients With HIV

Key Points

Question  What are the long-term outcomes of patient and graft survival in kidney and liver transplant among HIV-positive patients?

Findings  In this cohort study looking at the long-term outcomes of kidney and liver transplant in 119 HIV-positive patients who were propensity matched to 655 HIV-negative patients, results showed similar long-term graft survival in kidney transplant and similar long-term patient survival in liver transplant. Additionally, HIV-positive kidney transplant recipients who had at least 1 episode of acute rejection had significantly reduced kidney graft survival.

Meaning  Kidney and liver transplant in HIV-positive patients may be an appropriate use of transplant resources, with comparable patient and graft survival.

Importance  Kidney transplant (KT) and liver transplant (LT) in HIV-positive patients have become more widely adopted. Data looking at long-term outcomes of patient and graft survival are lacking.

Objective  To compare the long-term outcomes of KT and LT in HIV-positive recipients with matched HIV-negative recipients.

Design, Setting, and Participants  Retrospective, single-center, cohort, study using data from 2000 to 2019. Patients were observed until death, or graft failure requiring retransplant. All HIV-positive patients who underwent KT and/or LT between 2000 and 2019 were included. Propensity matching was performed to the corresponding HIV-negative cohort, which was obtained from the University of California, San Francisco’s transplant recipient registry. The data were analyzed from 2020 to 2021.

Exposures  HIV infection.

Main Outcomes and Measures  Patient and graft survival for KT and patient survival for LT. Incidence of acute rejection and its association with KT graft survival.

Results  For KT, 655 HIV-negative recipients (mean [SD] age, 52.3 [13.6] years; 450 [68.7%] were men) and 119 HIV-positive recipients (mean [SD] age, 51.7 [9.4] years; 86 [72.3%] were men) were included. Patient survival was 79.6% (95% CI, 73.6%-86.1%) and 53.6% (95% CI, 38.9%-74.0%) at 15 years posttransplant, respectively. Graft survival was 57.0% (95% CI, 47.8%-68.0%) and 75.0% (95% CI, 65.3%-86.2%) at 15 years posttransplant, respectively. Diagnosis of HIV was not associated with worse graft survival (hazard ratio, 1.09; 95% CI, 0.61-1.97; P = .77). For LT, 80 HIV-positive recipients (mean [SD] age, 52.6 [8.2] years; 53 [66.3%] were men) and 440 HIV-negative recipients (mean [SD] age, 54.6 [12.8] years; 291 [66.1%] were men) were included. Patient survival was 75.7% (95% CI, 71.8%-79.8%) for HIV-negative LT recipients and 70.0% (95% CI, 60.6%-80.8%) for HIV-positive LT recipients at 15 years posttransplant. Diagnosis of HIV was not a statistically significant predictor of patient survival (hazard ratio, 1.36; 95% CI, 0.83-2.24; P = .22). In KT, HIV-positive patients with at least 1 episode of acute rejection had a graft survival of 52.8% (95% CI, 38.4%-72.5%; P < .001) at 15 years posttransplant, compared with 91.8% in those without AR.

Conclusions and Relevance  In this single-center cohort study, KT and LT in HIV-positive patients had comparable long-term outcomes with those in matched HIV-negative patients. The high incidence of acute rejection was associated with reduced graft survival. The findings support providing transplant to HIV-positive patients, which may be an appropriate use of transplant resources and provides equitable access for HIV-positive patients.

HIV infection evolved from a fatal disease to a chronic condition in the 1990s as a result of combined antiretroviral therapy.^1-3 Consequently, people infected with HIV were no longer experiencing progression to AIDS but dying from end-stage kidney and liver disease secondary to the comorbidities of HIV.^4,5 The transplant community was slow in accepting these transplants owing to concerns for immunosuppression in a cohort of recipients who were immunocompromised from HIV infection. During the late 1990s, the community was confronted with considerable increases in the number of referrals for HIV-infected candidates for liver transplant (LT) and kidney transplant (KT) related to the high local prevalence of HIV.⁶ The University of California, San Francisco (UCSF) began a pilot trial in 2000 to evaluate the feasibility of LT and KT in HIV-positive patients who had previously been excluded from transplantation.⁷ This pilot trial demonstrated early safety, and importantly, no evidence of HIV progression to AIDS. On the basis of the safety and efficacy demonstrated in the pilot trials, the National Institutes of Health funded a prospective multicenter trial that ultimately confirmed excellent early patient and graft survival.⁸ However, at this time, a paucity of data still exists for long-term outcomes of KT and LT in HIV-positive patients, which are particularly of interest given the known increased incidence of early allograft rejection in this cohort.^8-11

