The short term beneficial effect of highly active antiretroviral therapy (HAART) on survival and development of an AIDS-defining illness is well established [1–3]. Not only has the life expectancy of HIV-infected patients increased but also their quality of life has improved [4,5]. In addition the spectrum of causes of mortality is changing as the number of deaths related to opportunistic infections has diminished [6,7].
A significant problem associated with HAART is the adverse effects of drugs which may result in increasing morbidity and a reduced quality of life [4,8,9]. In addition, low adherence to drug regimens may lead to sub-optimal therapy thereby jeopardizing the likelihood of maintaining viral suppression [10–12], which in turn may lead to therapy failure and drug resistance [13,14]. It is therefore currently recommended that antiretroviral treatment is deferred in patients who have high CD4+ T-cell counts or relatively low plasma levels of HIV-RNA or both . Another debated issue is structured therapy interruptions (STI) which, by enhancing HIV-specific immune responses, may contribute to maintaining virologic suppression [16–19].
Several studies have assessed prognostic markers related to progression to death or AIDS. Baseline CD4+ T-cell counts are highly predictive for (AIDS-free) survival in both untreated and treated patients whereas baseline HIV-RNA levels have little additional predictive value in treated patients [20–23]. Increases in CD4+ T-cell counts during treatment with HAART and reaching undetectable levels of viral load are also significantly associated with a better prognosis [22,24,25].
In the present study prognostic markers for survival and progression to AIDS among HIV-infected patients in the ATHENA observational cohort  were included in a multivariate hazards model. This model was used to estimate 5-year survival probabilities for patients at initiation of HAART. We know of only one previous study that estimated survival probabilities for patients treated with HAART . These probabilities are important when answering questions regarding life expectancy when patients begin demanding and life-long anti-HIV treatment. In addition, we assessed the effect of HAART taken continuously or discontinuously, which was used as a surrogate marker of therapy success. This allowed us to predict the effect on disease outcome of unstructured treatment interruptions or deferred treatment. Finally we studied the changes in HIV-related and non-related mortality and associated risk factors.
Data used in this study were selected from the ATHENA observational cohort . This cohort consists of HIV-infected patients living in The Netherlands (n = 3908 by 31 July 2001) and using highly active antiretroviral therapy (HAART). At the time of inclusion patients were at least 18 years old and gave informed consent to participate. Monitoring of patients was undertaken in 22 hospitals across the country. At each visit data were collected on case report forms and entered into the database on site. Data were then sent to the central database where consistency checks were performed. Source document verification was performed for 10% of the data and mistakes or missing data were reported back to each site and corrected.
Prospective data collection started in May 1998. From patients who started HAART before that time data were collected retrospectively, including those patients who had died before 1 May 1998. Demographic data were collected at entry in ATHENA. Clinical data focused on HIV-1 infection related events according to the Centers of Disease Control definition , the therapeutic and prophylactic drugs used to treat opportunistic infections and the combination and dose of the antiretroviral drugs used and their side-effects. In addition, data on plasma HIV-1 RNA concentration and CD4+ and CD8+ T-cell counts were collected.
The start date T0 of HAART, which in our study defined the start of follow-up, was determined from the data on antiretroviral drug use. HAART was defined as a combination of at least three drugs from at least two classes (nucleoside and non-nucleoside reverse transcriptase inhibitors and protease inhibitors). Patients whose HAART regimen was not recorded before closure of the database (31 July 2001) were excluded from this study (n = 166). Therapy interruptions after T0 were assessed by counting the number of weeks NHAART in which HAART was used in the preceding 24 weeks. In the first 24 weeks after start of the first HAART regimen we assumed a 100% usage of HAART (NHAART = 24). After each period of 24 weeks of follow-up NHAART was adjusted to its most recent value.
CD4+ T-cell counts and HIV-RNA load were measured on average every 12 weeks. Baseline (i.e. at start of HAART) CD4 cell count and viral load were determined by taking the value closest to the start of HAART measured between 24 weeks prior to T0 and 1 week thereafter for CD4 cell counts and 0 weeks after T0 for viral load. Disease status at baseline was defined as the most serious CDC-event (category B or C) in the year prior to T0 and the first 4 weeks thereafter. Additional information was collected on the transmission route (homosexual sex, heterosexual sex, intravenous drug use and other), gender and age at T0. Groups of patients were compared using Wilcoxon Mann–Whitney and χ2 non-parametric tests.
