Up to September of 2001, 222 356 cases of AIDS had been reported in Brazil . Initially, most cases were among men who have sex with men and persons with blood exposure through needle sharing or receipt of blood products. More recently, heterosexual transmission has become increasingly important, with nearly as many cases in women as men in some areas, and poorer segments of the population have been increasingly affected [2,3]. Since the mid-1980s, Brazil has benefited from strong efforts in both the governmental and non-governmental sectors in the areas of prevention, surveillance, and care for HIV/AIDS.
Brazil has led the way among developing countries in providing access to antiretroviral drugs for its entire population through its national health care system (Sistema Único de Saude-SUS) that was established in 1988. As early as 1991, zidovudine was made available through the public health care system, followed by several other reverse transcriptase inhibitors in 1993. Beginning in late 1996, protease inhibitors were also added to the public formulary. More recently, new drugs have become available through public clinics in Brazil at about the same time that they are released on the international market. Provision of drugs was accompanied by a national laboratory network that began providing general access to CD4 testing in 1997, viral load testing in 1998, and genotypic testing for drug resistance in 2002. Currently, about 115 000 persons with AIDS are receiving antiretroviral treatment.
A developing country such as Brazil has only been able to manage the high cost of providing antiretroviral treatment through a combination of political commitment and the adoption of controversial policies such as local generic drug production and aggressive negotiation for deep discounts in the price of imported drugs. These policies have produced reactions ranging from condemnation to admiration. Besides issues of international trade relations, many have questioned whether it is possible for a developing country that lacks an ideal health infrastructure to provide antiretroviral treatment effectively on a broad scale. Not only can these drugs have significant toxicity, but their effective use requires sophisticated laboratory monitoring, high levels of consistent drug availability, and adherence to complicated treatment regimens to produce a full therapeutic effect and minimize the emergence of drug resistance.
Studies in developed countries have demonstrated substantial benefits from antiretroviral treatment in terms of survival and quality of life for patients with AIDS [4–12]. No studies have yet demonstrated whether similar gains can be achieved on a national level in a developing country. The last major study of survival among adult Brazilian AIDS patients showed that those diagnosed prior to 1989 had a median survival of only 5.1 months [13,14]. We undertook a national study of adults diagnosed with AIDS in Brazil in 1995 and 1996 to determine whether and to what extent survival time has increased with widespread access to antiretroviral drugs. A second goal was to examine the predictors of survival for Brazilian AIDS patients in the current setting of free universal access to treatment.
This retrospective cohort study examined survival among patients diagnosed with AIDS in 1995 and 1996. These two years of diagnosis were chosen as the earliest in which a substantial proportion of patients would have received highly active antiretroviral therapy and the latest for which follow-up time would be sufficient to determine median survival. The sample was drawn from 40 587 adult AIDS cases that had been reported to the Brazilian National AIDS Control Program as having been diagnosed in 1995 and 1996 at the time that the study started (this number has since grown to 41 622) . First, the sample was stratified based on four macro-regions of Brazil: north-east, south, south-east, and other (north and central-east). We purposively selected seven states to represent these four regions, and 18 cities within these states. Cities were chosen to give a mix of state capitals and smaller cities. Cities were excluded from consideration if they had less than 40 reported AIDS cases per year. We also examined the quality and completeness of AIDS reporting data received at the national level and excluded some locations from potential inclusion when it appeared that poor reporting might compromise the quality of the study.
Cases in these 18 cities were randomly selected at the central level for inclusion in the final sample based on the following criteria. First, they needed to reside in the selected city and be at least 13 years old at the time of AIDS diagnosis, which had to be in 1995 or 1996. Cases with a date of death listed in the national database that was less than 7 days from the date of diagnosis (including those that were diagnosed at the time of death) were excluded because these cases provided limited data on the potential impact of antiretroviral treatment. Sampling fractions in the selected cities were weighted to provide a demographically representative sample of all reported Brazilian cases in terms of macro-region and distribution between larger and smaller cities. Sampling fractions ranged from 100% in some of the smaller cities to 11% in Sao Paulo.
