Modeling radiation pneumonitis risk with clinical, dosimetric, and spatial parameters

A Nomogram to Predict Radiation Pneumonitis, Derived From a Combined Analysis of RTOG 9311 and Institutional Data

“…[6] The best WU model was a poor performer when applied to the RTOG dataset. Applying the WU model directly to RTOG 9311 will yield r = 0.06.…”

“…We had previously shown that a three parameter model of tumor location (inferiorly-located tumors with higher risk), maximum dose, and D35 was superior to either V20 or MLD in our own database. [6]. At the same time, we obtained permission to test this model within the RTOG 9311 database.…”

“…The lack of effect from chemotherapy is consistent with other publications. [11,12,6,13,1] Figure 5 shows a histogram of the most frequently selected model sets on bootstrap refitting simulations. The twoparameter model using MLD and S-I tumor position is clearly superior.…”

“…In a recent analysis, Hope et al demonstrated that a three-parameter model that included tumor position (inferiorly-located tumors had higher risk of RP), maximum dose, and D35 was a better predictor for RP than either V20 or MLD alone in our database. [6] However, substituting V20 or MLD for D35 achieved nearly as good a statistical performance. To further validate or modify these results, we present here an expanded analysis, in which we combine WU data with data extracted from RTOG 9311 trial treatment planning archives maintained by the Image-Guided Therapy QA Center.…”

“…A description of this patient population was previously published. [6] Inclusion criteria required a minimum follow up of 6 months, or the observation of RP prior to 6 months. From the RTOG 9311 dataset, we were able to extract full 3 dimensional (3D) treatment planning data from 129 of the 179 patients enrolled.…”