Plasma HIV RNA levels and CD4+ lymphocyte counts are established markers of HIV disease progression and quantify an individual's risk for AIDS and death. Current guidelines for the initiation of antiretroviral therapy among asymptomatic HIV infected individuals in both the developed  and developing  world are based upon measurements of these markers. These guidelines have been established based upon data from prospective epidemiological studies that have linked single marker measurements and longitudinal marker trends with the incidence of clinical events. [3–5]
Obtaining CD4+ cell counts and HIV RNA levels may be impractical for countries with scarce resources, due to the expense of the assays and technical factors such as access to the proper equipment [6,7]. Therefore there is a need for prognostic markers that can be used in resource-limited settings. Two markers that may fulfill this need are total lymphocyte count (TLC), and hemoglobin (Hgb) level. Both markers are relatively inexpensive and straightforward to measure. TLC has been shown to correlate with CD4+ cell counts, particularly among symptomatic individuals [6,8] and single measurements have been shown to have some prognostic information for survival [7,9,10] and progression to AIDS . Furthermore, World Health Organization guidelines for developing countries  suggest the use of TLC for disease monitoring if CD4+ cell counts are not known. Hgb levels, often examined as a dichotomous anemia variable, have also been shown to be associated with progression to AIDS [12–14].
Longitudinal patterns of averaged TLC and Hgb measurements have been described at the population level [15,16], but detailed descriptions of the trajectories of markers within individuals have not been published. Identifying patterns within individuals capitalizes on the repeated observations of the same individual and extends findings at the population level. Furthermore, by analyzing individual patterns, one can assess the timing of important marker patterns over the course of disease and the suitability of the marker for monitoring disease progression in a given person.
In this study, we analyzed longitudinal measurements of TLC and Hgb among men enrolled in the Multicenter AIDS Cohort Study (MACS) obtained before the widespread use of highly-active antiretroviral therapy (HAART). We related these markers to the development of clinically defined AIDS as well as to the time of T-cell homeostasis failure, which we have previously shown to be an important immunological milestone related to HIV disease pathogenesis [15–20]. Using approaches previously applied to the investigation of changes in T-cell homeostasis [17,18], we estimated the prevalence, magnitude, and timing of within-individual changes in TLC and Hgb levels.
Materials and methods
Study population and laboratory methods
The MACS was initiated in 1983 to study the natural history of HIV infection among homosexual and bisexual men in the United States. The design of the MACS has been described  and only aspects pertinent to this study are presented here. In 1984–85, 4954 HIV-positive and HIV-negative men were enrolled in Baltimore/Washington DC, Chicago, Los Angeles and Pittsburgh, with an additional 625 men enrolled between 1987 to 1991. Men with AIDS and men younger than 18 years were excluded from enrollment. At semi-annual visits, men returned to the clinics to provide specimens for laboratory analyses, undergo a physical examination, and complete self-administered data forms and an interviewer-administered questionnaire.
T-cell subset levels on all men at each visit were measured in peripheral blood stained with monoclonal antibodies by a whole blood lysing method and analyzed by two-color flow cytometry [22,23] and monoclonal antibodies specific for CD3+, CD4+, and CD8+ lymphocytes. Absolute numbers of cells per microliter of blood were calculated using the complete blood count with automated 10 000 cell differential. Hemoglobin concentrations were measured using standard techniques.
The individuals included in the present study consisted of MACS participants with at least six visits up to November 1996 (prior to widespread HAART use in MACS) of which a minimum of four visits had to be free of AIDS. Requiring six visits was necessary for our segmented regression methods , and the latter requirement was designed to ensure adequate numbers of data points prior to AIDS. The visits occurring in the first year after seroconversion, or entry into the study if the individual was seroprevalent for HIV, were omitted from the analysis to minimize the variability that is often seen in marker levels during the acute phase of HIV infection. For individuals developing AIDS, at least two visits within 1.5 years preceding AIDS were required to ensure that the marker patterns would be discernible at the time nearest to AIDS. Lastly, individuals who developed AIDS after October 1994 were excluded to allow an unbiased estimate of the interval between the point in time when the change in marker trajectory is most likely to occur (termed the inflection point, described below) and AIDS. If AIDS cases after 1994 had been included in the analysis, the inflection points would have been forced to occur before the onset of AIDS due to insufficient follow-up after AIDS, because the statistical methods require several visits before and after the inflection points to determine well-defined marker slopes.
