Scolaris Content Display Scolaris Content Display

Predictive factors for BK polyomavirus infection in solid organ transplant recipients

Collapse all Expand all

Abstract

Objectives

This is a protocol for a Cochrane Review (prognosis). The objectives are as follows:

To (a) define the prognostic factors that predict the development of BKPyV infections in kidney and kidney‐pancreas transplant recipients, and (b) explore sources of heterogeneity.

Background

Description of the health condition and context

BK polyomavirus (BKPyV) is a ubiquitous deoxyribonucleic acid (DNA) virus that remains quiescent in renal and uroepithelial tissue after primary infection and is generally harmless in immunocompetent hosts. Most of the population has been exposed during childhood and BKPyV antibodies (indicating prior exposure) can be detected in 80% to 90% of the general population (Antonsson 2010Knowles 2003). There have been isolated reports of BKPyV affecting the native kidneys of immunosuppressed patients (Al Zein 2019Shah 2019) but the overwhelming majority of infection is seen in the allografts of kidney transplant recipients. With the advent of increasingly potent immunosuppressive regimens after kidney transplantation, BKPyV infection has become a major cause of graft dysfunction and loss. 

The earliest manifestation of the infection is detectable BKPyV in the urine (viruria). About 50% of patients with viruria will progress to develop BKPyV in their blood (viraemia) (Laskin 2010) and without intervention, the virus will propagate and may result in infiltration of the allograft leading to significant graft dysfunction and potentially graft loss. Biopsy‐proven disease is known as BKPyV‐associated nephropathy (BKPyVAN). The prevalence of BKPyVAN in the post‐kidney population is approximately 1% to 10% (Dharnidharka 2009Nickeleit 2000Schold 2009) but the prevalence of viruria and viraemia is much higher, with 25% to 35% (Brennan 2005Manzano 2019) and 10% to 20% (Brochot 2019Madden 2018) of recipients having detectable BKPyV in their urine and blood respectively, at some point post‐transplant. Graft dysfunction in the setting of high viral loads in the urine or blood is often presumed to be due to BKPyV infection even without confirmatory biopsy‐proven disease. Initial BKPyVAN outcomes were dismal with up to 50% of patients losing their graft within five years of diagnosis (Vasudev 2005), but recent outcomes have improved (Manzano 2019) temporally corresponding with the widespread adoption of screening protocols to detect BKPyV at an earlier, pre‐clinical stage. 

Because of the asymptomatic nature of BKPyV infection, and the severe consequences if left untreated, most transplant centres have adopted screening protocols using highly sensitive polymerase chain reaction (PCR) testing to detect early low‐level BKPyV in the urine or blood before overt nephropathy develops. Infection is most likely to occur in the first year post‐transplant (Dharnidharka 2009Hocker 2019) when cellular immunity is most compromised due to high initial doses of immunosuppression. As such, current guidelines recommend that screening protocols are initiated one month post‐transplant and continue periodically for two years (Hirsch 2019KDIGO 2009). PCR testing of blood is preferred over urine due to its superior specificity for clinically relevant infection (Hirsch 2002) and efficacy in monitoring response to therapy (Brennan 2005). 

The mainstay of BKPyV treatment is immunosuppression reduction (Hirsch 2019KDIGO 2009). To date, there are no pharmacological therapies that have been proven effective at improving graft outcomes in those with BKPyV infection. Given that both BKPyV infection and allograft rejection are most common in the early stages post‐transplant, immunosuppression reductions in response to BKPyV infection increase the risk of allograft rejection (Hardinger 2010). In this context management decisions usually involve trying to determine the optimal balance between immunosuppression reduction facilitating viral clearance without triggering allograft rejection and graft dysfunction. The degree of immunosuppression reduction is dependent upon the severity of the infection and the individual patient’s immunological risk for rejection. 

