|Home | About | Journals | Submit | Contact Us | Français|
To examine long-term effects of antiretroviral therapy (ART) on kidney function, we evaluated the incidence and risk factors for chronic kidney disease (CKD) among ART-naive, HIV-infected adults and compared changes in estimated glomerular filtration rates (eGFR) before and after starting ART.
Multicenter observational cohort study of patients with at least one serum creatinine measurement before and after initiating ART. Cox proportional hazard models, and marginal structure models examined CKD risk factors; mixed-effects linear models examined eGFR slopes.
Three thousand, three hundred and twenty-nine patients met entry criteria, contributing 10 099 person-years of observation on ART. ART was associated with a significantly slower rate of eGFR decline (from −2.18 to −1.37 ml/min per 1.73 m2 per year; P = 0.02). The incidence of CKD defined by eGFR thresholds of 60, 45 and 30 ml/min per 1.73 m2 was 10.5, 3.4 and 1.6 per 1000 person-years, respectively. In adjusted analyses black race, hepatitis C coinfection, lower time-varying CD4 cell count and higher time-varying viral load on ART were associated with higher CKD risk, and the magnitude of these risks increased with more severe CKD. Tenofovir and a ritonavir-boosted protease inhibitor (rPI) was also associated with higher CKD risk [hazard odds ratio for an eGFR threshold <60 ml/min per 1.73 m2: 3.35 (95% confidence interval (CI) = 1.40–8.02)], which developed in 5.7% of patients after 4 years of exposure to this regimen-type.
ART was associated with reduced CKD risk in association with CD4 cell restoration and plasma viral load suppression, despite an increased CKD risk that was associated with initial regimens that included tenofovir and rPI.
Treatment with potent combination antiretroviral therapy (ART) is associated with lower incidence of HIV-associated nephropathy (HIVAN)  and improved kidney function [2–5], but information about the long-term benefits of ARTon kidney function is lacking. HIV-infected individuals have a greater risk of chronic kidney disease (CKD) than the general age-matched population and kidney function may continue to decline despite ART [6,7]. As in the general population, HIV-infected patients with CKD have higher risk of cardiovascular disease and all-cause mortality [8,9].
Although ART may improve kidney function, nephrotoxicity has been observed with some drugs. Many studies have identified nephrotoxicity in association with tenofovir [10–18]. Nephrotoxicity also was associated with atazanavir [17,19], whereas some but not all studies have described enhanced nephrotoxicity when tenofovir was combined with atazanavir or a ritonavir-boosted protease inhibitor (rPI), suggesting additive toxicity or interactions between drugs [20–24]. However, most previous studies examined the effects of individual drugs rather than drug regimens on kidney function.
To better understand the long-term benefits and risks of ARTon kidney function among the broader spectrum of HIV-1 infected patients in routine care, we sought to determine the incidence of moderate and severe CKD in association with demographic factors, comorbid conditions, CD4 cell count, and plasma HIV-1 viral load during treatment, and we compared changes in estimated glomerular function (eGFR) before and after ART-initiation in a large multicenter cohort of ART-naive patients from the Center for AIDS Research Network of Integrated Clinical Systems (CNICS) Cohort . To further explore nephrotoxicity associated with tenofovir, and to specifically examine whether a rPI may enhance this toxicity, we also compared CKD risk by initial ART regimens that were categorized into four mutually exclusive regimen-types according to whether they included a rPI or a nonnucleoside reverse transcriptase (NNRTI), with or without tenofovir.
CNICS comprises a large, diverse population of HIV-1 infected patients receiving care at eight clinical sites that are affiliated with the Centers for AIDS Research, an interdisciplinary collaboration of basic and clinical investigators distributed across the United States . Comprehensive clinical data captured in point-of-care electronic health records, including medications, laboratory results, and diagnoses undergo rigorous quality assessment and standardization, and are integrated into a single data repository. Patients without prior ARTof any kind, with at least one serum creatinine before and after initiating ART between April 1996 and July 2009 were included. ART was defined as three or more drugs, including a nonnucleoside reverse transcriptase inhibitor or a rPI. To enhance the generalizability of our findings with current ART treatment guidelines, patients were excluded whose initial regimen included indinavir, two or more protease inhibitors (not including ritonavir), a protease inhibitor and a NNRTI, or an unboosted protease inhibitor. Because of small numbers of initial regimens that included raltegravir and maravaroc these regimens were also excluded. Initial regimens were categorized into four mutually exclusive regimen-types as tenofovir and a rPI; tenofovir and a NNRTI; a rPI without tenofovir; and a NNRTI without tenofovir, to which the other three regimens were compared.