End points in this study were either death or AIDS events occurring as of 4 weeks after start of HAART. A CD4 cell count below 200 × 106 cells/l in isolation was not considered as an AIDS-defining event . Patients who did not die or develop AIDS were censored at 3 months after the last follow-up visit, at the date of closure of the database, or, when analysing progression to AIDS, at the date of death, whichever came first. The period of 3 months corresponded with the median time between two consecutive visits.
Death cases were scored by a panel of three independent physicians (P.R., I.C.G., K.B.) as HIV-related, non-HIV-related (including therapy-related cases) or unknown. This score was based on clinical data at time of death reported by the treating physician and the patient's history of CDC and adverse events. Cases about which the physicians disagreed were discussed by the panel until a consensus score was reached. Mortality was calculated per 100 person-years of follow-up after T0. Poisson's distribution was used to calculate 95% confidence intervals for rates. Expected death rates were calculated for an age and gender matched group from the general Dutch population .
A multivariate hazards model was used to find the set of covariates that best predicted the time from start of HAART to onset of AIDS and death. For each patient the hazard for death or AIDS, h(t), after t years of follow-up after T0 is modelled as the product of an underlying hazard h0(t) common to all patients and a function containing patient-specific covariates. In order to estimate the underlying hazard in models with time-dependent covariates we used a discrete-time generalized linear model  in which for each patient the follow-up time was split in 3-week intervals, such that the Poisson probability of having two or more events per patient in the same time interval could be neglected. The hazard h(ti) of dying in time interval i is then given by the expression
where the sum runs over all covariates j with corresponding hazard ratios exp[βj(ti)]. The survival probability S(ti) up to time interval i is given by S(ti) = exp[−H(ti)] with H(ti) the cumulative hazard at ti defined as H(ti) = ∑j ≤ih(tj).
Covariates considered for inclusion in the model were disease status at start of HAART, age, gender and transmission route. Antiretroviral pre-treatment before start of HAART was included as a dichotomous covariate indicating whether or not a patient was pre-treated more than 1 year prior to T0 (no significant difference in disease progression was found between therapy-naive patients and patients pre-treated less than 1 year). NHAART was stratified in three categories: continuous use of HAART (NHAART = 24), not continuous but 16 weeks or more (16 ≤ NHAART < 24) and less than 16 weeks (NHAART < 16). We also considered baseline CD4 cell count and viral load and a covariate indicating whether a baseline CD4 cell count was available. An analogous covariate for viral load was not included, as it would be severely biased because pre-treated patients were less likely to have a baseline viral load than therapy-naive patients (Table 1). Parameters were estimated by maximizing the partial likelihood. Covariates were excluded from the model via backward elimination if this did not yield a significantly worse model (P < 0.01, log-likelihood, χ2 test). Wald 95% confidence intervals were calculated for the hazard ratios.
Predicted survival probabilities
By extrapolating the model beyond the median follow-up time of the study group we could estimate 5-year survival probabilities after initiation of HAART using the hazards model as described. We estimated survival probabilities assuming various therapy scenarios: (a) continuous HAART (NHAART = 24 always), (b) HAART is interrupted for 4 weeks in each 24-week interval of follow-up (by definition NHAART = 24 in the first 24 weeks after T0 and NHAART = 20 thereafter), irrespective of the reason and the temporal pattern of the interruptions, and (c) instead of starting HAART at T0 patients waited for 48 weeks and started continuous HAART thereafter (`deferred HAART', NHAART = 24 in the first 24 weeks of follow-up, 0 in the next 48 weeks, and again 24 thereafter). In this latter scenario T0 is not the start of HAART but the time by which patients were eligible for treatment. Baseline values for the covariates in the hazards model were still determined at this time to allow a direct comparison between the three scenarios. Bootstrapping methods were used to obtain 95% confidence intervals on S(t) and to test the significance of differences between survival probabilities. A risk calculator based on our model will become available via our website (http://www.hiv-monitoring.nl).
