Anemia burden in tuberculosis patients across Southeast Asia: a systematic review and meta-analysis

Abstract

Background

Anemia represents a significant comorbidity in tuberculosis (TB) patients, affecting treatment outcomes and mortality. Despite Southeast Asia accounting for 46% of global TB cases, no comprehensive regional synthesis of anemia burden exists, limiting evidence-based policy development.

Objectives

This systematic review and meta-analysis aimed to estimate the pooled prevalence of anemia among tuberculosis patients across Southeast Asia and to examine variations in prevalence by geographic, demographic, and clinical characteristics.

Methods

We searched PubMed, Embase, Scopus, and Web of Science for observational studies published from 2000 onwards that reported anemia prevalence in drug-susceptible TB patients from Southeast Asian countries. Random-effects meta-analysis was performed to calculate pooled prevalence estimates with 95% confidence intervals. Subgroup analyses were conducted by country, HIV co-infection status, treatment status, study setting, COVID-19 period, and World Bank income classification. Publication bias was assessed using Egger’s and Begg’s tests.

Results

Twenty-one studies comprising 6,090 TB patients from five Southeast Asian countries were included. The pooled prevalence of anemia was 72% (95% CI: 68%-76%), with substantial heterogeneity (I²=89.9%). These regional prevalence estimates are higher than the global pooled estimate of 53.9% reported by Barzegari et al. (2019), reflecting the distinct epidemiological context of Southeast Asia. Subgroup analysis revealed significantly higher prevalence in HIV-positive patients (88%, 95% CI: 78%-94%) compared to HIV-negative patients (70%, 95% CI: 65%-74%). Studies conducted during COVID-19 showed elevated prevalence (84%, 95% CI: 75%-90%) compared to pre-COVID (72%, 95% CI: 70%-73%) and post-COVID periods (72%, 95% CI: 68%-83%, p = 0.03). Lower-middle income countries showed higher prevalence (73%, 95% CI: 69%-77%) than upper-middle income countries (67%, 95% CI: 63%-70%). No significant publication bias was detected (Egger’s p = 0.721, Begg’s p = 0.537).

Conclusions

This first regional synthesis demonstrates that anemia affects three-quarters of TB patients in Southeast Asia, with significant variation by HIV status, income level, and pandemic period. Unlike previous global estimates (53.9%), our region-specific findings (72%) and documentation of COVID-19 impact (84% during pandemic) provide actionable evidence for Southeast Asian TB programs to integrate anemia screening and management, particularly for vulnerable populations during health system disruptions.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12890-026-04404-x.

Introduction

Anemia, characterized by reduced hemoglobin concentration or red blood cell mass below physiological thresholds, represents one of the most prevalent hematological disorders worldwide, affecting approximately 1.92 billion people globally [1]. The World Health Organization defines anemia as hemoglobin levels below 13 g/dL in adult men and below 12 g/dL in non-pregnant adult women [2]. This condition encompasses multiple etiologies including nutritional deficiencies such as iron, folate, and vitamin B12 deficiency, chronic inflammatory diseases, infectious processes, and hematological disorders [3]. Anemia contributes significantly to morbidity and mortality, accounting for 68.4 million years lived with disability in 2019, representing 8.8% of all global disability [4]. Moreover, anemia can be a risk or a prognostic factor for other diseases, such as tuberculosis and heart failure [5].

Tuberculosis (TB), caused by Mycobacterium tuberculosis, remains one of the oldest and most persistent infectious diseases affecting humanity, primarily targeting the lungs but capable of involving virtually any organ system [6]. Despite being both preventable and curable with appropriate antibiotic therapy, TB continues to pose substantial challenges to global health systems, particularly in resource-limited settings where healthcare infrastructure and diagnostic capabilities remain inadequate [7]. A global total of 8.2 million people were reported as newly diagnosed with TB in 2023, up from 7.5 million in 2022 and 7.1 million in 2019 and far above the levels of 5.8 million in 2020 and 6.4 million in 2021 [8]. Despite decades of intensive control efforts through programs such as DOTS (Directly Observed Treatment, Short-course), TB incidence has declined by only 2% annually, falling short of the End TB Strategy targets [9]. The Southeast Asian region bears a disproportionate share of this burden, accounting for nearly 46% of global TB cases, with India alone contributing approximately 26% of the world’s TB cases, followed by Indonesia (10%), the Philippines (6.8%), Bangladesh (3.5%) [8], The region faces unique challenges including high population density, widespread poverty, malnutrition, increasing diabetes mellitus comorbidity, tobacco use, and healthcare access disparities that fuel TB transmission and complicate treatment outcomes [10]. The complex interplay of social determinants—including overcrowding, inadequate housing, food insecurity, indoor air pollution, and limited healthcare infrastructure—creates an environment conducive to TB transmission despite the availability of effective treatment regimens [11].

Anemia affects approximately 24.8% of the global population, with the highest prevalence observed in preschool-age children (47.4%) and pregnant women (41.8%), though all demographic groups experience significant burden across the lifespan [12]. Iron deficiency accounts for approximately 50% of all anemia cases globally, Globally, in 2021, 31.2% of women had anemia compared with 17.5% of men. The gender difference was more pronounced during the reproductive years, ages 15–49. In this age group, anemia prevalence in women was 33.7% versus 11.3% in men [1]. Anaemia caused 50 million years of healthy life lost due to disability in 2019. The largest causes were dietary iron deficiency, thalassaemia and sickle cell trait, and malaria [12], while other nutritional deficiencies, chronic diseases, and genetic hemoglobin disorders contribute to the remaining burden. Southeast Asia demonstrates particularly high anemia prevalence, with regional estimates suggesting that 30–40% of the general population experiences anemia, driven predominantly by nutritional deficiencies, parasitic infections including hookworm and malaria, and chronic inflammatory conditions [13]. Country-specific data reveal substantial variation, with anemia prevalence ranging from 23% in Thailand to over 50% in India and Bangladesh, reflecting differences in nutritional status, healthcare access, and underlying disease burden [14].

The intersection of tuberculosis and anemia creates a particularly concerning clinical scenario with significant implications for patient outcomes. Multiple studies have documented that 30–94% of TB patients present with concomitant anemia, though prevalence varies substantially across geographic regions, study populations, and healthcare settings [15, 16]. The mechanisms underlying anemia in TB are multifactorial and include anemia of chronic disease (characterized by inflammation-mediated disruption of iron homeostasis through hepcidin elevation), nutritional deficiencies exacerbated by TB-associated hypercatabolism and anorexia, direct bone marrow involvement in disseminated disease, gastrointestinal blood loss from intestinal TB or medication side effects, and hemolysis in some cases [17]. Anemia in TB patients has been associated with adverse clinical outcomes including increased mortality, delayed sputum conversion, longer time to culture negativity, and poorer treatment outcomes, underscoring its clinical significance [18, 19]. In Southeast Asia specifically, limited data from individual country studies suggest anemia prevalence ranging from 31% to 89% among TB patients, with particularly high rates observed in hospital-based studies from India (72–76%), Indonesia (63–80%), and Bangladesh (75%) [20–22]. However, no comprehensive regional synthesis exists to systematically quantify this burden, identify population-specific risk factors, characterize anemia severity distribution, examine temporal trends, or evaluate the impact of different TB treatment phases on anemia prevalence—critical knowledge gaps that impede evidence-based clinical management and public health policy development for integrated TB and anemia care in this high-burden region [23].

