The life expectancy of HIV-infected individuals in the developed world has improved substantially since highly active antiretroviral therapy (HAART) became widely available from 1997 [1,2]. As HIV-infected persons live longer, they are increasingly more likely to die from non-HIV related causes [3,4], such as the consequences of co-infection with hepatitis viruses or aging, and increased exposure to HAART may place individuals at risk of death from therapy-related toxicities, including cardiovascular side-effects .
For example, the progression of liver disease associated with hepatitis C (HCV) infection is known to be accelerated in HIV-infected persons . Due to the increasing survival of HIV-infected persons in the HAART era, mortality from HCV may take on a greater significance among HIV-infected drug users and haemophiliacs, in whom HCV infection is common. Chronic hepatitis B infection has also been shown to increase mortality among HIV-infected persons [7,8].
In the pre-HAART era, particularly in injecting drug users (IDU) and homosexual men, the risk of pre-AIDS mortality increased as the CD4 cell count decreased and HIV RNA level increased . Among HIV-infected persons treated with HAART, the risk of death has also been shown to be associated with the CD4 cell count . However, it is not clear whether these relationships are the same for all causes of death, or whether any relationship with the CD4 cell count or HIV RNA level may simply reflect a longer duration of infection in those with more advanced disease. The CASCADE cohort collaboration, which includes a large group of HIV-infected individuals with reliable estimated dates of seroconversion from all risk groups, provides a unique opportunity to describe in detail the incidence of death from different causes over the first 20 years following the start of the HIV epidemic, adjusting for the duration of HIV-infection.
We evaluated the impact of the introduction of HAART on cause-specific mortality and investigated the relationships between CD4 cell counts and HIV RNA levels and specific causes of death in this large collaboration.
Data from a collaboration of 22 HIV seroconverter cohort studies in Europe, Canada and Australia (CASCADE: Concerted Action on SeroConversion to AIDS and Death in Europe) were used for this analysis.
The CASCADE collaboration, which includes HIV-1 infected persons where a date of seroconversion can be estimated reliably, has been described in detail elsewhere . Seroconverters were either enrolled into one of the constituent cohorts while HIV negative or else when already infected but with the ability to estimate HIV seroconversion through the availability of previously stored blood samples. The majority (85%) of seroconversion dates were calculated as the mid-point between the last negative and first positive HIV antibody tests; individuals in whom this period was greater than 3 years were excluded from current analysis (91% having an interval less than 2 years). Analyses were restricted to individuals from 19 cohorts that collected information on specific causes of death (COD). Children under 15 year of age at seroconversion were excluded.
Specific causes of death
Causes of death were grouped into 11 categories. These included three AIDS-related categories [AIDS-related malignancy, AIDS opportunistic infections (OI), AIDS/HIV-related unspecified] and seven non-AIDS-related causes [hepatitis/liver-related, other infections, organ failure, non-AIDS-related malignancy, (un)-intentional death (including accidents, suicide, overdose), cardiovascular disease (CVD)/diabetes mellitus and other (including epilepsy, fits, blood disorders)]. The final category consisted of the unknown COD, which could include AIDS-related as well as non-AIDS-related deaths. Most cohorts reported the primary COD for individuals who died, while some reported multiple COD. When more than one COD was given, the most likely underlying COD was scored independently by a panel of three physicians. Causes about which the physicians disagreed were reviewed by the panel until a final consensus was reached.
Follow-up was calculated from the estimated date of seroconversion until the earliest of: death, loss to follow-up, the cohort censoring date (each cohort had their own censoring date reflecting the processes and timelines in place for ascertaining events) or 31 December 2003. Those who were enrolled retrospectively into the constituent cohorts were included in the risk set from the date of cohort enrolment (i.e. a correction for left truncation was applied). We used calendar time as a proxy for the introduction and widespread use of HAART and defined two calendar periods, pre-1997 and 1997 onwards, to reflect the pre-HAART and HAART eras, respectively.
