Persistent viremia in human immunodeficiency virus/hepatitis B coinfected patients undergoing long-term tenofovir: Virological and clinical implications
Download the PDF here
Download the PDF here
"Patients were selected from the French HIV-HBV Cohort Study.....A total of 111 HIV-HBV-infected patients undergoing TDF-containing antiretroviral therapy were prospectively followed.....At TDF initiation, 78 (70.3%) patients were HBeAg positive and 92 (82.9%) had detectable HBV-DNA with a median HBV-DNA VL at 5.19 log10 IU/mL (interquartile range [IQR]: 3.16-7.01). All patients had previously initiated ARV therapy (median duration: 6.6 years; IQR, 4.2-8.5), thus median CD4+ cell count was at 396/mm3 (IQR, 292-552) and 62 (55.5%) had undetectable HIV-RNA. Of the 101 patients (91.0%) with previous LAM exposure, median LAM duration was 50.3 months (IQR, 31.2-71.2) at TDF initiation. Patients initiated TDF in combination with LAM (n = 82; 73.9%), alone (n = 24; 21.6%), with emtricitabine (FTC; n = 3; 2.7%), or with LAM plus FTC (n = 2; 1.8%).
In this large, prospective study among patients on continuous TDF therapy for a median of 7 years, without ETV or Peg-IFN intensification, we demonstrate that 77.5% had optimal HBV suppression. In general, these patients were more likely to be either at later phases of HBV infection or have had more successful anti-HBV therapy preceding TDF initiation (higher proportion of HBeAg negative and lower ALT levels)."
"In conclusion, PV (persistent viremia) hinders important serological endpoints during TDF treatment in HIV-HBV coinfected patients, yet this does not necessarily translate to increased risk in liver-related morbidity or mortality. No specific mutations on the pol gene were associated with PV, yet other genetic aspects of HBV need to be more thoroughly evaluated. Whether this phenomenon is caused by inadequate adherence remains to be concretely evaluated; nevertheless, our data suggest that roughly half of patients with PV were possibly the result of poor adherence. Other host factors, particularly immunoregulation, merit further investigation."
Persistent viremia in human immunodeficiency virus/hepatitis B coinfected patients undergoing long-term tenofovir: Virological and clinical implications
advance pub June 20 2014
Anders Boyd,1 Joel Gozlan,2,3 Sarah Maylin,4,5,6 Constance Delaugerre,4,5,6 Gilles Peytavin,7
Pierre-Marie Girard,1,8,9 Fabien Zoulim,10,11,12 and Karine Lacombe1,8,9
This study has been funded, in part, by Sidaction (AO 19) and received additional grants from ANRS (Agence Nationale de Recherche sur le Sida et les Hepatites). Gilead Sciences, Inc., provided an unrestricted grant for the French HIV-HBV cohort. Roche Diagnostics gave additional funding for antigen quantification. Both Gilead Sciences, Inc., and Roche Diagnostics were not involved in any part of the data collection, analysis, and manuscript writing. A.B. was awarded a postdoctoral fellowship from the ANRS.
Tenofovir (TDF) is considered the ideal treatment for patients coinfected with human immunodeficiency virus (HIV) and hepatitis B virus (HBV). However, certain coinfected patients exhibit incomplete viral suppression, with persistent, and sometimes transient, bouts of HBV replication. The reasons for this, including clinical effect, are unclear. A total of 111 HIV-HBV-infected patients undergoing TDF-containing antiretroviral therapy were prospectively followed. Serum HBV-DNA viral load, hepatitis surface (HBsAg) and e antigen (HBeAg) status were obtained at baseline and every 6-12 months. Amino acid (aa) changes on the polymerase gene were assessed using direct sequencing after nested polymerase chain reaction in patients with persistent viremia (PV).
