Development and validation of risk models to predict chronic kidney disease among people living with HIV: protocol for a systematic review

Abstract

Introduction

Chronic kidney disease (CKD) is estimated to affect about 9.1% of the global population with a substantially increased risk of the condition (6.8%–17.2%) among people living with HIV (PLWH). This increased risk is attributed to HIV infection itself, antiretroviral therapy, coexisting viral infections, non-infectious comorbidities and traditional risk factors for CKD. Predictive models have been employed in the estimation of prevalent and incident CKD risk in both PLWH and the general population. A predictive model showing an individual’s risk of prevalent and/or progression to kidney failure is useful for initiating timely interventions that prevent further worsening of kidney function. This study will systematically review published prediction models developed and/or validated for prevalent and incident CKD in PLWH, describe their characteristics, compare performance and assess methodological quality and applicability.

Methods and analysis

Studies with predictive models of interest will be identified by searching MEDLINE, Web of Science, Cumulative Index to Nursing and Allied Health Literature, Cochrane library and Scopus from inception to May 2022. Title and abstract screening, full-text review and data extraction will be completed independently by two reviewers. Using appropriate tools designed for predictive modelling investigations, the included papers will be rigorously assessed for bias and applicability. Extracted data will be presented in tables, so that published prediction models can be compared qualitatively. Quantitative data on the predictive performance of these models will be synthesised with meta-analyses if appropriate.

Ethics and dissemination

The findings of the review will be disseminated in peer-reviewed journals and seminar presentations. Ethical approval is not required as this is a protocol for a systematic review.

PROSPERO registration number

CRD42021279694.

Correspondence to Dr Oluwatosin Olaseni Odubela; [email protected]

Strengths and limitations of this study

• This systematic review protocol follows the Preferred Reporting Items for Systematic Reviews and Meta-Analysis Protocols 2015 guidelines.

• This review addresses the knowledge gap regarding chronic kidney disease prediction models specific for people living with HIV.

• The review will also incorporate models predicting progression to kidney failure requiring kidney replacement therapy.

• This review is not limited to randomised controlled trials.

• In the absence of a sufficient number of studies/models, meta-analysis may not be performed.

Introduction

Chronic kidney disease (CKD) has become a global threat as it constitutes a sizeable proportion of morbidity and premature deaths.¹ CKD is estimated to affect 697.5 million people globally representing a global prevalence of 9.1%.¹ The traditional risk factors for CKD include increasing age, hypertension, diabetes mellitus and obesity. CKD is a predictor of poor quality of life as well as increased healthcare expenses.^{2 3}

There is a substantially increased risk of developing CKD (6.8%–17.2%) among people living with HIV (PLWH).^{4 5} This is corroborated by postulations of HIV infection and antiretroviral therapy (ART) regimens in the acquisition of CKD.^{6 7} CKD could arise among PLWH from the classic HIV-associated nephropathy or immune complex disease, non-infectious comorbidities (hypertension and diabetes), coexisting viral infections (Hepatitis B and C) and antiretroviral toxicity.^{8 9}

CKD is known to contribute to the increased morbidity and mortality among PLWH as the number of PLWH with kidney failure requiring kidney replacement therapy increases globally.^{2 3} CKD is an important non-infectious cause of morbidity and mortality among PLWH and could reverse the gains achieved via ART roll-out.^{10 11} Thus, risk assessments using predictive models remain a veritable tool for mitigating CKD in the populace (and HIV cohort).

Rationale

Predictive models have been developed to assist with CKD risk evaluation in both the HIV and general population. These predictive models have been employed to identify persons at greater risk of CKD (diagnosis or prognosis) and found to be helpful in clinical decision-making as well as public health interventions to mitigate against CKD.^{12 13} These models use biological markers (urinary protein and albumin, serum creatinine and cystine C, uric acid) along with other traditional risk factors for CKD (age, sex, blood pressure readings and diabetes mellitus) in their formulation. Early detection and management of hypertension, diabetes mellitus and CKD in this population can help improve renal and cardiovascular outcomes thereby regressing or slowing the progression towards kidney failure requiring kidney replacement therapy.