The cumulative incidences of all 11 categories of COD were calculated within a competing risks framework, for the pre-HAART and HAART eras separately . Non-parametric estimates of cumulative transition probabilities from seroconversion to one of the competing COD were calculated using the multiple decrement model transition programme.
We also performed multivariate regression analyses for the rate of progression to the different COD [11,13] taking into account competing risks, allowing for late entry and adjusted for potential confounders. An administrative censoring time was created for each person who died of a COD that was not the COD of interest as this would be the date the person would have theoretically been alive at. Therefore, persons were censored at 1 January 1997 if they had died before that date from a COD that was not the COD of interest or at the cohort censoring date if they had died after it.
We evaluated the effect of calendar time within each risk group, as well as the interaction between risk group and calendar time for each COD. Hazards ratios were only calculated for COD with more than five deaths. To reduce the chance of a type 1 error as a result of multiple testing, we also calculated 99% confidence intervals (CI). All models were adjusted for sex, age at seroconversion, and presentation during acute HIV infection (defined when the interval between the last HIV-negative and first HIV positive test was less than 1 month ).
To evaluate the impact of HAART on the relationships between the CD4 cell count, HIV RNA level and each specific COD, the cause-specific hazard ratios (CHR) were estimated using a Cox proportional hazards model for progression from HIV seroconversion to each cause. In this analysis these markers were evaluated as continuous variables (after square-root and logarithmic transformation, respectively) and were treated as time-dependent covariates. If the interval between two consecutive marker measurements or between the last measurement and either death or end of follow-up exceeded 18 months, individuals were temporarily removed from the risk set from the 18-month point until the date of their next available measurement.
Finally, a sensitivity analysis was conducted by excluding data from all cohorts in which more than 25% of the COD were unknown.
The median age of the 7680 seroconverters included in the analyses was 29 years [interquartile range (IQR), 24–35] and 81% were male. Sex between men was the most frequent exposure category (49%) followed by IDU (25%), sex between men and women (15%) and haemophilia (3%). A total of 1938 of the 7680 seroconverters died during follow up (26%); a specific COD was available for 1437 of these deaths (74%, Table 1). Of these 1938 deaths, 248 patients had more than one COD and 1424 of the deaths (72%) occurred in individuals who had been diagnosed with AIDS.
The cumulative incidence of all cause mortality was 0.10 at 5 years and 0.44 at 10 years following seroconversion in the pre-HAART era. The cumulative incidences of all cause mortality were 0.05 at 5 years and 0.15 at 10 years in the HAART era (Fig. 1). For those who became infected after the introduction of HAART, and were, therefore, potentially able to fully benefit from HAART, the 5-year cumulative incidence of all cause mortality was 0.02.
Cumulative incidences for each specific COD are shown in Fig. 2a and b. In the pre-HAART era, AIDS OI was the most common cause, followed by unknown causes, HIV/AIDS not further specified and other infections. A small proportion (3% at 15 years following seroconversion, 95% CI, 2–4) died from hepatitis/liver-related causes (Fig. 2a).
In contrast, in the HAART era (Fig. 2b), unknown deaths was the most common category. All AIDS-related COD decreased by at least 50% in the HAART era. At 15 years after seroconversion, the cumulative probability of dying from AIDS OI decreased from 20% (95% CI, 17–23) in the pre-HAART era to 5% in the HAART era (95% CI, 4–6). However, AIDS OI remained the most important known COD. HIV/AIDS not further specified became a less important COD in the HAART era. Although the percentage of persons dying from AIDS-related malignancies is small, it also halved in the HAART era, in comparison with the pre-HAART era. Although the probability of dying from hepatitis or liver-related causes did not change in the era after HAART was introduced, the probability of dying from other infections decreased from 8% at 15 years following seroconversion (95% CI, 6–9) in the pre-HAART era to 3% (95% CI, 3–4) in the HAART era. Death from organ failure also showed a large reduction, from 1% in the pre-HAART era to 0.4% in the HAART era.