Despite the recognized clinical significance of anemia in tuberculosis patients, current global and regional TB management guidelines provide limited specific recommendations for systematic anemia screening, diagnosis, and management within TB care programs [24]. The World Health Organization’s TB treatment guidelines mention anemia as a comorbidity but do not mandate routine hemoglobin testing as part of standard TB diagnostic workup or monitoring protocols [25]. In Southeast Asia, national TB programs vary considerably in their approach to anemia screening, with most countries lacking standardized protocols for identifying and managing anemia in TB patients [8]. India’s National Tuberculosis Elimination Program recommends baseline hemoglobin testing but does not provide clear treatment pathways for detected anemia, while Indonesia’s national guidelines include anemia assessment as part of nutritional screening but implementation remains inconsistent across health facilities [26, 27]. Recent evidence suggests that targeted interventions addressing anemia in TB patients—including iron supplementation, treatment of underlying causes, and nutritional support—may improve treatment adherence, reduce adverse drug reactions, shorten time to sputum conversion, and potentially decrease mortality [28]. However, optimal timing, formulation, and duration of such interventions remain unclear, particularly regarding concerns about iron supplementation potentially enhancing mycobacterial growth, though recent studies suggest this may not be clinically significant in the context of effective anti-TB therapy.

The absence of comprehensive regional data on anemia burden, risk factors, and severity distribution in TB patients across Southeast Asia represents a significant barrier to developing evidence-based screening protocols, treatment guidelines, and resource allocation strategies for integrated TB-anemia management programs, highlighting the critical need for systematic synthesis of available evidence to inform policy and practice in this high-burden region [29].

Rationale for this regional synthesis

While Barzegari et al. (2019) provided a global overview of anemia in TB patients (pooled prevalence 53.9%), their analysis included only limited data from Southeast Asia and did not examine region-specific factors that may influence anemia burden [23]. Southeast Asia, accounting for 46% of global TB cases, presents distinct challenges, including diverse income levels, varying healthcare system capacities, high rates of malnutrition, and different patterns of HIV co-infection compared to African and other settings [8]. Moreover, the COVID-19 pandemic caused unprecedented disruptions to TB diagnosis, treatment, and nutritional support programs across the region, with potential impacts on anemia prevalence that have not been systematically examined. National TB programs in Southeast Asian countries currently lack evidence-based, region-specific data to guide anemia screening and management protocols tailored to their unique epidemiological and resource contexts. Our systematic review addresses these gaps by providing the first comprehensive regional synthesis of anemia burden among drug-susceptible TB patients with detailed subgroup analyses by country, income level, HIV status, and pandemic period, generating actionable evidence for TB control programs in this high-burden region.

The specific objectives are

Primary outcome: determine the pooled prevalence of anemia among TB patients in Southeast Asian countries.

Additional outcomes:

• Country-specific and regional prevalence patterns across Southeast Asian nations.

• Assess the Anemia prevalence across different TB treatment status groups and identify regional patterns across different Southeast Asian nations.

Methodology

Protocol development and registration

This systematic review adheres to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [30]. The review protocol was prospectively registered with the International Prospective Register of Systematic Reviews (PROSPERO) under registration number CRD420251043794 on April 30, 2025 [31]. The review encompasses published research from April 2000 to November 2025, focusing exclusively on studies conducted in Southeast Asian countries, including Bangladesh, Bhutan, India, Indonesia, the Maldives, Myanmar, Nepal, Sri Lanka, Thailand, and Timor-Leste, as well as the Democratic Republic of Korea.

PICOS framework

The following PICOS algorithm principles guided the initial retrieval strategy:

Population (P)

Patients diagnosed with tuberculosis of all ages residing in Southeast Asian countries, including both treatment-naïve individuals and those receiving anti-TB treatment, encompassing all forms of tuberculosis (pulmonary and extrapulmonary), studied in both hospital-based and community settings.

Intervention/exposure (I)

Not applicable, as this is a prevalence study.

Comparison (C)

No comparison.

Outcome (O)

The primary outcome is the pooled prevalence of anemia among tuberculosis patients across Southeast Asia.

Study design and setting (S)

Observational study designs including cross-sectional studies, cohort studies (baseline data), surveillance studies, and population-based surveys conducted in hospital-based, community-based, or mixed healthcare settings.

Study population

• Patients with confirmed tuberculosis diagnosis of all ages.

• Residing in Southeast Asian countries (Bangladesh, Bhutan, India, Indonesia, Maldives, Myanmar, Nepal, Sri Lanka, Thailand, and Timor-Leste).

• Both treatment-naïve patients and those receiving anti-tuberculosis treatment.

• All forms of drug-sensitive tuberculosis (pulmonary and extrapulmonary).

• Patients from both hospital-based and community-based settings.

• Studies conducted in urban, rural, or mixed settings.

Study design

• cross-sectional study/retrospective study/cohort studies (baseline data)/surveillance data, hospital-based studies/community-based surveys/randomized control trials (baseline data)/English language publication/population-based surveys/time series analysis.

• Studies published from April 2000 to November 2025 to capture contemporary TB-anemia epidemiology in the era of standardized DOTS implementation and WHO diagnostic criteria.

Exclusion criteria

• Pregnant women with tuberculosis.

• Patients with drug-resistant tuberculosis (MDR-TB, XDR-TB, or any form of resistance)- Patients with drug-resistant tuberculosis (MDR-, pre-XDR-, and XDR-TB) were excluded to maintain clinical homogeneity and focus on drug-susceptible tuberculosis populations. Drug-resistant TB is associated with prolonged disease duration, repeated or failed treatment courses, extensive exposure to second-line anti-tuberculosis drugs, and higher cumulative inflammatory and nutritional burden, all of which independently affect anemia prevalence and severity. Inclusion of these populations would introduce substantial treatment-related and biological heterogeneity, potentially obscuring anemia patterns attributable to tuberculosis itself.

• Studies conducted outside Southeast Asia.

Study design exclusions

Regarding study design, we excluded case reports, case series with fewer than 10 participants, opinion pieces, narrative reviews, review articles, systematic reviews, meta-analyses, commentaries, editorials, letters, qualitative studies, and conference abstracts. Publications in languages other than English were excluded, as were studies without full-text availability or those with insufficient data for prevalence calculation and Studies without a clear definition of anemia or hemoglobin measurement.

Study outcomes: diagnostic criteria and cut-off values

The outcome variable in this systematic review and meta-analysis was the prevalence of Anemia among Tuberculosis patients in Southeast Asia. Anemia was defined according to the World Health Organization (WHO) criteria based on hemoglobin (Hb) concentration in Children (6–59 months): < 11.0 g/dL, Children (5–11 years): < 11.5 g/dL, and Children (12–14 years): < 12.0 g/dL. Adults: Men (15 years and above): < 13.0 g/dL, Women (non-pregnant, 15 years and above): < 12.0 g/dL [32, 33].

