Before the advent of combination antiretroviral therapy (cART), anemia was a common complication of advanced HIV infection1 associated with significantly shorter survival.1-5 Zidovudine (AZT), a highly effective component of many cART regimens, can cause or exacerbate anemia in some patients, although cART with or without AZT has proven to be an effective treatment for the anemia of HIV infection.6 AZT remains a recommended first-line antiretroviral agent in treatment guidelines for the United States7 and for resource-limited settings.8 A widely available and affordable drug, AZT's use in developed nations has due to concerns about anemia and other toxicities and the availability of equally potent but better tolerated agents. Among resource-limited nations, AZT is part of numerous national fixed-dose regimens, including in locations where the prevalence of anemia may be elevated because of malaria, malnutrition, and other non-HIV-related causes. Thus, in developed and resource-limited settings, AZT is an option that may be considered for patients with some degree of underlying anemia. To characterize the impact of AZT on anemia among HIV-infected patients better, we reviewed data from a large US observational cohort, describing changes in hemoglobin (Hb) that occurred after initial prescription of AZT-containing and non-AZT-containing regimens.
We analyzed data collected from medical records of HIV-infected persons observed in the Adult and Adolescent Spectrum of HIV Disease (ASD) project from January 1998 through June 2004. The methods used in the ASD project have been described.9 At more than 100 medical inpatient and outpatient facilities in 11 US cities, HIV-infected persons ≥13 years of age were selected at their first health care visit after the project was initiated. Data abstractors reviewed all available medical records for demographic characteristics, mode of HIV exposure, and occurrences of AIDS-defining opportunistic illnesses (AIDS OIs). Initial enrollment included information from the preceding 12 months on medications prescribed and laboratory values, with review of medical records every 6 months thereafter until death or loss to follow-up.
For these analyses, we used data from patients in the cohort who were presumed to be antiretroviral drug naive before initiation of cART. We consider patients to be antiretroviral naive if they had no documentation of previously prescribed antiretroviral therapy, if they had no history of AIDS OIs, if they initiated a cART regimen commonly prescribed as first-line therapy for HIV-infected patients according to treatment guidelines,7 and if their immunologic and virologic parameters within 180 days before cART initiation suggested they were plausible candidates for therapy (ie, viral load measurement >10,000 copies/mL and CD4 count <500 cells/μL). We excluded patients who had a history of treatment of anemia with erythropoietin or transfusion.
We used laboratory data to define anemia status at the time of initiation of antiretroviral therapy. We used gender-specific definitions1: men were considered to be nonanemic if their most recent Hb measurement was >14.0 g/dL, to have mild anemia if 10 < Hb ≤ 14 g/dL, to have moderate anemia if 8 < Hb ≤ 10 g/dL, and to have severe anemia if they had an Hb measurement ≤8.0 g/dL. Women were considered to be nonanemic if their most recent Hb measurement was >12.0 g/dL, to have mild anemia if 10 < Hb ≤ 12 g/dL, to have moderate anemia if 8 < Hb ≤10 g/dL, and to have severe anemia if their Hb was ≤8.0 g/dL. We defined improvement of anemia using follow-up Hb measurements; in each 6-month interval, the lowest and most recent Hb measurements were recorded. Improvement of anemia was defined as an increase in Hb from baseline of at least 1.0 g/dL, with a concomitant decrease in the severity category of anemia (eg, from moderate to mild).
We used Kaplan-Meier estimates to compare the cumulative probability of improvement of anemia among persons taking and not taking AZT-containing cART, stratified by anemia category at cART initiation. We censored observation at the time of first treatment with erythropoietin or transfusion after initiation of cART; at date of last follow-up; or on June 30, 2004. We used log-rank tests for equality over strata of AZT exposure to assess differences in cumulative probabilities of improvement by AZT status.
We used Cox proportional hazards regression to estimate the hazard of improvement of anemia, controlling for CD4 cell count, viral load, gender, baseline anemia category, and race/ethnicity, and censoring patients at the time of their last observation; on June 30, 2004; or at the time of treatment with erythropoietin or transfusion. The results were expressed as a hazard ratio (HR) with a 95% confidence interval (CI).10 We first ran regression models using data from all available patients in an intent-to-treat (ITT) analysis. We then compared the results of this ITT analysis with those of an as-treated (AT) analysis; however, the AT analysis could be conducted only on a subset of patients who initiated cART before changes in the ASD project software that collected precise dates of therapy initiation and dates of medication changes. To allow a direct comparison, we thus also conducted a subset ITT analysis using only the data from this same subset of patients whose data were collected after the software update. We report the hazard of improvement of anemia with a 95% CI. An HR <1.0 indicated a lower hazard of improvement (eg, a longer time to improvement), whereas an HR higher than 1.0 indicated a shorter time to improvement.
