Three randomized controlled trials have proved the failure of a less intensive maintenance therapy, as compared with prolonged induction therapy, to maintain viral load < 500 copies/ml [1–3]. In Trilège, an interim analysis in December 1997 demonstrated that the proportion of patients reaching the primary endpoint (two consecutive specimens of plasma HIV RNA > 500 copies/ml) was significantly higher in the zidovudine (ZDV) + lamivudine (3TC) (P < 0.001) and ZDV + indinavir (IDV) groups (P = 0.01) than in the continued triple-drug maintenance arm . Early clinical trials of protease inhibitor monotherapy suggested that the pathway to treatment failure was exclusively via drug resistance [4,5]. Results from a case–control study in Trilège, indicate that the first virological failure is indeed multifactorial and not the result of multidrug resistance alone . Undoubtedly, adherence to a treatment regimen is essential.
Although the interim analysis made after a median follow-up of 6 months provided an answer to the primary objective of the trial , patients were followed up to 18 months as planned in the protocol. The main objective of the present work is to investigate whether Trilège patients experiencing virological failure were able to reduce their plasma HIV RNA level < 500 copies/ml, according to randomized groups, post-failure treatment modifications, and baseline characteristics.
Study design and patients
In December 1997, randomization of antiretroviral-naive patients in the Trilège trial had been stopped on the basis of the first full interim analysis. The study design, which is described elsewhere , can be summarized as follows. In the 3-month induction phase, patients received open-label treatment with LZDV, 3TC, and IDV. Patients who experienced no severe adverse reactions and who had an HIV RNA level < 500 copies/ml at month 2, were randomly assigned at month 3 to either ZDV + 3TC + IDV, or ZDV + 3TC, or ZDV + IDV. During the maintenance phase, clinical and biological assessments were performed every 6 weeks from randomization. Virologic failure was defined as the first plasma HIV RNA value of ≥ 500 copies/ml (S1), confirmed in a second specimen 6 weeks later (S2). In these failing patients – defined as cases – the resumption of the triple-drug regimen was recommended by the protocol in patients randomized to dual therapy arms while alternative therapy were encouraged in patients from the triple-drug regimen.
Patients controlling viral replication (HIV RNA levels < 500 copies/ml) and defined as controls in both of the dual therapy arms, were encouraged to reinitiate the triple-drug regimen although other alternative therapies were conceivable. Controls in the ZDV + 3TC + IDV arm were prompted to continue the triple therapy. Modifications of treatment were recorded prospectively on the case report form. Treatment changes were grouped as combination of either: (i) nucleoside analogue reverse transcriptase inhibitors (NRTI) only; (ii) NRTI plus protease inhibitors (PI); (iii) NRTI plus PI plus non-nucleoside analogue reverse transcriptase inhibitors (NNRTI); (iv) NRTI plus NNRTI; or (v) no treatment.
Viral load measurements
Plasma HIV RNA levels were determined using the Amplicor HIV assay (Roche Laboratories, Alameda, California, USA; limit of detection, 200 copies/ml). Median viral rebound in the S1 sample (first plasma sample with HIV-1 RNA > 500 copies/ml) from cases was compared with baseline value to assess the extent of viral rebound.
This work focused on the dynamics of plasma HIV RNA in cases beyond the time of first virological failure. We analysed the number of subsequent viral load rebounds, defined in this study as any single plasma HIV RNA titre > 500 copies/ml occurring after the specimen that confirmed a virological failure (S2). Such viral load rebounds are considered to be recurrent virological events in the model using multiple failure time data as described below.
The analyses used an intent-to-treat approach that included all observations randomized in the Trilège trial. The times to virological failure were compared among the treatment groups by means of log-rank tests. Annual incidence rates of virological failure were the ratio of the number of failures to the corresponding number of person years. The Wilcoxon rank-sum test was used to compare for no differences in continuous variables by categorical variables. Fisher's exact test was used to compare association between categorical variables.
