With long-term follow-up for toxicity, we have demonstrated that higher doses of radiation than are traditionally administered (60 Gy–66 Gy) can be safely delivered to a majority of patients with NSCLC by use of 3D conformal techniques without inclusion of elective nodal irradiation to minimize the volume of lung treated and by quantifying the relationships among dose, volume irradiated, and risk of toxicity. In 83 patients (76%) who received more than 69.3 Gy (maximum dose of 103 Gy), no Grade 4 to 5 lung toxicity occurred. The occurrence of Grade 2 to 3 radiation lung toxicity was not directly associated with the prescribed tumor dose, but with lung-dosimetric factors such as MLD, V20, V13, and lung NTCP.

The clinically significant (Grade 2 to 3) radiation lung toxicity was not found to be associated with the prescription dose, although it is traditionally believed to be so. The traditional belief on total dose and toxicity was based on whole-organ radiation. Indeed, for whole-lung irradiation, lung tolerance decreases markedly with increasing dose after 8.2 Gy in a single fraction or 26.5 Gy in 20 fractions. When more uniform volume of lung was included by use of 2D planning, results from earlier RTOG trials showed a trend of increasing lung toxicity in the dose range of 40 to 60 Gy. The relationship between prescription tumor dose and lung tolerance is less predictable when 3D conformal radiation therapy is used, because the normal lung is nonuniformly irradiated to confine the dose to the region surrounding the tumor volume. That is, for 3D conformal treatments, the prescription dose, which is the dose to the tumor, may not always be correlated with the dose to the normal lung; patients with same prescription dose may have greatly varying volumes of normal lung irradiated, depending upon tumor size, location, and the beam arrangement used. Our finding is consistent with other recent studies  that total dose is not significantly associated with pneumonitis. However, as demonstrated in the current study, 3D tools allow assessment of dose distribution across the normal lung and assignment of maximum prescription dose on the basis of the dose–volume relationships in the lung itself, rather than on the tumor dose, regardless of lung dosimetry.

In the current trial, the patients who received higher doses of radiation were those with smaller effective volumes of irradiated lung, which suggests that irradiated-lung volume may be a more important factor than tumor dose in determination of the lung toxicity. Indeed, both pneumonitis and fibrosis were significantly associated with many volumetric factors, such as primary tumor volume, total-lung volume, and percent of lung volume that receives a certain dose (e.g., V13 and V20). The tumor and lung volumes were significant as they were correlated with the lung volume irradiated but became insignificant when lung-dosimetric factors were included in the same model. Results from this trial help confirm the independent associations between lung-dosimetric factors such as V20, V13, and MLD and the risk of radiation pneumonitis. The MLD model integrates the effects of the dose distribution over the lung volume and appears to work well for describing toxicity for these conformal lung treatments. Although V20 has been the most commonly used parameter, the literature is controversial regarding its exact relationship with pneumonitis Grade greater than 2: Graham reported 7% for patients with V20 of 22% to 31%, whereas Tsujino  had 51% for a V20 of 26% to 30%. This outcome is not surprising because V20 is a single-point dose–volume metric, which varies with the shape of the dose–volume histogram (DVH), which, in turn, is normally dependent on the pattern of beam arrangement of each institution. Results from the current study showed a more significant association between MLD and the combined incidence of Grade 2 or higher pneumonitis and fibrosis. One must note, however, that a statistically significant association is not the same as a good predictor of the toxicity. Results of our analysis also revealed that V13, V20, and MLD all have a similar suboptimal predictive ability for Grade 2 to 3 lung toxicity no significant difference occurs in predictive accuracy among the above models. The more complicated NTCP model was not superior to that of simple dosimetric factors in this study.

