Skip to main content
  • Research article
  • Open access
  • Published:

The predictive value of current haemoglobin levels for incident tuberculosis and/or mortality during long-term antiretroviral therapy in South Africa: a cohort study



Low haemoglobin concentrations may be predictive of incident tuberculosis (TB) and death in HIV-infected patients receiving antiretroviral therapy (ART), but data are limited and inconsistent. We examined these relationships retrospectively in a long-term South African ART cohort with multiple time-updated haemoglobin measurements.


Prospectively collected clinical data on patients receiving ART for up to 8 years in a community-based cohort were analysed. Time-updated haemoglobin concentrations, CD4 counts and HIV viral loads were recorded, and TB diagnoses and deaths from all causes were ascertained. Anaemia severity was classified using World Health Organization criteria. TB incidence and mortality rates were calculated and Poisson regression models were used to identify independent predictors of incident TB and mortality, respectively.


During a median follow-up of 5.0 years (IQR, 2.5-5.8) of 1,521 patients, 476 cases of incident TB and 192 deaths occurred during 6,459 person-years (PYs) of follow-up. TB incidence rates were strongly associated with time-updated anaemia severity; those without anaemia had a rate of 4.4 (95%CI, 3.8-5.1) cases/100 PYs compared to 10.0 (95%CI, 8.3-12.1), 26.6 (95%CI, 22.5-31.7) and 87.8 (95%CI, 57.0-138.2) cases/100 PYs in those with mild, moderate and severe anaemia, respectively. Similarly, mortality rates in those with no anaemia or mild, moderate and severe time-updated anaemia were 1.1 (95%CI, 0.8-1.5), 3.5 (95%CI, 2.7-4.8), 11.8 (95%CI, 9.5-14.8) and 28.2 (95%CI, 16.5-51.5) cases/100 PYs, respectively. Moderate and severe anaemia (time-updated) during ART were the strongest independent predictors for incident TB (adjusted IRR = 3.8 [95%CI, 3.0-4.8] and 8.2 [95%CI, 5.3-12.7], respectively) and for mortality (adjusted IRR = 6.0 [95%CI, 3.9-9.2] and adjusted IRR = 8.0 [95%CI, 3.9-16.4], respectively).


Increasing severity of anaemia was associated with exceptionally high rates of both incident TB and mortality during long-term ART. Patients receiving ART who have moderate or severe anaemia should be prioritized for TB screening using microbiological assays and may require adjunctive clinical interventions.

Peer Review reports


Although antiretroviral therapy (ART) reduces tuberculosis (TB) incidence rates by approximately 70% in people living with HIV [1], high TB incidence rates continue to be observed among patients receiving ART in sub-Saharan Africa, especially during the initial months of treatment [2-9]. Moreover, in patients receiving ART long-term, TB incidence rates remain substantially higher than background rates, even in those with good immune recovery [2]. Incident TB also remains an important independent risk factor for mortality during short-term and long-term ART [4,10]. To improve clinical outcomes in ART programmes in sub-Saharan Africa, it is important to identify risk factors and/or predictive markers associated with incident TB and mortality so that effective interventions may be designed and implemented.

Measurement of haemoglobin concentrations is low-cost, more widely available than CD4 cell count measurement in clinical settings in sub-Saharan Africa, and may have predictive value for both active TB disease and all-cause mortality. We have previously demonstrated that greater degrees of anaemia severity at ART initiation are associated with a high prevalence of undiagnosed TB, as well as death during the initial months of ART [11]. While it remains poorly defined, we hypothesized that time-updated haemoglobin concentrations may be associated with incident TB and/or death during long-term ART. Therefore we undertook this retrospective cohort analysis among South African patients receiving ART for up to 8 years to determine TB incidence and mortality rates stratified by time-updated anaemia severity. We also sought to determine whether time-updated anaemia severity was an independent predictor for incident TB and/or mortality even after adjustment for time-updated CD4 counts.


Study setting and data collection

A retrospective cohort analysis was undertaken among patients recruited for previously reported studies [2,10] and was conducted as part of ongoing research at the Hannan Crusaid Antiretroviral Centre in Gugulethu township, Cape Town, South Africa [6,12,13]. All consecutively recruited ART-naïve adult patients (age ≥16 years) who enrolled in the ART programme and subsequently started ART were eligible for study inclusion. All patients provided written informed consent. This study was approved by the human ethics committee of the University of Cape Town, Cape Town, South Africa and was reported in conformity with the STROBE Statement checklist for cohort studies [14].

A total of 2,000 patients enrolled between September 2002 and May 2006 (an interval for which very complete clinical records and time-updated TB, CD4 count and viral load data were available) and were followed up until censoring of observations on 01 January 2011. During this period, patients were eligible to start ART if they had either a CD4 count <200 cells/uL or a World Health Organization (WHO) stage 4 disease, which was in accordance with national ART guidelines at that time. First-line ART comprised a non-nucleoside reverse transcriptase inhibitor (predominantly efavirenz) with a nucleoside reverse transcriptase inhibitor backbone (stavudine or zidovudine and lamivudine), while the second-line regimen was protease inhibitor-based (lopinavir/ritonavir) with a nucleoside reverse transcriptase inhibitor backbone (zidovudine and didanosine). Information on the ART regimen was attained from electronic pharmacy records. All patients received daily trimethoprim-sulphamethoxazole prophylaxis. However, in accordance with local guidelines and practice at the time, patients did not receive isoniazid preventive therapy (IPT). All treatment was provided free of charge.