We compared the risk of dying from each specific COD in the pre-HAART era with the risk of dying from that COD in the HAART era. The effect of calendar time was not the same for each risk group. Therefore, the relative hazards for each risk group are shown separately in Fig. 3. Within each risk group the risk of dying in the HAART era is compared with the risk of dying from the same cause in the pre-HAART era (left column of Fig. 3). We also compared the risk of dying among IDUs, heterosexuals and haemophiliacs to that among men who have sex with men (MSM), separately for the pre-HAART and HAART eras (middle and right column).
Changes in risks of dying in the HAART era within each risk group
Among MSM, there was a reduction in the cumulative risk of dying from most COD in the HAART era in comparison with the pre-HAART era. Large cumulative risk reductions were seen for AIDS-related malignancies, AIDS OI and the unknown categories. All reductions in cumulative risks were significant at the 5% level, with the exception of organ failure. Although not significant, the cumulative risk of dying from organ failure strongly decreased. However, a small non-significant increase in the cumulative risk of dying from hepatitis/liver-related causes and CVD/diabetes mellitus was seen.
Among IDU, a significant reduction was observed in the cumulative risk of death from AIDS OI and a large significant reduction was seen in the cumulative risk of death from non-AIDS-related malignancy. The cumulative risks of dying from AIDS-related malignancies, other infections and organ failure were all reduced, but these reductions were not significant. For the organ failure category, the cumulative risk strongly reduced. However, the cumulative risk of dying from (un)-intentional death increased significantly, whereas a non-significant increase in the cumulative risk of dying from hepatitis/liver-related causes was seen.
In heterosexuals, strong reductions in the cumulative risk of dying from AIDS-related malignancies, AIDS OI, other infections, hepatitis/liver-related death and (un) intentional death were seen, although the effect was not significant for AIDS-related malignancies and hepatitis/liver-related causes. A non-significant increase was seen for the cumulative risks of dying from non-AIDS-related malignancies and CVD/diabetes mellitus.
Finally, among haemophiliacs, no significant reductions or increases in the risks of dying were seen, although there was a large increase in the cumulative risk of dying from hepatitis/liver-related causes, organ failure and non-AIDS-related malignancies.
Comparison of the risk of death between risk groups
In the pre-HAART era, we observed a significantly higher cumulative risk of death among IDUs in comparison with MSM, from AIDS OI, hepatitis or liver-related deaths, other infections and (un)-intentional death. A significantly cumulative lower risk was observed for AIDS and non-AIDS-related malignancies, unspecified HIV/AIDS-related deaths and unknown in IDUs. In the HAART era, IDU had a higher cumulative risk of death in comparison with MSM for all COD, except non-AIDS-related malignancies. The cumulative risks of death from AIDS OI, hepatitis/liver-related causes, (un)-intentional causes and unknown COD were significantly higher. Pre-HAART, heterosexuals had a significantly higher cumulative risk of death from other infections in comparison with MSM, whereas no significant differences were seen in the cumulative risks of death from other specific COD. In the HAART era, no significant differences were found in the cumulative risks of each specific COD between heterosexuals and MSM.
In the pre-HAART era, haemophiliacs had a significantly lower cumulative risk of death from AIDS-related malignancies, AIDS OI, other infections and unknown COD in comparison with MSM. The cumulative risk of dying from unspecified HIV/AIDS-related causes, hepatitis/liver-related causes and CVD/diabetes mellitus increased. After HAART became available, the cumulative risk of death from hepatitis/liver-related causes remained significantly higher.
Association between laboratory markers and specific causes of death
All COD showed a significant relationship with lower CD4 cell counts, except (un)-intentional death in the pre-HAART era (Table 2).
For almost all COD, the relationship with the CD4 cell count did not differ significantly in the HAART era, compared to the pre-HAART era. However, the relationships between the CD4 cell count and the risk of dying from unknown COD was significantly weaker in the HAART era.
Fewer HIV RNA data were available in the pre-HAART era, due to the lack of availability of these tests at that time. None of the relationships between the COD and higher HIV RNA levels differed between the pre-HAART and HAART eras, with the exception of the relationship between HIV RNA levels and unknown COD, which was weaker in the era of HAART (Table 2). The relationship between organ failure and HIV RNA levels became stronger in the era of HAART (P = 0.05).
