Efforts from observational cohorts to transpose results into useful recommendations for clinical management of HIV-infected patients
From the INSERM Unit 593 (ex. U330), Bordeaux, France.
See also p. 2227
Correspondence to Geneviève Chêne, INSERM Unit 593 (ex. U330), 146 rue Léo-Saignat, Case 11, 33076 Bordeaux cedex, France. E-mail: Genevieve.firstname.lastname@example.org
Received: 1 April 2003; accepted: 8 May 2003.
With the availability of highly active antiretroviral therapy (HAART), people with HIV can expect longer survival and less clinical events. However, the rate of clinical progression in treated patients is still highly variable  and the determinants important to identify. For this purpose, large observational cohorts or collaboration of cohorts are very useful but have to take two recent trends into account in their modelling exercise: (1) causes of death have changed as now about one-half of deaths are not AIDS-related, and (2) treatment changes are frequent and the principle of intent-to-treat analysis might not reflect different treatment strategies over a long-term follow-up period.
In their study, van Sigham et al.  examined HIV disease progression in the Dutch ATHENA observational cohort of patients who started HAART. After having classified causes of death as HIV- or non-HIV-related, they showed that, up to the year 2000, only HIV-related mortality decreased. In 2000, HIV-related mortality (0.8 per 100 person-years) was similar to non HIV-related mortality (0.9 per 100 person-years). This confirms that HAART is a dominant prognostic factor for mortality at the population level. Consistently, large surveys on causes of death report that one-half of deaths are attributable to HIV or AIDS and another half to other causes [3,4]. The study of cause-specific mortality raises several comments. Firstly, from a methodological point-of-view, Van Sigham et al. used a consensus scoring by three experienced physicians based on clinical records. Others have proposed an algorithm based on the International Classification of Diseases (ICD)-10  also used for the coding of death certificates in national surveillance systems. In any case, epidemiologists should recommend the development of an internationally recognized, standardized algorithm with details on clinical and biological data required to classify all cases with good performance. The algorithm should aim at yielding the lowest proportion in the unknown category and the lowest proportion misclassified according to specific mortality (HIV and non-HIV). Second, when causes of death are correctly classified, the spectrum of non-HIV related causes of death appears to be very broad, including end-stage liver disease and other complications of hepatitis infections, all types of cancers and lung cancer in particular, cardiovascular diseases, accidents and suicide among the most frequently cited causes [3,4]. Regarding prognostic factors, hypotheses might well be very different for mortality due to end stage liver disease or to cardiovascular diseases, for example. Clinical records will have to accommodate the wide spectrum of possible hypotheses to take these prognostic factors into account and models used for prediction will have to consider all causes of death as potentially competitive. Finally, from a pragmatic point-of-view, one should never forget that HIV-infected people are, as individuals, interested in surviving as long as possible, not on surviving a specific cause of death while dying more rapidly due to another cause. Identifying precise causes of death and their prognostic factors might be the only way to provide clinicians and researchers with results specific enough to be useful for establishing priorities that usefully lead to improvements in clinical management. Therefore, large cohort studies of HAART-treated, HIV-infected patients should nowadays yield results on both overall and specific mortality.
In another part of their paper, the efforts made by Van Sigham et al. to analyse the effect of treatment interruptions on further disease progression should be acknowledged. Indeed, the analysis of cohort studies using the intent-to-treat principle is becoming less and less acceptable considering the frequency of treatment changes. Van Sigham et al. propose the use of duration under treatment as a time-dependent variable, namely the number of weeks in each 24-week period of follow-up after the start of HAART, the first 24-week period being considered entirely under treatment whether or not there is a treatment change. They show that, in comparison with patients who did not interrupt treatment, the rate of clinical progression after interrupting HAART for up to 8 weeks is doubled and the rate after interrupting HAART for more than 8 weeks is multiplied by five. This is expected since these treatments have proved to be highly active and any unstructured interruption is a marker of failure. Consistently, the authors are very cautious that no conclusion from these data can be drawn concerning structured treatment interruptions that are being evaluated in current clinical trials. This modelling exercise of treatment strategies should be viewed as a basis for more sophisticated analyses. Indeed, the treatment covariate was used as a surrogate marker of treatment failure that reflects either immuno-virological failure or toxicity. It is possible that different circumstances of interruptions have a different impact on clinical progression: intolerance of treatment being less deleterious than poor adherence or early virological failure . Therefore the reason for changing could also be an important indicator of prognosis and should be taken into account to avoid confusion by indication in the estimation of treatment effect.
