Neutropenia in Human Immunodeficiency Virus Infection: Data From the Women's Interagency HIV Study

Background:  Neutropenia is well described in individuals infected with human immunodeficiency virus (HIV) and occurs in approximately 10% to 50% of cases. Neither the effect of highly active antiretroviral therapy (HAART) on neutrophil counts nor the significance of neutropenia in terms of survival has previously been evaluated.

Methods:  The prevalence of neutropenia among 1729 HIV-infected women, followed up as part of the Women's Interagency HIV Study, was evaluated. The CD4 lymphocyte counts, HIV-1 RNA levels, and complete blood cell counts, including absolute neutrophil counts, were obtained at 6-month intervals.

Results:  Neutropenia was common among HIV-infected women; at baseline, 44% had neutrophil counts less than 2000/μL, whereas 7% had counts less than 1000/μL. During 7.5 years of follow-up, neutrophil counts less than 2000/μL occurred on at least 1 occasion in 79%, whereas absolute neutrophil counts less than 1000/μL were documented in 31%. Worsening HIV disease parameters, such as lower CD4 cell counts (P<.001) and higher HIV-1 RNA levels (P<.001), were associated with development of neutropenia. Resolution of neutropenia was associated with higher CD4 cell counts (P<.001) and use of HAART (P=.007). We found that HAART, without zidovudine, was associated with protection against development of neutropenia. On multivariate analysis, neutropenia was not found to be associated with decreased survival among HIV-infected women.

Conclusions:  Worsening HIV disease parameters are associated with neutropenia in HIV-infected women. Treatment with HAART, without zidovudine in the regimen, protects against development of neutropenia, whereas HAART use and higher CD4 cell counts are associated with resolution of neutropenia. Neutropenia is not associated with decreased survival in HIV-infected women.

Infection with human immunodeficiency virus (HIV) is associated with profound effects on hematopoiesis, with anemia, thrombocytopenia, and neutropenia all well described in infected individuals.^1,2 Although neutropenia has been reported in approximately 10% to 50% of these patients, relatively little has been studied in terms of the correlates of neutropenia, its significance on overall survival, or the effects of highly active antiretroviral therapy (HAART) in either preventing or correcting the neutropenia associated with HIV.

The Women's Interagency HIV Study (WIHS) is an ongoing evaluation of the course of HIV infection among women. Study participants are evaluated every 6 months, with information obtained by structured interview regarding use of HIV-related medications, intercurrent illness, and other issues, and a comprehensive laboratory evaluation is also performed. Data regarding white blood cell and neutrophil counts and HIV-related laboratory evaluations have been collected.

We analyzed the neutrophil counts in 1729 HIV-infected women, collected every 6 months during a follow-up period of 7.5 years, to ascertain the relationship between neutropenia and the extent and severity of HIV infection, the role of HAART on development or resolution of neutropenia, and the significance of neutropenia in terms of the overall survival of these individuals.

Details of the WIHS have been described previously.³ In short, the WIHS is a multicenter, prospective study designed to examine the characteristics and course of HIV infection in US females. A total of 2059 HIV-infected and 569 matched controls were enrolled in the study between October 1994 and November 1995 at 6 different sites in the United States (Los Angeles, Calif; northern California; Washington, DC; Bronx and Manhattan, NY; Brooklyn, NY; and Chicago, Ill). Informed written consent was obtained from the study participants, and the study was approved by the institutional review boards at each of the 6 sites.

Study participants were interviewed and received a thorough physical examination at baseline and every 6 months during follow-up. A standardized instrument was used to collect information on demographics, clinical characteristics, antiretroviral therapy (ART), other medication use, and lifestyle behaviors, including smoking, alcohol use, drug use, and sexual practices. Laboratory specimens were collected at each visit. Flow cytometry was performed to assay lymphocyte subsets within 24 hours of specimen collection in laboratories participating in the National Institutes of Health quality assurance program. Enzyme immunoassay was used to detect serologic reactivity to HIV-1; reactivity was confirmed by Western blot analysis. Plasma HIV-RNA viral load was measured from freshly frozen plasma samples separated within 6 hours of phlebotomy using the isothermic RNA sequence-based amplification method (Organan-Teknika, Durham, NC) with a lower detection limit of 80 copies/mL. Viral load detection assays were performed in laboratories participating in the National Institutes of Allergy and Infectious Diseases virology quality assurance program.

