Journal of Thoracic Oncology:
Profferred Paper Abstracts: Session A2: Imaging - Prognostic Determinants: Monday, September 3: Imaging - Prognostic Determinants, Mon, 13:45 - 15:30
1 MAASTRO Clinic, Maastricht, The Netherlands 2 University Hospital Maastricht, Maastricht, The Netherlands 3 Maasland Hospital, Sittard, The Netherlands 4 Laurentius Hospital, Roermond, The Netherlands 5 Sint Jans Hospital, Weert, The Netherlands 6 Atrium Medical Center, Heerlen, The Netherlands 7 Siemens Medical Solutions, Philadelphia, PA, USA 8 MAASTRO Clinic, Heerlen, The Netherlands
Accurately predicting survival of NSCLC patients is important for treatment decision making. However, it is widely recognized that the UICC (TNM) staging system has its shortcomings when used for the risk stratification of inoperable NSCLC patients treated with (chemo)radiotherapy. Factors that are lacking and may be important for the outcome after (chemo)radiation include size of the tumor (which is not taken into account as a continuous variable), gender, performance status and radiation dose. In addition, the number of positive lymph nodes is not taken into account, whereas it is of prognostic value for surgical patients. As it is possible to assess the mediastinal lymph nodes with FDG-PET scans, we hypothesized that also for non-surgical patients, this would affect survival.
To investigate the prognostic value of tumor volume (assessed by CT), the number of positive lymph node stations (assessed by PET), gender, performance status and equivalent radiation dose corrected for time (EQD2,T) for overall survival, and compare this with UICC stage in patients with inoperable NSCLC treated with (chemo)radiotherapy.
Clinical data from 270 inoperable NSCLC patients, UICC stage I-IIIB, treated at MAASTRO clinic with (chemo)radiotherapy, was collected retrospectively. Diagnostic imaging was performed either with an integrated PET-CT or with CT-scan and PET-scan separately. Overall survival was calculated from the start of radiotherapy treatment. A logarithmic transformation was applied to obtain more symmetrically distributed data for the tumor volume. The Kaplan-Meier method as well as Cox regression were used to analyze the data. The likelihood ratio test (LR test) was applied to compare the performance of the models. The resulting p-values were reported. In addition, Akaike's information criterion (AIC) was calculated. The AIC takes into account how well the model fits the data as well as the complexity of a model, e.g. the number of estimated variables, thereby reducing the risk of overfitting. The preferred model is the one with the lowest AIC value. To assess the relative merits of a model the difference is interpreted as follows: 4-7 indicates less support, ≈10 indicates essentially no support for a model.
Univariate analysis showed that the number of positive lymph node stations, N-stage as well as T-stage were significantly associated with survival. However, in the final multivariate Cox regression, N-stage was no longer significant. A comparison was made between three multivariate models each consisting of gender, WHO-performance status, EQD2,T and only one of the following combinations: 1) total tumor volume and number of positive lymph node stations, 2) T-stage and N-stage, 3) UICC overall stage. The p-values of the LR test were <0.001, 0.004 and 0.99 respectively. The AIC of the models was 1965.8, 1989.9 and 2001.2 respectively. It was therefore concluded that model 1 was the most informative for prediction of overall survival.
The combination of total tumor volume, number of positive lymph node stations, gender, performance status and equivalent radiation dose corrected for time (EQD2,T) is the best predictor for survival in NSCLC patients, stage I-IIIB, treated with (chemo)radiation.
Copyright © 2007 by the European Lung Cancer Conference and the International Association for the Study of Lung Cancer.