Information sources

A comprehensive literature search was conducted across multiple electronic databases, including PubMed/MEDLINE, Embase, Scopus, and Web of Science, from April 2000 to November 2025. The search strategy incorporated both controlled vocabulary terms (Medical Subject Headings) and free-text keywords encompassing “tuberculosis,” “TB,” “Mycobacterium tuberculosis,” “pulmonary tuberculosis,” “extrapulmonary tuberculosis,” combined with “anemia,” “anaemia,” “hemoglobin,” “haemoglobin,” “hematocrit,” “haematocrit,” “iron deficiency,” “anemia of chronic disease,” and specific Southeast Asian country names including Bangladesh, Bhutan, India, Indonesia, Maldives, Myanmar, Nepal, Sri Lanka, Thailand, and Timor-Leste, Democratic republic of korea. Detailed search strategies for each database are documented in accordance with PRISMA guidelines. To ensure comprehensive coverage, we conducted a hand-search of reference lists from eligible studies, performed forward citation tracking of seminal papers, and consulted with regional tuberculosis specialists and haematology experts across Southeast Asia. Grey literature sources, including conference proceedings, dissertations, and trial registries, were also examined to identify potentially relevant unpublished studies (Supplementary Table 1).

To enhance comprehensiveness, we additionally searched the Index Medicus for the South-East Asia Region (IMSEAR); however, no additional studies meeting the inclusion criteria were identified.

The search methodology and documentation followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 guidelines to ensure transparency, reproducibility, and methodological rigor of the review process [34].

Study selection process

Retrieved records from all databases were exported and managed using Rayyan systematic review software to facilitate the identification and removal of duplicates. After deduplication, two independent investigators (V.S.) & (H.K.) conducted a comprehensive two-phase screening procedure. The initial phase involved screening titles and abstracts of all identified records against predetermined eligibility criteria. Studies were included if they reported prevalence data on anemia among tuberculosis patients in Southeast Asian countries, utilized observational study designs including cross-sectional, cohort, or surveillance studies, and included patients of all ages with confirmed tuberculosis diagnosis. Studies were excluded if they were conducted outside Southeast Asia, consisted of case reports or case series with fewer than ten participants, involved only animal or in vitro studies, or lacked sufficient data on anemia prevalence. In the second phase, full-text manuscripts of potentially relevant studies were retrieved and examined in detail for final inclusion determination. Any disagreements between the two primary investigators during either screening phase were addressed through detailed discussion to achieve consensus. When consensus could not be established through discussion, a third independent investigator (Y.M.) was consulted to provide the decisive evaluation. A standardized eligibility assessment form was utilized throughout the selection process to ensure consistency in applying inclusion and exclusion criteria.

Data extraction

A standardized data extraction form was developed and pilot-tested on a sample of five included studies before full implementation. Two independent reviewers extracted data from all included studies using this standardized form. The following information was systematically extracted from each study: first author name and publication year, study country and geographic setting, study design and sampling methodology, sample size and population characteristics, tuberculosis-related parameters including type of tuberculosis, anatomical site, and bacteriological confirmation status, hemoglobin measurement methods and anemia definitions, overall prevalence of anemia with numerators and denominators, demographic characteristics including age distribution and gender breakdown, clinical parameters such as HIV co-infection status and tuberculosis treatment status, reported risk factors or associations with anemia, and study quality indicators. Where studies reported anemia prevalence at multiple time points, baseline or pre-treatment data were preferentially extracted. When data were presented only in graphical format, numerical values were extracted using digital extraction tools. Any discrepancies in extracted data between the two reviewers were resolved through discussion and reference to the original publication. Missing data were managed through a systematic protocol. When critical information was unavailable from published reports, study authors were contacted via email (with up to two follow-up attempts) to request clarification or additional data. For studies lacking confidence intervals but reporting prevalence and sample sizes, confidence intervals were calculated using standard binomial methods. Studies were excluded only when essential data (overall prevalence, sample size, or diagnostic criteria) could not be obtained or calculated. Missing moderator variables were coded as “not reported” rather than excluded to maximize study retention.

Quality assessment

The methodological quality and risk of bias of included studies were assessed independently by two reviewers using appropriate quality assessment tools based on study design. For cross-sectional prevalence studies, the Joanna Briggs Institute Critical Appraisal Checklist for Studies Reporting Prevalence Data was utilized [35], which evaluates studies across nine domains, including appropriateness of sampling frame, adequacy of sample size, detailed description of study subjects and setting, sufficient coverage of identified sample, validity and reliability of outcome measurement, appropriateness of statistical analysis, adequacy of response rate, and appropriate reporting of prevalence with confidence intervals and subgroup analysis (Supplementary Table 2).

Two independent reviewers conducted quality assessments for each included study using the JBI framework. Each methodological criterion was scored as “1” for clearly satisfied requirements with adequate reporting or “0” for unmet or poorly described criteria. Total quality scores were calculated by summing individual criterion scores, yielding a range of 0–9 points per study. Studies were subsequently stratified into quality categories: low quality (0–3 points), moderate quality (4–6 points), and high quality (7–9 points) [36]. Additionally, we implemented GRADE (Grading of Recommendations Assessment, Development and Evaluation) staging to assess the overall certainty of evidence across studies.

Statistical analysis

All quantitative analyses were performed using R statistical software version 4.5.1, incorporating the ‘meta’ and ‘metafor’ packages for meta-analytical computations. A random-effects meta-analysis was conducted to pool prevalence estimates across included studies, accounting for expected heterogeneity in study populations, settings, and methodologies. The Sidik Jonkman method was used to calculate pooled prevalence estimates with corresponding 95% confidence intervals. The Sidik-Jonkman method was selected for estimating between-study variance (τ²) as it provides more accurate confidence intervals for prevalence meta-analyses, particularly in the presence of high heterogeneity, and has been shown to outperform traditional methods like DerSimonian-Laird in simulation studies (Langan et al., 2019) [37]. Individual study prevalence estimates were transformed using the Freeman-Tukey double arcsine transformation to stabilize variances before pooling, and back-transformed for presentation of final pooled estimates. Statistical heterogeneity across studies was quantified using the I-squared statistic, which represents the percentage of total variation across studies attributable to heterogeneity rather than chance, with I² values exceeding 50% and p-values below 0.05 signifying considerable heterogeneity, warranting additional exploration. Forest plots were generated to provide a visual representation of individual study estimates and pooled prevalence with confidence intervals.

Subgroup and sensitivity analyses

Predefined subgroup analyses were conducted to explore sources of heterogeneity and examine variation in anemia prevalence across different characteristics [38]. Subgroups were defined based on geographic location, patient demographics, including treatment status, HIV co-infection status, study setting, and sampling methodology, Sampling range, World Bank classification, and the COVID period. Pooled prevalence estimates were calculated separately for each subgroup, and statistical tests for subgroup differences were performed using the chi-square test. Meta-regression was conducted when sufficient studies were available to examine associations between study-level covariates and reported prevalence. Sensitivity analyses assessed the robustness of findings by sequentially excluding high-risk studies, small sample size studies, outliers identified through Galbraith plots, and studies not using the World Health Organization anemia criteria. Cumulative meta-analysis ordered by publication year examined temporal trends in prevalence estimates.

