Introduction

Renal transplantation is considered the optimal treatment for end-stage kidney disease. With the use of potent combinations of immunosuppressive agents, the survival rate of transplanted kidneys at the 1-year mark exceeds 95%.1 However, despite significant reductions in acute rejection rates, there has been a relatively modest progress in improving the long-term survival of the transplanted kidneys.2 The 10-year overall graft survival rate in deceased donor transplants was 42.3% from 1996 to 1999; it improved to 53.6% from 2008 to 2011.3

Several factors contribute to allograft failure, including acute T cell-mediated rejection (TCMR) or antibody-mediated rejection (ABMR), calcineurin inhibitor toxicity, BK virus nephritis, and interstitial fibrosis and tubular atrophy (IFTA) of unclear aetiology.4 IFTA of unclear aetiology is one of the primary causes of late allograft failure.5 The aetiology of IFTA is likely related to minor HLA mismatches between the donor and recipient, leading to immunological injury to the allograft manifesting as clinical or subclinical TCMR or ABMR6 among other contributory factors.

The term ‘subclinical rejection (SCR)’ of kidney allografts was coined in the 1990s.7 SCR is defined as the presence of histological findings meeting the criteria for acute rejection without accompanying kidney allograft dysfunction. Thus, SCR can be detected only in patients who undergo protocol biopsies. The incidence of SCR within the first year varies from 2.6% to 25%.8–10 There are several reasons for this discrepancy in the incidence of SCR. These stem from changes in immunosuppression over the years from cyclosporine-based regimens to tacrolimus-based regimens as well as changes from azathioprine to mycophenolate mofetil. Additionally, the differences in utilisation of induction agents as well as maintenance regimens and timing of biopsy impact the incidence of SCR. Due to this reason, the results from different centres are quite discrepant. Pooling of data from different centres may help identify some of these differences as well as provide insight into the development and progression of this relatively common condition.

Reports indicate that SCR may play a contributory role in the development of IFTA,11 suggesting an association with allograft dysfunction. Subclinical ABMR has been shown to have a negative impact on allograft survival in some studies,12 but data are limited to very few studies. The impact of subclinical TCMR on long-term allograft survival also remains a subject of debate. There have been a few studies that have looked at this but have been limited by duration of follow-up, differing biopsy protocols as well as immunosuppressive protocols.12–15 The difficulty in studying SCR also stems from a variety of other factors, including the paucity of institutions performing protocol biopsies, the demand for increased manpower in instituting a protocol biopsy programme as well as costs.16–18 This review will evaluate the association between SCR diagnosed with surveillance biopsy and long-term allograft survival after kidney transplant. Further stratification will be based on the type of rejection (subclinical ABMR, subclinical TCMR and borderline categories).

Methods and analysisLiterature search

This systematic review will be conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses Protocol.19 The protocol is registered in the International Prospective Register of Systematic Reviews (PROSPERO) with the identification number CRD42023463536.

The summary of the literature search and study selection process is encapsulated in figure 1. We will perform a systematic search of MEDLINE, EMBASE (Elsevier) and Cochrane Central from January 1995 to November 2023. A health sciences librarian with expertise in systematic review will develop all searches. The following keywords will be applied: renal transplantation, allograft function, SCR, subclinical inflammation, subclinical TCMR, subclinical ABMR, subclinical borderline rejection, graft survival, ABMR, cellular rejection and protocol biopsy. A combination of Medical Subject Headings terms and title, abstract and keywords will be used to develop the initial MEDLINE search, which will be validated against a known set of relevant studies. Boolean logic will connect the concept clusters, with proximity searching included for selected terms. Animal, paediatric, non-English and selected publication-type studies will be removed. The MEDLINE search will be adapted to the syntax of other databases. Grey literature will be covered through preprints in Pubmed and EMBASE, conference papers in EMBASE and trials in Cochrane Central. The Amsterdam Efficient Deduplication20 will be used during the search process to remove an initial set of duplicate records. In addition, bibliographies of included studies will be examined to identify additional relevant articles.

