To evaluate factors associated with durable viral suppression (DVS), low viral rebound (virus load of 500–5000 copies/mL), and high viral rebound (virus load above 5000 copies/mL), we studied 3736 patients who achieved an undetectable virus load (i.e., <500 copies/mL) while receiving an initial highly active antiretroviral therapy regimen containing a protease inhibitor (PI). A total of 2636 patients (71%) had DVS, 387 (10%) had low viral rebound, and 713 (19%) had high viral rebound. Factors associated with DVS were antiretroviral-naive status, choice of PI (indinavir or boosted PI), lower virus load and higher CD4+ cell count, shorter duration of therapy (<6 months), and larger CD4+ cell increment until an undetectable virus load was attained. These factors (except for receipt of indinavir) were also associated with low versus high viral rebound. Compared with patients with DVS, patients with high viral rebound were more likely to experience clinical failure (adjusted hazard ratio [aHR], 1.71; 95% confidence interval [CI], 0.97–3.02) and immunologic failure (aHR, 1.57; 95% CI, 1.34–1.83), and patients with low viral rebound were as likely to experience these types of failure (aHR for clinical failure, 1.18 [95% CI, 0.48–2.92]; aHR for immunologic failure, 1.07 [95% CI, 0.87–1.32]), suggesting that a low viral rebound while receiving HAART that contains a PI has no significant consequence on midterm clinical outcome.
Since the advent of potent antiretroviral combination therapies, it has been recognized that there is need for durable suppression of HIV replication, to obtain both optimal immunologic efficacy and a reduction in long-term HIV-related morbidity and mortality [1–4]. With this aim, some guidelines have recommended widespread use of potent antiretroviral drug combinations . The hope was that HIV replication would be completely suppressed by HAART and that the use of potent combinations would prevent the emergence of drug-resistant strains. However, more-recent recommendations argue for more-conservative approaches, with initiation of antiretroviral therapy recommended for patients with CD4+ cell counts of 200–350 × 106 cells/L and/or plasma HIV RNA levels of 50,000–100,000 copies/mL [5–7]. Indeed, the occurrence of antiretroviral resistance and drug-related toxicity have limited therapeutic options for a growing number of patients . In cohort studies [9, 10] and in clinical practice, some patients who have persistently detectable plasma HIV RNA levels while receiving HAART nonetheless have stable or improved CD4+ cell counts and no clinical progression of infection, which suggests that permanent undetectability of the virus (i.e., a plasma HIV RNA level less than the standard detection threshold value) may not always be essential for obtaining durable immunologic and clinical responses. Thus, identification of factors associated with the degree of antiviral efficacy after initial complete suppression of viral replication in HAART recipients, as well as determination of the corresponding clinical and immunologic outcomes, could be helpful in patient treatment.
We studied patients in whom the HIV load became undetectable during an initial course of HAART, to identify factors associated with the degree of viral rebound (none, low, or high) during the subsequent year. We also analyzed midterm clinical and immunologic outcomes according to the degree of antiviral efficacy.
Patients. The French Hospital Database on HIV is a large, prospective cohort study of HIV-infected patients who are aged ⩾15 years and who receive treatment at 68 French university hospitals. The enrollment criteria are documented HIV-1 or HIV-2 infection and provision of written informed consent. Trained research assistants use the French Ministry of Health DMI2 software to collect clinical and biological data (including plasma virus load measurements since July 1996) prospectively at study inclusion and at each visit or hospital admission for an HIV-related clinical event or a new treatment prescription, or at least every 6 months, on standardized forms.