This study, conducted among a large group of HIV-infected individuals with well-estimated dates of seroconversion followed in the pre-HAART and HAART eras, demonstrates clearly that the probability of dying from most causes of death decreased substantially after HAART became widely available. Nevertheless, although the cumulative incidence of death from all AIDS-related causes decreased in the era of HAART, AIDS OI still remained the most common COD.
Although individuals in the pre-HAART era may have been infected with HIV for longer periods of time than those in the HAART era, it is unlikely that our findings can be explained by differences in duration of infection between those followed in the in the two time periods because we were able to adjust our analyses for the time since seroconversion.
However, the population in which mortality can be estimated in the first years following the introduction of HAART does inevitable include many individuals infected in the pre-HAART era, some of whom would have had limited opportunity to benefit from HAART, for example due to very low CD4 cell counts at HAART initiation  or prior use of sub-optimal treatments . For those who did become infected in the HAART era, however, who had the opportunity to fully benefit from HAART, we found that mortality was even lower than the wider population at risk in the HAART era, suggesting that further reductions in mortality in HIV-infected persons may emerge over time .
A reduction in AIDS-related mortality at the population level following the widespread use of HAART has been shown in earlier studies [17–21]. However, our study is the first to show a general reduction in the risk of dying from specific causes in seroconverters in the era of HAART. Interestingly, the benefits of HAART appear to be less strong in IDU across the specific causes. Furthermore, the cumulative risk of dying from (un) intentional causes significantly increased among IDU in the HAART era. One reason could be temporal changes in other risk factors (e.g. injecting behaviour). However, as fewer IDU are now dying from the other specific causes investigated in our study, more IDU remain at risk of dying from (un) intentional causes, therefore in the HAART era more IDU will die from this cause leading to an increase in the cumulative risk. In line with a study from the pre-HAART era these deaths are not associated with having lower CD4 cell counts . All other COD, including non-AIDS-related COD, were associated with lower CD4 cell counts, with virtually no differences in the association with CD4 cell count between the HAART and pre-HAART eras. Hence all COD, other than (un)intentional deaths, appeared to be related in some way to the underlying immunodeficiency caused by HIV infection regardless of their putative association with AIDS itself. Similarly, the relationship between HIV RNA levels and the risk of all COD remained virtually unchanged in the HAART era, although this relationship did become slightly weaker for unknown COD, perhaps suggesting that many unknown causes in the pre-HAART era were, in fact, HIV-related.
Other studies have shown that hepatitis/liver-related deaths and malignancies are now the most common non-AIDS-related COD among HIV-infected individuals . Among HIV-infected individuals who died, the proportion of deaths accounted for by these COD and CVD have increased since 1996 . Although we found that the risk of dying from CVD increased, (un)-intentional death, other infections and hepatitis/liver-related deaths were the most common non-AIDS-related COD in the HAART era in our study. We believe this highlights one of the limitations of conducting studies based only on persons who have died, since these cannot account for the substantially larger number of persons surviving in the HAART era. Our study shows a trend towards a higher probability of dying from hepatitis/liver-related causes in the HAART era among most risk groups, in line with other studies [24,25], and most likely because large numbers of patients survive to experience this competing COD. However, this finding is not supported by Qurishi et al. , most likely due to survivorship bias not accounted for in their analyses .
Since HAART became generally available, several studies have demonstrated reductions in the incidence of AIDS-related malignancies among HIV-infected persons [28,29] and we found parallel reductions in the incidence of death from AIDS-related malignancies. However, we observed a small non-significant increase in the risk of death from non-AIDS-related malignancies in the HAART era among heterosexuals in our study. Furthermore, although cancer rates among those with HIV have not markedly increased in the HAART era , HIV-infected individuals do have an increased risk of non-AIDS-related cancers in comparison with the general population .
