The association between levels of HIV RNA and expected survival time and time to development of AIDS amongst HIV-1-infected individuals has been repeatedly demonstrated [1-6]. However, relatively less attention has been paid to the question of how HIV RNA and other prognostic markers relate to the risk of specific AIDS-defining opportunistic infections. This question is important because AIDS is not a single entity, but rather a syndrome comprised of many distinct conditions. Clearly, markers of HIV may relate differently to the risk of particular opportunistic infections, and the investigation of this issue amongst a cohort of individuals is one epidemiologic approach to better understanding the pathogenesis of AIDS. In this study, we considered three separate diagnoses: Pneumocystis carinii pneumonia (PCP), cytomegalovirus (CMV) infection, and Mycobacterium avium complex (MAC) disease. These three particular infections have been connected with a significant portion of the morbidity experienced by late-stage HIV patients in the United States . Our primary aims were to compare and contrast the prognostic information for these three outcomes attributable to age, HIV-related symptoms, CD4 cell counts, and plasma HIV RNA levels. Given the emphasis in current guidelines  upon the CD4 cell threshold of 200×106/l for evaluating the need for PCP prophylaxis, we closely investigated the data from this cohort regarding what additional prognostic information is contained in the plasma HIV RNA level.
Despite changing incidences in recent years , PCP is still the most common AIDS-defining illness in the United States. CMV attacks a wide variety of organs, but most commonly causes sight-threatening retinitis . MAC, a non-tuberculous mycobacterial infection, occurs relatively late in the course of HIV infection. Each of these three illnesses has been found to be associated with advanced immunodeficiency as measured by low CD4+ lymphocyte counts. Both PCP and MAC tend to be increasingly likely in individuals presenting with HIV-related symptoms, primarily candidiasis and fever [11,12].
The availability of HIV RNA quantification techniques justifies a renewed examination of the predictors of specific diseases, such as PCP, CMV and MAC. Recently, Swindells et al.  studied the association between baseline measurements of HIV RNA and CD4 cell count and the risk of these three conditions using data on treated individuals in relatively late stages of HIV infection, from three AIDS Clinical Trial Group (ACTG) studies. They reported that HIV RNA and CD4 cell count taken separately were both prognostic for all three outcomes. Since PCP, CMV and MAC were the three most common opportunistic infections in a large phase III trial, ACTG investigators also provided data on the risk of these conditions across two treatment arms, one of which represented highly active antiretroviral therapy (HAART) comprised of a protease inhibitor and two nucleoside reverse transcriptase inhibitors . The time to develop PCP and CMV was demonstrated to be longer for this three-drug arm, and larger observed increases in CD4 cell count were independently associated with reduced risk of developing an opportunistic infection.
Laboratory data from a prior study conducted by Multicenter AIDS Cohort Study (MACS) investigators  provide a unique parallel opportunity for addressing the potentially differential prognostic value of markers of HIV infection for different opportunistic infections outside the setting of clinical trials, amongst a large cohort of HIV-positive men. In this study, we focused attention upon pure natural history, in the sense that follow-up for the outcomes of interest was restricted to the time period prior to the widespread availability of antiretroviral therapies and prophylaxis against specific opportunistic infections. Such natural history studies are significant given the geography of the worldwide HIV epidemic, and the sobering fact that antiretroviral therapy remains largely inaccessible in many developing countries, as well as in parts of the developed world. For populations with access to therapies, the resulting information provides an important reference for comparison with future studies in cohorts of HIV-positive individuals. Such reference information becomes increasingly relevant as follow-up for these opportunistic infections further accrues during the era of more potent and ever-changing antiretroviral regimens and more regular monitoring of HIV RNA levels. At the population level, meaningful insights into the influence of antiretroviral treatment with respect to changing patterns in the incidence of opportunistic infections and in how markers of HIV relate to the risks of specific outcomes will best be gained via comparison with pure natural history studies in large cohorts.
