ORIGINAL INVESTIGATION. A Critical Assessment of the Prognostic Value of HIV-1 RNA Levels and CD4 + Cell Counts in HIV-Infected Patients

Transcription

1 ORIGINAL INVESTIGATION A Critical Assessment of the Prognostic Value of HIV- RNA Levels and CD + Cell Counts in HIV-Infected Patients Sabine Yerly, MS; Thomas V. Perneger, MD, PhD; Bernard Hirschel, MD; Olivier Dubuis, MD; Lukas Matter, MD; Raffaele Malinverni, MD; Hansjakob Furrer, MD; Luc Perrin, MD; for the Swiss HIV Cohort Study Objective: To determine to what extent human immunodeficiency type (HIV-) RNA levels and CD + cell counts predict clinical outcomes in a general HIV- infected population. Methods: Community-based prospective study (Swiss HIV Cohort Study) including 9 HIV- infected patients, randomly selected from strata of CD + cell counts (0 to 0.05, 0.05 to 0.0, 0.0 to 0.50, and /L). Levels of HIV- RNA, CD + cell counts, and other variables were evaluated from samples collected between 99 and 99 for their ability to predict death and clinical progression. Results: Patients were followed up on average for 9 months. Baseline HIV- RNA levels, CD + cell counts, clinical stage, and -microglobulin levels independently predicted survival, whereas only HIV- RNA levels and CD + cell counts independently predicted clinical progression. Multivariate relative hazards (RHs) for death ranged from.0 to 5. across quartiles of CD + counts, but only from.0 to.8 across quartiles of HIV- RNA. For clinical progression, gradients of risk were similar for CD + counts (.0-.) and for HIV- RNA (.0-.). In patients with CD + cell counts less than /L, HIV- RNA levels predicted neither death nor clinical progression. Finally, the number of HIV- RNA copies per CD + cell was the best predictor of death (multivariate RH, across quartiles) and clinical progression (multivariate RH,.0-.). Conclusions: Levels of HIV- RNA and CD + cell counts provided independent and complementary information on clinical outcomes. The RNA/CD + ratio was the best single predictor. In patients who had fewer than /L CD + cells, HIV- RNA levels had little prognostic value. Arch Intern Med. 998;58:7-5 From the Laboratory of Virology (Ms Yerly and Dr Perrin) and AIDS Center (Dr Hirschel), Division of Infectious Diseases, Geneva University Hospital, Geneva, Switzerland; Institute of Social and Preventive Medicine, University of Geneva (Dr Perneger); Institute of Medical Microbiology, University of Bern, Bern, Switzerland (Drs Dubuis and Matter); and AIDS Center, Medical Policlinic, Inselspital Bern (Drs Malinverni and Furrer). Members of the Swiss HIV Cohort Study are listed in the acknowledgments at the end of this article. VARIOUS BIOLOGICAL parameters predict clinical outcomes in patients infected with the human immunodeficiency virus type (HIV-). Until 995, CD + cell counts were considered to be the best predictor, but several recent studies suggest that a direct measurement of viremia (plasma HIV- RNA) may provide superior prognostic information. - However, currently available evidence presents some limitations, which we sought to address in this study. First, previous studies have involved selected patient populations: patients enrolled in clinical trials of antiviral treatments, 7- asymptomatic patients recently infected with HIV-,, or patients with hemophilia. Whether these results apply to HIV-infected patients encountered in general medical practice, in particular to those with advanced immunosuppression, is therefore uncertain. Second, most previous studies have focused on viremia, and treated other predictors, including CD + cell counts, as confounders. Little information is available on the respective contribution of each variable, and on the interaction between viremia and CD + cell counts in predicting clinical outcomes. Last, few studies -6 have examined separately the risk of death and the risk of clinical progression; predictors may vary for these outcomes. The main aim of this study was to assess the