Laboratory measurements were collected at the discretion of the clinician during routine care according to guidelines for HIV-disease management . We estimated eGFR by the four variable Modification of Diet and Renal Disease (MDRD) equation. In sensitivity analyses, eGFR was also measured using the CKD-Epidemiology equation . In accordance with recently proposed revisions to the classification of CKD, moderate CKD (stage 3) was subdivided by eGFR thresholds of less than 60, and less than 45 ml/min per 1.73 m2; severe CKD (stage 4) was defined by a threshold of less than 30 ml/min per 1.73 m2 [28,29]. Qualifying values included those measurements that fell below each threshold and persisted for at least 3 months, during which time the median value of all eGFR measurements also had to remain below these thresholds. Cases of moderate CKD also included those with more advanced stages. CNICS is approved by the Institutional Review Boards at each CNICS site.
We used standard Cox proportional hazards models to examine risk factors for CKD at each eGFR threshold. Patients were included who had an eGFR more than 60 ml/min per 1.73 m2 at the last pre-ART measurement, and did not have a prior diagnosis of kidney disease, determined by their physician. We examined the following baseline patient factors before ART initiation: AIDS-defining illness (ADI), pharmacologically treated hypertension and diabetes, chronic hepatitis B (hepatitis B surface antigen positive) and C (detectable hepatitis C RNA), nadir CD4 cell count, latest pre-ART HIV-1 viral load; and also examined time-varying CD4 cell count and viral load after ART initiation using the last value-carried-forward method for missing time-varying measurements. Time at risk began upon ART initiation and continued until the onset of CKD (determined by the first qualifying eGFR value) and patients were censored upon death, lost-to-follow-up, or at the conclusion of the study, whichever occurred first. Each risk factor was examined in a separate model adjusted for age, sex, race, last pre-ART eGFR and initial ART regimen-type. All patients were included in these analyses whether or not they remained on their initial ART regimen.
We used marginal structural pooled logistic regression models to examine associations between initial ART regimen exposure and CKD risk by censoring patients upon any change in ART, in addition to censoring for the reasons specified above, and adjusting for potential bias introduced by this informative censoring. Because initial ART regimens were not assigned at random, we also controlled for potential confounding by indication in these models using inverse probability treatment weights for differences between initial ARTregimen-types among the following baseline covariates: age, sex, race (black versus not black), prior (ADI), pharmacologically treated hypertension and diabetes, chronic hepatitis B and C infections, nadir CD4 cell count and the last pre-ART eGFR and viral load, using a random forest approach . Inverse probability weights for regimen change were estimated within uniform 12-week intervals to adjust for possible informative censoring using all of the above baseline variables and the initial ART regimen-type, and the following time-varying covariates after ART initiation: CD4 cell count and viral load [31,32]. Finally, we explored interactions between all baseline factors and the initial ART regimen-type and we used a cumulative incidence analysis to estimate the proportion of patients with CKD according to the initial ART regimen-type, in which death was treated as a competing risk .
Linear eGFR slopes before and after ART initiation were estimated by mixed-effects linear spline models that included all available serum creatinine measurements before and after ART initiation, with a knot at the time of ART initiation. These models were adjusted for initial ART regimen-type, chronic hepatitis B and C infections, pharmacologically treated hypertension and diabetes, time-varying CD4 cell count and viral load, and incorporated all eGFR measurements, whether or not patients remained on their original ART. They used random intercepts and slopes, and assumed an unstructured covariance matrix. Statistical significance was defined as P < 0.05 in all analyses, which used R (version 2.10.1), or SAS (version 9.0; SAS Institute Inc., Cary, North Carolina, USA).
Three thousand, three hundred and twenty-nine patients met study entry criteria and initiated 15 distinct ART regimens (categorized into four ART regimen-types as above) between April 1996 and July 2009 (Table 1) The median follow-up duration was 4.8 years [231 weeks; interquartile range (IQR) 121–364 weeks] including a median of 23 weeks (IQR 6–115 weeks) before, and 143 weeks after ART-initiation (IQR 61–255 weeks). Patients had a median of three (IQR 1–8) and 13 (IQR 6–25) creatinine measurements before and after initiating ART, respectively, with a cumulative follow-up duration of 10 990 person-years on ART. The majority of patients (2176; 64%) remained on their original ART regimen throughout follow-up, including 83 and 73% of patients, who received tenofovir and an NNRTI or a rPI, respectively, and 46 and 38% of patients, who received a NNRTI or a rPI, respectively, without tenofovir (P < 0.001 for comparisons between tenofovir plus a NNRTI versus the other three regimen types). Among patients, who had a change in their initial ART, this occurred after a median of 79 weeks (IQR 33–166 weeks). One hundred and sixty-seven patients died during follow-up a median of 122 weeks (IQR 53–212 weeks) after starting ART. Additional characteristics of the study cohort are summarized in Table 1.