We studied participants in the French Hospital Database on HIV who started receiving their first protease inhibitor (PI)–containing HAART regimen (i.e., regimen including ⩾3 antiretroviral drugs) from 1 January 1997 through 31 December 1998, who had no previous exposure to PIs or nonnucleoside reverse-transcriptase inhibitors (NNRTIs), who had not previously been included in a blinded clinical trial of antiretroviral therapy, and in whom the plasma HIV RNA level decreased to <500 copies/mL on at least 1 occasion during HAART. This threshold value of 500 copies/mL was chosen to overcome differences in the detection levels of the assays used at the different participating centers. Inclusion in the study was restricted to patients who started HAART on or after 1 January 1997, to avoid selecting long-term survivors . The date of inclusion in the study was the first date that a plasma HIV RNA assay revealed a value of <500 copies/mL while the patient was receiving HAART. Patients were excluded if CD4+ cell count or plasma HIV RNA data were unavailable within 3 months before initiation of HAART, if the CD4+ cell count was unavailable at the date of inclusion, or if the patient did not have ⩾1 follow-up visit in which the plasma HIV RNA level was assayed every 3 months during the year after inclusion. Patients initially prescribed dual-PI therapy (other than saquinavir hard-gel capsule [HGC] plus ritonavir [23 patients] or an NNRTI [107 patients]) were also excluded. The cutoff date for database analysis was 31 December 1999. A total of 3736 patients met these criteria.
Statistical analysis. We sought to identify factors associated with 3 levels of antiretroviral efficacy during the year after initial virus load undetectability while the patient was receiving HAART—namely, persistent undetectability (all plasma HIV RNA values were <500 copies/mL, or a single value was ⩾500 copies/mL), low viral rebound (at least 2 plasma HIV RNA values were 500–5000 copies/mL), and high-level viral rebound (at least 2 plasma HIV RNA values were >500 copies/mL, and at least 1 value was >5000 copies/mL).
We first compared patients with persistent undetectability and patients with viral rebound. Baseline predictors of persistent undetectability were identified using univariate and multivariate logistic regression models. All variables associated with persistent undetectability in univariate analysis (P < .20) were included in the multivariate model. Then, to identify factors associated with each level of viral rebound, groups of patients who had low and high viral rebound were compared with the group of patients who had persistent undetectability, and patients who had low viral rebound were compared with patients who had high viral rebound. All comparisons were based on multivariate logistic regression models.
Kaplan-Meier life tables were used to evaluate clinical and immunologic outcomes according to the 3 categories of antiretroviral efficacy after initial virus load undetectability. Clinical failure was defined as the occurrence of a new AIDS-defining event or death. Immunologic failure was defined as a decrease in the CD4+ cell count of ⩾20% from the value recorded 12 months after inclusion. Because levels of antiretroviral efficacy were defined according to the plasma HIV RNA values during the first year after initial virus undetectability during HAART, the time to clinical or immunologic failure was measured from month 12 after inclusion. Hazard ratios for clinical and immunologic failure were estimated with the Cox model after adjustment for other prognostic factors (age, transmission group, previous AIDS-defining events, CD4+ cell count and plasma HIV RNA level at HAART initiation, previous antiretroviral therapy, associated nucleoside reverse-transcriptase inhibitor [NRTI] combination, PI chosen at HAART initiation, time from HAART initiation to initial undetectable virus load, and CD4+ cell count increment between HAART initiation and initial undetectable virus load). Data were analyzed using SAS statistical software, version 6.12 (SAS Institute).
Patient population. The first PI prescribed was indinavir for 1728 (46.3%) of 3736 patients; nelfinavir, for 795 (21.3%); saquinavir HGC, for 696 (18.6%); ritonavir, for 391 (10.5%); and saquinavir HGC plus ritonavir, for 126 (3.4%) (table 1). At HAART initiation, 1540 patients (41.2%) were antiretroviral naive. The other 2196 patients (58.8%) had received antiretroviral therapy for a median of 19 months (interquartile range [IQR], 9–37 months). The median duration of follow-up after the date of inclusion (i.e., initial undetectable virus load) was 21 months (IQR, 16–25 months).