To describe the development of anemia among patients without anemia initiating cART, we used Kaplan-Meier estimates of anemia-free survival, log-rank tests for equality across strata of AZT exposure to assess differences in cumulative probability of developing anemia, and Cox proportional hazards regression to model the independent association of AZT with time to development of anemia in the full cohort, controlling for CD4 cell count, viral load, gender, and race/ethnicity. The results were expressed as HRs with 95% CIs. We also conducted subset analyses of time to anemia using ITT and AT methods, for the reasons described previously.
Among 2195 patients who initiated cART during the observation period, 491 were excluded from analysis because of insufficient Hb data and 84 were excluded for having received prior anemia treatment. Of the 1620 remaining patients included in the analysis, all had sufficient Hb measurements for analysis and were not treated for anemia with erythropoietin or transfusion before cART initiation. Of these persons, 543 (34%) had no anemia at cART initiation, 848 (52%) had mild anemia, 174 (11%) had moderate anemia, and 55 (3%) had severe anemia. The characteristics of the patients who were excluded because of insufficient follow-up Hb data and who were excluded because of treatment for anemia before cART initiation are shown in Table 1. There was differential exclusion with respect to age, race/ethnicity, and risk for HIV infection. For patients included in the analysis, the median length of follow-up was 21 months (range: 1 to 80 months), and the median number of Hb measurements during follow-up was 3 (range: 1 to 13 measurements).
Time to Improvement of Anemia
Overall, 581 (54%) of 1077 persons who initiated cART with anemia had improvement of anemia during follow-up. Among 848 persons initiating cART with mild anemia, 396 had improvement of anemia (288 [46%] of 631 initiating a non-AZT-containing regimen and 108 [50%] of 217 initiating an AZT-containing regimen). Among 174 persons initiating cART with moderate anemia, 137 had improvement of anemia (116 [79%] of 147 initiating a non-AZT-containing regimen and 21 [78%] of 27 initiating an AZT-containing regimen). Among 55 persons initiating cART with severe anemia, 48 had improvement of anemia (38 [86%] of 44 initiating a non-AZT-containing regimen and 10 [91%] of 11 initiating an AZT-containing regimen). Unadjusted Kaplan-Meier estimates of cumulative probability of improvement of anemia for persons with anemia at cART initiation indicated that inclusion of AZT in the initial regimen was associated with lower probability of improvement of mild anemia (Fig. 1A) but was not associated with differences in probability of improvement of moderate or severe anemia in univariate analyses (see Figs. 1B, C).
The unadjusted probability of improvement of anemia at 12 months was related to baseline anemia severity: persons with more severe anemia at baseline had higher probabilities of improvement. The 12-month probability of improvement of mild anemia among patients whose initial regimens did not include AZT was 34% (95% CI: 30 to 38), and for patients whose initial regimens included AZT, it was 27% (95% CI: 21 to 35); the log-rank test for equality across strata of AZT exposure indicated a significant difference (P = 0.0006). The 12-month probability of improvement of moderate anemia among patients whose initial regimens did not include AZT was 75% (95% CI: 68 to 83), and for patients whose initial regimens included AZT, it was 60% (95% CI: 32 to 80; P = 0.36 across AZT strata by log-rank test). The 12-month probability of improvement of severe anemia among patients whose initial regimens did not include AZT was 96% (95% CI: 85 to 100), and for patients whose initial regimens included AZT, it was 77% (95% CI: 49 to 96; P = 0.33 across AZT strata by log-rank test).