ability of biological parameters, including HIV- RNA levels and CD + cell counts, to predict death and clinical progression in a general population of HIV-infected patients. Secondary aims were to assess the predictive value of HIV- RNA levels and CD + cell counts at various levels of disease severity, to evaluate whether predictors differ for death and clinical progression, and to identify the most efficient way of expressing the prognostic information conveyed by these biological parameters. ARCH INTERN MED/ VOL 58, FEB 9, American Medical Association. All rights reserved. Downloaded From: on 0/06/08

2 METHODS STUDY DESIGN AND VARIABLES This prospective study examined the value of baseline HIV- RNA levels, CD + and CD8 + cell counts, -microglobulin levels, immune complex dissociated (ICD) p antigen levels, clinical stage, patient age, patient sex, mode of HIV transmission in predicting death (HIV related or not), and clinical disease progression. Clinical progression was defined as occurrence of Centers for Disease Control and Prevention (CDC) stage B or C diseases (CDC 99 criteria ) of equal or greater severity than the baseline clinical status, including HIV-related death. Patients were enrolled between January 99 and December 99, with follow-up until April 996. STUDY POPULATION Patients enrolled in the Swiss HIV Cohort Study in Bern and Geneva, Switzerland, who had a frozen plasma sample (stored at 75 C) collected between January 99 and December 99 were eligible. Of 6 eligible patients, 00 were selected at random, stratified on the basis of initial CD + cell counts ( 0 9 /L), with 00 patients in each of strata (0 to 0.05, 0.05 to 0.0, 0.0 to 0.50, and 0.50). Proportions of patients selected within the eligible population according to strata of CD + cell counts were 78%, 65%, 7%, and 78%, respectively. When several plasma samples were available for a given patient, the first one was used as the baseline sample. LABORATORY METHODS Levels of HIV- RNA were measured by polymerase chain reaction using the Amplicor HIV Monitor assay (Roche, Basel, Switzerland). The lowest detection threshold of the assay was 00 copies per milliliter. CD + and CD8 + cell counts were determined by flow cytometry (Coulter EPICS IV, Instrumente Gesellschaft AG, Basel). - Microglobulin levels were determined by enzyme immunoassay (IMX Diagnostic System, Abbott Laboratories, North Chicago, Ill). The level of ICD p antigen was measured in heat-denatured plasma using DuPont HIV- p Antigen and DuPont ELAST ELISA Amplification System (Du- Pont, Boston, Mass). 5 Results of HIV- RNA and ICD p antigen measurements were expressed on a logarithm scale (base 0). STATISTICAL ANALYSIS Relationships between baseline predictor variables (categorized by quartile) and outcome variables were assessed using Kaplan-Meier analyses, log-rank tests for linear trend, and proportional hazards models. 6 The predictive value of HIV- RNA level was also examined across quartiles of CD + cell counts; in these analyses, log HIV- RNA/mL was treated as a continuous variable; similarly, the predictive value of CD + cellcounts(onacontinuousscale)wasexaminedacross quartiles of HIV- RNA. In the latter models, an interaction term was used to determine whether the risk of event associated with one predictor changed significantly across quartiles of the other predictor. The RNA/CD + ratio was calculated by dividing log HIV- RNA/mL by CD + cells/ml. P values less than.05 were considered significant. RESULTS PATIENT CHARACTERISTICS Six patients were withdrawn from the original sample: HIV- RNA levels could not be measured in patients due to inhibition of polymerase chain reaction, and follow-up data were not available for other patients. The 9 remaining patients were, on average, 5 years old (range, 8-66 years), and 07 (78%) were men. Risk stratification included men who have sex with men (5%), 07 persons infected through heterosexual contact (7%), 67 intravenous drug users (7%), and with