All statistical analyses were carried out using SAS version 8.00 (SAS Institute, Cary, North Carolins, USA).
The number of patients in the ATHENA database that initiated HAART was 3742 of which 18 patients were excluded as their age at start of HAART was unknown. Thus our study contained 3724 patients who received HAART of whom 346 died during follow-up. The total follow-up time was 12 503 person-years. Characteristics of the patients at start of HAART are shown in Table 1. Among the 3708 patients who survived more than 4 weeks 459 patients were diagnosed with AIDS during follow-up. Of the 16 patients who did not survive more than 4 weeks nine died whilst the remaining seven patients were censored.
Development of AIDS
Covariates associated with progression to AIDS are listed in Table 2. The underlying hazard function h0 was modelled as a constant plus the log-transformed time of follow-up which was significantly associated with a reduced risk of developing AIDS . For pre-treated patients the risk of progression to AIDS was 1.91 times larger than for patients who had no or less than 1 year of previous treatment. Each unit increase in log-transformed baseline CD4 cell count reduced the risk of progression to AIDS by a factor 0.62. Patients without a measured baseline CD4 cell count had a hazard ratio of 3.11 compared to those who had one. Gender, transmission category and viral load and age at baseline were not significantly associated with development of AIDS.
The effect on disease progression of having been diagnosed with AIDS in the year prior to start of HAART was the only one that was found to be dependent on time of follow-up after start of HAART. Compared to patients who were not diagnosed with AIDS before initiation of HAART the hazard ratio was 4.22 in the first half year after T0 decreasing to 1.83 in the 2.5 years thereafter, while after 3 years following start of HAART the hazard ratio was no longer significant. A Centers of Disease Control (CDC) category-B event in the year prior to start of HAART was not significantly associated with progression to AIDS after start of HAART. Interrupting HAART up to 8 weeks (16 ≤ NHAART < 24) or more than eight weeks (NHAART < 16) of the preceding 24 weeks was associated with a hazard ratio of 2.03 and 4.94, respectively, compared to continuous treatment with HAART.
In addition to the covariates associated with progression to AIDS two extra covariates were associated with survival, transmission via intravenous drug use and age at start of HAART (Table 2). In contrast to progression to AIDS, survival was not directly associated with time of follow-up. The hazard ratio of having experienced a CDC category-C event in the year prior to start of HAART was 14.5 in the first half year after initiation of HAART decreasing to 2.39 in the 2.5 years thereafter. Hazard ratios associated with treatment interruptions were 4.00 (16 ≤ NHAART < 24) and 7.28 (NHAART < 16).
Estimated 5-year survival probabilities as a function of baseline CD4 cell count and age for therapy-naive patients who are not intravenous drug users are given in Figure 1. Figures 1a and 1b show survival probabilities for patients without and with an AIDS-defining disease, respectively, in the year prior to the start of HAART who use HAART continuously (therapy scenario a). For patients younger than 50 years with CD4 cell counts above 10 × 106 or 150 × 106 cells/l (see Figure 1a and 1b), respectively, the predicted survival probabilities are above 90%. Estimated 5-year survival probabilities for patients without prior AIDS-defining event who interrupt HAART during 4 weeks of each 24-week interval of follow-up (scenario b) are only above 90% when CD4 cell counts at start of treatment are above 450 × 106 cells/l (Figure 1c). When the same group of patients defers HAART for 48 weeks and uses continuous HAART thereafter patients with CD4 counts above 110 × 106 cells/l have a 90% 5-year survival probability (scenario c).
HIV-related and non-related mortality
Of the 346 deaths in our study group HIV-related mortality was scored as the most probable cause of death in 191 (55%) cases whereas in 97 (28%) cases the cause of death was non-HIV-related of which seven (2%) were therapy-related. In 58 (17%) cases the cause of death could not be determined. HIV-related mortality per 100 person-years of follow-up since start of HAART was 3.8 [95% confidence interval (CI), 2.4–5.9] in 1996 and decreased to 0.7 (95% CI, 0.5–1.1) in 2000 (P < 0.01) (Fig. 2). Non-HIV-related mortality did not change over time, 0.4 (95% CI, 0.0–1.3) and 0.9 (95% CI, 0.6–1.3), respectively (P = 0.25).
