Share this article on:

Modeling the Time Course of CD4 T-Lymphocyte Counts According to the Level of Virologic Rebound in HIV-1–Infected Patients on Highly Active Antiretroviral Therapy

Kousignian, Isabelle; Abgrall, Sophie; Duval, Xavier; Descamps, Diane; Matheron, Sophie; Costagliola, DominiqueClinical Epidemiology Group of the French Hospital Database on HIV

JAIDS Journal of Acquired Immune Deficiency Syndromes: September 1st, 2003 - Volume 34 - Issue 1 - p 50-57
Clinical Science

Objective To study the influence of the level of virologic rebound during combination antiretroviral therapy on the time course of the CD4 count.

Methods Between January 1997 and December 1999, we enrolled 3736 patients from the French Hospital HIV Database who had an undetectable viral load on a first course of highly active antiretroviral therapy (HAART). Four levels of virologic rebound were defined on the basis of viral load values during the year following initial undetectability on HAART: group 1, all viral loads <500 copies/mL; group 2, all viral loads <5000 copies/mL; group 3, all viral loads <10,000 copies/mL; and group 4, at least 1 viral load >10,000 copies/mL. We developed a continuous time-homogeneous Markov process with 5 reversible stages defined by CD4 count intervals.

Results CD4 counts increased continuously over time in each group. The smaller the virologic rebound, the stronger was the increase in the CD4 count (P < 0.0001). The mean CD4 cell count increments between months 2 and 6 were 26, 20, 11, and 2 cells/mm3 in groups 1, 2, 3, and 4, respectively. The rate of gain fell after month 6 and was almost nil in group 4.

Conclusion After achieving an undetectable viral load on HAART, immunologic reconstitution is possible whatever the subsequent level of viral replication, except among patients with high-level rebound, meaning that in patients with a long history of antiretroviral therapy and a reduced choice of antiretroviral drugs due to acquisition of resistances, delay in antiretroviral therapy switch can be possible in patients with low or intermediate rebound.

From the INSERM EMI 0214, Paris (Drs Kousignian, Abgrall, and Costagliola); and Services des Maladies Infectieuses et Tropicales A (Drs Abgrall and Matheron), Services des Maladies Infectieuses et Tropicales B (Dr Duval), and Service de virologie, Hôpital Bichat Claude Bernard, Paris, France (Dr Descamps).

Received for publication March 11, 2003; accepted June 16, 2003.

The French Hospital Database on HIV is supported by the Institut National de la Santé et de la Recherche Médicale (INSERM), the Agence Nationale de la Recherche sur le Sida (ANRS), the Fondation pour la Recherche Médicale, and the French Ministry of Health.

Presented in part at the 23rd Annual Conference of the International Society for Clinical Biostatistics (ISCB), Dijon, France, September 9–13, 2002 (presentation P67) and at the 10th Conference on Retroviruses and Opportunistic Infections (CROI 2003), Boston, Massachusetts, February 10–14, 2003 (abstract O-12).

Written informed consent was obtained from patients enrolled in the French Hospital Database on HIV.

Address correspondence and reprint requests to Isabelle Kousignian, INSERM EMI 0214, 56 Boulevarde Vincent Auriol, B.P. 335, 75625 Paris Cedex 13, France. (e-mail:

The principal factor associated with a good immunologic and clinical response to highly active antiretroviral therapy (HAART) is effective and durable viral suppression. 1–5 In cohort studies, 6,7 however, some patients have persistently detectable plasma HIV-1 RNA, whereas on HAART, they nonetheless have stable or increasing CD4 cell counts and no clinical progression. Better knowledge of immunologic outcome associated with different levels of viral replication after initial undetectability on HAART could help to manage patients with a discrepancy between their immunologic and virologic responses, especially when antiretroviral resistance and drug-related toxicity limit therapeutic options. Indeed, it is interesting to evaluate the level of virologic rebound when a switch in antiretroviral therapy is necessary to preserve a satisfactory clinical and immunologic outcome in patients with a reduced choice of antiretroviral drugs.