Assessment of publication bias

Publication bias was assessed using multiple complementary approaches. Funnel plots of prevalence estimates against standard errors were visually inspected for asymmetry. Egger’s regression test and Begg’s rank correlation test were performed as statistical assessments, with p-values less than 0.05 considered significant. When publication bias was detected, trim-and-fill analysis estimated the number of missing studies and provided adjusted pooled estimates. The Doi plot and Luis Furuya-Kanamori index were generated as alternative assessments, with index values exceeding plus or minus two indicating major asymmetry [39]. Given that publication bias is less concerning for prevalence studies, these assessments were interpreted cautiously alongside other sources of asymmetry, such as heterogeneity and methodological quality differences.

Results

Literature search and identified results

A comprehensive systematic literature search across multiple electronic databases yielded a total of 954 records. The search was conducted in PubMed (n = 236), Scopus (n = 182), Embase (n = 147), Web of Science (n = 289), and Google Scholar (n = 100). Following the removal of duplicate records [40], 343 unique records remained for screening. During the initial title and abstract screening phase, 316 records were excluded as they did not meet the predetermined eligibility criteria, leaving 27 records for full-text retrieval and detailed assessment. Of these, two records could not be retrieved despite attempts to contact authors and access through institutional libraries. The remaining 25 full-text articles underwent a comprehensive eligibility assessment against all inclusion and exclusion criteria. During this phase, four reports were excluded for specific reasons: two studies were excluded because they did not utilize appropriate World Health Organization diagnostic criteria for anemia classification, and two studies were excluded due to the absence of reported prevalence data in the published manuscript. Ultimately, 21 studies met all eligibility criteria and were included in the qualitative synthesis and quantitative meta-analysis. This systematic selection process, conducted in accordance with PRISMA guidelines & MOOSE reporting guidelines [41], ensured that only high-quality, relevant studies examining anemia prevalence among tuberculosis patients in Southeast Asian countries were included in the final review (Fig. 1).

Fig. 1.

PRISMA Flowchart of the selection of studies for anemia prevalence among TB patients in Southeast Asia: a systematic review and meta-analysis

Features of the included studies

The systematic review included 21 studies conducted across Southeast Asia between 2000 and 2025, encompassing 6,090 tuberculosis patients. 9 Studies were predominantly from India (n = 14), followed by Indonesia (n = 4), with single studies from Bangladesh, Thailand, and Nepal. Sample sizes ranged from 41 to 1,695 participants, with most studies (n = 16) having 10–500 participants. The majority of studies were conducted in lower-middle-income countries (n = 17) with weak surveillance systems according to Joint External Evaluation criteria. Hospital-based settings predominated (n = 16), with urban and mixed populations most commonly represented. Consecutive or systematic sampling was employed in 16 studies. Regarding tuberculosis treatment status, 13 studies included patients on anti-tuberculosis treatment, four focused on treatment-naive patients, and four included mixed populations. HIV co-infection status varied, with eight studies examining HIV-negative patients, 11 studies including patients with unknown HIV status, and two studies focusing on HIV-positive patients. Quality scores ranged from 5 to 9 out of 10, indicating moderate to high methodological quality (Supplementary Table 7).

Meta-analysis results

The pooled analysis of 21 studies involving 6,090 tuberculosis patients across Southeast Asia revealed a pooled anemia prevalence of 72% using the random-effects model (95% CI: 68% to 76%). The common effect model yielded an identical pooled prevalence of 72% (95% CI: 71% to 73%). The prediction interval ranged from 52% to 87%, indicating considerable variation across different settings. Individual study prevalence estimates varied widely, ranging from 31% (Rathi et al. 2023) to 92% (Yogesh M et al. 2025), with most studies reporting estimates between 60% and 80%. The largest study by Bhargava et al. 2013, including 1,695 participants, reported 73% prevalence and contributed the greatest weight to the pooled estimate. Statistical heterogeneity was substantial with an I-squared value of 89.9% and tau-squared of 0.1752, indicating that approximately 90% of observed variation was due to true differences between studies rather than sampling error. The chi-square test for heterogeneity was highly significant (p < 0.0001), confirming considerable heterogeneity and justifying the use of random-effects meta-analysis and further subgroup exploration (Fig. 2).

Fig. 2.

Meta-analysis of forest plot

Cumulative meta-analysis

A cumulative meta-analysis was performed to investigate the historical progression of anemia prevalence estimations by successively incorporating papers from 2000 to 2025. The initial study conducted by Karyadi et al. in 2000 indicated a prevalence of 59% (95% CI: 43% to 72%). After adding Sahiratmadja et al. in 2007, the overall estimate rose to 63% (95% CI: 58% to 67%), and there was no variation (I² = 0%). The addition of three studies from 2013 gradually raised the estimate, and the big Bhargava et al. study kept it at 68% (95% CI: 61% to 75%, I² = 83.4%).

Starting in 2014, the pooled prevalence stayed the same, only going up and down between 70% and 73%. Studies conducted from 2021 to 2023 sustained prevalence estimates between 72% and 73%. However, Rathi et al. (2023) found a significantly lower prevalence of 31%, which briefly adjusted the pooled estimate to 71% and markedly heightened heterogeneity (I² = 90.4%). However, other investigations in 2024 and 2025 found that the stability was still approximately 70–73%. The final cumulative estimate was 73% (95% CI: 67% to 77%, I² = 89.9%), which is almost the same as the usual meta-analysis result, suggesting the 72–73% estimate is robust and unlikely to change substantially with future studies (Supplementary Fig. 1).

Subgroup analysis by country

When we looked at anemia prevalence across different Southeast Asian countries, some interesting patterns emerged. India, contributing the most of studies with 14 research papers involving 5012 patients, showed a pooled prevalence of 72% (95% CI: 71% to 74%). Indonesian studies, though fewer with just four articles covering 583 patients, revealed a slightly lower but still substantial prevalence of 67% (95% CI: 57% to 76%). Single studies from Bangladesh, Thailand, and Nepal reported prevalences of 75%, 80%, and 73% respectively, though these estimates should be interpreted cautiously given the limited data from each country. Despite India’s notably high heterogeneity (I² = 92.5%), the differences between countries weren’t statistically significant when using the random-effects model (p = 0.55), suggesting that anemia remains a widespread problem across the region regardless of national boundaries (Supplementary Fig. 2).

Subgroup analysis by covid-period

The substantially higher anemia prevalence observed during the COVID-19 period likely reflects delayed tuberculosis diagnosis, worsening nutritional insecurity, and healthcare disruptions. The return to pre-pandemic prevalence levels in post-COVID studies suggests partial recovery of TB services, although the persistently high burden underscores unresolved structural vulnerabilities (Supplementary Fig. 3).