During the year after inclusion, 2636 patients (70.6%) had persistently undetectable virus loads, as defined in Methods. Of these patients, 555 (21.1%) had a single plasma HIV RNA value of ⩾500 copies/mL. A viral rebound was observed in the remaining 1100 patients (29.4%); low viral rebounds occurred in 387 patients (10.4%), and high viral rebounds occurred in 713 patients (19.0%).
Factors associated with the degree of antiretroviral efficacy after initial undetectable virus load. Table 2 shows patient characteristics, according to virologic outcome, during the year after their initial undetectable virus load while receiving HAART. The results of multivariate logistic regression analyses of persistent undetectability are also shown in table 2. After adjustment, the likelihood of persistent undetectability was lower in patients with AIDS and patients with higher plasma RNA levels at HAART initiation, and it was higher in patients with higher CD4+ cell counts. Compared with antiretroviral-naive patients, the likelihood of persistent undetectability was 50% lower for patients who had already received 1 or 2 NRTIs before HAART initiation, and it was 66% lower for patients who had already received >2 NRTIs. Compared with patients who were prescribed indinavir, those who were prescribed saquinavir HGC, nelfinavir, or ritonavir had a lower likelihood of persistent undetectability, whereas patients who were prescribed saquinavir HGC plus ritonavir had the same likelihood of persistent undetectability as patients who were prescribed indinavir. Late occurrence of virus load undetectability (>6 months after HAART initiation) was associated with a lower likelihood of persistent undetectability, whereas a larger CD4+ cell count increment after HAART initiation was associated with a higher likelihood of persistent undetectability.
Factors associated with low and high viral rebound are shown in table 3. Factors associated with a lower likelihood of persistent undetectability in previous analyses (diagnosis of AIDS, lower CD4+ cell count and higher plasma HIV RNA level at HAART initiation, previous antiretroviral therapy, late viral suppression, and no CD4+ cell count increment) were associated with high viral rebound. In contrast, treatment with saquinavir HGC, nelfinavir, or ritonavir (associated with a higher risk of viral rebound than was treatment with indinavir; see above) was predictive of low viral rebound (the risk of high viral rebound was 39%, 45%, or 34% lower, respectively, than the risk of low viral rebound), showing that viral rebound occurring in patients treated with indinavir was more likely to be high level than low level.
Similar results were obtained when we excluded the 555 patients who were categorized as having persistent virus load undetectability but who had a single plasma HIV RNA value of ⩾500 copies/mL during the first year of HAART. Similar results were also obtained when treatment-naive and treatment-experienced patients were analyzed separately (data not shown).
Clinical and immunologic progression. Changes in plasma HIV RNA levels and CD4+ cell counts in the 3 groups during the 27 months of follow-up are shown in figure 1. Values were compared with baseline (inclusion) values.
As shown in figure 1 and table 4, most patients with persistently undetectable virus loads had plasma HIV RNA levels of <500 copies/mL at all times throughout follow-up. Likewise, most patients with low viral rebound during the first year of follow-up subsequently had low-level or undetectable viral replication, and most patients with high viral rebound subsequently had high-level viral replication.
As expected, the highest median CD4+ cell count increment occurred in patients with persistent virus load undetectability, and the lowest median increment occurred in patients with high viral rebound (figure 1). Among patients with available CD4+ cell count measurements at 27 months after inclusion, the median change relative to baseline was +163 × 106 cells/L (IQR, 38–282 × 106 cells/L) for the 720 patients with persistent undetectability, +120 × 106 cells/L (IQR, 1–247 × 106 cells/L) for the 113 patients with low viral rebound, and +59 × 106 cells/L (IQR, -28 to 163 × 106 cells/L) in the 182 patients with high viral rebound (for persistent undetectability vs. low viral rebound, P = .03; for persistent undetectability vs. high viral rebound, P = .001; and for low viral rebound vs. high viral rebound, P = .01; all by Mann-Whitney nonparametric test). As shown in figure 1, median CD4+ cell counts continued to increase throughout follow-up among patients who had high viral rebound during the first 12 months after initial virus load undetectability.