As a result of the early acceptance of solid organ transplant at UCSF, we have the unique opportunity to study the long-term outcomes of 122 KT recipients and 85 LT recipients. To our knowledge, there are no long-term reports (> 10 years) on outcomes in the HIV-positive population. Given our experience with transplant in HIV-positive patients, the study’s aim was to show KT and LT outcomes of HIV-positive patients since 2000, with an appropriately matched HIV-negative cohort. We hypothesized that with comparison with an appropriately matched HIV-negative cohort, long-term graft and patient survival would be similar irrespective of HIV status.

A retrospective cohort analysis of all HIV-positive patients who underwent KT and/or LT between 2000 and 2019 was performed. The institutional review board at UCSF approved this study, and a combination of written and oral informed consent was obtained from all patients for study participation. We had a total of 122 KT recipients and 83 LT recipients. For the matching HIV-negative cohort, we searched University of California, San Francisco's transplant recipient registry over the same study period, to total 2949 HIV-negative KT recipients and 2027 HIV-negative LT recipients for propensity score (PS) matching. All patients met the Milan criteria for LT.¹² Patients with hepatocellular carcinoma (HCC) were managed with transarterial chemoembolization and radiofrequency ablation as per protocols, and the diagnosis was radiologic because pretransplant biopsies were not performed based on the risk of tumor spread. If the tumor size/burden increased beyond the Milan criteria, the patients did not proceed to transplant.

Of 3071 patients who had KT, HIV-positive patients (n = 122) were matched to HIV-negative patients (n = 2949) using optimal full PS matching. A maximum of 10 HIV-negative patients were each matched with an HIV-positive patient. Matching was restricted to patient observations that had PS in the extended common support region (0.003-0.740), which extends the common support region by 0.25 times a pooled estimate of the common SD of the logit of the PS. Standardized differences and variance ratios for the PS model covariates were used to assess sample balance after matching. Acceptable balance was defined by a maximum of 0.2 for the absolute value of standardized difference and by values within the 0.5 to 2.0 range for variance ratio. The same matching method was used on a sample of 2110 patients who received LT to match HIV-positive patients (n = 83) to HIV-negative patients (n = 2027). The PS in the extended common support region ranged from 0.002 to 0.70. The PSMATCH procedure in SAS, version 9.4 (SAS Institute) was used to perform optimal full PS matching.

The PS model for the KT cohort included age at transplant, sex, race, hepatitis C virus (HCV) status, time of transplant (2000-2006, 2007-2013, or 2014-2019), donor type (living or deceased), donor gender, donor age, and primary diagnosis leading to end-stage kidney disease (ESKD). Age at transplant did not meet the sample balance variance ratio criteria for KT patient survival and therefore was included in the propensity matching and additionally adjusted for in the Cox proportional hazards model performed on the matched sample. The matched sample included 119 HIV-positive and 655 HIV-negative patients. The PS model for the LT cohort included age at transplant, gender, race, HCV status, time of transplant, donor type (living or deceased), listing Model for End-stage Liver Disease (MELD), donor gender, donor age, transplant for HCC, and multiple organ transplant status. MELD did not meet the sample balance variance ratio criteria and therefore was included in the propensity matching and additionally adjusted for in the Cox proportional hazards model performed on the matched sample. The matched sample included 80 HIV-positive patients and 440 HIV-negative patients.

R, version 3.6.2 (R Foundation) was used. For demographics, Wilcoxon rank sum testing was performed for continuous variables and χ² testing for categorical variables. Death-censored Kaplan-Meier survival analysis of KT patient and graft survival, and LT patient survival was performed. Survival curves were created using the ggplot2 and survminer packages and stratified by HIV status. The study team performed stratified log-rank testing by HIV status to appropriately account for the propensity-matched data set.¹³ Cox proportional hazards regression was performed with KT patient and graft survival as the outcomes of interest. For KT patient survival, the covariates in the model included HIV status, but also age at transplant because this variable could not be accounted for with propensity matching as described above. For KT graft survival, the study team adjusted for HIV status. For LT patient survival, the study team adjusted for HIV status and MELD.

