
Association Between Tenofovir Exposure and Reduced Kidney Function in a Cohort of HIVPositive Patients: Results From 10 Years of Followup



Download the PDF here
Clin Infect Dis (15 February) 2013
Claudie Laprise,1,2 JeanGuy Baril,3 Serge Dufresne,3 and Helen Trottier1,2
1Department of Social and Preventive Medicine, University of Montreal, 2SainteJustine Hospital Research Center, and 3Clinique Medicale du Quartier
Latin, Montreal, Quebec, Canada
"In this cohort, TDF exposure was associated with reduced kidney function, but the loss in eGFR attributable to TDF is relatively mild in a longterm perspective."
"There has been debate about the association between TDF exposure and renal dysfunction and about the clinical impact of the loss in eGFR due to TDF exposure. Our study shows that the association was not of a high magnitude and that the quantified loss in eGFR attributable to TDF is relatively modest after many years of exposure. Importantly, the loss attributable to TDF seems to occur during the first year of exposure and stabilizes after that. Although the loss is maintained, it does not seem to further deteriorate with additional years of exposure. The clinical impact of this association need to be analyzed, taking into account the efficacy of TDF, but it is highly plausible that TDF exposure, although associated with reduced kidney function, has no severe adverse effects over the long term for most HIVpositive patients."
Adjusted hazard ratios (HR) and odds ratios (OR) for the association between TDF and kidney dysfunction (defined as eGFR <90 mL/min/1.73 m2) were calculated using the Cox proportional hazards model and generalized estimating equations. Mean loss in eGFR attributable to TDF by cumulative years of exposure was estimated using linear regressions.
Abstract
Background. Some studies have shown that tenofovir disoproxil fumarate (TDF), a drug widely used in highly active antiretroviral therapy, is associated with kidney dysfunction, but the magnitude of the effect and its clinical impact is still being debated. Our objective was to evaluate the association between longterm TDF exposure and kidney dysfunction in a cohort of 1043 human immunodeficiency viruspositive patients followed up for 10 years and to quantify the loss in estimated glomerular filtration rate (eGFR) in patients exposed to TDF in comparison with those exposed to other antiretroviral therapies.
Methods. Adjusted hazard ratios (HR) and odds ratios (OR) for the association between TDF and kidney dysfunction (defined as eGFR <90 mL/min/1.73 m2) were calculated using the Cox proportional hazards model and generalized estimating equations. Mean loss in eGFR attributable to TDF by cumulative years of exposure was estimated using linear regressions.
Results. Tenofovir exposure increased the risk of kidney dysfunction by 63% (HR, 1.63; 95% confidence interval, 1.262.10). The cumulative eGFR loss directly attributable to TDF after 1, 2, 3, and 4 years of TDF exposure was 3.05 (P = .017), 4.05 (P = .000), 2.42 (P = .023), and 3.09 mL/min/1.73 m2 (P = .119), respectively, which shows that most of the loss occurred during the first years of exposure.
Conclusions. In this cohort, TDF exposure was associated with reduced kidney function, but the loss in eGFR attributable to TDF is relatively mild in a longterm perspective.
Kidney dysfunction is associated with morbidity and mortality in human immunodeficiency virus (HIV)positive patients and clinical followup is therefore important [1]. HIV infection itself has been identified as a risk factor for kidney dysfunction [2], as have many other nonHIVrelated factors, such as baseline glomerular filtration rate <90 or <60 mL/min/1.73 m2, high HIV load, low CD4 cell count, female sex, older age, black race, low weight and comorbid conditions (diabetes, hepatitis B or C, hypertension, proteinuria, albuminuria) [37]. Atazanavir, didanosine, indinavir, lopinavir/ritonavir, other nephrotoxic drugs, and recreational drugs, particularly cocaine, have also been identified as risk factors for loss of kidney function [6, 812]. Many studies have shown that tenofovir disoproxil fumarate (TDF), a drug widely used in highly active antiretroviral therapy (ART) is associated with kidney dysfunction, but the magnitude of the effect and its clinical impact are still being debated [13, 14]. Discrepancies in the literature can be explained by lack of power or limited sample size, dissimilar study populations, short followup, and different analysis strategies.