Finally, the authors addressed the question of deferring treatment using model predictions. Since they had no data on deferred treatment in their cohort, these data should be interpreted with caution. Although it is obvious that the difference between continuous use of HAART and deferred treatment is small when patients have high CD4+ cell count, the result about treatment deferral until CD4+ counts as low as 110 × 106 cells/l is surprising. This result has to be confronted with a recent analysis of the MAC Study supporting the recommendation that HAART be started when CD4+ cell count reaches the threshold of 200 × 106 cells/l. In patients with CD4+ cell count higher than 200 × 106 cells/l, the recommendation to start is also dependent on HIV RNA .
HIV disease is a rather unique infectious disease as its evolution lasts over a very long period of time and is punctuated by complications that are not all attributable to HIV. What is the relative impact of HIV infection itself, ageing, co-morbidities such as HCV infection, behaviours such as smoking and adverse effects of HAART in the progression of treated HIV-infected people ? This question is at the cornerstone of clinical decisions for HIV-infected people: for example, if the impact of HIV infection itself is still important, we need to intensify HAART but if adverse effects of HAART are important, we need to defer HAART until the benefit : risk ratio changes in favour of the use of treatment. Large observational studies are particularly suited to add to this knowledge that should then be translated into priorities in the complex and comprehensive clinical case management of the people treated in developed countries. It should also be used to anticipate situations that developing countries will face now with the rapid introduction of HAART.
1. Egger M, May M, Chêne G, Phillips AN, Ledergerber B, Dabis F, et al
. and ART Cohort Collaboration. Prognosis of HIV-1-infected patients starting highly active antiretroviral therapy: a collaborative analysis of prospective studies. Lancet
2. van Sighem AL, van de Wiel MA, Ghani AC, Jambroes M, Reiss P, Gyssens IC, et al
. Mortality and progression to AIDS after starting highly active antiretroviral therapy. AIDS
3. Mocroft A, Brettle R, Kirk O, Blaxhult A, Parkin JM, Antunes F, et al
. Changes in the cause of death among HIV-positive subjects across Europe: results from the EuroSIDA study. AIDS
4. Lewden C, Bevilacquas S, Bonnet F, Heripret L, Salmon D, Morlat P, et al. Causes of death in HIV-infected adults in the era of HAART: the French Survey ‘Mortalité 2000'. Antivir Ther
5. Le Moing V, Chêne G, Leport C, Lewden C, Duran S, Garré M, et al
. Impact of discontinuation of initial protease inhibitor on further virological response in a cohort of human immunodeficiency virus-infected patients. Clin Infect Dis
6. Phair JP, Mellors JW, Detels R, Margolick JB, Munoz A. Virologic and immunologic values allowing safe deferral of antiretroviral therapy. AIDS
© 2003 Lippincott Williams & Wilkins, Inc.
What does "Remember me" mean?
By checking this box, you'll stay logged in until you logout. You'll get easier access to your articles, collections,
media, and all your other content, even if you close your browser or shut down your
To protect your most sensitive data and activities (like changing your password),
we'll ask you to re-enter your password when you access these services.
What if I'm on a computer that I share with others?
If you're using a public computer or you share this computer with others, we recommend
that you uncheck the "Remember me" box.
Data is temporarily unavailable. Please try again soon.