Due to the lack of therapeutic options and the poor graft outcomes, once BKPyVAN is established, a thorough understanding of risk factors for BKPyV infection assumes additional importance. If potentially modifiable risk factors can be reliably identified, avoidance of these factors may prevent the development of BKPyV in at‐risk patients and improve outcomes throughout the transplant population by allowing for a more personalised approach to the management of kidney transplant recipients.

Description of the prognostic factors

After primary infection, BKPyV remains dormant in renal and uroepithelial cells (Chesters 1983Drachenberg 1999Heritage 1981). The impaired cellular immune response due to immunosuppression in the early post‐transplant period allows for the uncontrolled replication of the virus. As BKPyV overwhelmingly affects transplanted allografts (as opposed to native kidneys), additional factors that induce local inflammation at the site of transplantation (Hashim 2014), and donor characteristics (Verghese 2015), are likely to play a key role in the pathogenesis of this condition. 

Over 140 factors have been reported in association with the development of BKPyV infection (Demey 2018) but the majority of these are based on small, retrospective studies with frequently conflicting results. Almost all immunosuppressive agents have been associated, however, the strongest evidence supports a link between tacrolimus usage and BKPyV (Hirsch 2013Mengel 2003) with the proposed pathogenesis involving activation of the FKBP‐12 pathway (Hirsch 2016). Male sex and increasing age have consistently been shown to increase the risk of BKPyV infection (Dharnidharka 2009Kayler 2013Thangaraju 2016) likely due to the reduction in the antibody response to BKPyV infection with increasing age (Gossai 2016Kean 2009). 

The most clinically relevant prognostic factors have been selected to be investigated in this review. These can be categorised as relating to the recipient, the donor, or the transplant itself. Apart from age, these factors are static and will be determined by their respective values at the time of transplantation or in the immediate days thereafter. With the exception of delayed graft function, all other factors are known at the time of transplant. Delayed graft function refers to poor function of the transplanted kidney within the first week of the post‐transplant period and is recorded as a dichotomous outcome at that point. ABO‐incompatibility and the number of human leukocyte antigen (HLA) mismatches indicate the degree of immunological difference between donor and recipient and will not change over time. 

Of particular interest are potentially modifiable transplant factors. If strong associations are found, transplant recipients’ management could be tailored according to the individual patient’s risk profile to avoid certain factors resulting in a reduction in the risk of BKPyV development. 

  • Recipient factors 

    • Age, sex and ethnicity

    • Primary cause of kidney failure 

    • Patient’s first or subsequent graft

  • Donor factors 

    • Age, sex and ethnicity

  • Transplant factors

    • Ureteric stent placement at the time of transplant 

    • Initial immunosuppression regimen at the time of transplant

    • Delayed graft function

    • ABO‐incompatibility

    • Number of HLA mismatches 

Health outcomes

Why it is important to do this review

BKPyV infection is an increasingly common cause of graft loss and currently, no effective drug treatments exist. Vast heterogeneity in treatment practices exists amongst transplant clinicians and the evidence regarding when, and how, to treat BKPyV is sparse and inconclusive. Even with all appropriate measures, BKPyV infection can still result in graft loss due to uncontrolled viral replication or graft rejection secondary to the obligatory immunosuppression reduction. Despite widespread screening for BKPyV resulting in improvements in graft outcomes, rates of graft loss remain unacceptably high. 

Findings from this review will directly impact clinical practice. Should positive associations be found, this will allow transplant clinicians to risk stratify patients according to the presence or absence of certain factors. Should a patient be deemed extremely high risk (e.g. previous graft loss due to BKPyV) then immunosuppression could be tailored, certain factors avoided and donors with appropriate characteristics chosen to minimise the future risk of BKPyV infection. Modifiable factors could be tested in prospective research and become standard clinical practice.

Objectives

To (a) define the prognostic factors that predict the development of BKPyV infections in kidney and kidney‐pancreas transplant recipients, and (b) explore sources of heterogeneity.