One must keep in mind that the results of this study are limited by the number of events for toxicity and patients who received chemotherapy. The results and conclusions are limited by the study population and method of dose prescription. The lung-volume–dose–lung-toxicity relationship demonstrated by this study, particularly those with the point-dosimetric factors such as V20 and V13, should not be used as a practice guideline, as they are highly dependent on the shape of DVH, which is a result of treatment technique/beam arrangement. The cutoff value for MLD is also limited to patients with NSCLC, who are treated with radiation or sequential chemoradiation by the use of 3D conformal techniques, with beam arrangements primarily located in 1 lung. For patients treated with concurrent chemoradiation, the steepness of the relationship between MLD and lung toxicity may change, although a recent study from the M.D. Anderson Cancer Center reported that MLD of 20 Gy still appears to be a reasonable limit for these patients . Further, the results of all these studies should not be extrapolated to other more encompassing beam arrangements  or dose-computation techniques without great care. Although modeling Grade 2 to 3 toxicity is useful, the important message from this study is the safety of high-dose radiation when the lung volume included is limited. Although its predictive sensitivity is limited, MLD cut-off of 20 Gy had an acceptable negative predictive value (89%) and excellent specificity (95%)  which suggest the safeness of this limit for clinical application under similar treatment regimens in patients with NSCLC.

Of note, we tried to analyze symptomatic (Grade 2 to 3) fibrosis separately from and jointly with pneumonitis. As both fibrosis and pneumonitis belong to radiation lung toxicity, the logic of the combined analysis is obvious. The reasons for performing a separate analysis were (1) fibrosis is a later radiation lung toxicity, which differs from pneumonitis in molecular regulation, hisopathologic presentation, radiographic appearance, and clinical presentation, and (2) fibrosis may or may not be a direct sequela of pneumonitis and, thus, may have different predictors than pneumonitis. Fibrosis is traditionally considered as a radiographic finding and was shown to have a pattern of positive dose-dependent relationships radiographically in voxel level, with remarkable individual differences). A study from the M.D. Anderson Cancer Center demonstrated that the probability of radiographic fibrosis was a function of dose and volume, with a threshold dose of 30 Gy to 40 Gy. Radiographic fibrosis may not have clinical significance. Data on modeling the clinically meaningful fibrosis (Grade ≥ 2) are lacking in the literature, although an increased incidence of thoracic radiation–related death was reported in the presence of fibrosis on chest X-ray. As previously noted by others, this study suggests that clinical symptomatic fibrosis may be related to moderate radiation pneumonitis. Similarly, lung-dosimetric factors such MLD, V20, and NTCP were all significant variables. One should also note that symptomatic fibrosis was less predictable than pneumonitis, with remarkably wider 95% confidence interval in risk prediction Although V13 was significantly associated with pneumonitis (p < 0.001), it was not a significant factor for symptomatic fibrosis. Additionally, lung volume and functional capacity such as FVC may also be associated with the clinical symptoms from fibrosis, as suggested by our earlier analysis. The reasons for above results are unknown. Future studies with larger number of patients and that use multiple variables of functional reserve, dosimetric, and individual biologic factors may improve the predictive ability.

Finally, on the basis of the results from this trial and several other dose-escalation trials, the current standard practice of prescribing tumor doses of 60 Gy or slightly higher  (largely because of concerns of radiation lung toxicity) are likely overly conservative for a majority of patents with NSCLC, particularly when radiation is delivered in a conformal fashion. In our trial, 83 of 109 patients (76%) received more than 66 Gy without any Grade 4 to 5 lung toxicity. In RTOG 9311, dose was escalated to 83.8 Gy in patients with V20 less than 25% without severe toxicity Other studies have treated to 84 Gy in patients with NTCP less than 25%  or relative MLD less than 0.12  and to 90 Gy with concurrent chemotherapy without limiting lung-dosimetric parameters . Although Grade 2 to 3 toxicity is clinically significant, as it may have an impact on the patient’s quality of life, the common occurrence of Grade 1 toxicity is of minimal clinical significance. For patients with NSCLC treated with radiotherapy, the major problem continues to be local-regional failure, with the majority of patients still dying of their disease With accumulating data on the positive dose–response effect, the standard use of a radiation dose of 60 Gy  may need to be reconsidered. A randomized controlled study is still needed to confirm the benefit of high-dose (>60 Gy) radiation. As in the conformal radiation with concurrent chemotherapy treatment era, patients differ greatly in their risk of developing moderate lung toxicity for a given tumor-prescription dose, individualized high-dose radiation based on lung dosimetry would provide an attractive strategy to improve the therapeutic ratio in patients with NSCLC. If the mean lung dose is limited, individualized high-dose radiation may be safely delivered to a majority of patients, although further lung-toxicity modeling is needed to generate safe cutoff values for patients treated with concurrent chemoradiation, which is the current standard of care for Stage III NSCLC.