Prospective study of

backgroundOn conventional diagnostic imaging, the features of radiation necrosis (RN) are similar to those of local recurrence (LR) of brain metastases (BM).

methodsThe eligible subjects had enlarging contrast-enhanced lesions (>1 cm) on MR imaging after any form of radiotherapy for BM, suggesting LR or RN. However, it was difficult to differentiate LR from RN in these cases. From August 2013 to February 2017, MET-PET was performed for 37 lesions in 32 eligible patients. Tracer accumulation in the regions of interest was analysed as the standardised uptake value (SUV) and maximal lesion SUV/maximal normal tissue SUV ratios (LNR) were calculated. The cut-off value for LNR was provisionally set at 1.40. Salvage treatment strategies determined based on MET-PET diagnosis and treatment results were investigated. The diagnostic accuracy of MET-PET was evaluated by receiver operating characteristic (ROC) curve analysis.

resultsThe median interval from primary radiotherapy to MET-PET was 19 months and radiotherapy had been performed twice or more for 13 lesions. The MET-PET diagnoses were LR in 19 and RN in 18 lesions. The mean values and standard deviation of LNRs for each diagnostic category were 1.70 ± 0.30 and 1.09 ± 0.25, respectively. At the median follow-up time of 18 months, final diagnoses were confirmed histologically for 17 lesions and clinically for 20 lesions. ROC curve analysis indicated the optimal LNR cut-off value to be 1.40 (area under the curve: 0.84), and the sensitivity and specificity were 0.82 and 0.75, respectively. The median survival times of patient groups with LR and RN based on MET-PET diagnosis were 14.8 months and 35.1 months, respectively (P = 0.035, log-rank test).

conclusionsMET-PET showed apparently reliable diagnostic performance for distinguishing between LR and RN. The provisional LNR cut-off value of 1.4 in our institution was found to be appropriate. Limitations of diagnostic accuracy should be recognised in cases with LNR close to this cut-off value.