Latent class analysis of CT patterns in pulmonary non-tuberculous mycobacterial infection

Background: Pulmonary Nontuberculous mycobacterial infection (pNTM) is a challenging condition which is increasingly prevalent. The clinical course is variable and may be difficult to predict. Previous studies have identified radiological features associated with an adverse prognosis. Aims: The aim of this study was to use latent class analysis (LCA) as an unbiased method of grouping subjects with pNTM based on their CT features and to compare the clinical characteristics of these groups. Methods: Subjects with pNTM were recruited and a contemporaneous CT thorax obtained. CT scans were scored as described previously (Zoumot et al. Respirology 2014 19:714) and groups identified using LCA. Results: CT scans were obtained for 89 subjects. LCA identified three groups with distinct disease patterns. Group one (19.1% of subjects) was characterised by severe cavitation, consolidation and extensive bronchiectasis. Aspergillomas were present in 47%, but absent in the other two groups. Group two (43.8%) was characterised by mild bronchiectasis and nodules. Group three (37.1%) had the most extensive bronchiectasis and the highest frequency of tree-in-bud opacification, but nodules and cavitation were rare. Subjects in group one were older, with a higher CRP and platelet count and lower serum albumin and BMI. Over a median 126 weeks follow-up a significantly higher mortality was seen in group one (35.3%) compared to groups two (16.1%) and three (9.1%) ( P = 0.049). Conclusion: Subjects with pNTM formed three groups with different radiological patterns of disease. These patterns are associated with significant differences in clinical variables and survival.