Lopinavir/Ritonavir Or Efavirenz Plus Two

Antiviral Therapy 11:609–618 Lopinavir/ritonavir or efavirenz plus two nucleoside analogues as first-line antiretroviral therapy: a non-randomized comparison Andrea De Luca 1*, Alessandro Cozzi-Lepri 2, Andrea Antinori 3, Mauro Zaccarelli 3, Marco Bongiovanni 4, Simona Di Giambenedetto1, Patrizia Marconi 3, Paola Cicconi 4, Francesco Resta5, Benvenuto Grisorio6, Maria Ciardi 7, Roberto Cauda1 and Antonella d’Arminio Monforte4 on behalf of the I.Co.N.A, UCSC, IMIT- MI and INMI cohorts†

1Institute of Clinical Infectious Diseases, Catholic University of the Sacred Heart, Rome, Italy 2Royal Free and University College Medical School, UCL, London, UK 3National Institute for Infectious Diseases ‘Lazzaro Spallanzani’, Rome, Italy 4Institute of Infectious and Tropical Diseases, University of Milan, Milan, Italy 5SS Annunziata Hospital, Taranto, Italy 6Ospedali Riuniti, Foggia, Italy 7University La Sapienza, Rome, Italy †See Additional file for a complete list of the members of the cohort studies and their affiliations, available on the web

*Corresponding author: Tel: +39 06 3015 4945; Fax: +39 06 3054 519; E-mail: [email protected]

Background: Although efavirenz (EFV) and lopinavir/ difference in the adjusted hazards of virological failure ritonavir (LPV/r) are both recommended antiretroviral (LPV/r versus EFV relative hazard [RH] 1.16, 95% confi- agents for combination therapy in drug-naive HIV-infected dence intervals [CI]: 0.58–2.32, P=0.67), CD4 recovery patients, no randomized comparison of their efficacy and (RH=0.93, 95% CI: 0.66–1.30, P=0.66), clinical progres- tolerability is available yet. A multi-cohort prospective sion (RH=1.64, 95% CI: 0.70–3.84, P=0.25), drug observational comparative study was performed. discontinuation for toxicity (RH=0.92, 95% CI: 0.51-1.64, Methods: Efficacy was examined comparing time to viro- P=0.76) and for any reason, and rates of liver enzyme and logical failure, CD4 recovery and clinical progression. total/low density lipoprotein (LDL) cholesterol elevation. Tolerability was examined comparing time to treatment In contrast, the rate of triglycerides elevations (>1 NCEP discontinuation for any reason and for toxicity and time Adult Treatment Panel III category increase) was higher in to liver enzymes or lipid alterations. Survival analysis was the LPV/r group (RH=1.69, 95% CI: 1.14–2.50; P=0.01). conducted by an intent-to-treat principle using the Models weighted for the inverse of conditional probability Kaplan–Meier method, and standard and weighted Cox of receiving either drug applied to the efficacy endpoints regression models. yielded similar results. CD4 recovery with both drugs was Results: A total of 674 antiretroviral-naive patients also similar in the lowest CD4 strata. starting a two nucleoside reverse transcriptase inhibitor Conclusions: Our analysis suggests similar efficacy and regimen plus either EFV (n=481) or LPV/r (n=193) were tolerability for EFV- or LPV/r-based first-line antiretro- examined. At baseline, patients starting LPV/r had higher viral regimens. LPV/r was associated with higher rates of HIV RNA and lower CD4+ T-cell counts. There was no hypertriglyceridaemia.

Introduction

The primary goal of antiretroviral treatment is to reduce inhibitor (PI) [3] or a non-nucleoside reverse transcrip- AIDS-related morbidity and mortality by maximally and tase inhibitor (NNRTI) [4–6]. For antiretroviral-naive durably inhibiting HIV replication [1,2]. The most effec- patients, treatment guidelines recommend tive regimens in treatment-naive patients are those based lopinavir/ritonavir (LPV/r) or efavirenz (EFV) as the on combinations of two nucleoside reverse transcriptase preferred drugs for PI-based or NNRTI-based highly inhibitors (NRTIs) and either a boosted protease active antitiretroviral therapy (HAART) regimens,