Primary care clinics, TB clinics and antenatal clinics referred patients for ART, and upon a patient’s first visit to the clinic their demographic details were recorded. All patients received routine clinical review at a screening visit, ART initiation, after 4, 8 and 16 weeks of ART, and at least every 4 months thereafter. Haemoglobin concentrations as well as blood CD4 cell counts and plasma HIV viral load levels were measured routinely before starting ART, every 4 months during ART and as clinically indicated. Laboratory results were transferred on a monthly basis to an electronic database. All information related to TB diagnoses and treatment was obtained from prospectively maintained structured clinical records.

TB screening and diagnosis

All patients were screened for TB prior to starting ART using a multi-symptom screening questionnaire to identify patients who required further TB investigations. Several investigations were available to diagnose TB either pre-ART or during ART, including sputum induction, sputum smear fluorescence microscopy, automated liquid culture of sputum using mycobacterial growth indicator tubes (MGIT 960, Becton Dickinson, Sparks, Maryland, USA), chest radiology, ultrasonography and fine needle lymph node aspiration cytology. TB case ascertainment approaches were similar for the duration of the study period, and all follow-up was completed prior to the national scale-up of Xpert MTB/RIF assay in South Africa. All TB cases fulfilled the WHO diagnostic criteria for resource-constrained settings with high HIV prevalence. While the large majority of cases were culture-proven, any culture-negative TB diagnoses were defined by strong clinical or histological evidence of active TB, radiological abnormalities consistent with active TB and the clinician’s decision to treat with a full course of anti-TB therapy. All microbiological specimens were analysed in accredited laboratories using standardized methods, and all positive cultures of mycobacteria were speciated by polymerase chain reaction. TB treatment was rifampicin-based and administered via a local network of community clinics using the directly observed treatment short-course (DOTS) strategy.


Baseline and time-updated haemoglobin levels were used to classify anaemia severity according to WHO criteria [15]: no anaemia (haemoglobin concentration ≥13.0 g/dL for males, ≥12.0 g/dL for females), mild anaemia (11.0-12.9 g/dL for males, 11.0-11.9 g/dL for females), moderate anaemia (8.0-10.9 g/dL for males and females) or severe anaemia (<8.0 g/dL for males and females). ‘Baseline haemoglobin level’ was defined using a haemoglobin concentration within 90 days of starting ART for those without a known TB diagnosis prior to ART initiation (and currently receiving anti-TB therapy) and within 30 days for those with known TB who were currently receiving anti-TB therapy, as it is known to be associated with rapid changes in haemoglobin concentration [16-18].

A ‘past history of TB’ was defined by those who had previously been diagnosed with TB and had completed a full course of anti-TB therapy prior to study enrolment. ‘Prevalent TB’ was defined by any patient with a known TB diagnosis who was currently receiving anti-TB therapy at the time of ART initiation. Any new clinical episode of TB diagnosed after starting ART, irrespective of date of symptom onset, was defined as ‘incident TB.’ ‘Recurrent TB’ diagnoses were made with reference to any previous episodes of TB, including those made prior to ART enrolment (that is, past history of TB, prevalent TB or incident TB). Patients who had multiple incident TB diagnoses completed full courses of anti-TB therapy and had complete resolution of symptoms between episodes. Thus, patients undergoing retreatment after failure of initial treatment or those returning after treatment default did not count towards a case of incident TB or recurrent TB.

Clinical outcomes

We determined clinical outcomes of patients during routine follow-up within the ART service as previously described [19]. Patients’ outcomes were classified into one of four mutually exclusive groups: alive within the programme and receiving ART, transfer to another ART service, death from any cause, or lost to follow-up (LTFU). LTFU was defined by those receiving ART who were more than 12 weeks late for a scheduled clinic appointment or who failed to return to the ART programme, and could not be located by active community-based follow-up.

Data analysis

Data were analysed using Stata version 12.0 (College Station, TX). Analysis was restricted to those who had at least one haemoglobin result available from any point either prior to ART initiation or during ART. If either a haemoglobin level or a CD4 cell count was missing for 1 or more intervals, a mean of the values for the preceding and subsequent intervals was used (about 1.5% intervals). Chi-squared tests and Kruskall-Wallis tests were used to compare proportions and medians, respectively. Since we analysed existing data, we did not undertake formal sample size calculations. However, the net sample size of n = 1,521 provided a statistical power of close to 1.0 for most expected outcomes, and 0.88 for the extreme case of finding an only twofold difference (10 versus 5/100 person-years, PY) in TB incidence between equal groups with and without anaemia. All statistical tests were two-sided at α = 0.05.

Person-time was divided into intervals predefined by haemoglobin results at the start of each interval that were done at least 4-monthly and started accruing from the date of ART initiation and continued until death, transfer-out, LTFU, or data censorship on 01 January 2011. Haemoglobin levels and CD4 counts were coded as time-dependent covariates, and these values were updated at least every 4 months as a repeated measure. Only intervals with both a time-updated haemoglobin level and CD4 result counted towards person-time accrued. TB incidence and mortality rates with corresponding 95% confidence intervals stratified according to time-updated WHO anaemia severity and CD4 cell strata were calculated. This meant that a patient could be reclassified at each time-updated interval on the basis of their current anaemia severity or CD4 count and that a case of incident TB or death was only counted for that person-time observation if it occurred during that period. Person-time accrued during treatment of prevalent or incident TB was excluded when calculating TB incidence rates. Multivariable models to determine factors associated with incident TB and mortality during ART were constructed using Poisson regression. A gamma distributed random-effects model was used to adjust for repeated measurements in patients with incident TB. A priori covariates (time-updated anaemia severity and time-updated CD4 cell count) and variables meeting the predefined cutoff of P ≤ 0.1 in the univariable model were included in the multivariable model. Likelihood ratio tests were used to investigate interactions between variables and test statistical hypotheses. Sensitivity analyses were undertaken to determine if the relationship between time-updated anaemia severity and mortality rates (unadjusted and adjusted) was affected by patients LTFU where it was assumed that all patients LTFU had died.