We selected 3930 cases for medical record review. Prior to conducting record reviews, the study protocol was approved by the ethical review board of the Sao Paulo State STD/AIDS Reference and Training Center (CRT-DST/AIDS), following standards of the National Commission on Ethics in Research (CONEP). Medical record reviews were carried out by senior health professionals in each selected city, who received standardized training based on a detailed field manual describing all study procedures. Data collection took place between April 2000 and January 2002. Data collected in the chart review included a revised date of diagnosis. This was defined as the date at which the patient first met any of the following AIDS diagnostic criteria that are used in Brazil: (1) the CDC modified case definition of 1992; (2) the Rio de Janeiro/Caracas case definition (also known as the PAHO-Caracas case definition); or (3) a CD4 cell count of less than 350 × 106 cells/l . Data were also collected on socio-demographics, transmission category, opportunistic diseases diagnosed at the time of AIDS presentation, laboratory data, and types of antiretroviral and prophylactic treatment received throughout the course of care. Date of death was based on the medical record, local AIDS reporting data, and the national surveillance database. Date of last clinical follow-up was used as a date of censor for all patients not known to have died.
All medical record abstraction forms were reviewed by trained field supervisors on the local level and by the principal investigator upon receipt at the central level. Incomplete or improbable data were returned for verification prior to entry. Data were entered throughout the data collection period by trained personnel using EpiInfo version 6.0  and were subjected to range and logic checks.
Of 3930 cases in the original sample, a total of 1109 were excluded for the following reasons:
252 had a revised date of AIDS diagnosis based on medical record review that was before or after the 1995–1996 period of study.
65 were found never to have met the diagnostic criteria for AIDS.
110 did not survive at least 7 days after the revised date of diagnosis.
20 were duplicates.
292 cases had no medical record in the city where they were reported.
165 had medical records that did not include sufficient data to complete the chart review form.
160 were excluded because data collection had not been completed by the end date of the study.
45 cases were excluded for other miscellaneous reasons.
This left 2821 cases that were included in the statistical analysis, which was conducted using STATA version 7.0 . Survival curves, survival at various intervals, and median survival were calculated using the Kaplan–Meier method. Univariate and multivariate predictors of survival were examined using Cox proportional hazards analysis. All variables associated with survival in univariate analysis (likelihood ratio test P < 0.05) were entered into the multivariate model with the exception of CD4 cell count. This was excluded because it was often missing, and its inclusion would have eliminated too many cases from the model.
Log–log plots and Schoenfeld residuals were used to check the validity of the proportional hazards assumption. Exposure category, treatment, year of diagnosis, and Pneumocystis carinii pneumonia (PCP) prophylaxis violated this assumption; the hazard ratios for these variables should thus be interpreted as the average effect (geometric mean) over the entire period of follow-up . For the multivariate analysis, we examined various models incorporating time-dependent covariates for variables violating the proportional hazards model. None of these substantially changed the association between the remaining predictors and survival; we therefore present only the simpler multivariate model without time-dependent covariates.
Data from the study of cases diagnosed in 1995–1996 were compared with data from a previous national study of survival among 2135 adult Brazilian AIDS patients diagnosed in 1982–1989. The methods for this study have been described previously . In brief, this study used a national sample of cases that was not limited to selected cities. Medical records were reviewed by local personnel without any special training; data were obtained for 68% of cases in the original sample. Cases were included whether or not they were diagnosed at the time of death or had a minimum survival time. We recalculated the survival curves presented in this article using the original data. We also conducted a separate calculation of median survival excluding cases that died within 7 days of diagnosis to determine the extent to which this difference in inclusion criteria affected the results.
Characteristics of cases
Of 2821 AIDS patients studied, 763 (27%) were female. The median age was 33 years, and 2127 (75%) had elementary education or less. The largest transmission category was heterosexual, followed by blood transmission (90% of which was injection drug use). Most cases were diagnosed as AIDS based on the PAHO-Caracas case definition, with the most common defining opportunistic infections being tuberculosis, pneumocystis pneumonia, and neurotoxoplasmosis. Only 30% had a documented CD4 cell count within 3 months of the diagnosis of AIDS. The proportion of patients receiving antiretroviral treatment before meeting diagnostic criteria for AIDS was 9.6% in 1995 and 15.6% in 1996 (Table 1).
The largest group of patients (38%) received triple antiretroviral therapy (26% of those diagnosed in 1995 and 48% of those diagnosed in 1996). Most of the remaining patients received double therapy or monotherapy that did not include a protease inhibitor. Rates of prophylaxis for PCP and tuberculosis documented in the medical record were 44 and 5%, respectively, for all patients; we did not attempt to determine for which patients such prophylaxis was clinically indicated. During the period of follow-up, 1339 patients were known to have died.
