|Home | About | Journals | Submit | Contact Us | Français|
ANRS 127 was a randomized pilot trial involving naïve patients receiving two dual-boosted protease inhibitor (PI) combinations. Virological response, defined as a plasma HIV RNA level of <50 copies/ml at week 16, occurred in only 41% patients. Low baseline plasma HIV RNA level was the only significant predictor of virological response. The purpose of this study was to investigate the impact on virological response of pretherapy mutations in cleavage sites of gag, gag-pol, and the gag-pol frameshift region. The whole gag gene and protease-coding region were amplified and sequenced at baseline and at week 16 for 48 patients still on the allocated regimen at week 16. No major PI resistance-associated mutations were detected either at baseline or in the 26 patients who did not achieve virological response at week 16. Baseline cleavage site substitutions in the product of the gag open reading frame at positions 128 (p17/p24) (P = 0.04) and 449 (p1/p6gag) (P = 0.01) were significantly more frequent in those patients not achieving virological response. Conversely, baseline cleavage site mutation at position 437 (TFP/p6pol) was associated with virological response (P = 0.04). In multivariate analysis adjusted for baseline viral load, these 3 substitutions remained independently associated with virological response. We demonstrated here, in vivo, an impact of baseline polymorphic gag mutations on virological response in naïve patients receiving a combination of two protease inhibitors. However, it was not possible to link the substitutions selected under PI selective pressure with virological failure.
Since the accession of highly active antiretroviral therapy (HAART), guidelines for first-line HIV-1 infection therapy have recommended the use of a combination of two nucleoside reverse transcriptase inhibitors (NRTI) associated with a protease inhibitor (PI) or a nonnucleoside reverse transcriptase inhibitor. Problems of therapeutic failure, adherence, and toxicity had justified other antiretroviral regimen alternatives. Improvements in the determination of pharmacokinetic interactions had led to a boost in protease inhibitor bioavailability by concomitant use of low doses of ritonavir (boosted protease inhibitor), increasing protease inhibitor efficacy.
The combination of two protease inhibitors boosted with ritonavir, associated or not associated with other antiretroviral drug classes, has been used in response to therapeutic failure among multiresistant viruses or in NRTI-sparing therapeutic regimen strategies (19a, 30, 43, 44a). Resistance to protease inhibitors follows an accumulation of mutations in the protease (PR)-coding region in first-time major mutations, which lead to resistance alone at one or several protease inhibitors and diminish viral fitness, and in second-time minor mutations, which improve the processing efficiency of the resistant enzyme, thus partially restoring viral replication (14, 24, 32, 42, 49). In first-line regimens containing a boosted protease inhibitor, mutations in the protease-coding region are very rarely detected at the time of virological failure (18, 24). One hypothesis for explaining the absence of protease inhibitor mutations at the time of virological failure might be that mutations are selected not in the protease-coding region but at a distance in the gag gene.
The protease interacts with seven amino acids at each cleavage site. A specific nomenclature was established for the five amino acids upstream and the five amino acids downstream of the site where the protease cleaves the Gag and Gag-Pol polyproteins, called P5 to P1 and P1′ to P5′, respectively (40). Protease activity leads to the cleavage of Gag and Gag-Pol polyprotein precursors. In the product of the gag open reading frame, Gag polyproteins are cleaved at five cleavage sites into p17 (MA), p24 (CA), p2 (SP1), p7 (NC), and p6gag. In the product of the gag-pol open reading frame, Gag-Pol polyproteins are cleaved at eight cleavage sites into p17 (MA), p24 (CA), p2 (SP1), p7 (NC), transframe protein (TFP), p6pol, protease, reverse transcriptase, and integrase (10, 13, 19, 39). In HIV-1, as in many other retroviruses, frameshifting is required to synthesize Gag and Gag-Pol polyprotein precursors. Ribosomal −1 frameshifting mediated by a downstream signal is required for translation of the mRNA in these precursors. The downstream signal is composed of a heptameric X XXY YYZ consensus slippery sequence (U UUU UUA in HIV) and a downstream secondary RNA structure which causes the ribosome to pause at the p7/p1 junction (13, 37). The downstream secondary RNA structure is a stem-loop or hairpin structure in HIV. Frameshifting is a rare controlled event, occurring only for 1 of 10 to 20 ribosomes, and allows for a correct Gag/Gag-Pol ratio for optimal virus activity (15, 22, 36, 37). The RNA folding and stability of the gag-pol frameshift region can be evaluated by the measure of hairpin free energy.