One limitation of our study is the lack of uniform classification of COD within cohorts that participate in the CASCADE collaboration with each cohort using their own classification. Differences in COD distribution were observed when comparing the cohorts that reported one COD with the cohorts that reported multiple causes. In particular, cohorts with one COD were more likely to report unspecified AIDS/HIV-related causes than cohorts with multiple COD, suggesting that deaths were allocated a COD according to a hierarchy. This difference in process might cause some bias and it is possible that the proportion of deaths due to AIDS/HIV-related causes may be underestimated. A recent study has shown that the lack in precise information about the COD might result in misinterpretation of results , so a standardized classification of COD among HIV-infected persons is needed The recent CoDe project provides such a uniform system for collecting and classifying the COD (www.cphiv.dk/CoDe) and this may help to further improve the quality of reporting in the future and allow monitoring of trends in COD. Another limitation of our study is the relatively large proportion of deaths from unknown causes. However, a sensitivity analysis restricted to cohorts in which less than 25% of the deaths were from unknown causes showed no major differences in findings, except that in the era of HAART, the cumulative incidence of hepatitis/liver-related deaths became somewhat higher. This suggests that hepatitis/liver-related deaths might become even more important as classification of deaths improves in the future.
In conclusion, both overall mortality as well as cause-specific mortality has substantially decreased following the introduction of HAART. However, AIDS OI remain the most common COD in the era of HAART, suggesting that AIDS-related events will continue to be an important cause of death in the future. Since the time after HAART became available is limited and the majority of HIV-infected individuals in our study became infected in the pre-HAART era, it might be too early to observe an increase in non-AIDS-related death due to therapy-related toxicities. Therefore monitoring mortality trends among HIV-infected individuals using standardized international guidelines for the coding of deaths among HIV-infected persons will be essential for the future. Furthermore, the development of prevention measures and treatment guidelines for non-AIDS-related conditions aimed specifically at HIV-infected individuals is warranted.
The authors wish to thank the panel of physicians: Court Pedersen, Karen Lindenburg, Cees Das.
1. Porter K, Babiker A, Bhaskaran K, Darbyshire J, Pezzotti P, Porter K, et al
. Determinants of survival following HIV-1 seroconversion after the introduction of HAART. Lancet 2003; 362:1267–1274.
2. Palella FJ, Delaneny 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 1998; 338:853–860.
3. 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 1995–1999. Clin Infect Dis 2001; 32:1487–1493.
4. 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 2002; 16:1663–1671.
5. Friis-Moller N, Weber R, Reiss P, Thiebaut R, Kirk O, d'Arminio MA, et al
. Cardiovascular disease risk factors in HIV patients–association with antiretroviral therapy. Results from the DAD study. AIDS 2003; 17:1179–1193.
6. Telfer P, Sabin C, Devereux H, Dusheiko G, Lee C. The progression of HCV-associated liver disease in a cohort of haemophilic patients. Br J Haematol 1994; 87:555–561.
7. Konopnicki D, Mocroft A, de Wit S, Antunes F, Ledergerber B, Katlama C, et al
. Hepatitis B and HIV: prevalence, AIDS progression, response to highly active antiretroviral therapy
and increased mortality in the EuroSIDA cohort. AIDS 2005; 19:593–601.
8. Thio CL, Seaberg EC, Skolasky R Jr, Phair J, Visscher B, Munoz A, et al
. HIV-1, hepatitis B virus, and risk of liver-related mortality in the Multicenter Cohort Study (MACS). Lancet 2002; 360:1921–1926.
9. Prins M, Sabin CA, Lee CA, Devereux H, Coutinho RA. Pre-AIDS mortality and its association with HIV disease progression in haemophilic men, injecting drug users and homosexual men. AIDS 2000; 14:1829–1837.
10. Lewden C, Raffi F, Cuzin L, Cailleton V, Vilde JL, Chene G, et al
. Factors associated with mortality in human immunodeficiency virus type 1-infected adults initiating protease inhibitor-containing therapy: role of education level and of early transaminase level elevation (APROCO-ANRS EP11 study). The Antiproteases Cohorte Agence Nationale de Recherches sur le SIDA EP 11 study. J Infect Dis 2002; 186:710–714.
