Predictive Value of Anemia for Tuberculosis in HIV-Infected Patients in Sub-Saharan Africa

An Indication for Routine Microbiological Investigation Using New Rapid Assays

Abstract

Background: 

The relationship between anemia and undiagnosed tuberculosis (TB) in patients living with HIV in sub-Saharan Africa is incompletely defined. We assessed the prevalence of TB among those with HIV-related anemia and evaluated new means of rapid TB diagnosis.

Methods: 

Blood hemoglobin levels were measured in unselected antiretroviral treatment–naive patients in Cape Town, South Africa, and anemia was classified according to World Health Organization criteria. All patients were screened for TB by testing paired sputum samples using liquid culture (reference standard), fluorescence microscopy, and Xpert MTB/RIF. Urine samples were tested for lipoarabinomannan (LAM) using the Determine TB-LAM diagnostic assay.

Results: 

Of 602 adults screened, 485 had complete results. Normal hemoglobin levels were found in 44.5% (n = 216) of patients, and mild, moderate, or severe anemia were present in 24.9% (n = 121), 25.4% (n = 123) and 5.2% (n = 25) of patients, respectively. Culture-confirmed pulmonary TB was diagnosed in 8.8% (19/216) of those without anemia compared with 16.5% (20/121), 26.0% (32/123), and 40.0% (10/25) among those with mild, moderate, or severe anemia, respectively (P < 0.001). Anemia was a strong independent predictor of TB. The sensitivities of diagnostic assays were much higher among those with moderate/severe anemia compared with those with no/mild anemia using sputum microscopy (42.9% vs 15.4%), urine LAM (54.8% vs 0%), sputum microscopy plus urine LAM (71.4% vs 15.4%), and sputum Xpert (73.8% vs 41.0%) (P < 0.01 for all).

Conclusions: 

A very high prevalence of undiagnosed TB was found in patients with moderate or severe anemia. Such patients should be prioritized for routine microbiological investigation using rapid diagnostic assays.

BACKGROUND

Anemia is the most common hematologic abnormality associated with HIV.^1–3 It is frequently observed among people living with HIV (PLWH), especially among those with evidence of advanced HIV disease.⁴ Clinical consequences associated with HIV-associated anemia include severe fatigue,⁵ poorer quality of life for PLWH,^6,7 and possibly an increased rate of HIV disease progression.^2,8,9 Anemia is independently associated with increased mortality risk^10,11 and remains a common problem among ART-naive patients in sub-Saharan Africa.¹²

Anemia may also be related to HIV-associated tuberculosis (TB), which remains the leading cause of death among PLWH worldwide.¹³ Anemia is an independent predictor of early incident TB among those initiating ART in sub-Saharan Africa^12,14 and is also associated with increased mortality in those with HIV-associated TB.^15–18 For these reasons, some have recommended routine TB investigations among PLWH with low hemoglobin levels living in high TB incidence settings, before starting ART.¹⁹

We conducted this study to further assess whether there is a need to investigate for TB in PLWH who are enrolling for ART and are found to be anemic. In this study of ART-naive patients enrolling in a large community-based clinic in Cape Town, South Africa, we report on the prevalence of anemia, the prevalence of active TB and the relationship between the two. We also evaluated novel tools and approaches to diagnosis of TB in those with anemia to identify effective means of systematic screening in this patient group.

METHODS

Patient Characteristics

We have previously described in detail the ART service in Gugulethu township in Cape Town, and the high TB burden of TB among its patients.^20–23 Those eligible for this study were ART-naive, HIV-infected adults, who were without a current TB diagnosis and were consecutively recruited from among patients newly referred to the clinic for initiation of ART between March 2010 and April 2011. Written informed consent was provided by all patients, and the study was approved by the research ethics committees of the University of Cape Town and the London School of Hygiene and Tropical Medicine, United Kingdom.

At their first visit to the clinic, patients completed standard symptom screening that included the World Health Organization (WHO) symptom screen (the presence of more than 1 of the following symptoms: cough of any duration, fever, weight loss, or night sweats) for HIV-associated TB.²⁴ They were clinically characterized and clinical samples were obtained. Whenever possible, 2 sputum samples were obtained; the first sample was a spot specimen followed by a second induced sample.²⁵ Urine samples were collected and stored at −20°C. Blood CD4 cell counts and plasma viral load were measured. Hemoglobin levels were determined using ADVIA 2120 hematology analyzer (Siemens Healthcare Diagnostics, Erlangen, Germany). Chest radiographs were obtained when possible and were reported by a reader experienced in using the Chest Radiographic Reading and Recording System.²⁶

Laboratory Procedures

Sputum samples were processed in an accredited laboratory according to standardized protocols with external quality assurance procedures. Decontaminated and centrifuged sputum samples were examined for acid-fast bacilli using auramine O fluorescent stain, tested using the Xpert MTB/RIF assay, and cultured in liquid media as previously described.^27–29 Positive cultures were speciated. Determine TB-LAM lateral-flow assay was used to retrospectively analyze defrosted frozen urine samples for the presence of lipoarabinomannan (LAM). A test band equal to or greater than the intensity of the weakest band on the reference card (grade 1) defined a positive LAM result. Technologists blinded to the outcomes of the other assays read the results of all tests.