other or unknown risks (%). At baseline, 50 patients (8%) had disease that was in clinical stage A, (9%) in stage B, and 0 (%) in stage C. All predictor variables were widely distributed (Table ). In 8 patient samples (%), HIV- RNA was below the detection limit of the assay, and these patients were assigned the threshold value of 00 copies/ml. Quartiles of log HIV- RNA were less than.87,.87 to.5,.6 to.9, and.9 to 6. copies/ ml. Levels of HIV- RNA were significantly associated with CDC stages, CD + and CD8 + cell counts, - microglobulin levels, ICD p antigen levels, and age (data not shown). Fifty-five percent of patients received antiretroviral treatment at any time during follow-up, mainly zidovudine, didanosine, or zalcitabine as monotherapy, Table. Baseline Variables of HIV- Infected Patients (N=9)* Variable Mean±SD Median (Range) CD + count, 0 9 /L 0.0± ( ) CD8 + count, 0 9 /L 0.79± (0.0-.9) HIV- RNA, log copies/ml.5±0..86 (.00-6.) -Microglobulin, nmol/l ±0 97 (0-9) ICD p Ag, log pg/ml.0±0..6 ( ) *HIV- indicates human immunodeficiency virus type ; ICD p Ag, immune complex dissociated p antigen. and 7% received prophylaxis for Pneumocystis carinii pneumonia. PATIENT SURVIVAL During follow-up (mean, 9 months; range, months), 69 patients (%) died; 8 deaths were related to HIV and were not (suicides, homicides, or other causes). The overall mortality rate was 8 per 00 personyears (95% confidence interval [CI], 6-0). Levels of HIV- RNA strongly predicted survival (log-rank P.00) (Figure, A). Mortality rates in increasing quartiles of HIV- RNA were, 7, 7, and deaths per 00 person-years. Similarly, CD + cell ARCH INTERN MED/ VOL 58, FEB 9, American Medical Association. All rights reserved. Downloaded From: on 0/06/08

3 00 A 00 B Survival, % 0 Survival, % C 00 D Survival Free of Clinical Progression, % Months of Follow-up Survival Free of Clinical Progression, % Months of Follow-up Figure. Relation between human immunodeficiency virus type (HIV-) RNA levels or CD + cell counts and clinical outcomes. Kaplan-Meier curves for survival stratified by quartiles of log HIV- RNA/mL (A) and CD + cell counts (B), and for clinical progression stratified by quartiles of HIV- RNA/mL (C) and CD + cell counts (D). Quartiles of log HIV- RNA/mL:,.87;,.87-.5;,.6-.9; and, Quartiles of CD + ( 0 9 /L):, 0.50;, 0.0 to 0.50;, 0.05 to 0.0; and, counts strongly predicted survival (log-rank P.00) (Figure, B). Mortality rates for decreasing quartiles of CD + cell counts were, 6,, and deaths per 00 person-years. In univariate analysis, patient sex and transmission mode were not associated with death (both P=.8). All other predictors, including patient age (relative hazards,.; 95% CI,.-.5 per 0 years older), significantly predicted death (Table ). In multivariate analysis, only HIV- RNA levels, clinical stage, CD + cell counts, and -microglobulin levels predicted death. The strongest gradient of risk was observed for CD + cell counts, for which relative hazards ranged from.0 to 5.. In contrast, relative hazards ranged only from.0 to.8 for HIV- RNA. A significantly increased risk of death was also observed for patients with -microglobulin levels greater than 05 nmol/l (.6 mg/l), and for those who had the acquired immunodeficiency syndrome (AIDS) at baseline. CLINICAL DISEASE PROGRESSION During follow-up, 70 patients (69%) experienced clinical progression (06 CDC stage B events in patients who did not have AIDS at baseline, 6 AIDSdefining events, and 8 HIV-related deaths without prior recorded clinical progression). The overall clinical progression rate was per 00 person-years (95% CI, 0-6). Low HIV- RNA levels were significantly associated with fewer events (log-rank P.00) (Figure, C). Progression incidence rates in increasing quartiles of HIV- RNA were 6, 7, 65, and 9 per 00 personyears. In decreasing quartiles of CD + cell counts, progression incidence rates were 7, 6, 70, and 08 per 00 person-years (log-rank P.00) (Figure, D). In univariate analysis, patient sex and transmission mode were not associated with clinical progression (P=. and P=.9, respectively). All biological markers, clinical stage, and patient age (relative hazards,.; 95% CI,.