Changes in survival over time
AIDS cases diagnosed in 1995 had a median survival of 18 months. Survival at 1 and 2 years were 60 and 45%, respectively. Cases diagnosed in 1996 had much better survival: median survival was 58 months with 1- and 2-year survival of 72 and 63%. For cases diagnosed in 1995 and 1996 combined, the median survival was 36 months (Fig. 1).
The study of cases diagnosed in 1982 to 1989 found a median survival of 5.1 months and survival at 1 and 2 years of 32 and 21%, respectively. When we re-analyzed these data excluding cases that did not survive at least 7 days after diagnosis (for better comparability with the 1995–1996 data), median survival increased to 6.8 months.
Predictors of survival
In univariate analysis, the type of antiretroviral treatment received was a very strong predictor of survival. Demographic predictors of longer survival included female sex, higher level of education, and a transmission category other than blood transmission or unknown. Other predictors of longer survival included being diagnosed in 1996 rather than 1995, being diagnosed with AIDS based on CD4 cell count, having a higher CD4 cell count at diagnosis, and receiving PCP prophylaxis. Age and presenting diagnosis were not significant predictors of survival (Table 2).
The multivariate analysis yielded fewer independent predictors of survival. Type of antiretroviral treatment remained a strong predictor of survival. Criteria for AIDS diagnosis and transmission category also remained significant predictors.
This national study of survival among adult Brazilian AIDS patients diagnosed in 1995 and 1996 demonstrates a substantial increase in survival time compared to AIDS patients diagnosed in the 1980s. Most of this increase appears to have taken place recently, with patients diagnosed in 1996 having a median survival three times longer than those diagnosed in 1995. This improvement coincides with the widespread availability of triple antiretroviral treatment in Brazil. Type of antiretroviral treatment was a very strong predictor of survival. Its inclusion in the multivariate model (along with other predictors) increased the P-value for year of diagnosis from < 0.001 to 0.467, suggesting that the improvement in survival between 1995 and 1996 was mostly or entirely due to antiretroviral treatment. Although no other recent studies have directly measured the survival time of Brazilian AIDS patients, these results are consistent with the perceptions of health care professionals involved in the care of AIDS patients and the documented recent decline in deaths attributed to AIDS in Brazil [20,21].
The demographic and clinical characteristics of the patients studied are of interest because this study is the first in Brazil to verify such data for a large national sample of AIDS patients based on medical record review. When compared with earlier studies, our results confirm that Brazilian AIDS cases are increasingly likely to be heterosexually transmitted and to occur among females and persons of low education [2,3]. Tuberculosis has replaced Pneumocystis as the most common presenting diagnosis, perhaps in part because of increasing rates of prophylaxis for Pneumocystis among HIV-infected patients prior to AIDS diagnosis. Relatively few cases were diagnosed based on CD4 cell count because this test was not yet routinely available in 1995 and 1996.
Our results are similar to those from many industrialized countries that have demonstrated substantial increases in survival among AIDS patients since 1995. For example, AIDS patients diagnosed in San Francisco in 1995 to 1996 had a median survival of 31 month [22,23]. AIDS patients diagnosed in Australia in 1996 had a median survival of 28 months . At 36 months, 67% of US AIDS patients diagnosed in 1996 were still alive , compared with our result of 58% for patients diagnosed in the same year in Brazil. Corresponding numbers for Europe ranged by region from 78 to 88% . Many studies of survival do not include individual data on antiretroviral treatment, but those that do [8,11,22,24–27] match our findings that triple therapy is a key predictor of increased survival.
Proper use of antiretroviral drugs is not easy for patients or clinicians. Some have suggested that patients of low socio-economic status cannot properly adhere to complex treatment regimens and that improper use of these drugs by health care systems with inadequate infrastructure will blunt their effectiveness and favor the emergence of antiretroviral resistance . Brazil has neither a health care system that meets the standards of richer countries nor a high socio-economic level for the population in general or AIDS patients in particular. For example, CD4 and viral load testing were not widely available during most of the period we studied, and testing for antiretroviral resistance is only now being generally implemented. It is therefore notable that Brazil has achieved gains in survival on a national level that appear to be similar to those seen in more developed countries.