The four keys to choosing a model for the analysis of multiple events are: risk intervals, baseline hazard, risk set, and within-subject correlation . We choose a conditional model suggested by Prentice, Williams and Peterson , also referred to as the PWP model using gap time, common hazard, and restricted risk set. Gap time is the time from the previous event: the clock restarts after each event; for example, a hypothetical patient having three events at times 6, 9 and 16 is at risk of the first event during (0,6], and of the second and third events during (0,3] and (0,7], respectively. With a restricted risk set, contributions to the k th risk set is restricted to include only the k th event risk intervals of those subjects who have experienced (k − 1) events. Choice of the appropriate model was driven mainly by the objectives of the study rather than statistical considerations although authors recommended the use of this model. We used the marginal approach to compute the robust variance using a ‘sandwich’ estimator although the naive variance estimates (unadjusted variance) provide similar results. To quantify the predictive effect of treatment changes in the model we created four time-dependent variables to capture all of the different kinds of change. We investigated separately changes to a combination of either: (i) NRTI only; (ii) NRTI + PI other than the original triple-drug regimen; (iii) the original triple-drug therapy (ZDV + 3TC + IDV); or (iv) NRTI + NNRTI + PI and NRTI + NNRTI. Time-dependent variables take the value 1 from the time of treatment changes and the value 0 otherwise. All reported P values are two-sided.
In December 1997 the main results of Trilège were based on 58 virological failures occurring among 279 randomized patients (8/92 in the triple-drug regimen, 29/93 in ZDV + 3TC, and 21/94 in ZDV + IDV) . Follow-up of all patients from randomization to 31 December 1997 was 154 person-years giving an annual incidence rate of virological failure of 0.16, 0.57, and 0.41 in the triple-drug regimen, ZDV + 3TC, and ZDV + IDV arms, respectively. From January 1998 to the end of the study, virological failure was observed in a further 25 patients of whom 23 were still receiving the randomized treatment (one patient in each of the less intensive regimens were not under the randomized treatment at time to failure). Annual incidence rates of failure for the latter period were 0.03, 0.28, and 0.10 in the triple-drug regimen, ZDV + 3TC, and ZDV + IDV arms respectively. Then, there were 83 patients experiencing a virological failure: 10 patients in the ZDV + 3TC + IDV, 46 patients in the ZDV + 3TC arm, and 27 patients in the ZDV + IDV arm. Log-rank tests confirmed the previous results (P < 0.001 and P = 0.002, respectively, for ZDV + 3TC and ZDV + IDV versus the triple-drug regimen). Baseline median viral load was 48 000 copies/ml in the 83 cases, with no statistical difference among randomized arms. At the time of failure (S1), the median viral load was significantly higher in the triple-drug regimen arm (16 000 copies/ml) than in the ZDV + 3TC arm (1300 copies/ml;P < 0.001) and the ZDV + IDV arm (3400 copies/ml;P = 0.03). The corresponding extent of viral rebound (HIV RNA at failure minus baseline value in log10 copies/ml) were −0.35, −1.49, and −1.13 log copies/ml in patients in the triple-drug regimen, ZDV + 3TC, and ZDV + IDV arms, respectively. Median viral loads (viral rebound) at S2 were 11 500 (−0.46 log copies/ml), 1800 (−1.35 log copies/ml), and 11 800 (−0.54 log copies/ml) copies/ml in the triple-drug regimen, ZDV + 3TC, and ZDV + IDV arms, respectively.
Considering all patients at month 18, whatever the treatment ultimately received, there was no statistical difference in the proportions of patients with less or more than 500 copies/ml for both pair-wise comparisons between less intensive combination therapy and the triple-drug regimen (Table 1). Among cases, there was a significantly higher proportion of patients recovering an HIV RNA titre < 500 copies/ml in the ZDV + 3TC arm than in the triple-drug maintenance arm (80% versus 25%;P = 0.002), while there is only a trend in the ZDV + IDV group (66% versus 25%;P = 0.14).
In cases, median delay between time to virological failure and time to modification of treatment, if any, was 12 weeks (interquartile range, 8.4–34 weeks) with no statistical difference among the randomized arms. Fig. 1 shows the Kaplan–Meier curve of the probability of being under the randomized treatment for each maintenance arms (log-rank test, P < 0.001). From December 1997 to the end of the study, follow-up of patients under the randomized treatment were 50.9, 13, and 9.7 person years corresponding to annual incidence rates of virological failures of 0.04, 1.23, and 0.51 in the triple-drug regimen, ZDV + 3TC, and ZDV + IDV arms, respectively. Follow-up of patients under alternative therapy were 10.4, 47.5, and 52.4 person years with annual incidence rates of 0, 0.02, and 0.02, respectively.