Covariates associated with either HIV-related or non-related death are shown in Table 3. No significant association was found between intravenous drug use and HIV-related death while CD4 cell counts at baseline were not associated with non-HIV-related death. Pre-treatment with antiretroviral drugs, a CDC category-C event in the year prior to start of HAART as well as discontinuous use of HAART were associated with an increased risk of HIV-related and non-related death.
Our study shows, in accordance with previous findings, that clinical markers associated with a higher survival probability and a slower progression to AIDS are a high baseline CD4+ T-cell count, absence of CDC category-C events before start of HAART and no or limited prior treatment with antiretroviral drugs. Age and intravenous drug use are significant predictors for progression to death but not for development of AIDS. In addition, continuous HAART is associated with slower progression to death and AIDS in comparison with interrupted HAART. The large difference between the AIDS and survival model in parameter values associated with an AIDS-diagnosis in the year prior to initiation of HAART is largely explained by the inclusion of time of follow-up in the first model. To avoid complexity in our model we only included prognostic variables that are measured in routine patient care. Thus our model could easily be applied to other observational cohorts.
In accordance with previous studies [22,23] we did not find a significant association between HIV-RNA levels at start of HAART and subsequent disease progression. Other studies showed that reaching undetectable plasma HIV-RNA levels during HAART is much more strongly related to risk of death and AIDS than pre-HAART RNA levels are [22,24]. In our analyses, however, we did not include viral load during HAART as we aimed to predict survival probabilities at the time of initiating HAART. Furthermore, we did not include haemoglobin and transaminase levels which have been shown to be associated with survival probability  since in our data these values were only recorded during follow-up.
By using our model to estimate survival probabilities we found that for young patients starting continuous HAART whilst in a less advanced stage of HIV-infection the 5-year survival probabilities were above 90%. This agrees with the 3-year survival probabilities that were previously found for a similar group of patients and varied between 95.9% and 99.8% . Longer term predictions could be made but since the long-term effects of HAART are still unknown, these predictions cannot be validated with currently available data and might be too optimistic considering the risk of toxicities and resistance.
The effect of treatment on disease outcome is assessed by studying therapy interruptions, which reflect both failure to suppress viral load or to increase CD4 cell counts and toxicity-driven switches. This analysis does not aim at drawing conclusions about the effect of a treatment strategy adopted at start of HAART. The high survival probabilities for patients with high baseline CD4 counts using interrupted HAART confirm that occasional treatment interruptions shorter than 3 months do not increase the risk of death . Our results are also compatible with STI studies which show that short-term disease progression does not change or may even improve slightly [16–18]. However, therapy interruptions in our cohort are largely unstructured and occur for a longer period of time (median duration 2 months, data not shown) than in previous reports. Moreover, therapy may have been interrupted because of toxicity or therapy failure, the latter two being associated themselves with an increased risk of death and AIDS. Other studies have shown that therapy failure is the reason for about 10% of the therapy interruptions within 1 year after start of HAART, and toxicity for about 30%. This latter percentage increases with each calendar year following the introduction of HAART [26,32,33]. Untangling the pure effect of STI from these other therapy interruptions is difficult in the settings of an observational cohort.
A possible underestimation of the effect of therapy interruptions is caused by patients who die within 24 weeks after initiating HAART as, by construction, therapy interruptions are not counted. We argue, however, that the effect of interruptions on disease progression is not instantaneous but only manifests itself after some time.
The optimum time to initiate HAART in asymptomatic patients is an issue that is hard to resolve [15,27]. The beneficial effect on immune restoration should be balanced against the risk of drug-related toxicities and development of drug resistance [15,34]. Our model could predict survival probabilities for patients who deferred initiation of HAART for 1 year after becoming eligible for treatment, although these predictions were not based on a dedicated study on deferring treatment. We approximated the decrease in CD4 counts in this year by an increased risk of death due to the absence of treatment. This risk is probably overestimated as treatment interruptions in our data were always during HAART. Compared to patients of the same age and CD4 cell count who initiated HAART immediately after becoming eligible for treatment, the difference in survival probabilities was a few percentage points when CD4 cell counts were above 110 × 106 cells/l. Although small, this difference was significant and compatible with previous findings that initiating HAART in patients with CD4 cell counts above 350 × 106 cells/l significantly delayed clinical progression .