In an epidemiologic setting, multistate Markov models are useful for longitudinal studies describing an infectious disease process over time through different stages of infection. 8,9 These models are particularly suited to analyze ordered clinical processes and can be applied to estimating and predicting rates of HIV disease progression. Markov models are now extensively used in the analysis of HIV infection. 10–14 The specificities of these models, and their inherent difficulties, mainly reside in the definition of disease stages and stage transitions. 15–17

The aim of this study was to evaluate the influence of the level of virologic rebound in patients who previously achieved a viral load below 500 copies/mL during HAART on the dynamics of the CD4 T-lymphocyte count. We used a multistate Markov model to compare HIV disease progression based on CD4 cell count states of each level of virologic rebound.

Back to Top | Article Outline


Study Population


The French Hospital Database on HIV (FHDH) is a large prospective cohort study of HIV-infected patients aged at least 15 years and managed in 68 French university hospitals. The enrollment criteria are documented HIV-1 or HIV-2 infection and written informed consent. The FHDH methodology has been described in detail elsewhere. 18

Back to Top | Article Outline


This study focused on HIV-1–infected FHDH patients in whom the plasma HIV-1 RNA level became undetectable (<500 copies/mL) on at least 1 occasion during a first HAART regimen (at least 3 antiretroviral drugs, including a protease inhibitor [PI]) between January 1, 1997 and December 31, 1998, and who were monitored for at least 12 months after this initial undetectability. Patients were not eligible if they had previously been exposed to PIs or to nonnucleoside reverse transcriptase inhibitors (NNRTIs). We also put aside NNRTI-treated patients (n = 107), because drug resistance seems to occur more rapidly with NNRTI therapy and because of their limited number. 19

Patients were excluded if CD4 cell count or plasma HIV-1 RNA values were unavailable in the 3 months preceding HAART initiation or at the date of initial HIV RNA undetectability, or if they did not attend follow-up visits (with plasma HIV-1 RNA assay) every 3 months during the year following initial undetectability. All enrolled patients had at least 2 CD4 cell counts during the 12-month study period. Moreover, patients who were initially prescribed regimens containing 2 PIs other than the saquinavir hard-gel capsule (hgc) and ritonavir were also excluded because of the low number of concerned patients (n = 23). The cutoff date for database analysis was December 31, 1999. A total of 3736 patients met these criteria.

To study the impact of the level of virologic suppression on the immune response, we modeled CD4 cell count dynamics using a Markov model that takes into account different levels of virologic rebound. Four levels were defined according to plasma HIV-1 RNA values during the year following initial undetectability on HAART:

  • Group 1: “durable virologic suppression” (all plasma HIV-1 RNA values <500 copies/mL, but a single value above 500 copies/mL was tolerated)
  • Group 2: “low-level rebound” (all values <5000 copies/mL, with at least 2 values above 500 copies/mL)
  • Group 3: “intermediate-level rebound” (all values <10,000 copies/mL, with at least 2 values above 500 copies/mL and at least 1 value above 5000 copies/mL)
  • Group 4: “high-level rebound” (at least 2 values above 500 copies/mL, with at least 1 value above 10,000 copies/mL)

The viral load threshold value of 500 copies/mL was the highest detection limit of the different assays used in participating centers during the study period. The threshold values of 5000 and 10,000 copies/mL were the results of a first study on this database. 20 Indeed, a cutoff of 5000 copies/mL was arbitrarily chosen to distinguish low-level from high-level viral rebound. When the analyses were repeated with a cutoff of 10,000 copies/mL, patients with low viral rebound had an intermediate clinical and immunologic prognosis.

The time course of CD4 cell counts in these 4 groups was studied during the first year following initial undetectability on HAART.

Back to Top | Article Outline

Statistical Analyses

Markov Model Based on CD4 Cell Dynamic

The objective was to study and compare the evolutions of CD4 lymphocyte cell count according to several levels of virologic rebound. We considered the Markov approach to be appropriate to modeling the longitudinal CD4 cell count evolution with the virologic rebound group as an explicative covariate, where stages were defined as CD4 count intervals. This modeling allowed a comparison of the mean CD4 cell count increases according to the levels of virologic rebound.