The univariate analysis demonstrated a variety of significant predictors of survival, many of which did not remain significant in the multivariate analysis. This suggests that many of these predictors, such as year of diagnosis and level of education, may have been markers for the likelihood of receiving antiretroviral treatment. Predictors other than antiretroviral treatment that remained significant in the multivariate analysis included diagnostic criteria and transmission category. Regarding diagnostic criteria, it is not surprising that the few patients diagnosed based on CD4 cell count did relatively better or that patients meeting multiple diagnostic criteria fared worse [29,30]. Most of the heterogeneity observed by transmission category is because patients of unknown category had worse survival. This may be partly circular: the health care system had less opportunity to determine transmission category for patients with short survival. Similarly, the extremely poor survival among patients with no antiretroviral treatment may be because patients with short survival had less opportunity to receive treatment.
This study has several important limitations. For practical reasons, we excluded from our sample cities with few AIDS cases and areas known to have poor reporting systems. We were unable to obtain sufficient data for analysis from 19% of eligible patients in our original sample. Insofar as AIDS patients in the excluded areas or with limited medical record data may have poorer survival, this may bias our results toward longer survival. As is normal practice for survival studies, we used last date of follow-up as a censor date for those patients not known to have died. This may produce a bias toward longer survival if patients lost to follow-up have poorer survival, as would also the exclusion from our sample of patients who did not live at least 1 week from the date of diagnosis.
Although these factors may mean that our results overestimate overall survival, they are unlikely to invalidate any of the other major findings of the study. Furthermore, methods were identical for cases diagnosed in 1995 or 1996, the period that saw the greatest absolute improvement in survival. The basic approach used in the study of patients diagnosed from 1982–1989 was similar. It is possible that some of the improvement in survival between the 1980s and more recent years could be because patients are being diagnosed earlier (as opposed to living longer). Any such difference is difficult to quantify since CD4 testing was generally not available in Brazil in the 1980s. CD4 cell counts were not substantially different among the minority of patients who received such testing when diagnosed in 1995 and 1996. In any case, we do not believe that this or any differences in methods could reasonably explain the very large increase in survival observed.
The Brazilian policy of providing universal access to antiretroviral treatment resulted from the convergence of many unique factors. Its implications and the controversies that it has produced go far beyond the scope of this study and even beyond the medical and public health sectors to included issues of property and human rights and international relations. Many of the questions raised by this bold policy do not have simple answers, but one thing is certain. It has made a tremendous difference for the survival of Brazilians with AIDS.
We thank the Health Ministry – National STD/AIDS Program for their financial and technical support; Augusto Leopoldo Ayrosa Galvao Center for Studies of the Faculdade de Medicina da Santa Casa de Sao Paulo for administrative support; Sao Paulo State and Sorocaba Municipal STD/AIDS Programs for technical support; Dr Draurio Barrera, Dr Marco Vitoria, Rosemeire Munhoz, and Maria R. Otero for providing national data, helpful comments, and references; Ana Lucia C. Monteiro for revising the data; Dr Yone A. Guibu for general support; Sergio Giannelli for data entry; Jose E. Neto, Celia Landman, Angela J. Gadelha, Ana Costa, Fernando Proietti, Luiza Matida, Maria Ines Dourado for their expertise and critical review of the study protocol; and the coordinators, supervisors, and investigators from each of the participating State and Municipal STD/AIDS Programs for data collection.
Sponsorship: This study was funded by the Brazilian National AIDS Control Program (CN-DST/AIDS).