Statistical methods for characterizing trajectories and changes in longitudinal markers
The TLC and Hgb values were logarithmically transformed to improve the fit of the linear models. To provide a comprehensive description of the changes in marker levels at the individual level, segmented regression models were adapted from methods originally developed for investigating changes in CD3+ lymphocytes . The methods select, for every individual, a point in time when the change in marker slope is most likely to have occurred (termed the inflection point, IP). Briefly, candidate IPs are defined as the midpoints in time between any two consecutive visits. At least three marker values are required before and after each candidate IP to ensure that slopes are well defined. In addition, the marker levels are required to decrease during the period surrounding the candidate IPs (due to the premise is that a decrease in the marker levels precedes disease progression). The most probable time of inflection is identified by selecting from all candidate inflection points the point that produces the smallest overall residual error . This forces an IP to be identified for every individual to whom the method is applied. The parameters obtained from the application of these methods to individual trajectories include: (1) an IP that represents the time when a change in trajectory was most likely to have occurred, (2) the slope before the IP, (3) the slope after the IP, and (4) the level of the marker at the IP.
Using non-parametric Wilcoxon-rank statistics, we compared the distributions of these parameters, which describe the longitudinal trends in TLC and Hgb measurements, among the different groups. To examine whether this information could be used to discriminate between those who developed AIDS and those who remained AIDS-free, we computed receiver operating characteristic (ROC) curves to describe the sensitivity and specificity at each level of the parameters, and we compared the area under the curves to determine significant differences between ROCs . This enabled us to be able to identify a cutoff, which could be used to distinguish ‘true’ inflections from ‘false’ inflections in which trajectory changes occur due to the natural variability of the marker levels over time. Lastly, we also investigated the timing of inflections relative to the times of clinical AIDS, death, and T-cell homeostasis failure, identified using previously developed segmented regression models of the CD3+ T-cell counts .
Characteristics of study population
Of the 5579 individuals in MACS, 2195 were seroprevalent and 511 had seroconverted prior to 1996. Figure 1 illustrates the number of MACS participants who were included in the present study stratified by HIV status and progression to AIDS. The total number of men contributing data to the TLC analysis is 3299 and to the Hgb analysis is 3289 individuals. The median [interquartile range (IQR)] number of marker measurements per individual was 14 (10, 18). At the first visit included in the analysis, the seroprevalent population who did not progress to AIDS had median (IQR) CD4+ lymphocyte count of 652 (486, 882) cells × 106/l, a median TLC (IQR) of 1960 (1568, 2068) × 106 cells/l and a median (IQR) Hgb concentration level of 15.2 (14.5, 16.0) g/dl. The seroprevalent men who developed AIDS had median (IQR) levels which were slightly lower: a CD4+ lymphocyte count of 532 (392, 710) × 106 cells/l, a TLC of 1793 (1444, 2303) × 106 cells/l, and a Hgb level of 15.1 (14.5, 15.8) g/dl.
Longitudinal marker patterns
Figure 2 displays the distribution of the slopes (% change in marker per year) prior to (panels a and c) and subsequent to the IP (panels b and d) for the markers TLC and Hgb stratified by HIV and AIDS status. For both TLC and Hgb, there were no meaningful differences in the slopes prior to the IP among the groups of individuals (Fig. 2a and 2c). The slopes were essentially symmetrically distributed around 0 for TLC (Fig. 2a), and showed a very slight decline for Hgb (Fig. 2c).
In contrast, there were substantial differences in the TLC slopes after inflection between those progressing to AIDS and those remaining free of AIDS (Fig. 2b). Specifically, after the inflection, there was a rapid rate of decline in TLC for the group progressing to AIDS, which was substantially steeper (P < 0.0001) than the minimal post-IP declines among those remaining AIDS free and among those seronegative for HIV. Similarly, the post-IP Hgb annual changes were significantly steeper (P < 0.0001) among those progressing to AIDS than among the other two groups (Fig. 2d). These results were similar when investigating both seroprevalent and seroconverter groups separately, and there was little difference in post-IP slope distributions when stratifying results by quartiles of the total number of contributed visits.
The group developing AIDS had a slightly lower median TLC and Hgb levels (1710 × 106 cells/l and 14.4 g/dl) at the estimated IP than the group not developing AIDS (2032 × 106 cells/l and 14.8 g/dl) and the seronegative group (2223 × 106 cells/l and 15.3 g/dl). This is in part due to the lower levels of the marker at the first visit included in the analyses for individuals who developed AIDS, since both markers were relatively stable prior to IP.