A few studies have reported predictive models for prevalent and incident CKD among PLWH as well as systematic reviews of CKD predictive models in the general population.^14–17 However, none of these studies have provided systematic reviews on CKD predictive models with the HIV population as the focus.

This protocol is for a systematic review with or without meta-analysis of the spectrum of prevalent and/or incident CKD prediction models developed and/or validated in PLWH, to identify existing gaps and guide research endeavours in the future.

Objective

This study aims to conduct a systematic review to identify and characterise predictive models developed and/or validated on prevalent and incident CKD in adult PLWH, evaluate the performances of these models in PLWH and identify existing knowledge gaps as reported in the published literature.

Methods and analysis

The protocol will be conducted in alignment with the Preferred Reporting Items for Systematic Reviews and Meta-Analysis Protocols 2015 guidelines (see online supplemental appendix 1).¹⁸ In addition, the CHARMS checklist will be employed to frame the review questions while the patient population, intervention, comparator, outcome, timing, and setting (PICOTS) criteria will be used to determine inclusion and exclusion criteria.^{19 20}

Review questions

The review questions are as follows: (1) are there risk models specifically developed to predict prevalent and/or incident CKD among PLWH; and what are their characteristics? and (2) are there models for predicting prevalent or incident CKD which have been validated in PLWH and how do they perform in this population? The CHARMS checklist was used in the formulation of the review questions.¹⁹

The above questions will form the basis of our review in identifying both prognostic and diagnostic models for CKD, with all types of prediction modelling studies included (development with or without external validation in independent data and external model validation with possible model updates). The scope of the review will be to inform clinical decision-making as it relates to prevalent and incident CKD among PLWH. The review will include diagnostic and prognostic models for CKD as well as models predicting progression to kidney failure requiring kidney replacement therapy with reference to HIV diagnosis. PLWH who are ≥18 years of age will be our target population.

Inclusion and exclusion criteria

The PICOTS framework will be employed in defining the inclusion and exclusion criteria for this review.²⁰ The population of interest is adult (≥18 years) PLWH with emphasis on predictive models for prevalent and incident CKD among the population. The primary outcomes are predictive models for (1) prevalent CKD, (2) incident CKD and (3) progression to kidney failure requiring kidney replacement therapy in PLWH. Our secondary outcomes were the evaluation of the performance of the above three groups of models as well as the candidate variables employed in their derivation. Studies to be included in the review are cross-sectional, cohort, clinical controlled and randomised controlled trials with the intent for the models to be used for clinical decision-making and public health advocacy measures.

The exclusion criteria will include the following. Studies primarily among paediatric, adolescents and pregnant women, case–control studies, editorials, letters to editors, models generated from simulation and animal studies as well as those evaluating quality of life among patients with CKD.

Search strategy

From inception to May 2022, a systematic search of the following electronic databases of peer-reviewed journal articles and online search records will be conducted: MEDLINE, Web of Science, Cumulative Index to Nursing and Allied Health Literature (CINAHL), Cochrane library and Scopus. Keywords relating to population (patients with HIV, PLWH, PLWH/AIDS, HIV infected, ART experienced, ART-naive); disease (impaired renal function, impaired kidney function, CKD, chronic renal disease, chronic renal insufficiency, chronic kidney failure, end-stage kidney disease, end-stage renal disease); modelling (prevalence, prevalent, incidence, incident, predict, risk, risk scores, prediction models, prediction tools, risk assessment, risk engineering) and models predicting CKD prevalence and incidence among PLWH as well as factors associated with the validity of these models. To accommodate each database, the search phrases will be concatenated. Reference lists of relevant papers will be scanned for eligible studies. A search strategy using MEDLINE is attached (see online supplemental appendix 2). Grey literature (such as reports, conferences and workshop proceedings) will be searched using the Google Scholar search engine as well as important relevant websites such as African Journals Online (hand searches). EndNote reference manager will be used to export references and delete any duplications identified.