A continuous time-homogeneous Markov model with reversible stages was developed to model the time course of CD4 cell counts for each group of virologic rebound.

The Markov model with reversible stages allows an estimation of the transition intensities from stage i to stage i + 1 and inversely. The mathematic formulation of the Markov model based on these transition intensities is a straightforward extension of some previous models. 13,14

For each CD4 cell count evaluation, patients from each group were classified into 1 of 5 stages according to immunologic findings (Fig. 1). For a patient in the k-th virologic group, λk i,i+1 and λk i+1,i denote, respectively, the transition intensities from stage i to stage i + 1 and from stage i + 1 to stage i, for all i,k in {1,…,4}.



Bayesian inference was used to estimate the transition intensities, noted as λk *i,i+1, and was implemented using Markov Chain Monte Carlo (MCMC) simulations, via the Gibbs sampling algorithm. 16 After checking convergence by using classic techniques, 16 we used the estimated transition intensities for each virologic group k, at each iteration, to calculate functional estimated temporal transition probabilities Pkij * and the estimated mean transition time Tkij * required for transition from stage i to stage j, for all i,j in {1,…,5} and all k in {1,…,4}. 14 The results of the estimated transition intensities were obtained with 5000 Gibbs sampler iterations for each Markov model. This number of iterations was chosen after estimating the convergence criterion for the parameters of interest. The Student t test was used to compare the mean estimated parameters of interest.

Back to Top | Article Outline

Modeling the Time Course of CD4 Cell Counts

From the estimated transition probabilities, we derived the time course of CD4 cell count increases. Because the stages of our model were characterized by defined intervals of CD4 cell counts (except for the last stage), we chose the midpoint of each interval as being representative of each stage. For the last stage (CD4 cell counts above 650 cells/mm3), we chose a value of 900 cells/mm3 as representative of this class. Indeed, individuals classified in this stage at time 0 had a mean CD4 cell count of 889 cells/mm3. Therefore, the function of loss or acquisition of CD4 cells for the transition from stage i to stage j, noted νij for all i,j in {1,…,5}, is defined by the difference between the 2 representative values of each stage:MATH



For instance, the function of loss or acquisition of CD4 cells count for the transition from stage 1 to stage 2 (ν12) is equal to 175 cells/mm3 (275 − 100) and inversely equal to −175 cells/mm3 for the reverse transition.

Thus, if we consider V(Δt) as the random variable corresponding to the CD4 cell count increase in the time interval Δt, we thereby obtain an estimation of the mean temporal increase of the CD4 cell count for each virologic group k as the estimation of the expected value of V(Δt): MATH



where Pkij * (Δt) corresponds to the estimated transition probability from stage i to stage j in time interval Δt and ∐ki is the initial proportion of individuals in stage i for the virologic group k.

The Kolmogorov-Smirnov test was used to compare the time course of CD4 cell counts according to the degree of virologic rebound. Multiple comparisons between these distributions of CD4 cell counts were made using Bonferroni adjustments. The following groups were compared: 1 and 2, 2 and 3, and 3 and 4.

Back to Top | Article Outline


Baseline Characteristics

We applied our model to data on the 3736 selected HIV-1 infected individuals included in the FHDH. The study population was distributed as follows: group 1 (“durable virologic suppression”), 2636 subjects (70.6% of the overall population); group 2 (“low-level rebound”), 425 subjects (11.4%); group 3 (“intermediate-level rebound”), 163 subjects (4.3%); and group 4 (“high-level rebound”), 512 subjects (13.7%). Table 1 summarizes the characteristics of the patients and their treatments at HAART initiation according to the category of viral suppression after initial undetectability.



Figure 2 shows the proportions of individuals in each stage at inclusion in the study (ie, 1 year after the first plasma HIV-1 RNA value below 500 copies/mL), which occurred a median of 2.8 months after treatment initiation.



The distributions of the 5 CD4 cell count stages differed significantly among the 4 groups of virologic suppression (χ2 test , P < 0.0001). More patients in groups 3 and 4 than in groups 1 and 2 had CD4 cell counts below 350 cells/mm3 at baseline.