For both HIV-positive KT recipients and LT recipients, the study staff examined patient survival based on HCV status and era of transplant. Kaplan-Meier analysis were performed, stratifying by HCV status and transplant pre-2014 or post-2014, the time at which the institution instituted anti-HCV direct-acting agents (DAAs). Log-rank testing was performed to determine statistical significance. The cumulative incidence of acute rejection (AR) in HIV-positive KT and LT was also determined using the Kaplan-Meier method. Kidney transplant graft survival was plotted, stratifying based on episodes of AR, with log-rank testing. Univariate Cox proportional hazards regression analysis of the relevant clinical variables was performed to determine association with AR. Covariates that had a P value less than .05 in the univariate analyses were included in the final multivariate proportional hazards model to determine significant associations with AR.

In the KT analysis, we included 119 HIV-positive recipients and 655 HIV-negative recipients (following propensity matching). Matched KT recipient and donor demographics are presented in Table 1. After matching, noteworthy differences between groups included era of transplant, HCV status, months receiving dialysis, and select subgroups of race and ethnicity. In the LT analysis, 83 HIV-positive recipients and 468 HIV-negative recipients were included. Matched LT recipient and donor demographics are presented in Table 1. After matching, noteworthy differences between groups included age at transplant and transplantation for HCC.

Patient survival was 79.6% (95% CI, 73.6%-86.1%) for HIV-negative KT and 53.6% (95% CI, 38.9%-74.0%) for HIV-positive KT at 15 years posttransplant (P = .03; Figure 1A). Additionally, the stratified log-rank test¹³ comparing survival based on HIV status suggested that HIV status was associated with worse long-term survival. Cox proportional hazards regression was performed, adjusting for HIV status, with HIV infection being associated with worse survival (hazard ratio [HR], 2.8; 95% CI, 1.4-5.5; P < .001). Additionally, the proportional hazards model was also adjusted for age at transplant (HR, 1.1; 95% CI, 1.0-1.1; P < .001), as this variable was included in the propensity matching but did not meet the appropriate sample balance variance ratio criteria as previously described.

Graft survival was 57.0% (95% CI, 47.8%-68.0%) for HIV-negative KT and 75.0% (95% CI, 65.3%-86.2%) for HIV-positive KT at 15 years posttransplant (P = .77; Figure 1B). The stratified log-rank test based on HIV status,¹³ and the Cox proportional hazards model adjusting for HIV status (HR, 1.09; 95% CI, 0.61-1.97; P = .77) suggested that HIV status was not associated with graft survival.

Given previous data suggesting particularly poor outcomes in HIV-HCV coinfection,⁸ the study team examined 5-year patient survival comparing HIV-HCV coinfection based on transplant pre-2014 and post-2014, the time at which the institution introduced DAAs for HCV treatment. Notably, the worst survival among all subgroups was in HIV-HCV–positive coinfected patients prior to DAA use, as expected (Figure 1C), with survival at 5 years posttransplant down to 57.1% (95% CI, 39.5%-82.8%), which reached statistical significance (log-rank P = .045).

Patient survival was 75.7% (95% CI, 71.8%-79.8%) for HIV-negative LT, and 70.0% (95% CI, 60.6%-80.8%) for HIV-positive LT at 15 years posttransplant (Figure 2A), with a stratified log-rank test that was not statistically significant (P = .12).¹³ Cox proportional hazards regression was performed, adjusting for HIV status, and revealed similar findings to the stratified log-rank test as expected (HR, 1.36; 95% CI, 0.83-2.24; P = .22). In addition, the proportional hazards model was also adjusted for MELD at transplant (HR, 0.98; 95% CI, 0.9-1.0; P = .36), as this variable could not be included as part of the propensity matching as previously described.

The study team then examined the 5-year patient survival comparing HIV-HCV coinfection based on transplant pre-2014 and post-2014 (DAA use), as described previously for KT. Notably, the worst survival among all subgroups was in HIV-HCV–positive coinfected patients prior to DAA use (Figure 2B), as would have been expected, with survival at 5 years posttransplant down to 59.5% (95% CI, 45.6%-77.6%), which reached statistical significance (log-rank P = .04).

In HIV-positive LT recipients, explants showed HCC remaining in 22 recipients with the following characteristics: 4 well-differentiated, 8 moderately differentiated, 1 poorly differentiated, and 8 entirely necrotic tumors. Vascular invasion was present in 4 tumors, and 2 tumors had mixed findings of cholangiocarcinoma and HCC. Only 1 patient died of recurrence and metastatic disease, and their pathology was notable for a moderately differentiated tumor with vascular invasion.

