Incidence of impaired kidney function among people with HIV: a systematic review and meta-analysis

Background In the era of combination antiretroviral therapy (ART), the incidence, manifestations and severity of kidney diseases have dramatically changed in people living with HIV (PLWH). Little is known about the incidence of impaired kidney function (IKF) measured by serum creatine-based estimated glomerular filtration rate (eGFR) in PLWH. Methods In this systematic review and meta-analysis, we searched PubMed, Ovid, Medline, Embase and Web of Science for studies published before May 7th, 2021, with estimates of incidence of IKF among PLWH. We independently reviewed each study for quality by using the Newcastle-Ottawa scale. The incidence and 95% confidence intervals (CIs) were calculated using random-effects model. Results Sixty out of 3797 identifiable studies were eligible for the meta-analysis. A total of 19 definitions of IKF were described and categorized into three types: the threshold of eGFR, an absolute or percent decrease in eGFR, and certain eGFR threshold combined with decrement in eGFR. The eGFR< 60 ml/min/1.73m2 was the most widely used definition or criterion for IKF, by which the pooled incidence rate of IKF was 12.50 (95%CI: 9.00–17.36) per 1000 person years (PYs). The second most-studied outcome was a > 25% decrease in eGFR, followed by eGFR< 90 ml/min/1.73m2, eGFR< 30 ml/min/1.73m2 and a combination of eGFR threshold plus decreased eGFR. The reported incidence rates of IKF differ widely by different definitions of IKF. The highest pooled incidence was observed for those with > 25% decrease in eGFR, while the lowest was observed in those with eGFR < 30 ml/min/1.73m2. Substantial heterogeneity was identified across most estimates. Conclusion Our study provides a comprehensive summary of eGFR-based definitions and incidence rates of IKF in PLWH, not only promoting our understanding of IKF, but also underscoring needs for a concerted action to unify definitions and outcomes of IKF and their applications in AIDS care. Supplementary Information The online version contains supplementary material available at 10.1186/s12882-022-02721-x.


Introduction
There are approximately 37.7 million people living with HIV (PLWH) worldwide, with 1.5 million newly reported in 2020 [1]. At the end of December 2020, 27.5 million PLWH were accessing antiretroviral therapy (ART) globally [1]. ART has led to considerable improvement in the life expectancy of PLWH [1][2][3]. Accordingly, impaired kidney function (IKF) arises as a consequence of aging, therapy of HIV infection and its complications, which is not only associated with significant morbidity and mortality in PLWH, but also their quality of life [4][5][6].
IKF is a generic term of various kidney dysfunctions or diseases ranging from asymptomatic changes in kidney Open Access *Correspondence: nhe@fudan.edu.cn 2 Yiwu Research Institute of Fudan University, Shanghai, China Full list of author information is available at the end of the article function to severe end-stage kidney diseases (ESKD). The most typical type of the kidney disease by HIV infection before the availability of ART is HIV-associated nephropathy (HIVAN), which includes collapsing glomerulopathy, HIV-immune-complex kidney disease and thrombotic microangiopathy proved by kidney biopsy [7][8][9]. However, the wide use of ART over the past two decades has converted HIV infection to a chronic illness, with associated changes in the incidence, type and severity of kidney diseases in PLWH [10,11].
IKF is characterized by a low estimated glomerular filtration rate (eGFR) or elevated urinary albuminto-creatinine ratio (ACR) without treatment of renal replacement therapy [12,13]. In both research and clinical practice, eGFR is largely used as an alternative for GFR that is widely accepted as the best overall indicator of kidney function [14]. Serum creatinine concentration and/or serum cystatin C are routinely used to estimate the GFR for assessment of kidney function in PLWH [7]. Although cystatin C has been proposed to be a better indicator than creatinine for IKF, previous studies support the use of the creatinine using Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equation for routine clinical care among PLWH [15].
There is an increasing number of studies reporting the incidence of IKF measured by serum creatine-based eGFR (eGFRcreat) in PWLH across the world. However, the data have not been appropriately synthesized to provide an overview of IKF burden based on eGFRcreat in PLWH. Only one review summarized prevalence of chronic kidney disease (CKD) among PLWH [16]. To fill this gap, we synthesized available data to investigate the eGFRcreat-based incidence of IKF by different working definitions in PLWH.