respectively [7–9]. Nonetheless, results from randomized Branchburg, NJ, USA), or signal amplification b-DNA trials comparing the efficacy and tolerability of EFV and assay (Versant 3.0; Bayer Health Care Diagnostics LPV/r in treatment naive patients are not yet available. Division, Berkeley, CA, USA). The lower limit of detec- The aim of this study was to compare the long-term tion of these assays is 50 copies/ml. For values below efficacy and tolerability of potent antiretroviral combi- the limit of detection we imputed the value of 50 nations based on two NRTIs plus either EFV or LPV/r copies/ml. Peripheral blood CD4+ T-cell counts were in previously treatment-naive subjects enrolled in four measured using standard flow cytometry techniques. Italian observational cohorts. Statistical analysis Methods The characteristics of patients who started EFV or LPV/r were compared using the two-sample Wilcoxon Cohorts rank sum test for continuous measurements and the χ2 Patients were enrolled from four Italian prospective test for categorical variables. cohort studies. The ‘Italian Cohort of Naive for Patients were compared with respect to several Antiretrovirals’ (I.Co.N.A) is a multicentre observa- outcomes using ‘time-to-event’-type analyses. Dates of tional study started in 1997, enrolling HIV-1-positive the first of two consecutive viral load measurements individuals who, for whatever reason, were naive for >400 copies/ml after 6 months of therapy and of first antiretroviral drugs [10]. The other three cohorts are CD4+ T-cell count >200 cells/µl above pre-therapy from single clinical sites that enrolled HIV-positive count were used to define virological failure and individuals who were treated in major tertiary care immune reconstitution events, while the dates of new or infectious diseases clinics in Rome (Catholic University, recurring AIDS-defining conditions [11] or death for UCSC and National Institute for Infectious Diseases, any cause were used to analyse clinical progression INMI) and Milan (University of Milan, IMIT-MI). In events. For clinical or laboratory toxicity events defini- all four cohorts demographic, clinical and laboratory tions were: discontinuation of EFV and LPV/r because data and information on treatments used are collected of any toxicity; reaching grade III ALT elevation (that is, and prospectively updated for all participants, and >204 IU/ml in those with pre-therapy levels ≤40 IU/ml recorded in separate databases. Updates take place at or >3.5×pre-therapy levels in those with pre-therapy the occurrence of any clinical event and, in their level >40 IU/ml); and experiencing ≥1 step up in the absence, at least every 6 months. Reasons for discon- NCEP Adult Treatment Panel III category definitions tinuing therapy, according to treating clinicians, are for total cholesterol, LDL cholesterol and triglycerides also noted. Data were standardized, merged and [12]. For patients who did not experience such events, analysed centrally. follow-up was right-truncated at last visit. All analyses were conducted using an intent-to-treat approach by Study population and methods ignoring treatment switches during the follow-up. To be included in the present study, patients belonging The Kaplan–Meier method was used to compare to the four cohorts had to meet the following criteria: (i) time-to-event data in patients starting EFV or LPV/r in having started a potent antiretroviral regimen based on univariable analyses. Adjustment for confounders was the combination of two NRTIs and either LPV/r or EFV, conducted using two methods. The main analysis used without previously experiencing any treatment for HIV- a Cox proportional hazards model stratified by cohort: 1; (ii) having at least one measurement of HIV RNA for clinical and viro-immunological outcomes and after treatment initiation; and (iii) having no treatment discontinuation due to toxicity, models were adjusted with immunomodulating agents before baseline or for all variables listed in Table 1 except ALT and lipids. during follow-up. Information used for this study As additional covariates, the models with ALT as included: gender, age at enrollment, risk factors for the outcome used the pre-therapy ALT and serological transmission of HIV-1 infection, Centers of Disease markers of infection with hepatitis B and C virus, while Control and Prevention (CDC) stage, the date of HIV models with lipids as outcomes used pre-therapy values treatment initiation, the specific antiretroviral regimen of the lipid of interest. Moreover, for virological and used, the dates of changes of that regimen and the clinical progression outcomes, we used weighted Cox reasons for those changes, CD4+ T-cell counts, plasma proportional hazards models estimating the inverse HIV-RNA levels, alanine aminotransferase (ALT), total probability of weighting [13], by adapting a publicly and low density lipoprotein (LDL) cholesterol and available SAS programme [14]. The year of starting triglyceride levels at initiation of antiretroviral therapy therapy and the pre-therapy CD4+ T-cell count were and every 3 months thereafter. chosen as time-varying confounders. The weights were Plasma HIV RNA was measured using quantitative based on the inverse of each patient’s probability of RT-PCR (Amplicor 1.5; Roche Molecular Systems, being started with EFV or LPV/r, given all the other

Table 1. Characteristics of the study population at the time of therapy initiation according to the use of EFV or LPV/r

Treatment

Characteristic EFV (n=481) LPV/r (n=193) P-value

Age, years 37 (19–70) 38 (19–74) 0.15 Female, n (%) 136 (28.3) 49 (25.4) 0.45 Modality of HIV exposure, n (%) Injecting drug users 119 (24.7) 39 (20.2) – Male homosexuals 108 (22.5) 54 (28.0) – Heterosexuals 205 (42.6) 74 (38.3) 0.18 Other/unknown 49 (10.2) 26 (13.5) – CDC Class C, n (%) 190 (39.5) 80 (41.5) 0.64 NRTI pair employed, n (%) d4T/3TC 79 (16.4) 36 (18.7) – d4T/ddI 43 (8.9) 20 (10.4) – ddI/3TC 12 (2.5) 3 (1.6) – ZDV/3TC 254 (52.9) 104 (53.9) – ZDV/ddI 10 (2.1) 0 (0.0) – Tenofovir/other NRTI 28 (5.8) 11 (5.7) 0.95 Year of starting, n (%) 1999–2000 215 (44.8) 1 (0.5) – 2001 109 (22.7) 23 (11.9) – 2002 88 (18.3) 97 (50.3) 0.0001 2003–2004 69 (14.4) 72 (37.3) – Viral load, log copies/ml 4.81 (2.68–6.38) 4.94 (2.65–7.00) 0.02 CD4+ T-cell count, cell/ml 255 (4–1,368) 145 (1–1,025) 0.0001 ALT, IU/ml 30 (5–438) 29 (5–163) 0.12 Total cholesterol, mg/dl 157 (67–332) 149 (19–297) 0.05 LDL cholesterol, mg/dl 95 (27–186) 95 (41–172) 0.32 Triglycerides, mg/dl 114 (29–1,456) 121 (29–515) 0.41

Unless specified, reported values are medians (range). CDC, Centers of Disease Control and Prevention; 3TC, lamivudine; ALT, alanine aminotransferase; d4T, stavudine; ddI, didanosine; EFV, efavirenz; LDL, low-density lipoprotein; LPV/r, lopinavir/ritonavir; NRTI, nucleoside reverse transcriptase inhibitor; ZDV, zidovudine.

