Two authors (F.F., and V.D.) independently reviewed English-language citations from the National Library of Medicine's Medline database from 1989 through June 30, 2018. Data on HCV status were not available before 1989, when the first assay for HCV was manufactured. We conducted our search by four Medline databases engines (Embase, Grateful Med, Ovid, and PubMed). Our Medline search was limited to human studies. We applied the following algorithm in medical subject heading and in free text words: (“HEPATITIS C” or”HCV” or “HEPATITIS C VIRUS INFECTION” or “HCV INFECTION” or”ANTIBODY POSITIVE SEROLOGIC STATUS” or”ANTI-HCV”) and (“CHRONIC KIDNEY DISEASE” or “CKD” or “END-STAGE RENAL DISEASE” or “ESRD” or “GLOMERULONEPHRITIS” or”GN” or”LOW GLOMERULAR FILTRATION RATE” or”low GFR” or”KIDNEY DISEASE” or”KIDNEY EVENTS”) and (“INTERFERON” or”IFN” or”pegylated-INTERFERON” or”peg-IFN” or “RIBAVIRIN” or”RBV”) and (”DIRECT-ACTING ANTIVIRAL AGENTS” or “DAAs”) and (“SUSTAINED VIROLOGICAL RESPONSE” or”SUSTAINED VIRAL RESPONSE” or”SVR” or “CURE”) and (“HAZARD RATIO” or “HR” or “INTERNATIONAL RATE RATIO” or “IRR” or “ODDS RATIO” or”OR”). An additional search was performed with electronic searches of the Cochrane Library; manual searches of selected specialty journals were done to identify all pertinent literature. Reference lists from qualitative topic reviews and published clinical studies were also searched. It was previously demonstrated that a Medline search alone might not be sensitive enough10. Data on study design, study period, patient characteristics, HCV prevalence, antiviral therapy towards HCV, and kidney disease outcomes were abstracted. Authors of selected papers were contacted to obtain missing data and only data from individuals with known HCV status were included in the meta-analysis. Consensus was achieved for all data. Studies were compared to eliminate duplicate reports for the same patients, which included contact with investigators when necessary. Eligibility and exclusion criteria were pre-specified. Our search was limited to human studies that were published in the English literature.

Studies were included if they met the following inclusion criteria: (1) they presented original data from cohort and longitudinal studies; (2) the outcome of interest was clearly defined as the risk of occurrence (or deterioration) of kidney disease after antiviral therapy among HCV-positive patients; (3) they provided adjusted risk estimates and their confidence intervals. We considered both case-control studies and cohort studies as eligible for inclusion in the analysis. If data on the same population were duplicated in more than one study, the most recent study was included in the analysis.

Studies were excluded if they reported inadequate data on the rate of kidney disease in the HCV- infected population after antiviral therapy towards HCV (e.g., incomplete information on HCV status or antiviral therapy or renal outcomes). Unpublished studies, studies that were only published in abstract form or as interim reports were excluded; letters and review articles were not considered for this systematic review.

The quality of the 15 studies was appraised using a scale adapted from the ‘Newcastle/Ottawa Scale (NOS)’11. The Newcastle-Ottawa scale is a scoring system that assesses every aspect of an observational epidemiologic study from a methodological point of view. When a study included relevant information that could be associated with the NOS, one point was added. Seven items in cross-sectional studies and eight items in cohort and case-control studies that could be related to the NOS were identified. Therefore, cross-sectional studies assigned 8-10, 6-7, 4-5, or 0-3 points (stars) were evaluated as very good, good, satisfactory or unsatisfactory studies, respectively. Similarly, cohort/case-control studies with 7-9, 5-6, 4 and 0-3 points (stars) were identified as very good, good, satisfactory or unsatisfactory, respectively. We carried out subgroup analyses based on those studies provided with very good quality. Data extraction and quality scoring were performed independently by two reviewers (F.F. and V.D.) and the results were merged by consensus. The complete protocol for quality scoring is available on-line (see Supplemental File 2).

