Abstract
BACKGROUND: Early antimicrobial therapy is crucial regarding the prognosis of vertebral osteomyelitis, but early pathogen diagnosis remains challenging. OBJECTIVE: In this study, we aimed to differentiate the types of pathogens in iatrogenic vertebral osteomyelitis (IVO) and native vertebral osteomyelitis (NVO) to guide early antibiotic treatment. METHODS: A total of 145 patients, who had confirmed spinal infection and underwent metagenomic next-generation sequencing (mNGS) testing, were included, with 114 in the NVO group and 31 in the IVO group. Using mNGS, we detected and classified 53 pathogens in the 31 patients in the IVO group and 169 pathogens in the 114 patients in the NVO group. To further distinguish IVO from NVO, we employed machine learning algorithms to select serum biomarkers and developed a nomogram model. RESULTS: The results revealed that the proportion of the Actinobacteria phylum in the NVO group was approximately 28.40%, which was significantly higher than the 15.09% in the IVO group. Conversely, the proportion of the Firmicutes phylum (39.62%) in the IVO group was markedly increased compared to the 21.30% in the NVO group. Further genus-level classification demonstrated that Staphylococcus was the most common pathogen in the IVO group, whereas Mycobacterium was predominant in the NVO group. Through LASSO regression and random forest algorithms, we identified 5 serum biomarkers including percentage of basophils (BASO%), percentage of monocytes (Mono%), platelet volume (PCT), globulin (G), activated partial thromboplastin time (APTT) for distinguishing IVO from NVO. Based on these biomarkers, we established a nomogram model capable of accurately discriminating between the two conditions. CONCLUSION: The results of this study hold promise in providing valuable guidance to clinical practitioners for the differential diagnosis and early antimicrobial treatment of vertebral osteomyelitis.