Berhane, Kiros PhD*; Karim, Roksana MSc*; Cohen, Mardge H MD†; Masri-Lavine, Lena MSc*; Young, Mary MD‡; Anastos, Kathryn MD§; Augenbraun, Michael MD∥; Watts, D Heather MD¶; Levine, Alexandra M MD*
Anemia is common in HIV-infected individuals, occurring in approximately 30% of those with asymptomatic HIV infection and in 75% to 80% of those with clinical AIDS.1–3 Anemia has been associated with decreased quality of life in HIV-infected persons4 and has also been associated with decreased survival in several studies, most of which are composed primarily of men.3,5–9 These studies have also demonstrated that recovery from anemia may be associated with a prolongation in survival.3,5,6 Thus, in an observational study, Moore and colleagues5 showed that administration of erythropoietin led to an increase in hemoglobin (Hb) levels and was associated with a significantly decreased hazard of death in a group of approximately 500 HIV-infected persons with anemia.
The presence of anemia in HIV-infected women is known to correlate with worsening HIV disease parameters, including a CD4 count <200 cells/μL, higher HIV-1 RNA levels in plasma, and a history of a clinical AIDS-defining illness.10 Recently, the use of highly active antiretroviral therapy (HAART) has been associated with improvements in immune function and significant declines in HIV-1 RNA levels. HAART has been widely used in the United States since 1996 and has led to a major decrease in AIDS-defining conditions,11 with significant prolongation in survival.12,13 HAART therapy consists of antiretroviral drugs from several classes, often including a protease inhibitor (PI).14 Of interest, ritonavir, a PI, has recently been shown to stimulate the growth of hematopoietic progenitors in vitro and to inhibit apoptosis of these progenitors.15 It is possible that the use of HAART might be associated with protection against anemia or correction of preexisting anemia as a result of HIV disease. A recent study by Mocroft and colleagues7 was consistent with this hypothesis, indicating that HAART therapy for 6 months or more was associated with a decreased likelihood of anemia. In a recent cohort of 66 HIV-infected patients, including 1 woman, Huang and colleagues16 demonstrated that the use of HAART was associated with an increase in white blood cells and platelets. Although an initial decrease in red blood cells occurred, Hb levels subsequently increased after 3 months of HAART use.8,16
Anemia is more common in women than in men, and its significance in terms of prognosis in the era of HAART has not yet been fully ascertained. Therefore, in an attempt to evaluate the relationship between anemia and survival in HIV-infected women on HAART and/or other antiretroviral regimens, we studied 2056 HIV-infected women enrolled in the Women’s Interagency HIV Study (WIHS). These women were followed over a period of approximately 4 years, from October 1994 through September 1998, in an attempt to ascertain the relationship between anemia and survival within the first 2 years of HAART use.
The WIHS is a multicenter prospective study of HIV-1 infection and associated diseases in women in the United States. From October 1994 through November 1995, 2056 HIV-1–infected women were enrolled in the WIHS at 6 sites throughout the United States (Los Angeles, CA; Washington, DC; San Francisco, CA; New York City/Bronx, NY; Brooklyn, NY; and Chicago, IL). Women were recruited from HIV clinics, street outreach, referral from other studies, and word of mouth. Informed consent was obtained from all participants as approved by the site-specific institutional review boards. WIHS methods and cohort characteristics have been described previously.17 Study participants were seen every 6 months and underwent a standardized history, physical examination, gynecologic examination, and blood tests at each of these visits. Demographic and clinical characteristics of the cohort are provided in Table 1. At a median of 3.5 years of follow-up, the retention rate of HIV-positive women was 80.4%, excluding deaths. During the 3.5-year follow-up period, a maximum of 8 study visits was possible for each participant. In fact, the median number of visits per participant was 6 (range: 1–8 visits). The majority of participants (1657 of 2056 participants [approximately 80%], excluding those who died) had their last visit after the widespread introduction and use of HAART. The presence of other intercurrent medical illnesses, pregnancy, menstrual and gynecologic history, and use of AIDS- and non–AIDS-related medications was recorded at each study visit. The WIHS is an observational study, however, and participants were not uniformly evaluated in terms of the potential cause(s) or specific types of therapy for anemia.
