Impaired Hematological Status Increases the Risk of Mortality among HIV-Infected Adults Initiating Antiretroviral Therapy in Tanzania

ABSTRACT

Background

Hematological status may predict HIV disease progression and mortality among adults initiating highly active antiretroviral therapy (HAART).

Objectives

We aimed to examine the relation of anemia and iron status at HAART initiation with survival and morbidity outcomes.

Methods

We conducted a case-cohort study of 570 HIV-infected adults initiating HAART who were enrolled in a trial of multivitamins in Tanzania. Hemoglobin, serum ferritin, and hepcidin concentrations were assessed at HAART initiation and participants were followed up monthly. We adjusted serum ferritin for inflammation using a regression correction method to characterize hematological status. Cox proportional hazards models were used to estimate HRs for mortality and incident clinical outcomes.

Results

We found an 83% prevalence of anemia, 15% prevalence of iron deficiency anemia, and 66% prevalence of anemia of chronic diseases (ACD). The prevalence of elevated iron was 33% and 19% had iron deficiency (ID). After multivariate adjustment, severe anemia (HR: 2.57; 95% CI: 1.49, 4.45) and ACD (HR: 4.71; 95% CI: 2.91, 7.62) were associated with increased risk of mortality as compared with nonanemic participants. In addition, both ID (HR: 2.65; 95% CI: 1.08, 7.78) and elevated iron (HR: 2.83; 95% CI: 2.10, 3.82) were associated with increased risk of mortality as compared with normal iron concentrations. Severe anemia and elevated iron concentrations were associated with incident wasting and >10% weight loss (P values <0.05).

Conclusions

Anemia and both ID and elevated iron were associated with increased mortality among HIV-infected adults initiating HAART. Safety and efficacy studies including anemia etiology, timing of HAART initiation, and dose of iron supplementation among HIV patients appear warranted.

This trial was registered at clinicaltrials.gov as NCT00383669.

Introduction

Anemia is the most frequent hematological complication in HIV-infected patients, with studies documenting prevalences ranging from 11% to 69% (1–3). The causes of anemia in HIV are multifactorial and include nutritional deficiencies, opportunistic infections (OIs), antiretroviral treatment, and direct effects of HIV infection on bone marrow (4). Anemia has been shown to predict HIV disease progression and mortality independently of CD4 cell count, viral load, and other prognostic factors (1, 2, 5). In addition, iron deficiency (ID) and inflammation-induced iron maldistribution and elevation may be associated with a cycle of impaired immunity, treatment failure, and disease progression in HIV-infected individuals (6–8). The potential risks of impaired hematological status, including anemia and high or low iron, are of particular relevance for HIV treatment programs in sub-Saharan Africa (SSA) where the largest global burdens of HIV and anemia coexist (9).

Many studies have determined that anemia is associated with adverse HIV-related outcomes including suboptimal immune recovery, HIV viral persistence, and increased susceptibility to infections (1, 2, 10, 11). Evidence from a large cohort study of HIV-infected adults in Europe demonstrated a 13-fold increased risk of death among adults with severe anemia at the initiation of highly active antiretroviral therapy (HAART) (12). Further findings from 10 collaborative HIV cohort studies of >12,000 HIV-infected adults in North America and Europe indicated that anemia severity at HAART initiation was associated with increased risk of mortality (13). Although HIV-related anemia is common and is an important predictor of mortality, limited guidance on anemia treatment among HIV-infected patients exists for resource-limited settings. Oral iron supplementation is widely available and effective in managing iron deficiency anemia (IDA) outside the context of chronic inflammation (14). However, iron supplementation has also been associated with growth of microbial pathogens, increased risk of OIs including tuberculosis (TB), and it could potentially worsen HIV disease (9, 15). Nevertheless, data are sparse on the relative contributions of iron status and anemia (in terms of types and severity) at HAART initiation in SSA and the implications for HIV progression and mortality.

To address the need for a comprehensive evaluation of hematological status in HIV in the context of SSA, we undertook a case-cohort study among HIV-infected adults enrolled in a trial of multivitamins in Tanzania. We examined the association of impaired hematological status at HAART initiation with mortality and incident morbidity outcomes.