ANRS 127 was a randomized pilot trial involving antiretroviral therapy (ART)-naïve patients receiving two dual-boosted protease inhibitor-nucleoside reverse transcriptase inhibitor-sparing regimens, involving fosamprenavir-atazanavir boosted with ritonavir (fosamprenavir-atazanavir/r) (n = 30) or saquinavir-atazanavir/r (n = 31). Virological response was defined as a plasma HIV-1 RNA level of <50 copies/ml at week 16, which occurred in only 41% of patients. No protease inhibitor resistance mutation in the protease-coding region was detected, regardless of the double-boosted protease inhibitor combination at the time of virological failure. The only factor predictive of virological success was baseline viral load (30). The aim of this study was to explore, for patients included in the 2IP ANRS 127 assay, the impact on virological response of mutations at baseline in gag cleavage sites and in the gag-pol frameshift region, involving residues at the main positions in gag described so far to be associated with decreased efficacy of protease inhibitors.
(This work was presented in part at the 16th Conference on Retroviruses and Opportunistic Infections, 8 to 11 February 2009, Montreal, Canada [30a], and at the 18th International HIV Drug Resistance Workshop, 9 to 13 June 2009, Fort Myers, FL [30b].)
In this virological substudy of the 2IP trial (30), we focused on 48 patients still on the allocated regimen at week 16 (26 patients with suboptimal virological response [viral loads of ≥50 copies/ml at week 16] and 22 patients with viral loads below 50 copies/ml at week 16).
Viral RNA was extracted from plasma (1 milliliter) with the automated nucleic acid extractor Magnapure and the Total NA large-volume program (Roche Diagnostics, GmbH, Mannheim, Germany) and stored at −80°C for 55 patients at inclusion and 26 patients at week 16. First, 10 μl of RNA was used for reverse transcription-PCR (RT-PCR) (Titan one-tube RT-PCR kit; Boehringer, Manheim, Germany) with forward and reverse primers 5′-GCGGCGACTGGTGAGTACGCC-3′ and 5′-GGCAAATACTGGAGTATTGTATG-3′, respectively, in 10 μl of 5× RT-PCR buffer, 2.5 μl of dithiothreitol (DTT) solution, 4 μl of deoxynucleoside triphosphate (dNTP) mix, 2 μl each of 10 μM primers, 1 μl of RNase inhibitor, 1 μl of Titan enzyme mix, and 17.5 μl of PCR-grade water. The cycling parameters were as follows: 50°C for 30 min; 94°C for 2 min; 10 cycles at 94°C for 30 s, 55°C for 30 s, and 68°C for 2 min; 25 cycles at 94°C for 30 s, 55°C for 30 s, and 68°C for 2 min, with a cycle elongation of an additional 5 s for each cycle; and a final step at 68°C for 7 min. RT-PCR products were stored at −20°C. Second, 5 μl of cDNA was amplified by nested PCR using forward primer 5′-AGCAACCCTCTATTGTGTGCAT-3′ and reverse primer 5′-AATGCTTTTATTTTTTCTTCTGTCAATGGC-3′ in 5 μl of 10× high-fidelity PCR buffer, 0.4 μl of deoxynucleotide triphosphates (100 μM each; Roche Diagnostic, Amersteim, Germany), 2 μl of MgSO4 (50 mM), 1 μl each of 10 μM primers, 0.4 μl of Platinum Taq high-fidelity DNA polymerase (1.0 U; Invitrogen, Carlsbad, CA), and 37 μl of PCR-grade water (Eurobio, les Ulis, France). The cycling parameters were as follows: 94°C for 2 min; 40 cycles at 94°C for 30 s, 60°C for 30 s, and 68°C for 2 min; and a final step at 68°C for 7 min.
Subsequently, population sequencing was performed using a BigDye Terminator version 1.1 cycle sequencing kit (Applied Biosystems, Paris, France), with forward primer 5′-AGCAACCCTCTATTGTGTGCAT-3′ and reverse primer 5′-CTCTTTAACATTTGCATGGCTGC-3′ for gag cleavage site p17/p24 and with forward primer 5′-ATAATCCTGGGATTAAATAAAATAG-3′ and reverse primer 5′-CTAATACTGTRTCATCTGCTCC-3′ for gag cleavage sites p24/p2, p2/p7, p7/p1, p1/p6gag, TFP, TFP/p6pol, and p6pol/PR. Sequencing was performed with an ABI Prism 3130 sequencer (Applied Biosystems, Foster City, CA). Nucleotide sequences were aligned using Sequence Navigator software (Applied Biosystems), and differences in amino acid sequence relative to the wild-type HXB2 virus were noted. Mutations in the protease were listed according to the IAS-USA 2009 expert panel (http://www.iasusa.org/). All substitutions observed in gag cleavage sites were noted.