Materials and methods
MACS is an ongoing prospective study into which 4954 homo-/bisexual men were enrolled between March 1984 and April 1985. Participants were primarily Caucasian (88%), were between the ages of 18 and 60 years at the start of the study, and were enrolled at one of four study centers (located in Baltimore, Chicago, Los Angeles, and Pittsburgh). They return for semiannual visits that include collection of laboratory specimens, a physical examination, and completion of detailed questionnaires. Specific information about the full cohort and the data collected have been presented previously .
Mellors et al.  measured HIV RNA in stored plasma samples from 1604 men who were AIDS-free at a baseline visit (MACS visit 3 or 4). Each participant‚s sample was originally collected during this baseline visit, which took place during the years 1984-1985. For the present analysis, we identified 734 of these individuals for whom the concomitant (at visit 3 or 4) CD4 cell count was ≤500×106/l. This restriction was implemented to focus attention on a fairly homogeneous group of men with some evidence of immunodeficiency, hence excluding those who would be judged on that basis to be at minimal risk of developing opportunistic infections in the short term. Of these 734 individuals, 694 were HIV-seropositive at enrollment into MACS, and 40 seroconverted between the time of enrollment and the baseline visit at which the sample for HIV RNA quantification was collected.
As described previously , HIV RNA was quantified in heparinized plasma samples using a branched DNA assay (Chiron Corporation, Emeryville, California, USA) with a quantification limit of 500 copies/ml. This assay has been described in detail elsewhere . Previous studies, including one study conducted in MACS, indicate that plasma viral load quantification is not compromised when dealing with samples that have been stored in heparin [16,17]. Absolute CD4+ T-lymphocyte counts were obtained by multiplying percentages derived via flow cytometry by absolute lymphocyte counts .
Variables for study
Outcomes (PCP, CMV, MAC) for this study were determined based on a combination of active and passive surveillance utilizing self-reports confirmed by medical records, together with frequent contact of severely immunodeficient men and men previously diagnosed with AIDS . PCP was diagnosed according to standard criteria, and MAC was defined as disseminated atypical non-tuberculous mycobacterial infection. CMV was defined as retinitis, or as histopathologically documented infection of an organ other than the liver, spleen, or lymph nodes.
The major prognostic variables considered were baseline determinations of age, HIV RNA concentration, CD4 cell count, and the self-reported occurrence of oral candidiasis or fever for a period of 2 weeks or more since the prior visit. The effects of geographic location and baseline hemoglobin measures, which were previously found to be associated with the development of MAC disease in very late stages of HIV , were also evaluated; however, these were not found to be prognostic over this study period and are not further discussed. The cut-off date for analysis selected for this study was 31 July 1988. As previously described , this strategy permitted a pure natural history study by concentrating on follow-up that occurred prior to any significant use of antiretrovirals and PCP prophylaxis.
Spearman correlations were calculated amongst the continuous variables age, HIV RNA, and CD4 cell count at baseline, and relevant descriptive statistics were compiled. Baseline HIV RNA levels were divided into four categories: 0-29999, 30000-59999, 60000-89999, and ≥90000 copies/ml. Baseline CD4+ lymphocyte count was categorized as ≥350×106/l, 200-349×106/l, or <200×106/l. Age at baseline was dichotomized near the median (<35 versus ≥35 years).
Cumulative incidence rates for PCP, CMV and MAC were compared across the categories of HIV RNA, and overall differences in risk were assessed using the log-rank test. The baseline categories of age, candidiasis or fever, CD4 cell count, and HIV RNA were subsequently assessed jointly for discrepancies in risk via Cox proportional hazards regression. To assess the effect of a prior AIDS diagnosis (based on the 1987 Centers for Disease Control and Prevention definition), this variable was included as a time-dependent covariate in separate models. Individuals were censored at death or at the time of loss-to-follow-up. If loss-to-follow-up occurred after the cut-off date for analysis of 31 July 1988, then individuals were censored at this date.
To investigate in more detail the implications of HIV RNA levels with respect to the risk of PCP among men who were not severely immunosuppressed, a slight modification of the aforementioned HIV RNA categories was adopted, with 75000 copies/ml utilized as a cut-off point in place of 60000 and 90000 copies/ml. After thus reducing to three HIV RNA categories, numbers were sufficient to compare relative hazards (RH) of developing PCP across all nine baseline HIV RNA/CD4 cell subgroups. A referent group was selected consisting of a subset of those subjects with baseline CD4 cell counts below 200×106/l, and Cox regression analysis was employed to estimate and compare RH.