Table 2 summarizes the number of patients remaining or modifying their randomized treatment among cases and controls. Modification of treatment occurred in 80% (8/10) of cases receiving the triple-drug regimen, 93% (43/46) of cases receiving ZDV + 3TC, and 96% (26/27) of cases receiving ZDV + IDV. Among controls 11% (9/82), 70% (33/47) and 79% (53/67) of patients modified their treatment in the triple-drug therapy, ZDV + 3TC, and ZDV + IDV arms, respectively. Among controls modifying their treatment, 78% (7/9), 79% (26/33), and 98% (52/53) of patients had HIV RNA levels < 500 copies/ml in the visit prior to modification of treatment in the triple-drug regimen, ZDV + 3TC, and ZDV + IDV, respectively. Among patients modifying their treatment, 77% [(14+47)/(26+53)] of patients in the ZDV + IDV arm restarted the original triple-drug therapy compared with 45% [(18+16)/(43+33)] of patients in the ZDV + 3TC arm.
Over the 18-month follow-up, 81 (88%), 51 (55%), and 78 (83%) patients in the triple-drug regimen, ZDV + 3TC, and ZDV + IDV arms, respectively, used only the three original drugs prescribed in Trilège (Table 3). Patients in the ZDV + 3TC used significantly more drugs than patients in both IDV-containing regimens (Table 3, P < 0.001).
Multivariate modelling of multiple HIV RNA concentrations
Table 4 highlights the persistency of certain cases to maintain a plasma HIV RNA level > 500 copies/ml. For example, in the triple-drug regimen 60% of cases had at least three HIV RNA plasma measurements > 500 copies/ml after their first virological failure, compared with only 22% of cases in both of the less intensive treatment arms.
We first fitted to the data a series of univariate models including the following variables: randomized treatment (pair-wise comparison of both less intensive regimens versus the triple-drug regimen), baseline HIV RNA, baseline CD4 cell count, HIV RNA level at time of virological failure, viral load rebound at time of virological failure, and treatment modifications using a time-dependent variable taking the value zero before time to modification and one after modification. Patients in both of the less intensive treatment arms had a lower hazard ratios of experiencing a recurrent HIV RNA level > 500 copies/ml than patients in the triple-drug regimen (P = 0.05 and P = 0.01 in ZDV + 3TC and ZDV + IDV, respectively). Baseline characteristics were not statistically significant whereas both higher HIV RNA levels and greater extent of viral load rebound were associated with occurrence of recurrent virological events with P < 0.001 and P = 0.002, respectively. Modification of treatment to receive a combination of NRTI + PI or re-initiation of the triple therapy were associated with a lower hazard ratios of experiencing recurrent virological events:P = 0.03 and P = 0.11, respectively. All variables for which the univariate P value was < 0.20 were used in a step-wise regression to determine independent prognostic factors of recurrent virological events in a final multivariate model (Table 5). A 1.0 lower HIV RNA level in log10 copies/ml at time to first virological failure was associated with a lowering to 0.59 in the hazard ratio for recurrent virological events, i.e., a 41% reduction in the risk of recurrent events. Both modifications of treatment (NRTI + PI or ZDV + 3TC + IDV), had an independent significant beneficial effect on the risk of recurrent events in the final multivariate model.
These results confirmed that, in HIV-infected patients initiating antiretroviral therapy, the strategy of maintenance of less intensive antiviral therapy after 3 months of ZDV + 3TC + IDV induction therapy is less effective than continuation of triple-drug therapy to sustain a plasma HIV RNA titre < 500 copies/ml [1,2]. In both less intensive regimens, incidence rates of virological failure under the randomized treatment increased over time after the interim analysis carried out in December 1997 (from 0.57 to 1.23 in ZDV + 3TC, and from 0.41 to 0.51 in ZDV + IDV). In these latter groups, the large proportions of patients modifying their randomized treatment probably avoided an even larger number of failures.