After month 12, 1041 patients either had a CD4+ cell count decrease or experienced a clinical event: 977 patients had only a CD4+ cell count decrease, 38 had only a clinical event, and 26 had both. After month 12, the 15-month probabilities of clinical failure were 2.4% (95% CI, 1.7%–3.4%) among patients with persistent undetectability, 2.2% (95% CI, 0.9%–5.3%) among patients with low viral rebound, and 5.8% (95% CI, 3.6%–9.2%) among patients with high viral rebound. The corresponding 15-month probabilities of immunologic failure were 37.2% (95% CI, 34.5%–40.0%), 39.7% (95% CI, 33.2%–46.6%), and 51.9% (95% CI, 46.5%–57.3%), respectively (figure 2). As shown in figure 2, hazard ratios for clinical and immunologic failure were similar among patients with persistent undetectability and patients with low viral rebound, and they were significantly higher among patients with high viral rebound. Similar results were obtained when treatment-naive and treatment-experienced patients were analyzed separately (data not shown).
Factors linked to persistent virus load undetectability and to low-level viral rebound (compared with high viral rebound) in this study were AIDS-free status, higher CD4+ cell count and lower plasma HIV RNA level at HAART initiation, no antiretroviral drug exposure before HAART, a rapid virologic response to HAART, and a larger CD4+ cell count increment between HAART initiation and initial virus load undetectability. With regard to the choice of PI, indinavir and the saquinavir HGC/ritonavir combination were both associated with persistent undetectability or, when viral rebound occurred, with a high-level rebound. The median CD4+ cell count continued to increase, whatever the secondary virologic response, with the smallest increment in patients with high viral rebound. Clinical and immunologic outcomes were similar in patients with persistent virus load undetectability and patients with low viral rebound, but they were poorer in patients with high viral rebound.
Like all observational studies, ours may have suffered from biases, such as exclusion of patients for whom plasma HIV RNA levels were not measured every 3 months during the first year of HAART. Thus, data for patients who were lost to follow-up, patients who died, and patients who may not have attended outpatient visits because of slow disease progression were not analyzed, and this may have led us to underestimate or overestimate the risk of clinical failure. Other observational studies [9, 10, 12] have suggested that many AIDS-defining events occur within the first months of HAART and are not associated with therapeutic failure but with the clinical expression of underlying diseases. Our evaluation of clinical outcome after 12 months of HAART permitted us to identify failures of HAART itself.
Information was lacking in this analysis on possible switches in antiretroviral therapy, adherence to antiretroviral therapy, and preexisting nucleoside analogue resistance in antiretroviral-experienced patients. However, despite these limitations, the large size of the study cohort would limit this weakness. Furthermore, the analysis was conducted on an intent-to-treat basis, to control for the fact that some patients could have had several switches in antiretrovirals during the study.
Undetectable virus load was defined by a cutoff of 500 copies/mL for this analysis, and it is conceivable that different results would have been obtained with a lower threshold. Likewise, a cutoff of 5000 copies/mL was arbitrarily chosen to distinguish low from high viral rebound. However, when the analysis was repeated with a cutoff value of 10,000 copies/mL, patients with low viral rebound had an intermediate clinical and immunologic prognosis.
Clinical stage of infection and CD4+ cell count and plasma HIV RNA level at HAART initiation were independent predictors of persistent virus load undetectability in this study, as elsewhere , and these factors were also predictive of low viral rebound in patients in whom viral rebound occurred. Furthermore, earlier suppression of viral replication during HAART was predictive of better midterm virologic outcome.
A CD4+ cell count increment of >50% between the time of HAART initiation and the first assay result showing undetectable virus load was associated with persistent undetectability and with low viral rebound in patients in whom rebound occurred, after adjustment for other prognostic factors. It is possible that patients who had unsatisfactory immunologic responses, despite initial virus load undetectability during HAART, had low residual levels of circulating virus . In such patients, a relatively small increase in the CD4+ cell count could be explained by residual viral replication and could thus be associated with higher viral rebound.