Prognostic variables for PCP, CMV and MAC
The number of cases of PCP, CMV, and MAC observed over the time period under study were 138, 25, and 25, respectively, with 119, seven, and nine of these being initial AIDS diagnoses. The large discrepancies between the overall number of cases and the proportion of these cases that were initial AIDS- defining illnesses for PCP as opposed to CMV and MAC were expected, particularly given the fact that follow-up occurred prior to significant use of PCP prophylaxis. Less than 3% of the study population reported any use of PCP prophylaxis over the study period, and almost none of this use occurred prior to the last 6 months of follow-up. Reported use of prophylaxis for CMV or MAC was essentially non-existent (less than 0.5%).
Weak correlations between baseline age and CD4 cell count (Spearman‚s r=0.09) and age and HIV RNA (r=-0.07), and a moderate negative correlation between CD4 cell count and HIV RNA (r=-0.28, P=0.001) were observed. Table 1 provides baseline descriptive statistics (medians, first and third quartiles) for age, HIV RNA, and CD4 cell count, and percentages reporting candidiasis or fever. This summary is reported both overall and for groups stratified according to whether each event (PCP, CMV, MAC) was observed over the course of the study period. The results (Table 1) showed a clear tendency towards higher HIV RNA levels and lower CD4 cell counts amongst those eventually developing PCP, CMV or MAC. Also noteworthy was the relatively higher median age (37.4 years) amongst those who developed CMV and the higher baseline prevalence of candidiasis or fever (45.8%) amongst those who contracted MAC disease.
Of the 734 individuals in the study population, 439, 105, 73, and 117 persons were members of the four baseline HIV RNA categories defined by <30000, 30000-59999, 60000-89999, and ≥90000 copies/ml, respectively. The percentages of those in the four categories who experienced events over the course of follow-up were as follows: PCP, 5.9%, 23.8%, 41.1%, 48.7%, respectively; CMV, 1.1%, 1.9%, 6.8%, 11.1%, respectively; MAC, 0.7%, 1.0%, 8.2%, 12.8%, respectively. Log-rank tests were significant (P<0.001) for each of the three outcomes, indicating a clear overall difference in time to event according to the plasma viral load at baseline (Fig. 1).
Table 2 shows the results of Cox regression analysis to assess prognostic variables for the various risk factor categories. For PCP, a significant increase in RH for all non-reference categories of CD4 cell count and HIV RNA was found, with a pronounced trend towards increasing risk as the viral load increased. The adjusted RH estimates for the most severely immunosuppressed and most highly viremic categories were 3.72 and 11.13, respectively (P<0.001). With respect to CMV as the outcome, larger adjusted RH than those for PCP were observed with respect to the CD4 cell categories (RH,7.93 for CD4 cells <200×106/l; P<0.001). For MAC, a report of candidiasis or fever at baseline was a highly significant risk factor (RH, 4.26; P<0.005). The adjusted RH involving the highest two HIV RNA categories were largest for MAC as the outcome (RH,14.97 for HIV RNA >90000 copies/ml; P<0.001).
Table 3 shows the results of Cox regression models which were adjusted additionally for the effect of a prior clinical diagnosis of AIDS as a time-dependent covariate. As one would expect, such an event significantly elevated the risk for each of the three outcomes, most notably for CMV and MAC (adjusted RH, 8.72 and 4.26, respectively). The adjusted RH estimates relating CD4 cell count and HIV RNA to the risk of PCP were almost identical to those in Table 2, but those relating CD4 cell count and HIV RNA to the risk of CMV and MAC were noticeably attenuated after adjustment for prior AIDS.
