A Nomogram Predictive Model for Drug-Resistant Tuberculosis Detection by GeneXpert MTB/RIF Assay in Paediatric Patients.

OBJECTIVE

To develop and validate a clinical-imaging integrated nomogram predictive model for DR-TB risk in pediatric patients (≤18 years), and to evaluate the diagnostic performance of the GeneXpert MTB/RIF assay in this context.

METHODS

This retrospective study included 223 patients with TB aged ≤18 years hospitalised between 1 July 2018 and 31 December 2023. Drug resistance profiles were analysed, and the clinical/imaging features of DR-TB were compared with those of drug-susceptible TB (DS-TB). Multivariate logistic regression was used to identify DR-TB risk factors.

RESULTS

Of the 223 patients, 73.5% had DS-TB and 26.5% had DR-TB (including 13.5% with multidrug-resistant TB). Resistance rates to first-line drugs were as follows: isoniazid 22.4%, rifampicin 16.1%, streptomycin 8.1% and ethambutol 3.6%. Independent DR-TB risk factors were retreatment (odds ratio [OR] = 5.303, 95% confidence interval [CI]: 1.378-20.414), smoking history (OR = 4.129, 95% CI: 1.233-13.825), right middle lobe involvement (OR = 3.004, 95% CI: 1.148-7.863) and cavity formation (OR = 2.950, 95% CI: 1.325-6.567) (all P < 0.05). A nomogram model was developed, with an area under the curve of 0.776, showing effective predictive performance and clinical utility. The GeneXpert MTB/RIF assay had 87.5% sensitivity and 90.48% specificity for detecting rifampicin resistance, with good agreement with drug susceptibility testing (kappa 0.751) (P < 0.001). No significant differences in T-lymphocyte subsets were found between the DR-TB and DS-TB groups.

CONCLUSION

In children and adolescents, TB is more common in older individuals and consists primarily of new cases. The nomogram model has good predictive value for DR-TB risk. Key predictors include retreatment, smoking, right middle lobe involvement and cavity formation. The GeneXpert MTB/RIF assay is a potential early screening tool for DR-TB. Further validation with larger samples is needed.