Patient Outcomes

Patients were followed up within the routine ART service. Those diagnosed as having TB were referred to local TB treatment clinics. ART service patient records were reviewed to determine vital status at 90 days.

Definitions and Analysis

Mycobacterium tuberculosis cultured from 1 or more sputum samples was used to define TB cases. Using hemoglobin values before ART initiation, all patients were categorized according to WHO criteria³⁰ as having no anemia (≥13.0 g/dL for men, ≥12.0 g/dL for non–pregnant women, and ≥11.0 g/dL for pregnant women), mild anemia (11.0–12.9 g/dL for men, 11.0–11.9 g/dL for non–pregnant women, and 10.0–10.9 g/dL for pregnant women), moderate anemia (8.0–10.9 g/dL for men and non–pregnant women and 7.0–9.9 g/dL for pregnant women), or severe anemia (<8.0 g/dL for men and non–pregnant women and <7.0 g/dL for pregnant women).

Medians were compared using either Wilcoxon rank-sum tests or Kruskal–Wallis tests where appropriate. The χ² tests or Fisher exact tests were used as indicated to compare proportions. The sensitivity, specificity, and predictive values of the different TB diagnostic assays were calculated [with corresponding 95% confidence intervals (CIs)] for patient groups stratified by severity of anemia. Logistic regression analyses were used to identify factors independently associated with HIV-associated anemia. All variables in the univariable model meeting a cutoff of P ≤ 0.1 were included in the multivariable model. Statistical tests were 2-sided at α = 0.05.

RESULTS

Anemia Diagnoses

Of 602 patients recruited, 485 had complete results of tests performed on blood, sputum, and urine samples. Among those included in the analysis (n = 485), the median age was 33.6 [interquartile range (IQR), 27.9–40.7], 63.5% were women and the median CD4 count was 169 cells per microliter (IQR, 96–231). The median blood hemoglobin level was 12.0 g/dL (IQR, 10.6–13.4).

Anemia was diagnosed in 269 patients (prevalence, 55.5%; 95% CI: 50.9 to 60.0) and the remaining 216 (44.5%) had normal hemoglobin levels. Anemia was classified as mild in 121 (24.9%), moderate in 123 (25.4%), and severe in 25 (5.2%). Patients with greater severity of anemia were more likely to be women, have lower body mass index and have higher white cell counts and absolute neutrophil counts (Table 1). Such patients also tended to have lower CD4 cell counts, higher HIV viral load, pulmonary TB, and either stage 3 or 4 HIV disease at program enrollment. However, pregnancy was not associated with degree of anemia.

TABLE 1: Patient Characteristics Among All HIV Patients Independent of TB Status (n = 485)

Prevalence of TB Among Patients With HIV-Associated Anemia

Overall, 81 patients had culture-positive TB diagnosed (prevalence, 16.7%; 95% CI: 13.5 to 20.3) and 404 (83.3%) patients had negative sputum cultures. Among the 269 patients with any degree of anemia, 62 had TB (prevalence, 23.0%; 95% CI: 18.2 to 28.5). The prevalence of TB was strongly and directly correlated with degree of anemia (Fig. 1). However, the prevalence of TB did not differ by gender, as 33.3% (95% CI: 15.6 to 55.3) of men with moderate or severe anemia had TB compared with 27.4% (95% CI: 19.8 to 36.2) among women with moderate or severe anemia (P = 0.556). A positive WHO symptom screen was found in 66 (81.5%) of overall patients with TB and in 88.1% of the subset of TB patients with either moderate or severe anemia compared with 75.0% among those with no or mild anemia (P = 0.0094).

FIGURE 1: The proportion (with 95% CIs) of HIV-infected patients (n = 485) who have TB, stratified by degree of anemia.

Risk Factors for HIV-Associated Anemia

We next used logistic regression to define whether TB was an independent risk factor associated with anemia. Both univariable and multivariable analyses demonstrated that a number of variables were associated with moderate or severe anemia (Table 2). TB remained strongly associated with moderate and severe anemia after adjustment for all other variables.