Figure 1

Overview of literature search, study selection process and data handling. This figure provides a visual representation of the methodology employed in conducting the literature search, selecting relevant studies and data handling. It outlines the systematic approach used to identify pertinent research articles and the subsequent process of study selection.

The search strategy including filters and limits is included as online supplemental file.

We will assess meta-bias with funnel plots, Begg and Mazumdar test and Egger’s test. In addition, we plan to conduct univariable meta-regression analyses to examine the effect of study size, age and sex distribution, causes of end-stage kidney disease, living donor status and HLA mismatches on outcomes.

Study selection

The inclusion criteria for studies will be as follows: prospective and retrospective observational studies or randomised controlled studies, including conference papers or posters; studies involving adult patients (>18 years old) who received kidney transplants (deceased or living donor), multiorgan transplants (eg, liver and kidney, pancreas and kidney, lung and kidney, heart and kidney, multivisceral transplant—intestine, liver, pancreas, kidney); studies that investigate patients who exhibited SCR detected through surveillance biopsies; studies that follow allograft outcomes, including subsequent allograft rejection or inflammation and allograft loss; studies that followed patients for at least 2 years post-transplant. Search will be limited to only English articles. There will be no geographical locations or economic status restrictions. Exclusion criteria are as follows: review articles or case reports; studies that include only ‘for-cause’ biopsies, which are those requested because of new-onset post-transplant renal dysfunction; studies involving non-human subjects; studies that lack available data needed to calculate HRs for the study groups, even after contacting the authors.

Study selection will be completed in DistillerSR (DSR) V.2023.8.2 (Evidence Partners, 2022). All study selection processes will be independently performed by project members.

All titles, abstracts and full-text articles will be screened by two screeners. In case of disagreement on exclusion, consensus will be reached through discussion by the two screeners. If consensus cannot be reached, a third project member (RM) will serve as the final adjudicator.

The results of the screening and study selection process will be reported in the final manuscript and presented in a flow diagram.21 A list of studies excluded at the full-text screening level (sorted by reason for exclusion) will be available as online supplemental material. Our study will not include any articles that involve executed prisoners, or that lack IRB approval or informed consent.

Outcomes

The primary outcome of this study will be death-censored allograft loss. The secondary outcome will include the development of subsequent rejection (TCMR or ABMR or mixed TCMR-ABMR) either clinical or subclinical and allograft loss from any cause (death or graft failure).

Data extraction and quality assessment

Prior to proceeding to the data extraction stage, each study selected for inclusion will be searched for in Retraction Watch (http://www.retractionwatch.com) and PubMed to identify retraction notices or notices of correction.

Data extraction will be completed in DSR. TY will develop a standardised form for data extraction, as well as a detailed instruction manual for training reviewers. All reviewers will perform data extraction; inter-rater reliability will be assessed and disagreements between reviewers will be resolved through discussion. In the event of uncertainty about missing or unclear study data, we will contact the study authors to clarify.

For assessment of risk of bias in randomised controlled trials, we will use the Cochrane Collaboration’s revised tool for assessing the risk of bias in randomised trials (RoB 2). RoB 2 includes sequence generation, allocation concealment, blinding, incomplete outcome data and selective outcome reporting. The judgements were categorised as ‘low risk’, ‘high risk’, or ‘unclear’ based on the extracted information.

For cohort studies, we will use the Newcastle-Ottawa Scale to assess the risk of bias. The Newcastle-Ottawa Scale was developed to assess the quality of non-randomised studies. It uses a ‘star system’ in which studies are judged on three aspects: selection of the study groups; comparability of the groups and ascertainment of either the exposure or outcome of interest for case–control or cohort studies, respectively. Each domain will be evaluated and rated based on the extent of bias present. The overall risk of bias in the study will be determined by considering the ratings across all domains. The overall risk of bias will be judged based on the answers to the signalling questions, using an inbuilt algorithm. We will exclude studies judged as ‘high risk of bias’ and ‘very high risk of bias’.