In an effort to inform the current debate, our objective was to evaluate the association between longterm TDF exposure and kidney dysfunction and to quantify the loss in estimated glomerular filtration rate (eGFR) attributable to TDF in a cohort of HIVpositive patients exposed to ART and followed up between January 2002 and March 2012.
DISCUSSION
In this study, we evaluated the association between TDF exposure and kidney function and quantified the mean loss in eGFR attributable to TDF through the cumulative years of exposure. The KaplanMeier curve shows that the cumulative incidence of reduced kidney function is higher among patients exposed to TDF than among those exposed to other ARVs (P = .024). We used 2 different regression models (GEE and Cox modeling) and found that there was a consistently higher risk of reduced kidney function associated with exposure to TDF than with any other ARVs. Cox modeling analysis showed that TDF exposure increased the risk of reduced kidney function by 63% (adjusted HR, 1.63; 95% CI, 1.262.10). Other researchers have made comparable findings. For example, in a large cohort of 10 841 HIVinfected patients with a median followup of 3.9 years, Scherzer et al [11] used a similar strategy with Cox modeling, although the outcome was defined by an eGFR <60 mL/min/1.73 m2. They found a 33% higher risk of chronic kidney disease, defined as 2 consecutive measurements of eGFR <60 mL/min/1.73 m2 ≥3 months apart, calculated with the MDRD formula. They also found that 11% of patients newly exposed to TDF had a rapid annual decline in eGFR (defined as a loss of ≥3 mL/min/1.73 m2 in 2 consecutive years).
We also analyzed determinants of renal function with GEE modeling, which allows multiple events (persistent kidney dysfunction) to be taken into account. We confirmed that TDF exposure increases the risk of renal dysfunction (adjusted OR, 1.63; 95% CI, 1.481.79). We observed no effect modification (interaction) for age, PI use, diabetes or hypertension in the association between TDF and renal function. For example, older patients (aged ≥50 years) exposed to TDF were at no greater risk of kidney dysfunction than those <50 years old.
We also found that adjusted loss in eGFR directly attributable to TDF exposure was 3.05 mL/min/1.73 m2 after 1 year of treatment. In a systematic review of 17 studies (including 9 randomized controlled trials), Cooper et al [13] found that the pooled loss attributable to TDF was 3.72 mL/min/1.73 m2.
Importantly, we found that adjusted loss in eGFR directly attributable to TDF exposure was 3.05 mL/min/1.73 m2 after 1 year of treatment and that this loss stayed relatively constant through the following years of exposure. Losses were 4.05(P = .000), 2.42 (P = .023), 3.09 (P = .119), 0.12 (P = .946), and 0.32 mL/min/1.73 m2 (P = .898) after 2, 3, 4, 5, and ≥6 years of exposure, respectively. This seems to indicate that the loss induced by TDF occurred mainly during the first year of exposure and stayed relatively constant afterward. Although comparison has to be made with caution as the CIs overlap, it is also possible that a few losses occurring during the first 2 years of exposure to TDF were recovered in the following years (the mean loss attributable to TDF was lower after 3 years than after 1 or 2 years).
This analysis calculated the mean loss for the entire cohort; one might wonder about the possible individual variability, that is, whether the effect might be large in a small group. However, the standard deviation, median, and IQR for the exposed and unexposed groups were comparable. For example, after ≥6 years of TDF, a mean loss of 10.46/min/1.73 m2 was observed in exposed patients (SD, 14.00, median, 8.76; IQR, 18.51 to 3.31), whereas in patients exposed to other ARVs it was 9.42 mL/min/1.73 m2 (SD, 12.50; median, 8.56; IQR, 17.62 to 1.20).
As in other studies, we found an increased risk of renal dysfunction in older patients, those with a lower eGFR at baseline, and those exposed to PIs, as well as a reduced risk with NRTI use [3, 5, 7, 10, 6]. Surprisingly, however, in the GEE models, we found a negative association with black race and smoking as well as a positive association with INSTI use, contrary to what is reported in the literature [1923]. One possible reason for the negative association with black race is the adjustment for race in the formulas used to estimate the glomerular function rate, such as CKDEPI or MDRD. Black patients are scored higher than those of other races to compensate for their lower glomerular function rate. It is therefore to be expected that regression models using eGFR as the dependant variable will find a negative association with black race. The same results were observed recently in the large D:A:D cohort study [12].