In the overall Cox proportional hazards regression models using data from all available patients as an ITT analysis, time to improvement of anemia was longer for patients whose initial regimen contained AZT and who were non-Hispanic blacks, whereas time to improvement was shorter for patients who had moderate or severe anemia at baseline or who had a CD4 count < 200 cells/μL (Table 2). The magnitude of the hazard of including AZT in the initial regimen was equivalent to the hazard for non-Hispanic black race/ethnicity. The interaction term for AZT*baseline anemia status was not significant when we calculated the hazard of AZT prescription for patients with baseline mild anemia compared with patients with baseline moderate (P = 0.6) or severe (P = 0.8) anemia. This finding suggests that, in contrast to the findings from the unadjusted Kaplan-Meier analyses, including AZT in the initial regimen was associated with longer time to improvement for patients with all severities of anemia at baseline. Results were similar in the ITT (n = 890) and the AT (n = 856) subset models (data not shown), except that in the AT analysis, the 95% CI was somewhat wider for the hazard of improvement for patients whose initial regimen contained AZT (HR = 0.7, 95% CI: 0.4 to 1.1).
Time to Development of Anemia
Among 543 persons initiating cART without anemia, 263 developed anemia during follow-up (181 [44%] of 408 initiating a non-AZT-containing regimen and 82 [61%] of 135 initiating an AZT-containing regimen). Kaplan-Meier estimates of anemia-free survival among patients without anemia at cART initiation indicated that inclusion of AZT in the initial regimen was not associated with the cumulative probability of developing anemia (see Fig. 1D). The 12-month probabilities of developing anemia were similar: 62% (95% CI: 57% to 67%) for patients whose initial regimens did not include AZT and 58% (95% CI: 49% to 67%; P = 0.40 across AZT strata by log-rank test) for patients whose initial regimens included AZT. In the Cox proportional hazards regression model of time to development of anemia, inclusion of AZT in the initial regimen was not associated with time to failure in the full cohort and subset in the ITT models; however, in the subset AT model, time to development of anemia was associated with inclusion of AZT in the initial regimen (HR = 1.8, 95% CI: 1.2 to 2.6; full model results not shown).
It is well established that treating HIV infection with antiretroviral drugs generally improves anemia, even when specific treatment for anemia is not provided.6 Our results confirm and extend this finding, but suggest that inclusion of AZT in the initial cART regimen may prolong the time to improvement of anemia by up to 30%. Our findings with respect to hazard of developing anemia were less conclusive: using univariate and multivariate ITT methods, inclusion of AZT in the initial antiretroviral regimen was not associated with an increased hazard of developing anemia, whereas using multivariate AT methods, we found that inclusion of AZT was associated with an increased hazard of developing anemia.
We found that improvement of anemia after cART initiation was more common for, and occurred more quickly for, patients with moderate or severe anemia or who were more immunosuppressed. In patients with HIV infection in the United States, most anemia is attributable to anemia of HIV infection (anemia of chronic disease).11 The anemia of HIV infection is mediated by elaboration of myelosuppressive cytokines and changes in the marrow microenvironment; cytokine production is associated with HIV replication.12 Our finding that improvement of anemia was more rapid for patients with low CD4 cell counts and high viral loads suggests that interruption of viral replication associated with cART initiation may lead to more rapid resolution of anemia.
We also found that the hazard of improvement of anemia at cART initiation was lower for non-Hispanic blacks than for other racial/ethnic groups and that the magnitude of this hazard was comparable to having AZT in the initially prescribed regimen. General population surveys and other samples have established lower average values for the measures of anemia among US blacks13-16 and among Africans.17,18 The differential impact of anemia among HIV-infected persons of black race has been noted in previous studies.1,2,19,20 It is likely that additional mechanisms other than HIV infection may contribute to the overall greater prevalence of anemia among HIV-infected blacks, including sickle cell disease,21 thalassemia, and iron deficiency22 and that anemia in HIV-infected blacks from these and other non-HIV-related causes may not be ameliorated by control of HIV replication, thus leading to a lower hazard of improvement of anemia among blacks in our cohort.
We observed an increased hazard of developing anemia for nonanemic patients initially prescribed an AZT-containing regimen, but only when using AT analytic methods. The reason for this finding may be that when patients without anemia initiate an AZT-containing regimen and experience early decreases in Hb, clinicians change the regimen to a non-AZT-containing regimen. In this case, the intention to treat with an AZT-containing regimen would not predict development of anemia because of misclassification of persons without long-term AZT exposure as AZT exposed; actual aggregate exposure to an AZT-containing regimen did predict development of anemia.
