Introduction
Tuberculosis (TB) remains a major global health concern, affecting approximately one-third of the world’s population and causing around 2 million deaths annually due to
Mycobacterium tuberculosis infection [
1]. In 2019 alone, global TB cases were estimated between 8.9 and 11 million, with Indonesia ranking second-highest in the world [
2]. TB risk factors are categorized as intrinsic, such as diabetes mellitus, malnutrition, and malignancy, and extrinsic, including high population density, smoking, and alcohol consumption [
3].
The prevalence of diabetes mellitus has risen dramatically, with the World Health Organization (WHO) reporting an increase from 4.7% in 1980 to 8.5% in 2014. Diabetes mellitus has become a leading global non-communicable disease, projected to affect approximately 700 million individuals by 2045 [
3]. Patients diagnosed with both diabetes mellitus and TB experience notably worse outcomes, including higher risks of severe clinical presentations, treatment failure, relapse, and mortality [
4,
5]. Epidemiological studies report that between 5% and 30% of TB patients also have comorbid diabetes mellitus, and individuals with diabetes mellitus face a threefold increased risk of developing TB compared to those without diabetes mellitus [
6]. Diabetes mellitus adversely impacts immune function and influences TB progression by elevating bacterial load and altering the absorption of anti-TB medications, leading to increased treatment complexity [
7].
Emerging evidence suggests chronic inflammation significantly contributes to the TB-diabetes mellitus relationship. Type 2 diabetes mellitus (T2DM) is characterized by persistent systemic inflammation, which is integral to its complications. Studies utilizing inflammatory marker scores indicate that patients with T2DM exhibit elevated systemic inflammatory responses compared to healthy individuals [
8-
10]. Likewise, TB induces excessive immune activation, predominantly driven by neutrophilic responses, leading to lung tissue damage and worsened clinical outcomes. The overlapping inflammatory mechanisms between diabetes mellitus and TB imply that systemic inflammation may underpin the heightened susceptibility to TB observed in patients with diabetes [
11].
Despite the well-established association between TB and diabetes mellitus, existing screening methods remain insufficient, frequently failing to identify patients with diabetes who are at high risk. Many patients with diabetes mellitus and TB present atypically, contributing to higher rates of treatment failure, relapse, and latent TB infections [
12,
13]. Conventional diagnostic tools—such as sputum smear microscopy, nucleic acid amplification tests, and chest X-rays—have limitations, particularly in identifying latent or early-stage TB among patients with diabetes who often exhibit altered immune responses [
14]. Moreover, poor glycemic control complicates TB risk prediction and management, adding complexity to clinical outcomes [
15].
To address this gap, Isfandiari et al. [
16] developed the tuberculosis predictive index (TPI) score, a novel predictive tool incorporating biological, social, and environmental risk factors to estimate TB risk among patients with diabetes. This scoring system provides a structured method for identifying high-risk individuals, potentially improving early detection and targeted screening, particularly in TB-endemic regions like Indonesia. By evaluating the TPI score’s effectiveness, the present study aims to enhance risk stratification and inform prevention strategies for TB in populations with diabetes mellitus.
Materials and Methods
Study Design and Setting
This study used an observational case-control design. Data collection took place at Fatmawati General Hospital from May to August 2024. Medical records from the Internal Medicine Outpatient Clinic at Fatmawati General Hospital, spanning from 2021 to 2024, were reviewed. Additionally, primary data were collected directly from patients through questionnaires administered during follow-up visits in July and August 2024.
Participants
The study targeted adult patients (aged over 18 years) diagnosed with diabetes mellitus, with or without pulmonary TB, who attended the Internal Medicine Outpatient Department at Fatmawati General Hospital between 2021 and 2024. Participants were selected using a consecutive sampling method.
The inclusion criteria were: (1) adult patients aged over 18 years, (2) diagnosis of T2DM, and (3) presence or absence of clinically or bacteriologically confirmed pulmonary TB. Exclusion criteria included: (1) patients with immunocompromised conditions or autoimmune diseases receiving post-transplant care, chemotherapy, or immunosuppressive therapy, and (2) patients with incomplete medical records. Continuous follow-up monitoring was conducted for eligible patients throughout the study period.
