What is the impact of human leukocyte antigen mismatching on graft survival and mortality in renal transplantation? A meta-analysis of 23 cohort studies involving 486,608 recipients

Abstracts Background The magnitude effects of human leukocyte antigen (HLA) mismatching on post-transplant outcomes of kidney transplantation remain controversial. We aim to quantitatively assess the associations of HLA mismatching with graft survival and mortality in adult kidney transplantation. Methods We searched PubMed, EMBASE and the Cochrane Library from their inception to December, 2016. Priori clinical outcomes were overall graft failure, death-censored graft failure and all-cause mortality. Results A total of 23 cohort studies covering 486,608 recipients were selected. HLA per mismatch was significant associated with increased risks of overall graft failure (hazard ratio (HR), 1.06; 95% confidence interval (CI), 1.05–1.07), death-censored graft failure (HR: 1.09; 95% CI 1.06–1.12) and all-cause mortality (HR: 1.04; 95% CI: 1.02–1.07). Besides, HLA-DR mismatches were significant associated with worse overall graft survival (HR: 1.12, 95% CI: 1.05–1.21). For HLA-A locus, the association was insignificant (HR: 1.06; 95% CI: 0.98–1.14). We observed no significant association between HLA-B locus and overall graft failure (HR: 1.01; 95% CI: 0.90–1.15). In subgroup analyses, we found recipient sample size and ethnicity maybe the potential sources of heterogeneity. Conclusions HLA mismatching was still a critical prognostic factor that affects graft and recipient survival. HLA-DR mismatching has a substantial impact on recipient’s graft survival. HLA-A mismatching has minor but insignificant impact on graft survival outcomes. Electronic supplementary material The online version of this article (10.1186/s12882-018-0908-3) contains supplementary material, which is available to authorized users.


Background
Compared with dialysis, renal transplantation is a more preferred option for end-stage renal disease (ESRD) [1]. In recent report of global database on donation and transplantation (http://www.transplant-observatory.org), about 80,000 renal transplants were performed annually [2]. However, in 2016 United States Renal Data System (USRDS) Annual Data Report, the long-term survival benefit remained unsatisfactory, with ten-year graft survival probabilities of 46.9% for deceased donor transplant [3].
Human leukocyte antigen (HLA) was important biological barrier to a successful transplantation and has substantial impact on the prolongation of graft survival [4]. However, the emergency of modern immunosuppressive agents minimized the effect of HLA compatibility. The US kidney allocation system was extensively modified to eliminated HLA-A similarity in 1995 [5] and HLA-B similarity in 2003 [6]. In the revised United Kingdom kidney allocation scheme, HLA-A matching is no longer considered [7]. But the latest European Renal Best Practice Transplantation Guidelines still recommended that matching of HLA-A, -B, and -DR whenever possible, while gave more weight to HLA-DR locus [8]. So far, the current kidney allocation guideline recommendations were inconsistent in term of HLA compatibility. Besides, for the primary aim to make the kidney last as long as possible, all the current kidney allocation systems were not perfect. Here, we sought to conduct a meta-analysis to assess the magnitude effect of HLA mismatching in adult kidney transplantation, with a particular focus on graft survival and recipient mortality.

Methods
The study was registered in the PROSPERO international prospective register of systematic reviews (CRD42017071894). Details of protocol are described in Additional file 1: Supplemental Methods. The metaanalysis was performed in accordance with the Metaanalysis of Observational Studies in Epidemiology (MOOSE) protocol [9] (Additional file 2: Table S1) and the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) guideline [10] (Additional file 3: Table S2).

Literature search strategy
We searched PubMed, EMBASE and the Cochrane Library from their inception to December, 2016, without language restriction. We used the following combinations of Medical Subject Heading (MeSH) terms and corresponding text words: "kidney transplantation", "renal transplantation", "human leukocyte antigen", "HLA" and all possible spellings of "survival". Further details are described in Additional file 1. Reference lists of articles were manually screened to identify further relevant studies. The literature search was performed independently by two investigators (XMS and XHZ). Differences were resolved by consensus.

Study selection
We included studies that (1) included a study cohort comprising adult post-kidney transplant recipients; (2) were cohort studies/trials reporting associations between HLA mismatching and post-transplant survival outcomes; and (3) provided effect estimates of hazard ratios (HRs) with 95% confidence interval (CIs). Studies reporting data on children or animals or in vitro research were excluded. Besides, reviews, meta-analyses, case reports, case series and technical descriptions with insufficient data or unrelated topics were also excluded. For studies covered overlapping data, we included the most recent and informative one. XMS and XHZ independently screened the titles and abstracts for eligibility. Discrepancies were resolved by consensus.

