Purpose: The aim of this study was to estimate 201Tl SPECT and CT-MRI cut-off values that lead to a validated prognostic classification for the end-point overall survival, in order to discriminate glioma patients with good and poor prognosis at an early stage during chemotherapeutic treatment.
Methods: We studied patients who underwent 201Tl SPECT and CT-MRI before and after two courses of chemotherapy. Cut-off values were retrieved from the Cox model. Patients were classified according to the computed cut-off values, creating subgroups of patients with different prognosis in terms of survival [tumour regression (TR); stable disease (SD); tumour progression (TP)]. The differences between the subgroups were assessed by Kaplan-Meier analyses. The predictive performance of the classification procedure was evaluated by a leave-one-out cross-validation method.
Results: 201Tl SPECT classified 41% of the patients as SD, 25% as TR and 34% as TP. CT-MRI classified 82% of the patients as SD, and only 4% and 14% as TR and TP, respectively. Of those patients with a relatively long overall survival (i.e. > or =16 months), cross-validation estimates of 201Tl SPECT classification rates were 50% TR and 50% SD, and cross-validation estimates of CT-MRI classification rates were 7% TR, 72% SD, and 21% TP.
Conclusion: We constructed a 201Tl SPECT model that makes it possible to identify glioma patients with a good or a poor prognosis at an early stage during chemotherapeutic treatment. With this model, accurate predictions can be made with regard to the expected duration of survival.