Impact of Persistent Anemia on Systemic Inflammation and Tuberculosis Outcomes in Persons Living With HIV

Abstract

Tuberculosis (TB) is associated with systemic inflammation and anemia, which are aggravated in persons living with HIV (PLWH). Here, we characterized the dynamics of hemoglobin levels in PLWH coinfected with TB undergoing antitubercular therapy (ATT). We also examined the relationships between anemia and systemic inflammatory disturbance as well as the association between persistent anemia and unfavorable clinical outcomes. Data on several blood biochemical parameters and on blood cell counts were retrospectively analyzed in a cohort of 256 TB/HIV patients from Brazil during 180 days of ATT. Multidimensional statistical analyses were employed to profile systemic inflammation of patients stratified by anemia status (hemoglobin levels <12 g/dL for female and <13.5 g/dL for male individuals) prior to treatment and to perform prediction of unfavorable outcomes, such as treatment failure, loss to follow up and death. We found that 101 (63.63%) of patients with anemia at pre-ATT persisted with such condition until day 180. Such individuals exhibited heightened degree of inflammatory perturbation (DIP), which in turn was inversely correlated with hemoglobin levels. Recovery from anemia was associated with increased pre-ATT albumin levels whereas persistent anemia was related to higher total protein levels in serum. Multivariable regression analysis revealed that lower baseline hemoglobin levels was the major determinant of the unfavorable outcomes. Our findings demonstrate that persistent anemia in PLWH during the course of ATT is closely related with chronic inflammatory perturbation. Early intervention to promote recovery from anemia may improve ATT outcomes.

Keywords: HIV; anemia; inflammation; treatment outcome; tuberculosis.

FIGURE 1

Differential change in biomarkers between anemic and non-anemic patients. A Heatmap was designed to depict the overall pattern of complete blood counts (CBC) (A) and biochemical markers (B) at all timepoints in anemic and non-anemic at different study timepoints of anti-tubercular treatment. A two-way hierarchical cluster analysis (Ward’s method) was performed. Dendrograms represent Euclidean distance. Expression scale represents Z-score normalization from the median at each timepoint and group. To define anemia according to baseline hemoglobin, the cut-off point of 12 g/dL for women and 13.5 g/dL for men was used. A log₁₀ of fold-change was calculated and statistical analyses were performed using the Mann–Whitney U adjusted test. Significative differences (p < 0.05) between anemic and non-anemic patients for each time point are highlighted in red bars. Data are from 191 patients who had complete information on cell counts and biochemical measurements at all study timepoints.

FIGURE 2

Spearman correlation analysis of cells and biochemical parameters versus hemoglobin in blood of TB-HIV prior to antitubercular treatment initiation. Plots from statistically significant Spearman correlations between biochemical parameters and hemoglobin levels at study baseline (pre-ATT) are shown (n = 256). To define anemia according to baseline hemoglobin, the cut-off point of 12 g/dL for women and 13.5 g/dL for men was used. ALT, Alanine Aminotransferase; AST, Aspartate Aminotransferase.