Figure 3 shows the observed TLC and Hgb levels over time and the fitted segmented regression model to three representative individuals among those HIV-positive developing AIDS (Fig. 3a), HIV-positive without onset of AIDS (Fig. 3b), and HIV-negative (Fig. 3c). The individuals selected for this figure were those with pre- and post-IP slopes closest to the median pre- and post-IP trajectories for both markers within each of the three groups.
Figure 3 illustrates the need to distinguish ‘true’ inflection from trajectory changes estimated in the presence of the natural variability of the markers, since the algorithm assigns an IP for every individual. Based on the ROC analysis (where the gold-standard of disease was defined by an AIDS diagnosis), the post-inflection slopes were the best parameters (for both TLC and Hgb) to discriminate between those developing AIDS and those remaining free of AIDS. In other words, the ROC for this parameter had the largest area under the curve when compared to the ROC from the pre-inflection slope and the level at the IP.
To quantify the relative sensitivity and specificity of TLC and Hgb for development of AIDS, we compared the resulting ROC curves for the post-inflection slopes to the corresponding curve for the post-IP slope of the CD3+ lymphocyte count , a marker of the failure of T-cell homeostasis and an established predictor of AIDS [16,17,20] (Fig. 4). The cutoffs which maximized the sensitivity and specificity for these markers were annual declines of 10.2, 2.2, 10.9% for the TLC, Hgb and CD3+ lymphocyte count, respectively. At these cutoffs, 76.9, 78.6, and 80.2% of the men with AIDS onset, would be classified as true inflectors in the respective markers as compared to only 23.1, 21.5, and 19.7% of the men remaining free of AIDS (Table 1). The calculated areas under the curve (AUC) for TLC and Hgb were slightly smaller than that of the CD3+ lymphocyte count, showing that TLC and Hgb were about as strong as CD3+ lymphocyte counts to distinguish those developing AIDS from those not developing AIDS (Table 1).
We investigated the robustness of the markers in distinguishing between individuals progressing to AIDS and individuals remaining free of AIDS in two ways. First, we restricted our analysis to individuals who had their IP prior to AIDS, excluding 87 individuals from the TLC ROC analysis and 79 from the Hgb ROC analysis. This improved the sensitivity and specificity slightly and yielded estimated optimal annual post-IP declines of 11.1% for TLC, 2.6% for Hgb, and 11.5% for CD3+ lymphocyte count. The AUC also improved slightly for TLC, Hgb, and for CD3+ lymphocyte count, with the result that there were no significant differences between the abilities of the three markers to distinguish those developing AIDS onset from those remaining free of AIDS (Table 1).
Second, we were concerned about the influence of zidovudine use, because anemia has been shown to occur in approximately 25% of individuals initiating zidovudine  and there may be selection biases associated with initiation of therapy as individuals become ill. Among those with data available and at risk for initiating zidovudine, 18.2 and 22.8% initiated zidovudine in the year immediately before or after the estimated Hgb IP, respectively. When we restricted the ROC analysis to those not receiving zidovudine in the year prior to the Hgb IP, the year after the IP, or prior to AIDS, all inferences remained the same; specifically, a rapid decline in Hgb could distinguish those progressing to AIDS from those remaining free of AIDS. In fact, the maximum sensitivity and specificity improved slightly to between 80.5 and 81.4%, and the AUC improved to approximately 0.85. These results suggest that the influence of zidovudine initiation on our results was minimal.
Timing of rapid declines in markers
By examining inflection points on the individual level, we could evaluate the timing of CD4+ cell count changes and inflections in TLC, Hgb and CD3+ lymphocytes relative to the times of AIDS and death (Table 2). Our results showed that the inflections in TLC and Hgb of those who developed AIDS occurred on average 0.25 years after CD4+ cell counts first fell below 350 × 106 cells/l, 0.75 years before CD4+ cell counts first fell below 200 × 106 cells/l, 1.6 years before clinically-defined AIDS, and 2.8–2.9 years before death in this essentially untreated cohort. These intervals were similar whether the analysis was restricted to the population of seroconverters or to the seroprevalent population, rather than the entire seropositive population.
As expected, the timing between the TLC IP and CD3+ IP was strongly concordant: 70.3% of individuals had their CD3+ IP concurrent with their TLC IP, demonstrating that the TLC IP primarily reflects the failure of T-cell homeostasis [16,17]. The Hgb IP also occurred very close to the CD3+ IP (IQR, 0.96 years prior; 0.81 years subsequent), although only 18.3% were precisely concurrent.