New onset stage 3 or greater CKD, defined by an eGFR less than 60 ml/min per 1.73 m2, developed in 106 patients a median of 45 weeks after ART initiation (IQR 11–137 weeks), with an incidence of 10.1 cases per 1000 person-years [95% confidence interval (CI) = 8.3–12.3]. Factors associated with significantly higher risk for CKD defined by this eGFR threshold included: black race, hepatitis C coinfection, pharmacologically treated hypertension, a prior ADI, lower CD4 nadir and lower time-varying CD4 cell count, higher last pre-ART viral load and higher time-varying viral load on ART in adjusted models (Table 2). The initial ART regimen-type was not significantly associated with increased CKD risk in these models.
Seventy-two patients developed stage 3 or greater CKD while on their original ART regimen. In marginal structural models controlling for informative censoring upon regimen change and confounding by indication, patients who received tenofovir and a rPI had a significantly higher risk of CKD than those treated with a NNRTI without tenofovir [hazard odds ratio (HOR) 3.35 (95% CI, 1.40–8.02), P = 0.006] (Table 3). Tenofovir and a NNRTI was not significantly associated with increased risk of CKD in these models [HOR 1.59 (95% CI, 43–5.87), P = 0.48]. Tenofovir and a rPI was associated with a similar trend towards higher CKD risk when eGFR was estimated by the CKD-Epidemiology equation [HOR: 2.44 (95% CI, 0.99–6.01), P = 0.052]. The cumulative incidence of CKD in patients, who received tenofovir and a rPI was 5.7% after 4 years.
Although patients were censored upon any change of initial ART in this analysis, similar associations with initial ART regimen-types were evident when we included cases that began within 6 months after changing ART [HOR for rPI and tenofovir 2.75 (95% CI, 1.20–6.28)]. Among rPI regimens, the risk of CKD with atazanavir was similar to that associated with protease inhibitors not including atazanavir [HOR: 2.23 (95% CI, 0.82, 6.06), P = 0.12 and 5.40 |(95% CI, 1.94–15.05), P = 0.001 with atazanavir–ritonavir, and rPI not including atazanavir, respectively]. Among baseline factors, only race modified the association between CKD and initial ART [HOR for CKD in patients, who received tenofovir and a rPI versus a NNRTI without tenofovir: 9.12 (95% CI, 2.06–40.32), and 1.08 (95% CI, 0.04–29.77) among nonblack, and black patients, respectively, P-value for interaction = 0.02].
Thirty-four patients developed stage 3 or greater CKD, defined by an eGFR threshold of 45 ml/min per 1.73 m2 [incidence 3.3 cases per 1000 person-years (95% CI, 2.3–4.6)]. Black race, hepatitis C virus (HCV) coinfection, pharmacologically treated hypertension, a prior ADI, lower nadir and time-varying CD4 cell counts, and higher time-varying viral load remained significantly associated with this event, and the magnitude of risk for stage 3 or greater CKD was higher than when defined by a 60 ml/min per 1.73 m2 threshold (Table 2). Initial ART regimens were not associated with increased risk of CKD defined by this threshold among all patients (Table 2), among the 20 patients who remained on their original regimen at the onset of this event (Table 3), or when including cases whose initial onset was within 6 months after changing ART (data not shown). However, 90% of stage 3 CKD events were lost in the marginal structural logistic regression analysis when the CKD threshold was changed from 60 to 45 ml/min per 1.73 m2 among patients whose initial ART regimen contained tenofovir and a rPI, compared with only 50% of events lost among patients whose initial ART regimen contained tenofovir and a NNRTI, limiting the ability to estimate this effect.
Sixteen patients developed stage 4 or greater CKD after ART initiation [incidence 1.6 cases per 1000 person-years (95% CI = 0.8–2.4 per 1000 person-years)]. Consistent with the previous analyses, the risk of stage 4 CKD remained significantly higher among blacks, in patients with HCV coinfection, with lower time-varying CD4 cell count and higher time-varying viral load, and again, the magnitude of risk increased compared to CKD defined by higher eGFR thresholds (Table 2).