Search strategy
In this study, a systematic literature search was performed in PubMed, Ovid Medline, EMBASE, and the Web of Science up to May 7th, 2021 to identify observational studies that reported the incidence of IKF in PLWH. A combination of the keywords and of HIV infection, IKF and incidence and their synonyms were used (Supplementary 1). We also performed a manual search for cited reference lists in the original articles and reviews.

Study selection
Two investigators independently reviewed the identified studies for further review and meta-analysis according to preset inclusion and exclusion criteria. Duplicate records were removed before title and abstract screening. The title and abstract were first screened to confirm whether an article was potentially relevant. Then, full-texts of the articles were assessed to determine if they met the eligibility criteria. Any disagreements were addressed by the third investigator discussing with these two reviewers.
Studies were included if they fulfilled all of the following criteria: 1) reporting a certain eGFR threshold and/or eGFR decline determined by serum creatinine; 2) with a prospective or retrospective cohort study design; 3) available data for calculation; 4) with a sample size of more than 200; 5) published in English. Studies using Cockcroft-Gault formula would also be excluded given that calculation would depend on a participant's body weight and BMI, which is different from CKD-EPI and Modification of Diet in Renal Disease (MDRD) equations.

Quality assessment and data extraction
Two investigators independently performed quality assessment and data extraction, with discrepancy resolved by a third reviewer. Study quality was evaluated using the Newcastle-Ottawa Scale (NOS), which included three sections that covered different methodological perspectives. Scores on the NOS ranged from zero to nine points, where the higher scores indicated the better quality (Table S2). A study with a score of ≥7 was considered have high quality, with a score of 5 to 6 was considered to have moderate quality.
The following data for each study were extracted from full-text articles: first author, publication year, country or region of study, study period, male proportion, age, race composition, ART status at baseline, definition of outcome, eGFR calculation equation, frequency of eGFR measurement, sample size, duration of follow-up and numbers of incident cases.

Statistical analysis
All definitions of IKF were summarized. The incidence of IKF was calculated by data extracted from included studies. The cumulative incidence was calculated for all studies per available incident cases and sample size. The incidence rate was only calculated for studies with available person-time. A random-effect model employing Hartung-Knapp method to adjust test statistics and confidence intervals (CIs) with Sidik-Jonkman estimator used for the between-study variance (HKSJ method) was performed to pool incidence rates, as it was proved that the HKSJ method performs better than Der Simonian-Laird (DL) method, especially when heterogeneity was present [17]. The overall incidence rates were only pooled for five most-studied definitions of IKF, including eGFR< 90 ml/min/1.73m 2 (n = 3 studies), eGFR< 60 ml/min/1.73m 2 or CKD (n = 28 studies), eGFR< 30 ml/min/1.73m 2 (n = 6 studies), > 25% decrease in eGFR (n = 4 studies) and combined eGFR (short for the combination of a confirmed eGFR < 60 mL/min/1.73m 2 with > 25% decrease in eGFR) (n = 4 studies), respectively. Forest plots were used to describe pooled incidence rates. We also pooled incidence rates of different stages of CKD by including all studies that defined IKF based on the cut-off values of eGFR< 60, < 50, < 45, < 30 and < 15 ml/min/1.73m 2 , respectively, in sensitive analysis.
χ2 -based Cochran's Q test and I 2 statistic were applied to evaluate the statistical heterogeneity across eligible studies. Cut-off values of I 2 statistic of 25, 50, and 75% were used to define low, moderate, and high heterogeneity, respectively. When substantial heterogeneity was detected, subgroup and univariate metaregression analysis were performed to examine sources of heterogeneity stratified by variables including WHO regions, income level, race composition, median or average age, male proportion, ART status, sample size and calculation equation. Bonferroni-adjusted P values were applied to univariate meta-regression for comparison of effect estimates between subgroup. Subgroup analysis, meta-regression analysis, summary of risk factors, assessment of publication bias and sensitive analysis were only conducted for pooling studies in which eGFR< 60 ml/min/1.73m 2 was used to define IKF, as we required a minimum of ten independent studies to justify the analysis. The nonparametric "trim-andfill" method was undertaken to assess publication bias. Publication bias was also evaluated by Begg's (nonparametric rank correlation test) and Egger's tests (regression-based test). P < 0.05 and asymmetric funnel plot indicated that there existed significant publication bias. Moreover, sensitivity analysis was performed to evaluate the influence of a single study on the overall pooled estimate by deleting one study at each step. P < 0.05 was considered as statistical significance. All the statistical analysis were performed using Stata 17.0 (Stata Corporation, College Station, TX, USA).