Our analysis has some important limitations. Although our study population was large and included persons receiving care in diverse settings in the United States, it was not representative of all persons in care for HIV infection. For analysis of improvement of anemia, we were concerned that there could be misclassification of AZT status if clinicians took patients off of AZT soon after early signs of anemia but found that the results of AT analysis were similar to the results of ITT analysis in most cases. We believe that our findings are important to consider in settings such as Africa, where fixed-dose combinations play an important role and where the background prevalence of anemia is high. In different geographic settings, however, there are numerous factors that may influence the underlying prevalence and causal mechanisms of anemia and might also have an impact on outcomes of anemia after cART initiation. Specifically, prevalence of comorbid conditions such as malaria and tuberculosis, differences in the prevalence of structural differences in Hb (eg, sickle trait, sickle disease), and differences in nutritional status of patients may be important contextual factors. Thus, we recognize that data from the United States have limited relevance to clinical recommendations in Africa and that data collection in Africa and other settings is important to the development of the evidence base for local treatment recommendations.23
The definition of “presumed antiretroviral naive” also may have led to some misclassification. The definition was constructed to maximize specificity; for example, even patients initiating cART according to ASD project data were not counted as “presumed naive” unless their regimen was a typical first-line regimen and their CD4 count and viral load were consistent with an indication for starting therapy. This may have excluded some patients who were truly initiating cART using an unusual first-line regimen or with an extremely favorable immunologic or virologic status. Further, although chart abstractors attempted to collect information on historical treatment at the time of first observation in the ASD project, some patients may have had undocumented experience with cART before coming to care in their ASD project facility and may have been incorrectly classified as being treatment naive when they were, in fact, experienced.
We also did not include data on patients who had a history of treatment for anemia, and data from patients who were given erythropoietin or transfusion after cART was initiated were censored at the time of treatment for anemia. Thus, our findings cannot be generalized to patients prescribed erythropoietin or transfusion before or together with antiretroviral therapy.
AZT was associated with longer time to improvement of anemia for all anemic patients, including those with mild anemia at baseline. It is likely that many patients with mild chronic anemia are not symptomatic for the condition. For those uncommon patients with symptomatic mild anemia or for patients with more severe anemia in whom it is clinically important to resolve anemia as quickly as possible, a non-AZT-containing initial antiretroviral regimen may be recommended. Implementation of effective cART (with or without AZT) would be expected to result in resolution of anemia of many causes, including anemia of HIV infection (anemia of chronic disease). There are other causes of anemia, such as nutritional deficiencies and genetically determined anemias, that would not be expected to be affected by initiation of cART, however. Therefore, appropriate diagnostic evaluation of anemia is important in all cases so that treatable causes can be identified and resolved. For patients with refractory anemia, treatment with recombinant human erythropoietin may be useful in improving Hb concentration, reducing transfusion requirements and improving quality of life, especially for patients prescribed AZT;24 however, a recent meta-analysis of treatment of anemia for persons with HIV infection did not find definitive evidence that treatment with erythropoietin improved survival.25
Based on our findings, we conclude that on a population basis, the use of AZT in an initial antiretroviral regimen is a medically rational choice, even for patients with preexisting anemia. For most patients with HIV infection and anemia, initiating cART with or without AZT results in improvement of anemia. Alternative regimens should be considered for patients in whom rapid resolution of anemia is clinically important or for whom clinically relevant anemia develops after initiation of an AZT-containing regimen.
1. Sullivan PS, Hanson DL, Chu SY, et al. Epidemiology of anemia
in human immunodeficiency virus (HIV
)-infected persons: results from the multistate Adult and Adolescent Spectrum of HIV
Disease surveillance project. Blood
2. Moore RD. Human immunodeficiency virus infection, anemia
, and survival. Clin Infect Dis
3. Sullivan P. Associations of anemia
, treatments for anemia
, and survival in patients with human immunodeficiency virus infection. J Infect Dis
. 2002;185(Suppl 2):S138-S142.
4. Mocroft A, Kirk O, Barton SE, et al. Anaemia is an independent predictive marker for clinical prognosis in HIV
-infected patients from across Europe. AIDS
5. Lundgren JD, Mocroft A. Anemia
and survival in human immunodeficiency virus. Clin Infect Dis
. 2003;37 (Suppl 4):S297-S303.