Methods

This review will be conducted within the framework of the Cochrane Kidney and Transplant review group and reported in line with the PRISMA guidelines (Page 2021) and the protocol will follow the guidance of the CHARMS checklist (Moons 2014).

The 'Methods' section is based on the exemplar Cochrane Prognosis Review protocol for prognostic factors (Hayden 2019), and the general protocol template of the Cochrane Prognosis Methods Group.

Criteria for considering studies for this review

The criteria for considering studies for inclusion in this review will follow the PICOTS method (Moher 2009).

Types of studies

Inclusion criteria 

  • Studies that investigate the association between any risk factor and BKPyV as an outcome (viruria, viraemia, nephropathy)

  • Studies with a minimum duration of follow‐up of six months with no maximum duration of follow‐up

  • Observational study designs: all non‐randomised cohort studies (retrospective and prospective), registry data studies, prognostic studies and case‐control studies where the prognostic factors are well‐defined

  • Randomised study designs: studies where associations between risk factors and BKPyV outcomes are clearly reported

  • Studies in any language

  • Full peer‐reviewed studies.

Exclusion criteria 

  • Abstracts without peer review due to the inability to sufficiently assess information for quality assessment or outcomes 

  • Grey literature such as unpublished abstracts, dissertations, articles or theses.

Targeted population

  • Adults and children of all ages 

  • Kidney transplant and kidney‐pancreas transplant recipients (excluding other multiorgan recipients)

  • First or subsequent grafts. 

Types of predictive factors

Index

  • Recipient factors that have been associated with BKPyV: age, sex, ethnicity and primary kidney disease leading to kidney failure

  • Donor factors previously associated with BKPyV: age and sex

  • Transplant factors are potentially modifiable risk factors that have previously been linked with BKPyV infection: initial immunosuppressive regimen including choice of calcineurin inhibitor (CNI), anti‐metabolite or mammalian target of rapamycin‐inhibitor (mTOR‐I) use. The number of HLA mismatches between donor and recipient, ABO incompatibility, graft number for the individual recipient, ureteral stent placement at the time of transplantation and delayed graft function.

Timing

  • All predictive factors will be measured at the time of transplantation or in the first week thereafter.

Outcomes of interest

  • Viruria: detectable BKPyV in the urine using polymerase chain reaction (PCR) testing.

  • Viraemia: detectable BKPyV in the blood using PCR testing. 

  • BKPyVAN: biopsy‐proven infiltration of kidney tissue by BKPyV causing graft dysfunction in accordance with The Banff Classification (Nickeleit 2018).  

  • Presumed nephropathy: graft dysfunction that has been attributed to BKPyV in a study where the BKPyV outcome (viruria, viraemia or BKPyVAN as defined above) isn't specified. 

Timing

  • Studies that have a minimum of six months of follow‐up post‐transplant will be included and there will be no limit to the maximum duration of follow‐up.

Search methods for identification of studies

Electronic searches

The following databases will be searched in consultation with the Cochrane Kidney and Transplant Information Specialist, with no date or language restrictions.

  1. MEDLINE via Ovid (1946 – search date), using the strategy in Appendix 1.

  2. EMBASE via Ovid (1974 – search date), using the strategy in Appendix 1.

  3. ClinicalTrials.gov and The World Health Organization (WHO) International Clinical Trials Registry Platform (ICTRP) using keywords based on the strategies for MEDLINE and EMBASE as in Appendix 1.

Searching other resources

We will search for additional relevant studies in the reference lists of included studies and any relevant systematic reviews identified in the search. 

Data collection

Selection of studies

Two authors (RG, CC) will independently screen titles and abstracts to determine the eligibility of studies identified by the electronic searches and clearly irrelevant articles will be removed. Full‐text articles will then be retrieved for those articles deemed potentially relevant or where relevance cannot easily be established. Two authors (RG, CC) will independently assess for eligibility and discuss any differences in the determination of article eligibility. If there is a difference of opinion regarding the inclusion of particular studies, then a third author (GW) will be consulted for further clarification. 