TABLE 2: Univariable and Multivariable Analysis of Factors Associated With Moderate or Severe Anemia Among HIV-Infected Patients (n = 485)

Characteristics and Outcomes of Patients With HIV-Associated TB and Anemia

Among patients with HIV-associated TB (n = 81), those with lower hemoglobin levels were more likely to be female, have higher absolute neutrophil counts, lower CD4 cell counts, and higher HIV viral loads (data not shown). Radiological abnormalities were generally not associated with anemia with the exception that those with more severe anemia were more likely to have mediastinal lymphadenopathy. Patients with a greater severity of anemia were much less likely to be retained in program after 90 days due to death or loss to follow-up (Fig. 2). Patients with TB who died during follow-up (n = 5) all had moderate or severe anemia (P = 0.026).

FIGURE 2: The proportion (with 95% CIs) of patients with HIV-associated TB (n = 81) who either died or were lost to follow-up within 90 days of ART initiation, stratified by degree of anemia.

Diagnosis of TB Among Patients With HIV-Associated Anemia

We next assessed the diagnostic accuracy of a range of microbiological assays for TB used during routine systematic screening (Tables 3–5). It was striking that for each diagnostic assay, the sensitivities were significantly greater among those with moderate or severe anemia compared with those with mild or no anemia (Table 3). This increment ranged from 27.5% for sputum microscopy to 54.8% for Determine TB-LAM. Sputum Xpert MTB/RIF and Determine TB-LAM each detected a majority of TB cases among those with moderate or severe anemia, whereas sputum smear microscopy did not. Combining Determine TB-LAM with either sputum smear microscopy or Xpert MTB/RIF increased the diagnostic sensitivity among those with moderate or severe anemia to more than 70% and 80%, respectively.

TABLE 3: Diagnostic Utility of Novel Diagnostics for the Detection of TB Among Patients With HIV-Associated Anemia; Sensitivity Among Patients With Culture-Confirmed TB

Both sputum Xpert MTB/RIF and Determine TB-LAM correctly identified TB in all of those who died in the first 90 days of clinical follow-up (n = 5), whereas sputum microscopy only diagnosed TB in 2 of the 5 cases. The positive predictive value was maintained above 90% for all assays among those with moderate or severe anemia (Table 4). The specificity of all assays either in isolation or in combination was greater than 98% and 97%, respectively, and did not differ according to the severity of anemia (Table 5).

TABLE 4: Diagnostic Utility of Novel Diagnostics for the Detection of TB Among Patients With HIV-Associated Anemia; PPV and NPV Among the Overall Cohort

TABLE 5: Diagnostic Utility of Novel Diagnostics for the Detection of TB Among Patients With HIV-Associated Anemia; Specificity Among Patients Without Culture Confirmed TB

DISCUSSION

This study found a high prevalence of anemia among treatment-naive patients enrolling to start ART and that a substantial proportion of those with anemia had underlying TB, which was a strong independent risk factor for anemia. The prevalence of TB among those with moderate or severe anemia was so high and their clinical outcomes were so poor that it suggests the need for routine microbiological investigations for TB in this patient group. A majority of TB among those with moderate or severe anemia could be rapidly diagnosed using Determine TB-LAM and/or Xpert MTB/RIF.

The prevalence of TB among ART-naive patients with anemia in developing countries is poorly defined and is likely to vary greatly between settings. We carefully documented this relationship in a South African cohort, showing that a higher prevalence of TB was directly associated with lower hemoglobin levels (reaching as high as 40% among those with severe anemia) and that TB was a strong independent risk factor for HIV-associated anemia. These data are consistent with the findings of previous studies from Malawi,³¹ Rwanda,³² and India.³³ Multivariable analysis additionally found that while low CD4 cell counts, high viral loads, and low body mass indexes also demonstrated associations, female sex was the factor most strongly associated with moderate or severe anemia. This may be explained, for example, by menstrual blood loss and low dietary iron intake.

The mechanisms underlying anemia in patients with HIV-associated TB remain incompletely defined and are likely to be multiple. The most common mechanism is likely anemia of chronic disease.^34,35 Iron-deficiency anemia^35,36 may also occur as a result of insufficient dietary intake or blood loss from the gastrointestinal tract due to mucosal involvement with TB.³⁷ TB may also disseminate to the bone marrow^38,39 and impair all hematopoietic cell lines, including red blood cells. Other reported mechanisms may include autoimmune hemolysis^40,41 and nutritional deficiencies of folate, selenium,⁴² and rarely vitamin B12 secondary to malabsorption caused by ileal TB involvement.⁴³ Additionally, HIV-associated anemia is associated with worsening HIV disease parameters,⁴⁴ including higher HIV viral load and low CD4 counts^2,45 and the risk for active TB disease increases with greater severity of immunosuppression.

