PV QA 4 - Poster Viewing Q&A 4
Purpose/Objective(s): We previously described a competing risk model of first failure site in locally advanced non-small cell lung cancer (LA-NSCLC) patients incorporating baseline PET-scan data. Herein, we seek to validate the model in an independent single-institution series.
Materials/Methods: Failure-type specific outcomes in consecutive patients with LA-NSCLC treated with curative chemoradiation between January 2000 and December 2016 were retrospectively analyzed. After excluding patients with trimodality therapy or the absence of pre-treatment PET/CT or dosimetric data, 231 patients met inclusion criteria. Failures inside the planning target volume (PTV) alone were considered local failures (LF); failures outside the PTV or distant metastases (DM) were classified as non-local failures (NLF) per model specifications. Multivariate cause-specific Cox regression analysis of patient demographic, disease and treatment variables along with competing risk analysis (CMPRSK) for failure type (LF vs NLF) was performed with Fine and Gray’s test, followed by a chi-square goodness-of-fit analysis.
Results: The population was predominantly stage IIIA (62%), male (56%), white (61%), adenocarcinoma (ACA, 53%) and performance status 0-1 (82%) with median age 65y (range, 30-88y). Median SUVmax of primary tumor (T) and node (N) was 13.31 (range, 1.8-41.66) and 8.1 (range, 1.3-48.3), respectively. Eighty-eight percent (n=203) of patients received concurrent chemotherapy with a median radiation dose of 61.2 Gy (range 20-81.6). Median overall survival (OS) and time to first failure (TTF) was 23m and 11m, respectively. Failure patterns defined as alive with no evidence of disease (alive NED), LF, NLF, and dead with no evidence of disease (dead NED) for ACA/squamous cell carcinoma (SCC) demonstrated 23%/24%, 6.5%/13%, 58%/40%, and 12.5%/23%, respectively (p=0.020). Intracranial DM were observed in 16% of ACC and 7% of SCC patients, respectively (p=0.03). On binary logistic regression, LF increased with increasing SUVmax (p=0.002, OR 1.3, 95% CI 1.1-1.4), but not T vs N location (p=0.27); however, all but one LF occurred in the T location. Two-year expected vs observed rates of first failure are shown in Table 1, with lower rates of LF and higher rates of NLF in our cohort compared to the model (p<0.001), possibly due to wider use of PET/CT versus CT thorax in follow-up.
Conclusion: In patients with LF, we demonstrate that increasing SUVmax is associated with increasing rates of LF, with all but one LF in the T site. However, our CMPRSK analysis rejected the fit of the 2-year patient level failure model with fewer LF and more NLF in our cohort. Future directions include elucidation of differences in patient and restaging patterns between the two cohorts, detailed analysis of failure sites, and ongoing multi-institutional validation.
|First Failure||Expected %||Observed % (95% CI)||p-value|
|Dead NED||10||16 (11-21)|
The asset you are trying to access is locked. Please enter your access key to unlock.