Complement-activating donor-specific anti-HLA antibodies in solid organ transplantation: systematic review, meta-analysis, and critical appraisal

Introduction Several studies have investigated the impact of circulating complement-activating anti-human leukocyte antigen donor-specific antibodies (anti-HLA DSAs) on organ transplant outcomes. However, a critical appraisal of these studies and a demonstration of the prognostic value of complement-activating status over anti-HLA DSA mean fluorescence intensity (MFI) level are lacking. Methods We conducted a systematic review, meta-analysis and critical appraisal evaluating the role of complement-activating anti-HLA DSAs on allograft outcomes in different solid organ transplants. We included studies through Medline, Cochrane, Scopus, and Embase since inception of databases till May 05, 2023. We evaluated allograft loss as the primary outcome, and allograft rejection as the secondary outcome. We used the Newcastle-Ottawa Scale and funnel plots to assess risk of bias and used bias adjustment methods when appropriate. We performed multiple subgroup analyses to account for sources of heterogeneity and studied the added value of complement assays over anti-HLA DSA MFI level. Results In total, 52 studies were included in the final meta-analysis (11,035 patients). Complement-activating anti-HLA DSAs were associated with an increased risk of allograft loss (HR 2.77; 95% CI 2.33-3.29, p<0.001; I²=46.2%), and allograft rejection (HR 4.98; 95% CI 2.96-8.36, p<0.01; I²=70.9%). These results remained significant after adjustment for potential sources of bias and across multiple subgroup analyses. After adjusting on pan-IgG anti-HLA DSA defined by the MFI levels, complement-activating anti-HLA DSAs were significantly and independently associated with an increased risk of allograft loss. Discussion We demonstrated in this systematic review, meta-analysis and critical appraisal the significant deleterious impact and the independent prognostic value of circulating complement-activating anti-HLA DSAs on solid organ transplant risk of allograft loss and rejection.


Introduction
Antibody-mediated rejection has been identified as the main cause for allograft loss (1) and the prognostic role of circulating antihuman leukocyte antigen donor-specific antibodies (anti-HLA DSAs) has been extensively assessed across different solid organ transplants (2-5). One key characteristic of anti-HLA DSAs is their ability to undergo class-switch recombination and activate complement by fixing complement fractions. Several studies have been conducted to evaluate the impact of complement-activating anti-HLA DSAs on allograft outcomes. The reported results were heterogeneous with some studies demonstrating a strong association of complementactivating anti-HLA DSA with adverse allograft outcomes (6, 7) while others showed no or weak associations (8,9).
As a consequence, our team previously performed a systematic review and meta-analysis to study the role of complementactivating anti-HLA DSAs on adverse allograft outcomes (10) and showed that circulating complement-activating anti-HLA DSAs increased the risk of allograft loss and rejection. However, since the publication of the review in May 2018, major studies assessing the effect of circulating complement-activating anti-HLA DSAs on allograft outcomes have been conducted (11,12).
In addition, the quality and risk of bias of the previous and recent studies have not been evaluated and a critical appraisal remains to be performed. The Sensitization in transplantation: Assessment of Risk (STAR) working group have recently highlighted several gaps regarding whether ancillary complementbased assays (C1q, C3d, C4d) provide additional useful clinical information compared to mean fluorescence intensity (MFI) values provided by single antigen bead (SAB) pan-IgG assay (6,8,13,14). Therefore, STAR working group recommends verify the role of complement binding assays in vivo as potential markers for adverse outcomes before recommending its use in clinical practice.
Therefore, the aim of this article was to provide a comprehensive up-to-date systematic review, meta-analysis and critical appraisal of studies testing the effect of circulating complement-activating anti-HLA DSAs on allograft outcomes and to evaluate and adjust for risk of bias.

Methods
This study was an incremental update of a systematic review and a meta-analysis (10), supplemented by a critical appraisal. The study was reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) (15).

Data sources
A comprehensive search strategy was conducted on Medline, Cochrane, Scopus and Embase since inception of databases till January 31, 2018 (10). For the period between the closing date of the previous review (10) and May 05, 2023 we created a search strategy using a complementary combination of two PubMed search strategies: 1) narrow Boolean which consists of the main Medical Subject Heading (MeSH) for the population combined with the main MeSH for the intervention (see Supplementary for details), and 2) ranking strategy which consisted of screening all the studies listed under the "similar articles" feature on PubMed of the three largest and three newest studies included in the previous review (16). We opted for a PubMed-only database search for this period because the included articles of the previous review (10), whose search strategy was comprehensive and included multiple databases, were all indexed in PubMed (17). This search strategy was further complemented by a manual search for potential additional studies.