Subgroup analysis by World Bank classification

A subgroup study based on the World Bank income classification revealed intriguing socioeconomic trends in anaemia prevalence. The bulk of studies (n = 17, 5,453 patients) came from lower-middle-income countries. The pooled prevalence was 73% (95% CI: 69%-77%), with a significant level of heterogeneity (I² = 91.3%). There were fewer studies in upper-middle-income countries (n = 5, 637 patients), and they found a lower prevalence of 69% (95% CI: 60% to 78%). The results were moderately heterogeneous (I² = 75.4%). The common effect model indicated statistically significant differences among income groups (p = 0.0018), whereas the random-effects model revealed no significant difference (p = 0.46). This indicates that although point estimates seem disparate, the range within each income level is sufficiently significant that we cannot conclusively assert that economic position solely dictates anemic burden (Supplementary Fig. 4).

Subgroup analysis by study area

We examined whether anemia prevalence differed between hospital and community settings. Hospital-based studies (n = 16, 2,252 patients) showed a prevalence of 73% (95% CI: 66% to 79%), while community-based studies (n = 4, 1,415 patients) reported virtually identical results at 70% (95% CI: 62% to 77%). Two studies covering both settings found a 73% prevalence. The differences between healthcare settings were not statistically significant (p = 0.66), indicating that anemia affects tuberculosis patients equally whether they’re hospitalized or receiving community-based care (Supplementary Fig. 5).

Subgroup analysis by surveillance system

We analyzed anemia prevalence based on countries’ surveillance system strength using Joint External Evaluation criteria [42]. Studies from countries with weak surveillance systems (n = 6, 1,024 patients) reported a prevalence of 70% (95% CI: 63% to 76, with moderate heterogeneity (I² = 75.1%). The single study from a country with strong surveillance (n = 1, 54 patients) reported 80% prevalence (95% CI: 66% to 89%). Most studies (n = 14, 5,012 patients) came from settings where surveillance was not formally assessed, showing 73% prevalence (95% CI: 68% to 78%) with high heterogeneity (I² = 92.5%). This indicates that the prevalence of anemia is elevated among tuberculosis patients, irrespective of national monitoring infrastructure, though the limited data from strong surveillance settings prevent definitive comparisons (Supplementary Fig. 6).

Subgroup analysis by sampling method

Research utilizing consecutive (n = 16, 5,115 individuals) revealed a 73% prevalence of anemia (95% CI: 68% to 77%). In contrast, studies employing convenience sampling (n = 5, 975 patients) indicated a prevalence of 71% (95% CI: 63% to 77%). So that means the sample strategies were not statistically significant (p = 0.63), which means that the estimations of anemia prevalence remain steady, no matter whether which sampling method is used. (Supplementary Fig. 7).

Subgroup analysis by sample range

Studies were grouped by sample size: small (< 100 patients, n = 4, 805 patients) showed 71% prevalence, medium-sized (100–500 patients, n = 15, 2,862 patients) reported 73% prevalence, one large study (501–1000 patients, n = 1, 728 patients) found 74% prevalence, and one very large study (> 1000 patients, n = 1, 1,695 patients) showed 73% prevalence. No significant differences existed between sample size categories (p = 0.91), confirming that anemia prevalence estimates are consistent regardless of study size (Supplementary Fig. 8).

Subgroup analysis by population

Studies that looked at both adults and teens (n = 5,789 patients) found that 68% of them had anemia (95% CI: 61% to 75%). Studies that only looked at adults (n = 16, 5,301 patients) found that 74% of them had anemia (95% CI: 69% to 78%). Even if there seems to be a difference, statistical testing showed that there was no significant difference between age groups (p = 0.20). This means that the burden of anemia is the same among adult tuberculosis patients, whether or not adolescents are included (Supplementary Fig. 9).

Subgroup analysis by treatment status

Treatment-naive patients (n = 6, 903 patients) exhibited a 69% prevalence of anemia (95% CI: 66% to 72%), whereas individuals undergoing anti-TB treatment (n = 11, 2,549 patients) demonstrated a prevalence of 73% (95% CI: 65% to 80%). Research involving heterogeneous populations (n = 4, 2,638 patients) indicated a prevalence of 74% (95% CI: 72% to 75%). There were no significant differences in treatment status groups (p = 0.19), indicating that anemia impacts tuberculosis patients similarly, regardless of whether they are newly diagnosed or currently on treatment. Anemia prevalence was compared across treatment status groups rather than assessed longitudinally (Supplementary Fig. 10).

Subgroup analysis by HIV Co-infection status

Subgroup analysis by HIV status revealed important patterns. HIV-negative tuberculosis patients (n = 8, 1,478 patients) showed 70% anemia prevalence (95% CI: 65% to 74%), while studies with unknown HIV status (n = 11, 4,356 patients) reported a prevalence of 72% (95% CI: 63% to 78%). Notably, HIV-positive patients (n = 2, 256 patients) had the highest anemia burden at 88% (95% CI: 78% to 94%). The differences between HIV status groups were statistically significant in the common effect model (p < 0.001) but not in the random-effects model (p = 0.01). This finding underscores that HIV co-infection substantially increases anemia risk in tuberculosis patients, likely due to the combined effects of both infections on hematopoiesis, immune function (Supplementary Fig. 11).

While the above subgroup analyses provided category-specific prevalence estimates with direct clinical applicability, we additionally performed meta-regression to quantify how much variance each moderator independently explained and to assess their relative importance in contributing to overall heterogeneity.

Meta regression

We conducted meta-regression to explore which study characteristics might explain the substantial heterogeneity in anemia prevalence estimates. Among all variables examined, only study setting emerged as a statistically significant predictor of anemia prevalence (p < 0.05). Other factors, including publication year (p = 0.30), country, World Bank income classification (p = 0.43), surveillance capacity (p = 0.48), COVID period, sampling method, sample size, TB treatment status, and HIV co-infection status (p = 0.08), were not statistically significant predictors.

However, when examining the adjusted R² values to identify potential sources of heterogeneity, three variables stood out: study setting, COVID period, and HIV co-infection status. All showed substantial R² values, suggesting they may contribute to the observed heterogeneity despite not all reaching statistical significance. The study setting’s significant p-value, combined with meaningful R², confirms it as a key source of variation (Supplementary Table 3).

Sensitivity analysis: leave-one-out method

We tested whether any single study was driving our results by systematically removing each study one at a time and recalculating the pooled prevalence. The analysis showed impressive stability—the pooled estimate barely budged, staying between 70% and 73% across all iterations. The biggest change came from removing Yogesh M et al. 2025, which dropped the estimate to 70%, while excluding Rathi et al. 2023 pushed it up to 73%. Most exclusions changed the pooled prevalence by less than 1%.

Heterogeneity remained high throughout (I² = 57.6% to 90.4%), with Rathi et al. showing the highest when removed (I² = 90.4%). The fact that removing any single study—whether small or large, from India or Indonesia, with high or low prevalence—produced such minimal changes proves our findings are rock-solid. No single study dominated the results, and the 72% pooled prevalence stands firm regardless of which studies are included or excluded (Supplementary Table 4).