Phylogenetic analyses were performed by estimating the relationships among pol sequences and reference sequences of HIV-1 genetic subtypes and circulating recombinant forms obtained from the Los Alamos Database (http://hiv-web.lanl.gov/). Nucleotide sequences were aligned with the CLUSTAL W program, version 1.7.29. Phylogenetic reconstruction was performed using a Kimura 2-parameter model and the neighbor-joining method with 500 bootstrapped data sets.
RNA folding and the stability of the hairpin structure of the gag-pol frameshift region were determined using measurement of free energy in accordance with Turner's rules (CombFold, RNAsoft [http://www.rnasoft.ca/cgi-bin/RNAsoft/CombFold/combfold.pl]). The sequence analyzed was from nucleotide 2093 to nucleotide 2135 with respect to the whole HXB2 genome (http://www.hiv.lanl.gov/content/sequence/HIV/COMPENDIUM/2009/hiv1dna.pdf).
The population of interest is the population under treatment. Intent-to-treat (ITT) analyses were also performed for sensitivity analyses (results not shown).
Differences in the frequency of amino acid sequences compared to HXB2 for cleavage sites in the product of the gag open reading frame and for TFP, TFP/p6pol, and p6pol/PR in the product of the gag-pol open reading frame were studied according to subtype using Chi-square or Fisher exact tests.
Differences in the frequency of amino acid sequences relative to HXB2 for cleavage sites in the product of the gag open reading frame and for TFP, TFP/p6pol, and p6pol/PR in the gag-pol open reading frame were studied according to virological response at week 16 using Chi-square or Fisher exact tests. Mutations yielding P values of <5% in the univariate analysis were analyzed in a logistic regression model with adjustment for log10 viral load at baseline, which was highly predictive of the virological response at week 16.
Baseline RNA folding and stability were compared according to virological response at week 16, subtype, and baseline viral load (<50,000 versus ≥50,000 copies/ml) using the Wilcoxon rank sum test.
At baseline, sequences of the protease were available for 55 patients, of whom 48 were still on the allocated treatment at week 16. No major PI resistance mutations were detected at baseline. Sequences of gag cleavage sites and the gag-pol frameshift region were available for 48 patients still on the allocated regimen at week 16, except for p6pol/PR, for which sequences were available for 47 out of 48 patients. The polymorphism percentage value was defined as the proportion of virus exhibiting mutations at each position among all patients at baseline. We noted polymorphisms of more than 20% for some positions of gag cleavage sites in the product of the gag open reading frame at p17/p24 (position 128), p2/p7 (positions 373 to 376 and 380), and p1/p6gag (position 449) and in the product of the gag-pol open reading frame at TFP/p6pol (positions 437, 441, 443, and 444) and p6pol/PR (positions 484 and 486 to 488). The highest polymorphic cleavage site was p2/p7, with three positions (positions 373 to 375) having polymorphisms higher than 30% for all the patients. At position 380, substitutions were present for more than 30% of patients with plasma HIV-1 RNA levels above 50 copies/ml at week 16. In univariate analysis for any dual-boosted PI combination, cleavage site substitutions in the product of the gag open reading frame at positions 128 (p17/p24) (P = 0.04) and 449 (p1/p6gag) (P = 0.01) were associated with virological failure (Fig. (Fig.1).1). There was no difference according to PI regimen. Conversely, substitutions in the product of the gag-pol open reading frame at position 437 (TFP/p6pol) (P = 0.04) were associated with virological response (Fig. (Fig.1).1). In multivariate analysis, these three substitutions remained independently associated with virological outcome, regardless of the level of baseline viral load.
Among the 26 patients with viral loads above 50 copies/ml at week 16, 24 sequences of the protease were available (Fig. (Fig.2).2). No major PI resistance mutations were selected between baseline and week 16 (Fig. (Fig.2).2). In contrast, minor PI mutations were selected during this period for three patients under the allocated treatment (L10I [n = 1], V11I [n = 1], and A71V [n = 1]) (Fig. (Fig.22).