Outcome measures
Our primary clinical endpoint was overall graft failure; secondary clinical endpoints were death-censored graft failure and all-cause mortality. The European Renal Best Practice Transplantation Guidelines and Kidney Disease: Improving Global Outcomes Guidelines was used to evaluate the incidence of measured outcomes [11,12].
Data extraction and quality assessment Data were extracted from predefined protocol, then recorded in a standardized Excel form, including the first author's name, publication date, study location, study design, cohort size, recipient age, sex distribution, duration, donor source, data source (multi-centered or singlecentered), follow-up, unadjusted and adjusted HRs of overall graft failure, death-censored graft failure and allcause mortality per HLA-mismatch increased, and adjusted covariates in reported multivariable analysis. We contacted libraries abroad or corresponding author of relevant articles by email when detailed data for pooling analysis was unavailable. The methodological quality of included studies was described using the Newcastle-Ottawa Scale. High-quality studies were defined by a score of > 5 points [13]. Disagreements in the scores were resolved by consensus between XMS, XHZ and JD.

Statistical analysis
Hazard ratios (HRs) with corresponding 95% confidence intervals (CIs) were directly retrieved from each study. We chose HRs as the statistic estimates because they correctly reflect the nature of data and account for censoring. Cochran's Q test and I 2 statistic were applied to assess heterogeneity between studies. The following criteria were used: I 2 < 50%, low heterogeneity; 50-75%, moderate heterogeneity and > 75%, high heterogeneity [14,15]. When significant heterogeneity was found between studies (P < 0. 10 or I 2 > 50%), the effect estimates were calculated using a random-effects model and the DerSimonian-Laird method [16]; otherwise, a fixed-effects model with the Mantel-Haenszel method was used [17]. Subgroup analyses included recipient sample size (≥10,000 vs < 10,000), the nature of data (univariable-unadjusted vs multivariableadjusted effect estimates), donor source (deceased vs living and deceased), data source (multi-centered vs singlecentered) and geographical locations (Europe, North America, Asia and Oceania). A sensitivity analyses was performed by omitting one study at a time and then reanalyzing the data to assess the change in effect estimates. To further explore heterogeneity, a random-effects univariate meta-regression was conducted when at least 10 studies were available. For outcomes of at least 10 studies included, publication bias was assessed by funnel plot and Egger test [18]. Egger test with two-sided P < 0.10 was considered to be statistically significant. Analyses were performed using STATA software, version 13.0 (STATA Corporation, College Station, Texas, USA).

Primary outcomes HLA per mismatch and overall graft failure
Eleven studies (289,987 adult recipients) reported data on HLA mismatching and overall graft failure. The pooled analysis revealed that each incremental increase of HLAmismatches was significant associated with a higher risk of overall graft failure, both in univariable-unadjusted summary estimates (HR: 1.14; 95% CI: 1.04-1.26; P = 0. 008; Fig. 2) and multivariable-adjusted summary estimates (HR: 1.06; 95% CI: 1.05-1.07; P < 0.001; Fig. 2). The heterogeneity was low (I 2 = 24.8 and 27.4%, respectively). Detailed predefined subgroup analyses were listed in Table 2. The effect estimates did not changed significantly after stratification for sample size (≥10,000 vs < 10,000), data source (multi-centered vs single-centered), donor source (cadaveric vs living and cadaveric), geographic locations (European, North America, Asia and Oceania) and year period (prior to 1995 vs not prior to 1995). In sensitivity analysis, the summary estimates were not modified after excluding one study at a time. Subsequent univariate metaregression indicated that these factors did not significantly change the overall effect (Additional file 5: Fig. S1). Publication bias was not significant (Additional file 6: Fig. S2).

HLA-DR mismatches and overall graft failure
Eight studies with 152,105 adult recipients were analyzed to investigate the association between HLA-DR mismatching and overall graft failure. The pooled results revealed an unadjusted HR of 1.44 (95% CI: 0.86-2.41; P = 0.160) with moderate heterogeneity (I 2 = 70.0%). After adjustment, each incremental increase of HLA-DR mismatches was significant associated with 12% higher   Table 2). The effect estimates remained stable after excluding one study at a time. Considering only 8 studies included in meta-analysis, we did not perform a meta-regression.

HLA-B mismatches and overall graft failure
Associations of HLA-B epitope and overall graft failure were reported in 4 studies with 146,019 recipients. The pooled analysis demonstrated that each incremental increase of HLA-B mismatches was not associated with higher risk of overall graft failure (HR: 1.01; 95% CI: 0.90-1.15; P = 0.834; Fig. 3), with moderate heterogeneity (I 2 = 66.0%). Sensitivity analysis with a fixed-effects model obtained similar effect estimates (HR: 1.01; 95% CI: 0.89-1. 14; P = 0.079). In addition, the effect estimates did not changed significantly after stratification for sample size (≥10,000 vs < 10,000) of cohorts.