Clinical Data and Laboratory Methods
Using a standardized instrument, trained interviewers collected information on demographic and clinical manifestations, medication use, and sexual and drug use behaviors at baseline and at study visits every 6 months. Questions regarding medication usage included specific queries regarding the antiretroviral regimen currently being followed as well as the medications taken over the preceding 6 months. Lymphocyte subsets were assayed within 24 hours of collection using standard flow cytometry techniques at local laboratories participating in the National Institutes of Health, Division of Acquired Immune Deficiency Syndrome Quality Assurance Program. Serologic reactivity to HIV-1 was determined by enzyme immunoassay (EIA) and confirmed by Western blot analysis using standard techniques. Plasma HIV RNA levels were measured using freshly frozen specimens separated within 6 hours of phlebotomy utilizing the isothermal NASBA method (Organon-Teknika, Durham, NC), with a minimum value of 4000 copies/mL. These assays were performed in laboratories participating in the National Institutes of Allergy and Infectious Diseases virology Quality Assurance Program.
Using standard techniques, data on Hb, hematocrit (Hct), and mean corpuscular volume (MCV) were derived in each local laboratory from samples collected at the initial and 6-month visits. Anemia was defined as an Hb level <12 g/dL.18 Normal MCV was defined as 80 to 100 fL.
Definition of HAART
All participants were queried about their current use of specific antiretroviral medications at the time of each visit (every 6 months) and about the medications taken during the preceding 6 months.
The definition of HAART was guided by the International AIDS Society–United States of America (IAS-USA) Panel12 and the Department of Health and Human Services (DHHS)/Kaiser Panel19 guidelines as follows: (1) 2 or more nucleoside reverse transcriptase inhibitors (NRTIs) with at least 1 PI or 1 nonnucleoside reverse transcriptase inhibitor (NNRTI) (91% of observations classified as HAART), (2) 1 NRTI with at least 1 PI and at least 1 NNRTI (6%), (3) ritonavir and saquinavir in combination with 1 NRTI and no NNRTIs (2%), and (4) an abacavir-containing regimen of 3 or more NRTIs in the absence of both PIs and NNRTIs (1%). Combinations of zidovudine (AZT) and stavudine (d4T) with either a PI or an NNRTI were not considered HAART.
The most frequent use of monotherapy included 1 NRTI (92%). Of the other monotherapy users, taking only PIs accounted for 7%, whereas taking only NNRTIs accounted for 1%.
All other regimens were classified as combination therapy. The 3 most frequent types of combination therapy were: (1) 2 NRTIs (37%); (2) 3 or more NRTIs without abacavir, PIs, or NNRTIs (13%); and (3) 1 NRTI and 1 PI (9%).
Descriptive statistical analyses were used to study the baseline characteristics of the HIV-positive participants of the WIHS. Initial anemia status was analyzed using χ2 analysis on proportions and mean comparisons.20 Longitudinal data from the HIV-positive participants in the WIHS (from the baseline visit in approximately October 1994 up to the end of visit 8 in September 1998) were then analyzed to evaluate: (1) the correlates of anemia; (2) the correlates of resolution of anemia; (3) the independent effect of anemia on overall survival; and (4) the role of HAART therapy on overall survival, development of anemia, and resolution of anemia.
The data were categorized into 6-month time intervals as follows: October 1994 to March 1995, April 1995 to September 1995, October 1995 to March 1996, April 1996 to September 1996, October 1996 to March 1997, April 1997 to September 1997, October 1997 to March 1998, and April 1998 to September 1998. Six-month time intervals were used, because each participant was seen at some point within this designated time frame. Medical and medication-related data included information pertinent at the specific date of the study visit as well as information covering the 6 months since the last study visit. At any time interval, anemia was defined as having an Hb level of <12 g/dL. Resolution of anemia was defined as an Hb level of ≥12 g/dL in a subject who was previously anemic.