In the KT cohort, 42 patients had at least 1 episode of AR. At 1 year posttransplant, AR incidence was greater than 20%, and by 3 years posttransplant was greater than 30%, both notably higher than national data among all KT recipients, where the 1-year cumulative incidence of AR is less than 10%.¹⁴ In the LT cohort, 24 patients had at least 1 episode of AR. At 1 year posttransplant, AR incidence had nearly peaked to just less than 30%, which is notably higher that national data among all LT recipients, where the 1-year cumulative incidence of AR is just greater than 10%.¹⁵

The study team also looked at the outcome of AR on long-term graft survival. HIV-positive KT patients who had at least 1 episode of AR had a graft survival of 52.8% (95% CI, 38.4%-72.5%) at 15 years posttransplant, compared with 91.8% in those recipients without rejection (Figure 3). Cox proportional hazards univariate analysis revealed that the only statistically significant associations of AR with KT were age at transplant (HR, 0.96; 95% CI, 0.9-1.0; P = .02), delayed graft function (DGF) (HR, 2.61; 95% CI, 1.41-4.85; P = .002), induction immunosuppression used (basiliximab HR, 0.41; 95% CI, 0.19-0.89; P = .03; thymoglobulin HR, 0.25; 95% CI, 0.11-0.60; P = .002), and integrase inhibitor use (HR, 0.50; 95% CI, 0.25-1.00; P = .05). When these variables were put into the multivariate model, only induction with thymoglobulin (HR, 0.21; 95% CI, 0.09-0.53; P < .001) and DGF (HR, 3.12; 95% CI, 1.63-5.97; P < .001) remained associated with AR in KT, with thymoglobulin associated with lower AR risk and DGF associated with greater AR risk. Univariate proportional hazards analysis of AR in LT did not reveal any statistically significant associations (Table 2).

The results of this study show that KT and LT in HIV-positive patients have comparable long-term graft and patient survival to HIV-negative patients. Additionally, we saw improved outcomes since the advent of anti-HCV DAA. Finally, rates of AR in the cohorts were similar to previously described experiences, and notably higher than national data for both HIV-negative KT and LT, which predicted worse graft survival in KT.

The first prospective National Institutes of Health trial that examined outcomes of KT and LT in HIV-positive patients showed comparable findings to HIV-negative patients at 1 year and 3 years, with KT patient survival of 95% and 88%, respectively, while allograft survival was 90% and 74%, respectively.⁸ In a matched analysis on multiple covariates, compared with HIV-negative LT, HIV-positive LT had a higher hazard of both graft failure and death; however, the absolute difference in the proportion of deaths was only 6.7% in the risk-matched analysis.¹⁶ The first long-term analysis in HIV-positive LT used the Scientific Registry of Transplant Recipient data and compared 180 HIV-positive LT recipients with matched HIV-negative controls over 10 years.¹⁷ The authors found that compared with matched HIV-negative controls, HIV-positive LT recipients had a 1.68-fold increased risk for death, and that these differences persisted independent of HCV status. However, in the post-2008 era, risk for death was similar between monoinfected and uninfected LT recipients, while in contrast, independent of transplant era, coinfected LT recipients still had increased risk for death.

Acceptance of HIV-positive KT has been more widespread. While studies have shown similar outcomes in HIV-positive patients compared with HIV-negative patients,^8,18 long-term outcomes are still lacking. The present work highlights comparable long-term graft survival outcomes in KT, irrespective of HIV status. Of note, the study team found reduced long-term patient survival among the HIV-positive cohort; we believe this is likely related to long-term cardiovascular HIV/AIDS–related comorbidities,^19,20 as studies have shown elevated risk of these complications in HIV-positive patients.²¹ It is important to evaluate appropriate use of donor organs given their status as a limited resource. While there is no universally accepted life expectancy where a potential recipient will become ineligible, we prefer to perform transplants in patients who are expected to live at least 5 years posttransplant. We showed that overall survival was still comparable even up to 10 years post-KT. Additionally with the advent of the HIV Organ Policy Equity Act, HIV-positive patients have access to HIV-positive donors and are waiting less time for those organs. These organs can only be used in HIV-positive patients, expanding the donor pool and benefiting all transplant waitlist patients, irrespective of HIV status.^22,23