Using segmented regression models and the extensive natural history data available in the MACS, we have demonstrated that both TLC and Hgb markers demonstrate a period of stability for many years after HIV infection, followed by a period of rapid decline beginning 1 to 2 years before the onset of AIDS. The results of our analysis explain why the correlation between TLC and CD4+ counts in asymptomatic individuals has been reported to be lower than the correlation found for symptomatic persons [6,8]: since CD4+ cell counts generally decline steadily over the course of HIV disease, they will show low correlation with TLC markers during the time when TLC is stable, but higher correlation later in disease when TLC is also declining. Additionally, our results demonstrate why these markers are more prognostic at low CD4+ cell counts. Until the time of inflection, TLC and Hgb markers are stable and provide little information about the risk of disease progression, but late in disease (when CD4+ cell counts are generally between 200 and 350 × 106 cells/l) the markers demonstrate rapid decline and thus distinguish those individuals who will soon progress to AIDS or death.
Our results also indicate that the decline in TLC and Hgb begins concomitantly with the declines in CD3+ lymphocytes that signals the failure of T-cell homeostasis, approximately 1.56 years prior to AIDS [16–18]. The sensitivity and specificity post-IP slopes of TLC demonstrated that TLC discriminated nearly as well as post-IP slopes of CD3+ lymphocyte counts. This disparity may reflect the variation in TLC due to non-T lymphocytes, which obscures to some extent the changes in the CD3+ lymphocytes. However, this disparity appears to be minor, as there were only slight differences between the ROC curves for TLC and Hgb compared to the ROC curve for CD3+ lymphocyte cell count, as shown by the area under the curve.
The reason why Hgb begins to decline close to the time of the failure of T-cell homeostasis is not immediately clear. There are several hypothesized causes of HIV-associated anemia, including effects of plasma HIV on bone marrow stromal cells [26,27], and disturbances of cytokines that affect hematopoiesis [14,28]. We have previously shown on both the population and individual level that HIV RNA levels remain stable and then show accelerated increases prior to AIDS [5,29], possibly due to the expansion of HIV variants that use the CXCR4 co-receptor . If the magnitude of HIV on stromal cells or other mechanisms of hematopoiesis is related to HIV RNA, then Hgb decline may occur after control of HIV replication has been lost.
Our methods, which estimate change-points on an individual level, extend methods which evaluate changes between two successive interval slopes . While the segmented regression methods force the estimation of an inflection point for every individual, some inflections will occur by chance. This motivated the use of receiver-operator methods to evaluate the ability of post-inflection point slopes to distinguish those with and without incipient disease progression. Our results showed that we were able to distinguish these groups, validating the importance of the changes in TLC and Hgb trajectories. Our analysis excluded men who did not contribute at least six laboratory marker measurements, four of which occurred prior to AIDS. This excludes fast progressors whose period of TLC and Hgb stability may be either relatively short or non-existent. As a result of this exclusion, estimates of the timing of marker inflections and clinical disease may be slightly longer than the timing among all HIV-infected individuals. We feel that the impact is negligible given the small number of individuals who progress within the first 3 years after infection and the variability in inflection points as noted in Table 2.
As we have pointed out using slightly different methods , the initial decline in TLC and Hgb generally occurs between an individual's first CD4+ lymphocyte count of 350 and 200 × 106 cells/l, the time when the current guidelines recommend initiating antiretroviral therapy [1,2]. Anemia has been demonstrated to resolve after initiation of highly-active antiretroviral therapy ; the pattern of TLC over this time is less clear. Furthermore, in conjunction with previous studies showing the association of changes in TLC and Hgb with disease progression [7,9–14], our results suggest that TLC and Hgb may also be useful for pathogenesis studies of HIV infection, since factors associated with the onset of rapid declines can be evaluated at the biologically important time point identified in this analysis.
Our results also suggest that TLC and Hgb may be suitable markers for monitoring and evaluating HIV-infected individuals, which has been an identified challenge in countries with limited resources . TLC and Hgb levels are relatively inexpensive to measure and are easily obtained with little sophisticated technology. Thus, our results suggest that one could monitor these markers in patients over time, with little risk of clinical AIDS up to the point of rapid decline. Even if patients could not be monitored prior to the inflection point, our results demonstrate that rapid declines in markers rarely occur among those with no immediately incipient clinical HIV-related disease. Thus, individuals seen with declines of the magnitude demonstrated in this paper would be at highest risk for clinical progression. This would be particularly useful for clinicians in developing countries.
