A Metaanalysis of the Efficacy and Safety of Unboosted Atazanavir

Home , Zalcitabine

A meta-analysis of the efﬁcacy and safety of unboosted atazanavir compared with ritonavir-boosted protease inhibitor maintenance therapy in HIV-infected adults with established virological suppression after induction

J Baril,1 B Conway,2 P Giguère,3 N Ferko,4 S Hollmann4 and JB Angel3 1Hospital of the University of Montreal, Montréal, QC, Canada, 2The University of British Columbia, Vancouver, BC, Canada, 3Ottawa Hospital Research Institute, Ottawa, ON, Canada and 4Cornerstone Research Group, Burlington, ON, Canada

Objectives Treatment simplification involving induction with a ritonavir (RTV)-boosted protease inhibitor (PI) replaced by a nonboosted PI (i.e. atazanavir) has been shown to be a viable option for long-term antiretroviral therapy. To evaluate the clinical evidence for this approach, we conducted a systematic review and meta-analysis of randomized controlled trials (RCTs) evaluating efficacy and safety in patients with established virological suppression. Methods Several databases were searched without limits on time or language. Searches of conferences were also conducted. RCTs were included if they compared a PI/RTV regimen to unboosted atazanavir, after induction with PI/RTV. The meta-analysis was conducted using a random effects model for the proportion achieving virological suppression (i.e. HIV RNA < 50 and <400 HIV-1 RNA copies/mL), CD4 cell counts, lipid levels and liver function tests. Dichotomous outcomes were reported as risk ratios (RRs) and continuous outcomes as mean differences (MDs). Results Five studies (n = 1249) met the inclusion criteria. The meta-analysis demonstrated no statistically significant difference in efficacy (i.e. HIV RNA < 50 copies/mL) between PI/RTV and unboosted atazanavir [RR = 1.04; 95% confidence interval (CI) 0.99 to 1.10], with no heterogeneity. Findings were similar in a subanalysis of studies where atazanavir/RTV was the only PI/RTV used during induction. Additional efficacy results support these findings. A significant reduction in total cholesterol (P < 0.00001), triglycerides (P = 0.0002), low-density lipoprotein (LDL) cholesterol (P = 0.009) and hyperbilirubinaemia (P = 0.02) was observed with unboosted atazanavir vs. PI/RTV. Conclusions The meta-analysis demonstrated that switching patients with virological suppression from an RTV-boosted PI to unboosted atazanavir leads to improvements in safety (i.e. blood parameter abnormalities) without sacrificing virological efficacy. Keywords: atazanavir, HIV, meta-analysis, ritonavir, unboosted. Accepted 30 October 2013

Introduction HIV type 1 (HIV-1) infection can be successfully treated with a broad range of antiretroviral therapies. Current Correspondence: Dr Jean-Guy Baril, Centre Hospitalier de l’Université de Montréal, 264, René-Lévesque Blvd. E, Montréal, QC, H2X 1P1, Canada. guidelines recommend that initial regimens consist of two Tel: 514 285 5500; fax: 514 285 2226; e-mail: [email protected] nucleoside reverse transcriptase inhibitors (NRTIs), plus a

301 302 J Baril et al.