Discussion
This is the first meta-analysis to evaluate the magnitude effect of HLA mismatching on post-transplant survival outcomes of adult kidney transplantation. The analysis included 23 studies with a large sample of subjects (totally 486,608 recipients). The results indicated that each incremental increase of HLA mismatches was significantly associated with higher risks of overall graft failure, death- The effect estimates were stratified for sample size (≥10,000 vs < 10,000), data source (multi-centered vs single-centered), donor source (cadaveric vs living and cadaveric), geographic locations (European, North America, Asia and Oceania) and year period (prior to 1995 vs not prior to 1995) a P value for heterogeneity censored graft failure and all-cause mortality. The pooled results also indicated that HLA-DR mismatches were significantly associated with a 12% higher risk of overall graft failure. We also observed that HLA-A per mismatch was associated with a 6% higher risk of overall graft failure, but the association was insignificant. There was no significant association between HLA-B mismatching and graft survival. All included studies were in high methodological quality and the heterogeneity between studies was acceptable in each pooling analysis. In addition, we found that sample size or recipient ethnicity may be potential sources of heterogeneity.
Human HLA genes are located on chromosome 6 and code for 3 major class I alleles (HLA-A, -B, -C) and 3 major   [42,43]. As closely HLA-matched graft is less likely to be recognized and rejected, HLA mismatching has a substantial impact on prolongation of graft survival. With the emergence of potent immunosuppressive agents that steadily improved the graft survival rates, the impact of HLA compatibility seems to be minimized [42,44]. But the recent Australia and New Zealand Dialysis and Transplant Registry (ANZDTR) survey with 12,662 recipients still demonstrated that each incremental increase of HLA mismatches was significantly associated with higher risk of graft failure and rejection [27]. Another recent survey from Massie et al. [19] with 106,019 recipients from the Scientific Registry for Transplant Recipients (SRTR) database revealed that HLA-B and -DR mismatches were all significant associated with worse graft survival outcomes. Using multivariable-adjusted data (adjusting for other determinant confounders such as donor and recipient age, gender, combined disease, serum creatinine levels, ischemic times, etc.), the present analysis indicated that HLA per mismatch was associated with an increased risk of overall graft failure (9%), death-censored graft failure (6%) and all-cause mortality (4%). The pooled results were in favor of recommendations of the latest European Renal Best Practice Transplantation Guidelines, which recommended that matching of HLA-A, -B, and -DR whenever possible [8].
The meta-analysis suggested that HLA-DR per mismatch was significant associated with a 12% higher risk of overall graft failure. Besides, a subsequent analysis suggested that compared with 0 DR-mismatches, 1 and 2 mismatches were significant associated with 12 and 15% higher risk of overall graft failure, respectively. The pooled results were in favor of the kidney allocation guideline recommendations in almost all countries, such as the current US kidney allocation system, the revised United Kingdom kidney allocation scheme, and the latest European Renal Best Practice Transplantation Guidelines, which all highlighted the importance of HLA-DR testing [5][6][7][8].
Notably, the present analysis revealed a tendency that HLA-A mismatching had an impact on overall graft survival as there were only 3 studies included with a pooled HR of 1.06 (95% CI: 0.98-1.14). However, we did not observe a significant association between HLA-B mismatching and overall graft survival (HR: 1.01; 95% CI: 0.90-1. 15). Our pooled results were inconsistent with the recommendations of the revised United Kingdom kidney allocation scheme, which eliminated the impact of HLA-A similarity instead of HLA-B similarity [7]. Moreover, miscellaneous factors can result in inferior outcomes [45]. For instance, inferior graft outcomes could be related to high risk for rejection particularly antibody-mediated rejection [45][46][47]. Inferior patient survival could partly be associated with consequences of enhanced immunosuppression [45]. Consequently, the pooled results should be cautiously interpreted and further studies should be conducted to investigate the impact of HLA-A mismatching on graft and recipient survival outcomes.
Subgroup analysis and meta-regression was conducted to explore heterogeneity between studies. In subgroup analysis of the association between HLA per mismatch and overall graft failure, we found that after stratification for donor source (cadaveric vs living and cadaveric), the heterogeneity decreased to insignificant (I 2 = 0 and 16.4, respectively). But subsequent meta-regression analysis revealed that donor source did not change the overall effect significantly. In subgroup analysis of the association between HLA-DR Fig. 5 Forest plots of the association between HLA per mismatch and all-cause mortality mismatching and overall graft failure, we found that ethnicity and recipient sample size were potential source of heterogeneity. Large sample size of cohorts usually demonstrated more stable results. Besides, ethnic diversity was a potential source of heterogeneity probably because of varying HLA polymorphisms in the genetic makeup of the geographically distinct cohorts.

Strengths and limitations
Strengths of our meta-analysis are large sample of subjects (totally 486,608 recipients) and strict study design. Besides, we used multivariable-adjusted data for pooling analysis, which adjusted for some primary determinant confounders. However, the present meta-analysis had some limitations. Firstly, the absence of randomized controlled trials was the biggest limitation of this meta-analysis. Secondly, several studies have suggested that other HLA loci, such as HLA-C and -DQ locus, may contribute to poorer graft outcomes [48][49][50], but this meta-analysis only included the HLA-A, -B and -DR loci. Thirdly, heterogeneity is inevitable in some outcomes. We conducted several subgroup and metaregression analyses to explore the potential source of heterogeneity, and used random-effects models to incorporate heterogeneity between studies. Fourthly, few studies included could provide data about induction agent, maintenance agent or PRA, so that it cannot be achieved to do the stratified analysis.