Study selection and quality assessment
A total of 3769 articles were identified through systematically searching four databases, and 28 were obtained from other sources. Of 3797 articles, 1272 articles were removed after screening for duplication. After removing 2157 unrelated articles, the 383 full-text articles were assessed for eligibility. Eventually, a total of 60 studies were processed for final review and meta-analysis. The flow chart of study selection was described in Fig. 1.
In this work, the NOS scores of most included studies ranged from six to nine points, suggesting that the quality of included studies was moderate to high (Table S2).

Characteristics of included studies
The main characteristics of included studies are summarized in Table 1. There were 60 studies with 358,221 participants. The 60 studies were conducted in World health organization (WHO) regions as the following: African region (n = 7), American region (n = 17), European region (n = 15), South-east Asia region (n = 2), Western pacific region (n = 14) and multi-region studies (n = 5).
Thirty-nine (65.0%) studies had a sample size > 1000. Fifty-eight articles reported male proportion, with a median of 81.3% (range: 4.6-76.5%). Fifty-eight (96.7%) articles reported the median or mean age, among which 23 studies had a median or mean age of ≥40 year-old. Regarding ART status at baseline, 44 articles had more than 80% of ART-experienced patients. Forty-seven articles reported the proportion of tenofovir (TDF) use, with a median of 61.0% (IQR: 48.3-100.0). Thirty-nine articles reported the proportion of protease inhibitors (PIs) use, with a median of 35.9% (IQR: 17.0-52.6).
Methods for estimates of GFR included the CKD-EPI equation (n = 30 studies) and the MDRD (n = 30 studies). More than half of included studies had two or more eGFRs to confirm the occurrence of the outcome. More information about these studies such as TDF use, PIs use at baseline or during the follow-up was presented in Table S1.

Different definitions and incidence of IKF
Of included studies, a total of 19 definitions of IKF were described and reported ( Table 1, Table S1 and Table S3), which were categorized into three types: a certain threshold of eGFR, an absolute decrease in eGFR or percent decline of eGFR and the combination of certain eGFR threshold with decrement in eGFR. The reported incidence rate of IKF differs widely by different definitions. The lowest incidence rate was observed in "eGFR< 30 ml/ min/1.73m 2 ", a 0.14 per 1000 person-years (PYs) reported by Suzuki S [62]. While the highest was observed in ">25% decrease in eGFR", with incidence rate of 190.1 per 1000 PYs [64,78]. Specifically, 16 studies reported two or more definitions of IKF at same time.
Four studies reported the result of combined eGFR with a study period of more than five years. The cumulative incidence of combined eGFR ranged from 2.4 to 15.3%. The overall incidence rate of it was 16.55 per 1000 PYs (95% CI: 2.99-91.57, I 2 = 99.79%, p < 0.001), which was higher than that of eGFR< 60 ml/min/1.73m 2 but without statistical significance (Fig. 3).

Decrease in eGFR > 25%
> 25% decrease in eGFR was the second most-studied IKF in included studies (n = 7). Seven studies reported a range of cumulative incidence from 3.1 to 34.0% (Table 1). All studies were conducted in countries in Western pacific region, four were in Japan. All studies had relatively small sample size of less than 1000 and short study period. Compared to all other definitions of pooled incidence rates, eGFR > 25% showed highest result, with 119.63 per 1000 PYs (95% CI: 83.23-171.94, I 2 = 90.67%, p < 0.001).