Galbraith plot analysis

The Galbraith plot, which looks at normalized residuals and precision (the opposite of standard error), found seven studies that might not be typical. Yogesh M 2025, Chhabra et al. 2021, Sahiratmadja et al. 2007, and Rathi et al. 2023 were all very far distant from the other data points. They were shown in red because they were more than three standard deviations apart from the pooled effect. This means they were a long way from the 95% confidence intervals. The orange lines show that Ramanna Nataraja et al. 2023, Karyadi et al. 2000, and Leon et al. 2024 were not too far off. The blue line marks the zero line, which is where the other 14 efforts came extremely close to. Rathi et al. 2023 was the most extreme outlier since it had a very high precision but a standardized residual of about − 10, which suggested that it was exceedingly uncommon (only 31% of the time). On the other side, Yogesh M 2025 had the highest positive residual, which was clear in 92% of cases. The sensitivity analysis showed that these outliers didn’t have a big effect on the overall pooled estimate, which suggests our results are strong (Fig. 3).

Fig. 3.

Enhanced Galbraith Plot

Sensitivity analysis excluding extreme outliers

To evaluate the influence of extreme outliers on our findings, we performed a sensitivity analysis omitting the four papers designated as extreme outliers in the Galbraith plot: Yogesh M 2025, Chhabra et al. 2021, Sahiratmadja et al. 2007, and Rathi et al. 2023. After excluding these investigations, the analysis comprised 17 trials with a total of 5,198 patients. Using the common effect model, the pooled prevalence was 73% (95% CI: 72% to 74%), while using the random-effects model, it was 73% (95% CI: 71% to 75%). This is almost the same as the original estimate of 72%. More crucially, heterogeneity was reduced a lot, from 89.9% to 40.4% (Tau² = 0.0102, p = 0.04). This shows that the extreme outliers were definitely big contributors to the observed heterogeneity. Even though this made the groups less different, the overall prevalence stayed the same. This shows that the 72–73% estimate is a good approximation of how many people with tuberculosis in Southeast Asia have anemia, no matter if whether outlier studies are included or not (Supplementary Fig. 12).

Risk of bias

Publication bias assessment

We assessed publication bias using visual and statistical methods. The funnel plot showed a relatively symmetric distribution of studies around the pooled estimate, with most studies clustering appropriately within the pseudo-confidence region, though some minor asymmetry was visible. Statistical tests confirmed the absence of significant publication bias. Egger’s regression test yielded a non-significant result (t = 0.3626, p = 0.721, SE = 0.3005), and Begg’s rank correlation test similarly showed no evidence of bias (z = 0.6171, p = 0.537). Both p-values substantially exceeded the 0.10 threshold typically used for publication bias detection. The consistency between visual inspection and statistical tests indicates that our pooled prevalence estimate is unlikely to be affected by selective publication of studies, providing confidence that the 72% estimate represents an unbiased summary of anemia burden in tuberculosis patients across Southeast Asia (Fig. 4 and Supplementary Table 5).

Fig. 4.

Funnel plot

Doi plot and LFK index assessment

As an alternative method for assessing publication bias that is less affected by heterogeneity than traditional funnel plots, we generated a Doi plot and calculated the Luis Furuya-Kanamori (LFK) index. The Doi plot displayed a relatively symmetric distribution of studies, with most points clustering around the central vertical line at position 1. The LFK index value was − 0.312, which falls well within the acceptable range of -1 to + 1, indicating no evidence of major asymmetry or publication bias (Fig. 5).

Fig. 5.

Doi Plot and LFK index

Quality of evidence: GRADE assessment

We evaluated the certainty of evidence using the GRADE approach. The evidence started as high quality from observational studies but was downgraded by two levels due to serious inconsistency (I² = 89.9%) and indirectness. No downgrades were applied for risk of bias, publication bias, or imprecision. The overall certainty of evidence was rated as low, indicating moderate confidence that the true anemia prevalence is close to 72%, though substantial variation exists across different settings and populations (Table 1).

Table 1.

GRADE assessment

Outcome	No. of Studies	Study Design	Risk of Bias	Inconsistency	Indirectness	Imprecision	Publication Bias	Effect (95% CI)	Certainty
Anemia prevalence in tuberculosis patients	21 studies (6090 participants)	Observational (cross-sectional and cohort (Case Control)	Serious	Very serious (I² = 89.9%)	Serious (predominantly hospital-based studies)	Not serious (narrow CI: 66.7–77.7%)	Not Detected (Egger’s p = 0.362, Begg’s p = 0.54)	~ 72.5% (CI: 66.7–77.7%)	⬤⬤◯◯ Low

Indirectness was downgraded by one level due to the predominance of hospital-based studies contributing the majority of participants and analytical weight, which may limit generalizability to tuberculosis patients managed exclusively in community settings, despite inclusion of some community-based studies

Discussion

This is the first systematic review and meta-analysis to synthesize anemia prevalence specifically among TB patients across Southeast Asia. The only prior global meta-analysis (Barzegari et al., 2019) [23] lacked region-specific stratification, COVID-period analysis, temporal trends, or subgroup analyses by HIV co-infection and income classification, all critical determinants in Southeast Asian settings. The pooled prevalence of 72% (95% CI: 68%–76%) from 21 studies across 6,090 patients represents the most comprehensive regional estimate available and addresses a documented evidence gap for a region accounting for 45% of global TB incidence.

Finally, methodological differences, including stricter adherence to World Health Organization hemoglobin cut-offs and a larger cumulative sample size, may further contribute to the observed discrepancy in prevalence estimates. Moreover, the universally applied WHO hemoglobin-based diagnostic criteria for anemia across all included studies ensured diagnostic consistency, a key quality indicator for prevalence meta-analyses. Studies lacking a clear WHO-aligned definition were excluded during the eligibility assessment phase, further strengthening the internal consistency of our pooled estimate.

The remarkable consistency of our pooled estimate across sensitivity analyses—remaining 70–73% regardless of which studies were excluded—provides compelling evidence of genuine disease burden. This aligns with individual country reports: Mukherjee et al. (2017) documented 73% in India [43], Monjur et al. (2014) found 75% in Bangladesh [22], and Sahiratmadja et al. (2007) observed 63% in Indonesia [44]. Despite substantial heterogeneity, the stability of pooled estimates across multiple sensitivity analyses supports the reliability of the overall prevalence estimate.

Geographic patterns and socioeconomic factors

The 72% pooled prevalence was remarkably consistent across countries despite economic differences—India (72%), Indonesia (67%), Bangladesh (75%), Thailand (80%), Nepal (73%)—suggesting TB-associated anemia is primarily driven by disease pathophysiology rather than national income levels. Across countries and income classifications, anemia prevalence remained consistently high, suggesting that tuberculosis-associated anemia is driven more by shared biological and structural determinants—such as chronic inflammation, malnutrition, and delayed care—than by national economic status alone. The absence of significant differences between lower-middle income countries (73%) and upper-middle income countries (69%, p = 0.46) further suggests that individual-level poverty and healthcare access barriers may be more important than national economic status. Importantly, no studies from Myanmar or Timor-Leste met the inclusion criteria, representing critical evidence gaps in countries with substantial TB burdens.