Selection of studies

The title and abstract of each paper will be evaluated by two independent reviewers before being included in the review. The final selection of papers to be included in the review will be decided after reading the full texts of eligible articles. Disagreements will be resolved through discussion and consensus or consultation with a third reviewer.

A model/risk assessment tool that predicts prevalent and/or incident CKD, as well as models for progression to kidney failure requiring kidney replacement therapy, must be derived in an adult human population. The area under the receiver operating characteristic curve (AUC-ROC) or C-statistic, reclassification percentage, net reclassification improvement (NRI) or integrated discrimination improvement index (IDI) are some analyses being proposed for assessing the qualities of the models.

Assessment of studies

Diagnostic models are those designed for prevalent purposes while prognostic models are designed for incident purposes.

PROBAST (Prediction model Risk Of Bias ASsessment Tool) will be employed to evaluate the quality of all models. PROBAST evaluates both the risk of bias and the application of multivariable prediction (diagnostic and prognostic) models created or validated in primary research. The PROBAST tool will also be involved in the systematic reviews of predictive models for prevalent and incident CKD as well as progression to kidney failure requiring kidney replacement therapy in PLWH.^{21 22}

Disagreements in the use of the above tool for the review will be resolved following consultation with the third reviewer.

Data extraction

The following details will be used to extract data from chosen studies; study specifics (first author, journal name, year of publication, country of study); study population (sample size, age range, sex distribution, number of PLWH, ART-naive or experienced); population characteristics (pre-existing comorbid conditions: hypertension, diabetes mellitus, stroke, myocardial infarction, dyslipidaemia, congestive heart failure).

In addition, model characteristics (number of participants in the derivation and validation cohorts, number of participants with the outcome of interest, number of candidate variables stated as predictors, number and list of variables included in the model, type of statistical analysis used in generating model); study designs employed in deriving the models (cross-sectional, cohort, controlled clinical trials, randomised controlled trials); models’ outcome data (prevalence and incidence of CKD, number of multivariable prediction scores/models, definition of CKD, CKD equations used) and performance (discriminatory (AUC or C-statistic), calibration (difference between observed and predicted rates of hypertension, p value of corresponding test statistic), reclassification (NRI and IDI values with their accompanying 95% CIs and p values)) will also be documented. The source of funding and study limitations will be captured. The extracted data will be presented in a tabular (data) form.

Data analysis

The data will be summarised in general (globally) as well as by the WHO regional designation, gender and study population (PLWH, ART-naive or experienced). Meta-analysis will be conducted for CKD predictive models in the presence of an adequate number of models otherwise a narrative review will be undertaken. The meta-analysis would focus on the performance measures of the identified models with reference to the AUC-ROC, Kolmogorov-Smirnov test and other appropriate statistical tests. The random effects will be determined using inverse variance weighting and 95% CIs of pooled estimates while heterogeneity will be determined using the inconsistency index (I²).^{23 24} Publication bias will be assessed using funnel plots and when found to be significant, further analysis (Egger’s and Begg’s tests) will be conducted.²⁵

Patient and public involvement

This review does not require patient or public involvement as it will be based on published works. It will involve published articles from the above listed databases and search engines. The selected models predicting CKD would not primarily involve patients.

Ethics and dissemination

Ethical approval is not required as this is a protocol for a systematic review. The findings of this review will be published in peer-reviewed journals and included as a chapter in a PhD thesis at the University of Cape Town. Furthermore, the findings of this evaluation will be shared with relevant agencies through seminars, conferences and policy development meetings.

We appreciate the support of the Clinical Sciences Department, Nigerian Institute of Medical Research towards the actualisation of the review.

