Nonlinear association between advanced lung cancer inflammation index and latent tuberculosis infection risk: threshold effects and predictive value of a novel biomarker.

Introduction
Tuberculosis (TB) is an infectious disease caused by Mycobacterium tuberculosis and continues to be one of the leading causes of mortality globally (1). In 2023, it was estimated that there were approximately 10.8 million cases and 1.25 million deaths attributed to TB worldwide (2). Latent TB infection (LTBI) is characterized by a persistent immune response to Mycobacterium tuberculosis in the absence of clinical symptoms or evidence of active disease. According to the World Health Organization (WHO), approximately 23% of the global population, or around 1.7 billion individuals, are estimated to have LTBI. A small proportion of this group (approximately 5–10% of those with LTBI) may progress to active TB during their lifetime (3,4), which may increase the risk of transmission, accelerate the spread of the disease, and significantly raise the risk of death. Most active TB results from reactivation of LTBI (5). Therefore, enhancing the detection and treatment of LTBI can contribute to reducing the reservoir of potential TB cases, thereby effectively decreasing TB incidence and ultimately supporting global efforts toward TB elimination (6). However, standard diagnostic tests like interferon (IFN)-gamma release assays (IGRAs), while effective for detection, are costly and do not adequately stratify this risk, creating a need for accessible, low-cost biomarkers to help identify individuals with higher susceptibility who could be prioritized for preventive therapy.
The immune response to M. tuberculosis is strongly shaped by the interaction between nutritional status and systemic inflammation. Malnutrition compromises cell-mediated immunity, a critical component of TB control (7), while chronic inflammation promotes immune dysregulation that favors pathogen persistence (8). Although individual biomarkers have been used to assess these domains—such as the neutrophil-to-lymphocyte ratio (NLR) for inflammatory balance and serum albumin for long-term nutritional status (9-11)—their single-dimensional nature limits their ability to reflect overall physiological resilience. A composite index integrating both pathways may therefore provide a more robust and comprehensive assessment of immunocompetence. The advanced lung cancer inflammation index (ALI), initially developed in oncology to predict prognosis, represents such an integrative measure (12). It combines indicators of nutritional reserve [body mass index (BMI), albumin] with markers of acute inflammation (neutrophil and lymphocyte counts). Despite its origins in cancer research, the conceptual framework linking nutritional and inflammatory status to resilience is directly applicable to chronic infectious diseases like TB (13-19).
Therefore, this study aims to examine the relationship between ALI and LTBI using data from the National Health and Nutrition Examination Survey (NHANES) 2011-2012. We hypothesize that an elevated ALI is associated with a reduced risk of LTBI. We present this article in accordance with the TRIPOD reporting checklist (available at https://jtd.amegroups.com/article/view/10.21037/jtd-2025-1508/rc).

Methods

Data source and study participants
The NHANES, conducted biennially by the USA Centers for Disease Control and Prevention (CDC), is a nationally representative cross-sectional survey designed to assess the health and nutritional status of non-institutionalized USA civilians using a complex stratified sampling approach. This comprehensive program integrates in-home interviews, physical examinations conducted at mobile examination centers, and laboratory analyses of biological specimens (including blood samples) to gather data on demographics, dietary intake, chronic disease prevalence, biomarkers, and environmental exposures. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments, and all collected data were de-identified by the National Center for Health Statistics before being made publicly available for research purposes.
The complete data related to TB were collected during the 2011–2012 period, which is the most recent cycle that included both QuantiFERON®-TB Gold In Tube (QFT-GIT) and tuberculin skin test (TST) for measuring LTBI. Consequently, this study exclusively utilized publicly available data from NHANES 2011–2012. In this dataset, a total of 9,756 participants were included. Participants under the age of 20 years, those with self-reported active TB (aTB), and those lacking data on TST, QFT-GIT, ALI, and covariates were excluded. Finally, our study included 3,010 participants (Figure 1).

Definition of LTBI
In our study, LTBI was diagnosed based on either a positive TST or QFT-GIT result. For the TST procedure, NHANES-trained phlebotomists administered tuberculin antigen injections on the palmar side of participants’ arms, with reactions assessed 46–76 hours later by physicians unaware of participants’ medical history or TB exposure. A positive TST was defined as induration ≥10 mm according to USA adult criteria, excluding individuals with special risk factors. Regarding QFT-GIT interpretation, we strictly followed NHANES guidelines requiring three criteria: (I) Nil control value ≤8.0 IU IFN-γ/mL; (II) TB antigen value minus Nil value ≥0.35 IU IFN-γ/mL; and (III) this differential value ≥25% of the Nil value (9,20).