Locke et al¹⁸ in particular had highlighted the adverse outcomes of HIV-positive KT in HCV-coinfected patients. Sawinski et al²⁴ also found similar post-KT outcomes irrespective of HIV status, although also showed worse outcomes in HCV coinfection. The study team shows above that since the advent of HCV DAA, outcomes have been comparable irrespective of HCV coinfection. Concerns about transplant in HIV-HCV–coinfected patients should therefore be thoroughly reevaluated in this new era of anti-HCV DAA. Unfortunately, because the institution only began implementing HCV DAA in 2014, the survival analysis could only be carried out to 5 years posttransplant. Future work will need to look at the effects of HCV DAA over a longer period of time. Additionally, the increased incidence of AR in HIV-positive KT is well established.^8-11 This higher incidence of AR is thought to be multifactorial. Initial hypotheses considered the expansion of memory T cells in HIV infection and the concomitant enhancement of alloimmunity, along with prior infections generating memory alloreactive T cells as a result of cross-reactivity.⁸ Given the common use of protease inhibitors (PIs), such as ritonavir, a potent CYP3A4 inhibitor, it was thought that drug-drug interactions between PIs and calcineurin inhibitors may have been responsible for inadequate trough levels, resulting in inadequate immunosuppression and thus higher AR rates.²⁵ This has led to some centers transitioning HIV-infected patients from PI-based regimens in anticipation of transplant,²⁶ with a preference for integrase-based regimens.²⁷ The data showed that in the center, at least 1 episode of AR was associated with a statistically significant decrease in graft survival. The management and prevention of AR in HIV-positive KT will therefore continue to be a key component in the care of these patients.

In contrast to KT, LT in HIV-positive patients has been less widely accepted, with even less data available to help guide decision-making. Although Roland et al¹⁶ found a higher relative hazard of graft loss and death in HIV-positive LT recipients, the study shows comparable long-term outcomes in LT, irrespective of HIV status. In comparison with this study, the study team observed patients over the longer term and included the analysis of patients over multiple eras of treatment, including the post-DAA era. The results also contrasted to the findings from Locke et al,¹⁷ where HIV-positive LT recipients had an increased risk for death, although it is worth noting that the authors looked specifically at the Scientific Registry of Transplant Recipient national data, while this work is exclusively from the single-center results.

The study is limited by its retrospective nature. Although the work is also partially limited owing to being from a single center, the study team believes this may have helped prevent some heterogeneity in outcomes that could have resulted from a multicenter study. The study team were also partly limited by the propensity matching. Although the study team attempted to match on multiple variables, the number of variables that could be matched on were limited. As reported in Table 1, the study had overall good matching among all matched variables; those that were not matched on were accounted for in the Cox proportional hazards models.

In this cohort study, the study data show that over the long-term, outcomes of graft and patient survival in HIV-positive KT and LT were comparable with those of HIV-negative patients. Given the complexity of HIV-positive patients, it was important to match patients based on multiple variables to account for these clinical comorbidities. Successful treatment of HCV coinfection was the last hurdle to achieve comparable results to the HIV-negative–transplant recipient. The findings, in combination with previously published work, support providing organ transplant to HIV-positive patients, which the study team believes is an appropriate use of transplant resources and provides equitable access to these resources for HIV-positive patients.

Accepted for Publication: October 25, 2021.

Published Online: January 5, 2022. doi:10.1001/jamasurg.2021.6798

Corresponding Author: Arya Zarinsefat, MD, Department of Surgery, University of California, San Francisco, 513 Parnassus Ave, S-321, San Francisco, CA 94143 (arya.zarinsefat@ucsf.edu).

Correction: This article was corrected on February 9, 2022, to fix an error in the Abstract.

Author Contributions: Dr Zarinsefat had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Zarinsefat, Ascher, Stock.

Acquisition, analysis, or interpretation of data: Zarinsefat, Gulati, Shui, Braun, Rogers, Hirose, Stock.

Drafting of the manuscript: Zarinsefat, Gulati, Shui, Hirose, Stock.

Critical revision of the manuscript for important intellectual content: Zarinsefat, Shui, Braun, Rogers, Hirose, Ascher, Stock.

Statistical analysis: Zarinsefat, Shui.

Obtained funding: Stock.

Administrative, technical, or material support: Braun, Rogers, Stock.

Supervision: Zarinsefat, Rogers, Hirose, Ascher, Stock.

Conflict of Interest Disclosures: None reported.

Funding/Support: Dr Zarinsefat was funded by a National Institutes of Health T32 training grant for this work.

Role of the Funder/Sponsor: The funder had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.