FIGURE 3

The majority of the anemic patients at baseline persist with low levels of hemoglobin after initiation of anti-tubercular treatment. To define anemia according to baseline hemoglobin, the cut-off point of 12 g/dL for women and 13.5 g/dL for men was used. (A) Hemoglobin levels at different time points of antitubercular therapy in the longitudinal population (n = 191) as well as in the groups of patients with or without anemia at baseline of treatment are shown. Anemic group presented statistically significant difference with p < 0.001 between baseline and timepoints after 2 months (D60 and D180) using Wilcoxon rank sum test with corrections. On Jonckheere-Terpstra permutation test, where an increase in one variable results in an increase or decrease in another variable, both groups presented p = 0.001 to “increase” hypothesis, with number of permutation equal to 1000. Using Jonckheere-Terpstra asymptotic test, p-value of non-anemic group was 0.851 and p-value of anemic group was <2.2e–16. Green dots represent non-anemic TB patients at baseline and purple dots represent anemic TB patients at baseline. (B) To define a patient as recovered from anemia, were considered normal levels (above the cut-off) of hemoglobin in any time point after D0. Chi-square test comparing D0 and D180 in both groups returned p < 0.00001. Green bars represent non-anemic TB patients at baseline and purple bars represent anemic TB patients at baseline. (C) Of the 161 patients who had anemia before starting treatment, 37.3% (n = 60) increased the values to normal hemoglobin levels at some time point. Of these, 95% were completely recovered (n = 55), so that 35% (n = 19) were recovered early (D60), and 60% (n = 36) were recovered late (D180). Finally, 5% (n = 5) of the patients who were anemic at study baseline presented a transient recovery (recovered at day 60 but were once again anemic at D180). 36.6% (n = 11) of the 30 patients without anemia at baseline developed anemia in D60, but three of them recovered normal hemoglobin values at D180. (D) Hemoglobin levels at different time points of antitubercular therapy in the in the population of anemic patients at baseline, divided according to the time of recovery. Using Jonckheere-Terpstra asymptotic test and Wilcoxon rank sum test with corrections, only the transient recovery group (that showed higher levels of hemoglobin at time 60 but had anemia at 180) did not exhibit a significant p-value between the timepoints. (E) Logistic binomial regression model was used to test independent associations between biochemical and clinical parameters and total recovery from anemia status at baseline, early recovery (recovery from anemia in ≤60 days from baseline) or late recovery (recovery from anemia in >60 days from baseline). The condition persistent anemia (anemia from baseline to day 180) was used as reference to test associations. Only parameters which remained with p ≤ 0.2 in univariate analysis (Supplementary Table 2 for details) model were inputted in the adjusted model. (95%CI, 95% confidence interval). Associations reported in (E) are for increases in 1 unit in plasma concentrations of the indicated markers. Data are from 191 patients who had complete information on cell counts and biochemical measurements at all study timepoints.

FIGURE 4

Patients who persists with anemia exhibit substantially higher degree of inflammatory perturbation. (A) Degree of inflammatory perturbation (DIP) is based on Molecular Degree of Perturbation (16), but instead of using gene expression, we use biochemical and cellular markers. DIP was calculated was calculated using the median and standard deviation of the control group as a starting point. Then, the Z-score was calculated for all groups, a cut-off point was established and, finally, an average disturbance calculation was performed for each sample. This figure is adapted https://mdp.sysbio.tools/about. (B,C) Left panels: Histograms show the single sample degree of inflammatory perturbation (DIP) score values relative to the non-anemic at baseline group between patients recovered from anemia and with persistent anemia at baseline (B) and day 180 (C). Right panels: Scatter plots of the summary data for each group are shown. DIP score values were compared between non-anemic at baseline (control group), recovered from anemia and persistent anemia at baseline (B) and day 180 (C). Lines in the scatter plots represent median values. Data were compared using the Mann–Whitney U test. *P < 0.05; ***P < 0.0001. Data are from 191 patients who had complete information on cell counts and biochemical measurements at all study timepoints.

FIGURE 5

Unfavorable outcome occurs mainly in patients with anemia. (A) Among all 191 patients who had complete laboratory data from all the study timepoints, only 18 had unfavorable treatment outcomes (death, failure or loss to follow up), whereas 173 were successfully treated (cure). (B) 14 of the 18 patients had anemia at baseline and 14 also had anemia at the end of the treatment. (C,D) A Heatmap was designed to depict the overall pattern of cells and biochemical markers at all timepoints in favorable and unfavorable outcome of anti-tubercular treatment. A two-way hierarchical cluster analysis (Ward’s method) was performed. Dendrograms represent Euclidean distance. Expression scale represents Z-score normalization from the median at each timepoint and group. It is possible to observe that patients who had an unfavorable outcome have a profile opposite to that of patients with a favorable outcome. In (C) we can identify that patients with a favorable outcome had an increase in most of the blood count parameters over time, and also a decrease in neutrophils, platelets and leukocytes, returning to normal levels of these components. In patients with an unfavorable outcome, the parameters remain more similar to the baseline, with the exception of neutrophils and platelets, which also decrease. In (D) the biochemical markers of both groups do not seem to change much over time, but they visibly present different profiles when comparing favorable and unfavorable. Patients with a favorable outcome have higher levels of albumin, creatinine, urea and bilirubin throughout the treatment. (E) Logistic binomial regression model to significant biochemical parameters was used to test independent associations between biochemical and clinical parameters at baseline and treatment outcome after 180 days. The treatment outcome unfavorable (failure, death or loss to follow up) was used as reference to test associations. Only parameters which remained with p ≤ 0.2 in univariate analysis model were inputted in the adjusted model. (95%CI, 95% confidence interval). Associations reported in (E) are for increases in 1 unit in plasma concentrations of indicated markers.