Absolute neutrophil counts (ANCs) were obtained by multiplying the total white blood cell count by the percentage of neutrophils obtained from the differential counts of white blood cells. For this analysis, neutropenia was defined as a peripheral blood neutrophil count less than 2000/μL.

At each visit, participants were asked about their current ART use and ART use in the past 6 months. The definition of HAART regimens was guided by the 2000 versions of the US Department of Health and Human Services⁴ and the International AIDS Society–USA Panel antiretroviral guidelines.⁵ Women were considered to be receiving HAART if they were taking (1) 2 or more nucleoside reverse-transcriptase inhibitors (NRTIs) with at least 1 protease inhibitor (PI) or nonnucleoside reverse-transcriptase inhibitor (NNRTI); (2) 1 NRTI in combination with at least 1 PI and at least 1 NNRTI; (3) a regimen containing ritonavir and saquinavir in combination with 1 NRTI and no NNRTIs; or (4) an abacavir sulfate–containing regimen of 3 or more NRTIs in the absence of both PIs and NNRTIs. Monotherapy included either 1 NRTI (95%) or only PIs (4%) or only NNRTIs (1%). All other regimens were defined as combination therapy.

This analysis included 7.5-year follow-up data. The maximum possible number of visits was 16, which occurred between July 1, 2002, and September 30, 2002.

The baseline characteristics were summarized using descriptive statistics. Longitudinal data from 1729 HIV-positive women in whom a neutrophil count was available and who had at least 2 follow-up visits were analyzed to evaluate (1) the predictors of development of neutropenia, (2) the predictors of resolution of neutropenia, (3) the impact of neutropenia on overall survival, and (4) the role of HAART in the development and resolution of neutropenia.

For longitudinal analysis, 7.5-year follow-up data were categorized using 6-month intervals, corresponding to the WIHS visits. To account for the correlated observations arising from women who contributed multiple observations (longitudinal study visits) to the analysis, a generalized estimating equation approach was used to determine the correlates of development and resolution of neutropenia using these 6-month periods. For each outcome (development or resolution), we used a logistic regression model. Women contributed data to each outcome, depending on their neutropenia status at their prior visit. Therefore, if a woman had neutropenia at her previous visit, data for the current visit contributed to the resolution model; if a woman did not have neutropenia at her previous visit, data for the current visit contributed to the development model. We used third-order transition models⁶ that considered neutropenia status at any given period as an explicit function of the neutropenia status 2 and 3 visits ago. (Neutropenia status 2 and 3 visits ago was included in the model in addition to other covariates of interest.) Thus, study visits were included in the analysis only if the data were available for that visit and the prior 3 visits. We did not include the neutropenia status at the immediately prior visit as a covariate, since, for example, all women analyzed for development of neutropenia at a particular visit would not have had neutropenia at the immediately prior visit to be included in the development model. We also evaluated higher-order transition models; these did not provide any significantly better fit of the data. Of the 1729 HIV-positive women with a known neutrophil count and at least 2 follow-up visits, 1183 were included in the analysis of development of neutropenia, and 882 were included in the analysis of resolution of neutropenia. These women contributed 8896 visits (62% of total visits) to the development model and 5386 visits (38%) to the resolution model.

The list of risk factors included in both sets of analyses was as follows: current smoking, injection drug use, current pregnancy, thrombocytopenia (platelet count <150 000/μL at the current visit); anemia (hemoglobin level <12 g/dL at the current visit); CD4 lymphocyte count (<200, 200-349, 350-499, or ≥500/μL), HIV-1 viral load (<5000, 5000-9999, 10 000-49 999, 50 000-99 999, or ≥100 000 copies/mL), ART use (no use, monotherapy with or without zidovudine, combination therapy with or without zidovudine, or HAART with or without zidovudine), total duration of HAART use (none, 1 six-month interval, 2 six-month intervals, and ≥3 six-month intervals), any use of zidovudine, and AIDS-defining illness. All potential risk factors and confounders were modeled as time-dependent variables based on the woman's status at her current visit or ART use since her previous visit.