After a median of 74.7 months (interquartile range: 33.4-94.7),
virological response (VR; <60 IU/mL) occurred in 96 of 111 (86.5%) patients. Of these, 86 of 96 (89.6%) remained completely undetectable during follow-up (stabilized VR). The remaining 10 of 96 (10.4%) patients had a transient blip of detectable HBV-DNA (transient PV), during which time 9 of 9 (100%) with available samples had detectable plasma TDF. Low-level PV (LL-PV; 61-2,000 IU/mL) was observed in 11 of 111 (9.9%) patients, the majority of which had detectable plasma TDF (8 of 9; 88.9%). High-level PV (>2,000 IU/mL) was rare (4 of 111; 3.6%) and was associated with non adherence.
At TDF initiation, patients with stabilized VR had significantly higher nadir CD4+ count, compared to those with transient PV (P = 0.006) or LL-PV (P = 0.04). No consistent aa changes, other than those associated with lamivudine resistance, were observed in patients with persistent viremia.
Importantly, HBeAg loss, HBeAg seroconversion, and HBsAg loss only occurred in patients with stabilized VR.
Two patients with stabilized VR developed hepatocellular carcinoma and 2 with LL PV died, 1 of a liver-related cause. Conclusion: Suboptimal HBV control during TDF treatment has a negative effect on serological outcomes, but not necessarily clinical events. Immunoregulation may provide more insight into this phenomenon.
aa amino acid
aOR adjusted odds ratio
ALT alanine transaminase
AST aspartate transaminase
CI confidence interval
HBeAg hepatitis B e antigen
HBsAg hepatitis B surface antigen
HBV hepatitis B virus
HCC hepatocellular carcinoma
HCV hepatitis C virus
HDV hepatitis D virus
HIV human immunodeficiency virus
HL-PV high-level persistent viremia
IQR interquartile range
LL-PV low-level persistent viremia
LOWESS locally weighted scatterplot smoothing
NA nucleoside/nucleotide analog
PCR polymerase chain reaction
Peg-IFN pegylated interferon
PEIU Paul Ehrlich Institute units
PV persistent viremia
qHBeAg quantification of hepatitis B e antigen
qHBsAg quantification of hepatitis B surface antigen
RT reverse transcriptase
VL viral load
VR virological response
In patients infected with human immunodeficiency virus (HIV), hepatitis B virus (HBV) has emerged as a major cause of morbidity and mortality, accounting for substantial rises in cirrhosis, hepatocellular carcinoma (HCC), and end-stage liver disease. The incidence of these diseases can be greatly curbed by controlling HBV replication.[2, 3] There are currently a variety of antiviral drugs available for suppressing HBV, such as lamivudine (LAM), adefovir (ADV), entecavir (ETV), and tenofovir (TDF). The lattermost antiviral is ideal for treating coinfected patients as a result of its dual anti-HBV and -HIV activity as well as superior potency in suppressing HBV-DNA replication.[4, 5] Coupled by the fact that no mutations associated with TDF resistance have been consistently described in both HBV- and HIV-HBV- infected patients over extensive follow-up,[4, 6] it would then appear that optimal HBV suppression could be easily achieved in coinfected patients.
However, recent data have emerged suggesting that some patients exhibit suboptimal virological suppression after long periods of antiretroviral (ARV) therapy, where serum HBV-DNA could transiently, and even persistently, replicate. A number of reasons for persistent viremia have been broached in the literature. First, guidelines often point toward poor adherence as a major source of unfavorable virological outcomes, which may likely be the case given previous research.[6, 7] Second, HBV-specific T-cell response and intrahepatic T-cell activation are diminished in the presence of HIV, with incomplete reconstitution during prolonged ARV therapy.[9, 10] Immunological features of HIV-HBV coinfection, including degree of immunosuppression, might explain why certain patients exhibit suboptimal HBV suppression. Third, data from in vitro studies suggest that certain amino acid (aa) substitutions on the polymerase (pol) gene, in combination with precore mutations, can give rise to TDF resistance,[11, 12] despite strong clinical evidence against it.[13, 14]
The aim of the present study was then to thoroughly evaluate various virological and clinical characteristics in explaining suboptimal virological suppression in HIV-HBV coinfected patients undergoing extensive treatment with TDF-containing ARV therapy. We then develop our analysis by further exploring the effect of persistent viremia on HBV serological outcomes, which include two markers gaining substantial interest in predicting therapeutic success-hepatitis B surface antigen quantification (qHBsAg) and hepatitis B e antigen quantification (qHBeAg). Finally, we wished to determine whether patients experiencing persistent viremia are at risk of developing liver-related morbidity or mortality.