The primary end point was to provide adjusted estimates of the risk (and 95% CIs) of kidney disease among HCV-infected population after antiviral therapy for HCV. Multivariate analysis was made to estimate the independent effect of anti-HCV therapy on the frequency of kidney disease after adjustment for potential confounders (covariates) (e.g., age, gender, race/ethnicity, diabetes mellitus, and others). Cox proportional hazard regression analysis was carried out to assess the effect of antiviral therapy for HCV per se on the risk of kidney disease after adjustment for differential follow-up time and distribution of potential confounders.

We weighted the study-specific log hazard ratios by the inverse of their variance to obtain a pooled effect estimate and its 95% confidence intervals. For each study, we used the estimate of the effect measure that was adjusted for the largest number of confounders. We present both fixed-effects and random-effects pooled estimates but use and report the latter when heterogeneity was present. We used the random-effects approach, as described by DerSimonian and Laird12, Cochrane Q–test was used for quantifying the heterogeneity13. The I2 statistic, which is the percentage of total variation across studies due to heterogeneity rather than chance, was also calculated14. The null hypothesis of this test is the absence of heterogeneity. We explored the origin of heterogeneity by stratified analysis and meta-regression technique. Subgroup or stratified analysis and meta-regression were pre-specified. Stratified analysis was made by restricting the analysis to subgroups of studies defined by study characteristics (such as country of origin, response to antiviral therapy, and others). Meta-regression was performed in order to look at simultaneously the effects of multiple (continuous and categorical) covariates on the outcome of interest. We performed random-effects meta-regression using the method of moments or maximum likelihood approaches where appropriate. Publication bias was assessed by the Egger test for funnel-plot asymmetry. All analyses were done with the statistical package Comprehensive Meta-Analysis (CMA), version 2.0 (Biostat Inc., USA, 2005). The 5% significance level was adopted for α risk. Every estimate was given with its 95% Confidence Intervals.

As shown in Figure 1, we retrieved 4,533 articles and 197 full-text papers were assessed for eligibility. The list of the 197 full-text papers is reported in the Supplemental File 3. Fifteen studies met our inclusion criteria; these were published in 13 papers [Figure 1] and carried out in 3 continents (n=356,825 patients)15–27. Thus, some studies reported data on more than one kidney disease outcome, but each cohort was represented once in any meta-analysis. There was a 100% concordance between reviewers with respect to final inclusion and exclusion of studies reviewed based on the predefined inclusion and exclusion criteria.

Figure 1.

Flow diagram of study selection.

(0,2MB).

We included studies where the diagnosis of HCV infection was done by testing for anti-HCV antibody in serum15,16,18,21,22,26,27. Information on HCV serological status was collected at the time of enrolment. Surveys based on the diagnosis of HCV infection by administrative data (ICD-9-CM codes) were also considered17,19,20,23–25.

Some authors defined the presence of kidney disease by ICD-9-CM codes17,19,20,23–25; others adopted assessment of eGFR15,16,18,21,22,26,27.

Many studies adopted the Cox proportional hazard model to examine the independent association of antiviral therapy with outcomes; one study used the multivariable Poisson regression model26. Conventional regression analyses were conducted in two papers16,18. Some authors have performed the propensity score methods to obtain better control for confounding19,24,25,27.

Tables 1–4 report some salient demographic and clinical characteristics of subjects enrolled in the included studies. The mean age of patient cohorts ranged from 39.8 to 58 years. The gender distribution ranged from 49.6% to 97.1% male. Five studies were from Asia; six and four from U.S. and Europe, respectively. The average follow-up ranged between 1.8 to 8.2 years. Table 5 gave information on study design and anti-viral therapy.

As listed in Table 6, there was large heterogeneity in the definition of the outcome. The majority of studies addressed the incidence of kidney disease after antiviral therapy among patients with HCV infection15–17,19–27. One report evaluated the impact of antiviral therapy on the progression of kidney disease among HCV-infected individuals18.