Among all patients, whether or not they remained on their regimen at the onset of the event, the risk of stage 4 or greater CKD was not significantly associated with either of the two types of tenofovir-containing regimens in the Cox proportional hazards model (Table 2). There were too few patients (9) who developed stage 4 or greater CKD while on their original regimen to reliably estimate the risk of CKD by initial ART regimen-type.
One hundred and two patients began ART with preexisting kidney disease (defined by the last pre-ART eGFR <60, but ≥30 ml/min per 1.73 m2), of whom 18 developed stage 4 CKD [eGFR <30 ml/min per 1.73 m2; incidence 51.1 per 1000 person-years (95% CI = 30.2–80.8 per 1000 person-years)]. Initial ART regimen-types were not significantly associated with an increased risk of stage 4 or greater CKD, both among the 15 patients, who remained on their original regimen at the onset of this event or among all patients (data not shown).
eGFR slopes before and after ART initiation were approximately linear in models that used all creatinine measurements, whether or not patients remained on their original ART regimen. ART was associated with a significantly slower rate of eGFR decline [by 0.81 (95% CI, 0.03–1.59) ml/min per 1.73 m2 per year, P = 0.02] after adjusting for ART regimen-type, hepatitis B and C, pharmacologically treated hypertension and diabetes, and time-varying CD4 cell count and viral load (Table 4). When analyzed according to the initial ART regimen-type, eGFR improved compared with the corresponding pre-ART slope in association with each regimen-type, but this improvement was significant only for patients, who received tenofovir and a rPI (Table 4).
In this large multicenter cohort of HIV-infected patients in routine care with a median follow-up duration of 4.8 years on ART, we observed a significantly slower rate of eGFR decline associated with antiretroviral treatment. We also observed a significantly higher risk of developing CKD in association with lower time-varying CD4 cell count and higher plasma viral load. This suggests that effective ART lowered the rate of eGFR decline and lowered the risk of CKD, and the magnitude of this protective effect increased in a dose-dependent manner with increasing severity of kidney disease. We also observed a significantly higher risk of moderate CKD associated with tenofovir and a rPI (HOR 3.35 for eGFR less than 60 ml/min per 1.73 m2, compared with a regimen that included a NNRTI without tenofovir). Despite this high risk, however, moderate CKD was uncommon in association with this regimen-type, developing in 5.7% of patients after 4 years of exposure. Furthermore, regimens containing tenofovir and a rPI were not associated with a significantly increased risk of more severe kidney disease in the context of usual care that included regular toxicity monitoring, although the small number of cases may have limited our ability to detect such an association. Of note, we did not observe a significantly increased risk of CKD at any eGFR threshold in patients, who initiated tenofovir plus a NNRTI.
Previous studies have demonstrated kidney function improvements in patients with HIVAN [1,34–36], and in observational cohorts of HIV-infected with renal insufficiency [2,4–6,37]. In some studies, this improvement was associated with viral load suppression on ART [2,3]. Consistent with two large multicenter cohort studies showing higher risk of ESRD associated with lower last CD4 cell count on ART [38,39], we extend these findings by demonstrating a significantly higher risk of CKD with lower time-varying CD4 cell count and higher time-varying plasma viral load on ART. These results provide additional evidence for the benefits of ART on kidney function in HIV-infected patients.
Many studies have found modest nephrotoxicity associated with tenofovir [10–17,21]. In a meta-analysis of 17 studies, the magnitude of this effect was −3.9 ml/min over a median of 48 weeks . Atazanavir was also independently associated with nephrotoxicity in one large study, and with enhanced nephrotoxicity when combined with tenofovir. Ritonavir may enhance tenofovir-associated nephrotoxicity by increasing tenofovir plasma levels, possibly via enhanced intestinal absorption [40,41]. In two clinical trials and two observational cohorts, greater eGFR declines were observed over 48 weeks to 24 months with tenofovir and a rPI, versus a NNRTI [20–23]. In contrast to our study, however, six observational studies did not find evidence of increased nephrotoxicity with tenofovir when it was combined with a rPI, including two that examined CKD risk and four that examined eGFR changes [10,14,17,18,24,42]. In the present study, the CKD risk associated with atazanivir–ritonavir did not appear to be different from that of other rPIs, mainly lopinavir–ritonavir. Although the reasons for discrepancies between the present study and several others are not apparent, it is possible that by examining outcomes associated with initial ART regimens only, we avoided obscuring associations with the drug exposures of interest with residual nephrotoxicity due to previous ART exposure . Furthermore, we may have been better able to detect additive or synergistic toxicities by examining ART regimens within a small number of mutually exclusive categories rather than evaluating drug combinations or interactions between numerous individual drugs. Finally by excluding indinavir, the ART-associated nephrotoxicity that we observed may be more representative of current ART regimens.