-. per 0 years) significantly predicted clinical progression (Table ). However, only HIV- RNA levels and CD + cell counts were still statistically significant predictors in a multivariate model. Gradients of risk were similar for quartiles of CD + cell counts (.0-.) and HIV- RNA levels (.0-.). ARCH INTERN MED/ VOL 58, FEB 9, American Medical Association. All rights reserved. Downloaded From: on 0/06/08

4 Table. Relative Hazards of Death and Clinical Progression Associated With Clinical and Biological Markers* Death RH (95% CI) Clinical Progression RH (95% CI) Variable Univariate Multivariate Univariate Multivariate Log HIV- RNA, copies/ml (0.8-.) 0.7 (0.-.6). (.6-.6).7 (.-.6) (.5-.0). (0.7-.0). (.9-6.5).6 (.7-.0) (5.9-0.).8 (0.8-.8) 6. (.-9.6). (.0-.8) CD + counts, 0 9 /L (.-8.8) 5. (.-.7) 7. (.9-0.9). (.7-6.5) (.5-5.6).9 (.-6.7).5 (.0-6.7). (.-5.) (0.9-.6). (0.5-.8).6 (.0-.). (0.9-.) Clinical stage (CDC) A B.6 (.-6.).6 (0.9-.). (.6-.0)... C.5 (8.7-.0).5 (.8-6.7).5 (.6-.7)... -Microglobulin, nmol/l (.-.). (0.7-.).9 (.-.7) (.-9.).0 (.-.5).7 (.9-.8) (.9-8.7).7 (.0-.).9 (.0-.)... CD8 + counts, 0 9 /L (.-.7)....6 (.5-5.) (.-.0)....5 (.0-.) (0.9-.8).... (0.8-.8) Log ICD p Ag, pg/ml (.6-.7).... (.6-.) (.5-.5)....6 (.8-.9) (.-6.)....8 (.9-.)... *RH indicates relative hazards; CI, confidence interval; HIV-, human immunodeficiency virus type ; CDC, Centers for Disease Control and Prevention; and ICD p Ag, immune complex dissociated p antigen. P.00 except in multivariable model for -microglobulin for death RH ( P=.0). Adjusted for HIV- RNA levels, CD + cell counts, clinical stage, and -microglobulin levels. Adjusted for HIV- RNA levels and CD + cell counts. INTERACTIONS BETWEEN EFFECTS OF CD + CELL COUNTS AND HIV RNA LEVELS In the whole sample, a 0-times higher HIV- RNA level was associated with a -fold increase in mortality and a.6-fold increase in the risk of clinical progression. Similarly, a 0-times lower CD + cell count was associated with a 5-fold increase in mortality and a.8-fold increase in the risk of clinical progression. The predictive value of HIV- RNA differed across quartiles of CD + cell counts (test on interaction term: P=.00 for death and P=.0 for clinical progression). Among patients with initial CD + counts less than /L, HIV- RNA levels predicted neither death nor clinical progression (Figure, A). Similarly, a gradient of risk was observed for CD + cell counts across quartiles of HIV- RNA (Figure, B). The risk of death (P=.005) and clinical progression (P.00) also differed significantly according to HIV- RNA levels. The predictive value of CD + cell counts was greatest among patients who had low levels of HIV- RNA. A NEW MARKER: RNA/CD + RATIO The RNA/CD + ratio demonstrated to be an excellent predictor of death and clinical progression in the whole study population. This ratio varied between less than and 8.7 (mean,.9; SD,.5). In univariate analysis, the gradient of relative hazards went from.0 to 0. for death, and from.0 to 8.9 for clinical progression, across quartiles of the RNA/CD + ratio (Table ). These gradients were stronger than those of CD + cell counts or of HIV- RNA levels expressed by milliliters. In multivariate models, both RNA/CD + ratios and CD + cell counts were independent predictors of death and clinical progression (Table ). When added to a model with the RNA/CD + ratio, HIV- RNA expressed by milliliters was not a significant predictor (data not shown). COMMENT In this study, we assessed the role of HIV- RNA levels, CD + cell counts, and other markers as predictors of clinical outcomes in an HIV-infected population representative of patients seen in general medical practice. We complement previous studies - by showing that CD + cell counts provide important prognostic information about clinical outcomes even when HIV- RNA levels are taken into account. We also observed that the predictive value of clinical and biological markers differed for death and clinical progression. In addition, analysis across quartiles demonstrated that both CD + cell counts and HIV- ARCH INTERN MED/ VOL 58, FEB 9, American Medical Association. All rights reserved. Downloaded From: on 0/06/08