confounders. This approach allowed more accurate EFV as third drug. The most frequently employed NRTI control for multiple confounding factors [15]. backbones were zidovudine/lamivudine (ZDV/3TC; Since only one patient of those who started therapy n=358, 53.1%), followed by stavudine/lamivudine before 2001 was treated with LPV/r, a sensitivity (d4T/3TC; n=115, 17.1%) and stavudine/didanosine analysis was conducted after excluding patients who (d4T/ddI; n=63, 9.4%). Patients receiving a LPV/r- initiated therapy before 2001. In a further sensitivity based first-line regimen had, on average, lower baseline analysis, the frequency of HIV RNA measurements was CD4+ T-cell counts and a higher viral load compared included as an additional covariate in the multivariable with those receiving EFV-based regimens. Also, the analysis of the virological outcome. LPV/r regimens were initiated at later calendar years compared with the EFV-based regimens (see Table 1). Results Virological failure Baseline patients characteristics HIV RNA was measured with a median frequency of The analysis focused on 674 patients who met study 3.95 times per year (interquartile range [IQR]: eligibility criteria: 418 from I.Co.N.A (62.0%), 98 2.76–5.80) in the EFV group and 5.94 times per year from UCSC (14.5%), 80 from INMI (11.9%) and 78 (IQR: 3.59–9.84) in the LPV/r group (Wilcoxon two- from IMIT-MI (11.6%). Baseline characteristics of the sample P=0.0001). Over a cumulative virological study population according to the third drug started in follow-up of 997.2 years, 94 patients experienced a addition to the NRTI pair are given in Table 1. LPV/r virological failure. The Kaplan–Meier estimate of the was given to 28.6% of patients, while 71.4% received time to virological failure according to which third

Antiviral Therapy 11:5 611 A De Luca et al.

drug was originally started is illustrated in Figure 1A. CD4+ T-cell count response The rate of virological failure in the group of patients CD4+ T-cell counts were measured with a median started on an LPV/r-based regimen was remarkably frequency of 4.20 times per year (IQR: 2.86–6.26) in similar to that of those started on an EFV-based the EFV group and 6.12 times per year (IQR: regimen: the percentages of patients with virological 3.57–9.84) in the LPV/r group (Wilcoxon two-sample failure by 72 weeks were 15.3% in the EFV group P=0.0001). Over a cumulative follow-up of 613.0 versus 15.7% in the LPV/r group (log-rank P=0.87). In years, 323 patients (47.9%) achieved a CD4 gain of at the multivariable analysis, the use of EFV or LPV/r as least 200 cells/µl above pre-therapy levels. The median the third drug remained not associated with differing time to achieve this CD4 gain was 61 weeks (95% CI: risks of virological failure (see Table 2). Independent 56–69) in the EFV group versus 62 (95% CI: 51–75) in predictors of shorter time to virological failure were: the LPV /r group (log-rank test P=0.53). There was still having acquired HIV through intravenous drug use as no significant difference in the adjusted analysis compared with heterosexual contacts, and the use of (RH=0.93, 95% CI: 0.66–1.30, P=0.66). Higher pre- an initial NRTI backbone of d4T/ddI compared with therapy viral load (RH=1.35 per log10, 95% CI: ZDV/3TC. Interestingly, there was no evidence that 1.14–1.61, P=0.0007), higher pre-therapy CD4+ T-cell the association of d4T/ddI with virological failure was count (RH=1.06 per 100 cells/µl, 95% CI: 1.00–1.12, dependent on the third drug used (relative hazard P=0.05) and more recent calendar year of starting [RH]=2.05, 95% CI: 1.07–3.91 in patients starting therapy (RH=1.14 per year, 95% CI: 1.00–1.30, EFV and RH=2.24, 95% CI: 0.51–9.77 in those P=0.05) were independently associated with a greater receiving LPV/r, P-value for interaction=0.99). Similar chance to achieve this CD4+ T-cell count increase. We results were obtained by using the weighted Cox also evaluated whether the chance of CD4+ T-cell count model (see Table 2). Results were unchanged after increase in the EFV and LPV/r-treated groups was restricting the analyses to patients starting therapy different according to the pre-therapy CD4+ T-cell after 2001 (LPV/r versus EFV standard Cox: adjusted counts and we found no evidence for such an interac- RH=1.35, 95% CI: 0.61–3.00, P=0.47; weighted Cox: tion (P=0.83). The adjusted RH comparing LPV/r with RH=1.55, 95% CI: 0.73–3.30, P=0.26). In an addi- EFV in patients with pre-therapy CD4+ T-cell count tional sensitivity analysis, the yearly frequency of ≤100 cells/µl was 1.01 (95% CI: 0.52–1.94). plasma HIV-RNA measurements was added as an additional covariate to the multivariable Cox model Clinical outcome and there was still no significant difference between The Kaplan–Meier estimates of the time to clinical treatment groups (LPV/r versus EFV: adjusted progression according to the initial treatment group RH=1.37, 95% CI: 0.54–3.51, P=0.51). are shown in Figure 1B. There was a significantly

Figure 1. Kaplan–Meier curves illustrating the cumulative proportion of patients with virological failure and clinical progression

A B 1 1 EFV EFV 0.9 0.9 LPV/r LPV/r 0.8 0.8 0.7 0.7 0.6 0.6 0.5 0.5 0.4 0.4 0.3 0.3 0.2 0.2 0.1 0.1 0 0 0 12 24 3648 60 72 84 0 12 24 36 48 60 72 84 Weeks from starting therapy Weeks from starting therapy

(A) cumulative proportion of patients with virological failure (>400 copies/ml) and (B) cumulative proportion of patients with clinical progression (new or recurrent AIDS-defining conditions or death), according to treatment with efavirenz (EFV; solid lines) or lopinavir/ritonavir. (LPV/r; dotted lines; log-rank P=0.003).