In this large, prospective study among patients on continuous TDF therapy for a median of 7 years, without ETV or Peg-IFN intensification, we demonstrate that 77.5% had optimal HBV suppression. In general, these patients were more likely to be either at later phases of HBV infection or have had more successful anti-HBV therapy preceding TDF initiation (higher proportion of HBeAg negative and lower ALT levels).
Poor adherence is often considered the first cause of persistent viremia to suspect among patients undergoing NA therapy with a high genetic barrier to resistance. Although 18 of the 21 patients with persistent viremia and available samples had detectable plasma TDF at comparable levels to those with undetectable HBV-DNA, it may not give the overall picture of adherence over longer periods of time. Some clues can be inferred from rates of detectable HIV-RNA, where adherence reaching as low as 75% for some ARV classes is associated with effective HIV control. Assuming that overall rates of undetectable HIV-RNA greater than 70% represent optimal adherence (HIV-RNA replication being explained by other factors, such as HIV resistance to non-nucleoside RT inhibitors or protease inhibitors), PV could then be the result of fairly poor adherence in an estimated 11 of 25 patients. Still, the optimal adherence level for HBV suppression has not been determined; nevertheless, preliminary data suggest it to be much higher than HIV, from anywhere above 95%-100%.[7, 27]
Another potential reason of PV could be the genetic variability of HBV. The aa change at position rtL217R, which has already been identified with poor VR to ADV, was observed in 4 patients. To the best of our knowledge, this specific mutation has not yet been phenotypically analyzed vis-a-vis tenofovir. A conserved site change at position rtR274W was also observed in 1 patient. A substitution located at the same position (rtR274Q/R) had been identified in 1 patient from the GS-US-174-0103 Study as the result of TDF discontinuation.
Interestingly, a highly mixed group of quasi-species was also found in this patient, whereas phenotypic evaluation of rtR274Q revealed a TDF-sensitive strain. Our study group has also identified a TDF-treated patient, not included in the present study, who harbored the possible TDF-resistance mutation, rtA194T, yet exhibited controlled HBV suppression. Taken together, we found no dominant virological characteristic on the pol gene, at both baseline and follow-up, that was consistently associated with persistent replication. Wider ranges of sequencing would be necessary in determining whether mutations on the preS/S and preC/C regions of the HBV genome are also associated with ongoing replication. Moreover, one cannot exclude that PV may favor the occurrence of compensatory mutations, especially considering that higher variability of HBV quasi-species has been previously observed with poor response.
Interestingly, patients with transient PV had significantly lower CD4+ cell counts, compared to other replication groups, whereas nadir CD4+ cell count was significantly lower among patients with PV overall. Because adaptive cellular mechanisms are intrahepatically impaired in HIV-HBV coinfection, it could be hypothesized that the loss of HBV-specific immunological control in the liver might have elicited bouts of transient or low-level replication. Any changes in genetic variability observed during follow-up would likely be mere reflections of the quasi-species originating from a persistent reservoir, such as within hepatocytes. Of course, more-exhaustive data would be needed to address this hypothesis. It should be noted that nadir CD4+ cell count was not associated with undetectable HBV-DNA in univariable analysis. Yet, this analysis does not differentiate detectable HBV-DNA during normal treatment response versus PV, underscoring the importance of defining HBV replication groups.