Study selection
The inclusion criteria were studies evaluating the effect of complement-activating anti-HLA DSAs on allograft loss and rejection in adult and paediatric solid organ transplant recipients. Two independent reviewers (SAA and AB) screened the titles and abstracts of the studies and any disagreement was resolved by consensus.

Data extraction
We collected the same data variables as the previous review: "author name, year of publication, study size, mean or median follow-up time, mean age of population, type of complementactivating anti-HLA DSA, comparison used (patients with complement-activating anti-HLA DSAs were either compared to patients without complement-activating anti-HLA DSAs, patients with non-complement activating anti-HLA DSAs detected, or a mixed group of patients without anti-HLA DSAs and with noncomplement activating anti-HLA DSAs), effect sizes (HR and/or OR) and their 95% confidence intervals, potential confounding factors, and unadjusted and adjusted estimated risks of graft loss or graft rejection." (10).

Critical appraisal
We used the Newcastle-Ottawa Scale (NOS) to assess the risk of bias in observational studies (18). A high NOS score (≥ 6) represents high methodological quality. Using this quality score, each study is judged on eight items which are divided into three components: selection of the study groups (up to four points), confounding variables adjustment quality (up to two points) and the outcome studied (up to three points). (see Supplementary for details).
Extraction of data and assessment of risk of bias was done by two independent reviewers (SAA and AB) and any disagreement was resolved by consensus.

Data synthesis and analysis
We performed the meta-analysis through a random-effects model with restricted maximum likelihood approach using an inversevariance to incorporate a measure of the anticipated heterogeneity into the weight of the studies (19). The index group was complementactivating anti-HLA DSA positive patients. They were compared to either complement-activating anti-HLA DSA negative patients, anti-HLA DSA negative patients, or a mixed group of both. The pooled effect size, study weights and amount of study heterogeneity were represented by forest plots for allograft loss and rejection.

Statistical heterogeneity and smallstudy effects
We evaluated statistical heterogeneity using I² index which reflects the percentage of variability in the effect size caused by heterogeneity rather than by chance alone. An I² above 50% represented substantial heterogeneity (19).
We used a funnel plot to visually assess for the presence of small-size effects which occurs when smaller studies show different, often more pronounced effect size. We statistically assessed any asymmetry in the funnel plot with the Egger's test (20). If this test was significant, we adjusted for small-study effects by using the precision-effect test (PET). This method provided an estimate of the effect size in a study with a hypothetical infinite sample size and thus eliminating small-study effects bias (21).
We tested for publication bias by using a contour-enhanced funnel plot (22). If a bias was observed, we adjusted by using the puniform* selection model which assumes that studies with statistically non-significant p-values are published with the same probability as statistically significant results (21).

Subgroup analyses
We performed the following subgroup analyses to address potential sources of heterogeneity in studies assessing graft loss:

High versus low methodological quality of studies
We separately meta-analysed higher quality studies (NOS scores ≥ 6) (23) versus lower quality studies (NOS score ≤ 5).

Comparator group used
We separately meta-analysed studies comparing complement activating anti-HLA DSA positive patients with complement activating anti-HLA DSA and studies comparing complement activating anti-HLA DSA positive patients with complement activating anti-HLA DSA negative patients and anti-HLA DSA negative patient.

Type of organ transplanted
We separately meta-analysed studies based on the type of the transplanted organ (kidney, liver, lung, heart, pancreas and intestine). We also separately meta-analysed kidney transplants versus all other organs based on the assumption that a low number of studies are available per organ.

Timing of antibody detection
We separately meta-analysed studies testing patients with preformed anti-HLA DSAs (defined as antibodies positive before or at the time of transplantation), de novo anti-HLA DSAs (defined as antibodies positive only after transplantation), or a combined group of de novo both.

Sensitivity analysis
We separately meta-analysed the newly identified studies since the publication of the previous review in 2018 and assessed the association of complement activating anti-HLA DSA with the risk allograft loss and allograft rejection.