We used a Cox proportional hazards model to evaluate the effect of neutropenia on survival. For the survival analysis,⁷ neutropenia was further categorized as a neutrophil count greater than 2000, 1000 to 2000, and less than 1000/μL. Deaths and follow-up that occurred within 6 months of a laboratory neutrophil count were included in the survival analysis. Deaths that occurred more than 6 months beyond a laboratory neutrophil count were censored at the death date. The multivariate Cox proportional hazards model for neutropenia included age, race, CD4 cell count, and HIV-RNA viral load as covariates. Except for race, all covariates were time dependent. All 1729 HIV-positive women with a known neutrophil count and at least 2 follow-up visits were included in the survival analysis. All analyses used SAS statistical software, version 8.01 (SAS Institute Inc, Cary, NC).

A total of 1729 HIV-positive women had neutrophil counts available for the analysis. At baseline, the median age of the participants was 37 yearas (age range, 17-73 years; Table 1). Most women (56%) were African American,had completed high school (63%), and were current smokers (55%). The median neutrophil count was 2193/μL (range, 10-14 014/μL). Of these 1729 women, 754 (44%) were neutropenic at baseline, when neutropenia was defined as a neutrophil count less than 2000/μL, whereas 120 women (7%) had neutrophil counts less than 1000/μL at baseline. During 7.5 years of follow-up, and occurring on at least 1 occasion, a neutrophil count less than 2000/μL was documented in 79%, whereas neutrophil counts less than 1000/μL occurred in 31%.

Among women who were free of neutropenia (neutrophil count >2000/μL) at their previous visit, transition models revealed that low CD4 cell count and high HIV-1 RNA viral load were significant independent risk factors for development of neutropenia during the 7.5-year follow-up period (Table 2). Compared with women with a CD4 cell count of 500/μL or higher, women with a CD4 cell count less than 200/μL were approximately 2 times more likely to develop neutropenia (odds ratio [OR], 1.82; 95% confidence interval [CI], 1.32-2.53). Relative to women with a viral load of less than 5000 copies/mL, women having 100 000 copies/mL or higher of HIV-1 RNA virus were 1.47 times more likely to develop neutropenia (OR, 1.47; 95% CI, 1.10-1.97). There was an increasing trend in the risk of development of neutropenia with decreasing CD4 cell count and increasing viral load (both P values for trend <.001).

Both anemia (OR, 1.45; 95% CI, 1.24-1.70) and thrombocytopenia (OR, 1.60; 95% CI, 1.27-2.02) were associated with a higher risk of developing neutropenia. The HIV-positive women using HAART without zidovudine were approximately 20% less likely to develop neutropenia (OR, 0.83; 95% CI, 0.67-1.01), whereas HAART regimens that contained zidovudine did not confer protection. In fact, the effect of HAART on protecting against development of neutropenia varied in a statistically significant manner, depending on whether zidovudine was included in the HAART regimen (P<.001) (Table 2). Use of combination ART without zidovudine was significantly protective against development of neutropenia (OR, 0.68; 95% CI, 0.51-0.92) (Table 2). The HIV-positive women who were pregnant at the study visit were also less likely to develop neutropenia (OR, 0.32; 95% CI, 0.14-0.72).

Women who were neutropenic at 2 visits before a given visit were approximately 3 times more likely (OR, 2.69; 95% CI, 2.26-3.21; P<.001) and at 3 visits before a given visit were 2.5 times more likely (OR, 2.55; 95% CI, 2.15-3.02; P<.001) to develop neutropenia again.

Using a conservative cut point for defining neutropenia (neutrophil count <1000/μL), low CD4 cell count, high HIV-1 RNA viral load, and the presence of anemia and/or thrombocytopenia maintained significance as a risk factor for the development of neutropenia. However, none of the ART-related variables were associated with the development of neutropenia (neutrophil count <1000/μL) (data not shown).

Among women with a neutrophil count less than 2000/μL at their previous visit, current pregnancy, CD4 cell count, combination ART without zidovudine, and HAART with or without zidovudine were significantly associated with resolution of neutropenia (Table 2). Women who were currently pregnant at any given visit were 4.8 times more likely to correct neutropenia compared with those who were not pregnant (OR, 4.77; 95% CI, 2.18-10.40). Resolution of neutropenia was significantly less likely to occur if anemia or thrombocytopenia was also present.

Compared with women with a CD4 cell count of 500/μL or higher, women with a CD4 count less than 200/μL were 32% less likely to have resolution of neutropenia (P for trend <.001). There was no clear trend in the ORs for viral load categories of 5000 or more copies/mL. Compared with those without any ART, HIV-positive women receiving HAART with or without zidovudine and those receiving combination ART without zidovudine were significantly more likely to resolve neutropenia.