5 RH for 0 Higher HIV- RNA RH for 0 Lower CD + Cell Counts Death Clinical Progression < < < Quartiles of CD + Cell Counts, 0 9 /L < Quartiles of HIV- RNA, log copies/ml Figure. Relative hazards (RH), with 95% confidence intervals in logarithmic scale, of death and clinical progression for 0-times higher human immunodeficiency virus type (HIV-) RNA levels across quartiles of CD + cell counts (A) and for 0-times lower CD + cell counts across quartiles of HIV- RNA (B). A B RNA levels have a lower predictive value in patients with more advanced disease. Finally, the RNA/CD + ratio, which combines information from both markers, was the best single predictor of clinical progression and death. Although HIV- RNA was an excellent predictor of clinical outcome, CD + cell counts remained a powerful predictor even when HIV- RNA level was accounted for. In earlier studies, - performed in patients with a narrower range of CD + levels, the predictive value of CD + cell counts was marginal. However, these studies were not designed to compare the predictive value of HIV- RNA and CD + cell counts, and have emphasized the predictive value of HIV- RNA. Moreover, predictors provided different information according to the outcome analyzed. In multivariate models, CD + cell counts had a similar predictive value for both clinical progression and death, whereas HIV- RNA levels were better at predicting clinical progression than death. This finding is similar to previous results seen in a different patient population, 5 and suggest that viremia may affect survival and clinical progression in different ways. For example, if the effect of HIV- RNA on survival (but not on clinical progression) was mostly mediated by CD + cell depletion, the weak ability of HIV- RNA to predict death in CD-adjusted models would be due to overadjustment. On the other hand, high levels of HIV- RNA might be associated with an enhanced replication of other pathogens and thus affect clinical progression independently of CD + counts. 7-9 The predictive value of CD + cell counts and HIV- RNA levels varied across quartiles of the other predictor. Both biological markers lost partially (CD + cell counts) or completely (HIV- RNA) their predictive ability among patients with more advanced disease (HIV- RNA.5 copies/mlandcd + cellcount /L,respectively). Several mechanisms may explain this finding. In patients with very low CD + cell counts, the availability of target cells might limit HIV- replication and affect HIV- RNA levels. If so, HIV- RNA levels would not adequately reflect clinical prognosis in CD + -depleted patients. Another possibility is that among the most CD + -depleted patients, extrinsic factors, such as exposure to pathogens, may be more important for prognosis than intrinsic factors, such Table. Relative Hazards of Death and Clinical Progression Associated With Quartiles of RNA/CD + Ratio and CD + Cell Counts* Death RH (95% CI) Clinical Progression RH (95% CI) Variable Univariate Multivariate Univariate Multivariate RNA/CD + ratio, copies/cell (.0-5.6).8 (0.7-.6). (.5-.5).9 (.-.0) (.5-.6) 5. (.9-.) 6. (.-9.7).0 (.-6.7) ( ) 9.7 (.-7.5) 8.9 (5.8-.). (.-7.7) CD + counts, 0 9 /L (.-8.8).8 (.-0.) 7. (.9-0.9). (.-.) (.5-5.6). ( ).5 (.0-6.7).0 (.-.) (0.9-.6). (0.5-.9).6 (.0-.). (0.7-.7) *RH indicates relative hazards; CI, confidence interval. P.00 except in multivariate models for CD + counts for death RH ( P=.006) and clinical progression ( P=.00). Adjusted for RNA/CD + ratio and CD + cell counts. Downloaded From: on 0/06/08 ARCH INTERN MED/ VOL 58, FEB 9, American Medical Association. All rights reserved.