HIV-related and non-related mortality
From 1996 until 2000 the mortality amongst HIV-infected persons in the ATHENA cohort declined. A similar reduction has been observed in other studies and is largely explained by the introduction of HAART [2,3]. However, part of the decline might be attributed to changes in the infected and treated population during these years, as it shifted towards one in a less advanced stage of HIV-1 infection at the start of HAART with a growing fraction of therapy-naive patients. Mortality rates in 2001 were unreliable as not all follow-up data are available yet.
Non-HIV-related mortality was two to three times higher than in the general population . Part of this excess can be explained by the seven proven and approximately 25 possibly therapy-related causes of death. Moreover, the HIV-infected and treated population is not representative of the general Dutch population, as there are relatively more intravenous drug-users who have a higher risk of dying of non-HIV-related causes.
Although misclassification of causes of death may limit interpretation of the results, the stable incidence of non-HIV-related causes over the years suggests that toxicity has not yet become a major cause of death. This may, however, change in the future when the long-term consequences of HAART become more clearly recognized. Thus, the recognition and careful recording of therapy-related toxicity and deaths remains of utmost importance.
In conclusion, using only covariates known at baseline, we could predict survival probabilities for patients initiating HAART under various therapy scenarios. Our model predicted that using HAART continuously is the best scenario. Deferring HAART for 1 year after becoming eligible for treatment and using continuous HAART thereafter is, according to our model, better than using interrupted HAART. However, absolute differences between 5-year survival probabilities in these three different scenarios were small for asymptomatic patients younger than 50 years with high levels of CD4 cell counts. Considering the risk of toxicities and adherence problems deferring treatment might therefore be acceptable.
Sponsorship: This study was supported by the National Health Insurance Council (grant number 97-46486), Amstelveen, The Netherlands. A.C.G. is supported by The Royal Society, London, United Kingdom. F.D.W. is supported by the Wellcome Trust, London, United Kingdom.
1. Palella FJ, Delaney KM, Moorman AC, Loveless MO, Fuhrer J, Satten GA, et al
. Declining morbidity and mortality among patients with advanced human immunodeficiency virus infection.N Engl J Med
2. Egger M, Hirschel B, Francioli P, Sudre P, Wirz M, Flepp M, et al
. Impact of new antiretroviral combination therapies in HIV infected patients in Switzerland: prospective multicentre study.BMJ
3. Mocroft A, Vella S, Benfield TL, Chiesi A, Miller V, Gargalianos P, et al
. Changing patterns of mortality across Europe in patients infected with HIV-1.Lancet
4. Cohen C, Revicki DA, Nabulsi A, Sarocco PW, Jiang P and the Advanced HIV
Disease Ritonavir Study Group. A randomized trial of the effect of ritonavir in maintaining quality of life in advanced HIV disease.AIDS
5. Revicki DA, Moyle G, Stellbrink HJ, Barker C. Quality of life outcomes of combination zalcitabine-zidovudine, saquinavir- zidovudine, and saquinavir-zalcitabine-zidovudine therapy for HIV-infected adults with CD4 cell counts between 50 and 350 per cubic millimeter.AIDS
6. Valdez H, Chowdhry TK, Asaad R, Woolley IJ, Davis T, Davidson R, et al
. Changing spectrum of mortality due to human immunodeficiency virus: analysis of 260 deaths during 1995
7. Lewden C, Raffi F, Chêne G, Sobel A, Leport C, and The APROCO Study Group. Mortality in a cohort of HIV-infected adults started on a protease inhibitor-containing therapy.J Acquir Immune Defic Syndr