In the triple-drug regimen, the incidence rate of failure observed under randomized therapy beyond the interim analysis was low. As shown recently by other authors, resistance to IDV is rarely observed in such patients , confirming and reinforcing our previous hypothesis that, in this group, virological failure was more probably associated with poor adherence . It seems that those patients were also at higher risk of subsequent failures even with new potent regimens, suggesting that they probably had chronic difficulties in adherence to therapy. We also confirmed that virological failure in the triple-drug regimen was associated with a larger extent of viral rebound compared with rebounds observed in the other two groups .
After 18 months of follow-up, whatever the treatment ultimately received, a large proportion (87%) of patients had an HIV RNA concentration < 500 copies/ml with no statistical difference between randomized arms. Among cases, less intensive maintenance therapy during the randomized part of the trial did not prevent patients from controlling viral replication beyond the time to first virological failure. Indeed, although the triple-drug regimen had the lowest incidence rate of failure, a lower proportion of cases in this group had a virological control at month 18 compared with the ZDV + 3TC group. Although patients randomized to this dual therapy arm had the highest rate of virological failure, 48% did not experience further virological rebound after the first failure and only 29% experienced two or more virological rebounds beyond the first failure, as compared with 80% of patients who failed on ZDV + 3TC + IDV. This has potential important implications and demonstrates that patients experiencing first virological failure on dual ZDV + 3TC therapy can in most cases achieve virological success when they are switched to a potent antiretroviral combination . The greater ease of salvaging the two-NRTI arm compared with the PI + two-NRTI arm emphasizes the importance of using a new class of drug, whenever possible, when changing therapy after virological failure. Although resistance mutations are usually not demonstrated in early failures under triple combination including IDV, they will ultimately develop after few weeks or months of viral replication . At the time of the study, few PI were available and a high rate of cross-resistance to IDV might therefore have been expected in patients harbouring IDV-resistant viruses . In addition, genotypic testing was not available, and treatment changes were made according to antiretroviral history and individual clinical expertise. With genotypic testing now available and having demonstrated a benefit for the selection of antiretroviral agents to be used when changing the regimen [12,13], it is likely that a smaller number of drugs would have been used.
Patients in the ZDV + IDV arm were more frequently put back on the original triple-drug regimen than patients in the ZDV + 3TC arm. This finding was especially true in control patients who sustained a low level of viral replication. Possible reasons for such a differential attitude could not be assessed within the trial but it can be hypothesized that in the setting of ZDV + IDV virological failure, the addition of 3TC was considered as intensifying the regimen and of potential benefit, as 3TC can restore, in part, susceptibility to ZDV . Indeed, appearance of the 3TC-associated mutation M184V resensitizes in vitro the virus to ZDV by significantly reducing pyrophosphorolysis and partially reversing the effect of ZDV-associated mutations . On the other hand, in virological failures on ZDV + 3TC, reintroduction of IDV without change of the nucleosides was less frequent as it could have been considered as an IDV monotherapy by many investigators . Indeed, resistance to PI emerges rapidly when these drugs are administered as monotherapy or as part of suboptimal regimens [4,17].
Among controls, reasons for modifying the randomized treatment were probably motivated by recommendations made by the scientific committee rather than based on patient's individual viral load as a high proportion of these patients had an HIV RNA level < 500 copies/ml before the visit inducing the modification of treatment. Because of this situation we cannot determine whether in patients with suboptimal antiretroviral regimen (i.e., dual therapy), it is better to change to a more potent regimen when viral load is still undetectable in order to avoid viral rebound or to wait until viral rebound occurs, in order to spare drugs and lessen treatment burden and inconvenience.
Randomization to a less intensive regimen of ZDV + 3TC or ZDV + IDV was not detrimental, because treatment modification, either to the original triple regimen or to a different regimen was successful. These findings were confirmed by the significant lower hazard ratios of recurrent virological rebounds in patients in both of the less intensive therapy arms in the univariate analysis. In the final model, recurrent virological rebound was independently predicted by the plasma viral load at the time of virological failure and by two distinct modifications of the randomized treatment to receive a combination of NRTI + PI. The absence of the indicator variable representing the randomized group in the final model is explained by the presence of the plasma HIV RNA level at failure. Indeed, this level was higher in the ZDV + 3TC + IDV arm than in the ZDV + IDV arm, and even more so than in the ZDV + 3TC arm; the effect of the randomized group is mediated in the model through the plasma HIV RNA level at failure.