Much ambiguity surrounds the clinical effect of persistent HBV replication in HIV-HBV coinfected patients. Data from the REVEAL-HBV study observed that among patients with HBV replication consistently at <104 copies/mL, the incidence rate of HCC was very low, at 73.4 per 100,000 person-years. This association could be considered no different in coinfection, given the lack of increased risk in HCC between HIV-HBV- and HBV-infected patients.[31, 32]Indeed, inasmuch as our data were limited by the low incidence of clinical events, we could observe no evidence indicating that PV per se was linked to increased liver-related morbidity and mortality. Nevertheless, there was a clear link between consistently undetectable HBV-DNA and HBeAg loss, HBeAg seroconversion, and HBsAg loss, all of which are ideal therapeutic endpoints and correspond to decreased viral activity and improved liver-related outcomes. This is in line with a previous study of HBV-monoinfected, NA-treated patients, where HBsAg loss only occurred in patients who were able to achieve HBV-DNA <300 copies/mL.
One important and disconcerting observation was the significantly shorter duration of follow-up among patients with LL-/HL-PV, when compared to other groups. When matching follow-up of patients with LL-/HL-PV to those with stabilized VR or transient PV in a sensitivity analysis (Supporting Methods), we observed that 12.5% of the latter patients were reclassified as having some form of LL/HL virological persistence. Defining persistent replication at time points with short follow-up appears to be prone to misclassification. These changes did not substantially modify the analysis, especially with respect to serological endpoints (Supporting Tables 1-3); however, the 9 patients newly classified as HL-PV had characteristics typical of coinfected patients at risk of delayed response,including high HBV-DNA and rather high increases in transaminase levels. This puts into question previous studies where detectable VL at 6-12 months is used as the hallmark of nonresponse or delayed response,[7, 30, 34] even as the majority of these patients eventually achieve VR.
Our study has several limitations that need to be addressed. First, the threshold for transient HBV-DNA replication was particularly low and HBV VLs were only evaluated every 6-12 months. Patients with stabilized VR profiles could have been misclassified. Second, qHBeAg was obtained by two different assays and rendered similar using a post-hoc approach. Some bias could have been generated for certain measures, especially the 12-month change in qHBeAg. Finally, we did not record HIV resistance mutations during follow-up, which may have helped explain certain instances during which HIV-RNA was detectable.
In conclusion, PV hinders important serological endpoints during TDF treatment in HIV-HBV coinfected patients, yet this does not necessarily translate to increased risk in liver-related morbidity or mortality. No specific mutations on the pol gene were associated with PV, yet other genetic aspects of HBV need to be more thoroughly evaluated. Whether this phenomenon is caused by inadequate adherence remains to be concretely evaluated; nevertheless, our data suggest that roughly half of patients with PV were possibly the result of poor adherence. Other host factors, particularly immunoregulation, merit further investigation.
Baseline Description of the Study Population
At TDF initiation, 78 (70.3%) patients were HBeAg positive and 92 (82.9%) had detectable HBV-DNA with a median HBV-DNA VL at 5.19 log10 IU/mL (interquartile range [IQR]: 3.16-7.01). All patients had previously initiated ARV therapy (median duration: 6.6 years; IQR, 4.2-8.5), thus median CD4+ cell count was at 396/mm3 (IQR, 292-552) and 62 (55.5%) had undetectable HIV-RNA. Of the 101 patients (91.0%) with previous LAM exposure, median LAM duration was 50.3 months (IQR, 31.2-71.2) at TDF initiation. Patients initiated TDF in combination with LAM (n = 82; 73.9%), alone (n = 24; 21.6%), with emtricitabine (FTC; n = 3; 2.7%), or with LAM plus FTC (n = 2; 1.8%).