Cumulative meta-analysis
We conducted a cumulative meta-analysis to show the change of hazard ratio of allograft loss as each study is added to the pool (24), which allowed to assess the stability of evidence i.e., whether additional studies change the overall effect of complement-binding anti-HLA DSAs on the outcome, and the sufficiency of evidence i.e., whether additional studies were needed to establish the same conclusion (25). The cumulative meta-analysis was represented on a forest-plot and the studies were arranged in a chronological order by year and month of publication.
2.10 Added prognostic value of complement-activating anti-HLA DSA status over anti-HLA DSA MFI level We identified studies that showed a correlation between complement-activating anti-HLA DSA status and pan-IgG anti-HLA DSA defined by MFI levels. Then, we identified and separately meta-analysed studies that conducted multivariable analyses adjusting complement-activating anti-HLA DSA status on pan-IgG anti-HLA DSA defined by MFI levels to assess the prognostic value of complement-activating anti-HLA DSA over standard SAB pan-IgG assays.
In addition, to assess the added prognostic value of complementactivating anti-HLA DSA over EDTA treated SAB assays, we identified studies that pre-treated sera with ethylenediaminetetraacetic acid (EDTA) as means to overcome complement interferencea shortcoming of SAB assays caused by complement activation which usually results in underestimating or completely masking strong DSAs (26).

2.11
Added prognostic value of complement-activating anti-HLA DSA status over anti-HLA DSA class We identified and separately meta-analysed studies that performed multivariable models adjusting complement-activating anti-HLA DSA status on DSA class to assess the independent prognostic value of complement-activating anti-HLA DSA.
The meta-analyses were conducted on R 4.1.1. All tests were two-sided, and a p-value lower than 0.05 was considered significant.

Study identification
The search strategy identified 1,112 potential studies. After removing duplicates (n=91), studies with non-human data or not written in English (n=102), studies with non-solid organ transplant data (n=475), studies with non-complement binding anti-HLA DSAs (n=400), non-original articles (n=19), and studies with different outcomes or without hazard ratio/odds ratio (n=10), 15 new studies were identified, corresponding to 3,099 patients ( Figure 1). The previous review (10) included 37 studies, therefore, in this incremental update, 52 studies in total were included in the final meta-analysis, corresponding to 11,035 patients. A descriptive summary of all the included studies is shown in Table 1.

Risk for allograft loss
Patients with complement-activating anti-HLA DSAs had a 2.77-fold increase in risk for allograft loss (95% CI 2.33-3.29, p<0.001; I²=46.2%) compared to patients without complementactivating anti-HLA DSA, patients without anti-HLA DSAs, and a mixed group of both ( Figure 2).

Small-study effects
Visually, the funnel plot presented in Figure 4 showed an asymmetry which was confirmed by Egger's test (p=0.01) indicating the presence of small-study effects. When adjusting using the PET method, the hazard ratio remains significant (HR=1.5, p<0.001) indicating that in a hypothetical infinite sample size, complement-binding anti-HLA DSAs would still increase the risk for allograft loss (Supplementary Figure 3).
Publication bias, as a potential cause for small-study effects, was assessed using the contour-enhanced funnel plot presented in Figure 5 which showed that more studies lie in the statistically significant side of the graph. We adjusted for this bias by using the p-uniform* selection model which yielded a hazard ratio of 2.46 (p=0.01) indicating that taking into account studies with nonsignificant p-values, complement-binding anti-HLA DSAs would still increase the risk of allograft loss. Table 2 summarizes the effect sizes for each subgroup.

Effect of complement-activating anti-HLA DSAs in high methodological quality studies
Analysis done on high methodological quality studies (NOS≥6) showed a significantly increased risk of allograft loss in complement activating anti-HLA DSAs positive patients with a pooled HR of 2.79 (95% CI 2.33-3.35, p<0.001, I 2 = 45.7%). Studies with lower methodological quality (NOS ≤ 5) also showed an increased risk of allograft loss with a HR of 2.46 (CI 1.28-4.70, p<0.001) however, as expected, the heterogeneity level between the lower methodological quality studies was higher (I 2 = 60.5%).

Effect of the complement-activating anti-HLA DSAs using different comparators
The association between complement-activating anti-HLA DSAs and risk of allograft loss remained significant using different comparator groups. When comparing complementactivating anti-HLA DSAs positive patients to complementactivating anti-HLA DSAs negative patients, the pooled HR was 2.56 (95% CI 1.99-3.30, p<0.001, I 2 = 54.2%). When comparing, complement-activating anti-HLA DSA positive patients to a mixed group of complement-activating anti-HLA DSA negative patients and anti-HLA DSA negative patients, the pooled HR was 3.58 (95% CI 2.70-4.74, p<0.001; I 2 = 4.1%).

