A competing risk nomogram to predict severe late toxicity after modern re-irradiation for squamous carcinoma of the head and neck

Publication date: March 2019

Source: Oral Oncology, Volume 90

Author(s): Matthew C. Ward, Nancy Y. Lee, Jimmy J. Caudell, Alexander Zajichek, Musaddiq J. Awan, Shlomo A. Koyfman, Neal E. Dunlap, Sara J. Zakem, Comron Hassanzadeh, Samuel Marcrom, Drexell H. Boggs, Derek Isrow, John A. Vargo, Dwight E. Heron, Farzan Siddiqui, James A. Bonner, Jonathan J. Beitler, Min Yao, Andy M. Trotti, Nadeem Riaz

Abstract

Purpose

Severe late toxicity is common after re-irradiation for recurrent or second primary (RSP) squamous carcinoma of the head and neck. However, many patients experience complications from tumor progression before manifesting late effects. We constructed a nomogram to examine this relationship between late toxicity and competing risks.

Methods and materials

Patients with RSP squamous carcinoma originating in a field previously irradiated to ≥40 Gy and treated with IMRT-based re-irradiation to ≥40 Gy were collected. Grade ≥3 late toxicity developing ≥90 days after re-irradiation was collected. A multivariable competing-risk model was fit to the actuarial risk of late toxicity with progression or death as the competing risk. The final bootstrap optimized model was converted into a nomogram.

Results

From 9 institutions, 505 patients were included. The 2-year incidence of grade ≥3 late toxicity was 16.7% (95% CI 13.2–20.2%) whereas progression or death was 64.2% (95% CI 59.7–68.8%). The median freedom from late toxicity, progression or death was 10.7, 5.5 and 3.2 months for RPA class I-III patients respectively, whereas the median OS was 44.9, 15.9 and 7.9 months, respectively. The final model included six clinical factors. Notably, dose, volume and fractionation did not significantly impact toxicity.

Conclusions

After re-irradiation, the risk of progression or death is approximately four times the risk of radiation-related severe late toxicity. The risk of late toxicity may be more dependent on patient and disease factors than modifiable treatment factors. This model is useful for patient selection, pre-treatment consent and post-treatment survivorship following re-irradiation.

http://bit.ly/2Dp9azW