Determinants of Undetectable HBV-DNA
Median follow-up was 74.7 months (IQR, 33.4-94.7). Risk factors of undetectable HBV-DNA VL over time are given in Table 1. Increased baseline HBV-DNA levels, longer duration of HBeAg-positive serology over time, and higher ALT levels over time were factors with significantly lower odds of undetectable HBV-DNA VL, whereas longer TDF duration was significantly associated with higher odds. Of note, neither cumulative LAM duration nor FTC duration were associated with undetectable HBV-DNA VL in multivariable analysis (adjusted odds ratio [aOR]/year = 1.00, 95% confidence interval [CI]: 0.99-1.01; aOR/year = 0.98, 95% CI: 0.97-1.01, respectively).
Distribution and Description of HBV Replication Profiles
Undetectable HBV-VL at the end of follow-up occurred in 96 (86.5%) patients, with 86 (77.5%) remaining completely undetectable during treatment (stabilized VR; Fig. 1). Among them, HIV-RNA was undetectable for a median 90.0% of study visits (IQR, 75.0%-100.0%). In the other 10 patients, detectable HBV replication was observed after a median of 8.8 months (range, 4.5-29.5) after having achieved undetectable HBV-VL (transient PV), during which time median peak HBV-DNA VL was 1,155 IU/mL (range, 68-7,162). Only 2 of these patients had a concomitant peak in qHBsAg during blip in HBV replication (Supporting Fig. 1). At the same time, plasma HIV-RNA was undetectable in 6 (60%) patients, low in 3 (30%; HIV-RNA VL range: 116-1,953 copies/mL), and fairly high in the remaining patient (HIV-RNA VL: 139,189 copies/mL).
In patients with detectable HBV-VL at the end of follow-up, LL-PV was observed in 11 (9.9%) patients, of whom 3 had achieved HBV VL <60 at least once during follow-up. In this group, median HBV-DNA VL was 111 IU/mL (range, 78-860) at the last visit. Meanwhile, HIV-RNA VL was undetectable for 7 (63.6%) patients and high for the remaining 4 (36.4%; range HIV-RNA VL: 3,421-590,000 copies/mL). HL-PV was found in 4 (3.6%) patients, among whom HBV-DNA ranged from 3.33 to 5.00 log10 IU/mL at the last visit. Only 1 patient achieved an undetectable HBV-VL during the course of follow-up. At the last visit, HIV-RNA was undetectable for 2 patients, whereas the remaining 2 had low HIV-RNA VLs (1,070 and 1,100 copies/mL).
Baseline Characteristics and HBV Replication Profiles
Table 2 describes the baseline characteristics of patients between profile groups. Because of small numbers of patients, LL-PV and HL-PV groups were combined for analysis. Among those with detectable baseline HBV-DNA, median HBV-DNA was higher for those with stabilized VR and LL-/HL-PV profiles than transient PV (P = 0.09 and 0.01, respectively). Patients with stabilized VR, as compared to all PV profiles, were significantly more likely to be HBeAg negative (P = 0.001) and have higher CD4+ nadir count (P = 0.004). Finally, stabilized VR and transient PV profiles had significantly lower baseline ALT/AST levels, when compared to LL-/HL-PV (P < 0.04).
On-Treatment Virological and Serological Characteristics per HBV Replication Profile
In patients with detectable baseline HBV-DNA, median time until undetectable HBV-VL was not different between patients with stabilized VR versus transient PV profiles (12.6 vs. 23.2 months, respectively; Fig. 2A). In HBeAg-positive patients, it would appear that qHBeAg decline was faster in patients with stabilized VR and LL-/HL-PV profiles (Fig. 2B). These profile groups also had slightly steeper 12-month ∇qHBeAg, compared to transient PV profiles, with more patients exhibiting >1.0 log10 12-month PEIU/mL decreases (Table 3); however, there were no significant differences. Overall, faster 12-month decline in qHBsAg (Fig. 2C; Table 3) was also observed among patients with stabilized VR and LL-/HL-PV profiles versus transient PV (P = 0.15 and P = 0.04, respectively). Importantly, HBeAg-loss, HBeAg seroconversion, and HBsAg loss only occurred in patients with stabilized VR.