Variables
The primary independent variable was the TPI score of patients from the Internal Medicine Outpatient Department at Fatmawati General Hospital. Dependent variables included age, sex, body mass index (BMI), history of contact with TB patients, duration of diabetes mellitus, symptoms resembling pulmonary TB, hemoglobin A1c (HbA1c) levels, housing environment, and psychological well-being.
The TPI score was calculated using the following formula [
16]: −3.218+0.867×age (≤55 years)+1.339×sex (male)+1.493×contact with previous TB patients (yes)+1.089×glycemic control (HbA1c >7%)+1.622×clinical complaints resemble TB symptoms (>3 symptom)+1.183×BMI (≤18.5 kg/m
2)+0.891×duration of T2DM (≤10 years)+0.454×house ventilation width (≤10% house floor width)+0.583×psychological well-being (poor).
Clinical symptoms resembling TB included at least 3 of the following: cough lasting over 2 weeks, hemoptysis, fever, night sweats, unintended weight loss, and reduced appetite. A TPI score below 0 indicated low TB risk among patients with T2DM, whereas a score of 0 or higher indicated higher TB risk.
Diabetes mellitus was diagnosed according to the American Diabetes Association 2024 criteria [
17], defined as fasting plasma glucose ≥126 mg/dL, 2-hour plasma glucose ≥200 mg/dL in an oral glucose tolerance test, HbA1c ≥6.5%, or random plasma glucose ≥200 mg/dL in the presence of hyperglycemic symptoms. Glycemic control was classified as good (HbA1c ≤7%) or poor (HbA1c >7%).
History of TB contact was determined by exposure to an individual diagnosed with TB within the previous 2 years. Psychological well-being was assessed using the WHO-5 Well-Being Index, with scores ≤50 indicating poor well-being and scores >50 indicating good well-being. Age was recorded at data collection, while sex referred to biological sex. BMI was calculated by dividing weight (kg) by height squared (m²). The duration of T2DM reflected the period from the diagnosis to the interview date.
Data Measurement
Data were collected from medical records and patient interviews. In cases of discrepancies between self-reported data and medical records, priority was given to medical records for objective measures such as diabetes status, HbA1c levels, BMI, and TB contact history. For subjective variables like TB-like symptoms and psychological well-being, patient-reported data were accepted unless inconsistencies indicated potential recall bias, in which case validation was conducted through physician confirmation.
Individuals were classified based on diabetes status, recorded as controlled diabetes (0) or uncontrolled diabetes (1). History of TB contact was determined from medical records and patient history, categorized as prior contact with a TB patient (1) or no contact (0). Presence of TB-like symptoms was assessed via medical records and patient history, classified as the presence of 3 or more symptoms (1) or fewer symptoms (0). Random blood glucose levels were measured in mg/dL, and HbA1c was categorized as ≤7% (0) or >7% (1).
Age was documented from medical records and grouped as ≤55 years (1) or >55 years (0). Sex was categorized as male (1) or female (0). BMI, calculated from weight and height using the Asia-Pacific classification, was categorized as ≤18.5 kg/m² (1) or >18.5 kg/m² (0). Housing conditions were evaluated using the Indonesian Healthy House Card criteria, categorizing ventilation as ≤10% of the floor area (1) or >10% (0). Psychological well-being, assessed using the WHO-5 Index, was categorized as poor (≤50; score=1) or good (>50; score=0). Diabetes duration, determined from medical records, was categorized as ≤10 years (1) or >10 years (0).
Bias
Potential bias arose from interviewing patients with both diabetes and TB regarding their TB-related symptoms 2 months after initiating anti-TB treatment, at which point symptoms may have resolved, affecting reporting accuracy. To minimize this bias, symptom evaluation was explicitly limited to the period preceding initiation of anti-TB therapy. This ensured that reported symptoms accurately represented patients' pre-treatment conditions.
Study Size
The sample size calculation aimed to determine the proportion of TB among patients with diabetes mellitus using a formula for unpaired numerical data (
Figure 1). Given the global prevalence of diabetes among TB patients is approximately 13.73% [
9], with a 95% confidence interval (CI) and a 5% type I error rate, the necessary sample size was calculated using a combined standard deviation of 40. Using this formula, the minimum sample size required was determined to be 69 subjects per group.