Table 2. Relative hazards of virological failure >400 copies/ml from fitting a traditional Cox proportional hazards model and a weighted Cox regression

Crude analysis Adjusted analysis* Adjusted analysis†‡

Factor RH (95% CI) P-value RH (95% CI) P-value RH (95% CI) P-value

LPV versus EFV 0.81 (0.45–1.45) 0.47 1.16 (0.58–2.32) 0.67 1.20 (0.60–2.43) 0.60 Pre-therapy viral load 0.94 (0.71–1.25) 0.94 1.10 (0.80–1.52) 0.57 0.98 (0.67–1.43) 0.91

(per 1 log10 higher) Pre-therapy CD4+ T-cell count 1.10 (1.00–1.20) 0.04 1.09 (0.99–1.20) 0.08 — — (per 100 cells/ml higher) CDC Class C No 1.00 – 1.00 – 1.00 – Yes 1.29 (0.85–1.98) 0.24 1.18 (0.75–1.85) 0.48 1.16 (0.70–1.93) 0.57 Age (per 10 years older) 0.82 (0.64–1.05) 0.11 0.94 (0.71–1.25) 0.69 0.79 (0.56–1.13) 0.20 Gender Male 1.00 – 1.00 – 1.00 – Female 1.08 (0.69–1.70) 0.73 1.22 (0.72–2.08) 0.46 1.16 (0.63–2.14) 0.63 Mode of transmission Heterosexuals 1.00 – 1.00 – 1.00 – Injecting drug users 2.22 (1.38–3.57) 0.001 2.44 (1.44–4.13) 0.0009 4.85 (2.68–8.80) 0.0001 Male homosexuals 1.07 (0.61–1.88) 0.82 1.14 (0.59–2.18) 0.70 1.20 (0.56–2.58) 0.63 NRTI pair ZDV/3TC 1.00 – 1.00 – 1.00 – d4T/3TC 1.55 (0.91–2.63) 0.11 1.53 (0.89–2.62) 0.12 1.61 (0.84–3.12) 0.15 d4T/ddI 2.29 (1.33–3.96) 0.003 2.16 (1.23–3.82) 0.008 5.64 (2.86–11.1) 0.0001 Year of starting (per more 0.77 (0.62–0.95) 0.01 0.81 (0.62–1.04) 0.10 — — recent year)

*Traditional adjustment. †Weighted adjustment. ‡Time varying confounders CD4+ T-cell count and calendar year of starting therapy are not included in the final pooled regression analysis as requested by the method. 3TC, lamivudine; CDC Centers of Disease Control and Prevention; CI, confidence interval; d4T, stavudine; ddI, didanosine; EFV, efavirenz; LPV, lopinavir; NRTI, nucleoside reverse transcriptase inhibitor; RH, relative hazard; ZDV, zidovudine. higher cumulative proportion of patients with clinical Discontinuation of LPV/r- or EFV-based initial progression in the group starting an LPV/r-regimen therapy (10.1% by 72 weeks) compared with the group A total of 164 patients interrupted EFV or LPV/r for any initially receiving an EFV-based regimen (4.6% by 72 reason over the study period. By 48 weeks of starting weeks, log-rank P=0.003). Nevertheless, after strati- therapy, 19.2% (95% CI: 15.5–22.9) of patients who fying by cohort and adjusting for the other baseline started EFV had interrupted this drug while 22.2% characteristics there was no longer any evidence of a (95% CI: 15.0–29.4) of those starting LPV/r had inter- statistical difference between the two groups (see rupted (log-rank P=0.96, see Figure 2A). The adjusted Table 3). A similar result was found using the RH comparing the rate of discontinuation of LPV/r with weighted Cox regression analysis (Table 3). In the that of EFV was 0.91 (95% CI: 0.59–1.42, P=0.68) traditional multivariable model, having been diag- suggesting that no difference in the rate of discontinua- nosed with AIDS before the date of therapy initiation tion of these two drugs seems to exist. and having acquired HIV through intravenous drug Over a cumulative follow-up of 932.5 years, 93 use were the only two factors significantly associated patients discontinued LPV/r or EFV because of toxicity. with the higher risk of clinical progression. The The most common main reason for discontinuing weighted analysis found that older age was an addi- LPV/r (n=21) was gastrointestinal intolerance (n=14, tional independent predictor of clinical progression 66.7%); the most common reason for discontinuing (see Table 3). EFV (n=72) was central nervous system toxicity (n=21, Finally, results were similar when the analysis was 29.2%), while hypersensitivity reactions led to EFV repeated using only the 458 patients who started LPV/r discontinuation in 10 cases (13.9%). The time to LPV/r or EFV after the year 2001 (standard Cox adjusted or EFV discontinuation because of toxicity is illustrated RH=1.57, 95% CI: 0.63–3.90, P=0.33 and weighted in Figure 2B. There was no difference in the proportion Cox RH=1.23, 95% CI: 0.53–2.85, P=0.62). of patients who discontinued EFV (15.2% by

Antiviral Therapy 11:5 613 A De Luca et al.