Definition of ALI
The primary exposure variable of interest was ALI. ALI was calculated as BMI (kg/m2) × albumin (g/dL)/NLR. BMI = weight in kilograms/(height in meters)2. NLR = neutrophil counts/lymphocyte counts. The ALI was categorized according to tertiles: T1 group (ALI ≤30.82), T2 group (30.82< ALI ≤55.50), and T3 group (ALI >55.50). The calculation formulas for other biomarkers reflecting nutritional and inflammatory status are as follows:
Prognostic nutritional index (PNI) = 10 × serum albumin (g/dL) + 5 × lymphocyte count (109/L);
Systemic immune-inflammation index (SII) = (neutrophil count × platelet count)/lymphocyte count.

Covariates
This study, informed by prior research and clinical expertise, systematically incorporated covariates that could potentially influence the association between ALI and LTBI: age, gender (male, female), race (Mexican American, Other Hispanic, Non-Hispanic White, Non-Hispanic Black, other races), marital status (married/living with partner, never married, widowed/divorced/separated), education level (under high school, high school or equivalent, above high school), poverty income ratio (PIR), smoking (never, former, current), drinking (yes, no), hypertension (yes, no), diabetes (yes, no), physical activity (PA: high, low), and total cholesterol (TC, mg/dL).
Smoking status was categorized as never (<100 lifetime cigarettes), former (≥100 lifetime cigarettes but quit smoking), or current (≥100 lifetime cigarettes and still smoking) (21). Drinking status was classified as no (<12 lifetime drinks) or yes (≥12 lifetime drinks) (22). Hypertension was defined as a self-reported prior diagnosis, the use of antihypertensive medication, or a blood pressure reading of ≥140/90 mmHg. Diabetes was defined as a self-reported prior diagnosis, a fasting plasma glucose level of ≥7 mmol/L, a glycosylated hemoglobin A1c (HbA1c) level of ≥6.5%, a 2-hour post-load plasma glucose level of ≥11.1 mmol/L, or the use of insulin or oral hypoglycemic medications. We calculated the total PA score according to how many days per week and how much time per day participants did different intensities of PA. Subsequently, we divided the participants into three groups (tertiles) and categorized those in the upper tertile as the high PA group, while the remaining participants were classified as the low PA group.

Statistical analysis
Continuous and categorical variables were expressed as the median along with interquartile range (IQR) and counts with percentages. Differences between groups were evaluated using the Mann-Whitney U-test or the Kruskal-Wallis test for continuous variables, and the Chi-squared test or Fisher’s exact test for categorical variables. Two-sided P<0.05 indicates statistical significance. All statistical analyses were performed with R software Version 4.4.2.
Three distinct logistic regression models were constructed to examine the association between ALI and LTBI. Model 1 was unadjusted for covariates, Model 2 adjusted for age, gender, and race, while Model 3 further adjusted for education, marital status, PIR, drinking, smoking, PA, diabetes, hypertension, and TC. Results were expressed as odds ratios (OR) with corresponding 95% confidence intervals (95% CI). Additionally, potential nonlinear relationships between ALI and LTBI were investigated by fitting restricted cubic spline (RCS) curves. Threshold effect analysis was performed using the inflection point identified through RCS modeling. To strengthen the robustness of our findings, stratified analyses in conjunction with interaction tests were conducted to comprehensively evaluate potential interactions between ALI and stratification variables. Moreover, we analyzed the mediating role of ALI in the smoking-LTBI pathway and the PIR-LTBI pathway, and causal mediation analysis with 1,000 bootstrap resamples was performed to estimate mediating effects and 95% CIs. Finally, the clinical predictive value of ALI was evaluated by receiver operating characteristic (ROC) curve analysis in comparison with other similar indicators after adjusting for all covariates.
It is important to note that, given its right-skewed distribution, ALI was subjected to a log2 transformation in all subsequent regression analyses when modeled as a continuous variable. This approach aimed to enhance the accuracy and reliability of the results by reducing skewness and stabilizing variance.
As this was a secondary analysis of an existing dataset, a formal a priori sample size calculation was not conducted. However, a post-hoc power analysis was performed using G*Power software (version 3.1) to assess whether the final sample size provided sufficient statistical power for the primary analysis. Based on a two-tailed logistic regression model, a Type I error rate (α) of 0.05, the observed OR for the association between ALI and LTBI, the prevalence of LTBI, and the sample size, the achieved statistical power was calculated. The overall power was 99.26%, with subgroup power ranging from 5.70% to 99.70%, as detailed in Table S1.