non-nucleoside reverse transcriptase inhibitor (NNRTI), an integrated strand transfer inhibitor (INSTI) or a ritonavir Methods (RTV)-boosted protease inhibitor (PI) [1]. RTV is an PubMed, EMBASE, CENTRAL, Cochrane Reviews and DARE antiretroviral drug from the PI class that is co-administered (up to August 2012) were searched. Publically available at low doses with other agents (including other PIs) for proceedings from the International AIDS Society (IAS), its ‘boosting’ effect through inhibition of the cytochrome Interscience Conference on Antimicrobial Agents and P450 34A (CYP34A) enzyme, thus increasing drug levels Chemotherapy (ICAAC) and the Conference on Retroviruses and allowing for less frequent administration of medica- and Opportunistic Infections (CROI) were also searched tions [1,2]. (January 2009 to August 2012). For PubMed, the following Some patients, however, experience high levels of adverse terms were identified by two reviewers: (‘atazanavir’ AND events (e.g. hyperbilirubinaemia, unfavourable lipid profiles (‘ritonavir OR boost* OR unboosted’)), with the asterisk and increased gastrointestinal toxicity) on a boosted representing multiple appendices to ‘boost’. No language or PI-based triple combination regimen [3,4]. Studies have publication date restrictions were imposed. Reference lists of suggested that cardiovascular disease is associated with all records included for full text review were screened. antiretroviral therapy, with recent data demonstrating an Studies were considered eligible if they were randomized increased risk of myocardial infarction associated with a controlled trials evaluating the efficacy and safety of longer duration of PI use. This could be explained, in part, by unboosted ATV (400 mg per day) compared with RTV- alterations in lipid profiles associated with therapy [5–7]. In boosted ATV (300 mg ATV and 100 mg RTV per day) or particular, RTV has been shown to significantly increase another RTV-boosted PI in HIV-1-infected adults. Pati- plasma lipid levels (e.g. total cholesterol) [8–10], although ents were treatment experienced (i.e., during an induction recent data do not show any difference in myocardial phase, they had received a regimen including an RTV- infarction between RTV-boosted and unboosted PI regimens boosted PI and had achieved and maintained virological [11]. Options to address this include switching to an NNRTI- suppression). Records were independently assessed for eli- or INSTI-based regimen, or removing the RTV boosting. gibility by two blinded reviewers. Disagreements between Atazanavir (ATV) (ReyatazTM, Bristol-Myers Squibb, reviewers regarding study inclusion were resolved through Princeton, NJ, USA) was the first once-daily PI approved for consensus and, when necessary, through third reviewer the treatment of HIV infection and is the only PI that can be consultation. The search strategy and eligibility results are currently given without RTV boosting with a relatively low presented in Fig. 1. daily dose and pill burden [12–15]. ATV has been associated For included studies, details were extracted using a com- with a more favourable lipid profile when compared with prehensive data extraction form [17]. One reviewer con- other PI-based therapies [10]. Recent trials have investigated ducted the data extraction, which was cross-checked by a whether patients can effectively continue on unboosted ATV second reviewer. Regarding data missing from the publi- regimens after an induction phase that included RTV. Results cation, three of the five investigators contacted provided have suggested that such a regimen is noninferior to the unpublished data requested. and potentially safer than an RTV-boosted PI regimen for The primary efficacy outcome was the maintenance of patients with a suppressed HIV plasma viral load at the time virological suppression and was defined as the proportion of switching [3,14]. In contrast, when HIV plasma viral load of patients maintaining HIV-1 RNA levels below specified is detectable, as in treatment-naïve patients or in those thresholds (i.e. <50 and < 400 HIV-1 RNA copies/mL) failing treatment, unboosted ATV is not recommended during the study maintenance phase (i.e. after the induc- because its plasma drug concentrations are variable and tion phase). Results were also extracted for change in CD4 may fall below the minimum threshold for maximal activity cell counts. Regarding safety, data were extracted on [15]. In fact, Malan et al. showed that RTV-boosted ATV change in values from the beginning to the end of the was associated with a lower virological failure rate than maintenance phase for mean lipid levels [i.e. total choles- unboosted ATV in treatment-naïve patients [16]. terol, triglycerides, low-density lipoprotein (LDL) and high- To evaluate the clinical evidence, a systematic review density lipoprotein (HDL)], renal function parameters (e.g. and meta-analysis were conducted of randomized trials creatinine) and the occurrence of hyperbilirubinaemia (i.e. that evaluated the efficacy and safety of switching from an grades 2−4), jaundice and scleral icterus. RTV-boosted PI to unboosted ATV compared with continu- Risk of bias at the study and outcome level was assessed ing on an RTV-boosted PI regimen in adult HIV-1-positive using Cochrane guidelines [17]. The Cochrane tool consists patients after patients showed established virological of five items (i.e. sequence generation, allocation conceal- suppression through an induction phase of PI/RTV-based ment, blinding, incomplete outcome data and selective highly active antiretroviral therapy (HAART). outcome reporting) [17]. Final decisions depended on the