COVID-19 pandemic impact

Our most striking novel finding is the 84% anemia prevalence during COVID-19 (95% CI: 75–90%) versus 71% pre-COVID and 72% post-COVID (p = 0.03), representing a 13-percentage-point increase previously undocumented in TB-anemia literature. This likely reflects delayed diagnosis (WHO reported a 21% decline in Southeast Asian TB notifications during 2020 [25], selecting for more severe cases), pandemic-related nutritional deterioration (Asian Development Bank estimated 78 million pushed into extreme poverty), and possible direct SARS-CoV-2 hematological effects. The return to 72% post-pandemic demonstrates healthcare system resilience but indicates persistent underlying determinants requiring sustained attention.

HIV co-infection and high-risk groups

Although some subgroup differences were not statistically significant in random-effects models, the consistently higher anemia prevalence among HIV-co-infected patients supports a biologically plausible interaction between tuberculosis, HIV-related immune dysregulation, and impaired hematopoiesis. These findings should be interpreted as additive clinical risk rather than contradictory statistical outcomes.

Pathophysiological interpretation of anemia in tuberculosis

Our investigation primarily assesses the prevalence of anemia, however the significant burden seen is likely attributable to the interplay of many pathophysiological pathways associated with tuberculosis. Inflammation is a primary cause of anemia. It happens because the immune system is always active and interleukin-6 generates hepcidin too much, which makes iron less available and prevents the production of red blood cells even when there is adequate iron in the body. This technique is really crucial for patients who are very sick, have symptoms that last a long period, and have a lot of mycobacteria in their systems. These features are more common in groups that are based in hospitals.

In Southeast Asia, nutritional anemia makes things much worse. Micronutrient deficiencies (iron, folate, and vitamin B12) are common, and tuberculosis-related hypercatabolism, anorexia, and socioeconomic misery may make them worse. Bone marrow suppression due to chronic infection, cytokine-mediated regulation of erythroid progenitors, and, in some cases, direct marrow involvement in disseminated disease illustrate other mechanisms. These pathways frequently coexist, resulting in mixed anemia phenotypes rather than specific classifications.

The outcomes from our subset corroborate this biological framework. The markedly heightened prevalence of anemia in HIV co-infected individuals (88%) likely signifies a confluence of immune dysregulation, chronic inflammation, opportunistic infections, and marrow destruction linked to antiretroviral therapy or drug utilization. The heightened incidence observed during the COVID-19 era may signify delayed tuberculosis diagnosis, exacerbated disease severity at presentation, higher inflammatory burden, and greater nutritional insecurity. These data suggest that the prevalence of anemia in tuberculosis should be considered not merely as a comorbidity count, but as a clinical marker of systemic disease severity and inflammatory burden.

We acknowledge that differentiation between anemia of chronic disease (ACD) and iron-deficiency anemia (IDA) is clinically relevant, as the two entities require distinct therapeutic approaches. However, this level of characterisation was not systematically reported in the primary studies available for this region and time period, and could not therefore be synthesised in the meta-analysis.

Treatment status and healthcare settings

We found no significant difference in anemia prevalence between treatment-naïve (69%) and on-treatment patients (73%, p = 0.19). However, this cross-sectional comparison cannot assess the true impact of anti-TB treatment on anemia resolution. The lack of difference may reflect cross-sectional sampling at various treatment time-points rather than indicating that treatment does not affect anemia. Longitudinal studies, such as Lee et al. (2006), have demonstrated anemia improvement with complete anti-TB therapy [45]. The paradoxical similarity in our data may also reflect anti-TB medication hematological effects (rifampicin-induced hemolysis, isoniazid B6 interference), persistent inflammation despite microbiological control, and unaddressed nutritional deficiencies.

Heterogeneity and robustness of findings

The substantial heterogeneity (I² = 89.9%) observed is expected and informative in prevalence meta-analyses synthesizing data across diverse real-world settings. Variability reflects genuine differences in anemia burden driven by HIV co-infection (88% vs. 70% in HIV-negative), healthcare setting, and COVID-19 disruptions (84% vs. 71% pre-COVID). Meta-regression identified study setting as a significant moderator (p < 0.05), with HIV status and COVID period contributing meaningfully to explained variance. The random-effects model appropriately accounts for this true between-study variation. Critically, leave-one-out sensitivity analyses demonstrated remarkable stability — pooled prevalence remained 70–73% regardless of which study was excluded, confirming no single study drove the results.

Sampling methodology

Consecutive or systematic sampling was used in 16 of 21 studies, reflecting standard practice in TB clinic-based research. This was not a selective inclusion decision; we applied exhaustive eligibility criteria and included all available studies meeting those criteria. Consecutive sampling (enrolling all eligible patients over a defined period) minimizes selection bias within facilities and differs from convenience sampling (opportunistic selection). All studies used WHO hemoglobin criteria, ensuring diagnostic consistency. Studies lacking WHO-aligned definitions were excluded during eligibility screening.

Furthermore, the meta-analysis pools data from 6,090 patients across 21 studies, providing substantial cumulative statistical power for prevalence estimation. The issue of study-level power is less critical in prevalence meta-analyses than in intervention studies, since the objective is to estimate a prevalence rather than detect a treatment effect across heterogeneous patient populations.

Limitations

There are some limitations with this systematic review that need to be pointed out. The predominance of Indian studies (14 of 21, 67%) and the absence of data from several Southeast Asian countries, including Myanmar, Laos, Cambodia, and Timor-Leste, limit the geographical representativeness of the pooled estimate. The findings are therefore most directly applicable to lower-middle-income, hospital-based settings resembling the Indian and Indonesian healthcare contexts that dominate the evidence base.

The cross-sectional design of the majority of included studies (17 of 21) precludes causal inference regarding the direction of the TB-anemia relationship — specifically, whether anemia precedes TB (as a susceptibility factor), develops concurrently with TB (as a pathophysiological consequence), or worsens during treatment (as a drug-related adverse effect). Longitudinal studies are needed to characterise anemia trajectories across the TB treatment continuum.

The inability to distinguish between anemia of chronic disease and nutritional deficiency anemia in the meta-analysis is a recognised limitation, arising from non-reporting of iron indices and inflammatory markers in the primary literature rather than from methodological decisions of this review. And eleven studies (covering 71.5% of participants) did not report HIV status, precluding HIV-stratified analysis for these studies, and the restriction to English-language publications may have excluded relevant regional studies published in Indonesian, Bengali, or Thai-language journals. These limitations are transparently reported and should guide the design of future primary research in this area.

The exclusion of drug-resistant TB (MDR-, pre-XDR-, XDR-TB) represents a deliberate methodological trade-off: while it reduces treatment-related heterogeneity and focuses the estimate on drug-susceptible TB, it narrows external validity in settings where drug-resistant TB is a clinically relevant and growing subpopulation. Given the higher inflammatory burden and hematological toxicity of second-line regimens, anemia prevalence in drug-resistant TB populations is likely higher than the 72% reported here, representing an important evidence gap for future research.

The persistent heterogeneity (I² = 89.9%) after subgroup analyses reflects both clinical and methodological sources. Clinical heterogeneity is driven by genuine differences in HIV co-infection rates, TB disease severity, nutritional status, healthcare setting, and pandemic period — evidenced by the 18-percentage-point prevalence gap between HIV-positive and HIV-negative patients. Methodological heterogeneity arises from differences in hemoglobin measurement timing, laboratory methods, and comorbidity reporting across 25 years of studies. The pooled estimate should therefore be interpreted as a regional burden indicator rather than a universally applicable point estimate.

Clinical and policy implications

The finding that nearly three-quarters of TB patients in Southeast Asia present with anemia carries substantial implications for regional TB elimination programmes. Current WHO TB treatment guidelines recommend hemoglobin assessment only when clinically suspected — a threshold that is evidently inadequate given a 72% background prevalence. A patient presenting with TB in Southeast Asia has an approximately 3 in 4 probability of having concurrent anemia, rendering ‘clinical suspicion’ an insufficient trigger for testing. Given the 72% pooled prevalence, there is a strong epidemiological rationale for integrating routine point-of-care hemoglobin measurement into the initial TB diagnostic workup across Southeast Asia, including in community-based and primary care settings where the majority of TB patients first present. However, a positive screen must be followed by a structured management pathway that accounts for the predominant etiology of anemia in this population. Since anemia of chronic disease — driven by TB-associated inflammation and hepcidin elevation — represents the most common anemia subtype in TB patients, its primary correction depends on effective anti-TB treatment itself rather than empirical iron supplementation. Indiscriminate iron supplementation without confirming iron deficiency is not recommended, as it carries theoretical risks of mycobacterial growth enhancement and may be ineffective when inflammation-mediated iron sequestration is the underlying mechanism.

Where laboratory capacity permits, baseline assessment of serum iron, ferritin, folate, and vitamin B12 should be incorporated into TB workup protocols to identify co-existing nutritional deficiency anemias that require specific micronutrient supplementation alongside anti-TB therapy. In resource-limited community settings where such testing is unavailable, a pragmatic tiered approach is recommended: universal hemoglobin screening at TB diagnosis, empirical nutritional supplementation for patients with moderate-to-severe anemia and confirmed or suspected nutritional deficiency risk factors, and referral for further evaluation where hemoglobin fails to improve following treatment initiation. These recommendations should be regarded as provisional, pending longitudinal and interventional studies that robustly evaluate the clinical impact of structured anemia management within TB care pathways.

The 88% anemia prevalence in HIV-TB co-infected patients further identifies this subgroup as requiring intensified haematological monitoring. Given that anemia has been independently shown to predict mortality in HIV-TB patients beyond CD4 count (Kerkhoff et al., 2016), haemoglobin trajectory should be incorporated as a monitoring parameter in HIV-TB co-management protocols within the region’s integrated HIV-TB services.

The COVID-period findings (84% prevalence vs. 71% in pre-COVID studies) provide evidence that health system disruptions disproportionately worsen anaemia burden in TB patients, likely through mechanisms including reduced nutritional supplementation, delayed diagnosis and treatment initiation, and interruption of community health worker support. This has direct implications for emergency preparedness planning within national TB programmes and should be incorporated into the WHO emergency response guidelines for TB during health system crises.

India’s National TB Elimination Programme (NTEP) and Indonesia’s National Strategic Plan both recommend baseline hemoglobin testing, but lack clear management pathways for identified anemia. The current evidence supports the development of integrated TB-anaemia management protocols that specify haemoglobin thresholds for nutritional supplementation, iron therapy, and specialist haematology referral within TB programme operational guidelines across the region.

Future research should prioritize standardized reporting of key variables including: (1) comprehensive nutritional assessment (BMI, dietary intake, micronutrient levels); (2) systematic comorbidity screening (HIV, diabetes, chronic kidney disease, helminthic infections); (3) detailed TB disease characteristics (extent of disease, cavitation, bacterial load); (4) standardized laboratory methods with quality control; and (5) socioeconomic indicators (education, income, food security, healthcare access). Development of a minimum dataset for TB-anemia research would facilitate more robust meta-analyses and enable better understanding of mechanistic pathways.

We recommend prospective studies incorporating ferritin, transferrin receptor, and inflammatory markers to enable mechanistic classification and inform targeted therapy.

Conclusion

Our systematic review and meta-analysis, the first to comprehensively synthesise anemia burden among TB patients specifically across Southeast Asia, establishes that anemia affects nearly three-quarters of pulmonary TB patients in this high-burden region. The substantial heterogeneity observed reflects genuine epidemiological variation across settings, patient populations, and healthcare contexts rather than methodological inconsistency, and is meaningfully explained by HIV co-infection status, study setting, and the COVID-19 pandemic period. Sensitivity analyses confirm the robustness and stability of the pooled estimate. These findings provide a critical, evidence-based foundation for revising TB diagnostic protocols to include universal anemia screening, strengthening HIV-TB co-management guidelines, and developing integrated TB-anemia management pathways across Southeast Asian national TB programmes. They also identify a clear research agenda: prospective studies characterising anemia type and mechanism, longitudinal assessment of haemoglobin trajectories during treatment, and primary data collection from currently unrepresented countries in the region.

Abbreviations

TB

Tuberculosis

Hb

Hemoglobin

CI

Confidence Interval

WHO

World Health Organization

PRISMA

Preferred Reporting Items for Systematic Reviews and Meta-Analyses

PROSPERO

International Prospective Register of Systematic Reviews

PICOS

Population, Intervention, Comparison, Outcome, Study Design

HIV

Human Immunodeficiency Virus

ART

Antiretroviral Therapy

ACD

Anemia of Chronic Disease

IDA

Iron Deficiency Anemia

MDR-TB

Multidrug-Resistant Tuberculosis

XDR-TB

Extensively Drug-Resistant Tuberculosis

DOTS

Directly Observed Treatment, Short-course

NTEP

National Tuberculosis Elimination Program

JBI

Joanna Briggs Institute

GRADE

Grading of Recommendations Assessment, Development and Evaluation

I²

I-squared (heterogeneity statistic)

LFK

Luis Furuya-Kanamori

JEE

Joint External Evaluation

BMI

Body Mass Index

COVID-19

Coronavirus Disease 2019

SARS-CoV-2

Severe Acute Respiratory Syndrome Coronavirus 2

HDI

Human Development Index

IMSEAR

Index Medicus for the South-East Asia Region

MOOSE

Meta-analysis Of Observational Studies in Epidemiology

Authors’ contributions

V.S. was responsible for the conception and design of the study; V.S., H.A., and D.S.performed data acquisition; V.S. performed data analyses; V.S. interpreted the results; H.K. authored the work, and all authors gave it significant revisions before approving the finished draft.

Funding

The authors received no specific funding for this work

Data availability

The datasets generated and/or analyzed during the current study are not publicly available to protect the privacy of the study. Still, they are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

This systematic review and meta-analysis were based on previously published studies and did not involve direct human participants, interventions, or the collection of primary data. Therefore, ethical approval and individual patient consent were not required.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.