The fact that after first virological failure further virological outcome is predicted by the level of plasma viral RNA at failure, and that ZDV + 3TC failures are associated with a low plasma viral load rebound, might have important implications for sequencing antiretroviral and long-term management of antiretroviral therapy. Indeed, we have shown previously  that despite almost constant presence of the M184V reverse transcriptase mutation, the level of viral rebound is low in cases of ZDV + 3TC failure.
Moreover, our results emphasize that treatment modifications should probably be made as soon as virological failure is documented, as waiting too long will increase the risk of accumulation of resistance mutations  in the face of continuous drug pressure together with a progressive increase of viral load. Also a higher viral replication rate might be associated with a higher risk of resistance mutation selection . The main information from this study is that in cases of virological failure due to a lack of antiretroviral potency (i.e. ZDV + 3TC or ZDV + IDV), achieving virological success in the medium-term, i.e., 18 months, is possible in almost all patients as long as treatment modification or intensification is done relatively early (to avoid emergence of predominant mutant viruses) and when the viral load is still relatively low, i.e., < 5000 copies/ml.
On the other hand, when the virological failure is due to poor adherence, as we demonstrated previously in the triple regimen of ZDV + IDV + 3TC , change of treatment is less likely to achieve virological success. In such a situation, the patient's education, compliance counselling and other interventions such as alternative simplified, although potent, treatment regimen, should be considered. Drug resistance testing at the time of virological failure should help to improve therapeutic strategy in these situations .
The drugs were kindly provided by Glaxo Wellcome and Merck Sharp & Dohme Chibret.
1. Pialoux G, Raffi F, Brun-Vezinet F. et al
. A randomized trial of three maintenance regimens given after three months of induction therapy with zidovudine, lamivudine, and indinavir in previously untreated HIV-1 infected patients. N Engl J Med 1998, 339: 1269–1276.
2. Havlir D, Marschner I, Hirsch M. et al
. Maintenance antiretroviral therapies in HIV-infected subjects with undetectable plasma HIV RNA after triple-drug therapy. N Engl J Med 1998, 339: 1261–1268.
3. Reijers M, Weverling G, Jurriaans S. et al
. Maintenance therapy after quadruple induction therapy in HIV-1 infected individuals: Amsterdam Duration of Antiretroviral Medication (ADAM) study. Lancet 1998, 352: 185–190.
4. Condra J, Schleif W, Blahy O. et al
. In vivo emergence of HIV-1 variants resistant to multiple proteases inhibitors. Nature 1995, 374: 569–571.
5. Molla A, Korneyeva M, Gao Q. et al
. Ordered accumulation of mutations in HIV protease confers resistance to ritonavir. Nature Med 1996, 2: 760–766.
6. Descamps D, Flandre P, Calvez V. et al
. Mechanisms of virologic failure in previously untreated HIV-invected patients from a trial of induction-maintenance therapy. JAMA 2000, 283: 205–211.
7. Kelly P, Lim L. Survival analysis for recurrent event data: an application to childhood infectious diseases. Stat Med 2000, 19: 13–33.
8. Prentice R, Williams B, Peterson A. On the regression analysis of multivariate failure time data. Biometrika 1981, 68: 373–379.
9. Gallego O, de Mendoza C, Perez-Elias MJ. et al
. Drug resistance in patients experiencing early virological failure under a triple combination including indinavir. AIDS 2001, 15: 1701–1706.
10. Lange J, Hill A, Ait-Khaled M, et al. Does the M184V mutation affect the efficacy of HAART? Evidence from sequential addition of protease inhibitors or abacavir to zidovudine/lamivudine or stavudine/lamivudine. Antiviral therapy
(Suppl 1):79–80 [abstract 116].
11. Kemper CA, Witt MD, Keiser PH. et al
. Sequencing of protease inhibitor therapy: insights from an analysis of HIV phenotypic resistance in patients failing protease inhibitors. AIDS 2001, 15: 609–615.