Other definitions of IKF
There were many other definitions of IKF in included studies, each of which contained less than three studies. However, different baseline eGFRs were applied even in studies using same definition of IKF. Threshold of eGFR included 15, 45, 50 and 70 ml/ min/1.73m 2 . More than half of these studies had sample size of more than 1000. The cumulative incidence was relatively low regardless of the study period, most of which were more than five years (Table 1). Less than half of studies reported the incidence rate but only cumulative incidence. Absolute value decreased in eGFR included 3, 10 and 20 ml/min/1.73m 2 , and most studies had sample size of less than 1000 and had high proportion of male. Two studies from Japan showed high cumulative incidence of decrease in eGFR > 10 ml/min/1.73m 2 (77.4 and 80.6%).
Other definitions of IKF included percent decrease in eGFR and threshold of eGFR combined with percent decrease or absolute decrease in eGFR (Table 1). Most studies were from high income countries. The highest incidence rate was observed in confirmed decrease in eGFR > 20%, yielding an incidence rate of 68.0 per 1000 PYs [75].

Assessment of publication bias
Funnel plots and result of "trim and fill" suggested no publication bias for reporting data on the incidence rate of eGFR< 60 ml/min/1.73m 2 (Fig. S1). Although the reporting data might underestimate the incidence rate compared with imputed result (12.50, 95%CI: 9.13-17.11 vs. 17.77, 95% CI: 12.68-24.91, per 1000 PYs), no significant difference was detected. The results were confirmed by the formal Egger test(p = 0.139) and Begg's tests (p = 0.195).

Sensitivity analysis
Sensitivity analysis for eGFR< 60 ml/min/1.73m 2 by omitted one study at a time showed the stability of effect estimates (Fig. S2). Additionally, we pooled the incidence rates of all stages of CKD, including all studies with outcome defined as eGFR< 15, < 30, < 45, < 50, or < 60 ml/min/1.73m 2 , yielding an overall incidence rate of 8.43 (95% CI: 5.68-12.51), per 1000 PYs. This result was lower than the pooled result of eGFR< 60 ml/ min/1.73m 2 but without any statistical significance.

Discussion
To our knowledge, this is the first attempt to summarize incidence estimates for IKF determined by eGFRcreat in PLWH across the world. Findings from this study revealed that various definitions of IKF based on eGFR were applied in PLWH. IKF included threshold of eGFR, absolute or percent decrease in eGFR, and certain eGFR threshold combined with decrement in eGFR. Among these outcomes, eGFR< 60 ml/min/1.73m 2 was the moststudied and most-represented, followed by decrease in eGFR > 25%. The overall incidence rate of eGFR< 60 ml/ min/1.73m 2 was 12.50 (95%CI: 9.00-17.36) per 1000 PYs. The highest pooled incidence rate was observed in the definition of decrease in eGFR > 25%, while the lowest was observed in the eGFR < 30 ml/min/1.73m 2 , despite with substantial heterogeneity.
The eGFR< 60 ml/min/1.73m 2 was treated as threshold of diagnosis criteria of CKD. Findings from our study suggested that most researchers still concerned about the accepted IKF---CKD. It has been suggested that the disease burden measured by eGFR< 60 ml/min/1.73m 2 could be underestimated by only using the eGFR in comparison with studies defining the outcome as eGFR < 60 ml/ min/1.73 m 2 combined with other indicators, such as proteinuria [79]. However, our results highlighted that PLWH borne the heavy burden of IKF even if using the eGFR < 60 ml/min/1.73 m 2 . And such burden was more considerable in the late-HAART era, suggested by Jespersen, N. A., et al. [80]. The cumulative incidence and pooled incidence rate of most-reported eGFR< 60 ml/ min/1.73m 2 were approximate to the corresponding Table 2 Subgroup analysis for incidence rate of eGFR < 60 mL/min/1.73m 2 among PLWH Note: IR, incidence rate; ART, antiretroviral therapy; MDRD, modification of diet in renal disease; CKD-EPI, Chronic kidney disease Epidemiology collaboration † the proportion of any race≥60% was recognized as predominant race in one study. The second most-studied definition of IKF was decrease in eGFR > 25%. The highest incidence rate was observed in decrease in eGFR > 25%, which suggested that the occurrence of decrease in eGFR > 25% was most frequent in PLWH. The decrease in eGFR > 25% is more an indicator of acute kidney injury (AKI) than that of CKD [81], which is usually applied to measure the TDF-associated IKF. AKI and CKD are usually inter-connected: AKI is attributable to the development and progression of CKD; CKD is known to predispose patients to AKI [82]. However, previous study found that a large proportion of PLWH used TDF developed AKI, but only a minority of TDF users progressed to CKD [83]. In this study, we did not distinguish the CKD from AKI deliberately. Our finding suggested that the application of different definition of IKF by eGFR could reflect the predisposition of IKF in some circumstances.