Methods

We conducted a case-cohort study of hematological status among HIV-infected men and women initiating HAART and who were enrolled and followed for 2 y in the Trial of Vitamins and HAART in HIV Disease Progression, conducted in Dar es Salaam, Tanzania, during 2006–2010 (NCT00383669) (16). Detailed methods of the parent trial have been published elsewhere (16). Briefly, this double-blind, randomized controlled trial assessed the effect of daily oral supplements of vitamins B-complex, C, and E, at high compared with standard levels of the RDA, on HIV disease progression or death. The criteria for enrollment were the following: age 18 y or older, HIV-infected, initiated HAART at enrollment, and intended to stay in Dar es Salaam for ≥2 y. Women who were pregnant or lactating were excluded from the study. Individuals with WHO clinical stage IV disease or CD4+ T-cell count <200 cells/μL, or with WHO clinical stage III disease and CD4+ T-cell count <350/μL, were initiated on HAART as recommended by the Tanzania national HIV treatment guidelines at the time of this study (17). First-line HAART combinations included stavudine (d4T), lamivudine, and nevirapine (NVP). Zidovudine was substituted for d4T for individuals who had peripheral neuropathy or could not tolerate d4T; for patients who could not tolerate NVP, efavirenz was provided as a substitute. Treatment of OIs and cotrimoxazole prophylaxis for patients with CD4+ T-cell count <200 cells/μL were provided as standard of care.

Case-cohort sampling procedure

A total of 3418 participants consented and were enrolled in the parent trial. Among them, a random sample of 1103 participants were later selected for baseline biomarker assessment. Using simple randomization, we further selected a subcohort of 500 participants among those with archived samples. We then enriched the sample with 70 deaths among participants with samples not in the subcohort (Figure 1). The sampling fraction of the subcohort was 0.45 based on using the Cai and Zeng (18) sample size calculations under the following assumptions: 80% power, 40% prevalence of ID, mortality risk of 13%, and an HR of 1.5 for the relation of ID with mortality.

FIGURE 1.

Flowchart for the selection of subcohort participants and cases included in this study.

Baseline covariate assessment

At enrollment, a full clinical examination was conducted, and a structured interview was completed to collect information on demographic characteristics. A medical examination was performed by study physicians and HIV disease stage was assessed in accordance with the WHO guidelines. Blood specimens were collected at baseline and every 4 mo postenrollment for determination of absolute CD4 T-cell count (FACSCalibur flow cytometer, Becton Dickinson) and complete blood count (AcT5 Diff AL analyzer, Beckman Coulter). All samples were collected during the daytime and further details on the assays have been published elsewhere (16, 19). Height and weight measurements were obtained by trained research assistants using calibrated instruments and standard operating procedures.

Exposure assessment and definitions

A total of 570 baseline HAART initiation samples were tested for iron biomarkers and hemoglobin concentration. The iron storage biomarker, serum ferritin, was quantified using electrochemiluminescence immunoassay (Roche Diagnostics) at Boston Children's Hospital in Boston, USA. Serum samples were first coagulated for 30 min at room temperature, then extracted and centrifuged at 1800 × gfor 10 min at room temperature, then the supernatant was mixed with biotinylated and ruthenium-labelled ferritin-specific antibodies to capture the sample ferritin in a sandwich complex. The immune complexes were magnetically entrapped using an electrode, then unbound reagents as well as sample were washed away. Voltage was applied to the electrode stimulating an illuminescent reaction, the intensity of the light being directly proportional to the amount of ferritin present in the sample. Day-to-day precision (defined as the CV expressed as a percentage) at various concentrations of ferritin ranged from 4.3% to 6.4%. Concentrations of high-sensitivity C-reactive protein (hsCRP) and soluble transferrin receptor (sTfR) were determined using an immunoturbidimetric assay on the Roche P Modular system (Roche Diagnostics). Hepcidin concentrations were measured using a hepcidin-25 bioactive enzyme immunoassay (R & D Systems). Day-to-day variability for the hsCRP, sTfR, and hepcidin assays ranged from 1.9% to 3.8%, 1.4% to 2.2%, and 6.2% to 11%, respectively.

We applied the Biomarkers Reflecting Inflammation and Nutritional Determinants of Anemia (BRINDA) method regression-correction approach to adjust ferritin concentrations for hsCRP (20). This approach uses linear regression to adjust ferritin by the hsCRP concentrations on a continuous scale and has been found to better reflect the relation of ferritin concentrations with inflammation than the other methods (21). Iron status was defined using adjusted ferritin concentrations obtained from the regression-correction method as follows: [ferritin_adjusted = ferritin_unadjusted − β1 (hsCRP_obs − hsCRP_ref)] (20–22). A reference limit for hsCRP was set at 10 mg/L (6, 9). Using this approach, we first log transformed ferritin and hsCRP to approximate normality based on regression diagnostics. Second, we estimated the linear regression coefficients for hsCRP with unadjusted ferritin as the outcome. Third, an ln-hsCRP reference value was subtracted from the ln-hsCRP value in the regression equation. The correction was therefore only applied to individuals with hsCRP greater than hsCRP_ref to avoid over-adjustment (20).