11. Babiker A, Darbyshire J, Pezzotti P, Porter K, Rezza G, Walker SA, et al
. Changes over calendar time in the risk of specific first AIDS-defining events following HIV seroconversion, adjusting for competing risks. Int J Epidemiol 2002; 31:951–958.
12. Tai BC, Machin D, White I, Gebski V. Competing risks analysis of patients with osteosarcoma: a comparison of four different approaches. Stat Med 2001; 20:661–684.
13. Fine JP, Gray RJ. A proportional hazards model for the subdistribution of a competing risk. J Am Stat Assoc 1999; 446:496–509.
14. CASCADE. The relationships between the HIV test interval, demographic factors and HIV disease progression.Epidemiol Infect
15. Mocroft A, Ledergerber B, Katlama C, Kirk O, Reiss P, d'Arminio MA, et al
. Decline in the AIDS and death rates in the EuroSIDA study: an observational study. Lancet 2003; 362:22–29.
16. Yamashita TE, Phair JP, Munoz A, Margolick JB, Detels R, O'Brien SJ, et al
. Immunologic and virologic response to highly active antiretroviral therapy
in the Multicenter AIDS Cohort Study. AIDS 2001; 15:735–746.
17. Smit C, Geskus R, Uitenbroek D, Mulder D, Van den Hoek A, Coutinho RA, et al
. Declining AIDS mortality in Amsterdam: contributions of declining HIV incidence and effective therapy. Epidemiology 2004; 15:536–542.
18. Chiasson MA, Berenson L, Li W, Schwartz S, Singh T, Forlenza S, et al
. Declining HIV/AIDS mortality in New York City. J Acquir Immune Defic Syndr 1999; 21:59–64.
19. Hogg RS, O'Shaughnessy MV, Gataric N, Yip B, Craib K, Schechter MT, et al
. Decline in deaths from AIDS due to new antiretrovirals. Lancet 1997; 349:1294.
20. Keiser O, Taffe P, Zwahlen M, Battegay M, Bernasconi E, Weber R, et al
. All cause mortality in the Swiss HIV Cohort Study from 1990 to 2001 in comparison with the Swiss population. AIDS 2004; 18:1835–1843.
21. Detels R, Munoz A, McFarlane G, Kingsley LA, Margolick JB, Giorgi J, et al
. Effectiveness of potent antiretroviral therapy on time to AIDS and death in men with known HIV infection duration. Multicenter AIDS Cohort Study Investigators. JAMA 1998; 280:1497–1503.
22. Prins M, Hernandez A, I, Brettle RP, Robertson JR, Broers B, Carre N, et al
. Pre-AIDS mortality from natural causes associated with HIV disease progression: evidence from the European Seroconverter Study among injecting drug users.AIDS
23. Selik RM, Byers RHJ, Dworkin MS. Trends in diseases reported on U.S. death certificates that mentioned HIV infection, 1987–1999. J Acquir Immune Defic Syndr 2002; 29:378–387.
24. Serraino D, Boschini A, Carrieri P, Pradier C, Dorrucci M, Dal Maso L, et al
. Cancer risk among men with, or at risk of, HIV infection in southern Europe. AIDS 2000; 14:553–559.
25. Bica I, McGovern B, Dhar R, Stone D, McGowan K, Scheib R, et al
. Increasing mortality due to end-stage liver disease in patients with human immunodeficiency virus infection. Clin Infect Dis 2001; 32:492–497.
26. Qurishi N, Kreuzberg C, Luchters G, Effenberger W, Kupfer B, Sauerbruch T, et al
. Effect of antiretroviral therapy on liver-related mortality in patients with HIV and hepatitis C virus coinfection. Lancet 2003; 362:1708–1713.