Results

Descriptive results
A total of 3,010 participants were included in this study, with a median age of 45 years (IQR: 31–60 years), and 53.82% of participants (n=1,620) were male. Table 1 displays the baseline characteristics of the study population stratified by LTBI status, whereas Table 2 provides the baseline characteristics of each group categorized by tertiles of ALI levels. Compared with those without LTBI, individuals with LTBI were more likely to be older, male, Mexican American, Other Hispanic, or other races, with lower educational attainment, poorer economic status, be married or living with a partner, higher TC and ALI levels. They were also more often physically inactive, non-drinkers, and with diabetes and hypertension (all P<0.05). There were no significant differences in BMI, smoking, albumin level, neutrophil count, lymphocyte count, or platelet count (all P>0.05). Compared to the lowest ALI tertile, participants in the higher ALI tertiles were more likely to be older, non-Hispanic White, with lower educational attainment, poorer economic status, and have higher BMI, higher TC and albumin levels, and higher neutrophil, lymphocyte, and platelet counts. They were also more frequently widowed/divorced/separated, current or former smokers, and diagnosed with diabetes and hypertension (all P<0.05). We observed no significant differences in gender, drinking, and PA (all P>0.05).

Association between ALI and LTBI
We employed three different logistic regression models to examine the association between ALI and LTBI. The logistic regression results (Table 3) confirmed a significant inverse association between ALI and LTBI risk, which was statistically significant in all three models (Model 1: OR =0.709, 95% CI: 0.573–0.876; Model 2: OR =0.747, 95% CI: 0.593–0.941; Model 3: OR =0.765, 95% CI: 0.601–0.973). When ALI was divided into tertiles, compared to participants in the first ALI tertile, those in the third tertile had a lower risk of LTBI across all models (Model 1: OR =0.607, 95% CI: 0.464–0.790; Model 2: OR =0.627, 95% CI: 0.471–0.832; Model 3: OR =0.637, 95% CI: 0.474–0.852). Moreover, the OR for LTBI decreased with increasing ALI in each model (P for trend <0.05).

Analysis of nonlinear and threshold effect between ALI and LTBI
The RCS curves revealed a non-linear relationship between log2-ALI and LTBI risk starting with the unadjusted model (Figure 2A). The shape of the curve remained largely consistent in Model 2 (Figure 2B) and persisted in the fully adjusted model (Figure 2C, P for nonlinear =0.04). Although the P value (0.07) for the non-linear test in Model 2 did not reach statistical significance, it was close to the conventional threshold of 0.05, indicating a potential non-linear trend that warrants cautious interpretation. We then fit a piecewise logistic regression model based on the two inflection points (5.05; 5.90) identified from the RCS curve of the fully adjusted model. The segmented regression results (Table 4) showed that for log2-ALI values below 5.05, there was no statistically significant association between log2-ALI and the odds of LTBI (OR =1.628, 95% CI: 0.809–3.545, P=0.20). However, when log2-ALI was greater than 5.05 and less than 5.90, each unit increase in log2-ALI was associated with a 67.1% lower odds of LTBI (OR =0.329, 95% CI: 0.200–0.535, P<0.001). Each unit increase in log2-ALI indicated an 82.3% reduction in the odds of LTBI when log2-ALI was greater than 5.90 (OR =0.177, 95% CI: 0.056–0.480, P=0.001). The results of the log-likelihood ratio test demonstrated a significant difference between the standard model and the segmented model (P=0.002), thereby providing evidence for the presence of a threshold effect.

Association between ALI and LTBI stratified by gender, age, race, PIR, smoking status, drinking status, and history of hypertension and diabetes
Subgroup analyses and interaction tests based on select covariates were conducted to evaluate the robustness of our findings and to identify potential susceptible populations (Figure 3). Significant negative relationships were observed among males, participants without diabetes and those with hypertension (all P<0.05). In addition, no significant interaction was observed between ALI and the stratified variables for LTBI (P for interaction >0.05).

Mediation of ALI in the association of smoking and PIR with risks of LTBI
Given the associations of smoking and income level (PIR) with both ALI (as shown in Table 2) and LTBI, we conducted an exploratory analysis to investigate whether ALI might play a mediating role in these pathways. Consequently, we analyzed the mediating role of ALI in these two pathways. In the smoking-LTBI pathway, ALI played a mediating role in the unadjusted model (Figure 4A; P=0.02, mediation proportion: 6.2%). However, no significant mediating effect was observed in the subsequent adjusted models (Figure 4B,4C; P>0.05). Regarding the PIR-LTBI pathway, no mediating effect was found in the unadjusted model (Figure 4D; P=0.13). A significant mediating role was observed in the partially adjusted model (Figure 4E; P=0.03, mediation proportion: 2.6%), but this effect was not significant in the fully adjusted model (Figure 4F; P=0.25). Given the low mediation proportions and the lack of consistency across models, these findings should be interpreted with caution and are considered exploratory, providing limited support for a substantive mechanistic role of ALI.