8. Carr A, Cooper DA. Adverse effects of antiretroviral therapy.Lancet
9. Carr A, Samaras K, Thorisdottir A, Kaufmann GR, Chisholm DJ, Cooper DA. Diagnosis, prediction, and natural course of HIV-1 protease-inhibitor-associated lipodystrophy, hyperlipidaemia, and diabetes mellitus: a cohort study.Lancet
10. Haubrich RH, Little SJ, Currier JS, Forthal DN, Kemper CA, Beall GN, et al
. The value of patient-reported adherence to antiretroviral therapy in predicting virologic and immunologic response.AIDS
11. Gifford AL, Bormann JE, Shively MJ, Wright BC, Richman DD, Bozette SA. Predictors of self-reported adherence and plasma HIV concentrations in patients on multidrug antiretroviral regimens.J Acquir Immune Defic Syndr
12. Nieuwkerk PT, Sprangers MAG, Burger DM, Hoetelmans RMW, Hugen PWH, Danner SA, et al
. Limited patient adherence to highly active antiretroviral therapy for HIV-1 Infection in an observational cohort study.Arch Intern Med
13. Vanhove GF, Schapiro JM, Winters MA, Merigan TC, Blaschke TF. Patient compliance and drug failure in protease inhibitor monotherapy [letter].JAMA
14. Race E, Dam E, Obry V, Paulous S, Clavel F. Analysis of HIV cross-resistance to protease inhibitors using a rapid single-cycle recombinant virus assay for patients failing on combination therapies.AIDS
15. Yeni PG, Hammer SM, Carpenter CCJ, Cooper DA, Fischl MA, Gatell JM, et al
. Antiretroviral treatment for adult HIV infection in 2002, updated recommendations of the International AIDS Society
16. Ruiz L, Carcelain G, Martínez-Picado J, Frost S, Marfil S, Paredes R, et al
. HIV dynamics and T-cell immunity after three structured treatment interruptions in chronic HIV-1 infection.AIDS
17. García F, Plana M, Ortiz GM, Bonhoeffer S, Soriano A, Vidal C, et al
. The virological and immunological consequences of structured treatment interruptions in chronic HIV-1 infection.AIDS
18. Dybul M, Chun TW, Yoder C, Hidalgo B, Belson M, Hertogs K, et al
. Short-cycle structured intermittent treatment of chronic HIV infection with highly active antiretroviral therapy: effects on virologic, immunologic, and toxicity parameters.Proc Nat Acad Sci
19. Taffé P, Rickenbach M, Hirschel B, Opravil M, Furrer H, Janin P, et al
. Impact of occasional short interruptions of HAART on the progression of HIV infection: results from a cohort study.AIDS
20. Mellors JW, Muñoz A, Giorgi JV, Margolick JB, Tassoni CJ, Gupta P, et al
. Plasma viral load and CD4+ lymphocytes as prognostic markers of HIV-1 infection.Ann Intern Med.
21. De Wolf F, Spijkerman I, Schellekens PT, Langendam M, Kuiken C, Bakker M, et al
. AIDS prognosis based on HIV-1 RNA, CD4+ T-cell count and function: markers with reciprocal predictive value over time after seroconversion.AIDS
22. Ledergerber B, Egger M, Erard V, Weber R, Hirschel B, Furrer H, et al
. AIDS-related opportunistic illnesses occurring after initiation of potent antiretroviral therapy.JAMA
23. Sterling TR, Chaisson RE, Moore RD. HIV-1 RNA, CD4 T-lymphocytes, and clinical response to highly active antiretroviral therapy.AIDS
24. Ledergerber B, Egger M, Opravil M, Telenti A, Hirschel B, Battegay M, et al
. Clinical progression and virological failure on highly active antiretroviral therapy in HIV-1 patients: a prospective cohort study.
Lancet 1999, 353
25. Miller V, Mocroft A, Reiss P, Katlama C, Papadopoulos AI, Katzenstein T, et al
. Relations among CD4 lymphocyte count nadir, antiretroviral therapy, and HIV-1 disease progression: results from the EuroSIDA study.Ann Intern Med.
26. De Wolf F, Lange JMA, Bossuyt PMM, Dijkgraaf MGW, Burger DM, Nieuwkerk PT, et al
. The ATHENA cohort study: implications of the introduction of HAART for the course of HIV-1 disease, public health and health care as well as the economic costs and benefits. In: Monitoring of Human Immunodeficiency Virus Type 1 (HIV-1) Infection in The Netherlands.