From September 2002 to May 2006, 2,000 patients were enrolled into the ART service and 1,544 met inclusion criteria, although 23 patients were excluded due to lack of any haemoglobin measurements (Figure 1). Thus, 1,521 patients were included in the final analysis and accrued 6,459 person-years of follow-up in the ART service for a median of 5.0 years (IQR, 2.5-5.8; range 0.0-8.2). During ART, 192 (12.6%) patients died, 314 (20.6%) were LTFU, 205 (13.5%) transferred to a different ART service and 810 (53.3%) were alive at the time that data were censored.

Figure 1
figure 1

Study overview including diagnoses of incident TB and patient outcomes.

At baseline, patients had a median age of 33 years and more than two-thirds were female (Table 1). Advanced immunodeficiency was common; the median CD4 count was 101 cells/uL and more than 80% had stage 3 or 4 disease. The median haemoglobin was 11.2 g/dL (IQR, 9.8-12.5) and 72% had anaemia of which 44% had moderate or severe anaemia. Nearly half of all patients had a past medical history of TB upon enrolment and 30% were receiving anti-TB therapy at the time of ART initiation (Table 1). The baseline characteristics (including haemoglobin level and CD4 cell count) of those who were LTFU (n = 314) or transferred their care to another ART service (n = 205) did not differ from those who remained alive and in programme (n = 810), with the exception that patients LTFU tended to be slightly younger and have prevalent TB (Table 1).

Table 1 Baseline characteristics of patients stratified according to outcome (n = 1,521)

Incident TB diagnoses and TB incidence rates during ART

Overall, 476 episodes of incident TB were diagnosed in 423 (27.8%) patients during ART. Of these, 73.3% of episodes were culture-confirmed. After excluding 520.1 years of person-time accrued during TB treatment, the overall TB incidence rate during ART was 8.0 cases/100 PYs (95%CI, 7.3-8.8). TB incidence during the first 4 months of ART was more than fourfold higher (incidence rate ratio (IRR) = 4.1 [95%CI, 3.2-5.2]) than the rate during the remainder of ART follow-up (Table 2). Analysis restricted to culture-proven TB cases, yielded a slightly lower overall TB incidence rate (5.9 [95%, 5.3-6.6] cases/100 PYs).

Table 2 TB incidence rates during ART stratified according to time-updated anaemia severity and CD4 cell count

Relationship between TB incidence rates and time-updated anaemia severity

TB incidence rates were strongly and directly associated with time-updated anaemia severity (Figure 2A). Overall incidence rates ranged from 4.4 to 87.8 cases/100 PYs (20-fold difference) among those with no anaemia and severe anaemia, respectively (Table 2). TB incidence rates were 2.5-fold higher in patients with severe anaemia compared to incidence rates in patients with CD4 counts <100 cells/uL (Table 2). Although TB incidence rates were higher during the first 4 months of ART, rates were strongly and directly related to time-updated anaemia severity, regardless of ART duration (Table 2). The strong and direct association between TB incidence rates and time-updated anaemia severity was unchanged when analysis was restricted to culture-confirmed cases (data not shown).

Figure 2
figure 2

Overall (A) TB incidence rates and (B) mortality rates during ART stratified according to time-updated WHO anaemia severity (cases/100 PYs). The endpoints represent the corresponding 95% confidence intervals. The red, dashed line represents the overall rate among all patients (n = 1,521).

Independent predictors for incident TB during ART

We next sought to determine independent predictors for incident TB. In univariable analysis, several factors including time-updated anaemia severity were associated with incident TB during ART (Table 3). In multivariable analysis, time-updated moderate anaemia (adjusted IRR = 3.8 [95%CI, 3.0-4.8]) and severe anaemia (adjusted IRR = 8.2 [5.3-12.7]) were the strongest independent predictors of incident TB, although increasing age, time-updated CD4 cell counts, HIV viral load and a past medical history of TB were also independently associated with incident TB. When time-updated haemoglobin levels were coded as a continuous variable in the same multivariable model, they remained strongly predictive for incident TB (adjusted IRR = 1.29 [95%CI, 1.23-1.36] for each 1 g/dL decrease). Substituting baseline anaemia severity in place of time-updated anaemia severity in the above model revealed that baseline anaemia severity was not associated with incident TB during ART after correcting for potential confounders (data not shown).

Table 3 Poisson regression of risk factors for incident TB among patients receiving ART

Relationship between mortality and time-updated anaemia severity during ART

Overall, 192 (12.6%) patients died a median of 0.8 year after starting ART (IQR, 0.2-2.6). The overall mortality rate among patients receiving ART was 3.0 (95%CI, 2.6-3.5) cases/100 PYs. One-third (n = 64) of all deaths occurred in the first 4 months of ART, yielding a mortality rate of 14.0 (95%CI, 11.0-18.1) cases/100 PYs compared to 2.1 (95%CI, 1.8-2.6) cases/100 PYs thereafter (IRR = 6.5 [95%CI, 4.8-8.8]).