The median number of CD4 cell measurements per patient was 10 (interquartile range [IQR]: 7–13), and the median follow-up was 18.8 months (IQR: 14.7–21.3). There were 30,370 observed transitions contributing to the models (20,883 transitions in group 1, 3460 transitions in group 2, 1506 transitions in group 3, and 4521 transitions in group 4).

Back to Top | Article Outline

CD4 Cell Count Modeling

The estimated mean transition intensities (λk * i,i+1)i,k were all significantly different from one group to another (P < 0.0001) (Table 2). The estimated mean transition times, (T k * ij)i,j,k, obtained from these estimated transition intensities are reported in Table 3 for each virologic group k, together with the 95% posterior credibility interval (PCI) values. All these values were also significantly different from one group to another (P < 0.0001).





We observed that for the transitions between stages 1 and 2 and between stages 2 and 3 (CD4 cell counts below 500 cells/mm3), the acquisition of a higher CD4 cell count occurred more rapidly than the corresponding loss in every virologic group (P < 0.0001). For instance, in the “durable virologic suppression” group (group 1), the mean transition time from stage 3 to stage 2 was twice as long as the reverse transition from stage 2 to stage 3 (12.6 and 6.7 months, respectively; P < 0.0001). Starting from stage 3 (CD4 cell counts below 500 cells/mm3), the trend gets reversed in most of the virologic groups, and we observed acquisition times higher than loss times between stages 3 and 4 and stages 4 and 5.

Overall, the mean transition times from stage i to stage i + 1 (increase in CD4+ cell count) were shorter in group 1 than in group 2 and also shorter in group 3 than in group 4. In the same way, the reverse process (decrease in CD4 cell count) was slower in group 1 than in group 2 and also slower in group 3 than in group 4.

Figure 3 represents the estimated mean time course of CD4 cell count increase in each group. The time course of CD4 cell counts increased continuously in each group. These increases were larger in groups 1, 2, and 3 than in group 4, showing that the smaller the virologic rebound, the stronger was the CD4 cell increment (P < 0.0001). For instance, patients in group 2 had a larger increase in CD4 cell counts than patients in group 3 (P < 0.0001).



In groups 1, 2, and 3, the increases in the CD4 cell count between baseline and month 2 were 22, 18, and 19 cells/mm3, respectively, compared with only 5 cells/mm3 in group 4. The CD4 cell count increments between months 2 and 6 in groups 1, 2, 3, and 4 were 26, 20, 11, and 2 cells/mm3, respectively. The rate of increase fell after month 6, and the slope of the increase differed from one virologic group to another. For instance, the CD4 cell count increment between months 6 and 12 was 14 cells/mm3 in group 1, 7 cells/mm3 in groups 2 and 3, and 1 cell/mm3 in group 4. Finally, the increment between month 12 and month 24 was even smaller (5, 1, and 3 cells/mm3 in groups 1, 2, and 3, respectively, and 0 cells/mm3 in group 4).

Thus, the CD4 cell count increased on HAART, regardless of the level of virologic rebound, albeit more rapidly in patients with durable virologic suppression or low-level rebound than in patients with intermediate- or high-level rebound. Finally, the CD4 cell count stabilized after an initial increase, and the higher the level of virologic rebound, the earlier this stabilization occurred.

Back to Top | Article Outline


This study focused on the influence of virologic rebounds on the time course of the CD4 cell count in patients on HAART. Overall, CD4 cell counts tended to increase continuously for 18 months after an initial fall in viral load to below 500 copies/mL. This increase was observed regardless of the degree of virologic rebound, although it was minimal in patients in whom viral load rebounded to values above 10,000 copies/mL.

Because of the chosen exclusion criteria, data on 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 mean CD4 cell count increment. Another limitation of this study is the lack of information on treatment interruption. Nevertheless, lengthy treatment interruptions were likely to be uncommon during the study period (up to the end of 1999). In addition, short treatment interruptions due to adverse events were unlikely to affect our results, because we studied the influence of plasma viral load variations during the year following the first measured value below 500 copies/mL.