Variables
The eGFR < 30 ml/min/1.73m 2 was had lowest pooled incidence rate in this study, similar to ESRD directly diagnosed by medical record, with an incidence rate of 2.56 (95% CI, 2.33-2.81) per 1000 PYs in PLWH [55]. Previous study showed that PLWH with eGFR< 60 mL/min/1.73 m 2 had higher risk of eGFR < 30 mL/min/1.73 m 2 in comparison with those who with higher baseline eGFR [13]. These results stressed the importance of IKF defined by eGFR < 30 ml/min/1.73m 2 . For the eGFR< 90 ml/min/1.73m 2 , it met the criteria of mildly decreased kidney function according to the Kidney disease: Improving global outcomes (KDIGO) guidelines [14]. Mildly decreased kidney function is common in ART-experienced PLWH. However, PLWH could have improved kidney function if they switch nephrotoxic ART regimen to other regimens [84]. Moreover, they also had the possibility to transit from a mild renal impairment to a normal eGFR even they were receiving TDF [85]. However, since mild impairment can also progress to advanced stages, early detection and management of mildly IKF is essential to address the rising epidemic of kidney disease.
Of note, many other definitions of IKF were employed in studies in this review. Limited by the quantity of study, we were unable to pool the incidence rate of them, but just make a summary. Some definitions of IKF were significant indicators for certain clinical kidney disease, such as decline in eGFR ≥30%. It has been reported that the decline in eGFR≥30% was strongly associated with CKD progression to end-stage renal disease (ESRD) [86]. Therefore, it was worth paying attention to absolute or percent decrement in eGFR. However, the quantity of such studies aiming to certain definitions of IKF was small, and many of them contained only one study. More studies that investigated the present definitions of IKF were needed to confirmed their clinical significance.
Notably, in this study, there was a total of 32 selected studies from the American and European region, 14 studies from Western Pacific region, and only seven studies from Africa region where the burden of HIV infection is heavier than other regions [1]. More attention should be paid to the need for more information and validated measures of IKF in countries with heavy burden of HIV infection, especially in the context of non-communicable diseases.
The major strength of our study was the comprehensive inclusion of studies for the incidence of IKF. We searched four databases to find as many articles as possible that met our included criteria. Second, to make sure the homogeneity of study population as much as possible, we only included longitudinal studies. Any study with all participants being of coinfection or comorbidity was excluded. However, we did not limit this review to any time period (pre-highly active antiretroviral therapy (HAART) era vs. post-HAART era). Third, the quality of most of selected studies was rated as high, so our metaanalysis provided good evidence for the incidence of eGFR-based IKF. Forth, we performed the random-effect meta-analysis and the sensitivity analysis, the results of which revealed the stability of pooled incidence of IKF in PLWH.
There are limitations worth noting. First, we observed significant heterogeneity of our pooled estimates, which may be inherent to the heterogeneity between studies in terms of sample size, inclusion criteria, data collection methods, race composition and measurement of eGFR and could not be completely explained by investigated study-level characteristics. Second, as some of included studies did not provide follow-up duration, incidence rate was calculated only for those with person time. Third, limited by the number of included studies, we did not pool the incidence rate for all definitions of IKF, but just made a summary of basic characteristics and cumulative incidences. Fourth, the creatinine-based eGFR could be affected by muscle mass and diet, which might result in an overestimation of eGFR.

Conclusion
This study revealed the substantial definitions of IKF applied in PLWH, among which eGFR< 60 ml/ min/1.73m 2 was most-studied and most-represented, followed by decrease in eGFR > 25%. These definitions would promote our understanding of IKF. Our findings also highlight the need for a concerted action to provide more evidence on certain kind of kidney impairment and a more uniform diagnostic criterion that this condition entails. Our review also points out the critical need of data from every part of the world, especially from these areas with heavy burden of HIV infection, that would help to further characterize the magnitude of CKD burden in PLWH.