Table 3. Relative hazards of clinical progression from fitting a traditional Cox proportional hazards model and a weighted Cox regression

Crude analysis Adjusted analysis* Adjusted analysis†‡

Factor RH (95% CI) P-value RH (95% CI) P-value RH (95% CI) P-value

LPV versus EFV 0.81 (0.45–1.45) 0.47 1.64 (0.70–3.84) 0.25 1.51 (0.71–3.22) 0.29 Pre-therapy viral load 0.94 (0.71–1.25) 0.94 1.46 (0.86–2.47) 0.16 1.35 (0.77–2.36) 0.30

(per 1 log10 higher) Pre-therapy CD4+ T-cell count 1.10 (1.00–1.20) 0.04 0.96 (0.78–1.20) 0.74 – – (per100 cells/ml higher) CDC Class C No 1.00 – 1.00 – 1.00 – Yes 1.29 (0.85–1.98) 0.24 2.20 (1.07–4.51) 0.03 2.66 (1.27–5.58) 0.009 Age (per 10 years older) 0.82 (0.64–1.05) 0.11 1.42 (0.98–2.06) 0.06 1.95 (1.30–2.93) 0.001 Gender Male 1.00 – 1.00 – 1.00 – Female 1.08 (0.69–1.70) 0.73 0.87 (0.37–2.05) 0.75 1.09 (0.43–2.73) 0.86 Mode of transmission Heterosexauls 1.00 – 1.00 – 1.00 – Injecting drug users 2.22 (1.38–3.57) 0.001 2.62 (1.16–5.94) 0.02 3.12 (1.27–7.65) 0.01 Male homosexuals 1.07 (0.61–1.88) 0.82 1.10 (0.43–2.81) 0.84 1.06 (0.36–3.11) 0.91 NRTI pair ZDV/3TC 1.00 – 1.00 – 1.00 – d4T/3TC 1.55 (0.91–2.63) 0.11 1.41 (0.60–3.30) 0.43 1.30 (0.49–3.40) 0.59 d4T/ddI 2.29 (1.33–3.96) 0.003 1.24 (0.45–3.44) 0.67 2.03 (0.67–6.13) 0.21 Year of starting (per more 0.77 (0.62–0.95) 0.01 1.27 (0.89–1.80) 0.19 — — recent year)

*Traditional adjustment. †Weighted adjustment. ‡Crucial confounders CD4+ T-cell count and calendar year of starting therapy are not included in the final pooled regression analysis as requested by the method. 3TC, lamivudine; CDC, Centers of Disease Control and Prevention; CI, confidence interval; d4T, stavudine; ddI, didanosine; EFV, efavirenz; LPV, lopinavir; NRTI, nucleoside reverse transcriptase inhibitor; RH, relative hazard; ZDV, zidovudine.

72 weeks) or LPV/r (14.5% by 72 weeks, log rank HBV co-infected (RH=5.11, 95% CI: 1.24–21.10, P=0.65) because of toxicity. Results were confirmed by P=0.02) and HCV-coinfected patients (RH=5.28, 95% multivariable analysis (see Table 4). In this model, CI: 1.12–24.85, P=0.04) had a significantly fivefold female patients were at higher risk of treatment discon- higher risk of ALT elevation compared with HIV tinuation because of toxicity. mono-infected individuals. Overall, 307 patients experienced at least one cate- Elevation of laboratory parameters gory change up the total cholesterol ATP III definition Only 28 patients experienced a grade III ALT elevation, ladder [12]. By 72 weeks, fewer patients started on as defined in the methods. The incidences by 72 weeks EFV than those started on LPV/r had experienced such were 4.1% in the EFV group and 7.6% in the LPV/r an event (44.5% versus 57.0%, respectively, log-rank group (log-rank P=0.34). In the LPV/r group, 140 P=0.005). However, the adjusted RH of experiencing patients (72.5%) had test results available for hepatitis total cholesterol elevation for patients receiving LPV/r B virus surface antigen (HBsAg) and HCV antibodies as compared with those receiving EFV was 1.01 (95% (HCV-Ab); in the EFV group, 392 (81.5%) had results CI 0.66–1.55; P=0.98) suggesting no difference of both markers available. While there was no evidence between the two drugs for this outcome. Independent for a difference in proportion of HCV-Ab-positive indi- predictors of total cholesterol elevation were higher viduals between the groups (28.6% in the EFV versus pre-therapy viral load (RH=1.29 per log10 higher, 27.2% in the LPV/r group), HBsAg-positive individuals 1.00–1.66, P=0.05), female gender (RH=0.62 versus were more highly represented in the LPV/r group male, 95% CI: 0.42–0.93, P=0.02), having started (16.4% versus 7.6%, P=0.002). The adjusted estimate d4T/3TC (RH=1.75 versus ZDV/3TC, 1.16–2.64; of the RH of ALT elevation (LPV/r versus EFV) was P=0.008) and injecting drug use (RH=0.36 versus 1.98 (95% CI: 0.33–11.91, P=0.46). As expected, both heterosexuals, 0.23–0.58; P=0.0001).

Figure 2. Cumulative proportion of patients discontinuing treatment for any reason and for toxicity

A B

1 1 EFV EFV 0.9 0.9 LPV/r LPV/r 0.8 0.8 0.7 0.7 0.6 0.6 0.5 0.5 0.4 0.4 0.3 0.3 0.2 0.2 0.1 0.1 0 0 0 12 24 36 48 60 72 84 0 12 24 36 48 60 72 84 Weeks from starting therapy Weeks from starting therapy

(A) Cumulative proportion of patients discontinuing treatment for any reason and (B) for toxicity, according to treatment with efavirenz (EFV; solid lines) or lopinavir/ritonavir (LPV/r; dotted lines).