Despite these limitations, this assessment of immunologic outcome, being based on a large cohort and relatively lengthy follow-up, is likely to be more reliable than studies based on clinical trials involving smaller numbers of patients followed up for a maximum of 1 to 2 years after treatment initiation.

The Markov models are often used to model a longitudinal process, particularly to model the CD4 cell count dynamics. These models, contrary to regression models, do not require any preliminary parametric assumptions such as a preset time point for the change in the slope of CD4 cell count for linear models. These multistate models also proved useful in the context of a database with censored data.

Our study was based on a Markov assumption that the rate of progression from one CD4 cell count stage to the next is independent of the rate of progression through previous stages: the time-homogeneous assumption. This simplification is required to obtain tractable forms of the transition probabilities and, thus, the likelihood function.

The stages of our Markov model were defined by categorizing continuous changes in the CD4 cell count, which, as in all models based on categorization of continuous variables, inevitably entails a certain loss of information. We chose to model CD4 cell count kinetics on the basis of the following clinically relevant categories: <200, 200 through 349, 350 through 499, 500 through 649, and >650 cells/mm3.

We initially used a Markov model with 6 stages: stages 1 and 2 represented CD4 cell counts between 0 and 100 cells/mm3 and between 100 and 200 cells/mm3, respectively, and stages 3 through 6 represented the same count intervals as those used in the final model. The information available for estimating the transition intensities, particularly between stages 1 and 2, appeared to be insufficient in the 6-stage model, however. Stages 1 and 2 were thus combined into a single stage (<200 cells/mm3), yielding the final 5-stage model.

Our results show that it seems easier to reconstitute an immunologic response with a CD4 cell count below the 500 cells/mm3 threshold value. This is in line with the fact that the slope of CD4 cell count increase tends to become smaller over time. 21

Our results also suggest that the CD4 cell count continues to increase on HAART despite virologic rebound but that the rate of increase varies significantly with the degree of virologic rebound. No fall in mean CD4 cell count was observed in these virologic rebound groups, even among patients in whom viral load rebounded to more than 10,000 copies/mL.

Our results agree with those of Le Moing et al 22 for patients with low-level virologic rebound (<5000 copies/mL) but not for patients with high-level rebound (>10,000 copies/mL). Indeed, these authors found that the CD4 cell count continued to increase during the second year of follow-up when the plasma HIV-1 RNA level at the time of rebound was below 5000 copies/mL, whereas it stabilized when the plasma HIV-1 RNA level was between 5000 and 10,000 copies/mL and fell significantly when virologic rebound exceeded 10,000 copies/mL. This discrepancy could be related to treatment modification after rebound, viral resistance, viral fitness, or differences in the modeling approach.

Other studies showed an increase in CD4 cell count among patients who had detectable plasma HIV-1 RNA while on HAART. Deeks et al 9 suggested that patients who remained on even partially effective therapy had more durable CD4 cell count responses than patients who discontinued therapy and that this effect was independent of the level of viral replication. Kaufmann et al 23 reported that some patients adhering to HAART may show an increase in their CD4 cell count despite persistent viremia, suggesting that virologic and immunologic responses can diverge. Deeks et al 9 reported that among patients who failed to achieve durable viral suppression (<500 copies/mL), there was a median interval of 3 years between the onset of virologic failure and the return of the absolute CD4 cell count to the pretreatment level.

Our results do not necessarily hold for NNRTI-containing HAART. Indeed, virologic rebounds among NNRTI-treated patients quickly involve acquisition of mutations responsible for a complete resistance of the virus to NNRTIs. 19 Moreover, viral mutations and subsequent viral fitness are different when rebound of viral replication occurred on NNRTI therapy compared with PI therapy.

We found that the rate of increase in the CD4 cell count tended to be inversely proportional to the level of virologic rebound. Taken together, our findings suggest that in terms of further CD4 count increase, a threshold of 5000 copies/mL is a practical value with which to define virologic failure and to consider a treatment switch. The risk of viral mutation is unknown in patients with this level of viral load on HAART, however. In patients with low or moderate virologic rebound while on a PI regimen or HAART who have difficulties with treatment observance or have previously been exposed to multiple antiretroviral regimens, delaying the treatment switch may spare some therapeutic options for future use without markedly compromising the CD4 cell count.