Fig. 1 PRISMA flow diagram for the systematic review. combination of these factors and the individual character- measures for the other included study [19]. Finally, for one istics of each study. Risk of bias was assessed by two study, no variance measures were provided at all and independent reviewers. In addition, expert opinion was therefore the average variance from the other studies was further used to determine whether important clinical het- used to approximate the missing SD [17,18]. One study erogeneity existed between studies. reported the proportion of patients who experienced viro- The meta-analysis was performed according to a pre- logical rebound instead of virological control. This study specified protocol. Cochrane Collaboration software also reported the total number of patients experiencing (Review Manager, Version 5.0, The Nordic Cochrane Centre, treatment failure which included both virological rebound The Cochrane Collaboration, Copenhagen, Denmark, 2008) and treatment discontinuation [20]. For the meta-analysis, it was utilized for synthesis of the data for the meta-analysis. was assumed that the proportion of patients who did not Dichotomous outcomes, such as the proportion maintaining experience treatment failure must have experienced treat- virological suppression, were reported as risk ratios (RRs). ment success and this was subsequently used as the outcome Continuous outcomes (e.g. total cholesterol) were reported measure for virological control. as mean differences (MDs) between study groups from the The primary analysis compared any RTV-boosted PI beginning to the end of the maintenance phase. Outcomes (including ATV) vs. unboosted ATV. A secondary analysis were assessed over a 48-week period, with the exception of included only studies comparing RTV-boosted ATV with one study at 24 weeks [18]. A random effects model was used unboosted ATV. Sensitivity analyses were conducted for to conduct the meta-analyses with aggregate data. A χ2 test the primary outcome of virological suppression (i.e. < 50 and the I 2 measure were used to assess heterogeneity of copies/mL) while studies at higher risk of bias were included studies. excluded. For liver function, the analysis was only com- Data imputation was conducted according to principles pleted for studies that involved RTV-boosted ATV and outlined by Cochrane [17]. For lipid outcomes, not all studies had complete data [3,14]. For renal function, because of reported all variance measures [e.g. standard deviation (SD) inconsistency across definitions of outcome measures and or standard error (SE)] for baseline (i.e. end of induction/ restricted data, only qualitative results are reported. beginning of maintenance), final (i.e. end of maintenance) and change values. Therefore, a correlation coefficient Results (CORR) was calculated based on the complete available variance data for one included study [14,17]. This CORR Of the 1600 unique records identified, 1584 records were value was used to impute the remaining missing variance excluded (Fig. 1). The full-text publications of 16 studies

were retrieved for detailed review, and 11 studies were For lipid parameters, results showed that unboosted ATV subsequently excluded. Five studies involving a total of was associated with significantly improved or comparable 1249 patients met the inclusion criteria [3,14,18–20]. outcomes vs. an RTV-boosted PI regimen. Specifically, a All studies evaluated antiretroviral treatment-experienced significant reduction in total cholesterol (MD = −14.7 mg/ patients and measured the primary efficacy endpoint at dL; 95% CI −20.96 to −8.49; P < 0.00001), triglycerides the end of a maintenance phase. A summary of the study (MD = −51.15 mg/dL; 95% CI −78.36 to −23.94; P = 0.0002) characteristics is presented in Table 1. (Fig. 3) and LDL cholesterol (MD = −5.56 mg/dL; 95% Of the five studies, two studies compared regimens CI −9.71 to −1.41; P = 0.009) (Fig. 3) was observed with including a PI/RTV combination (i.e. lopinavir/RTV [19], or unboosted ATV compared with PI/RTV. There were no lopinavir/RTV, indinavir/RTV or saquinavir/RTV [20]) vs. significant differences in HDL cholesterol (Fig. 3). These unboosted ATV, and three studies compared regimens con- results were supported in a subanalysis of only studies taining ATV/RTV vs. unboosted ATV [3,14,18]. In one study including