Ethics statements

Patient consent for publication

Not applicable.

Footnote

Contributors This systematic review was conceived and designed by OOO, NO and APK. This protocol was initially drafted by OOO, NP, AZM, NO, BLS and APK. Revisions and comments made to the protocol were approved by all authors. All authors have approved the submission.

Funding This project is part of the EDCTP2 programme supported by the European Union (grant number TMA2017GSF-1962—CaDERAL). The content hereof is the sole responsibility of the authors and does not necessarily represent the official views of the SAMRC, NIMR or the funders.

Competing interests None declared.

Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.

Provenance and peer review Not commissioned; externally peer reviewed.

Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.

References

¹ GBD Chronic Kidney Disease Collaboration. Global, regional, and national burden of chronic kidney disease, 1990-2017: a systematic analysis for the global burden of disease study 2017. Lancet 2020; 395: 709–33. doi:10.1016/S0140-6736(20)30045-3 http://www.ncbi.nlm.nih.gov/pubmed/32061315

² Hill NR, Fatoba ST, Oke JL, et al. Global prevalence of chronic kidney disease–a systematic review and meta-analysis. PLoS One 2016; 11: e0158765. doi:10.1371/journal.pone.0158765

³ Stanifer JW, Jing B, Tolan S, et al. The epidemiology of chronic kidney disease in sub-Saharan Africa: a systematic review and meta-analysis. Lancet Glob Health 2014; 2: e174–81. doi:10.1016/S2214-109X(14)70002-6

⁴ Ekrikpo UE, Kengne AP, Bello AK, et al. Chronic kidney disease in the global adult HIV-infected population: a systematic review and meta-analysis. PLoS One 2018; 13: e0195443. doi:10.1371/journal.pone.0195443 http://www.ncbi.nlm.nih.gov/pubmed/29659605

⁵ Bertoldi A, De Crignis E, Miserocchi A, et al. HIV and kidney: a dangerous liaison. New Microbiol 2017; 40: 1–10. http://www.ncbi.nlm.nih.gov/pubmed/28217815

⁶ Naicker SS, Rahmanian S, Kopp JB. HIV and chronic kidney disease. Clin Nephrol 2015; 83: S32–8. doi:10.5414/CNP83S032

⁷ Islam FM, Wu J, Jansson J, et al. Relative risk of renal disease among people living with HIV: a systematic review and meta-analysis. BMC Public Health 2012; 12: 234. doi:10.1186/1471-2458-12-234 http://www.ncbi.nlm.nih.gov/pubmed/22439731

⁸ Ekrikpo UE, Kengne AP, Akpan EE, et al. Prevalence and correlates of chronic kidney disease (CKD) among ART-naive HIV patients in the Niger-Delta region of Nigeria. Medicine 2018; 97: e0380. doi:10.1097/MD.0000000000010380 http://www.ncbi.nlm.nih.gov/pubmed/29668591

⁹ Chukwuonye II, Ogah OS, Anyabolu EN, et al. Prevalence of chronic kidney disease in Nigeria: systematic review of population-based studies. Int J Nephrol Renovasc Dis 2018; 11: 165–72. doi:10.2147/IJNRD.S162230 http://www.ncbi.nlm.nih.gov/pubmed/29861638

¹⁰ Kalyesubula R, Hau JP, Asiki G, et al. Impaired renal function in a rural Ugandan population cohort [version 3. peer review: 2 approved] view view 2018.