Back to Top | Article Outline


1. Autran B, Carcelain G, Li TS, et al. Positive effects of combined antiretroviral therapy on CD4+ T cell homeostasis and function in advanced HIV disease. Science. 1997; 277:112–116.
2. Carpenter CC, Cooper DA, Fischl MA, et al. Antiretroviral therapy in adults: updated recommendations of the International AIDS Society-USA Panel. JAMA. 2000; 283:381–390.
3. Katzenstein DA, Hammer SM, Hughes MD, et al. The relation of virologic and immunologic markers to clinical outcomes after nucleoside therapy in HIV-infected adults with 200 to 500 CD4+ cell per cubic millimeter. N Engl J Med. 1996; 335:1091–1098.
4. Li TS, Tubiana R, Katlama C, et al. Long-lasting recovery in CD4+ T-cell function and viral-load reduction after highly active antiretroviral therapy in advanced HIV-1 disease. Lancet. 1998; 351:1682–1686.
5. Wood E, Yip B, Hogg RS, et al. Full suppression of viral load is needed to achieve an optimal CD4+ cell count response among patients on triple drug antiretroviral therapy. AIDS. 2000; 14:1955–1960.
6. Grabar S, Le Moing V, Goujard C, et al. Clinical outcome of patients with HIV-1 infection according to immunologic and virologic response after 6 months of highly active antiretroviral therapy. Ann Intern Med. 2000; 133:401–410.
7. Ledergerber B, Egger M, Opravil M, et al. Clinical progression and virological failure on highly active antiretroviral therapy in HIV-1 patients: a prospective cohort study. Swiss HIV Cohort Study. Lancet. 1999; 353:863–868.
8. Cox DR, Miller HD. The Theory of Stochastic Processes and Their Applications. New York: Krieger; 1980.
9. Deeks SG, Barbour JD, Grant RM, et al. Duration and predictors of CD4 T-cell gains in patients who continue combination therapy despite detectable plasma viremia. AIDS. 2002; 16:201–207.
10. Longini IM, Clark WS, Gardner LI, et al. The dynamics of CD4+ T-lymphocyte decline in HIV-infected individuals: a Markov modeling approach. J Acquir Immune Defic Syndr Hum Retovirol. 1991; 4:1141–1147.
11. Bwayo JJ, Nagelkerke NJ, Moses S, et al. Comparison of the declines in CD4+ counts in HIV-1-seropositive female sex workers and women from the general population in Nairobi, Kenya. J Acquir Immune Defic Syndr Hum Retrovirol. 1995; 10:457–461.
12. Hendriks JC, Satten GA, Longini IM, et al. Use of immunological markers and continuous-time Markov models to estimate progression of HIV infection in homosexual men. AIDS. 1996; 10:649–656.
13. Guihenneuc C, Richardson S, Longini IM. Modeling markers of disease progression by a hidden Markov process: application to characterizing CD4+ cell decline. Biometrics. 2000; 56:733–741.
14. Kousignian I, Autran B, Chouquet C, et al. Markov modeling of changes in HIV-specific cytotoxic T lymphocyte responses with time in untreated HIV-1 infected patients. Stat Med. 2003; 22:1675–1690.
15. Billard L, Zhao Z. Review and synthesis of the HIV/AIDS epidemic as a multi-stage process. Math Biosci. 1993; 117:19–33.
16. Gilks WR, Richardson S, Spiegelhalter DJ. Markov Chain Monte Carlo in Practice. London: Chapman & Hall; 1996.
17. Sharples LD. Use of the Gibbs sampler to estimate transition intensities between grades of coronary disease following cardiac transplantation. Stat Med. 1993; 12:1155–1169.
18. Tassie JM, Gasnault J, Bentata M, et al. Survival improvement of AIDS-related progressive multifocal leukoencephalopathy in the era of protease inhibitors. Clinical Epidemiology Group. French Hospital Database on HIV. AIDS. 1999; 13:1881–1887.
19. Delaugerre C, Rohban R, Simon A, et al. Resistance profile and cross-resistance of HIV-1 among patients failing a non-nucleoside reverse transcriptase inhibitor-containing regimen. J Med Virol. 2001; 65:445–448.
20. Abgrall S, Duval X, Joly V, et al. Clinical and immunologic outcomes according to different levels of virologic suppression (VS) after undetectability on HAART. Presented at the First IAS Conference on HIV Pathogenesis and Treatment, Buenos Aires, July 8–11, 2001.
21. Smith CJ, Sabin CA, Lampe FC, et al. The potential for CD4 cell increases in HIV-positive individuals who control viraemia with highly active antiretroviral therapy. AIDS. 2003; 17:963–969.
22. Le Moing V, Thiebaut R, Chene G, et al. Predictors of long-term increase in CD4(+) cell counts in human immunodeficiency virus-infected patients receiving a protease inhibitor-containing antiretroviral regimen. J Infect Dis. 2002; 185:471–480.
23. Kaufmann D, Pantaleo G, Sudre P, et al. CD4-cell count in HIV-1–infected individuals remaining viraemic with highly active antiretroviral therapy (HAART). Swiss HIV Cohort Study. Lancet. 1998; 351:723–724.
Back to Top | Article Outline