Effect of complement-activating anti-HLA DSAs according to the type of organ transplantation
Analysis done on kidney allograft recipients versus all other solid organ allograft recipients showed a significant increased risk of allograft loss with HRs of 2.77 (CI 2.25-3.41, p<0.001; I 2 = 49.2%) and 2.74 (CI 2.03-3.69, p<0.001; I 2 = 29.2%) respectively. Analysis specific to other Flow chart summarizing study identification and selection process.

Effect of complement-activating anti-HLA DSAs according to the timing of antibody detection
Analysis according to the time of antibody detection all showed significant associations with the highest HR of 3.53 for de novo DSAs (CI 2.63-4.74, p<0.001; I 2 = 26%).

Sensitivity analysis
The separate meta-analysis of the 15 newly identified studies since the publication of the previous review in 2018 showed that patients with complement-activating anti-HLA DSAs had a 2.21-fold increase in risk for allograft loss (95% Cl 1.61-3.04; p<0.001; I 2 = 58.8) (Supplementary Figure 1) and a 8.87-fold increase in risk for allograft rejection (95% CI 3.64-21.6; p<0.001; I 2 = 65.3%) compared to patients without complement-activating anti-HLA DSA, patients without anti-HLA DSAs, and a mixed group of both (Supplementary Figure 2). Association between complement-activating anti-HLA DSAs and the risk of rejection. The figure shows the forest plot of the association between complement activating anti-HLA DSAs and the risk of rejection for each study and overall (n = 17). Studies are listed by date of publication. The black square-shaped boxes represent the HR for each individual study. The black square-shaped boxes represent the HR for each individual study. The size of these boxes represents the weight of the study, and lines represent the 95% CI for individual studies. The diamond at the bottom represents the overall HR. CI, confidence interval; DSA, donor-specific antibody; HLA, human leukocyte antigen; HR, hazard ratio. Funnel plot representing the analysis for small-study effects. Each black dot represents a study; the x-axis represents the study effect size (hazard ratio), and the y-axis represents the standard error of the hazard ratio. The dashed vertical line represents the overall risk estimate and the black line represents the no intervention effect. Contour-enhanced funnel plot representing the analysis for publication bias according to the statistical significance of studies. Each black dot represents a study; the x-axis represents the study effect size (hazard ratio), and the y-axis represents the standard error of the hazard ratio.

Cumulative meta-analysis
The cumulative meta-analysis showed the effect of adding new studies in a chronological order on the overall effect size (Supplementary Figure 4). Starting at the second study in 2011 till the end of analysis, there is a consistent and statistically significant risk of allograft loss.
The cumulative meta-analysis demonstrated that adding new studies: i) narrowed the confidence intervals of the overall effect size, ii) reduced the already statistically significant p-values, iii) converged the overall effect size of complement-activating antibodies on allograft loss. DSA status on pan-IgG anti-HLA DSA defined by the MFI levels as opposed to a linear univariable correlation analysis. The multivariable analysis demonstrated that complement-activating anti-HLA DSA's presence was significantly and independently associated with an increased risk of allograft loss (HR 2.77; 95% CI 2.13-3.6, p=0.017; I 2 = 45.4%) ( Figure 6).
Seven (13.5%) studies pre-treated the sera of the studied population, or a sample of the studied population, with EDTA to uncover interfering substances and only 3 studies (5.8%) performed a multivariable analysis models adjusting complement-activating anti-HLA DSA status on EDTA treated pan-IgG anti-HLA DSA assays.
3.9 Added prognostic value of complement-activating anti-HLA DSA status over anti-HLA DSA class type Among the 29 (55.8%) studies that used multivariate analysis to evaluate the risk of allograft loss, only three (5.8%) studies included DSA class as a predictive variable. Among these three studies, two showed that HLA class II DR was significantly associated with graft loss. Complement activating anti-HLA DSA remained independently associated with an increased risk for graft loss HR=3.76 (CI=2.33-6.06; The black square-shaped boxes represent the HR for each individual study. The size of these boxes represents the weight of the study, and lines represent the 95% CI for individual studies. The diamond at the bottom represents the pooled HR. CI, confidence interval; DSA, donor-specific antibody; HLA, human leukocyte antigen; HR, hazard ratio. p=0.626; I2 = 0%) however, the results were statistically insignificant due to the low number of studies that included DSA type in the multivariable models.

Study overview
In this systematic review, meta-analysis, and critical appraisal including 11,035 solid organ recipients, we confirmed the increased risk of allograft failure and rejection associated with complementbinding anti-HLA DSAs. To the best of our knowledge, this is the first comprehensive systematic review and meta-analysis on the topic and the first in-depth critical appraisal assessing for the risk of bias, adjusting for it and providing several subgroup analyses to study the association of complement-binding anti-HLA DSAs with allograft outcomes. We also addressed the utility of complement-activating anti-HLA DSAs assessment over anti-HLA DSA MFI levels.