A total of 256 participants experienced at least one HIV RNA. Despite the unbalanced baseline character- category change up the LDL cholesterol ATP III defini- istics, results after substantial follow-up showed that tion ladder. The proportions experiencing LDL eleva- the two regimens were comparable in terms of virolog- tion by 72 weeks were 31.4% and 38.0% in the EFV ical, immunological and clinical efficacy. and LPV/r tretated groups, respectively (log-rank Since traditional adjustments are less effective in the P=0.08). The adjusted RH of LPV/r compared with presence of large imbalances in some of the patients EFV was 0.70 (95% CI: 0.35–1.38, P=0.30). pre-therapy characteristics [16], we used weighted Cox Triglyceride elevation was experienced by a total of proportional hazards models in a sensitivity analysis: 349 patients. The cumulative incidence was higher than this approach allowed us to control more accurately for that observed for the total cholesterol with 44.2% of multiple measured confounding factors [15]. Results patients who started treatment with EFV and 79.1% of confirmed and strengthened the findings obtained by those treated with LPV/r experiencing this event by the traditional model, showing that the different condi- week 72 (log-rank P=0.0001). Multivariable analysis tional probability of receiving LPV/r or EFV did not confirmed that patients receiving LPV/r were at have a major influence on the final outcomes. Results increased risk of experiencing triglycerides elevation were not affected by the different probability of compared with those receiving EFV (adjusted RH=1.69, switching LPV/r or EFV since there was no difference 1.14–2.50; P=0.01). Patients who started treatment in the risk of discontinuing these drugs and an on-treat- more recently were also at greater risk of triglyceride ment analysis showed similar results (not shown). elevation (RH=1.42 per more recent year, 1.21–1.68, Similar to the results of a randomized study (ACTG P=0.0001), while higher pre-therapy triglyceride level 384), the initial use of a NRTI backbone with d4T/ddI (RH=0.79 per 100 mg/dl higher 95% CI: 0.66–0.95, led to an increased hazard of virological failure as P=0.01) and female gender (RH= 0.26 versus male, compared with ZDV/3TC [4]. While in the ACTG 384 0.10–0.69; P=0.007) were protective factors. trial this difference was observed in combination with EFV but not with nelfinavir, we found that patients Discussion starting d4T/ddI had a poorer virological oucome irrepective of whether EFV or LPV/r were used in combi- In a multi-cohort observational study we compared nation with this backbone. This confirms that this back- EFV- with LPV/r-based first-line regimens in previously bone should be avoided in any first-line combination. antiretroviral-naive individuals. Patients receiving The intravenous drug use transmission group was LPV/r had more advanced HIV disease markers at also associated with virological failure: this is in agree- treatment initiation as compared with those receiving ment with previous observations linking this category EFV, as shown by lower CD4+ T-cell counts and higher to a lower rate of medication adherence [17,18].

Antiviral Therapy 11:5 615 A De Luca et al.

Table 4. Relative hazards of discontinuing LPV/r or efavirenz for toxicity from fitting a Cox proportional hazards model

Crude analysis Adjusted analysis

Factor RH (95% CI) P-value RH (95% CI) P-value

LPV versus EFV 0.94 (0.57–1.57) 0.81 0.92 (0.51–1.64) 0.76

Pre-therapy viral load (per 1 log10 higher) 0.76 (0.57–1.02) 0.06 0.80 (0.59–1.09) 0.15 Pre-therapy CD4+ T-cell count 1.07 (0.97–1.17) 0.19 1.10 (0.98–1.22) 0.09 (per 100 cells/ml higher) CDC Class C No 1.00 – 1.00 – Yes 1.20 (0.78–1.84) 0.42 1.39 (0.88–2.19) 0.16 Age (per 10 years older) 1.13 (0.91–1.41) 0.24 1.26 (0.99–1.59) 0.06 Gender Male 1.00 – 1.00 – Female 1.55 (1.01–2.37) 0.12 1.84 (1.10–3.04) 0.02 Mode of transmission Heterosexuals 1.00 – 1.00 – Injecting drug users 1.15 (0.70–1.90) 0.58 1.42 (0.82–2.46) 0.21 Male homosexuals 0.79 (0.46–1.35) 0.38 1.06 (0.57–1.98) 0.85 Other/unknown 0.73 (0.34–1.57) 0.42 0.80 (0.37–1.74) 0.58 NRTI pair ZDV/3TC 1.00 – 1.00 – d4T/3TC 0.57 (0.29–1.13) 0.04 0.55 (0.28–1.11) 0.09 d4T/ddI 1.06 (0.55–2.04) 0.32 1.02 (0.52–1.99) 0.96 Other pair 0.65 (0.29–1.45) 0.37 0.65 (0.29–1.45) 0.30 Year of starting (per one more recent year) 1.06 (0.90–1.26) 0.48 1.15 (0.94–1.40) 0.17

3TC, lamivudine; CDC, Centers of Disease Control and Prevention; CI, confidence interval; d4T, stavudine; ddI, didanosine; EFV, efavirenz; LPV, lopinavir; NRTI, nucleoside reverse transcriptase inhibitor; RH, relative hazard; ZDV, zidovudine.

The lack of a statistical difference in the rate of CD4+ Our data do not support a difference in the rate of T-cell count response of patients in the LPV/r and the LPV/r or EFV discontinuations because of toxicity, EFV group was expected, as the rate of virological failure although, in agreement with earlier studies using first was also not different in the two groups and is in agree- generation PIs [22], results show that female patients ment with previous findings indicating similar immune interrupt more easily for toxicity and older patients recovery while using PI-based or NNRTI-based regimens show a trend towards a higher risk for interruption. [19]. The stratified analysis (supported by a non-signifi- These findings are relevant since these categories are cant test for interaction) showed that patients with low not always represented in clinical trials and might at pre-therapy CD4+ T-cell count levels also achieved similar least in part explain the lower efficacy of antiretroviral immune recovery with EFV or LPV/r, suggesting that treatments in clinical practice as compared with both types of regimens can be employed, independently randomized studies. from the baseline immunological condition. No significant association was observed between the Results are partially different from those of two treatment group and most of the laboratory abnormal- smaller observational studies, showing similar virolog- ities that we have considered. However, patients who ical efficacy in one case [20], and better virological started an LPV/r-containing regimen had a higher risk response with EFV in the second study [21], while both of experiencing triglyceride elevation than patients referenced studies found a better CD4+ T-cell response treated with EFV-containing regimes. Moreover, with LPV/r. Discrepancies with our findings could be patients who started d4T/3TC as the NRTI pair had a explained by the smaller sample size of previous studies, higher risk of total cholesterol elevation than those by an unusual definition of the intent-to-treat principle starting a backbone with ZDV/3TC. This observation used for both viral load and CD4 outcomes in one is in agreement with higher lipid abnormalities study, which excluded cases with missing values from observed in d4T-containing regimens in another study the analysis [21], and by the type of analysis employed. [23]. In contrast, neither the treatment group nor any