Mortality rates during ART were strongly associated with greater degrees of time-updated anaemia severity (Figure 2B), and those with severe anaemia had higher mortality rates throughout ART compared to those with advanced immunodeficiency (Table 4). Similar to TB incidence, mortality rates were higher in the first 4 months of ART; however, higher mortality rates were associated with greater anaemia severity regardless of ART duration (Table 4). In a sensitivity analysis, when all LTFU patients (n = 314) were presumed to have died (n = 506 total), the overall mortality rate was 7.8 (95%CI, 7.2-8.6) and the graded association with time-updated anaemia severity remained. Patients survived a median of 49 days (IQR, 26-88) from their last laboratory measurement to death, and this did not differ by time-updated anaemia severity or CD4 counts (data not shown).

Table 4 Mortality rates stratified according to time-updated anaemia severity and CD4 cell count

Mortality rates following episode of incident TB

Of 192 total deaths, 34 (17.7%) occurred within 6 months of an incident TB diagnosis, yielding a mortality rate of 20.7 cases/100 PYs (95%CI, 14.8-29.7). Mortality rates within 6 months of an incident TB diagnosis were strongly associated with time-updated anaemia severity (Table 4).

Independent predictors for mortality during ART

Multivariable regression was undertaken to determine factors associated with mortality during ART (Table 5). Univariable analysis demonstrated several associations with mortality, including time-updated anaemia severity and incident TB. In multivariable analysis, several variables demonstrated independent associations with mortality during ART, including: increasing age, male gender, WHO disease stage at baseline, time-updated CD4 cell count, time-updated anaemia severity, time-updated viral load, prevalent TB at ART initiation and time-updated incident TB. However, the strongest independent predictor of mortality was time-updated severe anaemia (adjusted IRR = 8.0 [95%CI, 3.9-16.4]), while time-updated moderate anaemia (adjusted IRR = 6.0 [95%CI, 3.9-9.2]) and CD4 counts <100 cells/uL (adjusted IRR = 6.5 [95%CI, 2.8-15.0]) were also very strong independent predictors. Time-updated haemoglobin levels retained a strong relationship with mortality risk when coded as a continuous variable (adjusted IRR = 1.49 [95%CI, 1.36-1.63] for each 1 g/dL decrease). When a combined outcome of either mortality or LTFU (n = 506) was used, time-updated severe anaemia remained the strongest independent predictor (adjusted IRR = 4.2 [95%CI, 2.6-6.7]) and time-updated moderate anaemia (adjusted IRR = 2.5 [95%CI, 1.9-3.2]) and time-updated CD4 counts <100 cells/uL (adjusted IRR = 2.5 [95%CI, 1.7-3.8]) were also strong independent predictors, while time-updated HIV viral load, male gender, prevalent TB and incident TB also demonstrated smaller independent associations (data not shown).

Table 5 Poisson regression of risk factors for mortality among patients receiving ART


In this prospective cohort of HIV-infected patients receiving ART for up to 8 years in South Africa, haemoglobin levels had high predictive value for incident TB and death. Both TB incidence rates and mortality rates were strongly and directly associated with time-updated anaemia severity, regardless of ART duration. We found that time-updated moderate or severe anaemia during ART were the strongest independent predictors for both incident TB and death during ART and, notably, were of greater predictive value than CD4 cell counts.

Patients with a recent haemoglobin measurement (within at least the past 4 months) indicating either moderate or severe anaemia during ART were at very high risk for incident TB, in agreement with previous studies from South Africa and Cambodia [3,20]. Additionally, it has been found that anaemia prior to ART initiation is highly predictive for undiagnosed prevalent TB or early incident TB during ART [3,11,21]. However, because haemoglobin levels are dynamic and can change quickly, this may explain why this and other studies have found that baseline anaemia prior to or at ART initiation has poor predictive value for incident TB during ongoing ART [22,23]. Therefore, only current WHO anaemia severity appears to have predictive value for incident TB disease during ART.

TB incidence rates during ART were so high and the independent associations so strong, that routine microbiological investigation for TB among those found to have moderate/severe anaemia (haemoglobin concentration <11.0 g/dL) during ART in high incidence areas should be considered. Additionally, mortality rates within 6 months of an incident TB diagnosis were exceedingly high among those with recent moderate/severe anaemia. Thus, routine screening for TB among those with moderate/anaemia may help reduce mortality by expediting initiation of appropriate treatment. Recent studies have demonstrated that the sensitivities of Xpert MTB/RIF assay on sputum and urine, sputum AFB microscopy and urine lipoarabinomannan (LAM) lateral-flow assay for the diagnosis of HIV-associated TB are significantly greater among those with moderate/severe anaemia compared to those with no/mild anaemia [11,24,25]. When TB is confirmed among those with HIV-related anaemia, it is reassuring that most can expect haemoglobin recovery with combined ART and anti-TB treatment without specific additional interventions for anaemia [26].

While strong, independent relationships between anaemia and incident TB were found, the biology underpinning these relationships is complex. Haemoglobin levels were measured at regular intervals throughout ART (as opposed to TB investigations, which were undertaken when clinically indicated), and thus more severe anaemia was likely in large part due to yet undiagnosed active TB, resulting in pro-inflammatory cytokine up-regulation and subsequent anaemia of chronic disease. However, emerging evidence suggests that the dysregulated iron axis in HIV-related anaemia may provide highly favourable intracellular macrophage conditions for TB bacilli and the subsequent development of TB disease [27,28]. Thus, while anaemia during ART most likely represents a downstream consequence of active TB, we cannot exclude the possibility that it may also represent an upstream risk factor for active TB.