The following parameters were evaluated in all models in terms of their effect on onset and/or resolution of anemia and overall survival: baseline age, ethnicity, MCV (low: <80 fL, normal/high: ≥80 fL), CD4 count (low: <200/dL, high: ≥200/dL), HIV RNA level (low: <50,000 copies/mL, high: ≥50,000 copies/mL), antiretroviral therapy in the past 6 months, AZT use in the past 6 months (yes/no), injection drug use at any time while on study (yes/no), and history of AIDS-defining condition (yes/no). Additional categories were initially evaluated in terms of CD4 count and MCV but were later collapsed into the current variable (high/low) because of clinical and statistical power considerations. In models for overall survival, the time-dependent status of anemia (yes/no) was also included as a time-dependent yes/no variable.
To examine the effects of antiretroviral therapy on resolution and/or development of anemia separately, logistic regression models were fitted on data that were split into 2 groups based on initial anemia status. We used modeling techniques that properly account for the possible recurrent (time-dependent) status of anemia in any given participant and the intrawoman correlation between the multiple measurements from each participant. Specifically, we used transition models21 that model anemia status at any given time period as an explicit function of the anemia status in prior visits, in addition to other covariates of interest. To allow for any residual intrawoman correlation beyond what is accounted for by the transitional modeling, our inference was based on “empiric” standard errors (SEs) from the generalized estimation equations (GEE) approach.21 Our results are thus based on first-order transition models (ie, those that include the anemia status at the prior visit as a covariate), because transition models of higher orders did not provide a significantly better fit of the data. All potential risk factors and confounders were treated as time-dependent covariates whenever appropriate so as to reflect longitudinal trends. The duration of HAART use variable, for example, was categorized into none, 1 interval, or 2 or more intervals, with subjects moving between categories according to their time-dependent level of HAART use at any given visit. Both univariate and multivariable models were constructed, and results were reported as odds ratios (ORs) along with 95% confidence intervals and associated P values.21,22 Missing data were assumed to be missing at random (ie, the pattern of missing data was not associated with the outcome), but sensitivity analysis was also conducted to examine the pattern of loss to follow-up by initial anemia status.
During the 3.5-year follow-up period, a maximum of 8 study visits was possible for each participant. In fact, the median number of visits per participant was 6. Because the majority of participants (1657 of 2056 participants [approximately 80%], excluding those who died) had their last visit after the widespread introduction and use of HAART, we were able to examine the association of HAART with anemia. Our ability to study the impact of long-term use of HAART was limited, however, because the number of women on HAART for at least 18 months was only 18.7% (385 of 2056 participants), limiting the ability to examine an effect of HAART in this group. Full details on retention of the WIHS cohort have been reported elsewhere.23
The effects of anemia, HAART, and other potential risk factors on survival were examined using the Cox proportional hazards model.24,25 Results were reported as hazard ratios (HRs) along with corresponding 95% confidence intervals and P values. Proportional hazards models were fitted on the whole data set to examine the effects of anemia and other risk factors adequately. To study the effect of anemia on overall survival in the post-HAART period, the proportional hazards models were also fitted by adjusting the follow-up time to start from the onset of HAART therapy. All analyses were conducted using the SAS statistical package (SAS Institute, Cary, NC) and EPILOG software (EPICENTER Software, Pasadena, CA).
Characteristics of Study Population
The demographic and clinical characteristics are given in Table 1. At baseline, the HIV-positive women had a median age of 36 years (range: 17–73 years), and 41% gave a history of injection drug use. African Americans comprised 56% of the group, whereas 23% were Latina and 18% were white. A more complete baseline description of the WIHS cohort has been reported elsewhere.17
During the follow-up period, the overall composition of the participants in the study remained unchanged with respect to ethnicity, history of injection drug use, and other demographic factors (data not shown). The number and percentage on AZT were unchanged from baseline.
Prevalence of Anemia With HAART Use
At baseline, 36.7% of the HIV-infected women were anemic, with Hb levels <12 g/dL. At the end of the 3.5-year follow-up period, 47% of the women had been anemic during at least 1 study visit.
Before the introduction of HAART (visits 3 and 4, 1996), no significant trend was apparent in terms of the prevalence of anemia at each study visit (data not shown). Thereafter, however, women who used HAART for 2 or more intervals (at least 12 months) had a trend toward a lower likelihood of being anemic compared with those who received no HAART or those who used HAART for only 1 follow-up interval (6 months) (P = 0.12).