1. Ministry of Health, Brazil. AIDS Epidemiological Bulletin
. 2001, XV
2. Szwarcwald CL, Bastos FI, Castilho EA.The dynamics of the AIDS apidemic in Brazil: a space-time analysis in the 1987–1995. Braz J Inf Dis
3. Ministry of Health, Brazil. AIDS Epidemiological Bulletin
4. Detels R, Muñoz 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. JAMA
5. Hogg RS, Heath KV, Yip B, Craib KJP, O'Shaughnessy MO, Schechter MT, et al
. Improved survival among following initiation of antiretroviral therapy. JAMA
6. Dimitrakopoulus A, Kosmopoulou O, Foukaneli T, Papalambrou C, Kalapothaki V, Kordossis T. Survival of AIDS patients in an academic unit in Athens, Greece. AIDS
7. Kholoud P and CASCADE Collaboration. Survival after introduction of HAART in people with known duration of HIV-1 infection. Lancet
8. Sendi PP, Bucher HC, Craig BA, Pfluger D, Battegay M, for the Swiss HIV Cohort Study. Estimating AIDS-free survival in a severely immunosuppressed asyntomatic HIV-infected population in the era of antiretroviral triple combination therapy. J Acquir Immune Defic Syndr Hum Retrovirol
9. Li Y, McDonald AM, Dore GJ, Kaldor JM, for the National HIV Survillance Committee. Improving survival following AIDS in Australia, 1991-1996. AIDS
10. Pezzoti P, Napoli PA, Acciai S, Boros S, Urciuoli R, Lazzeri V, et al
. Increasing survival time after AIDS in Italy: the role of new combination antiretroviral therapies. AIDS
11. Moore RD, Chaisson RE. Natural history of HIV infection in the era of combination antiretroviral therapy. AIDS
12. Porta D, Rapiti E, Forastiere F, Pezzoti P, Perucci CA, for the Lazio AIDS Survillance Collaborative Group. Changes in survival among people with AIDS in Lazio, Italy from 1993
13. Chequer P, Hearst N, Hudes ES, Castilho E, Rutherford G, Loures L, et al
. Determinants of survival in adult Brazilian AIDS patients, 1982
14. Mocroft A, Johnson MA, Phillipis NA. Factors affecting survival in patients with the acquired immunodeficiency syndrome. AIDS
15. Ministry of Health, Brazil. AIDS Epidemiological Bulletin.
: issue 1.
16. Ministry of Health, Brazil. National STD/AIDS Program: Review of national AIDS case definition
. : Brasilia: Ministry of Health; 1998.
17. Epi Info, version 6.04c
. Stone Mountain, GA: USD, Inc.; 1998.
18. StataCorp. Stata Statistical Software: Release 7.0.
College Station, TX: Stata Corporation; 2001.
19. Allison, Paul D. Survival Analysis Using the SAS System: A Practical Guide.
In: Cox models with nonproportional hazards
. Cary, NC: SAS Institute Inc.; 1995. pp. 154–155.
20. Lowndes CM, Bastos FI, Giffin KM, Reis ACGV, d'Orsi E, Alary M. Differential trends in mortality from AIDS in men and women in Brazil (1984–1995). AIDS
21. Casseb J, Pereira LC, Silva GL, Medeiros LA. Decreasing mortality and morbidity in adults AIDS patients from 1995 to 1997 in São Paulo, Brazil. AIDS Patient care STDs
22. Schwarcz SK, Hsu LC, Vittinghoff E, Katz MH. Impact of protease inhibitors and antiretroviral treatments on acquired immunodeficiency syndrome survival in San Francisco, California, 1987
–1996. Am J Epidemiol
23. Katz MH, Hsu L, Lingo M, Woelffer G, Schwarcs SK. Impact of socioeconomics status on survival with AIDS. Am J Epidemiol
24. Lee LM, Karon JM, Selik R, Neal JJ, Fleming PL. Survival after AIDS diagnosis in adolescents and adults during the treatment era, United States, 1984
25. Chiesi A, Mocroft A, Dally LG, Miller V, Katlama C, Ledergerber B, et al
. Regional survival differences across Europr in HIV-positive people: the EuroSIDA study. AIDS
26. Conti S, Masocco M, Pezzoti P, Toccaceli V, Vichi M, Boros S, et al
. Diferential impact of combined antiretroviral therapy on the survival of Italian patients with specific AIDS- defining illnesses. J Acquir Immune Defic Syndr
27. Gebhardt M, Rickenbach M, Egger M, and the Swiss Cohort Study. Impact of antiretroviral combination therapies on AIDS surveillance reports in Switzerland. AIDS
28. Harries AD, Nyangulu DS, Hargreaves NJ, Kaluwa O, Salaniponi FM. Preventing antiretroviral anarchy in sub-Saharan Africa. Lancet
29. Chêne G, Binquet C, Moreau JF, Neau D, Pellegrin I, Malvy D, et al . Changes in CD4+ cell count and the risk of opportunistic infection or death after highly active antiretroviral treatment. AIDS
30. Chaisson RE, Gallant JE, Keruly JC, Moore RD. Impact of opportunistic disease on survival in patients with HIV infection. AIDS