We will outline the study selection process in a PRISMA study flow diagram (Moher 2009). 

Data extraction and management

We will extract data from included studies using a data extraction form based on the CHecklist for critical Appraisal and data extraction for systematic Reviews of prediction Modelling Studies (CHARMS) (Moons 2014); it will be piloted on five studies chosen at random with appropriate edits being made if required. Two authors (RG, CC) will independently extract the data and disagreements will be resolved through discussion, or if required, by consulting with a third author (GW).

List of data to extract

  • Data relevant to the prognostic factors: donor age, sex and ethnicity; recipient age, sex and ethnicity; primary disease leading to kidney failure; initial induction agent, CNI and anti‐metabolite; initial use of mTOR‐I use (Y/N); first or subsequent transplant; ureteral stent use and duration of the stent; delayed graft function; ABO‐incompatibility and number of HLA mismatches

  • Data relevant to the outcomes: effect estimates of the relevant prognostic factor e.g. odds ratios (OR), hazard ratios (HR) or risk ratios (RR); outcome being measured e.g. viruria, viraemia or nephropathy

  • Characteristics of the study: design; dates; timing of outcome post‐transplant; duration of follow‐up; location; inclusion and exclusion criteria; participant baseline characteristics

  • Using a standardised template, we will contact trial authors for missing data. We will allow 30 days for a response, after which time, we will only include published data for the purposes of this review.

  • Adjusted and unadjusted effect estimates will be extracted from each study. If possible, effect estimates will be converted into a standard form to facilitate the inclusion of as much data as possible and avoid potential selection bias. 

Assessment of risk of bias in included studies

As recommended by The Cochrane Collaboration Prognosis Methods Group (Moons 2018) a standardised approach to the assessment of bias in observational studies will be undertaken based on the QUality in Prognostic Studies (Hayden 2013) (QUIPS) tool (see Appendix 2). This is a modified version of the Cochrane Risk of Bias Tool (Higgins 2022) that has been designed for prognostic reviews. Risk of bias assessments will be performed independently by two authors (RG, CC) with disagreements being arbitrated by a third author (GW). A pilot test of the assessment tool will be performed using five randomly selected studies to ensure consistency across the review team. 

The QUIPS tool assesses study bias based on the following six domains:

  1. Study participation

  2. Study attrition

  3. Prognostic factor measurement

  4. Outcome measurement

  5. Study confounding

  6. Statistical analysis and reporting.

Each study included will be graded as high, medium, or low risk of bias in each domain. Detailed evidence to support the classification of each study into a particular category will be given. We will also judge the overall risk of bias, by defining studies as having a low risk of bias if they achieve a low risk of bias score in each separate domain.  Should information be lacking to allow appropriate decisions regarding the degree of bias then authors of the affected trials will be directly contacted for further details. 

Measures of association or predictive performance measures to be extracted

We will extract the unadjusted and adjusted prognostic effect estimates including HR, RR, OR or mean differences (MD) along with their corresponding measures of uncertainty such as standard errors (SE), variances and 95% confidence intervals (CI). ORs will be the preferred outcome measure, and we will convert RRs and HRs to ORs when possible. If estimates for the prognostic factors and outcomes are reported as dichotomous variables (frequencies), we will extract the data in the format of 2 x 2 tables and convert them to effect estimates. We will also note any non‐linear relationship between the prognostic factor and outcome, and whether the modelling assumptions such as the proportional hazard hold for each prognostic factor of interest (Moons 2014Riley 2019). For consistency, parameter estimates will be calculated and reported in the same direction. 

Dealing with missing data

Details related to missing data will be collected and reported including the number of participants with missing data and reasons for missing values.  If there is missing data we will contact the relevant authors using a standardised template and allow 30 days for responses. Missing data will not be imputed.