Quantitative Variables
Quantitative variables analyzed in this study included age, BMI, blood glucose levels, and HbA1c. Each variable was categorized and measured specifically to evaluate their associations with TB risk. To assess the effect of age on TB risk, patients were categorized into 2 groups: older than 55 years, and 55 years or younger. BMI was classified according to Asia-Pacific guidelines into 2 categories: greater than 18.5 kg/m² and 18.5 kg/m² or less, to determine the relationship between weight status and TB risk. Blood glucose levels were measured in milligrams per deciliter, while HbA1c values were reported as percentages and categorized as either less than 7% or 7% or higher to evaluate long-term glycemic control. These categorizations were selected based on established clinical thresholds to ensure relevance and consistency when analyzing the impact of these quantitative variables on TB risk.
Statistical Methods
All baseline data and study outcomes were stored electronically for statistical analysis. Data were analyzed using SPSS version 25. Descriptive statistics were employed to determine frequency distributions, percentages, means, and ranges of observed variables. Bivariate analysis was performed to explore the distribution and associations between independent and dependent variables. Chi-square tests were applied to assess differences in TPI scores between the groups with and without TB.
To estimate the association between TB risk and predictor variables, logistic regression analysis was conducted. Both the composite TPI score and individual components—including age, sex, TB contact history, glycemic control, presence of TB-like symptoms, BMI, duration of diabetes, housing ventilation, and psychological well-being—were analyzed as predictors of TB risk. Odds ratios (ORs) with 95% CIs were calculated to quantify the strength of these associations. Cases with missing data were excluded from analyses to maintain statistical accuracy.
Ethics Statement
This study received approval from the Director of Human Resources, Education, and Research of Fatmawati General Hospital (approval number: DP.04.03/DXXI.2/7392/2024) on July 19, 2024. The research adhered to bioethical principles, including autonomy, beneficence, nonmaleficence, and justice. Written informed consent was obtained from all participants for publication of the study results.
Results
Participants
The minimum required sample size calculated for each group was 69 subjects. Initially, 70 control subjects and 79 case subjects were recruited. However, following a review of medical records, there was an insufficient number of eligible cases, prompting an adjustment in the case-to-control ratio to approximately 1:2. As a result, the final analysis included 70 controls (patients with diabetes) and 39 cases (patients with both diabetes and TB), totaling 109 participants. The study flow is detailed in
Figure 2.
Descriptive Data
Participants’ characteristics, including age, sex, TB contact history, presence of TB-like symptoms, HbA1c levels, duration of diabetes, BMI, housing ventilation quality, psychological well-being, and overall TPI scores, are summarized in
Table 1. Among the 39 patients with diabetes and TB, 41% were aged ≤55 years, compared to 74.3% among those without TB. There was a higher proportion of men among patients with diabetes and TB (53.8%) compared to patients with diabetes but not TB, who were predominantly female (64.3%). Both groups had a majority without a history of TB contact (56.4% in the TB group and 67.1% in the non-TB group). Notably, TB-like symptoms were much more frequent among patients with diabetes and TB (76.9%) compared to those without TB (20%).
HbA1c levels ≥7% were similarly common in both groups, present in 66.7% of patients with TB and 60% of patients without TB. Most patients had diabetes for more than 10 years (59% of patients with TB, 52.9% of patients without TB). Most participants in both groups had a BMI >18.5 kg/m² (79.5% of patients with TB, 92.9% of patients without TB). Good housing ventilation (>10%) was more common among patients with diabetes but not TB (91.4%) compared to those with TB (76.9%). Psychological well-being scores were positive (good) for 53.8% of patients with TB and 57.1% of patients without TB. Significantly, 82.1% of patients with diabetes and TB had high TPI scores, whereas 60% of patients with diabetes but not TB had low TPI scores.
TB Risk Based on TPI Score
The distribution of TB risk according to TPI scores is summarized in
Table 2. Patients were categorized into high- and low-risk groups based on their TPI scores. Among patients with diabetes and TB, 32 (82.1%) had high TPI scores, compared to only 28 of patients (40.0%) with diabetes but not TB. Conversely, low TPI scores were found in 7 patients (17.9%) with diabetes and TB and in 42 patients (60.0%) with diabetes but not TB. The association between high TPI scores and TB was statistically significant (OR, 6.8; 95% CI, 2.6–17.6;
p<0.001), confirming that patients with diabetes and TB were substantially more likely to have high TPI scores compared to those without TB. This finding aligns with prior evidence that diabetes significantly increases the risk of active TB [
9].