Factors Associated With Development and/or Resolution of Anemia
Among the HIV-positive women who were anemic at baseline based on a transition model of order 1 (where anemia status at the immediate prior visit is included as a covariate), several factors were independently associated with inability to resolve anemia. These included low (<200 cells/μL) CD4 count (OR = 0.56 [0.44, 0.72]; P < 0.001), MCV values <80 fL (OR = 0.40 [0.28, 0.56]; P < 0.001), and high levels of HIV-1 RNA (≥50,000 copies/mL) in plasma (OR = 0.65 [0.50, 0.84]; P < 0.001) as shown on Table 2. Injection drug use (P = 0.18) or development of clinical AIDS (P = 0.22) was not significantly associated with the persistence of anemia. Not surprisingly, being anemic during the prior time period was significantly associated with an inability to resolve anemia (OR = 0.19 [0.15, 0.25]; P < 0.001).
Among the HIV-positive women who were free of anemia at baseline, several factors were independently associated with development of anemia over time (see Table 2), including low (<200 cells/μL) CD4 lymphocyte counts (OR = 1.98 [1.61, 2.42]; P < 0.001), low MCV levels <80 fL (OR = 4.19 [2.85, 6.15]; P < 0.001), and having clinically diagnosed AIDS (OR = 1.36 [1.13, 1.64]; P = 0.001). Being anemic during the prior time period was also significantly associated with still being anemic during the subsequent time period (OR = 6.99 [5.65, 8.64]; P < 0.001). A history of injection drug use at any time during follow-up was not associated with development of anemia (P = 0.38).
Association Between HAART Use and Anemia
In a first-order transition model that adjusted for all other risk factors among the women who were anemic at baseline, use of HAART was independently associated with a higher likelihood of resolution of anemia even when used for only 6 months (OR = 1.45 [1.07, 1.95]; P = 0.01; see Table 2). The likelihood of resolution of anemia was slightly higher for ≥12 months of HAART use (OR = 1.47 [1.14, 1.91]; P = 0.004). Of interest, the use of AZT was associated with an inability to correct anemia (OR = 0.80 [0.64, 0.99]; P = 0.04).
The association between HAART use and development of anemia was also assessed in the same multivariate model that adjusted for all other factors (see Table 2). Among the women who were free of anemia at baseline, use of HAART was associated with decreased odds of developing anemia among those who used HAART for at least 12 months, demonstrating a significant protective effect (OR = 0.71 [0.57, 0.88]; P = 0.002). There was also a similar protective but statistically nonsignificant effect of developing anemia associated with use of HAART for only 6 months (OR = 0.84 [0.65, 1.08]; P = 0.17). The number of women who were on HAART for at least 18 months was rather small at 18.7% (385 of 2056 participants), limiting the ability to examine an effect of HAART in this group. Of interest, the use of AZT either separately or as part of the antiretroviral regimen was significantly associated with increased odds of developing anemia (OR = 1.69 [1.42, 2.01]; P < 0.001).
Association Between Anemia and Survival
Among all HIV-positive women in the study (Table 3), multivariate analyses revealed that being anemic at any time during the study was independently associated with decreased survival (HR = 2.58 (1.94, 3.43]; P < 0.0001). Other factors independently associated with decreased survival included low CD4 counts of <200 cells/μL (HR = 5.83 (4.12, 8.25); P < 0.0001), high HIV-1 RNA level in plasma of >50,000 copies/mL (HR = 2.12 [1.59, 2.83]; P < 0.0001), and clinical diagnosis of AIDS (HR = 2.89 [2.16, 3.85]; P < 0.0001). The use of HAART for at least 12 months was associated with longer survival (HR = 0.76 [0.43, 1.81]), although this effect was not statistically significant (P = 0.32). In a model that examined the effect of longer use of HAART therapy, there was a 2-fold increase in survival associated with HAART use for at least 18 months. This association was not statistically significant (HR = 0.47 [0.16, 1.35]; P = 0.16), perhaps as a result of lack of power, because only 18% of the women were on HAART therapy for at least 18 months. Ethnicity, abnormal MCV values, use of AZT, and age were not significantly associated with survival.