Mortality rates during ART were strongly associated with time-updated anaemia severity, and moderate or severe anaemia were the strongest independent predictors of death during ART. This is consistent with other studies from Africa that show that anaemia either before or during ART has high prognostic value [29-38] and suggests a need to develop standardized guidelines for the management of HIV-related anaemia to help reduce associated mortality. ART should be considered a primary intervention for HIV-related anemia, as it may resolve in anaemia in a majority of patients [26,39-43]. Additionally, management of those with anaemia during ART might include a greater frequency of follow-up in care, especially during the first 3-6 months of ART (given very high mortality rates), to allow for assessment of the patient’s current health and need for adjunctive interventions. Anaemia may serve as a crude marker for virological failure [44-46], and thus requesting an HIV viral load measurement may be warranted to determine if a patient is responding appropriately to ART. In addition to routine screening for TB and initiation of anti-TB therapy as described above, additional interventions a clinician may consider for those with anaemia during ART might include: treatment of comorbid infections that may contribute to anaemia (parasites, opportunistic infections), switching or stopping drugs that may be associated with anaemia (zidovudine, trimethoprim-sulphamethoxazole, among others), correction of nutritional deficiencies (iron, folate, vitamin B12), and investigations for AIDS-related malignancies. Of note, most patients survived for more than 45 days from the date of their final haemoglobin measurement to death, suggesting that there is a sufficient window of time to investigate the causes underlying HIV-related anaemia during ART and to implement appropriate interventions that may impact mortality risk.

Given the high predictive value of anaemia for both incident TB and mortality, this study demonstrates the potential utility of regular haemoglobin measurements both prior to ART initiation and at regular intervals during treatment. While automated haematologic analysers are the gold standard for determining haemoglobin concentrations, they are not routinely available for use in resource-limited settings. The WHO Haemoglobin Colour Scale is a simple, paper-based assay costing less than 15 cents per test [47], but the test has yielded mixed results in several studies, with overall moderate sensitivity and generally poor specificity [48]. The HemoCue system has been validated in resource-limited settings, is durable, easy to use and has shown good accuracy overall [49-51], but its cost and requirement for ongoing proprietary supplies may make its use inappropriate in some settings. In the near future, tests utilizing transcutaneous spectrophotometry may offer an inexpensive solution that allows for the routine measurement of haemoglobin levels in HIV-infected patients without a requirement for blood samples [52-54]. As methods for measuring haemoglobin become cheaper, more widely available and possibly less invasive, more frequent measurements during ART may be warranted and feasible.

This is the first study to report TB incidence and mortality rates during ART stratified according to time-updated WHO anaemia severity. Additional strengths include a large, consecutively recruited cohort followed up for a median of more than 5 years under routine programme conditions. While this study was conducted at a single site, it was a very large community-based cohort with comparable characteristics to other public sector ART cohorts in sub-Saharan Africa [7-9,55] and spanned a period of 8 years. Additionally, risk of selection bias was low due to limited inclusion/exclusion criteria and only about 1% of patients were excluded for not having haemoglobin results available. Our results are therefore likely generalizable to other ambulatory ART programmes in sub-Saharan Africa where there is a high TB burden.

There was a female predominance in our cohort; however, this likely reflects the epidemiology of HIV in sub-Saharan Africa, where the epidemic is well recognized as being feminized [56]. The proportion of female patients is entirely consistent with ART programmes throughout the region and would not be expected to affect the generalizability of our findings. Unfortunately, data were not available on pregnancy status, which may have led to a small minority of women having the severity of their anaemia misclassified during ART.

Also, we did not analyse the time lag between time-updated haemoglobin measurements and the diagnosis of incident TB; however, the vast majority of haemoglobin measurements were done at least 4-monthly. There was also a proportion of patients who were LTFU and for whom outcomes could not be ascertained, although this proportion is similar to other long-term ART cohorts in South Africa [3,57]. LTFU patient characteristics were very similar to those who remained alive throughout follow-up, and while LTFU patients may have a high mortality risk [57,58], this did not appear to apply to the present study. Furthermore, when such patients were presumed to have died, time-updated moderate or severe anaemia remained the strongest independent predictors of a poor clinical outcome.


Time-updated anaemia severity had very high predictive value for both incident TB and mortality during ART. Haemoglobin levels should be determined prior to ART initiation and at routine intervals during ART. Moderate or severe anaemia during ART should prompt systematic investigation of TB using microbiological assays as well as adjunctive clinical interventions given the very high risk of death among such patients.


  1. Suthar AB, Lawn SD, del Amo J, Getahun H, Dye C, Sculier D, et al. Antiretroviral therapy for prevention of tuberculosis in adults with HIV: a systematic review and meta-analysis. PLoS Med. 2012;9:e1001270.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  2. Gupta A, Wood R, Kaplan R, Bekker L-G, Lawn SD. Tuberculosis incidence rates during 8 years of follow-up of an antiretroviral treatment cohort in South Africa: comparison with rates in the community. PLoS One. 2012;7:e34156.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  3. van Rie A, Westreich D, Sanne I. Tuberculosis in patients receiving antiretroviral treatment: incidence, risk factors, and prevention strategies. J Acquir Immune Defic Syndr. 2011;56:349–55.

    Article  PubMed  PubMed Central  Google Scholar 

  4. Komati S, Shaw PA, Stubbs N, Mathibedi MJ, Malan L, Sangweni P, et al. Tuberculosis risk factors and mortality for HIV-infected persons receiving antiretroviral therapy in South Africa. AIDS. 2010;24:1849–55.