Next, we categorized iron status at baseline as low (adjusted ferritin <15 μg/L), normal (adjusted ferritin 15–300 μg/L in men or 15–200 μg/L in women), or elevated (adjusted ferritin >300 μg/L in men or >200 μg/L in women) (23, 24). Anemia was defined as a hemoglobin concentration <13.0 g/dL for men and <12.0 g/dL for women. Severe anemia was defined as a hemoglobin concentration <8.0 g/dL, for both sexes (25). Among those with anemia, IDA was defined based on the adjusted ferritin threshold <15 μg/L (23, 25). Among those with non-IDA, anemia of chronic diseases (ACD) was defined based on adjusted ferritin concentrations >336.2 μg/L. The sTfR/log(ferritin) ratio and hepcidin concentration were used to further differentiate ACD among those with non-IDA and adjusted ferritin ≤336.2 μg/L (ratio <1.0, or 1.0–2.0 and hepcidin >20 ng/mL) (8, 26, 27). An elevated hepcidin concentration was defined as >20 ng/mL, because this cutoff has shown good utility for predicting nonresponsiveness to oral iron therapy among patients with IDA (27, 28).

Outcome assessment and definitions

The primary outcome was all-cause mortality post–HAART initiation. Monthly clinic follow-up visits were conducted and during these visits participants received a clinical examination. Pulmonary TB was diagnosed according to Tanzanian National Tuberculosis and Leprosy Programme guidelines. Participants with symptoms of pulmonary TB were requested to provide a spot sputum specimen at the study visit during which symptoms were noted, an early morning specimen before a second clinic visit scheduled for the next day, and a third sputum specimen at the second clinic visit. Individuals received a diagnosis of pulmonary TB if ≥1 of the 3 sputum smears was positive for acid-fast bacilli, using Ziehl-Nielsen staining techniques, or when a chest radiograph showed features consistent with TB in the absence of positive sputum smear results (29). Pneumonia and oral thrush were diagnosed on the basis of symptoms reported by patients and signs assessed by study physicians. Malaria was diagnosed based on clinical symptoms and microscopic confirmation of any parasitemia in a peripheral blood smear.

At each scheduled visit, height and weight were measured by study nurses. Wasting was defined as a BMI <18.5 kg/m² (30). HIV-related weight loss >10% was also examined as an outcome (31). Immunological failure was defined using either the WHO definition (a decline of CD4+ T-cell count to <100 cells/μL ≥6 mo after initiating treatment with HAART) or the Tanzania HIV treatment guidelines definition [a decline of CD4+ T-cell count to less than the baseline (at HAART initiation) concentration, ≥6 mo after initiating treatment with HAART] (32, 33).

Statistical methods

Means and prevalences of study participant characteristics among the randomly selected subcohort were summarized. The analysis of hematological status with mortality was conducted among all cases (deaths) and the subcohort. Cox proportional hazards models with robust variance were used to investigate the relation of baseline iron status, anemia severity, and type of anemia with mortality. Prentice weights were used to account for oversampling of deaths by design in a case-cohort study (34). To assess the proportional hazards assumption, we included interaction terms of the predictors and the survival time in the model and used the likelihood ratio test.

Secondarily, Cox proportional hazard models with robust variance in the randomly selected subcohort data were used to investigate the relation between baseline hematologic status and first occurrence of immunological failure, pulmonary TB, pneumonia, malaria, oral thrush, wasting, and >10% loss of weight (35). Baseline prevalent outcomes were excluded from the respective models. Individuals without incident events were censored at the date of the last follow-up visit.

Confounders considered for multivariate models were selected based on literature on risk factors for impaired hematological status and the outcomes of interest (6, 10, 16, 19). Covariates entered into the multivariate models included sex, age, education, season, district, BMI, WHO HIV disease, CD4+ T-cell count, HAART regimen, and study regimen. The relation of baseline iron status, anemia severity, and anemia type (as exposures) with primary and secondary outcomes was analyzed using SAS PHREG with the robust variance estimator (36). The missing indicator method was used to handle missing covariate data (37). All P values were 2 sided, and a P value <0.05 was considered statistically significant. Statistical analyses were performed using SAS version 9.2 (SAS Institute).

Ethics statement

The trial protocol was approved by the institutional review boards of the Harvard TH Chan School of Public Health, Muhimbili University of Health and Allied Sciences, the Tanzania Food and Drug Authority, and the Tanzania National Institute of Medical Research.

Results

Table 1 presents baseline characteristics of the 500 participants in the subcohort. The mean ± SD age was 38 ± 8.0 y, with 67% of participants being women. Pre-HAART initiation, 80% of participants had CD4 T-cell counts ≤200 cells/μL and three-quarters of participants presented with WHO stage III or IV. The majority of the participants (417; 83%) were anemic at baseline, including 14% of participants who were severely anemic (hemoglobin < 8.0 g/dL). Almost half of the subcohort had normal iron status at baseline, whereas one-third presented with elevated concentrations of iron. ID and anemia due to other or mixed forms accounted for 15% and 3%, respectively. ACD was the most common type of anemia at HAART initiation and accounted for nearly 80% of all anemia cases at baseline.

TABLE 1.