Given that there may be residual confounding due to variations in CD4 and HIV RNA within categories, Table 4 shows comparative results in which these two markers were treated continuously (after log10- transforming HIV RNA copies/ml). These results may be compared with those in Table 2. The estimates presented were of adjusted RH corresponding to a 50×106/l deficit in CD4 cell count and to a 1log10 copies/ml increase in HIV RNA. Models treating CD4 cell count and HIV RNA continuously and adjusting for the prior AIDS diagnosis were also fit, but reflected qualitatively similar results to those shown in Table 3 (data not shown).
HIV RNA as a risk factor for PCP among men with CD4 cell counts above 200×106/l
Table 5 shows information on the 138 observed cases of PCP according to nine subgroups defined by baseline HIV RNA and CD4 cell count. For this analysis, we retained the same three categories of CD4 cell count, but consolidated the highest two HIV RNA categories by implementing a cut-off at 75000 copies/ml in place of the two previous cut-off values at 60000 and 90000 copies/ml. The resulting three categories (<30000, 30000-74999, and ≥75000 copies/ml) of HIV RNA relate to the quintiles of the study population, which were defined by 5296, 12930, 30530, and 74860 copies/ml, respectively. In particular, the first category essentially contained the lowest 60% of the subjects, and the latter two categories approximately comprised the upper two quintiles. The nine subgroups (Table 5) resulted from combining these three HIV RNA categories with the three original CD4 cell count categories. Here, we took those with baseline HIV RNA below 30000 copies/ml and baseline CD4 cell count below 200×106/l as the reference group.
Groups 1 and 2 consisted of those in the same CD4 cell count category as the reference group, but with higher HIV RNA levels (Table 5); hence, the significantly elevated RH were qualitatively consistent with the results for PCP (Tables 2 and 3). Groups 3-8, however, comprised individuals with CD4 cell counts above 200×106/l, and the RH associated with these groups were of particular interest. Groups 5 and 8 had a significantly lower risk of developing PCP than the reference group, as expected based on the fact that they were in the same HIV RNA category but had higher CD4 cell counts. However, groups 3, 4, 6 and 7 experienced equal or greater risk relative to the reference group. In the case of group 3 (HIV RNA ≥75000 copies/ml, CD4 cell count 200-350×106/l), the risk was significantly higher than for those with HIV RNA below 30000 copies/ml and CD4 cell count below 200×106/l (P=0.01). When this analysis was repeated with the entire group of those with baseline CD4 cell counts below 200×106/l as the reference, the estimated RH comparing group 3 to this reference group was close to (and not significantly different from) 1.0 (P=0.56; data not shown).
Because follow-up for this study (roughly encompassing early 1985 to mid-1988) occurred before significant use of antiretroviral therapy or PCP prophylaxis, these results extend those of an earlier report from MACS  about the natural history of HIV infection in terms of variables related to the risk of PCP amongst homo-/bisexual men. The primary extensions are the incorporation of plasma viral load as a prognostic marker and the comparisons drawn with respect to the information provided by baseline markers of HIV infection for future incidence of CMV infection and MAC disease.
Our study population was somewhat refined at baseline, being restricted to men with CD4 cell counts ≤500×106/l. However, some men who were excluded on the basis of CD4 cell counts did develop the outcomes of interest over the study period. In particular, amongst 869 men with CD4 cell counts above 500×106/l at baseline, 50 developed PCP, four developed CMV, and three developed MAC. For comparison, we repeated the analysis without the baseline CD4 cell restriction. Since the conclusions were no different regarding the markers that were independently predictive of each outcome, we only reported the results based on the more homogeneous study population.