Patients treated with indinavir and those treated with the saquinavir HGC/ritonavir combination were more likely to have persistently undetectable virus load, whereas those treated with saquinavir HGC, ritonavir, or nelfinavir were more likely to have low viral rebound when rebound occurred. The HGC formulation of saquinavir is less effective than are other PIs at driving virus load to less than the detection limit [9, 10, 13, 15, 16] because of its lower bioavailability . In this observational study, nelfinavir and ritonavir were also less effective than indinavir at maintaining virus loads of less than the detection limit, and the combination of saquinavir HGC and ritonavir was as effective as indinavir. These results might be explained by differences in intrinsic efficacy, by immediate or late intolerance reactions, or by poor adherence . When patients experienced a viral rebound after receiving a regimen containing indinavir or a combination of saquinavir and ritonavir, the rebound was more likely to be high level than low level. This supports findings of differences in antiviral potency among available PIs or poorer tolerance leading to complete interruption of HAART reported elsewhere .
The median CD4+ cell count continued to increase, regardless of the secondary virologic response, but the smallest increment was observed in patients with high viral rebounds. Other studies have shown that, despite persistently detectable viral replication during HAART, most patients have at least a midterm increase in the CD4+ cell count .
Clinical and immunologic outcomes were similar among patients with persistent virus load undetectability and patients with low viral rebound, but they were poorer among patients with high viral rebound. Although some studies have suggested that a durable virologic response is associated with better clinical  or immunologic  outcome, no statistical difference in clinical progression was found between patients who experienced a viral rebound and patients whose virus load remained less than the detection limit, even after adjustment for age and baseline CD4+ cell count . Other studies have reported the association between CD4+ cell count (but not virus load) at HAART initiation and survival or between CD4+ cell count and AIDS-free survival in treatment-naive patients [20, 21]. Our study shows the benefit, in terms of clinical and immunologic progression, not only of persistent virus load undetectability, but also of low viral rebound when rebound occurs, after adjustment for numerous prognostic factors, and it shows that patients with high viral rebound are particularly at risk of clinical progression. Whether these results are linked to a difference in viral fitness  or in host immunologic response against HIV between the 2 levels of viral rebound remains to be evaluated.
Although it has been suggested that immunologic recovery depends on the amplitude and duration of virus load suppression, there is controversy about the degree of viral suppression needed to improve CD4+ T cell function . Our study suggests that low viral rebound is compatible with satisfactory midterm immunologic outcome. Of interest, when the cutoff used to distinguish low from high viral rebound was raised from 5000 to 10,000 copies/mL, patients with low viral rebound had a poorer clinical and immunologic prognosis than did patients with a persistently undetectable virus load, and patients with high viral rebound had a much poorer prognosis (data not shown).
In cases of treatment failure involving indinavir-containing HAART regimens, the viral rebound mostly corresponds to wild-type protease and wild-type reverse transcriptase (except for the M184V mutation in patients with lamivudine therapy) at the time of the first rebound [18, 23, 24]. The selection of specific PI resistance mutations is more frequently detected at the time of first viral rebound for patients in whom nelfinavir-containing HAART fails . When viral rebound persists, viral mutations develop progressively . However, it has been reported that new, additional, major PI-associated mutations are detected in fewer than one-third of patients with low viral rebound . Thus, among patients with low viral rebound who are receiving PI-containing HAART, who have difficulties with treatment adherence, and who have previously been exposed to multiple antiretroviral regimens, delays in treatment switches may spare some therapeutic options for future use. However, this strategy should be evaluated with regard to long-term outcome and viral acquisition of drug resistance.