To assess whether there was a delayed association between anemia and survival, Cox proportional hazards models were fitted by analyzing anemia status at the prior visit (ie, 6 months earlier) and anemia status 2 visits ago (ie, 1 year). The associations were still significant (P < 0.001), indicating the independent association between anemia status over time and overall survival.
In a model that restricted the Cox regression to those participants who initiated HAART at some time during follow-up and started the follow-up period in April 1996 when HAART was first introduced, the results were similar to those given in Table 3. Thus, the following factors were independently associated with decreased survival: development of anemia (HR = 2.22 [1.45, 3.39]; P < 0.001), low CD4 count of <200 cells/μL (HR = 4.50 [2.74, 7.38]; P < 0.001), high HIV-1 RNA level in plasma of >50,000 copies/mL (HR = 2.41 [1.56, 3.70]; P < 0.0001), and clinical diagnosis of AIDS (HR = 3.14 [1.92, 5.15]; P < 0.001). The use of HAART for 6 months was associated, albeit nonsignificantly, with longer survival (HR of death = 0.75 [0.43, 1.32]; P = 0.32), as was the use of HAART for at least 12 months (HR = 0.83 [0.46, 1.50]; P = 0.54).
To examine whether there was any differential in the lost-to-follow-up pattern by initial anemia status, we compared the number of visits and lost-to-follow-up patterns between women who were anemic and those who were not anemic at baseline. The median number of visits in both groups was 6. The mean number of visits was 5.72 (SE = 0.06) and 4.92 (SE = 0.09) for those women who were nonanemic and anemic at baseline, respectively. Thus, the nonanemic women at baseline had a higher number of subsequent visits compared with those who were anemic at baseline. Therefore, we also conducted a Kaplan-Meier analysis on loss to follow-up, specifically, treating death or end of the follow-up period as the censoring mechanism. The lost-to-follow-up pattern was different in the 2 groups, with more loss to follow-up in the group that was anemic at baseline (log-rank test for group difference was significant; P = 0.03). The higher loss to follow-up in the women who were anemic at baseline would be expected to bias the effect of anemia on overall survival toward the null.
The current report represents the largest prospective study yet conducted on the significance of anemia in HIV-infected women. Of importance, the development or persistence of anemia was shown to be independently and significantly associated with shorter survival. Other series have also noted a significant relationship between anemia and survival in HIV-infected individuals.3,5,6,9 In Moore et al’s observational study,5 whereas receipt of erythropoietin was associated with resolution of anemia and decreased risk of death, correction of anemia by transfusions was associated with shorter survival. The study by Moore et al5 was conducted between July 1989 and December 1996, before the widespread use of HAART. In the study by Sullivan et al3 of 32,867 HIV-infected persons conducted before the HAART era, resolution of anemia was also associated with a prolongation in survival. Finally, the study by Semba et al9 of HIV-infected women noted a relationship between anemia and decreased survival. The reason for the poor prognostic impact of anemia in HIV disease remains speculative. It is highly unlikely that anemia, per se, is the actual cause of death in these cases. Instead, anemia may serve as an indirect and surrogate marker for more advanced systemic illness. In this regard, anemia has also been associated with decreased survival times in other chronic illnesses, such as cancer.26–28 Nonetheless, although the relationship between anemia and decreased survival in HIV infection may be indirect, all studies, including our own, have been consistent in noting this association. Further work is required to elucidate the precise mechanisms for this prognostically important association.
Other factors associated with decreased survival in our cohort included characteristics of more advanced HIV disease, such as a low CD4 lymphocyte count (<200 cells/μL), higher levels of HIV-1 RNA, and/or a diagnosis of an AIDS-defining clinical illness. Use of HAART for 6 months or longer was associated with a trend toward longer survival as has been reported by others.12 Of importance, the women in our cohort who initiated HAART earlier were more likely to include those who had more advanced HIV disease and were therefore at risk for a less optimal outcome.29 A total of 501 of 1054 HIV-infected women who initiated HAART in the WIHS had a history of clinical AIDS before HAART initiation. Our finding of a trend toward increased survival after short-term use of HAART (6 or 12 months) would thus indicate an advantage of HAART therapy in this group.