As shown in Table Table1,1, gag cleavage site sequences were obtained for 16 patients among the 26 patients with viral loads above 50 copies/ml at week 16. For gag cleavage site p6pol/PR, only 15 sequences out of the 26 samples were available. Compared to baseline sequences, substitutions in gag cleavage sites were observed for eight patients (in p2/p7 [n = 3], p1/p6gag [n = 3], and p6pol/PR [n = 3]). Reversions to wild-type sequences were observed for 10 patients (in p17/p24 [n = 3], p24/p2 [n = 1], p2/p7 [n = 3], p7/p1 [n = 2], p1/p6gag [n = 2], TFP/p6pol [n = 2], and p6pol/protease [n = 3]).
HIV-1 B subtypes were found in 29 patients. Nineteen patients were infected with HIV-1 non-B subtypes (A1 [n = 4], CRF_01AE [n = 2], CRF_02AG [n = 9], CRF_06cpx [n = 1], CRF_11cpx [n = 1], CRF_18cpx [n = 1], and a subtype for which data are not available [n = 1]). The highest number of polymorphic substitutions in gag cleavage sites was seen for non-B subtype viruses in comparison to B subtype viruses (Fig. (Fig.3).3). No impact of HIV-1 subtype (B versus non-B) on virological response was observed (P = 0.11).
The measure of baseline hairpin free energy was performed using hairpin gag-pol frameshift signal sequences for the 48 patients. No association between baseline hairpin free energy and virological response was evidenced (P = 0.61). Conversely, there was an association between HIV-1 subtypes and hairpin free energy. Indeed, as shown in Table Table2,2, the level of hairpin free energy was higher for HIV-1 B subtypes (n = 29; median, −25.6 kcal/mol) than for HIV-1 non-B subtypes (n = 19; median = −22.5 kcal/mol) (P = 0.0005). Moreover, the level of free energy was higher for HIV-1 B subtypes than for the HIV-1 CRF_02AG subtype (n = 9; median, −22.7 kcal/mol) (P < 0.0001).
The only predictive factor of virological outcome in the 2IP study was the level of baseline viral load (30). No major PI resistance-associated mutations were detected either at baseline or in the patients who did not achieve virological response at week 16. In this virological substudy, baseline cleavage site substitutions in the product of the gag open reading frame at positions 128 (p17/p24) and 449 (p1/p6gag) were significantly more frequent in those patients achieving suboptimal virological response at week 16. Conversely, baseline cleavage site mutations at position 437 (TFP/p6pol) were associated with virological response. These 3 substitutions remained independently associated with virological outcome in multivariate analysis, regardless of the level of baseline viral load and the dual-boosted PI combination regimen.
Several studies have shown the presence of gag cleavage site mutations in association with mutations in the protease-coding region (1, 2, 5, 8, 16, 17, 27, 29, 32, 33, 45, 47, 50). These mutations were mainly located in the P4-P3′ region of the cleavage site, probably because these residues interact directly with the protease. gag cleavage site mutations were found more frequently among protease inhibitor-experienced patients than among naïve patients, suggesting that these mutations selected under protease inhibitor selective pressure direct the mechanism of resistance to a compensatory mechanism to increase the activity of the mutant protease (1, 7, 8, 16, 17, 20, 26, 34, 46, 46a, 47). An increase of the affinity of the precursors for the protease mutant might enhance the cleavage of Gag and Gag-Pol precursors (1, 8, 14, 23, 28, 32, 41, 48). Indeed, when gag cleavage site mutations were present, a partial recovery of virus replicative capacity was achieved by increasing the cleavage of Gag and Gag-Pol precursors in comparison to the level for virus with mutated protease and no cleavage site-associated mutation (11, 32, 34, 38, 50). Moreover, a role for amino acids constituting the C terminus of the p6pol protein was also highlighted. These positions might intervene in regulating the activation and self-cleaving of the protease at the beginning of viral maturation (31).
An occurrence recently observed in vitro in cell culture with increasing doses of protease inhibitor and also in vivo is selection of substitutions in gag cleavage sites or in the gag-pol frameshift region without any selection of mutations in the protease-coding region. These substitutions alone were associated with an increase in the inhibitory concentrations for the protease inhibitors tested (9, 35).