of the other baseline parameters, with the exception of group. Therefore, both EFV and LPV/r together with HBV and HCV serostatus, predicted liver enzyme two NRTIs are very effective first-line regimens. elevations of grade 3 or higher. Our findings are in Regimen simplicity and toxicity profiles as well as agreement with cross-study comparisons from other plans for subsequent treatment lines after virological observational data [24,25]. failure should be balanced when making individual The major limitation of this study is the fact that treatment choices. patients were not randomly assigned to receive LPV/r or EFV. Observational studies comparing two treatment Acknowledgements regimens always suffer from the confounding-by-indication bias [15,16,26]. Nevertheless, in the analyses I.Co.N.A is supported by unrestricted educational regarding the efficacy endpoints, findings using the grants from GlaxoSmithKline Italy. conventional statistical methods could be repeated by means of an alternative analysis using inverse proba- Conflicts of interest bility of weighting, which better control for measured confounders. It must be stressed that none of the The authors disclose the following potential conflicts of methods employed in our analysis could replace interest: ADL was a member of advisory boards or randomization and that unmeasured confounders could received speakers honoraria from GlaxoSmithKline, still have played a role. Notwithstanding, these data are Boehringer Ingelheim and Abbott Virology; AC-L relevant since results from randomized comparative received speakers honoraria from GlaxoSmithKline studies are not yet available. Furthermore, observa- and Roche; AA was a member of advisory boards or tional studies reflect more closely the effect of a specific received speakers honoraria from GlaxoSmithKline, treatment in clinical practice, given that study patients Bristol-Myers Squibb, Gilead, Roche and Abbott are not selected as in randomized trials [23] and treat- Virology; A d’A M was a member of advisory boards ments are prescribed with a specific clinical indication. or received speakers honoraria from GlaxoSmithKline, Findings from this study can help in making individ- Bristol-Myers Squibb, Boehringer Ingelheim, Roche ualized treatment decisions. EFV has the advantage of and Abbott Virology a lower pill burden and our data suggest that the initial use of LPV/r may lead to a faster triglyceride elevation Additional files above normal levels. The advantages of EFV should be balanced against its lower genetic barrier compared The additional file ‘Complete cohort members list’ can with LPV/r. Drug resistance data were not available for be accessed via the Volume 11 Issue 5 contents page, this study, but observations from clinical trials and which can be found at www.intmedpress.com (by cohort studies show very little or an absence of PI resis- clicking on ‘Antiviral Therapy’ then ‘Journal PDFs’). tance and limited NRTI resistance in viral isolates from patients failing an LPV-based first-line regimen, while References failures with EFV-based first-line regimens are associ- 1. Mocroft A, Vella S, Benfield TL, et al. Changing patterns of ated with the selection of viral mutants showing higher mortality across Europe in patients infected with HIV-1. rates of cross-resistance [23,28,29]. Although this EuroSIDA Study Group. Lancet 1998; 352:1725–1730. study showed no difference in the outcome of surrogate 2. Vittinghoff E, Scheer S, O’Malley P, Colfax G, Holmberg SD, Buchbinder SP. Combination antiretroviral therapy and and clinical markers between the different initial treat- recent declines in AIDS incidence and mortality. J Infect ment strategies, a higher initial rate of drug resistance Dis 1999; 79:717–720. selection due to the use of drugs with lower genetic 3. Walmsley S, Bernstein B, King M, et al; M98-863 Study barrier could potentially lead to a more rapid exhaus- Team. Lopinavir-ritonavir versus nelfinavir for the initial treatment of HIV infection. N Engl J Med 2002; tion of drug options over a longer period of time. This 346:2039–2046. hypothesis deserves to be tested in further studies. 4. Robbins GK, De Gruttola V, Shafer RW, et al.; AIDS Clinical Trials Group 384 Team. Comparison of sequential three-drug regimens as initial therapy for HIV-1 infection. Conclusion N Engl J Med 2003; 349:2293–2303. 5. Gulick RM, Ribaudo HJ, Shikuma CM, et al.; AIDS Although patients were not randomized, this is the Clinical Trials Group Study A5095 Team. Triple-nucleoside regimens versus efavirenz-containing regimens for the largest analysis performed to date comparing safety initial treatment of HIV-1 infection. N Engl J Med 2004; and efficacy of EFV versus LPV/r in therapy-naive 350:1850–1861. patients. The analysis shows few differences between 6. van Leth F, Phanuphak P, Ruxrungtham K, et al.; 2NN the two drugs in terms of efficacy and discontinuation Study team. Comparison of first-line antiretroviral therapy with regimens including nevirapine, efavirenz, or both rates for toxicity. Our data, however, show a higher drugs, plus stavudine and lamivudine: a randomized open- risk of hypertriglyceridaemia in the LPV/r-treated label trial, the 2NN Study. Lancet 2004; 363:1253–1263.

Antiviral Therapy 11:5 617 A De Luca et al.