French Hospital Database on HIV Clinical Epidemiology Group

Scientific Committee

S. Alfandari, F. Bastides, E. Billaud, A. Boibieux, F. Boué, F. Bricaire, D. Costagliola, L. Cotte, L. Cuzin, F. Dabis, P. Enel, S. Fournier, J. Gasnault, C. Gaud, J. Gilquin, S. Grabar, D. Lacoste, J. M. Lang, H. Laurichesse, P. Leclercq, C. Leport, M. Mary-Krause, S. Matheron, M. C. Meyohas, C. Michelet, J. Moreau, G. Pialoux, I. Poizot-Martin, C. Pradier, C. Rabaud, E. Rouveix, P. Saïag, D. Salmon-Ceron, J. Soubeyrand, and H. Tissot-Dupont

Back to Top | Article Outline

DMI2 Coordinating Center

French Ministry of Health (B. Haury, V. Tirard-Fleury, and I. Tortay)

Back to Top | Article Outline

Statistical Analysis Center

INSERM EMI 0214 (S. Abgrall, D. Costagliola, S. Grabar, E. Lanoy, L. Lièvre, M. Mary-Krause, and V. Potard)

Back to Top | Article Outline

CISIH (Paris Area)

CISIH de Bichat-Claude Bernard (Hôpital Bichat-Claude Bernard: S. Matheron, J. P. Coulaud, J. L. Vildé, C. Leport, P. Yeni, C. Mandet, E. Bouvet, C. Gaudebout, B. Crickx, C. Picard-Dahan), CISIH de Paris-Center (Hôpital Broussais: L. Weiss, D. Tisne-Dessus; GH Tarnier-Cochin: D. Sicard, D. Salmon; Hôpital Saint-Joseph: J. Gilquin, I. Auperin), CISIH de Paris-Ouest (Hôpital Necker adultes: J. P. Viard, L. Roudière; Hôpital Laennec: W. Lowenstein; Hôpital de l'Institut Pasteur), CISIH de Paris-Sud (Hôpital Antoine Béclère: F. Boué, R. Fior; Hôpital de Bicêtre: J. F. Delfraissy, C. Goujard; Hôpital Henri Mondor: Ph. Lesprit, C. Jung; Hôpital Paul Brousse), CISIH de Paris-Est (Hôpital Rothschild: W. Rozenbaum, G. Pialoux; Hôpital Saint-Antoine: M. C. Meyohas, J. L. Meynard, O. Picard, N. Desplanque; Hôpital Tenon: J. Cadranel, C. Mayaud), CISIH de Pitié-Salpétrière (GH Pitié-Salpétrière: F. Bricaire, C. Katlama, S. Herson, A. Simon), CISIH de Saint-Louis (Hôpital Saint-Louis: J. M. Decazes, J. M. Molina, J. P. Clauvel, L. Gerard; GH Lariboisière-Fernand Widal: J. M. Salord, M. Diemer), CISIH 92 (Hôpital Ambroise Paré: C. Dupont, H. Berthé, P. Saïag; Hôpital Louis Mourier: E. Mortier, C. Chandemerle; Hôpital Raymond Poincaré: P. de Truchis), CISIH 93 (Hôpital Avicenne: M. Bentata, P. Berlureau; Hôpital Jean Verdier: J. Franchi, V. Jeantils; Hôpital Delafontaine: D. Mechali, B. Taverne)