¹¹ Heron JE, Bagnis CI, Gracey DM. Contemporary issues and new challenges in chronic kidney disease amongst people living with HIV. AIDS Res Ther 2020; 17: 11. doi:10.1186/s12981-020-00266-3 http://www.ncbi.nlm.nih.gov/pubmed/32178687

¹² Chang H-L, Wu C-C, Lee S-P, et al. A predictive model for progression of CKD. Medicine 2019; 98: e16186. doi:10.1097/MD.0000000000016186 http://www.ncbi.nlm.nih.gov/pubmed/31261555

¹³ Dai D, Alvarez PJ, Woods SD. A predictive model for progression of chronic kidney disease to kidney failure using a large administrative claims database. Clinicoecon Outcomes Res 2021; 13: 475–86. doi:10.2147/CEOR.S313857 http://www.ncbi.nlm.nih.gov/pubmed/34113139

¹⁴ Kshirsagar AV, Bang H, Bomback AS, et al. A simple algorithm to predict incident kidney disease. Arch Intern Med 2008; 168: 2466–73. doi:10.1001/archinte.168.22.2466 http://www.ncbi.nlm.nih.gov/pubmed/19064831

¹⁵ Echouffo-Tcheugui JB, Kengne AP. Risk models to predict chronic kidney disease and its progression: a systematic review. PLoS Med 2012; 9: e1001344. doi:10.1371/journal.pmed.1001344 http://www.ncbi.nlm.nih.gov/pubmed/23185136

¹⁶ Tangri N, Kitsios GD, Inker LA, et al. Risk prediction models for patients with chronic kidney disease: a systematic review. Ann Intern Med 2013; 158: 596–603. doi:10.7326/0003-4819-158-8-201304160-00004 http://www.ncbi.nlm.nih.gov/pubmed/23588748

¹⁷ Han WM, Bijker R, Chandrasekaran E, et al. Validation of the D: a: D chronic kidney disease risk score model among people living with HIV in the Asia-Pacific. J Acquir Immune Defic Syndr 2020; 85: 489–97. doi:10.1097/QAI.0000000000002464 http://www.ncbi.nlm.nih.gov/pubmed/33136750

¹⁸ Moher D, Shamseer L, Clarke M, et al. Preferred reporting items for systematic review and meta-analysis protocols (PRISMA-P) 2015 statement. Syst Rev 2015; 4: 1. doi:10.1186/2046-4053-4-1 http://www.ncbi.nlm.nih.gov/pubmed/25554246

¹⁹ Moons KGM, de Groot JAH, Bouwmeester W, et al. Critical appraisal and data extraction for systematic reviews of prediction modelling studies: the charms checklist. PLoS Med 2014; 11: e1001744. doi:10.1371/journal.pmed.1001744 http://www.ncbi.nlm.nih.gov/pubmed/25314315

²⁰ Debray TPA, Damen JAAG, Snell KIE, et al. A guide to systematic review and meta-analysis of prediction model performance. BMJ 2017; 356: i6460. doi:10.1136/bmj.i6460 http://www.ncbi.nlm.nih.gov/pubmed/28057641

²¹ Wolff RF, Moons KGM, Riley RD, et al. PROBAST: a tool to assess the risk of bias and applicability of prediction model studies. Ann Intern Med 2019; 170: 51–8. doi:10.7326/M18-1376 http://www.ncbi.nlm.nih.gov/pubmed/30596875

²² Moons KGM, Wolff RF, Riley RD, et al. PROBAST: a tool to assess risk of bias and applicability of prediction model studies: explanation and elaboration. Ann Intern Med 2019; 170: W1–33. doi:10.7326/M18-1377 http://www.ncbi.nlm.nih.gov/pubmed/30596876

²³ Cochran WG. The combination of estimates from different experiments. Biometrics 1954; 10: 101–29. doi:10.2307/3001666

²⁴ Higgins JPT, Thompson SG. Quantifying heterogeneity in a meta-analysis. Stat Med 2002; 21: 1539–58. doi:10.1002/sim.1186 http://www.ncbi.nlm.nih.gov/pubmed/12111919

²⁵ Egger M, Davey Smith G, Schneider M, et al. Bias in meta-analysis detected by a simple, graphical test. BMJ 1997; 315: 629–34. doi:10.1136/bmj.315.7109.629 http://www.ncbi.nlm.nih.gov/pubmed/9310563