ATV/RTV vs. unboosted ATV (data not shown). [20], the ATV arm was composed mainly of patients receiving The risk of grade 2−4 hyperbilirubinaemia was found an unboosted ATV regimen; however, a small proportion to be significantly lower with unboosted ATV compared of patients received ATV/RTV (9%) if tenofovir was co- with ATV/RTV (RR = 0.43; 95% CI 0.21 to 0.89; P = 0.02; administered. For this analysis, the proportion of patients on I2 = 0%) (Fig. 4). There were inadequate data to inform ATV/RTV was assumed to be small enough to be considered the meta-analysis for scleral icterus or jaundice. Renal not clinically meaningful and therefore the ATV arm was function parameters could not be pooled because of incon- analysed only as unboosted ATV. Overall, patient characteris- sistent measures across studies; however, the results of tics were comparable across studies, with some variations individual trials are noteworthy. In particular, Ghosn et al. in NRTI backbones. Regarding the unboosted ATV groups, [3] reported 0% grade 3 to 4 renal function abnormalities it is noted that a lamivudine and abacavir NRTI backbone (i.e. creatinine, blood urea nitrogen (BUN) and serum uric regimen was commonly used. In contrast, a regimen contain- acid) for both treatment groups. ing tenofovir was used much less frequently (Table 1). Results of a sensitivity analysis regarding study quality The results of the individual study quality assessments for the primary efficacy outcome (HIV RNA < 50 copies/mL) are summarized in Table 2. Study protocols were obtained, showed that findings did not change when each of the where available, to assess selective outcomes reporting. two studies deemed to have a potentially higher risk of bias All studies described statistical analyses of outcomes and was excluded individually [18,19]. However, when both addressed any incomplete data such as loss to follow-up. studies were excluded together, leaving only three studies, Studies were open label; however, outcomes evaluated were the results reached statistical significance (i.e. RR = 1.07; objective measures and it was assumed that nonblinding 95% CI 1.0 to 1.15; P = 0.05), indicating greater virological represented a low risk of bias. Two studies did not suffi- control in the PI/RTV arm. ciently report methods of sequence generation and allocation concealment, and supplementary data from the authors Discussion were not available [18,19]. Only one study was presented at a conference [18] and lacked clarity on a majority of Although RTV-boosted PI regimens have long-term efficacy categories. Overall, the risk of bias within the included and a high barrier to resistance, an association has been studies was assessed as adequate for three studies and found with metabolic abnormalities, including dyslipi- acceptable for inclusion in our analysis for the other two. daemia, related to RTV. As RTV is a potent inhibitor of Results from the meta-analysis demonstrated no statisti- CYP3A4 enzymes, patients are also more susceptible to cally significant difference in virological efficacy (i.e. HIV drug−drug interactions [15]. These concerns may be of RNA < 50 copies/mL) between PI/RTV and unboosted ATV particular importance in higher risk populations with a regimens (RR = 1.04; 95% CI 0.99 to 1.10), with no hetero- background cardiovascular risk, including those on certain geneity observed (I2 = 0%) (Fig. 2). Findings for virological non-ATV/RTV regimens, and elderly patients prone to drug- efficacy were similar in a subanalysis of studies where ATV related toxicities [21]. Therefore, given this unfavourable was the only PI with RTV during induction (RR = 1.04; 95% safety profile, a safer regimen that can maintain efficacy CI 0.98 to 1.10; I2 = 0%). Results for the proportion of may be preferred. patients achieving HIV RNA < 400 copies/mL also indicated The findings of the current report, being the most com- no statistically significant difference (RR = 1.05; 95% CI prehensive meta-analysis to date, support a switch from 0.99 to 1.11; I2 = 0%). The meta-analysis also showed no an RTV-boosted PI (e.g. lopinavir, indinavir or ATV) to significant differences between the two groups for change in an unboosted ATV regimen in patients with established CD4 counts (MD = 14.10; 95% CI −13.27 to 41.48; I2 = 53%). virological suppression after an induction period. Over 48