By focusing upon the untreated natural history, this investigation capitalizes on a major strength of cohort studies and complements ongoing clinical trials research [13,14] involving individuals on various antiretroviral therapy regimens. The results of our study re-emphasize the need to initiate treatment among previously untreated HIV-1-infected patients presenting with low CD4 cell counts with or without high HIV RNA levels, since these individuals were generally found to be at significantly higher risk (Tables 2 and 3). In addition to antiretrovirals, specific prophylaxis against PCP, CMV and MAC is available as a therapeutic option in such circumstances. In particular, specific recommendations have been made for the prevention of PCP and MAC, and oral ganciclovir has been suggested as prophylaxis for CMV infections in some situations . With respect to PCP, in our study 30% (30 out of 99) of those individuals with CD4 cells below 200×106/l at baseline also had HIV RNA levels below 30000 copies/ml. Compared with these, several groups of individuals with CD4 cell counts above 200×106/l were estimated to be at equal or greater risk of developing PCP over the study period, including one subgroup with higher HIV RNA levels that showed significantly higher incidence (Table 5). Among subjects with CD4 cell counts ≤500×106/l, the data in Table 5 indicate that PCP prophylaxis should be a consideration for all patients with plasma HIV RNA above 30000 copies/ml, in addition to all those with CD4 cells below 200×106/l. Factors with an impact on this decision include the level of monitoring of a particular patient, and, amongst those with access to potent antiretroviral therapies, the viremic response to such therapies. Because the risk of PCP in our study was defined over a 3.5-year period, closely monitored patients with intermediate CD4 cell counts at baseline may have declined below 200×106/l before developing disease and hence would have been prescribed PCP prophylaxis in a timely fashion according to current guidelines . According to present practice, those with relatively high baseline viral loads and access to treatment would be likely to have initiated potent antiretroviral therapy, and with a favorable response would have achieved a reduction in the risk of PCP subsequent to a decline in HIV RNA. Given the results of our investigation, however, those patients with high viral loads who may be less frequently monitored, or who do not tolerate or respond well to antiretrovirals, may be candidates for PCP prophylaxis despite CD4 cell counts above 200×106/l.
A somewhat different picture emerged from our analysis regarding the prognostic variables for PCP, CMV and MAC (Tables 2 and 3). For PCP, all non-reference categories of CD4 cell counts and HIV RNA were associated with elevated risk, but no significant associations were seen with age or the presence of candidiasis or fever at baseline despite the relatively large number of events. For CMV, there was a trend towards increasing risk with older age at baseline, with an adjusted RH estimate approaching 2.0. Further investigation of this trend revealed that it was not present when restricting analysis to CMV retinitis only, but was highly accentuated and significant in an analysis that excluded retinitis from the case definition and allowed less strict diagnostic criteria to apply in defining histopathologically documented CMV infection. More clearly, the risk of CMV was seen to be related to level of immunosuppression and viremia at baseline. For MAC, we observed the highest RH estimates with respect to increasing levels of viremia, and the report of symptoms (candidiasis or fever) at baseline was highly associated with increased risk. Although viral load provides prognostic information independently of the CD4 cell count in each case, Tables 2 and 3 showed somewhat different patterns of increasing risk across baseline categories of HIV RNA for the three separate outcomes.
The results in Table 3 suggested a tendency for some of the association between baseline HIV RNA and CD4 cell levels and the risk of CMV or MAC to be mediated via a prior diagnosis of AIDS. In our analysis, prior AIDS did not strictly qualify as a confounder since it may be part of a causal pathway relating baseline markers to the risk of specific opportunistic infections . Given such a possibility, it is recommended epidemiologic practice to consider and compare the results of analyses both adjusted and unadjusted for the potential intervening factor (in this case, prior AIDS). Hence, the analysis in Table 3 was considered in addition to, rather than in place of, the analysis unadjusted for AIDS (Table 2). The RH estimates and significance levels presented in Tables 2 and 3 suggest that, after adjusting for the intervening occurrence of AIDS, it is primarily immunosuppression that was associated with the risk of CMV and elevated viremia that was associated with MAC disease in this untreated study population. In the latter context, researchers have described a direct relationship between M. avium and HIV-1 viral load, as well as direct stimulatory interactions between M. avium and HIV-1 Tat, which could provide a mechanism for the association observed here [21-23]. The fact that prior AIDS was associated with increased risk of all three opportunistic infections makes intuitive sense, and is consistent with studies that have described the prognosis of AIDS amongst individuals with and without coexisting diagnoses [24,25].