In our study, gag cleavage sites and the gag-pol frameshift region were highly polymorphic at baseline. We found that positions 128 and 449 were associated with virological failure. Baseline substitutions at position 128 in p17/p24 have never been described until now as associated with virological failure in PI-experienced patients. Residue 128 is located far from the substrate binding subsite at the P5 position in the p17/p24 cleavage site and might not directly alter specific interaction with the protease enzyme but might modify the conformation of the substrate region in the polyprotein, probably favoring the cleavage step (14). L449P/F substitutions in p1/p6gag have already been described to occur in protease inhibitor-naïve and -experienced patients. The L449P mutation is polymorphic in protease inhibitor-naïve patients and has also been reported to occur with mutations in the protease (1, 32, 33). In several studies, L449F was observed only in protease inhibitor-experienced patients and with protease resistance mutations (1, 23, 33, 48). L449P/F substitutions might act as compensatory mutations allowing an increase of the cleavage activity of the mutant protease. Moreover, the L449F residue is located in the RNA “stem-loop” secondary structure, downstream of the native slippery sequence of the gag-pol frameshift signal, and created a new slippery sequence for ribosomes. This might potentially lead to an increase in pol translation and therefore an increase in viral replication (12, 19).
gag cleavage sites may have relatively high levels of polymorphism among different subtypes of HIV. Some polymorphisms might have a potential impact on virological response to treatment in protease inhibitor-experienced patients (32, 33). Several studies reported that the p2/p7 cleavage site had a large genetic polymorphism (27, 33). Selection of mutations at position 373 in p2/p7 was correlated with poor virological response (33). In the Monark trial, which evaluated first-line antiretroviral induction monotherapy with lopinavir, the number of baseline gag cleavage site mutations at the p2/p7 site had a negative impact on week 96 virological response (18a). Indeed, the presence of more than two mutations in the p2/p7 cleavage site was significantly associated with virological failure. In our study, p2/p7 was the more polymorphic gag cleavage site, mostly in patients harboring non-B subtype viruses. We did not observe that the number of substitutions in this gag cleavage site had any impact on virological response. However, the results of these two studies are difficult to compare because of the different protease inhibitor regimens used (fosamprenavir-atazanavir/r or saquinavir-atazanavir/r therapy versus lopinavir/r monotherapy) and the times of assessment of virological failure (week 16 versus week 96). Moreover, reduced susceptibilities to protease inhibitors of CRF01_AE strains, showing polymorphisms in the protease and gag cleavage sites, have been recently demonstrated (21).
For the product of the gag-pol open reading frame, we showed that the asparagine residue at position 437, located in the TFP, had a positive impact on virological response. In the complete genome database alignments (http://www.hiv.lanl.gov/content/index), we noticed that many of the compiled strains carried asparagine at position 437 in comparison to the HXB2 reference strain, allowing us to consider that asparagine might be the most representative circulating residue in naïve patients. In 2007, Callebaut et al., in in vitro experiments involving protease inhibitor mutation selection with a nonmarketed protease inhibitor, described that asparagine at position 437 could increase protease inhibitor phenotypic resistance in a specific genetic context involving the product of a gag-pol open reading frame harboring other mutations, including L441P (4). The eight amino acids of the TFP are part of two cleavage sites (p7/TFP or TFP/p6pol) and part of the gag-pol frameshift region. The potential mechanisms of resistance evocated by Callebaut et al. involved (i) an increase in protease mutant activity induced by a compensatory mechanism, (ii) a higher level of production of protease, and (iii) a decrease in inhibition induced naturally by the TFP. It might be necessary to conduct further studies with different protease inhibitors and larger numbers of patients to explore this phenomenon.
Several genotypic changes (selection or reversion to the wild type) were observed in gag cleavage sites between baseline and virological failure. In our study, among the substitutions and reversions selected under selective pressure, none were associated with virological outcome at week 16. Conversely, only baseline gag cleavage site mutations at positions 128 and 449 in the product of the gag open reading frame and at position 437 in the product of the gag-pol open reading frame were found to be associated with virological outcome at week 16.
