Purpose: No data is available on the different FDG PET and CT findings in the lymph nodes (LN) of patients with HIV and tuberculosis (TB) who respond compared with those who do not respond to anti-TB treatment by 4 months after initiation of TB treatment. These findings were the focus of our study.
Methods: PET/CT scans performed at 4 months after initiation of TB treatment in 20 consecutive HIV patients were analysed. SUVmax values were obtained for all regions of LN involvement. The diameter of the LNs was measured and the CT enhancement (LNs showing peripheral rim enhancement with central low attenuation, PRECLO, in comparison with homogeneously involved LNs) and the calcification patterns of involved LNs assessed. The relationship between the PET and CT findings and the clinical outcome, response or nonresponse, was evaluated.
Results: FDG PET identified 91 sites of LN involvement, 20 of which were not identified by CT. SUVmax values were significantly higher in nonresponders (8 patients, SUVmax 11.2 ± 4.0, mean ± SD) when compared to responders (12 patients, SUVmax 2.6 ± 2.3; p = 0.0001). In ROC analysis (AUC 0.952) a cut-off value of 4.5 for SUVmax yielded a sensitivity and specificity of 95% and 85% for discriminating nonresponding from responding LNs. LNs were significantly larger in nonresponders (1.9 ± 0.4 cm) than in responders (1.4 ± 0.4 cm; p = 0.0001); the AUC in the ROC analysis was 0.76. PRECLO LNs were significantly larger (2.2 ± 0.3 cm) than homogeneous involved LN basins (1.5 ± 0.4 cm) and LN basins with calcification (1.4 ± 0.5 cm; p = 0.001). Using the presence of at least one LN basin with PRECLO as a criterion for nonresponse, responders could be separated from nonresponders with a sensitivity of 88% and a specificity of 66%.
Conclusion: LNs responding to TB treatment could be differentiated from nonresponding LNs with a sensitivity and specificity of 95% and 85% using a SUVmax cut-off value of 4.5 and a sensitivity and specificity of 88% and 66% using the presence of at least one LN basin with PRECLO.