    Article  PubMed  PubMed Central  Google Scholar 

  5. Moore D, Liechty C, Ekwaru P, Were W, Mwima G, Solberg P, et al. Prevalence, incidence and mortality associated with tuberculosis in HIV-infected patients initiating antiretroviral therapy in rural Uganda. AIDS. 2007;21:713–9.

    Article  PubMed  Google Scholar 

  6. Lawn SD, Myer L, Bekker L-G, Wood R. Burden of tuberculosis in an antiretroviral treatment programme in sub-Saharan Africa: impact on treatment outcomes and implications for tuberculosis control. AIDS. 2006;20:1605–12.

    Article  PubMed  Google Scholar 

  7. Houlihan CF, Mutevedzi PC, Lessells RJ, Cooke GS, Tanser FC, Newell M-L. The tuberculosis challenge in a rural South African HIV programme. BMC Infect Dis. 2010;10:23.

    Article  PubMed  PubMed Central  Google Scholar 

  8. Hermans SM, Kiragga AN, Schaefer P, Kambugu A, Hoepelman AIM, Manabe YC. Incident tuberculosis during antiretroviral therapy contributes to suboptimal immune reconstitution in a large urban HIV clinic in sub-Saharan Africa. PLoS One. 2010;5:e10527.

    Article  PubMed  PubMed Central  Google Scholar 

  9. Dembélé M, Saleri N, Carvalho ACC, Saouadogo T, Hien AD, Zabsonre I, et al. Incidence of tuberculosis after HAART initiation in a cohort of HIV-positive patients in Burkina Faso. Int J Tuberc Lung Dis. 2010;14:318–23.

    PubMed  Google Scholar 

  10. Gupta A, Wood R, Kaplan R, Bekker L-G, Lawn SD. Prevalent and incident tuberculosis are independent risk factors for mortality among patients accessing antiretroviral therapy in South Africa. PLoS One. 2013;8:e55824.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  11. Kerkhoff AD, Wood R, Vogt M, Lawn SD. Predictive value of anemia for tuberculosis in HIV-infected patients in sub-Saharan Africa: an indication for routine microbiological investigation using new rapid assays. J Acquir Immune Defic Syndr. 2014;66:33–40.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  12. Bekker L-G, Myer L, Orrell C, Lawn S, Wood R. Rapid scale-up of a community-based HIV treatment service: programme performance over 3 consecutive years in Guguletu, South Africa. S Afr Med J. 2006;96:315–20.

    PubMed  Google Scholar 

  13. Lawn SD, Myer L, Orrell C, Bekker L-G, Wood R. Early mortality among adults accessing a community-based antiretroviral service in South Africa: implications for programme design. AIDS. 2005;19:2141–8.

    Article  PubMed  Google Scholar 

  14. von Elm E, Altman DG, Egger M, Pocock SJ, Gøtzsche PC, Vandenbroucke JP. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. Lancet. 2007;370:1453–7.

    Article  Google Scholar 

  15. World Health Organization. Haemoglobin concentrations for the diagnosis of anaemia and assessment of severity. Vitamin and mineral nutrition information system. Geneva, Switzerland: World Health Organization; 2011. (WHO/NMH/NHD/MNM/11.1). 2011

  16. Lawn SD, Obeng J, Acheampong JW, Griffin GE. Resolution of the acute-phase response in West African patients receiving treatment for pulmonary tuberculosis. Int J Tuberc Lung Dis. 2000;4:340–4.

    CAS  PubMed  Google Scholar 

  17. Lee JH, Chang JH. Changes of plasma interleukin-1 receptor antagonist, interleukin-8 and other serologic markers during chemotherapy in patients with active pulmonary tuberculosis. Korean J Intern Med. 2003;18:138–45.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  18. Friedland JS, Hartley JC, Hartley CG, Shattock RJ, Griffin GE. Cytokine secretion in vivo and ex vivo following chemotherapy of Mycobacterium tuberculosis infection. Trans R Soc Trop Med Hyg. 1996;90:199–203.

    Article  CAS  PubMed  Google Scholar 

  19. Lawn SD, Myer L, Harling G, Orrell C, Bekker L-G, Wood R. Determinants of mortality and nondeath losses from an antiretroviral treatment service in South Africa: implications for program evaluation. Clin Infect Dis. 2006;43:770–6.

    Article  PubMed  Google Scholar 

  20. Choun K, Thai S, Pe R, Lorent N, Lynen L, van Griensven J. Incidence and risk factors for tuberculosis in HIV-infected patients while on antiretroviral treatment in Cambodia. Trans R Soc Trop Med Hyg. 2013;107:235–42.

    Article  CAS  PubMed  Google Scholar 

  21. Lewis DK, Whitty CJM, Walsh AL, Epino H, Broek NRVD, Letsky EA, et al. Treatable factors associated with severe anaemia in adults admitted to medical wards in Blantyre, Malawi, an area of high HIV seroprevalence. Trans R Soc Trop Med Hyg. 2005;99:561–7.

    Article  PubMed  Google Scholar 

  22. Auld AF, Mbofana F, Shiraishi RW, Alfredo C, Sanchez M, Ellerbrock TV, et al. Incidence and determinants of tuberculosis among adults initiating antiretroviral therapy – Mozambique, 2004-2008. PLoS One. 2013;8:e54665.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  23. Seyler C, Toure S, Messou E, Bonard D, Gabillard D, Anglaret X. Risk factors for active tuberculosis after antiretroviral treatment initiation in Abidjan. Am J Respir Crit Care Med. 2005;172:123–7.