Table 1 Study and baseline characteristics.

HIV-1 RNA Median load at baseline Criteria to proceed

Reference Treatment age Male (log10 copies/mL) NRTI regimen Length of to maintenance; Primary efﬁcacy Secondary endpoints (study name) (n for ITT sample) (years) (%) (median) backbone induction length of maintenance endpoint deﬁnition§ analysed*

Gatell et al. 2007 I: PI/RTV (419) 40 84 1.69 3TC (88%), ABC (8%), d4T (31%), ≥3 months HIV RNA < 50 copies/mL Virological rebound (HIV • HIV RNA < 400 copies/mL (SWAN) [20] ddC (< 1%), ddI (15%), TDF for ≥ 3 months; 48 RNA ≥ 50 copies/mL) (9%), ZDV (47%) weeks M: PI/RTV (76) vs. 41 79 1.69 3TC (87%), ABC (6%), d4T (27%), ATV* (150)† ddC (1%), ddI (14%), TDF (8%), ZDV (54%) Ghosn et al. 2010 I: ATV/RTV (252) 35 72 4.86 3TC+ABC (53%), 3TC+d4T (1%), 26–30 weeks HIV RNA <50 copies/mL; Virological control (HIV • HIV RNA < 400 copies/mL (InduMa) [3] 3TC+ddI (7%), 3TC+ZDV (31%), 48 weeks RNA < 50 copies/mL) • CD4 count change ddI+FTC (7%), FTC+TDF (1%) • Lipids (TC, LDL, HDL, TG) M: ATV/RTV (85) vs. 35 75 4.86 3TC+ABC (51%), 3TC+d4T (1%), • Jaundice ATV (87) 3TC+ddI (8%), 3TC+ZDV (33%), • Scleral icterus ddI+FTC (7%), FTC+TDF (0) • Hyperbilirubinaemia Soriano et al. 2008 I: LPV/RTV (189) 40 51 Not reported 3TC, ABC, d4T, ddI, ZDV,TDF > 24 weeks HIV RNA <50 copies/mL; Virological control (HIV • HIV RNA < 400 copies/mL (SLOAT) [19] M: LPV/RTV (87) vs. 41 40 Not reported 3TC, ABC, d4T, ddI, ZDV; no 48 weeks RNA < 50 copies/mL) • Lipids (TC, LDL, HDL, TG) ATV (49) patients with TDF backbone were included for analysis in this arm Squires et al. 2010 I: ATV/RTV (515) 39 85 5.06 3TC+ABC 36 weeks HIV RNA <50 copies/mL Virological control (HIV • HIV RNA < 400 copies/mL (ARIES) [14] M: ATV/RTV (209) vs. 37 84 5.05 3TC+ABC and no virological RNA < 50 copies/mL; • CD4 count change ATV (210) failure; 48 weeks < 400 copies/mL) • Lipids (TC, LDL, HDL, TG) • Jaundice • Scleral icterus • Hyperbilirubinaemia Wohl et al.2012‡ I: ATV/RTV (data not 42 81 1.59 FTC+TDF ≥ 6 months Data not available; 24 Virological control (HIV • CD4 count change (ASSURE) [18] available) weeks RNA < 50 copies/mL) • Lipids (TC, LDL, HDL, TG) atazanavir unboosted of Meta-analysis M: ATV/RTV (97) vs. 44 78 1.59 3TC+ABC ATV (199) I Medicine HIV ITT, intention to treat; PI, protease inhibitor; RTV, ritonavir; NRTI, nucleoside reverse transcriptase inhibitor; LPV, lopinavir; TDF, tenofovir; ABC, abacavir; ZDV, zidovudine; 3TC, lamivudine; FTC, emtricitabine, TC, total cholesterol; LDL, low-density lipoprotein; HDL, high-density lipoprotein; TG, triglycerides; ATV, atazanavir; d4T, stavudine; ddC, zalcitabine; ddI, didanosine; I, induction phase; M, maintenance phase. *Results data were available for both the PI/RTV (or ATV/RTV) group and the ATV group separately (i.e. nonaggregate data) for input into the meta-analysis. †It was assumed that the ATV group could be considered an unboosted ATV group because the ATV/RTV composition was only 9% and considered not to make a large clinical impact. ‡Data from conference proceedings. § (2014), The following gives further details of the assessment of the primary outcome. Gatell et al. 2007: proportion of randomized patients who experienced virological rebound (i.e. patients with confirmed on-study HIV RNA loads ≥ 50 copies/mL or last on-study HIV RNA load ≥ 50 copies/mL followed by study discontinuation); Ghosn et al. 2010: proportion of patients with HIV RNA < 50 copies/mL sustained through week 48 (ITT analysis; noncompletion = failure). Soriano et al. 2008: all efficacy analyses were on an ITT basis; treatment failure defined on the basis of virological failure (HIV RNA > 50 copies/mL) or discontinuation

15 because of side effects. Squires et al. 2010: the primary efﬁcacy population was the intent-to-treat exposed population (i.e. at least one dose of study medication); other analyses for patients achieving HIV RNA levels below the lower limits of detection (< 50 and < 400 copies/mL) included observed and missing/discontinuation equal to failure (MD = F). Wohl et al. 2012: per protocol analysis (< 50 copies/mL). ,301 –310 305 306 J Baril et al.

Table 2 Qualitative risk of bias assessment summary

Sequence Allocation Incomplete outcome Free of selective Free of Trial generation concealment Blinding* data addressed reporting other bias

Ghosn et al. 2010 [3] Yes Yes Yes Yes Yes Yes Gatell et al. 2007 [20] Yes Yes Yes Yes Yes Yes Wohl et al. 2012 [18] Unclear Unclear Yes Yes Unclear Unclear Squires et al. 2010 [14] Yes Yes Yes Yes Yes Unclear Soriano et al. 2008 [19] Unclear No Yes Yes Yes Unclear

‘Yes’ indicates a low risk of bias. *Studies were open label; however, given objective outcome measures, it was determined that a low risk of bias was present and therefore ‘Yes’ is entered in the table.

Fig. 2 Forest plot of meta-analysis results comparing the proportions of patients achieving virological control (HIV RNA < 50 copies/mL) with a protease inhibitor/ritonavir (PI/RTV) regimen vs. an unboosted atazanavir (ATV) regimen. CI, conﬁdence interval; MH, Mantel Haenszel.