Our study has strengths and limitations. One of the strengths is the long followup period. The sample was also of considerable size. Furthermore, ours is a reallife cohort including older patients, making the results more generalizable to a clinicbased cohort of HIVpositive patients, although there was a low number of black and female patients. The cohort was essentially composed of white homosexual men and included a large proportion of IDUs. Among the limitations are also the observational design (nonrandomized) and the potential for the presence of confounding by indication. However, we made a very conservative adjustment for potential confounding in every multivariate model. Although it would have been interesting to include hepatitis C status (variable not collected), most of its potential confounding effect might be controlled with the inclusion of injection drug use in the models. We also controlled for monthly income and type of employment, which are known to reflect global health and lifestyle. Also, it would have been interesting to analyze the association between TDF and renal function in our cohort using an eGFR <60 mL/min/1.73 m2. However, few patients in our clinicbased cohort had this outcome, because eGFR seldom goes below 60 mL/min/1.73 m2 without changes in ART. Because we used a cut point at 90 mL/min/1.73 m2, our study demonstrated a loss of kidney function and not chronic kidney disease as formally defined by the National Kidney Foundation [18]. Moreover, it would have been interesting to have results on urinalysis or proteinuria, but these data were not available. In addition, although the sample was large, we did not have the power to look at the impact of every ARV taken individually.
Another limitation was the lack of power to evaluate the loss in eGFR attributable to TDF after >4 years of exposure. Although the loss in eGFR attributable to TDF seems to occur during the first year and then stay relatively stable after that, it is not possible to draw definite conclusion about the magnitude of the effect after 4 years of exposure. This needs to be elucidated. Furthermore, it is possible that we underestimated the loss in eGFR attributable to TDF, because patients in the TDFexposed group may have been more likely to experience significant changes in eGFR and to switch to alternative ARTs, leaving patients with more stable eGFRs in the analysis. Finally, a large number of patients were excluded from the analysis because they had not visited the clinic within the last 2 years. However, they were very similar at baseline to the active patients in term of age and sex (data not shown). CD4 cell counts (cells/mm3) were also similar (471 [IQR, 310680] for excluded patients vs 480 [260900] for included patients). The median viral loads at baseline differed slightly, at 2506 HIV RNA copies/mL (4927 851) for excluded patients and 1550 HIV RNA copies/mL (4942 285) for active patients. It is possible that excluded patients were less "healthy," which may have affected the results of the study.
There has been debate about the association between TDF exposure and renal dysfunction and about the clinical impact of the loss in eGFR due to TDF exposure. Our study shows that the association was not of a high magnitude and that the quantified loss in eGFR attributable to TDF is relatively modest after many years of exposure. Importantly, the loss attributable to TDF seems to occur during the first year of exposure and stabilizes after that. Although the loss is maintained, it does not seem to further deteriorate with additional years of exposure. The clinical impact of this association need to be analyzed, taking into account the efficacy of TDF, but it is highly plausible that TDF exposure, although associated with reduced kidney function, has no severe adverse effects over the long term for most HIVpositive patients.
RESULTS
Of 2352 patients, 2058 were followed up after January 2002, 630 were excluded because they were not actively followed up at the clinic (had not been seen at least once in the last 2 years), 242 were excluded because they had no eGFR measurements or missing information about race, which is required for the calculation of CKDEPI, and 143 were excluded because they were not exposed to any ARVs. The baseline characteristics of the 1043 patients included in the analysis are shown in Table 1. The median length of followup was 7.9 years. The median age at baseline was 39.3 years, and 3.8% of the cohort were women. The median number of eGFR measurements (per patient) for the TDFexposed group was 19 (interquartile range [IQR], 1030) and 25 (IQR, 1534) for the unexposed group. The median time between serum creatinine measurements was 97 days (IQR, 84125 days) for the exposed group and 98 days (IQR, 84120 days) for the unexposed group. The group exposed to TDF differed from the unexposed group (exposed to other ARVs) in terms of monthly income: 20.4% of those exposed have a monthly income <$1500, in contrast to 14.7% of those not exposed. The HIVRNA viral load at baseline was lower in the unexposed group than in the exposed group (median, 467 and 2198 HIV RNA copies/mL, respectively), and IDUs were more frequent in the unexposed group (41.2% vs 33.6%).
