Dosimetric Predictors of Radiation-induced Vaginal Stenosis in Rectal and Anal Cancer Patients

Abstract

While vaginal stenosis (VS) is a recognized toxicity in women who receive pelvic radiation therapy (RT), the relationship between dose/irradiation volume and extent of toxicity has not been analyzed. We modeled the relationship between post-pelvic RT VS and dose/volume metrics to suggest potential clinical guidelines. We evaluated 54 female patients, aged 29-78, who underwent pelvic RT for the treatment of rectal or anal cancer from 2009-2011 and were enrolled on a prospective study evaluating the incidence of VS and compliance with vaginal dilator use. Data at baseline prior to RT and 4 weeks after RT were collected on maximum dilator size inserted without difficulty. The severity of VS was measured by the ratio between maximum vaginal dilator size at one month post-RT and at baseline. Those with a ratio <1 were classified as having VS (n = 35), while those with a ratio ≥1 were classified as having no RT-induced VS (n = 19). Dose-volume parameters were extracted from the initial treatment plans for modeling using CERR software. On univariate logistic regression analysis, Spearman's correlation coefficients (Rs) were calculated to identify clinical and dose/volume variables associated with VS, Significant associations were tested using a logistic regression model, and the area under the receiver operating characteristic curve (AUC) and Rs were calculated to assess the strength of correlation between competing models and VS. Based on univariate analysis, 21 variables were further input into using multivariate logistic regression analysis. The best-fit model, consisting of Dmax (maximum dose), D80 (minimum dose that covers 80% of vaginal volume), and MOC5 (mean dose to the coldest 5% dose volume), was highly predictive of VS: Rs = 0.501 (p < 0.001) and AUC = 0.803. Mean MOC5 was 28.1 Gy and 12.6 for the VS and no VS groups, respectively. The VS and no VS groups had mean Dmaxs of 54.6 Gy and 51.9 Gy, respectively. For D80, the VS and no VS groups had mean D80's of 38.8 Gy and 28.1 Gy, respectively. Vaginal stenosis is influenced by multiple dose/volume characteristics, including the DVH high-dose region as well as low-dose sparing. A multi-metric model including Dmax, D80, and higher mean doses in the lowest dose regions of the vaginal tissue, as described by the MOC5, was highly predictive of VS. Dose constraints based on these findings -- in particular, limiting the Dmax to <52 Gy and maintaining 5% of the vaginal volume to <13 Gy -- may minimize the risk of VS and result in improved quality of life in women receiving pelvic RT for anal or rectal cancers. Validating these constraints will require analysis of independent datasets.