    Article  PubMed  Google Scholar 

  24. Lawn SD, Kerkhoff AD, Vogt M, Wood R. HIV-associated tuberculosis: relationship between disease severity and the sensitivity of new sputum-based and urine-based diagnostic assays. BMC Med. 2013;11:231.

    Article  PubMed  PubMed Central  Google Scholar 

  25. Kerkhoff AD, Meintjes G, Burton R, Nicol MP, Lawn SD. Rapid urine-based diagnosis of tuberculosis among HIV-infected hospital in-patients with moderate or severe anemia and poor prognosis. Barcelona, Spain: Union World Conference on Lung Health; 2014. Oral abstract OPA-209-30.

  26. Kerkhoff AD, Wood R, Cobelens FG, Gupta-Wright A, Bekker L-G, Lawn SD. Resolution of anaemia in a cohort of HIV-infected patients with a high prevalence and incidence of tuberculosis receiving antiretroviral therapy in South Africa. BMC Infect Dis. 2014;14:702.

    Article  Google Scholar 

  27. McDermid JM, Hennig BJ, van der Sande M, Hill AVS, Whittle HC, Jaye A, et al. Host iron redistribution as a risk factor for incident tuberculosis in HIV infection: an 11-year retrospective cohort study. BMC Infect Dis. 2013;13:48.

    Article  PubMed  PubMed Central  Google Scholar 

  28. Drakesmith H, Prentice AM. Hepcidin and the iron-infection axis. Science. 2012;338:768–72.

    Article  CAS  PubMed  Google Scholar 

  29. Toure S, Kouadio B, Seyler C, Traore M. Rapid scaling-up of antiretroviral therapy in 10,000 adults in Cote d’Ivoire: two-year outcomes and determinants. AIDS. 2008;22:873–82.

    Article  PubMed  PubMed Central  Google Scholar 

  30. O’Brien ME, Kupka R, Msamanga GI, Saathoff E, Hunter DJ, Fawzi WW. Anemia is an independent predictor of mortality and immunologic progression of disease among women with HIV in Tanzania. J Acquir Immune Defic Syndr. 2005;40:219–25.

    Article  PubMed  Google Scholar 

  31. Sieleunou I, Souleymanou M, Schönenberger A-M, Menten J, Boelaert M. Determinants of survival in AIDS patients on antiretroviral therapy in a rural centre in the Far-North Province, Cameroon. Trop Med Int Health. 2009;14:36–43.

    Article  PubMed  Google Scholar 

  32. Gupta A, Nadkarni G, Yang W-T, Chandrasekhar A, Gupte N, Bisson GP, et al. Early mortality in adults initiating antiretroviral therapy (ART) in low- and middle-income countries (LMIC): a systematic review and meta-analysis. PLoS One. 2011;6:e28691.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  33. Evans D, Menezes C, Mahomed K, Macdonald P, Untiedt S, Levin L, et al. Treatment outcomes of HIV-infected adolescents attending public-sector HIV clinics across Gauteng and Mpumalanga, South Africa. AIDS Res Hum Retroviruses. 2013;29:892–900.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  34. Giganti MJ, Limbada M, Mwango A, Moyo C, Mulenga LB, Guffey MB, et al. Six-month hemoglobin concentration and its association with subsequent mortality among adults on antiretroviral therapy in Lusaka, Zambia. J Acquir Immune Defic Syndr. 2012;61:120–3.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  35. Bussmann H, Wester CW, Ndwapi N, Grundmann N, Gaolathe T, Puvimanasinghe J, et al. Five-year outcomes of initial patients treated in Botswana’s National Antiretroviral Treatment Program. AIDS. 2008;22:2303–11.

    Article  PubMed  PubMed Central  Google Scholar 

  36. Etard J-F, Ndiaye I, Thierry-Mieg M, Guèye NFN, Guèye PM, Lanièce I, et al. Mortality and causes of death in adults receiving highly active antiretroviral therapy in Senegal: a 7-year cohort study. AIDS. 2006;20:1181–9.

    Article  PubMed  Google Scholar 

  37. Erikstrup C, Kallestrup P, Zinyama R, Gomo E, Mudenge B, Gerstoft J, et al. Predictors of mortality in a cohort of HIV-1-infected adults in rural Africa. J Acquir Immune Defic Syndr. 2007;44:478–83.

    Article  PubMed  Google Scholar 

  38. Lawn SD, Harries AD, Anglaret X, Myer L, Wood R. Early mortality among adults accessing antiretroviral treatment programmes in sub-Saharan Africa. AIDS. 2008;22:1897–908.

    Article  PubMed  Google Scholar 

  39. Johannessen A, Naman E, Gundersen SG, Bruun JN. Antiretroviral treatment reverses HIV-associated anemia in rural Tanzania. BMC Infect Dis. 2011;11:190.

    Article  PubMed  PubMed Central  Google Scholar 

  40. Moore RD, Keruly JC, Chaisson RE. Anemia and survival in HIV infection. J Acquir Immune Defic Syndr. 1998;19:29–33.

    Article  CAS  Google Scholar 

  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. 2001;26:315–9.

    Article  CAS  PubMed  Google Scholar 

  42. Sullivan PS, Hanson DL, Brooks JT. Impact on hemoglobin of starting combination antiretroviral therapy with or without zidovudine in anemic HIV-infected patients. J Acquir Immune Defic Syndr. 2008;48:163–8.

    Article  CAS  PubMed  Google Scholar 

  43. Berhane K, Karim R, Cohen MH, Masri-Lavine L, Young M, Anastos K, et al. Impact of highly active antiretroviral therapy on anemia and relationship between anemia and survival in a large cohort of HIV-infected women: Women’s Interagency HIV Study. J Acquir Immune Defic Syndr. 2004;37:1245–52.