12. Durant J, Clevenbergh P, Halfon P. et al
. Drug-resistance genotyping in HIV-1 therapy; the VIRADAPT randomized controlled trial. Lancet 1999, 353: 2195–2199.
13. Baxter J, Mayers D, Wentworth D. et al
. A randomized study of antiretroviral management based on plasma genotypic antiretroviral resistance testing in patients failing therapy. AIDS 2000, 14: F83–F93.
14. Larder BA, Kemp SD, Harrigan PR. Potential mechanism for sustained antiretroviral efficacy of AZT–3TC combination therapy. Science 1995, 269: 696–699.
15. Götte M, Arion D, Parniak M. et al
. The M184V mutation in the reverse transcriptase of human immunodeficiency virus type 1 impairs rescue of chain-terminated DNA synthesis. J Virol 2000, 74: 3579–3585.
16. Gass RJA, Shugarts D, Young R. et al
. Emergence of dual resistance to zidovudine (ZDV) and lamivudine (3TC) in clinical HIV-1 isolates from patients receiving ZDV/3TC combination therapy. Antiviral Therapy 1998, 3: 97–102.
17. Condra J, Holder D, Schleif W. et al
. Genetic correlates of in vivo resistance to indinavir, a human immunodeficeincy virus type 1 protease inhibitor. J Virol 1996, 70: 8270–8276.
18. Nijhuis M, Schuurman R, de Jong D. et al
. Increased fitness of drug resistant HIV-1 protease as a result of acquisition of compensatory mutations during suboptimal therapy. AIDS 1999, 13: 2349–2359.
19. Condra JH, Petropoulos CJ, Ziermann R, Schleif WA, Shivaprakash M, Emini EA. Drug resistance and predicted virologic responses to human immunodeficiency virus type 1 protease inhibitor therapy. J Infect Dis 2000, 182: 758–765.
20. Hirsch MS, Brun-Vezinet F, D'Acquila RT. et al
. Antiretroviral drug resistance testing in adult HIV infection. Recommendations of an International AIDS Society USA Panel.
JAMA 2000, 238: 2417–2426.
Coordinating trial centre
INSERM SC10: J.-P. Aboulker, B. Bazin, M. Delagnes, P. Flandre, F. Hakim, A. Halley, M. Harel, S. Hazebrouck, J. Martins, V. Meiffrédy, A. Prieur, S. Rivet, S. Robert, Y. Saïdi , S. Semole, J.-M. Vauthier).
Participating clinical and virologic centres
Hôpital Bichat Claude-Bernard, Paris: P. Yeni, V. Joly, N. Meslem, A. Villemant, C. Leport, G. Chirio, Z. Eid, P. Longuet, W. Nouioua, C. Tournerie, J.-L. Vildé, G. Walckenaer; J.-P. Coulaud, O. Bouchaud, P. Campa, R. Landman, Ch. Longuet, S. Masson, S. Matheron, E. Bouvet, K. Hamidi, M.-H. Prevot, C. Ruggeri, F. Brun-Vezinet, D. Descamps, F. Damond. Hôpital Cochin, Paris: D. Séréni, G. Bayol, J. Krulik, D. Sicard, A. Dufils, L. Finkielsztejn, O. Zak Dit Zbar; A. Krivine. Hôpital de l'Hôtel Dieu, Paris: L. Marsal, A. Compagnucci. Hôpital Necker, Paris: J.-P. Viard, V. Jubault, C. Rabian, C. Rouzioux, M. Burgard. Hôpital Lariboisière, Paris: J.-M. Salord, E. Badsi, J. Cervoni, M. Diemer, V. Vincent, M.-C. Mazeron. Hôpital de l'Institut Pasteur, Paris: G. Pialoux, B. Dupont, S. Lasry, H. Poncelet, M.-P. Treilhou, J.