We will calculate or estimate effect sizes using the available data including those reported in 2 x 2 frequency tables, graphs, and figures such as Kaplan‐Meier curves (Parmar 1998Tierney 2007).

Assessment of heterogeneity

We anticipate statistical heterogeneity due to clinical and methodological differences between studies. Hence, a random‐effects model will be used for the meta‐analysis. We will assess clinical heterogeneity by comparing important participant factors at a study level, and methodological heterogeneity by comparing the risk of bias of studies, taking into account study participation, participant attrition and outcome measurement factors across the studies. Variability of the estimates will be displayed using a forest plot (Riley 2019). We will quantify heterogeneity using the I² statistic and Tau².

Assessment of reporting deficiencies

We will conduct funnel plot asymmetry tests provided there are at least 10 studies to maintain sufficient power to distinguish the occurrence of chance from real asymmetry. We will also test for asymmetry at the 10% level using Egger’s test for HRs and Peters’ test for ORs (Debray 2018).

Data synthesis 

Data synthesis and meta‐analysis

Separate analyses will be conducted for the unadjusted and adjusted associations (Riley 2019). In the univariate analyses, each prognostic factor will be analysed separately with each outcome of interest. If appropriate, we will pool these effect estimates (HRs, ORs, or RRs) where the study designs, participants, and the prognostic factor definition are sufficiently homogenous to allow for meta‐analysis. The random‐effects approach will be used to account for between‐study heterogeneity.

If multivariable models from different studies adjust for the same variables (e.g. age, sex, ethnicity) we will combine the adjusted effect estimates to determine the overall size of the effect.

If data are not available in the text, tables, or from the author then we will graphically extract data using the free, open‐source software WebPlotDigitizer (Rohatgi 2021). Previous studies have shown the accuracy of this method when data is not available in traditional forms (Van de Mierden 2021). 

If the study results cannot be combined due to excess clinical heterogeneity or lack of appropriate data, we will present the results as a tabulated summary, and analyse them descriptively.

Subgroup analysis and investigation of heterogeneity

If appropriate, we will conduct meta‐analyses on subgroups to explore the causes of heterogeneity. However, we anticipate that the available data may be small and lack power to identify genuine causes of heterogeneity. Dependent upon the final distribution of studies, potential subgroup analyses that may take place include:

  • Multi‐organ (including kidney) versus single‐organ kidney transplantation (given the differing intensities of immunosuppression regimens between groups) 

  • Outcome measurement: viruria, viraemia, nephropathy and "likely" nephropathy

  • Age: children versus adults

  • According to geographical region.

Sensitivity analysis

If feasible, we will perform the following sensitivity analyses to examine the effect of certain factors on the magnitude of effect estimates.

  • Including only those studies with a low risk of bias

  • Excluding very large or prolonged studies (> 10 years duration) to assess for overwhelming influence of these studies on outcomes

  • Repeating the analyses separately according to study type (e.g. single versus multi‐centre, randomised versus non‐randomised).

Conclusions and summary findings

We will prepare a ‘Summary of findings’ table outlining the association between each prognostic factor and BKPyV infection including our confidence in the estimates of the effect and certainty of evidence using GRADE modified for prognostic factor studies (Foroutan 2020). The GRADE approach defines the certainty of a body of evidence as to the extent to which one can be confident that an estimate of effect or association is close to the true quantity of specific interest. This will be assessed by two authors (RG & CC). A summary of the GRADE approach is in Appendix 3.

For each risk factor, we plan to present the following outcomes in the ‘Summary of findings’ tables.

  • Viruria 

  • Viraemia 

  • Nephropathy

  • Presumed nephropathy 

We will rate the overall strength of evidence considering the risk of bias, inconsistency, indirectness, imprecision, publication bias, effect size, dose‐response gradient and the nature of plausible biases. We will rank evidence as high, moderate, low or very low. For observational evidence, we will begin with high certainty and reduce the grade as appropriate.