Baseline characteristics, hematological status, and iron biomarkers among a randomly selected subcohort of 500 HIV-infected adults initiating antiretroviral therapy in Dar es Salaam, Tanzania, 2006–2010¹

Characteristics	Values
Age, y	38.0 ± 8.0
Female sex	336 (67)
BMI, kg/m2	21.0 ± 3.9
Underweight (<18.5)	138 (28)
Normal (18.5–25.0)	285 (57)
Overweight (>25.0)	73 (15)
CD4 count,2 cells/μL	137 ± 99
<50	97 (20)
50 to <100	95 (20)
100 to ≤200	194 (40)
>200	100 (21)
WHO HIV disease stage
I or II	108 (24)
III	287 (63)
IV	61 (13)
HAART regimen prescribed at baseline
d4T,3TC,NVP	263 (60)
d4T,3TC,EFV	59 (13)
AZT,3TC,NVP	45 (10)
AZT,3TC,EFV	71 (16)
Hb concentration, g/dL	10.3 ± 2.3
Anemia status3
No anemia	83 (17)
Moderate anemia	347 (69)
Severe anemia	70 (14)
Serum ferritin (unadjusted), μg/L	457 ± 1275
Serum ferritin (adjusted),4 μg/L	382 ± 1247
Iron status (using adjusted ferritin)5
Low	94 (19)
Normal	239 (48)
High	167 (33)
Serum CRP,6 mg/L	188 ± 352
>10	400 (80)
Type of anemia7
No anemia	83 (17)
IDA	74 (15)
ACD	330 (66)
Others	13 (3)
Hepcidin,8 ng/mL	96 ± 106
High, >20	370 (74)
≤20 (among IDA)	54 (61)

¹Values are means ± SDs for continuous variables or n (%) for categorical variables. Percentages may not total 100 exactly, owing to rounding. ACD, anemia of chronic diseases; AZT, zidovudine; CRP, C-reactive protein; d4t, stavudine; EFV, efavirenz; HAART, highly active antiretroviral therapy; Hb, hemoglobin; IDA, iron deficiency anemia; NVP, nevirapine; 3TC, lamivudine.

²Total number of participants with CD4 counts is 486. CD4 counts were not available for 14 individuals.

³Severity of anemia was defined based on Hb concentration: severe anemia, Hb <8.0 g/dL; moderate anemia, Hb = 8.0 to <12 g/dL for women and 8.0 to <13 g/dL for men; no anemia, Hb ≥12 g/dL for women and ≥13 g/dL for men.

⁴Serum ferritin concentrations were adjusted for inflammation using the Biomarkers Reflecting Inflammation and Nutritional Determinants of Anemia regression-correction method. A correction factor was applied to values with corresponding CRP >10 mg/L.

⁵Iron status was defined using serum ferritin concentrations adjusted for inflammation using the regression-correction method [ferritin_adjusted = ferritin_unadjusted − β1 (CRP_obs − CRP_ref)]. A CRP reference limit was set at 10 mg/L. Adjusted ferritin <15 μg/L = low iron; adjusted ferritin 15 to 300 μg/L in men or adjusted SF 15 to 200 μg/L in women = normal iron status; adjusted ferritin >300 μg/L in men or adjusted ferritin >200 μg/L in women = high iron.

⁶CRP >10 mg/L is predictive of inflammation.

⁷Type of anemia was defined using Hb, adjusted serum ferritin concentrations, transferrin/log(ferritin) ratio, and hepcidin concentration: IDA = anemia with adjusted serum ferritin <15 μg/L; ACD = non-IDA with adjusted ferritin concentrations >336.2 μg/L, or non-IDA with transferrin/log(ferritin) ratio <1.0 (or 1.0–2.0 and hepcidin >20 ng/mL); other forms of anemia = non-IDA and adjusted ferritin ≤336.2 μg/L with transferrin/log(ferritin) ratio >2.0 (or 1.0–2.0 and hepcidin ≤20 ng/mL).

⁸The hepcidin >20 ng/mL cutoff was used to define elevated hepcidin concentrations and predicts nonresponsiveness to oral iron therapy among patients with IDA.

Table 2 presents the relation of anemia severity, iron status, and type of anemia with the risk of mortality. A total of 139 deaths occurred: 69 in the subcohort and 70 outside the subcohort. After multivariate adjustment, participants who presented with severe anemia had 2.57 times (95% CI: 1.49, 4.45; P < 0.001) the hazard of death as compared with participants without anemia at baseline. Participants with ID at HAART initiation had 2.65 times (95% CI: 1.08, 7.78; P = 0.04) the hazard of death compared with participants with normal iron status. We also found that elevated iron concentrations at baseline were associated with 2.83 times (95% CI: 2.10, 3.82; P < 0.001) the hazard of death compared with participants with normal iron status. Participants presenting with ACD at baseline had 4.71 times (95% CI: 2.91, 7.62; P < 0.001) the hazard of death compared with participants without anemia (Table 2).