Our study has shown that the use of HAART for as little as 6 months is statistically associated with resolution of anemia, whereas longer use is associated with more profound improvements. The association of HAART with amelioration of the anemia of HIV infection has been reported by others.7,16,30 In a study of almost 7000 HIV-infected patients from across Europe, Mocroft and colleagues7 found that more prolonged HAART use was associated with a greater likelihood of correcting anemia. Thus, 65.5% of the cohort was anemic before the use of HAART, 53.2% was anemic after 6 months of HAART, and 46.2% was anemic after 12 months of HAART. The impact of increasing Hb levels on overall survival times after institution of HAART was not evaluated.7
Our findings regarding the effects of HAART on resolution of anemia and HAART’s protective effect against development of anemia are unlikely to be caused by biases resulting from non-random missing data patterns.31 In a sensitivity analysis, we found that there was more loss to follow-up in the group that was anemic at baseline. From this observation, we conclude that our results remain valid, because the higher loss to follow-up occurred in the group that was more likely to stay anemic (baseline anemics). Thus, any bias that could result from this phenomenon on the effect of anemia on overall survival would be toward the null.
The mechanisms whereby HAART may be associated with correction of preexisting anemia are not yet fully understood. Most studies suggest that HIV-1 does not infect the hematopoietic progenitor.1,2,32 Nonetheless, abnormal growth of committed hematopoietic progenitor cells has been well described in HIV-infected persons.32 Further, HIV-1 infection of marrow stromal cells is sufficient to result in anemia and other cytopenias.33–35 A decrease in serum erythropoietin levels36; autoantibodies to erythropoietin37; or marrow suppression by opportunistic infections,38 tumors,39 or various medications40 may also contribute to the anemia commonly observed in HIV-infected persons. HAART may ameliorate many of these effects in an indirect manner simply by decreasing the HIV-1 viral burden.7,16,41 Additionally, however, Isgro et al42 have shown that HAART was associated with an increase in hematopoietic progenitor cell growth. Further, ritonavir, a PI, has been associated with decreased apoptosis of hematopoietic progenitors and direct stimulation of progenitor cell growth in vitro.5 These data would be consistent with our finding that HAART was associated with resolution of anemia.
Semba and colleagues9 have recently studied anemia in a cohort of HIV-infected and -uninfected women. Although the study design and patient population are similar to those of the WIHS, there are important differences in the modeling strategies, and thus in some of the conclusions drawn by each study. Semba et al9 examined risk factors for anemia in the pre-HAART and HAART eras, concluding that the risk of anemia was clearly decreased in the post-HAART era and that anemia was associated with a decrease in survival. By comparing the pre- and post-HAART eras alone, however, the effect of duration of HAART use on the development and/or resolution of anemia cannot be evaluated. Thus, we found that duration of HAART use was an important variable, with resolution of anemia seen after 6 months of HAART use and significant protection against development of anemia seen after 12 months of HAART use. Furthermore, we also examined those factors associated with the development or resolution of anemia separately. Additionally, in the Cox proportional hazards models, Semba et al9 used the baseline status of anemia as a risk factor instead of anemia status at any given visit. Because of the dynamic changes in Hb levels over time, we chose to consider anemia as a time-dependent covariate. Moreover, we chose to adjust for HAART use in the proportional hazards model. Neither the current study nor that of Semba and colleagues9 evaluated the specific causes of anemia in the 2 observational cohorts of women.
This study represents data from a large prospective cohort followed carefully over a 4-year period. Nonetheless, certain biases may have occurred. The HIV-positive women who received 3 or more intervals of HAART would have been those participants who had lived long enough to take the HAART for at least 18 months. On the other hand, HIV-positive women who began HAART early, thus having the opportunity for prolonged HAART use, would also be those participants with more advanced HIV disease at the time that HAART first became available. The first bias may have the potential of creating a tendency away from the null hypothesis, whereas the second has the potential of creating an inclination toward the null hypothesis. For this reason, the seemingly weak effect of HAART therapy on overall survival should be interpreted with caution. It is likely that the protective effects of HAART therapy may have been tempered by the fact that only those participants with more advanced HIV disease were given HAART at the time that it first became available. Further analyses will be conducted in the future to ascertain the role of long-term HAART use on anemia and survival times.