Applying our methods prospectively for identifying TLC and Hgb inflections in a clinical setting would not be without challenges, however. Differences in endemic disease, diet, host genetics, viral characteristics, and other factors might affect TLC and Hgb, possibly leading to misleading predictions of risk . Our results reflect the semi-annual visit schedule of the MACS, and in a different setting, important changes in markers could occur within this 6-month time interval. More frequent sampling of marker levels than was done in the MACS might aid in identifying inflections prospectively, but the utility of this approach remains to be validated. For these reasons, we believe further research in appropriate populations, is warranted to examine the feasibility of utilizing total lymphocyte count and hemoglobin concentrations as markers to monitor HIV disease progression in individuals in areas with limited resources.
Data in this manuscript were collected by the Multicenter AIDS Cohort Study (MACS) with centers (Principal Investigators) at The Johns Hopkins University Bloomberg School of Public Health (Joseph B. Margolick, Alvaro Muñoz), Howard Brown Health Center and Northwestern University Medical School (John Phair), University of California, Los Angeles (Roger Detels, Beth Jamieson), and University of Pittsburgh (Charles Rinaldo).
Sponsorship: The MACS is funded by the National Institute of Allergy and Infectious Diseases, with additional supplemental funding from the National Cancer Institute. UO1-AI-35042, 5-MO1-RR-00722 (GCRC), UO1-AI-35043, UO1-AI-37984, UO1-AI-35039, UO1-AI-35040, UO1-AI-37613, UO1-AI-35041.
Website located at http://statepi.jhsph.edu/macs/macs.html.
1. Panel on Clinical Practices for Treatment of HIV Infection. Guidelines for the Use of Antiretroviral Agents in HIV-Infected Adults and Adolescents.
Washington, DC: Department of Health and Human Services 4 February 2002.
2. World Health Organization. Scaling up antiretroviral therapy in resource-limited settings: Guidelines for a public health approach.
Geneva: World Health Organization; April 2002.
3. Polk BF, Fox R, Brookmeyer R, Kanchanaraksa S, Kaslow R, Visscher B, et al
. Predictors of the acquired immunodeficiency syndrome developing in a cohort of seropositive homosexual men.N Engl J Med
4. Mellors JW, Muñoz A, Giorgi JV, Margolick JB, Tassoni CJ, Gupta P, et al
. Plasma viral load and CD4+ lymphocytes as prognostic markers of HIV-1 infection.Ann Intern Med
5. Lyles RH, Muñoz A, Yamashita TE, Bazmi H, Detels R, Rinaldo CR, et al
. Natural history of human immunodeficiency virus type 1 viremia after seroconversion and proximal to AIDS in a large cohort of homosexual men Multicenter AIDS Cohort Study.J Infect Dis
6. Beck EJ, Kupek EJ, Gompels MM, Pinching AJ. Correlation between total and CD4 lymphocyte counts in HIV infection: not making the good an enemy of the not so perfect.Int J STD & AIDS
7. Schechter M, Zajdenverg R, Machado LL, Pinto ME, Lima LA, Perez MA. Predicting CD4 counts in HIV-infected Brazilian individuals: a model based on the World Health Organization staging system.J Acquir Immune Defic Syndr
8. Fournier AM, Sosenko JM. The relationship of total lymphocyte count to CD4 lymphocyte counts in patients infected with human immunodeficiency virus.Am J Med Sci
9. Montaner JSG, Le TN, Le N, Craib KJ, Schechter MT. Application of the World Health Organization system for HIV infection in a cohort of homosexual men in developing a prognostically meaningful staging system.AIDS
10. Lima LAA, May SB, Perez MA, Schechter MT. Survival of HIV-infected Brazilians: a model based on the World Health Organization staging system.AIDS