As previously indicated, a branched DNA assay was utilized to obtain baseline HIV RNA measurements for this study population. This assay typically yields lower copy numbers than alternative assays based on reverse transcriptase polymerase chain reaction (RT-PCR). Using the approximation provided by Mellors et al. , we estimate that the cut-off values employed for the branched DNA assay in our study of 30000, 60000, and 90000 copies/ml are roughly equivalent to 55000, 100000, and 150000 copies/ml, respectively, measured by RT-PCR. The cut-off of 75000 copies/ml employed for the further analysis of PCP risk corresponds to approximately 125000 copies/ml by RT-PCR.
In summary, this study of 734 HIV-infected men clarifies the joint prognostic implications of HIV RNA, CD4 cell counts, and HIV-related symptoms for assessing the risks of developing PCP, CMV and MAC disease. Because AIDS is comprised of many distinct conditions, such refinement of our current knowledge  about these markers of HIV-related outcomes is necessary in order to better determine prognosis and ensure valid decisions regarding the initiation of specific prophylaxis. Pure natural history studies to assess new markers of risk for specific opportunistic infections are possible because of valuable stored specimens collected during the follow-up of long-standing cohorts. They provide directly relevant information for populations for whom effective therapies are not available, in addition to important reference data for assessing the population-based impact of effective antiretrovirals. In cohorts such as MACS, continued follow-up of individuals exposed to HAART will allow valuable comparative studies of how these therapies alter the associations between viral load and CD4 lymphocyte counts and the risks of specific AIDS-defining illnesses.
1. Jurriaans S, Van Gemen B, Weverling GJ, et al.
: The natural history of HIV-1 infection: virus load and virus phenotype independent determinants of clinical course?Virology
2. Henrard DR, Philips JF, Muenz LR, et al.
: Natural history of HIV-1 cell-free viremia.JAMA
3. Mellors JW, Kingsley LA, Rinaldo CR Jr, et al.
: Quantitation of HIV-1 RNA in plasma predicts outcome after seroconversion.Ann Intern Med
4. O‚Brien TR, Blattner WA, Waters D, et al.
: Serum HIV-1 RNA levels and time to development of AIDS in the Multicenter Hemophilia Cohort Study.JAMA
5. Mellors JW, Rinaldo CR Jr, Gupta P, White RM, Todd JA, Kingsley LA: Prognosis in HIV-1 infection predicted by the quantity of virus in plasma.Science
6. Mellors JW, Muñoz A, Giorgi JV, et al.
: Plasma viral load and CD4+ lymphocytes as prognostic markers of HIV-1 infection. Ann Intern Med
7. Chan ISF, Neaton JD, Saravolatz LD, Crane LR, Osterberger J: Frequencies of opportunistic diseases prior to death among HIV-infected persons. Community Programs for Clinical Research on AIDS.AIDS
8. Prevention of Opportunistic Infections Working Group: 1997 USPHS/IDSA guidelines for the prevention of opportunistic infections in persons infected with human immunodeficiency virus disease: specific guidelines.Clin Infect Dis
9. Muñoz A, Schrager LK, Bacellar H, et al.
: Trends in the incidence of outcomes defining acquired immunodeficiency syndrome (AIDS) in the Multicenter AIDS Cohort Study.Am J Epidemiol
10. Pertel P, Hirschtick R, Phair J, Chmiel J, Poggensee L, Murphy R: Risk of developing cytomegalovirus retinitis in persons infected with the human immunodeficiency virus.J Acquir Immune Defic Syndr
11. Phair JP, Muñoz A, Detels R, Kaslow R, Rinaldo C, Saah A, and the Multicenter AIDS Cohort Study Group: The risk of Pneumocystis carinii pneumonia among men infected with human immunodeficiency virus type 1.N Engl J Med
12. Hoover DR, Graham NMH, Bacellar H, et al.
: An epidemiologic analysis of Mycobacterium avium complex disease in homosexual men infected with human immunodeficiency virus type 1.Clin Infect Dis
13. Swindells S, Currier JS, Williams P: Correlation of viral load and risk for opportunisticinfections.Fourth Conference on Retroviruses and Opportunistic Infections.
Washington, DC, January 1997 [abstract 359].