weeks (with one study reporting data to 24 weeks), viro- evidence of an association between cumulative exposure to logical efficacy was not significantly different between RTV-boosted ATV and myocardial infarction risk [11]. PI/RTV and unboosted ATV. Results remained consistent Several of these studies demonstrated favourable long- when other efficacy endpoints were evaluated and analyses term efficacy of unboosted ATV with no evidence of were restricted to studies that only used ATV/RTV for increased virological breakthrough [24–26]. For example, induction. However, in sensitivity analyses, when two in the largest study to explore the long-term efficacy studies deemed to potentially have a higher risk of bias of unboosted ATV-based regimens (the NEAT Unboosted were excluded (i.e. leaving three studies in the meta- Atazanavir Study Group), 886 patients with sustained viro- analysis), the results approached significance, indicating logical suppression were evaluated. The probability of viro- greater virological control in the PI/RTV arm. Furthermore, logical failure (HIV RNA > 50 copies/mL) in this study over the meta-analysis demonstrated significant improvements 1, 2 and 3 years was, respectively, 9.7, 16.1 and 20.1% [25]. in total cholesterol, triglycerides, LDL cholesterol, and Three-year virological failure rates were reduced to 9.3% hyperbilirubinaemia following a switch from an RTV- when the threshold was > 400 copies/mL. Caution must boosted PI to unboosted ATV. A recent meta-analysis be exercised, however, when comparing the results of the by Carey et al. provided preliminary evidence that RTV NEAT study with those of the current meta-analysis, given co-administration resulted in higher total cholesterol and the differences in study design (i.e. RCTs vs. a retrospective non-HDL cholesterol compared with unboosted ATV [10]. observational study) and time period. The results of our meta-analysis confirm the findings of Resistance testing and reporting were heterogeneous this published study, adding more recent and complete data across the included studies, and therefore our meta- on efficacy and safety. Data from several observational analysis did not evaluate mutation rates. The general cohort studies also support our findings, often showing conclusions from the studies suggested that there were either a significantly or non-significantly improved lipid no apparent differences in the rate of drug resistance profile with an unboosted ATV regimen compared with an mutations observed for the patients receiving ATV only RTV-boosted PI [15,22–24]. However, recent data show no compared with those patients receiving a PI in combination

Fig. 3 Forest plots of meta-analysis results comparing changes in the total lipid profile. (a) Total cholesterol; (b) triglycerides; (c) low-density lipoprotein (LDL) cholesterol; (d) high-density lipoprotein (HDL) cholesterol. ATV, atazanavir; CI, confidence interval; PI/RTV, protease inhibitor/ ritonavir; SD, standard deviation; IV, inverse variance. with RTV. However, given the limited data available (e.g. The goals of simplification strategies in antiretroviral missing baseline data) and other confounding variables therapy are to reduce pill burden, minimize side effects, (e.g. medication adherence), these conclusions should be and improve treatment adherence to prevent virological interpreted with caution. failure and preserve future drug options. In this context,

Fig. 4 Forest plot of meta-analysis results comparing the risk of hyperbilirubinaemia events (defined as grade 2 − 4) for an atazanavir/ritonavir (ATV/RTV) vs. an unboosted ATV regimen at the end of maintenance. Only two studies reported on hyperbilirubinaemia and were available for inclusion. Baseline values (i.e. at the beginning of maintenance) were not statistically significantly different between treatment groups of the two included studies and were therefore not used in the summary measure for the meta-analysis. CI, confidence interval; MH, Mantel Haenszel.

reducing toxicity appears to be most important, such that formally assessed because of inconsistencies in endpoints. intolerance to RTV accounts for close to half of patient Fourthly, in the SWAN study, a small proportion of those switches to unboosted ATV [22]. Our results support use on ‘unboosted ATV’ (9%) remained on RTV because their of an unboosted ATV regimen in that setting, and also regimens had a tenofovir-based NRTI backbone, with for patients who may not be able to take or have access to these patients considered as receiving ‘ATV only’, given RTV, as well as those at high risk of deleterious events that the absence of disaggregated data. However, sensitivity may be associated with RTV use. Discontinuation of RTV analyses remained robust when this study was excluded. from the regimen may also remove the requirement for Finally, this study focused on trials of 48-week duration, refrigeration of certain formulations. However, therapy which may not be predictive of the long-term efficacy with unboosted ATV may not be appropriate for patients of unboosted ATV regimens. On an individual study level, with a history of suboptimal response to previous anti- the ARIES 144-week extension study did demonstrate retroviral treatment, or for the majority of treatment-naïve sustained and similar efficacy of unboosted and boosted patients [15,27]. In situations where lower adherence may ATV regimens in patients who completed the extended be expected, or where concomitant medication use may follow-up [30]. At week 144, 77% vs. 73% of subjects in reduce its bioavailability, unboosted ATV should only be the unboosted and boosted ATV groups, respectively, main- prescribed with caution. tained HIV RNA <50 copies/mL. This is reassuring, but The choice of an NRTI backbone can be an important additional studies involving longer follow-up are required consideration in switching from boosted to unboosted ATV. to fully elucidate the long-term efficacy of unboosted ATV. The NEAT study showed that use of abacavir was associated with lower risk of virological failure vs. alternative NRTIs [25]. This finding supports the conclusion that abacavir, Conclusions usually with lamivudine as a fixed-dose combination There is increasing clinical experience with unboosted backbone, is the ideal companion drug for unboosted ATV ATV-based regimens in patients having achieved and [15]. In contrast, the co-administration of tenofovir with maintained virological suppression on other combinations, unboosted ATV may be problematic as a consequence of particularly those including RTV-boosted PIs. Although altered and less predictable ATV pharmacokinetic param- longer term follow-up in larger numbers of subjects should eters [28,29]. be accumulated, the results of this meta-analysis suggest The current study has several limitations. First, not all that this approach is evidence-based. It can be a legitimate data were available to inform variance inputs for the meta- option in patients who wish to change their medications to analysis. To address this, we utilized imputation techniques address issues of regimen complexity or safety. recommended by Cochrane [17] whereby assumptions were minimized through having complete variance data for at least one study. Secondly, for some outcome measures (e.g. Funding hyperbilirubinaemia), not all five studies had data available to be statistically combined, and so these results may not This work was supported by Bristol-Myers Squibb, who be as rigorous as for analyses with a larger sample size. provided funding to conduct the analysis and prepare Thirdly, renal function safety parameters could not be the manuscript. Data analysis and interpretation were