-D. Poveda, P. Barbot. Hôpital Pitié-Salpétrière, Paris: C. Katlama, L. Baril, F. Bricaire, C. Duvivier, N. Ktorza, M.-A. Valantin, A. Coutellier, M. Bonmarchand, F. Larue, M. Levy-Soussan, V. Calvez, M. Mouroux. Hôpital Rothschild, Paris: W. Rozenbaum, N. Adda, Ph. Mariot, E. Zafarana. Hôpital Saint-Antoine, Paris: J. Frottier, M.-C. Meyohas, D. Bollens, M.-J. Soavi. Hôpital Saint-Louis, Paris: E. Oksenhendler, L. Gérard, J. Modaï, S. Fournier, V. Garrait, J. Goguel, J.-M. Molina, D. Ponscarme, M.-N. Sombardier, F. Ferchal. Hôpital André Mignot, Le Chesnay: J. Doll, P. Colardelle, M Harzic. Hôpital Antoine Béclère, Clamart: F. Boué, P. Colson, L. Keros, A. Lazizi. Hôpital Avicenne, Bobigny: M. Bentata, M. Attia, A. Mosnier, B. Jarousse, J.-P. Pathe, P. Deny. Hôpital Bicêtre, Le Kremlin Bicêtre: J.-F. Delfraissy, C. Goujard, D. Peretti, D. Lecointe, A. M'Badi. Hôpital d'Angers, Angers: J.-M. Chennebault, P. Fialaire, J. Loison, F. Lunel, C. Payan. Hôpital Saint-Jacques, Besançon: H. Gil; Ch. Drobacheff, J.-M. Bastien, C. Derancourt, D. Devred, Y. Bourezane, B. Hoen, J.-M. Estavoyer, D. Bettinger, A. Bassignot. CHR Pellegrin, Bordeaux: J.-Y. Lacut; J.-M. Ragnaud, N. Kharlova, D. Neau, H. Fleury, B. Dumon. Hôpital Saint-André, Bordeaux: Ph. Morlat, D. Lacoste. CHU Côte de Nacre, Caen: C. Bazin, P. Hazera, M. Six, R. Verdon, F. Freymuth, A. Vabret. Hôpital du Bocage, Dijon: P. Chavanet, M. Buisson, M. Grappin, L. Piroth, A. Waldner, P. Pothier, E. Kohli. CHD Les Oudairies, La Roche sur Yon: Ph. Perré. Hôpital Hôtel Dieu, Lyon: L. Cotte, Ph. Rougier, I. Schlienger, J. Ritter. Hôpital Sainte Marguerite, Marseille: J.-A. Gastaut, A. Azzedine, T. Dinh, M.-P. Drogoul, M. Orticoni, C. Tamalet. CHU Gui de Chauliac, Montpellier: J. Reynes, P. Andre, V. Baillat, J.-P. Benezech, C. Favier, R. Le Stum, M. Segondy. Hôpital de Brabois, Vandoeuvre les Nancy: Th. Lecompte, C. Amiel, C. Burty, F. Brel, Th. May; A. Le Faou. Hôpital Hôtel Dieu, Nantes: F. Raffi, C. Allavena, E. Billaud, F. Bani-Sadr, N. Denis, C. François, S. Leauté, V. Reliquet, N. Tournemine, S. Billaudel, S. Auger. Hôpital De L'Archet, Nice: P. Dellamonica, Ph. Clevenbergh, P. Pugliese, Rahelin Irina, J. Cottalorda. Hôpital Pontchaillou, Rennes: C. Arvieux, F. Andrieux, F. Cartier, C. Michelet, F. Souala, J.-Y. Guillo, A. Ruffault. Hôpital Charles Nicolle, Rouen: F. Borsa-Lebas, F. Caron, Y. Debab, G. Humbert, C. Buffet-Janvresse. CHRU de Strasbourg, Strasbourg: J.-M. Lang, G. Kempf, V. Krantz, M. Nicolle, M. Partisani, D. Rey; M.-P. Schmitt. Hôpital Purpan, Toulouse: P. Massip, A. Bicart-See, E. Bonnet, J.-C. Carraro, M. Obadia, J. Peceny; J. Puel, I. Lecuyer. Hôpital de Tourcoing, Tourcoing: Y. Mouton, S. Ajana, M. Valette, L. Bocket. Hôpital Bretonneau, Tours: P. Choutet, J.-M. Besnier, F. Bastides, E. Didier, F. Barin. Hôpital Pierre Zobda-Quitman, Fort de France: G. Sobesky, S. Abel, A. Cabie, G. Comlan-Mayaud, G. Hillion, C. Thomas, O. Bera, M. Ouka.