Genotypic Characteristics of Patients With Persistent Viremia
Table 4 gives a complete description of the genotypic findings among patients with transient PV. Four of the six patients with available baseline sequences had LAM resistance mutations (rtM204V, n = 3; rtV173V/L+rtL180M+rtM204V: n = 1). At the time point where transient HBV replication occurred, the majority of mutations were associated with LAM resistance, with other aa changes observed at positions rtL217R (n = 2) and rtR274W (n = 1).
Among those with LL-PV (Table 5A), 3 of 4 patients with available baseline sequences had the triple rtV173L+rtL180M+rtM204V LAM resistance mutation, whereas 1 patient also harbored an HBV strain with aa changes at position rtL217R. During TDF treatment, no changes from baseline sequences were observed in these patients. Among those with HL-PV (Table 5B), all had LAM resistance mutations at baseline with other aa changes at positions rtL80I (n = 1) and rtL217R (n = 1). No other aa changes were noted during follow-up, except in 1 patient with a change at position rtV173L.
TDF Levels During PV
Among patients with transient PV, all 9 with available plasma samples had detectable TDF concentrations (median, 62 ng/mL; IQR, 54-67). In patients with LL-PV, 8 of 9 patients with available samples had TDF detected (median, 89 ng/mL; IQR, 82-121). Finally, in patients with HL-PV, 2 of 3 with available samples had plasma TDF <5 ng/mL, whereas the remaining patient was considered as having low TDF levels (30 ng/mL), according to the interval between last drug intake and sampling. Patients exhibiting PV did not have significantly different TDF concentrations, when compared to randomly selected patients with undetectable HBV-DNA (median, 65 ng/mL; IQR, 42-100; P = 0.6).
Transaminases and Liver-Related Outcomes
Overall, almost all patients observed declines in ALT levels, with a median maximum decline at -19 (IQR, -53, -8). Median maximum ALT decline only tended to be stronger in LL-/HL-PV than the transient PV group (P = 0.08; Table 3), which was expected given the high baseline transaminase levels in the LL-/HL-PV group. Otherwise, no other significant differences were observed in median maximum increase in ALT levels during follow-up or ALT levels at the last follow-up visit (Table 3).
No significant difference in Metavir equivalents (from the FibroMeter) was observed at baseline, yet no patient in the transient VR group had F3-F4 fibrosis (Table 2). In the subgroup with liver biopsies, taken after a median of 5.3 months of TDF (IQR, 0-27.7), there were no significant differences in Metavir fibrosis or activity level between stabilized VR and other PV groups: transient PV, 0.1 (95% CI: -0.7, 0.8; P = 0.9) or -0.1 (95% CI: -0.6, 0.4; P = 0.8); LL-/HL-PV, 0.5 (95% CI: -0.2, 1.2; P = 0.14) or 0.4 (95% CI: -0.1, 0.9; P = 0.09), respectively.
At TDF initiation, 3 patients had already experienced a severe liver-related event (n = 1, hepatic failure and n = 2, portal hypertension). Of these, 2 had stabilized VR and 1 had LL-PV. During TDF treatment, 2 patients developed HCC (both with stabilized VR replication profiles) and 2 died (both with LL-PV replication profiles). One death was the result of severe pneumonia, whereas the other was the result of a liver-related complication after the patient discontinued TDF therapy, resulting in an increase of HBV-DNA to 7.56 log10 IU/mL. No liver-related morbidity or mortality was observed in patients with HL-PV.