Comparison of the predictive value of inflammation/nutrition-related indicators for LTBI
Adjusted ROC curve analysis was employed to evaluate and compare the diagnostic efficacy of different inflammation/nutrition-based indicators in identifying LTBI. Table 5 illustrates that the area under the curve (AUC) for ALI in predicting LTBI was 0.7809 (95% CI: 0.7575–0.8044), which was marginally higher than the AUCs for PNI (0.7794, 95% CI: 0.7558–0.8031), SII (0.7797, 95% CI: 0.7561–0.8033), BMI (0.7798, 95% CI: 0.7562–0.8034), albumin (0.7802, 95% CI: 0.7566–0.8037), and NLR (0.7798, 95% CI: 0.7562–0.8035). DeLong’s test confirmed the absence of statistically significant difference. The potential value of ALI may reside less in its statistical performance and more in its conceptual framework. Unlike single biomarkers, ALI was designed to integrate multiple physiological dimensions by simultaneously capturing nutritional status and systemic inflammatory response. This integrative approach offers a more comprehensive assessment of a patient’s overall physiological state—a feature particularly relevant given the established roles of both nutrition and inflammation in the pathogenesis of LTBI. Therefore, despite its limited advantage in discriminative accuracy, ALI may serve as a more holistic biomarker, reflecting the complex interplay between nutritional status, inflammation, and LTBI risk.

