Agree that V20, MLD, and even sup-inf location are far more validated than V5.
Currently, I am not altering plans or doses
based on V5.
MDACC paper is a nice contribution to the literature on this topic, but unfortunately has several statistical methodoligic problems: (1) The authors of this study used a Cox proportional hazards model to compare the statistical significants of the covariates (lung volume, GTV, MLD, rV5-rV80, and aV5-aV80) in predicting clinical RP. Unfortunately, for the Cox model to be statistically valid, the covariates have to be independent of one another. MLD, rV5-rV80 and aV5-aV80 are not independent of one another. If the covariates are co-dependent, then the most statistically valid method for multivariate comparison is using Bootstrap modelling... which is very labor intensive. (2) The authors also failed to develop a "test" and "validation" cohort to confirm the statistical validity of their predictors (V5 specifically). Jeff Bradley's group at Wash U used Bootstrap modelling in their study in the Red J a few years ago. They also used a separate "test" cohort - from Wash U- and "validation" cohort- from RTOG 9311. In my opinion, this is the most statistically valid paper on this topic. (3) Finally, the authors of the MDACC/V5 paper failed to include sup-inf location in the Cox model. The Wash U/RTOG 9311 paper identified MLD and sup-inf location as the most significant predictors of pneumonitis (even more than V20). Leaving sup-inf location out further weakens the validity of the conclusions.