Statistical analysis

We will perform a meta-analysis using the ‘meta’ package (V.1.1–0) of R programming language (R Foundation for Statistical Computing, Vienna, Austria) and Comprehensive Meta-Analysis V.4 (Biostat, Englewood, New Jersey, USA). For time-dependent outcomes, we will prioritise using HRs and corresponding 95% CIs obtained directly from individual articles whenever available. If HRs are not reported, we will explore options for calculating HRs using Cox regression analysis based on available data. In cases where HRs are unavailable, we will calculate risk ratios based on the recorded events.

We will employ a random effects model.

We plan to employ a random effects model to estimate the overall risk. We will evaluate heterogeneity among the included studies using the probability value of the I2 variable. We will classify heterogeneity as low, moderate or high based on I2 values of 25%, 50% or 75%, respectively.

We plan to conduct univariable meta-regression analyses to explore the impact of study-level variables. Specifically, we will examine the effect of study size, age distribution, sex distribution, causes of end-stage kidney disease, living donor status and HLA mismatches on the association between subsequent rejection and allograft loss.

We will assess the potential presence of publication bias by examining the symmetry of funnel plots. Additionally, the Begg and Mazumdar rank correlation test and the Egger regression test will be used to quantitatively assess publication bias.

We will also perform subanalyses to differentiate the effects of subclinical ABMR versus subclinical TCMR in addition to differentiation between varying degrees of subclinical TCMR (ie, Banff IA vs borderline changes).

In order to assess the robustness of our findings and ensure the reliability of our conclusions, we will conduct a sensitivity analysis focusing on the quality of studies included in our review. The quality of each study will be evaluated using predefined criteria that encompass methodological rigour, sample size and the comprehensiveness of outcome reporting. By stratifying the studies based on these quality metrics, we can determine the influence of study quality on the overall results. This sensitivity analysis will help us identify whether conclusions are disproportionately driven by higher or lower quality studies, thereby providing a better understanding of the strength and validity of the evidence presented in our systematic review.

Patient and public involvement

None.

Ethics and dissemination

Ethics approval does not apply as no original data will be collected. The results will be disseminated through conference presentations and peer-reviewed publications.

Discussion

Since the early descriptions of SCR over 30 years ago, there have been many studies looking at short-term effects of SCR, including changes in chronicity on biopsies, changes in renal function and occurrence of donor-specific antibodies. However, there has been a paucity of data looking at the long-term impact of SCR. As outlined earlier, the number of centres performing protocol biopsies is relatively limited. To add to this, the variations in induction and maintenance protocols, timing of biopsy and variations in treatment have made it difficult to compare outcomes between institutions. Moreover, many studies report single-centre data and there is also variation in pathology interpretations of inflammatory infiltrates and their quantification.

We, therefore, feel that pooling of studies and studies of large numbers of patients would perhaps be the best way to answer this important question. It is imperative to understand that while SCR is an entity that is well described, there are certain nuances to be aware of while trying to decipher the true meaning of these infiltrates. For instance, the prognostic implications of subclinical ABMR are likely different from subclinical TCMR. Moreover, studies will need to be carefully screened to examine if they included borderline changes in their definition of SCR or whether rejections Banff IA or greater were included. We will try to tease out these finer differences depending on the findings of the data extraction. Lastly, it is important to mention that the term ‘subclinical inflammation’11 22 23 has often been used in place of SCR.

We hope that a systematic examination of all these factors would help stratify the different varieties and severities of SCR and their prognostic implications.

A major strength of such a study is the ability to examine outcomes from different institutions. The subsequent pooling of data helps increase the number of patients, especially given the low numbers of patients with SCR that have been systematically studied. Pooling of data from different centres does bring in challenges of different immunosuppressive regimens, inclusion of studies from different eras of immunosuppression, all of which would be limitations of such a study. Additionally, the inclusion of multiorgan transplants in addition to ‘kidney alone’ transplants is a limitation since the outcomes in these situations may be variable. However, the proportion of multiorgan transplants is expected to be small.

Ultimately, the real question then that needs to be answered is whether the implementation of a protocol biopsy programme would be effective in prolonging the longevity of the renal allograft. This systematic review and meta-analysis aims to provide valuable insights into the implications of protocol biopsy.