Discussion
In this study, we investigated the association between ALI and LTBI. Our findings suggest a complex, non-linear relationship between ALI and LTBI. After adjusting for relevant covariates, we found that higher levels of ALI were significantly associated with a reduced risk of LTBI, particularly among males, individuals without diabetes, and those with hypertension. The RCS analysis revealed a non-linear association between log2-ALI and LTBI, which was further confirmed by threshold effect analysis. Our exploratory analysis suggested that ALI partially mediated the associations of smoking and PIR with LTBI. The results of the ROC curve analysis showed that although ALI exhibited a marginally higher AUC compared to other indicators, this difference was not statistically significant.
Hematological parameters derived from routine blood tests, such as SII and NLR, are widely acknowledged for their capacity to reflect systemic inflammation and immune status. Huang et al. demonstrated that NLR was negatively correlated with LTBI and positively associated with all-cause mortality (11). Sheng and colleagues also identified a significant negative correlation between NLR levels and individual LTBI risk, indicating that NLR may serve as a potential biomarker for assessing LTBI risk (23). Another study utilizing the NHANES 2011–2012 database revealed a significant negative correlation between systemic inflammatory response index (SIRI) and LTBI (10). In this study, ALI-a novel biomarker reflecting nutritional inflammatory status calculated based on BMI, serum albumin, neutrophil, and lymphocyte counts-was used as the exposure variable. We observed a significant inverse association between ALI levels and LTBI risk, which aligns with the findings reported in the aforementioned studies. We also found that this significant inverse association persisted among men, participants without diabetes, and those with hypertension. Male participants were more likely to be current or former smokers, potentially leading to lower ALI levels and an increased risk of LTBI. Diabetes is known to directly elevate LTBI risk through chronic inflammation and immune dysfunction, which may obscure the protective role of ALI (24). After accounting for hyperglycemia-induced interference, the protective effect of ALI as a marker of inflammation and nutritional status on immune function became more pronounced. Compared with healthy individuals, the elevated level of inflammation in hypertensive patients might serve as one of the protective factors against the effects of Mycobacterium tuberculosis (23).
ALI and LTBI exhibited a non-linear negative relationship. The possible biological mechanism underlying this association can be explained from several aspects: firstly, the ALI is calculated by multiplying BMI by albumin and dividing by NLR, all of which are associated with nutritional and inflammatory status in LTBI patients. Patients with LTBI often exhibit malnutrition, inflammation, and metabolic disorders due to immune dysfunction. These characteristics underscore the critical need for developing a comprehensive index that integrates nutritional and inflammatory markers to optimize treatment strategies and improve prognostic assessment in LTBI patients. The ALI serves as precisely such a holistic and integrated indicator. BMI is one of the most widely used and straightforward methods for nutritional assessment in clinical practice, and a low BMI is generally associated with malnutrition. Badawi et al. found an inverse relationship between BMI and the prevalence of LTBI (25). Low BMI was a risk factor for progression from LTBI to active TB disease (26). The aforementioned studies suggested that a higher BMI exerted a protective effect. Consequently, we proposed that BMI played a crucial role in this non-linear inverse association. Maintaining an elevated yet healthy BMI may reflect improved nutritional status, which could contribute to enhanced immune function and subsequently mitigate inflammation and associated complications (27). Second, serum albumin is a well-established marker for assessing nutritional status and exhibits anti-inflammatory properties, playing a critical role in the pathogenesis of TB infection (28). Furthermore, previous studies had demonstrated that activated white blood cells released reactive oxygen species (ROS) via neutrophils and cytokines, thereby promoting systemic inflammation and endothelial dysfunction (29). ALI is directly proportional to BMI and albumin, and inversely proportional to NLR. A high level of ALI indicates a better nutritional status and a lower level of inflammation. On the one hand, good nutritional status can promote the secretion of interleukin-1β (IL-1β) and further modulate paracrine signaling pathways, thereby strengthening the role of macrophages in innate immune regulation (30). Active macrophages directly inhibit the resuscitation of Mycobacterium tuberculosis from the dormant state through the continuous production of bactericidal molecules [such as ROS/reactive nitrogen species (ROS/RNS), antimicrobial peptides] and autophagic degradation. On the other hand, the immune response to Mycobacterium tuberculosis is heavily reliant on cell-mediated immunity, and an elevated NLR may reflect relative lymphopenia (11). Consequently, the clinical progression of LTBI is likely influenced by analogous mechanisms.
The findings of this study offer several potential clinical implications. First, our analysis identifies ALI as a composite indicator associated with LTBI risk, particularly in males, individuals without diabetes, and those with hypertension. While not statistically superior to its individual components in predictive power, its value lies in integrating both nutritional and inflammatory status into a single metric. Crucially, our study demonstrated a significant non-linear relationship, indicating that the protective effect of ALI is not constant. This complex relationship suggests that LTBI risk is lowest within an optimal range of ALI values, rather than simply decreasing as ALI rises indefinitely. This finding highlights the importance of maintaining a balanced nutritional and inflammatory status, as reflected by ALI, for mitigating LTBI risk. Therefore, strategies that address the components of ALI, such as nutritional support or inflammation regulation (e.g., smoking cessation), warrant further investigation as a potential approach for LTBI prevention. Finally, given its low cost and holistic nature, ALI could serve as a complementary tool in the overall clinical assessment of individuals at risk for LTBI, providing additional physiological context alongside established tests like IGRA.
There are several limitations in this study. First, as NHANES employs a cross-sectional design, it cannot establish a causal relationship or determine the direction of association between ALI and LTBI risk. Second, although we adjusted for many relevant covariates, we cannot entirely exclude the possibility of residual confounding or the influence of unmeasured confounders, for example, environmental factors and genetic factors. Thirdly, although NHANES is representative, findings may not generalize to populations outside the USA. Future research should include diverse racial and regional cohorts to enhance generalizability. Fourth, the ROC curve analysis was conducted on the complete dataset without internal validation (e.g., bootstrap or cross-validation) or external validation in an independent cohort. The reported performance metrics may be overly optimistic due to potential overfitting; thus, our findings are descriptive of this cohort and cannot be generalized without further validation. Furthermore, although the overall sample size provided excellent statistical power, the power was inherently reduced in certain subgroup analyses. Therefore, findings from these subgroup comparisons should be regarded as exploratory and interpreted cautiously until replicated in larger, specifically designed studies. Lastly, the diagnosis of LTBI using both TST and QFT-GIT has limitations in specificity and sensitivity, which may affect the accuracy of the assessment. It is particularly important to note that TST cannot reliably distinguish individuals with LTBI from those vaccinated with Bacillus Calmette-Guérin (BCG), and the development of tests such as QFT-GIT is aimed to overcome this limitation of TST (31). However, in our study, the diagnosis of LTBI was based on a positive result from either TST or QFT-GIT, which may introduce a certain degree of misclassification and thereby impact the reliability of the findings. Future studies could adopt more precise diagnostic tools or combine multiple methods to improve the robustness of LTBI diagnosis.

Conclusions
This study demonstrates that higher ALI levels are significantly associated with reduced LTBI risk and exhibit a non-linear relationship. The findings link the integrated inflammation-nutrition status, as measured by ALI, to LTBI risk, suggesting that this pathway represents a potential target for preventative interventions.

Supplementary
The article’s supplementary files as