Women who were neutropenic 2 visits before a given visit were 49% less likely (OR, 0.51; 95% CI, 0.43-0.60; P<.001) and 3 visits before a given visit were 51% less likely (OR, 0.49; 95% CI, 0.41-0.58; P<.001) to experience resolution of neutropenia.

On multivariate Cox proportional hazards analysis, neutropenia was not statistically significantly associated with survival in HIV-positive women (Table 3). As expected, low CD4 lymphocyte count and high HIV-RNA viral load were significant predictors of survival in these women. Compared with women with CD4 lymphocyte counts greater than or equal to 500/μL, women who had CD4 cell counts between 200 and 349/μL were 2 times and those who had CD4 cell counts less than 200/μL were 8 times more likely to die (P = .04 and P<.001, respectively). Women with HIV viral loads between 50 000 and 99 999 copies/mL had 2.5 times greater risk of death, and women with 100 000 copies/mL or greater had a 3 times greater risk of dying, than women who had HIV-RNA viral loads of less than 5000 copies/mL (P<.001). On multivariate analysis, including age, race, CD4 cell counts, and HIV-RNA level, neither total number of visits with neutropenia nor recurrence of neutropenia in 2 or more successive visits (ie, persistent neutropenia) was associated with survival (data not shown).

This longitudinal study of a large number of HIV-infected women has demonstrated that neutropenia is clearly associated with more advanced characteristics of HIV disease, including low CD4 cell counts and high HIV-RNA levels in plasma. Moreover, the study is the first to demonstrate that HAART may prevent the development of neutropenia in women with normal neutrophil counts. Notably, we have also shown that HAART is statistically associated with resolution of neutropenia in HIV-infected women with neutrophil counts less than 2000/μL. Of interest, although ART was associated with prevention of neutropenia, inclusion of zidovudine in the antiretroviral regimen seemed to ameliorate this ability. These findings would be fully consistent with the myelosuppression known to occur with zidovudine.^8,9 The use of myelosuppressive medications is the most common cause of cytopenias, including neutropenia,^10,11 in the setting of HIV infection, and prior studies have also shown that HIV-infected patients with neutropenia are significantly more likely to be receiving zidovudine or trimethoprim-sulfamethoxazole than nonneutropenic patients.¹² In a study¹¹ of 87 consecutive HIV-infected patients with ANCs less than 2000/μL, only 3 individuals were not receiving drugs associated with neutropenia, and 66% were receiving 3 or more potentially myelosuppressive medications. The direct toxic effect of zidovudine on bone marrow function may dampen the ability of HAART to overcome HIV-induced neutropenia through its effects in reducing the viral load. Because of these data, it would seem prudent to avoid zidovudine-containing HAART regimens in patients with severe anemia (hemoglobin level <10 g/dL).

Other factors may also contribute to the neutropenia seen in the setting of HIV. Progenitor cell growth is clearly abnormal in HIV-infected individuals, affecting red cells and platelets in addition to neutrophils, with significant reduction in the number of granulocyte-macrophage colony-forming units demonstrated in vitro.^13-16 This finding would be consistent with the fact that both anemia and thrombocytopenia were risk factors for neutropenia among our study participants. The presence of soluble inhibitory substances, produced by HIV-infected cells, has been shown to suppress neutrophil production in vitro.¹⁶ Decreased serum levels of granulocyte colony-stimulating factor have been described in HIV-seropositive patients with neutropenia, indicating that a relative deficiency of this specific hematopoietic growth factor may contribute to persistent neutropenia in these individuals.^17,18 Autoimmune phenomena have also been described.¹⁹ Additionally, bone marrow infiltration by infectious organisms or by malignant cells may be operative in the development of neutropenia in HIV-infected individuals.

As discussed, many of the possible causes of neutropenia in the setting of HIV involve either direct or indirect effects of the virus itself against hematopoietic progenitors or growth factors. The ability of HAART to prevent the neutropenia of HIV infection or to correct this cytopenia, as demonstrated in the present study, would thus appear biologically plausible and consistent with the finding that HAART may normalize the function of hematopoietic progenitors.²⁰ Huang et al²¹ have shown a statistically significant increase in peripheral blood granulocytes after the use of effective ART in a group of 66 HIV-infected patients, 1 of whom was female. However, in contrast to our own results, Huang and colleagues found no relationship between neutropenia and CD4 cell count or HIV-1 viral load. This discrepancy may be related to the fact that the study by Huang and colleagues involved only 66 individuals followed up for 18 months, whereas the current study involved approximately 1700 individuals followed up for as long as 7.5 years.