7. Guidelines for the Use of Antiretroviral Agents in HIV- 19. Pulido F, Arribas JR, Miro JM, et al.; EfaVIP Cohort Study Infected Adults and Adolescents. 6 October 2005. A Group. Clinical, virologic, and immunologic response to Service of the U.S Department of Health and Human efavirenz-or protease inhibitor-based highly active anti- Services. Available at: http://aidsinfo.nih.gov (last access 28 retroviral therapy in a cohort of antiretroviral-naive November 2005). patients with advanced HIV infection (EfaVIP 2 study). 8. Yeni PG, Hammer SM, Hirsch MS, et al. Treatment for J Acquir Immune Defic Syndr 2004; 35:343–350. adult HIV infection: 2004 recommendations of the 20. Manfredi R, Calza L, Chiodo F. First-line efavirenz International AIDS Society-USA Panel. JAMA 2004; versus lopinavir-ritonavir-based highly active antiretro- 292:251–265. viral therapy for naive patients. AIDS 2004; 9. Gazzard B, BHIVA Writing Committee. British HIV 18:2331–2333. Association (BHIVA) guidelines for the treatment of HIV- 21. Torti C, Maggiolo F, Patroni A, et al.; MASTER Cohort. infected adults with antiretroviral therapy (2005). HIV Exploratory analysis for the evaluation of lopinavir/riton- Med 2005; 6(Suppl 2):1–61. avir-versus efavirenz-based HAART regimens in 10. Cozzi Lepri A, Phillips AN, d’Arminio Monforte A, et al.; antiretroviral-naive HIV-positive patients: results from the ICONA Study Group. When to start highly active anti- Italian MASTER Cohort. J Antimicrob Chemother 2005; retroviral therapy in chronically HIV-infected patients: 56:190–195. evidence from the ICONA study. AIDS 2001; 15:983–990. 22. d’Arminio Monforte A, Cozzi-Lepri A, Rezza G, et al. 11. From the Centers for Disease Control and Prevention. 1993 Insights into the reasons for discontinuation of the first revised classification system for HIV infection and highly active antiretroviral therapy (HAART) regimen in a expanded surveillance case definition for AIDS among cohort of antiretroviral naive patients. I.CO.N.A. Study adolescents and adults. JAMA 1993; 269:729–730. Group. Italian Cohort of Antiretroviral-Naive Patients. 12. Expert Panel on Detection, Evaluation and Treatment of AIDS 2000; 14:499–507. High Blood Cholesterol in Adults. National Cholesterol 23. Gallant JE, Staszewski S, Pozniak AL, et al. 903 Study Education Program: Adult Treatment Panel III Report. Group. Efficacy and safety of tenofovir DF vs stavudine in Bethesda, MD: National Heart, Lung and Blood Institute combination therapy in antiretroviral-naive patients: a 3- National Heart, Lung and Blood Institute; 2001. year randomized trial. JAMA 2004; 292:191–201. Publication 01-3095. Available at: http://www.nhlbi.nih.gov/guidelines/cholesterol/index.htm 24. Sulkowski MS, Thomas DL, Mehta SH, Chaisson RE, (last access 28 November 2005). Moore RD. Hepatotoxicity associated with nevirapine or efavirenz-containing antiretroviral therapy: role of 13. Robins JM, Hernan MA, Brumback B. Marginal structural hepatitis C and B infections. Hepatology 2002; models and causal inference in epidemiology. Epidemiology 35:182–189. 2000; 11:550–560. 25. Sulkowski MS, Mehta SH, Chaisson RE, Thomas DL, 14. Hernan MA, Brumback B, Robins JM. Marginal structural Moore RD. Hepatotoxicity associated with protease models to estimate the causal effect of zidovudine on the inhibitor-based antiretroviral regimens with or without survival of HIV-positive men. Epidemiology 2000; concurrent ritonavir. AIDS 2004; 18:2277–2284. 11:561–570. 26. Wood E, Hogg RS, Heath KV, et al. Provider bias in the 15. D’Agostino RB Jr. Propensity score methods for bias reduc- selection of non-nucleoside reverse transcriptase inhibitor tion in the comparison of a treatment to a non-randomized and protease inhibitor-based highly active antiretroviral control group. Stat Med 1998; 17:2265–2281. therapy and HIV treatment outcomes in observational 16. Phillips AN, Grabar S, Tassie JM, Costagliola D, Lundgren studies. AIDS 2003; 17:2629–2634. JD, Egger M. Use of observational databases to evaluate et al the effectiveness of antiretroviral therapy for HIV infection: 27. Madge S, Mocroft A, Wilson D, . Participation in clin- comparison of cohort studies with randomized trials. ical studies among patients infected with HIV-1 in a single EuroSIDA, the French Hospital Database on HIV and the treatment centre over 12 years. HIV Med 2000; Swiss HIV Cohort Study Groups. AIDS 1999; 1:212–218. 13:2075–2082 28. Kempf DJ, King MS, Bernstein B, et al. Incidence of resis- 17. Wood E, Montaner JS, Yip B, et al. Adherence and plasma tance in a double-blind study comparing lopinavir/ritonavir HIV RNA responses to highly active antiretroviral therapy plus stavudine and lamivudine to nelfinavir plus stavudine among HIV-1 infected injection drug users. Can Med Assoc and lamivudine. J Infect Dis 2004; 189:51–60. J 2003; 169:656–661. 29. Phillips AN, Dunn D, Sabin C, et al. UK Collaborative 18. Sterne JA, Hernan MA, Ledergerber B, et al. Swiss HIV Group on HIV Drug Resistance; UK CHIC Study Group. Cohort Study. Long-term effectiveness of potent antiretro- Long term probability of detection of HIV-1 drug resistance viral therapy in preventing AIDS and death: a prospective after starting antiretroviral therapy in routine clinical practice. cohort study. Lancet 2005; 366:378–384. AIDS 2005; 19:487–494.

Accepted for publication 6 April 2006