27. Sabin CA, Walker AS, Dunn D. HIV/HCV coinfection, HAART, and liver-related mortality. Lancet 2004; 364:757–758.
28. International Collaboration on HIV and Cancer. Highly active antiretroviral therapy and incidence of cancer in human immunodeficiency virus-infected adults.J Natl Cancer Inst
29. Clifford GM, Polesel J, Rickenbach M, Dal Maso L, Keiser O, Kofler A, et al
. Cancer risk in the Swiss HIV Cohort Study: associations with immunodeficiency, smoking, and highly active antiretroviral therapy
.[see comment]. J Natl Cancer Inst 2005; 97:425–432.
30. Grulich AE, Li Y, McDonald A, Correll PK, Law MG, Kaldor JM. Rates of non-AIDS-defining cancers in people with HIV infection before and after AIDS diagnosis. AIDS 2002; 16:1155–1161.
31. Soriano V, Garcia-Samaniego J, Valencia E, Rodriguez-Rosado R, Munoz F, Gonzalez-Lahoz J. Impact of chronic liver disease due to hepatitis viruses as cause of hospital admission and death in HIV-infected drug users. Eur J Epidemiol 1999; 15:1–4.
32. Mocroft A, Gatell J, Reiss P, Ledergerber B, Kirk O, Vella S, et al
. Causes of death in HIV infection: the key determinant to define the clinical response to anti-HIV therapy. AIDS 2004; 18:2333–2337.
Analysis and writing committee: Colette Smit, Ronald Geskus, Sarah Walker, Caroline Sabin, Roel Coutinho, Kholoud Porter, Maria Prins.
Steering committee: Valerie Beral, Roel Coutinho, Janet Darbyshire (project leader), Julia Del Amo, Noël Gill (chairman), Christine Lee, Laurence Meyer, Giovanni Rezza.
Co-ordinating centre: Kholoud Porter (scientific co-ordinator), Abdel Babiker, A. Sarah Walker, Janet Darbyshire, Krishnan Bhaskaran.
Collaborators: Aquitaine cohort, France: Francois Dabis, Rodolphe Thiébaut, Geneviève Chêne, Sylvie Lawson-Ayayi; SEROCO cohort, France: Laurence Meyer, Faroudy Boufassa; German cohort, Germany: Osamah Hamouda, Gabriele Poggensee; Italian Seroconversion Study, Italy: Benedetta Longo, Patrizio Pezzotti, Giovanni Rezza; Greek Haemophilia cohort, Greece: Giota Touloumi, Angelos Hatzakis, Anastasia Karafoulidou, Olga Katsarou; Edinburgh Hospital cohort, United Kingdom: Ray Brettle; Madrid cohort, Spain: Julia Del Amo, Jorge del Romero; Amsterdam Cohort Studies among homosexual men and drug users, the Netherlands: Liselotte van Asten, Akke van der Bij, Ronald Geskus, Maria Prins, Roel Coutinho; Copenhagen cohort, Denmark: Court Pedersen; Valencia IDU cohort, Spain: Ildefonso Hernández Aguado, Santiago Pérez-Hoyos; Oslo and Ulleval Hospital cohorts, Norway: Anne Eskild, Johan N Bruun, Mette Sannes; Royal Free haemophilia cohort, United Kingdom: Caroline Sabin, Christine Lee; UK Register of HIV Seroconverters, United Kingdom: Anne M Johnson, Andrew N Phillips, Abdel Babiker, Janet H Darbyshire, Noël Gill, Kholoud Porter; Swiss HIV cohort, Switzerland: Patrick Francioli, Philippe Vanhems, Matthias Egger, Martin Rickenbach; Sydney AIDS Prospective Study, Australia: David Cooper, John Kaldor; Sydney Primary HIV Infection cohort, Australia: David Cooper, John Kaldor, Tim Ramacciotti, Don Smith; Badalona IDU hospital cohort, Spain: Roberto Muga, Jordi Tor; Lyon Primary Infection cohort, France: Philippe Vanhems; South Alberta clinic, Canada: John Gill; Barcelona IDU cohort, Spain: Joan Cayla, Patricia Garcia de Olalla.