solely conducted by the authors, with Bristol-Myers Squibb 12 Bristol-Myers Squibb Canada. Product Monograph: remaining independent of this process and only providing ReyatazTM (Atazanavir sulfate) Azapeptide Inhibitor of HIV-1 minor editorial comments. Protease. Quebec, Canada. Prepared December 3, 2003 (Revised May 11, 2012). Control Number: 154449. Acknowledgements 13 Bristol-Myers Squibb. Highlights of the prescribing information: ReyatazTM (Atazanavir sulfate). New Jersey, The authors would like to acknowledge Viiv Healthcare for USA. Approved 2003. Revised March 2012. 1246226B2. providing additional study data for the ARIES trial. 687US11PBS05504. 14 Squires KE, Young B, Dejesus E et al. Similar efficacy and References tolerability of Atazanavir compared with Atazanavir/ 1 Thompson MA, Aberg JA, Hoy JF et al. Antiretroviral ritonavir, each with abacavir/lamivudine after initial treatment of adult HIV infection. 2012 recommendations on suppression with abacavir/lamivudine plus ritonavir-boosted the international antiviral society – USA Panel. JAMA 2012; Atazanavir in HIV-infected patients. AIDS 2010; 24: 308: 387–402. 2019–2027. 2 Panel on Antiretroviral Guidelines for Adults and 15 Focà E, Ripamonti D, Motta D, Torti C. Unboosted Adolescents. Guidelines for the use of antiretroviral agents Atazanavir for treatment of HIV infection: rationale in HIV-1-infected adults and adolescents. Department of and recommendations for use. Drugs 2012; 72: Health and Human Services. Available at: http://aidsinfo.nih 1161–1173. .gov/ContentFiles/AdultandAdolescentGL.pdf (accessed 15 16 Malan DR, Krantz E, David N et al. 96-week efficacy and May 2013). safety of atazanavir, with and without ritonavir, in a HAART 3 Ghosn J, Carosi G, Moreno S et al. Unboosted Atazanavir- regimen in treatment-naive patients. J Int Assoc Physicians based therapy maintains control of HIV type-1 replication as AIDS Care (Chic) 2010; 9: 34–42. effectively as a ritonavir-boosted regimen. Antivir Ther 17 Higgins JPT, Green S, eds. Cochrane handbook for 2010; 15: 993–1002. systematic reviews of interventions version 5.1.0. The 4 Cooper CL, van Heeswijk RP, Gallicano K, Cameron DW. Cochrane Collaboration. 2011. Available at: http://www A review of low-dose ritonavir in protease inhibitor .cochrane-handbook.org (accessed 20 August 2012). combination therapy. Clin Infect Dis 2003; 36: 1585–1592. 18 Wohl D, Bhatti L, Small CB et al. Simplification to 5 Iloeje UH, Yuan Y, L’Italien G et al. Protease inhibitor Abacavir/Lamivudine + Atazanavir From Tenofovir/ exposure and increased risk of cardiovascular disease in Emtricitabine + Atazanavir/Ritonavir Maintains Viral HIV-1 infected patients. HIV Med 2005; 6: 37–44. Suppression and Improves Bone Biomarkers. Oral Abstract 6 Friis-Moller N, Sabin CA, Weber R et al. Combination H-556c. 52nd ICAAC. San Francisco, CA. September 9-12, antiretroviral therapy and the risk of myocardial infarction. 2012. N Engl J Med 2003; 349: 1993–2003. 19 Soriano V, García-Gasco P, Vispo E et al. Efficacy and safety 7 DAD Study Group, Friis-Moller N, Reiss P et al. Class of of replacing lopinavir with Atazanavir in HIV-infected antiretroviral drugs and the risk of myocardial infarction. patients with undetectable plasma viraemia: final results N Engl J Med 2007; 356: 1723–1735. of the SLOAT trial. J Antimicrob Chemother 2008; 61: 8 Calza L, Manfredi R, Farneti B, Chiodo F. Incidence of 200–205. hyperlipidemia in a cohort of 212 HIV-infected patients 20 Gatell J, Salmon-Ceron D, Lazzarin A et al. Efficacy and receiving a protease inhibitor-based antiretroviral therapy. safety of Atazanavir-based highly activeantiretroviral Int J Antimicrob Agents 2003; 22: 54–59. therapy in patients with virologic suppression switched from 9 Shafran SD, Mashinter LD, Roberts SE. The effect of a stable, boosted or unboosted protease inhibitor treatment low-dose ritonavir monotherapy on fasting serum lipid regimen: the SWAN study (AI424-097) 48-week results. Clin concentrations. HIV Med 2005; 6: 421–425. Infect Dis 2007; 44: 1484–1492. 10 Carey D, Amin J, Boyd M, Petoumenos K, Emery S. Lipid 21 D’Arminio Monforte A, Reiss P, Ryom L et al. Atazanavir profiles in HIV-infected adults receiving Atazanavir and (ATV)-Containing Antiretroviral Treatment is not Associated Atazanavir/ritonavir: systematic review and meta-analysis of with an Increased Risk of Cardio- or Cerebro-Vascular randomized controlled trials. J Antimicrob Chemother 2010; Events (CVE) in the D:A:D Study. [Poster No. 823]. CROI 65: 1878–1888. 2012. 11 Monforte AD, Reiss P, Ryom L et al. Atazanavir is not 22 Giuntini R, Martinelli C, Ricci E et al. Efficacy and safety of associated with an increased risk of cardio- or boosted and unboosted atazanavir-containing antiretroviral cerebrovascular disease events. AIDS 2013; 27: regimens in real life: results from a multicentre cohort 407–415. study. HIV Med 2010; 11: 40–45.