Back to Top | Article Outline

CISIH (Outside Paris Area)

CISIH Auvergne-Loire (CHU de Clermont-Ferrand: H. Laurichesse, F. Gourdon; CHRU de Saint-Etienne: F. Lucht, A. Fresard), CISIH de Bourgogne-Franche Comté (CHRU de Besançon; CHRU de Dijon; CH de Belfort: J. P. Faller, P. Eglinger), CISIH de Caen (CHRU de Caen: C. Bazin, R. Verdon), CISIH de Grenoble (CHU de Grenoble), CISIH de Lyon (Hôpital de la Croix-Rousse: D. Peyramond, A. Boibieux; Hôpital Edouard Herriot: J. L. Touraine, J. M. Livrozet; Hôtel-Dieu: C. Trepo, L. Cotte; CH de Lyon-Sud: Médecine Pénitentiaire: P. Barlet), CISIH de Marseille (Hôpital de la Conception: I. Ravaux, H. Tissot-Dupont; Hôpital Houphouët-Boigny: J. P. Delmont, J. Moreau; Institut Paoli Calmettes: J. A. Gastaut; Hôpital Sainte-Marguerite: I. Poizot-Martin, J. Soubeyrand, F. Retornaz; CHG d'Aix-En-Provence: P. A. Blanc, T. Allegre; Center pénitentiaire des Baumettes: A. Galinier, J. M. Ruiz; CH d'Arles; CH d'Avignon: G. Lepeu; CH de Digne Les Bains: P. Granet-Brunello; CH de Gap: L. Pelissier, J. P. Esterni; CH de Martigues: M. Nezri, R. Cohen-Valensi; CHI de Toulon: A. Laffeuillade, S. Chadapaud), CISIH de Montpellier (CHU de Montpellier: J. Reynes; CHG de Nîmes), CISIH de Nancy (Hôpital de Brabois: T. May, C. Rabaud), CISIH de Nantes (CHRU de Nantes: F. Raffi, E. Billaud), CISIH de Nice (Hôpital Archet 1: C. Pradier, P. Pugliese; CHG Antibes Juan les Pins), CISIH de Rennes (CHU de Rennes: C. Michelet, C. Arvieux), CISIH de Rouen (CHRU de Rouen: F. Caron, F. Borsa-Lebas), CISIH de Strasbourg (CHRU de Strasbourg: J.M. Lang, D. Rey, P. Fraisse; CH de Mulhouse), CISIH de Toulouse (CHU Purpan: P. Massip, L. Cuzin, E. Arlet-Suau, M. F. Thiercelin Legrand; Hôpital la Grave; CHU Rangueil), CISIH de Tourcoing-Lille (CH Gustave Dron; CH de Tourcoing: S. Alfandari, Y. Yasdanpanah), CISIH de Tours (CHRU de Tours; CHU Trousseau)

Back to Top | Article Outline

CISIH (Overseas)

CISIH de Guadeloupe (CHRU de Pointe-à-Pitre), CISIH de Guyane (CHG de Cayenne: M. Sobesky, R. Pradinaud), CISIH de Martinique (CHRU de Fort-de-France), CISIH de La Réunion (CHD Félix Guyon: C. Gaud, M. Contant)


CD4 lymphocytes; virologic rebound; highly active antiretroviral therapy (HAART); Markov model; Bayesian inference

Copyright © 2003 Wolters Kluwer Health, Inc. All rights reserved.