14. Currier JS., Williams PL, Grimes JM, Squires KS, Fischl MA, Hammer SM for the ACTG 320 Study Team: Incidence rates and risk factors for opportunistic infections in a phase III trial comparing indinavir + ZDV + 3TC to ZDV + 3TC. Fifth Conferenceon Retroviruses and Opportunistic Infections.
Chicago, February 1998 [abstract 257].
15. Kaslow RA, Ostrow DG, Detels R, Phair JP, Polk BF, Rinaldo CR Jr: The Multicenter AIDS Cohort Study: rationale, organization, and selected characteristics of the participants. Am J Epidemiol
16. Todd J, Pachl C, White R, et al.
: Performance characteristics for the quantitation of plasma HIV-1 RNA using branched DNA signal amplification technology. J AcquirImmune Defic Syndr Hum Retrovirol
17. Kirstein L, Mellors JW, Giorgi JV, et al.
: Effect of processing delay, anticoagulant and assay (bDNA, RT-PCR) on plasma HIV RNA decay.12th World AIDS Conference.
Geneva, June 1998 [abstract 42144].
18. Giorgi JV, Cheng HL, Margolick JB, et al.
: Quality control in the flow cytometric measurement of T-lymphocyte subsets: The Multicenter AIDS Cohort Study Experience.Clin Immunol Immunopathol
19. Jacobson LP, Kirby AJ, Polk S, et al.
: Changes in survival after acquired immunodeficiency syndrome (AIDS): 1984-1991.Am J Epidemiol
20. Hennekens CH, Buring JE: Epidemiology in Medicine.
Boston: Little, Brown and Company; 1987.
21. Denis M: Tat protein from HIV-1 binds to Mycobacterium avium via a bacterial integrin.J Immunol
22. Denis M, Ghadirian E: Mycobacterium aviuminfection in HIV-1 infected subjects increases monokine secretion and is associated with enhanced viral load and diminished immune response to viral antigens.Clin Exp Immunol
23. Denis M, Ghadirian E: Interaction between Mycobacterium avium and human immunodeficiency virus type 1 (HIV-1) in bronchoalveolar macrophages of normal andHIV-1-infected subjects. Am J Respir Cell Mol Biol
24. Lundgren JD, Barton SE, Katlama C, et al.
: Changes in survival over time after a first episode of Pneumocystis carinii pneumonia for European patients with acquired immunodeficiency syndrome.Arch Intern Med
25. Luo K, Law M, Kaldor JM, McDonald AM, Cooper DA: The role of initial AIDS-defining illness in survival following AIDS.AIDS
The Multicenter AIDS Cohort Study
Baltimore: Joseph B. Margolick (Principal Investigator), Haroutune Armenian, Homayoon Farzadegan, Nancy Kass, Justin McArthur, Steffanie Strathdee, Ellen Taylor (The Johns Hopkins University School of Hygiene and Public Health).
Chicago: John P. Phair (Principal Investigator), Joan S. Chmiel, Bruce Cohen, Maurice O‚Gorman, Daina Variakojis, Jerry Wesch, Steven M. Wolinsky (Howard Brown Health Center and Northwestern University Medical School).
Los Angeles: Roger Detels (Principal Investigator), Barbara R. Visscher, Janice P. Dudley, John L. Fahey, Janis V. Giorgi, Andrew Kaplan, Oto Martínez-Maza, Eric N. Miller, Hal Morgenstern, Parunag Nishanian, John Oishi, Jeremy Taylor, Harry Vinters (University of California, UCLA Schools of Public Health and Medicine).
Pittsburgh: Charles R. Rinaldo (Principal Investigator), James T. Becker, Phalguni Gupta, Lawrence Kingsley, John Mellors, Sharon Riddler, Anthony Silvestri (University of Pittsburgh Graduate School of Public Health).
Data Coordinating Center: Alvaro Muñoz (Principal Investigator), Cheryl Enger, Stephen Gange, Lisa P. Jacobson, Cynthia Kleeberger, Robert Lyles, Steven Piantadosi, Sol Su (The Johns Hopkins University School of Hygiene and Public Health).
National Institutes of Health: Lewis Schrager (Project Officer, National Institute of Allergy and Infectious Diseases), Sandra Melnick (National Cancer Institute).