Subgroup analyses findings
This meta-analysis showed consistent results in multiple subgroup analyses. Complement-activating anti-HLA DSA were associated with an increased risk for allograft loss in higher quality studies, in different types of complement-activating anti-HLA DSAs (C1q, C3d, C4d and IgG3), at different times of evaluation for complement-activating anti-HLA DSA status (before and after transplantation) and at different MFI thresholds for complementactivating anti-HLA DSA positivity.

Cumulative meta-analysis findings
The cumulative meta-analysis further illustrated the significant overall effect of complement activating anti-HLA DSAs on allograft loss. Combining this finding with our findings from the subgroup analyses, we can perceive saturation of knowledge in particular in kidney transplant recipients and C1q evaluations. This is due to the fact that the majority of patients assessed were kidney recipients (78%) who were tested for C1q (54%) and therefore further research in this particular area could be redundant. However, there remains some areas that could benefit from further exploration, for instance, we did not identify any studies on the effect of complement-binding anti-HLA DSAs in pancreas and intestine transplants. In addition, more studies in liver, lung and heart recipients could be beneficial to confirm the initial findings by increasing the sample size and by comparing the risk of allograft loss across different organ transplants. Several studies in this meta-analysis and in the literature (53,71,72) indicated a strong correlation between complement-activating antibody status and anti-HLA DSA MFI level. Interestingly, studies included in this meta-analysis that performed multivariable analyses for the assessment of the independent prognostic value of complement-activating anti-HLA DSA adjusted on pan-IgG anti-HLA DSA defined by the MFI levels, showed that the association between C1q, C3d, C4d-binding tests or IgG3 test and allograft lost was independent of anti-HLA DSA MFI levels.
Although the absence of DSA complement binding antibodies should not be considered as a lack of the harmful effects of DSA in vivo, our meta-analysis supports a clinical utility of performing complement-binding assays. Indeed, the clinical impact remains significantly associated with graft loss independent of anti-HLA DSA MFI levels.
In addition to the uncertain association between the MFI levels and the clinical significance of an antibody, SAB pan-IgG assay remains a semi-quantitative test and technical limitations have been raised such as significant variations in repeated testing, between different laboratories (73), and due to various interfering substances (74). In addition, even though some studies addressed interfering substances by pretreatment of sera with EDTA (12,13,50,53,59), several limitations were noted; the EDTA concentrations were inconsistent across the studies, two studies only pretreated a small sample of the studied populations (4-8 patients), and the prognostic advantage of EDTA treated sera over complement assays was not demonstrated.
Therefore, our study shows that the use of complement binding anti-HLA DSA in clinical practice, in complement to MFI levels, which remains gold standard, could enhance risk stratification. We could not show independent association of complementactivating anti-HLA DSA status over HLA-DSA class due insufficient data published so far (only 3 studies). Further studies should therefore investigate the independent impact of class I or class II anti-HLA DSA regardless of their ability to activate complement, but also investigate the clinical impact of class I versus class II complement-activating anti-HLA DSA.

Implications
This study addresses several gaps highlighted by the STAR working group including the strong evidence regarding the prognostic role of complement-activating anti-HLA DSA in allograft rejection and loss, in complement to HLA-DSA titre and MFI assessment. This strongly supports a potential role for this test in clinical practice. and encourages interventional research regarding the role of certain drugs that target complement-dependent cytotoxicity as a prophylaxis and/or treatment of antibodymediated rejection and the value of a complement-activating anti-HLA DSA based strategy to monitor organ transplant patients to demonstrate clinical benefit and improvement of allograft survival.

Limitations
This study has the following limitations. First, we only included studies that provided a clear effect size for allograft loss or rejection (hazard or odds ratio). Second, No data was available from South America, Africa and Australia to reinforce the generalizability of the results. Third, all of the included studies were observational and retrospective. Finally, the review only included studies written in English.

Conclusion
The results of this systematic review, meta-analysis and critical appraisal support the significant and independent detrimental effects of complement-activating anti-HLA DSAs on allograft outcomes. This study highlights areas that need further exploration in complement-activating anti-HLA DSAs research, and encourages the clinical evaluation of complement-activating anti-HLA DSA testing to improve risk stratification and tailoring treatment regimens.

Data availability statement
The raw data supporting the conclusions of this article will be made available upon reasonable request.