July 2001. Amsterdam: HIV
Monitoring Foundation; 2001. pp. 18–51.
27. Egger M, May M, Chêne G, Phillips AN, Ledergerber B, Dabis F, et al
. Prognosis of HIV-1-infected patients starting highly active antiretroviral therapy: a collaborative analysis of prospective studies.Lancet
28. Centers for Disease Control and Prevention. 1993 revised classification system for HIV infection and expanded surveillance case definition for AIDS among adolescents and adults.Morb Mortal Wkly Rep
29. Statistics Netherlands. Available at: http://www.cbs.nl.
30. Van Houwelingen HC, Eilers PHC. Non-proportional hazards in models in survival analyis.
In: Bethlehem JG, Van der Heijden PGM (editors): COMPSTAT, Proceedings in Computational Statistics
. Physica-Verlag; 2000, pp. 151–160.
31. Rancinan C, Neau D, Savès M, Lawson-Ayayi S, Bonnet F, Marcié P, et al
. Is hepatitis C virus co-infection associated with survival in HIV-infected patients treated by combination antiretroviral therapy?AIDS
32. D'Arminio Monforte A, Cozzi Lepri A, Rezza G, Pezzotti P, Antinori A, Phillips AN, et al
. Insights into the reasons for discontinuation of the first highly active antiretroviral therapy (HAART) regimen in a cohort of antiretroviral naïve patients.AIDS
33. Mocroft A, Youle M, Moore A, Sabin CA, Madge S, Cozzi Lepri A, et al
. Reasons for modification and discontinuation of antiretrovirals: results form a single treatment centre.AIDS
34. Opravil M, Ledergerber B, Furrer H, Hirschel B, Imhof A, Gallant S, et al
. Clinical efficacy of early initiation of HAART in patients with asymptomatic HIV infection and CD4 cell count > 350 × 106/l.AIDS
Members of the ATHENA Project: Clinical and Epidemiological Working Group: W. Bronsveld, Medical Centre, Alkmaar; H. Weigel,* K. Brinkman, P. Frissen, Onze Lieve Vrouwe Gasthuis (OLVG), Amsterdam; J. ten Veen,* M. Hillebrand, S. Schieveld, OLVG, Prinsengracht; J. Mulder,* E. van Gorp, P. Meenhorst, Slotervaart Hospital, Amsterdam; A. van Eeden, Jan van Goyen Kliniek, Amsterdam; S. Danner,* F. Claessen,* R. Perenboom, Academic Hospital, Vrije Universiteit, Amsterdam; J. K. Eeftinck Schattenkerk, E. Gisolf, M. Godfried, J. van der Meer, J. Nellen, D. Notermans, T. van der Poll, M. van Praag, J. Prins, P. Reiss, M. Reijers, T. Ruys, M. van der Valk, A. Verbon, F. Wit, Academic Medical Centre, Amsterdam; C. Richter,* R. van Leusen, Hospital Rijnstate, Arnhem; R. Vriesendorp, Westeinde Hospital, The Hague; R. Kauffmann* and E. Kogger, Hospital Leyenburg, The Hague; B. Bravenboer, Catharina Hospital, Eindhoven; C. ten Napel,* K. Pogany, Medisch Spectrum Twente, Enschede; H. Sprenger,* G. Law, University Hospital, Groningen; R. W. ten Kate, Kennemer Gasthuis, Haarlem; M. Leemhuis, Medical Centre, Leeuwarden; F. Kroon,* E. Schippers, University Medical Centre, Leiden; G. Schrey,* S. van der Geest, A. van der Ven, University Hospital, Maastricht; P. Koopmans,* M. Keuter, D. Telgt, University Hospital, Nijmegen; M. van der Ende,* I. Gyssens, S. de Marie, Erasmus University Medical Centre, Rotterdam (EMCR); J. Juttmann,* C. van der Heul, St. Elisabeth Hospital, Tilburg; M. Schneider,* J. Borleffs, L. Hoepelman, C. Jaspers, University Medical Centre, Utrecht; W. Blok, Hospital Walcheren, Vlissingen. (*site co-ordinating physicians.)