The strong association between black race and severe CKD highlights the disproportionate and substantial burden of kidney disease among African Americans, as has previously been documented, particularly in association with advanced kidney disease [37,44]. The present study also identifies an important contribution by hepatitis C to the risk of severe CKD in persons with HIV disease, extending recent associations found with modest eGFR declines and heavy proteinuria among HCV coinfected patients [45–47]. It is possible that other factors associated with past or current injection drug use may confound associations between hepatitis C and CKD.
Consistent with a previous study, we observed a significantly slower rate of eGFR decline on ART . Although eGFR improved with each ARTregimen-type, such improvement was significant only in association with tenofovir and a rPI, a regimen-type that was also associated with significantly higher risk of CKD. A possible explanation for these findings may be that tenofovir and a rPI does not adversely affect kidney function in most patients, but may only affect a subset of patients. In support of this, despite the high relative risk of CKD associated with this regimen, CKD developed in less than 6% of patients after 4 years of exposure. Risk factors for tenofovir-associated renal tubular toxicity in previous studies include older age, lower body mass index, and lower eGFR at the start of therapy [17,48,49]. Among all baseline factors, we detected an interaction between ART regimen and race, which was of uncertain significance given the small sample size that warrants further study.
This study was conducted among patients in routine care whose initial ART regimen was not randomly assigned. Thus, differences in baseline characteristics among patients receiving different types of initial ART could contribute to confounding by indication, which has the potential to minimize apparent toxicities associated with tenofovir for example, if patients, who were at greater risk for kidney disease were also less likely to receive this drug. In addition, changes in ART regimens in routine care limit the ability to directly examine associations between drug-exposure and CKD risk. Therefore, we examined CKD risk associated with initial ART regimen using marginal structural logistic regression models to control for confounding by indication and potential bias due to informative censoring that may occur if ART changes were made before eGFR declines fell below event-defining thresholds, resulting in missed CKD events. Thus, this analysis provides more reliable estimates of CKD risk with initial ART than the standard Cox proportional hazards analysis. These models could only control for measured variables that were included in the analysis, however, and residual confounding from unmeasured factors may remain.
It is notable that we observed only 16 cases of severe incident CKD in this cohort with over 10 000 person-years of follow-up on ART. Although this small number of events limited our ability to detect associations with ART, it argues against a general risk of severe nephrotoxicity by any ART regimen-type in the context of usual care with regular toxicity monitoring, given the large sample size and the long follow-up duration of this cohort. Consistent with a recent large study from the Veteran’s Administration, we found that tenofovir-containing regimens were not associated with increased risk of more severe CKD defined by eGFR thresholds below 45 or 30 ml/min per 1.73 m2 .
In conclusion, we observed a significantly slower rate of eGFR decline in association with treatment with ART, and lower risk of CKD with higher time-varying CD4 cell count and lower time-varying plasma viral load on ART. Despite these benefits, initial ART regimens that included tenofovir and a rPI were associated with significantly higher risk of CKD compared with a regimen that included a NNRTI without tenofovir, but this developed in less than 6% of patients after 4 years of exposure to such a regimen. Many factors contribute to kidney disease in persons living with HIV infection. These findings suggest durable benefits of ART-associated immunologic and virologic improvements in reducing the risk of kidney disease, while also delineating contributions by ART-specific toxicities, demographic factors and comorbidities to this important complication.
The authors acknowledge the contributions of W. Christopher Mathews and Stephen Boswell for their role in data acquisition and for their critical review of the article.
Financial support: this research was supported by the CFAR Network of Integrated Clinical Systems-CNICS, a National Institutes of Health (NIH) funded program (R24 AI067039) that was made possible by the National Institute of Allergy and Infectious Diseases (NIAID), with additional support from NIH: AI036219, P30 AI027757, P30 AI27763, UL1 RR024131 R01-DA11602, R01-AA16893, K24-DA 00432, K01-AI071754 and the Veterans Administration: VISN10 Geriatric Research Educational and Clinical Centers, Louis Stokes Cleveland Veterans Administration Medical Center.
Conflicts of interest
R.C.K. and Z.K. received research support from Gilead Sciences; J.H.W. received research support from the Bristol-Myers Squibb, Pfizer, Tibotec Therapeutics, and Definicare, and consulted for Bristol-Myers Squibb and Gilead Sciences.
There are no conflicts of interest for B.L., R.N.M., H.M.C., B.R., R.A.S., Z.K., J.N.M., R.D.M., J.J.E., and M.M.K.