TABLE 2.

Risk of mortality by iron status and anemia at HAART initiation among 570 HIV-infected adults initiating HAART in Dar es Salaam, Tanzania, 2006–2010¹

Total		HR (95% CI)	P value	HR (95% CI)	P value
Anemia severity at baseline2
	No anemia (n/N = 10/89)	Moderate anemia (n/N = 88/392)		Severe anemia (n/N = 41/89)
Unadjusted	Ref	1.89 (1.35, 2.64)	<0.01*	5.10 (3.54, 7.32)	<0.01**
Adjusted3	Ref	1.24 (0.80, 1.91)	0.33	2.57 (1.49, 4.45)	<0.01**
Baseline iron status4
	Normal iron (n/N = 34/228)	Low iron (n/N = 15/86)		High iron (n/N = 90/256)
Unadjusted	Ref	1.11 (0.41, 2.98)	0.28	3.19 (2.57, 3.96)	<0.01**
Adjusted3	Ref	2.65 (1.08, 7.78)	0.04	2.83 (2.10, 3.82)	<0.01**
Type of anemia at baseline (ACD based on unadjusted ferritin, ferritin index, and hepcidin)5
	No anemia (n/N = 10/89)	IDA (n/N = 15/82)		ACD (n/N = 113/386)
Unadjusted	Ref	1.00 (0.63, 1.60)	0.99	2.74 (1.97, 3.82)	<0.01**
Adjusted3	Ref	1.43 (0.75, 2.73)	0.28	4.71 (2.91, 7.62)	<0.01**

¹Total deaths = 139. Based on Cox models including the random subcohort and the nonsubcohort deaths, weighted (Prentice weights), and with the robust variance. ^*,**Significant difference from reference: *P value <0.001, **P value <0.0001. ACD, anemia of chronic diseases; CRP, C-reactive protein; d4t, stavudine; EFV, efavirenz; HAART, highly active antiretroviral therapy; Hb, hemoglobin; IDA, iron deficiency anemia; n, number of deaths; N, number of participants in exposure category; NVP, nevirapine; 3TC, lamivudine.

²Severity of anemia was defined based on Hb concentration: severe anemia, Hb <8.0 g/dL; moderate anemia, Hb = 8.0 to <12 g/dL for women and 8.0 to <13 g/dL for men; no anemia, Hb ≥12 g/dL for women and ≥13 g/dL for men.

³Adjusted for sex (male, female) and baseline age (<30, 30 to <40, 40 to <50, ≥50 y), education (primary, secondary/college, informal), season (long rain, harvest, postharvest, short rain), district, BMI (<18.5, 18.5–25.0, >25.0 kg/m²), WHO HIV disease stage (stage I or II, stage III, stage IV), CD4+ T-cell count (<50, 50 to <100, 100 to ≤200, >200), HAART regimen (AZT 3TC NVP, d4T 3TC EFV, d4T 3TC NVP, missing HAART regimen), and study regimen. Models were adjusted for anemia at baseline with the exception of the second model on baseline anemia severity and mortality.

⁴Iron status was defined using serum ferritin concentrations adjusted for inflammation using the regression-correction method [ferritin_adjusted = ferritin_unadjusted − β1 (CRP_obs − CRP_ref)]. A CRP reference limit was set at 10 mg/L. Adjusted ferritin <15 μg/L = low iron; adjusted ferritin 15–300 μg/L in men or adjusted ferritin 15–200 μg/L in women = normal iron status; adjusted ferritin >300 μg/L in men or adjusted serum ferritin >200 μg/L in women = high iron.

⁵Type of anemia was defined using Hb, adjusted serum ferritin concentrations, transferrin/log(ferritin) ratio, and hepcidin concentration: IDA = anemia with adjusted serum ferritin <15 μg/L; ACD = non-IDA with adjusted ferritin concentrations >336.2 μg/L, or non-IDA with transferrin/log(ferritin) ratio <1.0 (or 1.0–2.0 and hepcidin >20 ng/mL); other forms of anemia = non-IDA and adjusted ferritin ≤336.2 μg/L with transferrin/log(ferritin) ratio >2.0 (or 1.0–2.0 and hepcidin ≤20 ng/mL). Totals exclude 13 participants (including 1 death) who had mixed or other forms of anemia.

We present the relation of hematological markers with diagnosis of incident clinical, hematological, and immunological outcomes among the subcohort in Tables 3 and 4 and Supplemental Table 1. After multivariate adjustment, participants with severe anemia had 3.34 times (95% CI: 1.22, 9.21; P = 0.02) the hazard of incident wasting and 5.48 times (95% CI: 1.56, 19.32; P < 0.01) the hazard of experiencing >10% weight loss as compared with participants without anemia at baseline. Elevated iron (HR: 2.13; 95% CI: 1.05, 4.34; P = 0.04) and IDA (HR: 3.84; 95% CI: 1.14, 12.98; P = 0.03) were significantly associated with >10% weight loss as compared with normal iron and no anemia, respectively.