In conclusion, this study has confirmed the independent association between anemia and decreased survival among HIV-infected women. Further, we found that HAART therapy may be associated with correction of preexisting anemia after use for 6 months or more and with protection against development of anemia when used for 12 months or more. Further follow-up is required to determine the long-term effects of HAART therapy on anemia and survival.
Data in this manuscript were collected by the WIHS Collaborative Study Group with centers (Principal Investigators) at the New York/Bronx Consortium (Kathy Anastos); Brooklyn, New York (Howard Minkoff); the Washington, DC Metropolitan Consortium (Mary Young); the Connie Wofsy Study Consortium (Ruth Greenblatt); the Los Angeles County/Southern California Consortium (Alexandra Levine); the Chicago Consortium (Mardge Cohen); and the Data Coordinating Center (Alvaro Munoz). The authors thank the editors and reviewers for their insightful comments. They also acknowledge helpful discussions with Wendy Mack and help in additional analyses from Jay Gravink.
1. Zon LI, Arkin C, Groopman JE. Hematologic manifestations of the human immunodeficiency virus (HIV). Semin Hematol
2. Levine AM. Anemia, neutropenia and thrombocytopenia: pathogenesis and evolving treatment options in HIV infected patients. In: King E, ed. HIV Clinical Management
. vol. 10. New York, NY: Medscape/WebMD; 2002. Available at: www.medscape.com.
3. Sullivan PS, Hanson D, Chu SY, et al. Epidemiology of anemia in human immunodeficiency virus infected persons: results from the Multistate Adult and Adolescent Spectrum of HIV Disease Surveillance Project. Blood
4. Abrams DI, Steinhart C, Frascino R. Epoetin alfa therapy for anaemia in HIV infected patients: impact on quality of life. Int J STD AIDS
5. Moore RD, Keruly JC, Chaisson RE. Anemia and survival in HIV infection. J Acquir Immune Defic Syndr
6. Moore RD. Human immunodeficiency virus infection, anemia, and survival. Clin Infect Dis
7. Mocroft A, Kirk O, Barton SE, et al. Anemia is an independent predictive marker for clinical prognosis in HIV infected patients from across Europe. AIDS
8. Lundgren JD, Mocroft A, Gatell JM, et al. A clinically prognostic scoring system for patients receiving highly active antiretroviral therapy: results from the EuroSIDA study. J Infect Dis
9. Semba RD, Shah N, Klein S, 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
10. Levine AM, Berhane K, Masri-Lavine L, et al. Prevalence and correlates of anemia in a large cohort of HIV infected women. Women’s Interagency HIV Study. J Acquir Immune Defic Syndr
11. Ledergerber B, Egger M, Erard V, et al. AIDS related opportunistic illnesses occurring after initiation of potent antiretroviral therapy: the Swiss HIV Cohort Study. JAMA
12. Lee LM, Karon JM, Selik R, et al. Survival after AIDS diagnosis in adolescents and adults during the treatment era, United States, 1984–1997. JAMA
13. Gange SJ, Barron Y, Greenblatt RM, et al. The Women’s Interagency HIV Study Collaborative Study Group: effectiveness of highly active antiretroviral therapy among HIV-1 infected women. J Epidemiol Community Health
14. Carpenter CCJ, Cooper DA, Fischel MA, et al. Antiretroviral therapy in adults: updated recommendations of the International AIDS Society-USA Panel. JAMA
15. Sloand EM, Maciejewski J, Kumar P, et al. Protease inhibitors stimulate hematopoiesis and decrease apoptosis and ICE expression in CD34+ cells. Blood
16. Huang SS, Barbour JD, Deeks SG, et al. Reversal of human immunodeficiency virus type 1-associated hematosuppression by effective antiretroviral therapy. Clin Infect Dis
17. Barkan SE, Melnick S, Preston-Martin S, et al. The Women’s Interagency HIV Study. Epidemiology
18. Petersdorf RG, Adams RD, Braunwald E, et al, eds. Harrison’s laboratory values of clinical importance. In: Principles of Internal Medicine
. 10th ed. New York: McGraw-Hill; 1983:A-3.
19. DHHS/Henry J. Kaiser Family Foundation Panel on Clinical Practices for the Treatment of HIV Infection. Guidelines for the use of antiretroviral agents in HIV infected adults and adolescents, January 2000 revision. Available at: http://hivatis.org.