Among those with HIV-associated TB, all deaths occurred among those with moderate or severe anemia, and patients were overall less likely to be retained in care after 3 months in program. Although the outcomes of those lost to follow-up are unknown, it is likely that those not retained in care were at high risk for mortality.^46,47 Our findings are consistent with other studies, which have demonstrated that lower hemoglobin levels are associated with decreased survival and may independently predict mortality among patients with HIV-associated TB.^15–18 Therefore, we suggest that routine investigation for TB among HIV patients with anemia may not only yield a large number of TB cases, but may identify many of those at greatest risk for significant TB-related morbidity and mortality.

In resource-limited settings, the WHO symptom screen for HIV-associated TB²⁴ is used to identify those who require further evaluation with a view to possible microbiological screening for TB. However, the sensitivity of this screen is incomplete²⁴ (as found in the present study). Moreover, due to its very low specificity, symptom screening identifies very large numbers of patients for whom it is simply not feasible to conduct microbiological investigations. However, the association between anemia and TB was so strong that this might reasonably be used as an absolute indication for microbiological screening. Thus, the presence of anemia readily identifies a subset of patients for whom investigations should be prioritized regardless of the presence or absence of symptoms.

We have previously shown that sputum Xpert MTB/RIF and urine Determine TB-LAM assays have higher diagnostic sensitivity in those with poorer prognostic characteristics, including anemia.⁴⁸ In the present study, we have now demonstrated the sensitivities of these assays used alone, together or in combination with smear microscopy among those with anemia classified according to WHO criteria. The observation that the sensitivities of these diagnostic assays were much higher among those with greater severity of anemia is likely to be related to the probability that both of these factors are associated with more advanced and disseminated disease and mycobacterial load.⁴⁸ Anemia may be particularly severe in patients with disseminated HIV-associated TB due to several factors. These include high levels of systemic inflammation (with upregulation of myelosuppressive proinflammatory cytokines) direct involvement of the bone marrow (leading to suppression of hematopoiesis), and gastrointestinal tract (leading to blood loss).

Determine-TB LAM is best prioritized for use in screening for TB among HIV-infected patients with CD4 cell counts <200 cells per microliter.^49,50 However, in many resource-limited settings, CD4 counts may not be available. Low hemoglobin levels may, however, represent an alternative simple trigger for appropriate TB testing among HIV-infected patients using LAM point-of-care assay. Further studies to assess the diagnostic accuracy and impact of TB screening using this and other rapid diagnostic assays in patients with moderate-to-severe HIV-associated anemia in resource-limited settings are prudent.

Among patients with moderate or severe anemia, the sensitivity of LAM point-of-care assay combined with sputum microscopy was very similar to that of a single sputum Xpert cartridge. This has important implications for resource-limited settings. Although Xpert is being scaled up and is becoming available in several sub-Saharan African countries, its cost and technical requirements will prohibit its implementation in some settings with high HIV-associated TB burdens.⁵¹ Microscopy is already widely available in most resource-limited settings and continues to be first line for TB diagnosis in most of these settings. LAM point-of-care assay is rapid, easy to use, and inexpensive and may be a very useful add-on test in resource-limited settings to increase the yield of TB diagnoses.

Strengths of this study include consecutive enrollment of patients who were well characterized. Sputum induction was used to obtain quality samples, and liquid culture was used as the reference gold standard and processed in an accredited laboratory according to standardized protocols and quality assurance procedures. A limitation of this study includes the lack of available red blood cell indices or iron studies. With only hemoglobin levels available, we could only classify patients according to the degree of anemia without being able to further investigate possible underlying mechanisms, such as iron deficiency or chronic inflammation. An additional limitation is that it was not possible to determine whether anemia in patients with TB was directly related to their TB disease, attributable to their HIV infection, or was simply a prevalent comorbidity unrelated to either their TB or HIV disease. Finally, the reference standard was determined by testing paired sputum samples using liquid culture. As extrapulmonary TB is more common in patients with advanced immunodeficiency, this may have underestimated the prevalence of active disease. Sampling multiple sites of disease for extrapulmonary TB may have therefore enhanced the reference standard.

In conclusion, HIV-associated anemia was common, and there was a very high prevalence of undiagnosed TB among those with moderate or severe anemia that was associated with very poor clinical outcomes. Sputum Xpert MTB/RIF and urine Determine TB-LAM were able to rapidly diagnose TB with useful sensitivity among such patients. PLWH with moderate or severe anemia in high burden settings should be investigated for TB using rapid microbiological assays regardless of symptoms.

ACKNOWLEDGMENTS

The authors are grateful to the Foundation for Innovative New Diagnostics (FIND), Geneva, Switzerland, for providing access to the Xpert MTB/RIFassay cartridges with preferential pricing.