The Clinical Epidemiology Group of The French Hospital Database on HIV
Scientific committee. Dr. S. Alfandari, Dr. F. Bastides, Dr. E. Billaud, Dr. A. Boibieux, Prof. F. Boué, Prof. F. Bricaire, D. Costagliola, Dr. L. Cotte, Dr. L. Cuzin, Prof. F. Dabis, Dr. P. Enel, Dr. S. Fournier, Dr. J. Gasnault, Dr. C. Gaud, Dr. J. Gilquin, Dr. S. Grabar, Dr. D. Lacoste, Prof. J. M. Lang, Dr. H. Laurichesse, Dr. P. Leclercq, Prof. C. Leport, M. Mary-Krause, Prof. S. Matheron, Prof. M. C. Meyohas, Dr. C. Michelet, Dr. J. Moreau, Prof. G. Pialoux, Dr. I. Poizot-Martin, Dr. C. Pradier, Dr. C. Rabaud, Prof. E. Rouveix, Prof. P. Saïag, Prof. D. Salmon-Ceron, Prof. J. Soubeyrand, and Dr. H. Tissot-Dupont.
DMI2 coordinating center. French Ministry of Health (Dr. B Haury, Dr. V. Tirard-Fleury, and I. Tortay).
Statistical data analysis center. INSERM EMI 0214 (Dr. S. Abgrall, D. Costagliola, Dr. S. Grabar, E. Lanoy, L. Lièvre, M. Mary-Krause, and V. Potard).
Paris-area Centre d'information et de soins sur l'immunodéficience humaine (CISIH). CISIH de Bichat—Claude Bernard (Hôpital Bichat—Claude Bernard: Prof. S. Matheron, Prof. J. P. Coulaud, Prof. J. L. Vildé, Prof. C. Leport, Prof. P. Yeni, C. Mandet, Prof. E. Bouvet, C. Gaudebout, Prof. B. Crickx, and Dr. C. Picard-Dahan), CISIH de Paris-Centre (Hôpital Broussais: Prof. L. Weiss and D. Tisne-Dessus; G. H. Tarnier-Cochin: Prof. D. Sicard and Prof. D. Salmon; and Hôpital Saint-Joseph: Dr. J. Gilquin and Dr. I. Auperin), CISIH de Paris-Ouest (Hôpital Necker adultes: Dr. J. P. Viard and Dr. L. Roudière; Hôpital Laennec: Dr. W. Lowenstein; and Hôpital de l'Institut Pasteur), CISIH de Paris-Sud (Hôpital Antoine Béclère: Prof. F. Boué and Dr. R. Fior; Hôpital de Bicêtre: Prof. J. F. Delfraissy and Dr. C. Goujard; Hôpital Henri Mondor: Dr. Ph. Lesprit and C. Jung; and Hôpital Paul Brousse), CISIH de Paris-Est (Hôpital Rothschild: Prof. W. Rozenbaum and Dr. G. Pialoux; Hôpital Saint-Antoine: Prof. M. C. Meyohas, Dr. J. L. Meynard, Dr. O. Picard, and N. Desplanque; Hôpital Tenon: Prof. J. Cadranel and Prof. C. Mayaud), CISIH de Pitié-Salpétrière (GH Pitié-Salpétrière: Prof. F. Bricaire, Prof. C. Katlama, Prof. S. Herson, and Dr. A. Simon), CISIH de Saint-Louis (Hôpital Saint-Louis: Prof. J. M. Decazes, Prof. J. M. Molina, Prof. J. P. Clauvel, and Dr. L. Gerard; GH Lariboisière—Fernand Widal: Dr. J. M. Salord and Dr. Diermer), CISIH 92 (Hôpital Ambroise Paré: Dr. C. Dupont, H. Berthé and Prof. P. Saïag; Hôpital Louis Mourier: Dr. E. Mortier and C. Chandemerle; and Hôpital Raymond Poincaré: Dr. P. de Truchis), and CISIH 93 (Hôpital Avicenne: Dr. M. Bentata and P. Berlureau; Hôpital Jean Verdier: J. Franchi and Dr. V. Jeantils; and Hôpital Delafontaine: Dr. Mechali and B. Taverne).