Association Between Tenofovir and Reduced Kidney Function
Figure 1 shows the cumulative incidence of reduced kidney function (defined as 2 consecutive measurements of eGFR <90 mL/min/1.73 m2 ≥3 months apart), over a 10year period, by exposure status. The cumulative incidence of reduced kidney function after 2, 5, and 10 years of exposure to TDF was 15.12% (95% CI, 11.1520.34), 31.47% (95% CI, 26.1737.54), and 52.29% (95% CI, 45.6559.26), respectively. In the unexposed group, the cumulative incidence after 2, 5, and 10 years was 10.27% (95% CI, 7.9413.23), 25.89% (95% CI, 21.7830.62), and 40.53% (95% CI, 34.8246.80), respectively. The logrank test showed that the difference is statistically significant (P = .024).
In total, there were 271 incident cases of reduced kidney function for 4285.00 personyears (incidence rate [IR], 63.2 per 1000 personyears; 95% CI, 56.271.2). In the TDFexposed group, 133 events occurred for 1809.66 personyears (IR, 73.5 per 1000 personyears; 95% CI, 62.087.1) and for the unexposed group, 138 events occurred for 2475.34 personyears (IR, 55.8 per 1000 personyears; 95% CI, 47.265.9).
Table 2 shows the results for the Cox proportional hazards regression model. Tenofovir exposure increased the risk of reduced kidney function by 63% (adjusted HR, 1.63; P = .000) compared with other ARV exposure. Other determinants associated with a significant increased risk were older age at baseline (HR, 1.03; P = .004; every additional year of age increased the risk by 3%), lower eGFR at baseline (HR, 0.93; P = .000; every additional mL/min/1.73 m2 of eGFR at baseline decreased the risk by 7%), and alcoholic status (HR, 2.04; P = .022). NRTI users (HR, 0.39; P = .019) had a lower risk than nonusers of NRTIs, and PI users (HR, 1.46; P = .018) had a higher risk than nonusers of PIs. The relation between TDF and kidney function was not affected by interaction with age, use of PIs, diabetes, or hypertension (data not shown).
Table 3 reports the results for the GEE logistic regression models. Exposure to TDF increased the risk of reduced kidney function by 63% (odds ratio [OR], 1.63; P = .000). Other determinants that increased the risk significantly were older age (OR, 1.06; P = .000; 6% greater risk for each additional year of age), lower eGFR at baseline (OR, 0.93; P = .000), alcoholic status (OR, 1.57; P = .020), INSTI use (OR, 1.49; P = .000), and PI use (OR, 1.82; P = .000). Finally, black patients had a lower risk than patients of other races (OR, 0.39; P = .019), and smokers had a lower risk than patients who had never smoked (OR, 0.73; P = .000).
Quantification of Mean Loss in eGFR
Table 4 quantifies mean loss in eGFR by exposure status. Cumulative mean loss in eGFR was calculated after 1, 2, 3, 4, 5, and ≥6 years of exposure to TDF or any other ARVs. For the group exposed to TDF, cumulative mean loss increased consistently over the years, from 3.31 mL/min/1.73 m2 after 1 year of exposure to 10.46 mL/min/1.73 m2 after ≥6 years. For the unexposed group, cumulative mean loss in eGFR started at 0.25 mL/min/1.73 m2 after 1 year of exposure and reached 9.42 mL/min/1.73 m2 after ≥6 years. The univariate models show the cumulative mean loss directly attributable to TDF exposure. The multivariate models show that the adjusted mean loss directly attributable to TDF exposure after 1 year of exposure was 3.05 mL/min/1.73 m2. However, there was no clear trend associated with the number of years of exposure; the adjusted cumulative mean loss were relatively stable over the years (3.05, 4.05, 2.42, and 3.09 mL/min/1.73 m2 after 1, 2, 3, and 4 years of exposure, respectively) indicating that most of the loss was acquired during the first year of exposure and stabilized after that.
 