6 as viremia or CD + counts. Last, a 0-fold difference in either HIV- RNA levels or CD + cell counts may have a different biological meaning depending on the absolute level of each predictor. Because HIV- RNA levels and CD + cell counts were correlated, stratifying on the first variable also partially stratifies on the second. Whichever explanation is correct, our findings suggest that measurement of HIV- RNA contributes little prognostic information in patients with very low CD + cell counts. The relative hazards reported in this study may be underestimated, because the sickest patients have preferentially received antiviral drugs, which in turn may improve their outcome. Because antiviral drugs were not administered at random, we did not examine whether antiviral treatment improved clinical outcome. 0 Replication of HIV is the main factor leading to CD + cell death, and HIV- RNA levels reflect the number of HIV-infected cells. Thus, the HIV- RNA/CD + ratio represents the contributions of these markers to patient prognosis. Indeed, the RNA/CD + ratio was the best single predictor of clinical disease progression and death in our population. The high predictive ability of the ratio may be due to the fact that it best reflects viral dynamics, which plays a central role in a patient s prognosis. Alternatively, the RNA/ CD + ratio may be a strong predictor because it combines information from variables (viral replication and immune response) that influence outcome independently. Regardless of the underlying mechanism, the RNA/CD + ratio was amuchbetterpredictorthanhiv-rnaexpressedbyplasma volume: the ratio predicted death even in CD + -depleted patients, and patients in the highest quartile of the RNA/ CD + ratio had a 0-fold increase in mortality, compared with the lowest quartile. Such patients may qualify for aggressive antiretroviral treatment even in the absence of evidence from controlled trials. The RNA/CD + ratio may also deserve an evaluation as surrogate marker of clinical outcomes in clinical trials. In summary, HIV- RNA level is a useful prognostic marker in HIV-infected patients, but its importance may have been overestimated in recent literature. In particular, viremia levels have virtually no predictive ability in patients with severe immunosuppression. Viremia should not be used in isolation, as CD + cell counts contribute important and independent prognostic information, particularly when predicting death. Finally, the HIV- RNA/CD + ratio is the most useful predictor of clinical outcome and deserves further attention. Accepted for publication June 6, 997. Presented in part at the th International Conference on AIDS, Vancouver, British Columbia, July 0, 996. We thank K. Zollinger for excellent technical assistance, V. Gabriel for database management, and O. Rutschmann, MD, A. Halfon-Poletti, MD, G. Figueras, MD, P. Lorenzi, MD, A. Mendoula, MD, and H. Wolff, MD, for reviewing all patient clinical charts. This study took advantage of the infrastructure of the Swiss HIV Cohort Study, whose members are M. Battegay, Ph. Bürgisser, R. Doorly, M. Egger, P. Erb (chair, Laboratories ), W. Fierz, M. Flepp (chair, Clinics ), P. Francoli (president), P. Grog, U. Grüninger, B. Hirschel (chair of the scientific board), B. Ledergerber, R. Lüthy, R. Malinverni, L. Matter, M. Opravil, F. Paccaud, L. Perrin, W. Pichler, M. Rickenbach (manager of the data center), O. Rutschmann, P. Vernazza, and J. von Overbeck. Reprints: Luc Perrin, MD, Laboratory of Virology, Geneva University Hospital, Geneva, Switzerland ( perrin@hcuge.ch). REFERENCES. Henrard DR, Phillips JF, Muenz LR, et al. Natural history of HIV- cell-free viremia. JAMA. 995;7: Mellors JW, Kingsley LA, Rinaldo CR, et al. Quantitation of HIV- RNA in plasma predicts outcome after seroconversion. Ann Intern Med. 995;: Mellors JW, Rinaldo CR, Gupta P, White RM, Todd JA, Kingsley LA. Prognosis in HIV- infection predicted by the quantity of virus in plasma. Science. 