Patients and Methods
Patients were selected from the French HIV-HBV Cohort Study. Briefly, this prospective study recruited 308 patients from seven centers located in Paris and Lyon, France, during May 2002-May 2003. Patients were included if they had HIV-positive serology confirmed by western blotting and HBsAg-positive serology for at least 6 months. Data collection continued until 2010-2011.
We retained individuals from a previous analysis of 143 patients undergoing a TDF-containing ARV regimen. Additionally, we decided to only include patients with ³12 months of follow-up, because virological response in patients undergoing nucleoside/nucleotide analog (NA) therapy may require extensive periods of follow-up to be fully evaluated, particularly in HIV-HBV coinfection. As a result, 13 patients were excluded. Second, follow-up was right censored for patients who discontinued TDF use or had pegylated interferon (Peg-IFN) intensification, resulting in the exclusion of, respectively, 7 and 5 patients because of short follow-up (<12 months). Finally, patients with baseline hepatitis C virus (HCV)- or hepatitis D virus (HDV)-positive serology had been excluded in the previous analysis. However, we additionally excluded patients with incident HCV infection (n = 2) or with HCV/HDV infection later discovered during follow-up (n = 5). Thus, 111 patients were included in the present analysis. All patients provided written informed consent to participate in the cohort and the protocol was approved by the Pitie-Salpetriere Hospital Ethics Committee (Paris, France) in accord with the Declaration of Helsinki.
Quantification of Serum Viral Loads
HBV-DNA viral loads (VLs) were obtained at TDF treatment initiation and every 6-12 months during follow-up. Commercial polymerase chain reaction (PCR)-based assays were used to quantify HBV-DNA VL (COBAS AmpliPrep/COBAS TaqMan HBV Test; detection limit: 12 or 38 IU/mL; COBAS Amplicor HBV Monitor Test; detection limit: 60 IU/mL; Roche Diagnostic Systems, Meylan, France). Because of varying detection limits, we decided to define undetectable HBV-DNA at the highest threshold (<60 IU/mL). HIV-1 VLs were measured using either a branched-DNA (Quantiplex 3.0; detection limit: 50 copies/mL; Bayer Diagnostics, Cergy Pontoise, France) or real-time PCR technique (COBAS AmpliPrep/COBAS TaqMan HIV-1 Test; detection limit: 40 copies/mL; Roche Molecular Systems).
HBV Replication Profiles
HBV replication profiles were defined a priori based on HBV-VL at the end of follow-up, where patients were initially classified on whether or not they had detectable HBV-VL. Among those with undetectable HBV-VL, patients were divided into two subgroups based on previous research7: stabilized virological response (VR; achieving and/or constantly maintaining HBV-DNA <60 IU/mL) or transient persistent viremia (PV; achieving <60 IU/mL, intermittently >60 IU/mL thereafter). Among those with detectable HBV-VL, patients were divided into two subgroups based on HBV-VL levels established from previous guidelines18: low-level persistent viremia (LL-PV; 60-2,000 IU/mL) or high-level persistent viremia (HL-PV; >2,000 IU/mL).
Plasma TDF Concentrations
Plasma TDF was measured using ultra performance liquid chromatography combined with tandem mass spectrometry (lower limit of quantification: 5 ng/mL; Waters Corporation, Milford, MA), with the intent of assessing treatment adherence. Plasma concentrations were only performed in patients with PV. For patients with LL-/HL-PV, TDF was quantified at the last study visit. For patients with transient PV, TDF was quantified at the visit during which the blip in HBV replication occurred. If there was a prolonged blip (i.e., more than one consecutive visit with detectable HBV-VL), TDF concentrations were determined at the visit with the highest HBV-VL. To determine whether TDF concentrations were sufficient, we also quantified TDF levels in randomly selected patients with undetectable HBV-DNA (n = 34) at various visits during follow-up and compared them to patients with PV.