23 Santos JR, Moltó J, Llibre JM et al. Unboosted atazanavir 27 Cohen C, Nieto-Cisneros L, Zala C et al. Comparison of plus co-formulated lamivudine/abacavir as a ritonavir- atazanavir with lopinavir/ritonavir in patients with prior sparing simplification strategy in routine clinical practice. protease inhibitor failure: a randomized multinational trial. HIV Clin Trials 2009; 10: 129–134. Curr Med Res Opin 2005; 21: 1683–1692. 24 Hocqueloux L, Le Moal G, De la Tribonnier X et al. 28 Atazanavir: new recommendations if combined with Maintenance therapy with unboosted atazanavir is tenofovir (Viread) – and warning on Viagra, Cialis, and comparable to boosted strategy in experienced patients, even Levitra. AIDS Treat News 2004; 399:5. with tenofovir: a multicentre propensity-adjusted cohort. 6th 29 Fournier C, Higgins N, Thomas R et al. Negligible effect IAS conference on HIV pathogenesis, treatment and of tenofovir on atazanavir trough concentrations and prevention. Rome, July 2011 [Abstract no. CDB359]. genotypic inhibitory quotients in the presence and absence 25 Pavie J, Porcher R, Torti C et al. Efficacy and safety of of ritonavir. Ther Drug Monit 2013; 35: 264–269. a switch to unboosted atazanavir in combination with 30 Squires K, Young B, DeJesus E et al. Similar efficacy and nucleoside analogues in HIV-1-infected patients with tolerability of atazanavir (ATV) compared to ATV/ritonavir virological suppression under antiretroviral therapy. (RTV), each in combination with abacavir/lamivudine J Antimicrob Chemother 2011; 66: 2372–2378. (ABC/3TC) after initial suppression with ABC/3TC + 26 Wentzell J, Kravcik S, Giguere P. Efficacy of boosted and ATV/RTV in HIV-1 infected patients: Final (144 weeks) unboosted atazanavir-based therapy in patients without prior results of the op. 6th IAS Conference on HIV Pathogenesis, antiretroviral treatment failure [abstract P079]. Can J Infect Treatment and Prevention. Rome, July 2011 [Abstract no. Dis Med Microbiol 2012; 23 (Suppl A): 66A. MOPE215].