External validation of the neoadjuvant rectal (NAR) score and Valentini prediction nomogram (VPN): A multicenter study.
2014; Lippincott Williams & Wilkins; Volume: 32; Issue: 15_suppl Linguagem: Inglês
10.1200/jco.2014.32.15_suppl.3532
ISSN1527-7755
AutoresSoundouss Raissouni, Jamison Mercer, Gillian Gresham, Aalok Kumar, Rachel Goodwin, Di Jiang, Alexander A. C. Leung, Daniel Yick Chin Heng, Patricia A. Tang, Corinne Doll, Anthony R. MacLean, Erin Powell, Julie A. Price Hiller, Jose Gerard Monzon, Winson Y. Cheung, Michael M. Vickers,
Tópico(s)Clinical practice guidelines implementation
Resumo3532 Background: The VPN [JCO 2011;29(23)] is a tool used for individualized prognostication in rectal cancer, and the recently described NAR score [Yothers et al. JCO 32, 2014 abst: 38] has been suggested as a surrogate endpoint for overall survival (OS) in clinical trials. However, these tools have not been validated outside of the clinical trial setting. We assessed these models in a retrospective multi-institution database, to determine their predictive ability for outcome in patients with rectal cancer treated with neoadjuvant chemoradiation (nCRT). Methods: Data from patients with locally advanced rectal cancer who received nCRT and had curative intent surgery from 2005 to 2012 were collected from Tom Baker Cancer Center, Cross Cancer Institute, BC Cancer Agency, Ottawa Hospital Cancer Centre and Dr. H. Bliss Murphy Cancer Centre. The NAR score was compared with pathologic complete response (pCR) status for OS and with the VPN for local control (LC), distant control (DC) and OS using Akaike’s information criterion (AIC – a tool for comparing models in a dataset). C-index was evaluated for VPN and NAR for OS. Results: 1,172 patients were included; pCR was achieved in 16.6%.The median NAR score was 17.36; NAR risk groups included low (22.6%), intermediate (42.6%), and high (34.9%). The 5-year OS in the NAR risk groups were 88.1% (low), 82% (intermediate – HR 1.89, 95% CI 1.13 – 3.1), 59.5% (high – HR 5.05, 95% CI 3.1 – 8.2), and 89.9% for pCR (HR 0.25, 95% CI 0.14 – 0.45) versus 73.1 % for no pCR (log rank p-value <0.0001 for both). AIC favoured NAR compared with pCR for OS (2,549.6 v 2,824.0).VPN as a continuous variable predicted LC (HR 0.97, 95% CI 0.95-0.99, p=0.015), DC (HR 0.98, 95% CI 0.98-0.99, p<0.0001) and OS (HR 0.96, 95% CI 0.95-0.96, p<0.0001). The AIC favoured VPN over NAR for LC (851.2 v 856.5), DC (2,154.5 v 2,160.6) and OS (2,455.3 v 2,549.6). C-index also favoured VPN over NAR for OS (0.697 v 0.66). Conclusions: Similar to recent clinical trial data, NAR outperformed pCR in predicting OS, and the VPN was significantly associated with LC, DC and OS in the non-clinical trial setting. VPN appeared to better predict LC, DC, and OS compared with the NAR score and is favoured for individual patient prognostication.
Referência(s)