TABLE 3.

Risk of incident clinical and immunological outcomes by baseline anemia, among 500 HIV-infected adults initiating HAART in Dar es Salaam, Tanzania, 2006–2010¹

	No anemia		Moderate anemia			Severe anemia
Outcome (values)	n/N		n/N	HR (95% CI)	P value	n/N	HR (95% CI)	P value
Pulmonary tuberculosis (20/493)
Unadjusted	1/69	Ref	16/343	1.34 (0.39, 4.58)	0.65	3/81	0.49 (0.05, 4.66)	0.53
Adjusted		Ref		1.89 (0.40, 9.10)	0.42		0.93 (0.08, 11.00)	0.95
Pneumonia (182/469)
Unadjusted	23/79	Ref	133/326	1.44 (0.92, 2.24)	0.11	26/64	1.77 (1.01, 3.11)	0.05
Adjusted		Ref		1.29 (0.79, 2.11)	0.32		1.59 (0.84, 3.02)	0.15
Malaria (144/463)
Unadjusted	26/75	Ref	98/327	0.90 (0.58, 1.39)	0.63	20/61	1.30 (0.73, 2.33)	0.38
Adjusted		Ref		0.85 (0.52, 1.38)	0.52		1.17 (0.60, 2.89)	0.65
Oral thrush (59/464)
Unadjusted	9/83	Ref	41/319	1.23 (0.60, 2.54)	0.57	9/62	1.79 (0.71, 4.51)	0.22
Adjusted		Ref		0.99 (0.40, 2.46)	0.98		1.78 (0.58, 5.46)	0.32
Wasting, BMI < 18.5 kg/m2 (52/343)
Unadjusted	7/65	Ref	29/237	1.09 (0.48, 2.48)	0.85	16/41	4.43 (1.82, 10.78)	0.001
Adjusted		Ref		0.82 (0.34, 1.95)	0.65		3.34 (1.22, 9.21)	0.02
>10% weight loss (61/470)2
Unadjusted	6/81	Ref	42/326	1.63 (0.69, 3.84)	0.26	13/63	3.27 (1.24, 8.61)	0.02
Adjusted		Ref		2.58 (0.89, 7.49)	0.08		5.48 (1.56, 19.32)	0.001
Immunological failure (23/403)3
Unadjusted	4/74	Ref	17/283	1.04 (0.35, 3.08)	0.95	2/46	0.74 (0.14, 4.05)	0.73
Adjusted		Ref		0.57 (0.14, 2.30)	0.43		1.99 (0.23, 16.97)	0.53

¹Severity of anemia was defined based on Hb concentration: severe anemia, Hb <8.0 g/dL; moderate anemia, Hb = 8.0 to <12 g/dL for women and 8.0 to <13 g/dL for men; no anemia, Hb ≥12 g/dL for women and ≥13 g/dL for men. Adjusted for sex (male, female) and baseline age (<30, 30 to <40, 40 to <50, ≥50 y), BMI (<18.5, 18.5–25.0, >25.0 kg/m²), WHO HIV disease stage (stage I or II, stage III, stage IV), CD4+ T-cell count (<50, 50 to <100, 100 to ≤200, >200), HAART regimen (AZT 3TC NVP, d4T 3TC EFV, d4T 3TC NVP, missing HAART regimen), and study regimen. d4t, stavudine; EFV, efavirenz; HAART, highly active antiretroviral therapy; Hb, hemoglobin; n, number of incident events; N, number of the subcohort participants without the event at baseline; NVP, nevirapine; 3TC, lamivudine.

²Percentage decrease from baseline weight measurement.

³Immunological failure was defined using the WHO definition (a decline of CD4+ T-cell count to <100 cells/μL 6 mo after initiating treatment with HAART) or Tanzania HIV treatment guidelines definition [a decline of CD4+ T-cell count to less than the baseline (at HAART initiation) concentration, 6 mo after initiating treatment with HAART].

TABLE 4.