At baseline, L-nucleoside-associated pol gene mutations at positions rt173, rt180, and rt204 and precore mutations at nucleotide 1896 were determined using DNA chip technology (bioMerieux, Marcy l'Etoile, France). During follow-up, genotypic analysis was performed in patients with PV. Samples were obtained at baseline and at the same visits used to quantify plasma TDF. aa substitutions of pol were examined by direct sequencing after nested PCR amplification of the reverse transcriptase (RT) encoding region, between aa 42-309, using the ANRS consensus method adapted from a previous work. Sequences were analyzed using a sequence editing program (Seqscape; Applied Biosciences, les Ulis, France) and were further submitted to the "Genotypic prediction of HBV drug resistance" tool (Geno2pheno[HBV] 2.0; available at: http://hbv.bioinf.mpi-inf.mpg.de).
Serostatus and Quantification of Serological Markers
Qualitative hepatitis B surface antigen (HBsAg), hepatitis B e antigen (HBeAg), and antibodies were detected at baseline and every yearly visit using a commercial enzyme-linked immunoassay (Diasorin, Antony, France).
Antigen levels were quantified at baseline and every 6-12 months during treatment. qHBsAg was performed using the Architect HBsAg assay (Abbott Laboratories, Rungis, France). qHBeAg was determined using the Architect assay (with Architect i2000 analyzer) during the first 3 years of follow-up and the Elecsys HBeAg assay thereafter (with Modular E170 analyzer; Roche Diagnostics). Using a previously designed protocol, signals from each assay were converted to Paul Ehrlich Institute units (PEIU)/mL. qHBeAg results from the Architect assay were then again converted into Elecsys units by an HIV-HBV-specific linear regression function. In other words, all qHBeAg units were expressed in PEIU/mL as if quantified from the Elecsys assay. Full details on the calculation of conversion factors are provided in the Supporting Methods (see the Supporting Information).
Assessing Liver-Related Parameters
Alanine transaminase (ALT) and aspartate transaminase (AST) were evaluated at baseline and every 6 months using standard methods. Baseline liver fibrosis levels were estimated using the noninvasive score FibroMeter, as detailed elsewhere, and converted into Metavir equivalents. Liver biopsies were available in 64 patients at some point during therapy (18 with two biopsies) and evaluated using Metavir classification. Any liver-related and non-liver-related clinical event was confirmed with the treating physician.
Risk-factor analysis for undetectable HBV-DNA over time was carried out using random-effects logistic regression, accounting for within-patient variance. Time-varying factors were assessed by including the product of TDF treatment duration and the covariable of interest, including its individual components, to the model. All risk factors with P ≤ 0.2 in univariable analysis were used to create a predictive, multivariable model. A forward step-wise selection process was then performed, removing any covariable with a P value greater than the above-described threshold upon adjustment.
In patients with detectable HBV-DNA at baseline, undetectable HBV-DNA rates at 12 and 24 months, including 12-month change in HBV-DNA, were evaluated. qHBsAg and qHBeAg during treatment were initially examined by locally weighted scatterplot smoothing (LOWESS) plot, whereas 12-month change, 12-month change >1.0 log10 in IU/mL (qHBsAg) or PEIU/mL (qHBeAg), and maximum change (defined as the nadir decrease from baseline) were also evaluated. Average HIV-RNA log10 copies/mL levels during follow-up (while using 50 copies/mL as a uniform detection threshold) and percentage of visits with undetectable HIV-RNA VLs overall were also summarized.
Comparisons between HBV replication profiles were performed for all clinical parameters at baseline and during follow-up using Kruskal-Wallis' test for continuous variables and Pearson's chi-square test or Fisher's exact test for categorical variables, unless otherwise stated. In the subset of patients with liver biopsies, liver fibrosis levels were compared between HBV replication groups using a mixed-effect linear regression adjusted for TDF duration, while accounting for within-patient correlation for those with two biopsies.
All statistical analyses were performed using Stata software (v12.1; StataCorp LP, College Station, TX), and significance was determined using a P value <0.05.