6. Moore RD, Forney D. Anemia
-infected patients receiving highly active antiretroviral therapy. J Acquir Immune Defic Syndr
7. Panel on Antiretroviral Guidelines for Adult and Adolescents. Guidelines for the use of antiretroviral agents in HIV
-infected adults and adolescents. Department of Health and Human Services. January 29, 2008. Available at: http://aidsinfo.nih.gov/contentfiles/AdultandAdolescentGL.pdf
. Accessed February 5, 2008.
9. Farizo KM, Buehler JW, Chamberland ME, et al. Spectrum of disease in persons with human immunodeficiency virus infection in the United States. JAMA
10. Kalbfleisch JD, Prentice RL. The Statistical Analysis of Failure Time Data
. New York: John Wiley & Sons; 1980.
11. Buskin SE, Sullivan PS. Anemia
and its treatment and outcomes in persons infected with human immunodeficiency virus. Transfusion
12. Koka PS, Reddy ST. Cytopenias in HIV
infection: mechanisms and alleviation of hematopoietic inhibition. Curr HIV Res
13. Jackson RT, Sauberlich HE, Skala JH, et al. Comparison of hemoglobin values in black and white male U.S. military personnel. J Nutr
14. Leshan L, Gottlieb M, Mark D. Anemia
is prevalent in an urban, African-American adolescent population. Arch Fam Med
15. Beutler E, West C. Hematologic differences between African-Americans and whites: the roles of iron deficiency and alpha-thalassemia on hemoglobin levels and mean corpuscular volume. Blood
16. Robins EB, Blum S. Hematologic reference values for African American children and adolescents. Am J Hematol
17. Ssali F, Stohr W, Munderi P, et al. Prevalence, incidence and predictors of severe anaemia with zidovudine
-containing regimens in African adults with HIV
infection within the DART trial. Antivir Ther
18. Lugada ES, Mermin J, Kaharuza F, et al. Population-based hematologic and immunologic reference values for a healthy Ugandan population. Clin Diagn Lab Immunol
19. Semba RD, Shah N, Klein RS, et al. Prevalence and cumulative incidence of and risk factors for anemia
in a multicenter cohort study of human immunodeficiency virus-infected and -uninfected women. Clin Infect Dis
20. Mildvan D. Prevalence of anemia
and correlation with biomarkers and specific antiretroviral regimens in 9690 human-immunodeficiency-virus-infected patients: findings of the Anemia
Prevalence Study. Curr Med Res Opin
21. Creary M, Williamson D, Kulkarni R. Sickle cell disease: current activities, public health implications, and future directions. J Womens Health (Larchmt)
22. McClung JP, Marchitelli LJ, Friedl KE, et al. Prevalence of iron deficiency and iron deficiency anemia
among three populations of female military personnel in the US Army. J Am Coll Nutr
23. Colebunders R, Kamya MR, Laurence J, et al. First-line antiretroviral therapy in Africa-how evidence-based are our recommendations? AIDS Rev
24. Henry DH, Beall GN, Benson CA, et al. Recombinant human erythropoietin in the treatment of anemia
associated with human immunodeficiency virus (HIV
) infection and zidovudine
therapy. Overview of four clinical trials. Ann Intern Med
25. Marti-Carvajal AJ, Sola I. Treatment for anemia
in people with AIDS. Cochrane Database Syst Rev
Adult/Adolescent Spectrum of HIV Disease Principal Investigators and Project Leads: Melanie Thompson and Ericka Sinclair (AIDS Research Consortium of Atlanta); David Cohn, Arthur Davidson, and Cornelius Rietmeijer (Denver Department of Health and Hospitals); Jane Turner and Amy Wohl (Los Angeles County Department of Health Services, Los Angeles); Anne Morse, Stephanie Broyles, and C. Lynn Besch (Louisiana Department of Health, Baton Rouge); Eve Mokotoff and Linda Wotring (Michigan Department of Community Health, Detroit); Judy Sackoff and Marie Antoinette Bernard (New York City Department of Health); Jose Otero, Robert Hunter, and Maria de los Angeles Gomez (University Central del Caribe, Bayamon); Sandra Miranda (Puerto Rico Department of Health, San Juan); Susan Buskin, Elizabeth Tesh Barash, and Sharon Hopkins (Public Health-Seattle and King County, Seattle); Sylvia Odem (Texas Department of Health, Austin); Philip Keiser (Parkland Hospital, Dallas); and Kaye Reynolds and Wes McNeely (Department of Health and Human Services, Houston).