Accessed October 11, 2000.
20. Kleinbaum DG, Kupper LL, Muller KE, et al. Applied Regression Analysis and Other Multivariable Methods
. 3rd ed. Pacific Grove, CA: Duxbury Press; 1998.
21. Diggle PJ, Liang K-Y, Zeger SL. Analysis of Longitudinal Data
. New York: Oxford University Press; 1994.
22. Agresti A. Categorical Data Analysis
. New York: John Wiley; 1990.
23. Hessol NA, Schneider M, Greenblatt RM, et al. Retention of women enrolled in a prospective study of human immunodeficiency virus infection: impact of race, unstable housing, and use of human immunodeficiency virus therapy. Am J Epidemiol
24. Cox DR, Oakes D. Analysis of Survival Data
. London: Chapman and Hall; 1984.
25. Cox DR. Regression models and life tables (with discussion). J R Stat Soc
. 2001;34:187–220. [Ser B]
26. Littlewood TJ, Bajetta E, Nortier JWR, et al. Effects of epoetin alfa on hematologic parameters and quality of life in cancer patients receiving nonplatinum chemotherapy: results of a randomized, double-blind, placebo-controlled trial. J Clin Oncol
27. Albain KS, Crowley JJ, LeBlanc M, et al. Survival determinants in extensive-stage non-small-cell lung cancer: The Southwest Oncology Group experience. J Clin Oncol
28. Ohlhauser C, Bulzebruck H, Ebert W, et al. Prognostic factors for survival in inoperable non-small-cell lung cancer: a multivariate regression analysis of 456 patients with radiation therapy. Oncology
29. Anastos K, Barron Y, Miotti P, et al. Risk of progression to AIDS and death in women infected with HIV-1 initiating highly active antiretroviral treatment at different stages of disease. Arch Intern Med
30. Moore RD, Forney D. Anemia in HIV infected patients receiving highly active antiretroviral therapy. J Acquir Immune Defic Syndr
31. Ahdieh L, Gange SJ, Greenblatt R, et al. Selection by indication of potent antiretroviral therapy use in a large cohort of women infected with human immunodeficiency virus. Am J Epidemiol
32. Koka PS, Jamieson BD, Brooks DG, et al. HIV-1 induced hematopoietic inhibition is independent of productive infection of progenitor cells in vivo. J Virol
33. Moses AU, Williams S, Henevild ML, et al. Human immunodeficiency virus infection of bone marrow endothelium reduces induction of stromal hematopoietic growth factors. Blood
34. Bahner I, Kerans K, Coutinho S, et al. 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
35. Scadden DT, Zeira M, Woon A, et al. Human immunodeficiency virus infection of human bone marrow stromal fibroblasts. Blood
36. Spivak JL, Barnes DC, Fuchs E, et al. Serum immunoreactive erythropoietin in HIV infected patients. JAMA
37. Sipsas NV, Kokori SI, Ionnidis JPA, et al. Circulating autoantibodies to erythropoietin are associated with human immunodeficiency virus type 1 related anemia. J Infect Dis
38. Frickhofen N, Abkowitz JL, Safford M, et al. Persistent B19 parvovirus infection in patients infected with human immunodeficiency virus type 1: a treatable cause of anemia in AIDS. Ann Intern Med
39. Seneviratne LC, Tulpule A, Espina BM, et al. Clinical, immunologic, and pathologic correlates of bone marrow involvement in 291 patients with AIDS related lymphoma. Blood
40. Walker RE, Parker RI, Kovacs JA, et al. Anemia and erythropoiesis in patients with the acquired immunodeficiency syndrome (AIDS) and Kaposi’s sarcoma treated with zidovudine. Ann Intern Med
41. Semba RD, Shah N, Vlahov D. Improvement of anemia among HIV infected injection drug users receiving highly active antiretroviral therapy. J Acquir Immune Defic Syndr
42. Isgro A, Mezzaroma I, Aiuti A, et al. Recovery of hematopoietic activity in bone marrow from human immunodeficiency virus type 1 infected patients during highly active anti-retroviral therapy. AIDS Res Hum Retroviruses
© 2004 Lippincott Williams & Wilkins, Inc.