In our study, no correlation was observed between the baseline hairpin free energy of the gag-pol frameshift signal and virological response. However, we showed that the level of hairpin free energy was higher in B subtypes than in non-B subtypes, particularly the CRF_02AG recombinant, the most prevalent HIV-1 non-B subtype in our study. These results are in favor of a better stability of this structure in HIV-1 B subtypes. HIV-1 non-B subtypes generally have higher numbers of gag cleavage site and gag-pol frameshift mutations. The similarity of the gag-pol frameshift signal structures of different HIV-1 group M viruses has been demonstrated (3, 6). HIV-1 A, F, and CRF_01AE subtypes had lower levels of thermodynamic stability than B, C, and D subtypes (6). It has been shown that a decrease in free energy in vitro was correlated with poor efficiency of change in the gag-pol open reading frame. In different studies, an absence of viral replication was observed for reductions in free energy between 35% and 60% (3, 44). It has recently been reported, for patients infected with HIV-1 B subtypes, that there were no significant differences in frameshift efficiency for sequences with and without gag cleavage site mutations, though significant differences in hairpin free energy were observed (25). However, Knops et al. reported a significant increase in frameshift efficiency when L449F (p1/p6gag) and I437V (p7/p1) associated with additional polymorphism were present. A decrease in the free energy of the RNA secondary structure of the gag-pol frameshift signal, inducing instability in this signal and thus a diminution of enzyme production, could be thwarted by an increased affinity of substrates for protease. This new method of gag-pol frameshift evaluation could be an additional parameter for a better understanding of the viral phenotypic characteristics.
In conclusion, we demonstrated here for the first time in vivo an impact of baseline polymorphic gag mutations on virological response in naïve patients receiving a combination of two protease inhibitors. However, it was not possible to link the substitutions selected under protease inhibitor selective pressure with virological failure. Further investigations are needed in order to better clarify the clinical relevance of such mutations. Although we have not underscored an impact of the hairpin free energy of the gag-pol frameshift signal on virological response, we pointed out differences in the distribution of free energy according to viral subtype.
The research leading to these results has received funding from the Agence Nationale de Recherche Sur le SIDA et les Hépatites Virales (ANRS), the European AIDS Treatment Network (NEAT, WP6) (grant no. 037570), and the European Community's Seventh Framework Programme (FP7/2007-2013) under the project “Collaborative HIV and Anti-HIV Drug Resistance Network” (CHAIN) (grant no. 223131). We have no conflicts of interest to declare.
We thank Christian Callebaut for helpful discussion. We thank Emmanuelle Morandi for her technical skills.
The ANRS 127 study group (ANRS, Paris, France) comprised the following individuals and locations. The members of the scientific committee were P. Yeni, R. Landman, F. Brun-Vezinet, D. Descamps, G. Peytavin, M. Bentata, C. Piketty, J. P. Aboulker, C. Capitant, C. Chazallon, M. J. Commoy, Y. Bennai, B. Hadacek, and D. Merah. The participating clinical departments (all in France) were Groupe Hospitalier Bichat, Paris (R. Landman and G. Fraqueiro), Groupe Hospitalier Pitié-Salpétrière, Paris (C. Katlama and M. Pauchard), Hôpital Tenon, Paris (G. Pialoux and C. Fontaine), Centre Hospitalier de la Région Annecienne, Annecy (C. Michon and M. Bensalem), Centre Hospitalier Départemental, La Roche Sur Yon (P. Perre and I. Suaud), Hôpital Saint-Louis, Paris (J. M. Molina and A. Rachline), Hôpital de Bicêtre, Le Kremlin-Bicêtre (C. Goujard and M. Môle), Hôpital Pontchaillou—CHRU, Rennes (C. Arvieux and M. Ratajczak), Hôpital Avicenne, Bobigny (M. Bentata and F. Touam), Hôpital Saint-Antoine, Paris (P. M. Girard and J. L. Lagneau), CHU Hôtel Dieu, Nantes (F. Raffi and H. Hue), Hôpital Foch, Suresnes (D. Zucman), Hôpital Raymond Poincaré-Vidal, Garches (P. de Truchis and H. Berthe), Hôpital Henri Mondor, Créteil (Y. Levy and C. Jung), Hôpital Européen Georges Pompidou, Paris (C. Piketty), CHU Hôpital Gui de Chauliac, Montpellier (J. Reynes and M. Vidal), Hôpital Saint-Louis, Paris (D. Sereni), and Hôpital Louis Mourier, Colombes (M. Bloch). M. L. Chaix, P. Flandre, and Y. Yazdanpanah were on the data and safety monitoring board. J. P. Aboulker, C. Capitant, N. Leturque, E. Netzer, V. Foubert, and C. Chazallon (statistics) and S. Izard and A. Arulananthan (data management) worked at the coordinating center (INSERM SC10, Villejuif, France).
Published ahead of print on 3 May 2010.