TB Risk Based on TPI Score Variables
All variables analyzed in
Table 3—age, sex, TB contact history, glycemic control (HbA1c), presence of TB-like symptoms, BMI, diabetes duration, housing ventilation, and psychological well-being—are components of the original TPI formula [
16]. Thus, the strong associations observed in this study are anticipated outcomes. For example, patients aged 55 or younger had an OR of 4.9 compared to those older than 55, and males had an OR of 4.0, reflecting the weights assigned to younger age and male sex in the original TPI calculation [
16]. Similarly, the very high odds ratio (OR) observed for TB-like symptoms (OR, 22.5; 95% confidence interval [CI], 7.0–71.0) and the elevated OR for poor glycemic control (HbA1c >7%) (OR, 2.8; 95% CI, 1.2–6.3) are consistent with their large coefficients in the predictive index formula [
16]. Additionally, shorter diabetes duration (≤10 years) was associated with a higher TB risk (OR, 2.5; 95% CI, 1.1–5.9), and poor housing ventilation (≤10% of floor area) showed a substantial association (OR, 6.5; 95% CI, 1.4–30.4), confirming their roles as significant predictors in the TPI model. Conversely, psychological distress was not significantly associated with increased TB risk (OR, 1.7; 95% CI, 0.8–3.2;
p=0.165), which aligns with its relatively low weighting in the TPI score. These findings confirm that each component of the TPI score behaves in accordance with its intended predictive role, thereby validating the overall design of the index rather than identifying novel independent risk factors [
16].
TB Risk Based Among Patients with Diabetes
Among individuals living with diabetes (
Table 4), three key factors were significantly associated with increased TB risk. Firstly, individuals reporting TB-like symptoms had markedly higher odds of TB (OR, 13.3; 95% CI, 5.1–34.3;
p<0.001) compared to those without symptoms. This strong association highlights the clinical value of symptom screening in detecting TB among people with diabetes, though it does not imply that symptoms cause TB. Indeed, clinical guidelines advocate vigilant monitoring of TB symptoms in this population as an essential case-finding strategy [
17–
20]. Secondly, undernutrition was a significant risk factor. Individuals with a body mass index (BMI) below 18.5 kg/m² had significantly increased odds of TB (OR, 3.3; 95% CI, 1.0–11.0;
p=0.039) compared to those with normal or higher BMI. This finding is consistent with systematic reviews showing that undernutrition can nearly double TB risk [
21–
23]. Thirdly, inadequate housing ventilation (≤10% of floor area) was associated with higher TB risk (OR, 3.2; 95% CI, 1.4–9.8;
p=0.008), reaffirming established relationships between substandard housing and airborne TB transmission [
24–
26]. Finally, poor housing ventilation (≤10%) was also associated with higher TB risk (OR, 3.2; 95% CI, 1.4–9.8;
p=0.008), reflecting established relationships between inadequate housing conditions and increased TB transmission [
9,
27,
28]. Collectively, these results emphasize the importance of prompt TB evaluation in patients with diabetes presenting with respiratory symptoms, and highlight the significance of addressing modifiable risk factors such as nutrition and living conditions for effective TB prevention within this vulnerable population.
Discussion
This study demonstrates that the TPI score is an effective and practical tool for assessing TB risk in patients with T2DM. Our findings confirm that the TPI score reliably differentiates high-risk patients with diabetes from those at lower risk, offering a valuable mechanism for early screening and timely intervention. Patients with diabetes diagnosed with TB had significantly higher TPI scores, corresponding to a 6.8-fold increase in the odds of TB compared to patients with diabetes but not TB. This underscores the potential clinical utility of the TPI score as a proactive screening tool, particularly among patients with diabetes who may not yet present overt TB symptoms but remain at considerable risk.
Standard TB screening among patients with diabetes generally depends on symptom-based assessments or bacteriological confirmation, potentially delaying diagnosis, especially in subclinical or early-stage cases. The TPI score overcomes this limitation by integrating multiple established risk factors—biological characteristics (e.g., age and sex), metabolic factors (e.g., glycemic control and BMI), and environmental conditions (e.g., housing ventilation)—into a comprehensive predictive model [
16]. Utilizing the TPI score in clinical practice could enhance TB detection among populations with diabetes, allowing more targeted investigations and earlier therapeutic interventions.