In our study, neutropenia (defined as an ANC of either <1000/μL or <2000/μL) was not an independent risk factor for survival, a finding not previously reported. Severe neutropenia per se is clearly a risk factor for bacterial infection, with ANCs less than 1500/μL statistically associated with an increased risk of serious bacterial infection among patients with acute leukemia, with increasing risk as the neutrophil count falls below this threshold.²² Other factors associated with infection include the duration of neutropenia²² and, in the setting of HIV infection, the level of CD4 cells.¹¹ Of interest, although neutropenia is relatively common among HIV-infected persons and despite the fact that neutrophil function is impaired in the setting of HIV,^23,24 serious bacterial infections is relatively uncommon in this setting. Moore et al¹² noted that 7.2% of 1638 HIV-infected individuals had neutrophil counts less than 1000/μL; there was no statistically increased risk of any individual bacterial infection among these individuals. However, when all infections were combined, the adjusted relative risk of bacterial infection was approximately 2 times higher than that seen in the nonneutropenic patients (P = .05).¹² In a group of 62 HIV-infected patients with ANCs less than 1000/μL, studied by Meynard et al,¹⁰ 24% developed infectious complications. Factors associated with infection included diagnosis of lymphoma, use of chemotherapy, neutropenia in the previous 3 months, use of central venous catheters, and a lower ANC. Thus, although neutropenia is clearly independently associated with development of infection,¹² the risk of such infection is still relatively low and less than that expected in other patient populations.²² The reasons for this discrepancy are unknown but may relate to the relatively short duration of neutropenia among HIV-infected patients, lasting a median of 13 days in the study by Moore et al,¹¹ or to the relatively moderate degree of neutropenia seen in this setting.^10,11 Nonetheless, despite low neutrophil counts, most HIV-infected patients do not develop serious bacterial infections, consistent with the fact that neutropenia was not associated with decreased survival in our study. These data would suggest that granulocyte colony-stimulating factors be used conservatively in patients with mild or moderate HIV-related neutropenia.

Our study was not able to elucidate the rate of bacterial infections that occurred during the 7.5 years of follow-up in study participants or the specific types of infections that occurred. Furthermore, we do not have full information regarding the complete medication history of participants. Nonetheless, this large, prospective study was powered to enable us to elucidate those factors associated with neutropenia among HIV-infected women and to define the factors associated with prevention of neutropenia or resolution of preexisting neutropenia.

In summary, neutropenia is common in HIV-infected women, occurring in 44% at baseline and in 79% during 7.5 years of follow-up. Neutropenia is associated with worsening HIV disease parameters and may be prevented or reversed by use of effective ART, which does not include zidovudine. Although bacterial infections may be increased among HIV-infected patients with neutropenia, this study has shown that overall survival is not adversely affected. Future study of the mechanisms that spare most of these individuals from developing severe bacterial infections would be of great interest.

Correspondence: Alexandra M. Levine, MD, USC/Norris Cancer Center, 1441 Eastlake Ave, Los Angeles, CA, 90033 (alevine@usc.edu).

Accepted for Publication: September 27, 2005.

Author Contributions: Dr Levine had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analyses.

Financial Disclosure: None.

Funding/Support: This work was supported in part by the Women's Interagency HIV Study, funded by the National Institute of Allergy and Infectious Diseases (Bethesda, Md), with additional supplemental funding from the National Institute of Child Health and Human Development (Rockville, Md), the National Institute on Drug Abuse (Bethesda), the National Institute of Dental and Craniofacial Research (Bethesda), the Agency for Health Care Policy and Research (Rockville), and by grants U01-AI-35004, U01-AI-31834, U01-AI-34994, U01-AI-34989, U01-HD-32632, U01-AI-42590, and N01-AI-35161 from the National Cancer Institute (Bethesda).

Role of the Sponsor: The various institutes within the National Institutes of Health that sponsored this study were involved, as noted in the authorship, in the design and conduct of the study; collection, management, analysis, and interpretation of the data; and preparation, review and approval of the manuscript, as per the policies of the Women's Interagency HIV Study.

Previous Presentation: A portion of this article was presented at the 14th International AIDS Conference; July 9, 2002; Barcelona, Spain.