We hypothesize that preferential and intentional dose reduction to regions of lung that are highly ventilated at baseline in radiation treatment plan optimization for stage III non-small cell lung cancer (NSCLC) patients will preserve lung ventilation and ultimately limit radiation-induced toxicities. To test this hypothesis, this study develops a novel treatment planning formulism combining anatomical and functional imaging and estimates the normal tissue complication probability (NTCP) improvement with this formulism. Fifty-eight stage III NSCLC patients from 2 institutions were retrospectively studied. In 21 patients from institution A, in-silico planning comparisons (ISPC) were conducted between PlanDV, using dose-volume (DV) constraints of organs-at-risk (OAR), and PlanDFV, using dose-function-volume (DFV) constraints derived from fractional regional ventilation (FRV) images. A multivariable logistic regression NTCP model was then developed using a cohort of 37 patients from institution B from whom clinical data on radiation fibrosis (RF) was available. Two hallmark characteristics of RF were recorded: volume loss and airway dilation. The NTCP model specifically incorporated the volume of functional lung exposed to irradiation. Other candidate variables included in the model were demographics (age, sex, etc.) and clinical treatment parameters (TNM stage, histology, etc.). We estimated the potential benefit of PlanDFV over PlanDV in reducing RF by integrating the results of the ISPC into the NTCP models. ISPC showed that with same target coverage, PlanDFV met OAR constraints such that all conventional DV metrics were indistinguishable (P>0.05). When DFV metrics were compared, 5-10% reductions of function-weighted V20 (P = 0.04) and function-weighted mean lung dose (P<0.01) were achieved with PlanDFV. The most predictive variables for volume loss were sV20 (lung voxels with relative function values above 60% receiving 20 Gy), smoking history and age (AUC = 0.95). NTCP modeling showed that lowering sV20 by 5% using PlanDFV, a 21% toxicity reduction could be achieved for 55 year-old non-smoker patients. Similarly, the most predictive variables for airway dilation were sV20, T stage and histology (AUC = 0.93). For squamous cell carcinoma, lowering sV20 by 5% using PlanDFV, toxicity could be reduced by 22% and 11% for T2 and T4 patients, respectively. This study showed that the inclusion of lung ventilation (in addition to lung volume), when optimizing radiation therapy plans for NSCLC patients, may allow for a clinically meaningful benefit for individual patients. A prospective trial is planned to validate these findings to determine if treatment-related morbidity and mortality can be reduced that might guide potential future dose escalation trials in NSCLC.