All variables used in the TPI calculation—including younger age (≤55 years), male sex, poor glycemic control, low BMI, TB-like symptoms, short diabetes duration, inadequate housing ventilation, and psychological distress—are integral, weighted components of the original predictive index [
16]. Thus, their significant associations with TB in this study confirm the validity of the TPI score rather than introducing new independent risk factors. Among these variables, the strongest association observed was with the presence of TB-like symptoms. However, this association likely indicates reverse causation, in which the presence of TB itself results in symptom manifestation rather than symptoms independently predicting future TB risk. Nevertheless, this symptom-based association remains clinically significant for effective case detection [
18,
19]. Our data reveal that 76.9% of patients with diabetes and TB presented with 3 or more TB-like symptoms, compared to only 20% of patients with diabetes but not TB. This highlights the clinical importance of symptom-based triage but also supports employing multifactorial tools like the TPI score for comprehensive screening to avoid missing asymptomatic or early cases [
18–
20].
Metabolic factors such as elevated HbA1c (>7%) and low BMI (≤18.5 kg/m²) were also significantly associated with increased TB risk. These findings align with previous research indicating hyperglycemia impairs immune responses and undernutrition increases susceptibility to TB [
17,
20]. Nonetheless, since these factors are integral to the TPI score itself, the associations observed here reinforce the score’s validity rather than identify novel independent risk factors.
Duration of diabetes was also a contributing factor. Patients with diabetes for ≤10 years exhibited lower odds of TB, though some of these individuals still scored highly on the TPI, suggesting early metabolic derangements in a subset of patients. This finding emphasizes the importance of proactive TB risk assessment early in the diabetes disease course [
22,
23].
Lastly, environmental conditions such as inadequate housing ventilation were strongly associated with TB, consistent with prior findings linking housing quality to TB transmission [
20,
24]. In contrast, psychological well-being, although included in the TPI score, did not show a significant association in this sample. This suggests it may have less predictive utility in similar clinical settings.
Environmental factors, particularly housing ventilation, significantly impacted TB risk. Patients with diabetes living in poorly ventilated homes (ventilation area ≤10% of floor area) had notably higher odds of TB, with an OR of 6.5 based on TPI score stratification. Within our study, 23.1% of patients with TB lived in poorly ventilated homes, compared to 8.6% of patients without TB (OR, 3.2; 95% CI, 1.0–9.8;
p=0.035). These findings are consistent with previous studies linking inadequate ventilation to increased airborne TB transmission [
27], underscoring the need to integrate environmental assessments into TB screening and preventive strategies.
A history of TB contact was also significantly associated with TB risk in patients with diabetes; those reporting previous TB exposure exhibited a 3.2-fold increased risk. While prior TB contact is an established risk factor, our findings suggest that metabolic and environmental factors contribute more substantially to TB susceptibility among patients with diabetes. Conversely, psychological well-being did not significantly correlate with TB risk in this study, despite its recognized role in chronic disease management. This finding suggests that mental health alone might have limited predictive utility regarding TB susceptibility in similar clinical populations.
Our study highlights the potential clinical value of implementing the TPI score. Routinely incorporating the TPI score in diabetes management could help healthcare providers efficiently identify patients at elevated TB risk who would benefit from closer monitoring, earlier diagnostic testing, or preventive interventions. If validated by future research, the TPI score could become an accessible and practical tool to optimize early TB detection, improve patient outcomes, and enhance preventive measures within diabetes care.
Despite the promising results, this study has limitations. First, data collection relied partially on self-reported questionnaires, introducing potential recall bias, especially concerning TB-like symptoms, previous TB exposure, and housing conditions. While attempts were made to mitigate bias through medical record cross-referencing, self-reporting remains susceptible to individual perception and recall issues. Second, as a case-control design, this study establishes associations but cannot demonstrate causality; longitudinal or cohort studies would be necessary to determine whether elevated TPI scores directly predict future TB incidence.
Additionally, the single-center setting limits generalizability. Although Fatmawati General Hospital serves a diverse patient population, multi-center validation studies are essential to confirm the TPI score’s applicability across various clinical and geographic contexts. Further, the TPI scoring system itself is novel and unvalidated externally; additional research is required to confirm its reliability and validity beyond the initial study setting. Finally, constraints such as limited sample size—particularly of patients with both diabetes and TB—and time limitations necessitated adjusting the study design to a 1:2 case-to-control ratio, which may influence the findings’ robustness.