996; 7: O Brien TR, Blattner WA, Waters D, et al. Serum HIV- RNA levels and time to development of AIDS in the Multicenter Hemophilia Cohort Study. JAMA. 996;76: Galetto-Lacour A, Yerly S, Perneger TV, et al. Prognostic value of viremia in patients with long-standing human immunodeficiency virus infection. J Infect Dis. 996;7: Ruiz L, Romeu J, Clotet B, et al. Quantitative HIV- RNA as a marker of clinical stability and survival in a cohort of 0 patients with a mean CD cell count of /l. AIDS. 996;0:F9-F. 7. Phillips AN, Eron JJ, Bartlett JA, et al. HIV- RNA levels and the development of clinical disease. AIDS. 996;0: Yerly S, Kaiser L, Mermillod B, Baumberger C, Hirschel B, Perrin L. Response of HIV RNA to didanosine as predictive marker of survival. AIDS. 995;9: O Brien WA, Hartigan PM, Martin D, et al. Changes in plasma HIV- RNA and CD+ lymphocyte counts and the risk of progression to AIDS. N Engl J Med. 996;: Welles SL, Jackson JB, Yen-Liebermen B, et al. Prognostic value of plasma human immunodeficiency virus type (HIV-) RNA levels in patients with advanced HIV- disease and with little or no prior zidovudine therapy. J Infect Dis. 996;7: Coombs RW, Welles SL, Hooper C, et al. Association of plasma human immunodeficiency virus type RNA level with risk of clinical progression in patients with advanced infection. J Infect Dis. 996;7: Katzenstein DA, Hammer SM, Hughes MD, et al. The relation of virologic and immunologic markers to clinical outcomes after nucleoside therapy in HIVinfected adults with 00 to 500 CD cells per cubic millimeter. N Engl J Med. 996;5: Center for Disease Control and Prevention. 99 Revised classification system for HIV infection and expanded surveillance case definition for AIDS among adolescents and adults. MMWR Morb Mortal Wkly Rep. 99;(RR-7): -9.. Ledergerber B, von Overbeck J, Egger M, Lüthy R. The Swiss HIV Cohort Study: rationale, organization and selected baseline characteristics. Soz Praventivmed. 99;9: Schüpbach J, Böni J, Tomasik Z, Jendis J, Kind C. Sensitive detection and early prognostic significance of p antigen in heat-denatured plasma of human immunodeficiency virus type -infected infants. J Infect Dis. 99;70: Collett D. Modelling Survival Data in Medical Research. London, England: Chapman & Hall; Sieczkowski L, Chandran B, Wood C. The human immunodeficiency virus tat gene enhances replication of human herpesvirus-6. Virology. 995;: Biggs BA, Hewish M, Kent S, Hayes K, Crowe SM. HIV- infection of human macrophages impairs phagocytosis and killing of Toxoplasma gondii. J Immunol. 995;5: Telfer PT, Brown D, Devereux H, Lee CA, DuSheiko GM. HCV RNA levels and HIV infection: evidence for a viral interaction in haemophilic patients. Br J Haematol. 99;88: Glesby MJ, Hoover DR. Survivor treatment selection bias in observational studies: examples from AIDS literature. Ann Intern Med. 996;: Wei X, Ghosh SK, Taylor ME, et al. Viral dynamics in HIV- infection. Nature. 995;7:7-.. Ho DD, Neumann AU, Perelson AS, Chen W, Leonard JM, Markowitz M. Rapid turnover of plasma virion and CD lymphocytes in HIV- infection. Nature. 995; 7:-6. ARCH INTERN MED/ VOL 58, FEB 9, American Medical Association. All rights reserved. Downloaded From: on 0/06/08