CISIH, outside Paris area. CISIH Auvergne-Loire (Centre Hospitalier Universitaire [CHU] de Clermont-Ferrand: Dr. H. Laurichesse and Dr. F. Gourdon; and Centre Hospitalier Régional Universitaire [CHRU] de Saint-Etienne: Prof. F. Lucht and Dr. A Fresard); CISIH de Bourgogne-Franche Comté (CHRU de Besançon; CHRU de Dijon; and CH de Belfort: Dr. J. P. Faller and P. Eglinger); CISIH de Caen (CHRU de Caen: Prof. C. Bazin and Dr. R Verdon), CISIH de Grenoble (CHU de Grenoble), CISIH de Lyon (Hôpital de la Croix-Rousse: Prof. D. Peyramond and Dr. A. Boibieux; Hôpital Edouard Herriot: Prof. J. L. Touraine and Dr. J. M. Livrozet; Hôtel-Dieu: Prof. C. Trepo and Dr. L. Cotte; and CH de Lyon-Sud: Médecine Pénitentiaire: Dr. P. Barlet), CISIH de Marseille (Hôpital de la Conception: Dr. I. Ravaux and Dr. H Tissot-Dupont; Hôpital Houphouët-Boigny: Dr. J. Moreau; Institut Paoli Calmettes; Hôpital Sainte-Marguerite: Prof. J. A. Gastaut, Dr. I. Poizot-Martin, Prof. J. Soubeyrand, and Dr. F. Retornaz; Hotel-Dieu; CHG d'Aix-En-Provence; Centre pénitentiaire des Baumettes; CH d'Arles; CH d'Avignon: Dr. G. Lepeu; CH de Digne Les Bains: Dr. P. Granet-Brunello; CH de Gap: Dr. L. Pelissier and Dr. J. P. Esterni; CH de Martigues: Dr. M. Nezri and Dr. R. Cohen-Valensi; and CHI de Toulon), CISIH de Montpellier (CHU de Montpellier: Prof. J. Reynes; and CHG de Nîmes), CISIH de Nancy (Hôpital de Brabois: Prof. T. May and Dr. C. Rabaud), CISIH de Nantes (CHRU de Nantes: Prof. F. Raffi and Dr. E. Billaud), CISIH de Nice (Hôpital Archet 1: Dr. C. Pradier and Dr. P. Pugliese; and CHG Antibes Juan les Pins), CISIH de Rennes (CHU de Rennes: Prof. C. Michelet and Dr. C. Arvieux), CISIH de Rouen (CHRU de Rouen: Prof. F. Caron and Dr. F. Borsa-Lebas), CISIH de Strasbourg (CHRU de Strasbourg: Prof. J. M. Lang and Dr. D. Rey; and CH de Mulhouse), CISIH de Toulouse (CHU Purpan: Prof. P. Massip, Dr. L. Cuzin, Prof. E. Arlet-Suau, and Dr. M. F. Thiercelin Legrand; Hôpital la Grave; and CHU Rangueil), CISIH de Tourcoing-Lille (CH Gustave Dron; and CH de Tourcoing: Dr. S. Alfandari and Dr. Y. Yasdanpanah), and CISIH de Tours (CHRU de Tours and CHU Trousseau).
CISIH overseas. CISIH de Guadeloupe (CHRU de Pointe-à-Pitre), CISIH de Guyane (CHG de Cayenne: Dr. M. Sobesky and Dr. R. Pradinaud), CISIH de Martinique (CHRU de Fort-de-France), and CISIH de La Réunion (CHD Félix Guyon: Dr. C. Gaud and Dr. M. Contant).
We thank all participants in the French Hospital Database on HIV and, especially, the research assistants, without whom this work would not have been possible.