The definition of kidney disease adopted by the authors was chronic kidney disease (n=7), end-stage renal disease (n=3), glomerular disease (n=3), and kidney events (n=2) (Table 6). Chronic kidney disease and end-stage renal disease were assessed according to the Kidney Disease Outcomes Quality Initiative/Kidney Disease: Improving Global Outcomes classification.

Some studies (n=7) expressed their results according to the viral response after anti-viral therapy [Table 7]. Pooling studies based on viral responses (n=7; 34,763 unique patients) demonstrated that the risk of kidney disease was greater in patients who did not obtain than those who obtained sustained viral response after anti-HCV treatment; the summary estimate for adjusted risk was 2.5 (95% CI, 1.41; 4.41) (P=0.0016) and between-studies heterogeneity was found (P value by Q test=0.004). According to the Egger's regression intercept, no publication bias occurred (P=0.58). The quality scores ranged between 4 and 8 stars.

Eight studies (n=333,312 patients) compared patient cohorts who received vs. those who did not receive antiviral treatment for HCV. The summary estimate for adjusted risk of lower incidence of kidney disease across the eight studies was 0.39 (95% CI, 0.25; 0.612) (P=0.0001) [Table 8]. Tests for homogeneity of the adjusted risk across the eight studies gave a Q value of 52,608 (P=0.0001), I2=86.6; the homogeneity assumption was rejected. The funnel plot concerning the publication bias is reported in Figure 2. The Egger test showed publication bias (P=0.015). The quality scores ranged between 5 and 8 stars.

Figure 2.

Funnel plot of precision by Log Ratio (cohorts who received antiviral therapy vs. those who did not; n=8 studies, n=333,312 unique patients).

(0,08MB).

Stratified analyses are shown in Tables 7–8, no substantial difference in pooled adjusted effect estimates across designs was found even if the homogeneity assumption was rejected in some subsets.

Aggregation of studies comparing treated vs. untreated cohorts (n=7; 321,522 unique patients) reported that the incidence of kidney disease was lower in patients who received antiviral therapy compared to those who had not received it, the summary estimate for adjusted HR of reduced incidence of kidney disease was 0.44 (95% CI, 0.25; 0.63) (P=0.0001) (Figure 3). Between-studies heterogeneity was noted (P value by Q test P=0.0001). According to the Egger's regression intercept, there was publication bias (P=0.037). Figure 4 shows the adjusted risk of lower incidence of CKD among HCV-infected patients receiving IFN-based therapy (n=6 reports).

Figure 3.

Estimated HR and 95% confidence intervals for each study (cohorts who received antiviral therapy vs. those who did not; n=7 studies, n=321,522 unique patients)

The vertical line represents the pooled aHR of risk of kidney disease (random-effects model) after comparison.

Adjusted HR 0.44 (95% CI, 0.25; 0.63) (P=0.0001). Heterogeneity statistics: Q-value (by Chi-squared test)=47.5, df=6, P=0.0001; I2=87.4%.

(0,13MB).

Figure 4.

Adjusted risk of lower incidence of CKD after the end of IFN-based antiviral therapy among HCV-infected patients (n=6 studies) (Odd man out).

(0,05MB).

We made meta-regression analysis in the subset of studies (n=8, n=333,312 unique patients) comparing patient cohorts who received antiviral therapy for HCV vs. those who did not [Table 2]. As listed in Table 9, meta-regression demonstrated a significant impact of frequency of cirrhosis (P=0.02), HBV infection (P=0.0001), arterial hypertension (P=0.002), DM (P=0.051), and gender (male) (P=0.05) on the outcome of interest (adjusted risk of reduced incidence of renal disease among HCV-positive patients after antiviral therapy ended). Other covariates included in the model (age, duration of follow-up, size, reference year) had no effect on the attitude of antiviral therapy in reducing the incidence of kidney disease among HCV-infected patients.

Figure 5 reports that the adjusted HR of lower incidence of kidney disease after antiviral therapy among HCV-infected patients (when adjusted for the other covariates) becomes more important when the frequency of diabetes increases.

Figure 5.

Meta-regression: Regression line (and 95% CI) of Log hazard ratio on Diabetes.

(0,09MB).