Risk of clinical, hematological, and immunological incident outcomes by baseline iron status, among 500 HIV-infected adults initiating HAART in Dar es Salaam, Tanzania, 2006–2010¹

	Normal iron		Low iron			High iron
Outcome (values)	n/N		n/N	HR (95% CI)	P value	n/N	HR (95% CI)	P value
Pulmonary tuberculosis (20/491)
Unadjusted	7/209	Ref	8/77	3.13 (1.13, 8.65)	0.03	5/205	0.80 (0.25, 2.51)	0.70
Adjusted		Ref		2.23 (0.68, 7.22)	0.17		0.61 (0.16, 2.27)	0.46
Pneumonia (182/467)
Unadjusted	70/202	Ref	33/73	1.39 (0.91, 2.10)	0.12	79/192	1.43 (1.03, 1.97)	0.03
Adjusted		Ref		1.24 (0.79, 1.95)	0.35		1.43 (0.97, 2.11)	0.07
Malaria (144/462)
Unadjusted	59/199	Ref	29/73	1.35 (0.87, 2.11)	0.19	56/190	1.11 (0.77, 1.60)	0.56
Adjusted		Ref		1.12 (0.68, 1.79)	0.68		1.41 (0.93, 2.13)	0.11
Oral thrush (59/462)
Unadjusted	26/202	Ref	15/74	1.59 (0.84, 3.00)	0.15	18/186	0.81 (0.45, 1.49)	0.50
Adjusted		Ref		1.14 (0.51, 2.55)	0.74		0.82 (0.39, 1.74)	0.61
Wasting, BMI <18.5 kg/m2 (52/352)
Unadjusted	18/163	Ref	13/53	2.22 (1.08, 4.53)	0.03	21/126	1.59 (0.86, 2.98)	0.15
Adjusted		Ref		2.08 (0.93, 4.61)	0.07		1.57 (0.76, 3.23)	0.22
>10% weight loss (61/469)2
Unadjusted	18/203	Ref	14/71	2.16 (1.07, 4.35)	0.03	29/195	1.87 (1.03, 3.37)	0.04
Adjusted		Ref		2.00 (0.89, 4.50)	0.09		2.13 (1.05, 4.34)	0.04
Immunological failure (23/403)3
Unadjusted	8/185	Ref	2/65	0.61 (0.13, 2.90)	0.54	13/153	1.90 (0.79, 4.60)	0.15
Adjusted		Ref		2.29 (0.31, 16.94)	0.42		0.46 (0.13, 1.66)	0.24

¹Iron status defined using serum ferritin concentrations adjusted for inflammation using the regression-correction method [ferritin_adjusted = ferritin_unadjusted − β1 (CRP_obs − CRP_ref)]. A CRP reference limit was set at 10 mg/L. Adjusted serum ferritin <15 μg/L = low iron; adjusted serum ferritin 15 to <300 μg/L in men or adjusted serum ferritin 15 to <200 μg/L in women = normal iron status; adjusted serum ferritin >300 μg/L in men or adjusted serum ferritin >200 μg/L in women = high iron. Adjusted for sex (male, female) and baseline age (<30, 30 to <40, 40 to <50, ≥50 y), BMI (<18.5, 18.5–25.0, >25.0 kg/m²), WHO HIV disease stage (stage I or II, stage III, stage IV), CD4+ T-cell count (<50, 50 to <100, 100 to ≤200, >200), HAART regimen (AZT 3TC NVP, d4T 3TC EFV, d4T 3TC NVP, missing HAART regimen), and study regimen. CRP, C-reactive protein; d4t, stavudine; EFV, efavirenz; HAART, highly active antiretroviral therapy; n, number of incident events; N, number of the subcohort participants without the event at baseline; NVP, nevirapine; 3TC, lamivudine.

²Percentage decrease from baseline weight measurement.

³Immunological failure was defined using the WHO definition (a decline of CD4+ T-cell count to <100 cells/μL 6 mo after initiating treatment with HAART) or Tanzania HIV treatment guidelines definition [a decline of CD4+ T-cell count to less than the baseline (at HAART initiation) concentration, ≥6 mo after initiating treatment with HAART].

Discussion

We found a high prevalence of impaired hematological status among Tanzanian HIV-infected adults initiating HAART. Individuals with impaired hematological status, including severe anemia, ACD, and deficient or elevated concentrations of iron, had increased risk of mortality post–HAART initiation. Severe anemia and elevated iron were associated with increased risk of wasting and weight loss. We did not identify a relation of baseline iron status, anemia severity, and type of anemia with pulmonary TB, oral thrush, or immunological failure.

To the best of our knowledge, this is the first study to systematically characterize anemia among HIV adult patients initiating HAART in SSA using the BRINDA correction method to adjust serum ferritin for inflammation (20). The prevalence of ID in our study (19%) was comparable with other studies of HIV-infected adults initiating HAART in low- and middle-income country settings (38, 39). However, we found the prevalence of elevated iron status (33%) (hyperferritinemia) was slightly less than the 40% and 48% found among Ugandan and Tanzanian HIV-TB co-infected patients, respectively (9, 38). There have been inconsistent findings on the proportion of IDA and ACD among HIV patients initiating HAART in SSA. Some studies classified about half of the anemia cases in HIV as IDA (6, 40), whereas other studies have shown ACD to be more prevalent, particularly among HIV-TB co-infected patients (27, 41). In our study, ACD accounted for almost 80% of anemia cases at HAART initiation, IDA 15%, and other forms of anemia 3%. In South Africa, a study designed to determine the contributions of ACD and ID to anemia in patients with HIV/TB co-infection found the prevalence of anemia to be 80%, of which 95% of cases were attributed to ACD (27). The predominance of ACD among anemic HIV-infected individuals may explained by high amounts of inflammation, HIV-associated upregulation of proinflammatory cytokines, and stimulation of C-reactive protein and hepcidin synthesis by hepatocytes (41).