11. Post FA, Wood R, Maartens G. CD4 and total lymphocyte counts as predictors of HIV disease progression.
QJM 1996; 89
12. Mocroft A, Kirk O, Barton SE, Dietrich M, Proenca R, Colebunders R, et al
. Anaemia is an independent predictive marker for clinical prognosis in HIV-infected patients from across Europe. EuroSIDA study group.AIDS
13. Moore RD, Keruly JC, Chaisson RE. Anemia and survival in HIV infection.J Acquir Immune Defic Syndr
14. Sullivan PS, Hanson DL, Chu SY, Jones JL, Ward JW. Epidemiology of anemia in human immunodeficiency virus (HIV)-infected persons: results from the Multistate Adult and Adolescent Spectrum of HIV Disease Surveillance Project.Blood
15. Margolick JB, Donnenberg AD, Chu C, O'Gorman MR, Giorgi JV, Muñoz A. Decline in total T cell count is associated with onset of AIDS, independent of CD4(+) lymphocyte count: implications for AIDS pathogenesis.Clin Immunol Immunopathol
16. Margolick JB, Muñoz A., Donnenberg AD, Park LP, Galai N, Giorgi JV, et al
. Failure of T-cell homeostasis preceding AIDS in HIV-1 infection. The Multicenter AIDS Cohort Study.Nat Med
17. Gange SJ, Muñoz A, Chmiel JS, Donnenberg AD, Kirstein LM, Detels R, Margolick JB. Identification of inflections in T-cell counts among HIV-1-infected individuals and relationship with progression to clinical AIDS.Proc Natl Acad Sci USA
18. Maas JJ, Gange SJ, Schuitemaker H, Coutinho RA, van Leeuwen R, Margolick JB.Strong association between failure of T cell homeostasis and the syncytium-inducing phenotype among HIV-1-infected men in the Amsterdam Cohort Study.AIDS
19. Shankarappa R, Margolick JB, Gange SJ, Rodrigo AG, Upchurch D, Farzadegan H, et al
. Consistent viral evolutionary changes associated with the progression of human immunodeficiency virus type 1 infection.J Virol
20. Galai N, Margolick JB, Astemborski J, Vlahov D. Existence and failure of T-cell homeostasis prior to AIDS onset in HIV-infected injection drug users.Clin Immunol Immunopathol
21. 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
22. Giorgi JV, Cheng HL, Margolick JB, Bauer KD, Ferbas J, Waxdal M, et al.Quality control in the flow cytometric measurement of T-lymphocyte subsets: the multicenter AIDS cohort study experience. The Multicenter AIDS Cohort Study Group.Clin Immunol Immunopathol
23. Schenker EL, Hultin LE, Bauer KD, Ferbas J, Margolick JB, Giorgi JV. Evaluation of a dual-color flow cytometry immunophenotyping panel in a multicenter quality assurance program.Cytometry
24. Hanley JA, McNeil BJ. The meaning and use of the area under a receiver operating characteristic (ROC) curve.Radiology
25. Richman DD, Fischl MA, Grieco MH, Gottlieb MS, Volberding PA, Laskin OL, et al.The toxicity of azidothymidine (AZT) in the treatment of patients with AIDS and AIDS-related complex. A double-blind, placebo-controlled trial.N Engl J Med
26. Bahner I, Kerans K, Coutinho S, Leonard EH, Kohn DB.Infection of human marrow stroma by HIV-1 is both required and sufficient for HIV-1 induced hematopoietic suppression in vitro: demonstration by gene modification of primary human stroma.Blood
27. Scadden DT, Zeira M, Woon A, Wang Z, Schieve L, Ikeuchi K, et al.Human immunodeficiency virus infection of human bone marrow stromal fibroblasts.Blood
28. Claster S. Biology of anemia, differential diagnosis and treatment options in human immunodeficiency virus infection.J Infect Dis
29. Gange SJ, Mellors JW, Lau B, Detels R, Phair JP, Muñoz A, Margolick JB.Longitudinal patterns of HIV type 1 RNA among Individuals with late disease progression.AIDS Res Hum Retroviruses
30. Detels R, English PA, Giorgi JV, Visscher BR, Fahey JL, Taylor JM, et al.Patterns of CD4+ cell changes after HIV-1 infection indicate the existence of a codeterminant of AIDS.J Acquir Immune Defic Syndr
31. Gange SJ, Lau B, Phair JP, Riddler S, Detels R, Margolick JB.Rapid declines in total lymphocyte count and hemoglobin in HIV infection begin at CD4 lymphocyte counts that justify antiretroviral therapy.AIDS
32. Moore RD, and Forney D. Anemia in HIV-infected patients receiving highly active antiretroviral therapy.J Acquir Immune Defic Syndr
33. Stephenson J. Cheaper HIV drugs for poor nations bring a new challenge: monitoring treatment.JAMA
2002; 288: 151–153.
Keywords:© 2003 Lippincott Williams & Wilkins, Inc.
epidemiology; natural history; statistics; haematology; markers