    Article  PubMed  Google Scholar 

  44. Meriki HD, Tufon KA, Afegenwi MH, Nyindem BA, Atanga PN, Anong DN, et al. Immuno-haematologic and virologic responses and predictors of virologic failure in HIV-1 infected adults on first-line antiretroviral therapy in Cameroon. Infect Dis Poverty. 2014;3:5.

    Article  PubMed  PubMed Central  Google Scholar 

  45. Labhardt ND, Lejone T, Setoko M, Poka M, Ehmer J, Pfeiffer K, et al. A clinical prediction score in addition to WHO criteria for anti-retroviral treatment failure in resource-limited settings - experience from Lesotho. PLoS One. 2012;7:e47937.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  46. Anude CJ, Eze E, Onyegbutulem HC, Charurat M, Etiebet M-A, Ajayi S, et al. Immuno-virologic outcomes and immuno-virologic discordance among adults alive and on anti-retroviral therapy at 12 months in Nigeria. BMC Infect Dis. 2013;13:113.

    Article  PubMed  PubMed Central  Google Scholar 

  47. World Health Organization. Haemoglobin colour scale: a practical answer to a vital need. Geneva, Switzerland: World Health Organization; 2001. p. 1–6.

  48. Critchley J, Bates I. Haemoglobin colour scale for anaemia diagnosis where there is no laboratory: a systematic review. Int J Epidemiol. 2005;34:1425–34.

    Article  PubMed  Google Scholar 

  49. Nkrumah B, Nguah SB, Sarpong N, Dekker D, Idriss A, May J, et al. Hemoglobin estimation by the HemoCue® portable hemoglobin photometer in a resource poor setting. BMC Clin Pathol. 2011;11:5.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  50. Medina Lara A, Mundy C, Kandulu J, Chisuwo L, Bates I. Evaluation and costs of different haemoglobin methods for use in district hospitals in Malawi. J Clin Pathol. 2005;58:56–60.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  51. Sanchis-Gomar F, Cortell-Ballester J, Pareja-Galeano H, Banfi G, Lippi G. Hemoglobin point-of-care testing: the HemoCue system. J Lab Autom. 2013;18:198–205.

    Article  PubMed  Google Scholar 

  52. Gayat E, Aulagnier J, Matthieu E, Boisson M, Fischler M. Non-invasive measurement of hemoglobin: assessment of two different point-of-care technologies. PLoS One. 2012;7:e30065.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  53. Crowley C, Montenegro-Bethancourt G, Solomons NW, Schűmann K. Validity and correspondence of non-invasively determined hemoglobin concentrations by two trans-cutaneous digital measuring devices. Asia Pac J Clin Nutr. 2012;21:191–200.

    CAS  PubMed  Google Scholar 

  54. PATH. Noninvasive technology for anemia detection. Seattle, WA, USA: PATH; 2013. p. 1.

  55. Egger M, Spycher BD, Sidle J, Weigel R, Geng EH, Fox MP, et al. West Africa and Southern Africa: correcting mortality for loss to follow-up: a nomogram applied to antiretroviral treatment programmes in sub-Saharan Africa. PLoS Med. 2011;8:e1000390.

    Article  PubMed  PubMed Central  Google Scholar 

  56. UNAIDS. World AIDS Day 2014 Report - Fact sheet. Geneva, Switzerland: UNAIDS; 2014. p. 1–7.

  57. Fox MP, Brennan A, Maskew M, MacPhail P, Sanne I. Using vital registration data to update mortality among patients lost to follow-up from ART programmes: evidence from the Themba Lethu Clinic, South Africa. Trop Med Int Health. 2010;15:405–13.

    Article  PubMed  PubMed Central  Google Scholar 

  58. Nglazi MD, Kaplan R, Wood R, Bekker L-G, Lawn SD. Identification of losses to follow-up in a community-based antiretroviral therapy clinic in South Africa using a computerized pharmacy tracking system. BMC Infect Dis. 2010;10:329.

    Article  PubMed  PubMed Central  Google Scholar 

Download references


SDL is funded by the Wellcome Trust, London, UK. RW is funded in part by the National Institutes of Health (NIH) through grants RO1 A1058736-01A1 and 5UO1A1069519-02. None of these sources played any role in the design, conduct, analysis, interpretation or decision to publish these data.

Role of the funding source

The study was funded by a Wellcome Trust grant held by SDL (grant number 088590). The funders had no role in the study design, data collection, data analysis, data interpretation or writing of this report or previous reports from this cohort. The opinions, findings and conclusions expressed in this manuscript reflect those of the authors alone.

Author information

Authors and Affiliations


Corresponding author

Correspondence to Andrew D Kerkhoff.

Additional information

Competing interests

The authors declare that they have no competing interests.

Authors’ contributions

ADK and SDL initiated and planned the study. RW, AGW, SDL and ADK collected the data. LGB and RW were responsible for the study cohort. ADK did the data analysis with input from AGW, SDL and FGC. ADK and SDL wrote the paper with input from RW, FGC, LGB and AGW. All authors approved the final version of the manuscript prior to submission.

Rights and permissions

This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited. The Creative Commons Public Domain Dedication waiver ( applies to the data made available in this article, unless otherwise stated.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Kerkhoff, A.D., Wood, R., Cobelens, F.G. et al. The predictive value of current haemoglobin levels for incident tuberculosis and/or mortality during long-term antiretroviral therapy in South Africa: a cohort study. BMC Med 13, 70 (2015).

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: