Diabetes and tuberculosis: a systematic review and meta-analyis of mendelian randomization evidence

Abstract

Background

Tuberculosis (TB) and diabetes mellitus (DM) are global health challenges, each imposing significant morbidity and mortality. Observational studies suggest an increased TB risk in individuals with DM, yet causal relationships remain unclear due to potential confounding factors. Mendelian randomization (MR) offers a method to assess causality by leveraging genetic variants as instrumental variables, mitigating biases from confounding and reverse causation. This systematic review aimed to consolidate existing MR evidence on the causal link between DM (types 1 and 2) and TB.

Methods

A comprehensive search was conducted in PubMed, Embase, Google Scholar, and Web of Science, identifying MR studies investigating the causal association between DM and TB. Studies were screened based on pre-specified inclusion criteria and assessed for quality using the STROBE-MR guidelines. Data extraction focused on study characteristics, MR methodology, and causal effect estimates. A meta-analysis was conducted estimate the pooled odds ratios for association between T2DM and TB.

Results

Four MR studies met the inclusion criteria, spanning East Asian and European populations. Findings indicated a consistent causal relationship between DM (particularly type 2 diabetes) and increased TB risk, with odds ratios (OR) ranging from 1.07 to 1.24 (p < 0.05). The pooled odds ratio (OR) was 1.2172 (95% CI: 1.1101–1.3347, p < 0.0001), indicating a significant positive association between T2DM and TB. One study identified pleiotropic effects, suggesting potential genetic overlap in DM and TB susceptibility. No reverse causal association was observed, indicating that TB does not increase the risk of DM.

Conclusion

This review highlights a causal association between DM and TB, emphasizing the need for integrated screening and management of DM within TB control programs, particularly in high-burden regions. Future MR studies should include diverse populations to enhance generalizability and explore genetic mechanisms underlying this association.

Background

Tuberculosis (TB) remains a significant global health challenge, with the World Health Organization (WHO) estimating that 7.5 million people were newly diagnosed with TB and officially notified as a TB case in 2022 with the total number of deaths caused by TB (including those among people with HIV) was 1.30 million down from best estimates of 1.4 million in both 2020 and 2021 and almost back to the level of 2019 [1]. The disease, caused by Mycobacterium tuberculosis, is a leading infectious cause of morbidity and mortality worldwide, disproportionately affecting low- and middle-income countries [2]. Despite concerted public health efforts, TB continues to pose a major burden due to its intricate association with several comorbidities, notably diabetes mellitus (DM) [3]. Diabetes, a non-communicable metabolic disorder, has been rising at an alarming rate globally, with an estimated 537 million adults living with the disease in 2021, killing 1 every 5 s [4]. This increasing prevalence of DM, particularly in regions heavily burdened by TB, has created an epidemiological synergy, complicating TB control efforts.

Emerging evidence suggests that diabetes significantly increases the risk of TB. Individuals with diabetes are approximately three times more likely to develop active TB than those without diabetes [5]. While previous studies have explored the potential risk factors and epidemiological linkages, the causal relationship between DM and TB remains poorly defined. Recent research has provided conflicting evidence on whether DM causally contributes to TB risk or whether their association is merely a reflection of shared risk factors [3, 6, 7]. Observational studies, including cohort and case-control studies, have consistently demonstrated an increased risk of TB among people with DM [3, 7]. However, such studies are susceptible to confounding and reverse causation. This is where Mendelian randomization (MR) approaches offer promise by providing a tool to assess causality especially when randomized controlled trials to examine causality are not feasible and observational studies provide biased associations because of confounding or reverse causality [8].

In the context of DM and TB, MR can help disentangle whether DM directly causes an increased risk of TB or whether the association is confounded by other factors such as obesity, poor glycemic control, or immune suppression. Mendelian randomization is a powerful epidemiological technique that uses genetic variants as instrumental variables (IVs) to estimate the causal effect of an exposure (in this case, diabetes) on an outcome (tuberculosis). The fundamental principle of MR is that genetic variants are randomly assorted at conception and thus not influenced by confounders that typically affect observational studies, such as socioeconomic status or health behaviors [9]. This random allocation of genetic variants mimics the randomization process in controlled trials, offering a more robust means to infer causality.

Our systematic review of Mendelian randomization studies offers a comprehensive means to synthesize the available evidence and evaluate the reported causal relationship between DM and TB. By systematically reviewing and exploring existing MR studies, this review aimed to provide a clearer evidence of the potential causal pathways linking these two diseases. Through this review, we aim to contribute to the growing body of literature on the dual burden of TB and DM.

Materials and methods

This systematic review was conducted in accordance with the guidelines set by the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) [10] and registered on PROSPERO, registration number CRD42024603115. The review aimed to assess the causal relationship between DM and TB from studies that employed MR. The approach involved a comprehensive search, rigorous inclusion and exclusion criteria, data extraction, and quality assessment.

Search strategy

A comprehensive literature search was conducted in PubMed, Embase, Google Scholar, and Web of Science. The search strategy was developed using a combination of keywords related to diabetes, tuberculosis, and Mendelian randomization. The following search terms were used: “Diabetes mellitus” OR “type 1 diabetes” OR “type 2 diabetes”, “Tuberculosis” OR “TB”, “Mendelian randomization” OR “causal inference” OR “genetic instruments”. Boolean term “AND” was used to narrow the search. The search was conducted without restrictions on publication date to include studies up to October 08, 2024. Only studies published in English were considered.

Inclusion criteria

• Study Design: Studies that used Mendelian randomization to assess the causal relationship between diabetes and tuberculosis.

• Population: Studies investigating human populations, including both type 1 and or type 2 diabetes.

• Outcome of Interest: Studies exploring tuberculosis in relation to genetic liability to diabetes.

• MR Methodology: Studies utilizing established MR methods, including two-sample MR, inverse-variance weighted (IVW), MR-Egger, or weighted median approaches.

• Published in Peer-Reviewed Journals: Only full-text articles published in peer-reviewed journals were included.

Exclusion criteria

• Non-Mendelian Randomization Studies: Studies that did not utilize MR methodology or focused on observational or clinical data without causal inference using genetic instruments.

• Animal Studies: Studies conducted in non-human populations.

• Language: Studies not published in English.

• Reviews, Editorials, and Commentaries: Non-original research articles, such as narrative reviews, editorials, or commentaries, were excluded.

• Lack of Full-Text Availability: Studies that were not available in full text or lacked sufficient data for extraction.

Study selection process

Two independent reviewers (IP and RR) conducted the literature search and screened the titles and abstracts of identified studies. Discrepancies between the reviewers were resolved by discussion, and a third reviewer was consulted if necessary.

The screening process was conducted in two phases: Articles were initially screened based on their titles and abstracts to determine potential eligibility then full texts of potentially eligible studies were retrieved and assessed against the inclusion and exclusion criteria. Studies meeting the criteria were included for data extraction. The search and selection process is illustrated in a PRISMA flow diagram (Fig. 1).

Quality assessment

For quality assessment, information was extracted based on a template developed from the Strengthening the Reporting of Observational Studies in Epidemiology using Mendelian Randomization (STROBE-MR) guidelines [11]. Table 1. The STROBE-MR checklist is structured into six sections: Title and Abstract, Introduction, Methods, Results, Discussion, and Other Information. It includes 20 core items and 30 sub-items. The checklist applies to both one-sample and two-sample Mendelian randomization (MR) studies, whether they explore one or several exposures and outcomes. Additionally, it addresses MR studies that follow a genome-wide association study and are presented within the same publication. It encourages authors to explain why MR is appropriate for their research question and to specify their causal hypotheses in advance.

Table 1 Quality assessment using STROBE-MR guidelines

Data extraction

Data from the included studies were extracted using a standardized data extraction form. The following key information was gathered from each study: Study, Ethnicity of the study population, Data source specifying the database from which genetic data were obtained; Genetic Instruments, detailing the specific SNPs or genetic variants used to proxy diabetes; Exposure, identifying whether the study focused on type 1 diabetes, type 2 diabetes; Sample Size, including the total number of cases and controls for both diabetes and tuberculosis; Specified Methods, outlining the specific MR techniques used in each study, such as two-sample MR, inverse-variance weighted (IVW) methods, MR-Egger, weighted median, and MR-PRESSO; and Effect Estimates of Inverse Variant Weighted, expressed as odds ratios (OR) with confidence intervals (CI) and associated p-values, capturing the causal relationship between diabetes and TB. The Conclusions section summarized each study’s overall findings regarding the causal link between diabetes and tuberculosis, including any identified pleiotropy or need for further investigation. (Table 2).

Table 2 Summary of included mendelian randomization studies

Statistical analysis

Meta-analysis was conducted for studies that reported T2DM to pool the causal effect estimates from the included studies using the inverse-variance weighted method. A random-effects model was applied to synthesize the causal effect estimates from the included Mendelian randomization studies. Heterogeneity among studies was quantified using the Q statistic, τ², and I² values. The Q statistic tested whether observed variations in effect estimates were consistent with random sampling error, with significant values indicating the presence of heterogeneity. τ², representing the between-study variance, was estimated using the restricted maximum-likelihood method, which minimizes bias in variance estimation. I² quantified the proportion of total variability attributable to heterogeneity, with values categorized to indicate low (25%), moderate (50%), or high (75%) heterogeneity. Confidence intervals for the pooled estimates were calculated using the Q-profile method, ensuring accurate interval estimation under the random-effects model. A forest plot was generated to visually represent the individual and pooled effect estimates.

Results

Screening and identification of studies for review

The systematic search strategy for this review identified a total of 1,320 relevant studies. After removing duplicates, 1,122 records were screened based on titles and abstracts leading to exclusion of 968 records. The primary reasons for exclusion included studies that did not focus on the specific relationship between diabetes and tuberculosis, utilized methodologies outside of Mendelian randomization, or were abstract-only publications. Following the title and abstract screening, 154 full-text articles were assessed for eligibility. Out of these, 150 articles were excluded for various reasons, including unclear definitions of diabetes or tuberculosis, irrelevant outcomes, and the absence of MR methodology, despite mentioning causal associations in the title or abstract.

Ultimately, this systematic review included four studies that investigated specific causal relationships relevant to our objectives, particularly focusing on diabetes and tuberculosis. These studies met our criteria of employing Mendelian randomization to analyze the impact of diabetes as an exposure variable. (Fig. 1).

Fig. 1

Preferred reporting items for systematic reviews and meta-analyses flowchart for the systematic review of the Mendelian Randomization Studies

Quality assessment of included MR studies

Employing the guidelines from STROBE-MR, we conducted a rigorous quality assessment to evaluate adherence to these standards. After converting the quality assessment scores into percentages, we categorized the 4 studies included in our analysis based on their risk of bias: scores less than 75% indicated a high risk of bias, scores between 75% and 85% suggested a medium risk, and scores exceeding 85% were associated with a low risk of bias Table 1. All 4 studies were of low risk of bias.

Distribution of ethnicity

Available literature have studies only focused on Genome Wide Sequence data from European countries (3) and East Asian countries (1) as highlighted in the map. Figure 2.

Fig. 2

World map showing regions of focus of the reviewed Mendelian Randomization studies

Summary of studies included in the review

The studies included in this review were published between 2023 and 2024, focusing on the causal relationships between DM and pulmonary tuberculosis. The sample sizes ranged from 178,671 to 424,357 participants for PTB and 210,865 to 433,540 participants for DM, with ethnic backgrounds primarily comprising East Asian and European populations. Various genetic instruments were employed, including multiple single nucleotide polymorphisms (SNPs) associated with type 1 and type 2 diabetes. The methodologies utilized were comprehensive, incorporating two-sample Mendelian randomization (TSMR), inverse variance weighted (IVW), and MR-Egger analyses. The findings consistently indicated significant causal relationships between DM and PTB, with odds ratios ranging from 1.07 to 1.24 and p-values indicating statistical significance (p < 0.05) across all studies but one of the 4 studies showed presence of pleiotropy. Reverse direction indicated no causal associations between PTB and DM, thus causal relatioship was primarily unidirectional. The included studies examined T1DM and T2DM separately, utilizing distinct sets of SNPs as genetic instruments. The T1DM study primarily relied on 74 completely unique SNPs, associated with autoimmunity, while T2DM studies used larger SNP panels (ranging from 92 to 273 SNPs) related to metabolic pathways (Table 2). Extracted SNPs from the studies showed no duplicates between the T1DM and T2DM studies, however, for the three T2DM studies, there were 108 duplicates and 415 unique SNPs. A complete list of SNPs used have been provided in Supplementary file S1.

A meta-analysis of the included Mendelian Randomization studies was performed to quantitatively synthesize the causal effect estimates for studies reporting the causal relationships between T2DM and pulmonary tuberculosis. The meta-analysis included three studies, all of which utilized data from BioBank Japan. Across these studies, 108 SNPs were duplicates and 415 were unique. It is possible that results with duplicated SNPs may have been combined during the meta-analysis, however, the random-effects model provided pooled estimates while accounting for variability between studies. The pooled odds ratio (OR) under random-effects models was 1.2172 (95% CI: 1.1101–1.3347, p < 0.0001), indicating a significant positive association between diabetes mellitus (DM) and tuberculosis (TB). Heterogeneity across studies was negligible (I² = 0.0%, τ² = 0), suggesting consistency in the effect estimates (Q = 0.12, p = 0.9407). The results are summarized in the forest plot (Fig. 3).

Fig. 3

Forest plot of pooled estimates

Discussion

This systematic review synthesized the current Mendelian randomization evidence on the causal association between diabetes mellitus, including both type 1 and type 2 diabetes mellitus, and pulmonary tuberculosis. By leveraging genetic instrumental variables, MR offers a powerful tool to evaluate causality and circumvent limitations inherent to observational studies, such as confounding and reverse causation. The findings across the included studies provide significant clues into the DM-TB nexus, reaffirming the clinical implications of DM on TB susceptibility and unveiling potential directions for future research.

We reviewed four studies that highlighted a consistent positive causal relationship between DM and TB, with odds ratios (ORs) ranging from 1.07 to 1.24 across studies [12,13,14,15], all indicating statistical significance (p < 0.05). The meta-analysis of the three studies on T2DM yielded a pooled odds ratio of 1.2172, affirming a significant causal association between diabetes mellitus and tuberculosis. The absence of heterogeneity highlights the consistency of findings across studies, further strengthening the evidence for T2DM as a risk factor for TB. Notably, individual studies reported similar effect sizes (OR range: 1.1910–1.2400), suggesting robustness of the causal relationship despite differences in some genetic instruments and study populations. These findings support the hypothesis that DM increases susceptibility to TB, a conclusion that aligns with prior observational evidence demonstrating a heightened TB risk among diabetic patients [16,17,18]. These findings advance our understanding of how chronic hyperglycemia and related immune dysfunction inherent in diabetes may compromise the host’s defenses against Mycobacterium tuberculosis [19]. Importantly, no study within our review suggested that PTB causally influenced DM, reinforcing that the observed association is primarily unidirectional. One included study on T1DM and PTB explicitly confirmed no reverse causal association between TB and DM [14]. Furthermore, the presence of pleiotropic effects, identified in one T2DM study, indicates a potential overlap in genetic loci influencing both DM and PTB, albeit without negating the causality [12]. It is also important to note that genetic instruments used in MR studies for T1DM and T2DM reflect distinct biological pathways: T1DM SNPs are predominantly linked to immune dysregulation, while T2DM SNPs are associated with insulin resistance and metabolic factors. These differences may influence TB susceptibility through varied mechanisms, such as impaired immune responses (T1DM) or chronic hyperglycemia (T2DM).

The causal association identified between T2DM and PTB stresses an urgent need to consider DM in TB prevention and control programs, especially in high TB-burden countries facing a concurrent rise in DM prevalence. Given the global rise in diabetes, especially within low- and middle-income countries (LMICs) that bear the brunt of the TB burden, integrating DM screening and management into TB control policies is increasingly essential as noted by Denise and colleagues in their review paper [20]. Effective glycemic control, coupled with early TB screening among diabetic patients, could mitigate TB risk, improve patient outcomes, and reduce disease transmission in the community [21,22,23].

This review leveraged MR’s robust methodological advantages to address confounding and reverse causation concerns typical of traditional observational studies. However, certain limitations persist. First, our review found only studies involving East Asian and European populations, limiting the generalizability of these findings to diverse ethnic groups, especially populations in sub-Saharan Africa and Latin America, where the TB-DM co-burden is significant. Future MR studies should therefore target more diverse populations to validate the external applicability of these findings. Another limitation concerns the reliance on summary-level data, which, while valuable, cannot completely account for the distinctions of individual-level variability. Additionally, pleiotropy, whereby genetic variants affect multiple traits, may obscure the causal estimates, as demonstrated by one included study by Chen and group [12], despite employing robust pleiotropy-adjustment methods like MR-Egger and weighted median estimators. This shows the need for advanced methodologies that can more precisely separate causative genetic factors from confounding variants, ensuring the reliability of causal inferences. The meta-analysis is is also limited by the possibility that duplicated SNPs, identified among the included studies, may have influenced the pooled results. While this does not negate the overall findings, it points out the need for caution when interpreting the results. Future studies should explicitly address and account for duplicated genetic instruments to ensure enhanced objectivity and reproducibility of findings.

Given the inherent limitations of MR, integrating functional genomics with MR approaches could elucidate precise causal pathways, especially regarding immune-modulatory genes [24]. Enhanced understanding of these pathways may enable the development of targeted therapies for DM patients to mitigate TB susceptibility. Future studies should also focus on exploring pleiotropy-free inverse variants and conducting sensitivity analyses across diverse genetic consortia to facilitate more accurate causal estimates.

Conclusions

This systematic review emphasizes a causal association between T2DM and PTB risk, highlighting the need for a concerted global health response to the dual epidemics of TB and DM. There is need for more MR studies outside European and East Asian population especially in Sub-Saharan Africa and Latin America where DM-TB burden is high.