Our finding that anemia and iron status were associated with mortality is consistent with prior studies (38, 41–43). Using fewer biomarkers (hemoglobin concentration and microcytosis) to characterize hematological impairment among 40,657 adult HIV-infected patients receiving HAART in Tanzania, our team has shown significantly higher mortality risk associated with anemia and IDA at HAART initiation and during follow-up with the magnitude of association being stronger among iron-supplemented users (43). Iron is required for several steps in the HIV-1 life cycle, including reverse transcription, HIV-1 gene expression, and capsid assembly (44). Iron status has been shown to predict treatment failure, with elevated iron being associated with increased reverse transcriptase activity and availability of iron-dependent enzymes that promote viral replication that can lead to impaired immunity, treatment failure, and increased risk of mortality from TB and other OIs (9, 45–47). In addition, we found a stronger association with mortality for participants with elevated iron status than for those with normal iron status (43). Consistent with our findings on ACD, increased risk of mortality could be explained by the iron redistribution occurring in the context of chronic inflammation and its effect on increasing viremia and OIs. We also found a strong association with mortality among participants with severe anemia as compared with those without anemia, which is consistent with existing knowledge on inflammation-mediated pathways for HIV-associated anemia (27, 38, 41).

About 60% of participants with IDA had low hepcidin concentrations (≤20 ng/mL) and therefore might respond to oral iron therapy. This proportion is higher than that found in HIV/TB co-infected adults in South Africa, where the proportion of IDA cases that might respond to oral iron supplementation was estimated to be 15% (27). Nonresponsiveness to oral iron supplementation among patients with IDA and elevated hepcidin suggests concurrent ACD, and that these patients are likely to benefit less from oral iron supplementation unless the accompanying proinflammatory response resolves (8, 27, 28). Although oral iron supplementation may be effective in treating IDA especially when inflammatory responses are relatively low, the safety and efficacy of iron supplements in the majority of HIV-infected patients initiating HAART have not been established. Increased dietary or supplemental iron may be associated with a range of side effects including gastrointestinal effects that may interfere with adherence to essential medications, e.g., HAART or anti-TB drugs (48). The safety and efficacy of oral iron therapy in HIV-infected adults initiating HAART remain unclear.

The study has several important limitations. First, despite adjusting for a number of potential confounders in the analysis, there is a possibility for residual and unmeasured confounding. In addition, we did not investigate additional etiologies of anemia, including noniron nutritional deficiencies (folate, vitamin B-12), infections, and genetic disorders. Therefore, we may have misclassified a small proportion of mixed forms of anemia or those resulting from sickle cell disease or thalassemia. We only assessed hematological biomarkers at the time of HAART initiation and therefore we were not able to assess the association of hematological status post–HAART initiation with mortality and disease progression.

In summary, we found a high prevalence of impaired hematological status among HIV-infected adults initiating HAART in Tanzania. ACD was the predominant anemia etiology and was a strong risk factor for mortality. We also found that elevated iron, ID, and severe anemia were associated with increased risk of mortality. Low hepcidin concentrations were common among participants with IDA and this suggests that these anemia cases may respond to oral iron therapy. We conclude that etiology-informed management of anemia in HIV presents an opportunity to improve survival and treatment outcomes for HIV-infected adults initiating HAART. Safety and efficacy of iron supplementation among HIV-infected adults need to be studied, and especially timing with respect to HAART initiation, etiology, and focus on IDA, as well as dose of the intervention. Further characterization of anemia persisting post–HAART initiation presents an opportunity for improved clinical monitoring, care, and survival of HIV patients on treatment.

Notes

Supported by National Institute of Child Health and Human Development grant R01 HD32257 (to WWF) and the Afya Bora fellowship sponsored by the President's Emergency Plan for AIDS Relief and the Office of AIDS Research (to RAN).

Author disclosures: The authors report no conflicts of interest.

Supplemental Table 1 is available from the “Supplementary data” link in the online posting of the article and from the same link in the online table of contents at https://academic.oup.com/jn/.

Abbreviations used: ACD, anemia of chronic diseases; BRINDA, Biomarkers Reflecting Inflammation and Nutritional Determinants of Anemia; d4T, stavudine; HAART, highly active antiretroviral therapy; hsCRP, high-sensitivity C-reactive protein; ID, iron deficiency; IDA, iron deficiency anemia; NVP, nevirapine; OI, opportunistic infection; SSA, sub-Saharan Africa; sTfR, soluble transferrin receptor; TB, tuberculosis.