Purpose: To determine whether a little dose to a large normal lung volume or a high dose to a small lung volume is more critical for induction of clinical pneumonitis. The second question is if dose-volume histogram (DVH) parameters are more reliable, if the lungs are analyzed as separate organs or as a whole organ.

Patients and methods: We analyzed the clinical and DVH data from 49 patients treated for a thoracic malignancy using 3-D conformal treatment plans. 18 patients had developed a clinical pneumonitis (CTC II or III). The majority of patients (n = 48) received radiochemotherapy for non-small-cell lung cancer (NSCLC) with a combination of paclitaxel and carboplatin. Patients were generally treated 5 fx/week, single dose 2 Gy, using a two-series approach (shrinking field) up to a total dose of 60-70 Gy. For every individual patient, the overall dose distribution was recalculated in the Helax-TMS by means of adding dose plans according to the total dose applied in each series. The lungs were defined both as separate organs and as a whole organ. Low-dose volume (< or = 10 Gy, Vlow), moderate-dose volume (> 10-40 Gy, Vmod) and high-dose volume (> 40 Gy, Vhigh), as well as V10-V40 and mean lung dose (MLD) were defined from the cumulative DVH. Dose-effect relationships were fitted with a logistic regression model.

Results: Manifestation of clinical pneumonitis was within 3 months from termination of irradiation in all cases. For the ipsilateral lung, the incidence of pneumonitis was closely correlated to Vhigh. The pneumonitis rate increased from 13% up to 60%. By contrast, with increasing Vlow the pneumonitis rate dropped to < 10%. A similar but less pronounced effect was seen for the total lung. The lung volumes Vlow, Vmod and Vhigh of the ipsilateral, contralateral and whole lung were significantly correlated to the corresponding MLD. The incidence of pneumonitis increased with increasing MLD for the ipsilateral lung with a D50 of 32 Gy and a gamma 50 of 0.98. For the whole lung, the observed increase was less steep. MLD showed a close correlation to NTCP calculated by the Kutcher model. However, NTCP calculation overestimated the pneumonitis risk for the ipsilateral lung and underestimated the risk for the whole lung due to the steeper gradient. The logistic regression curve for the DVH parameters V10-V40 showed an increase of steepness toward higher doses. From the logistic regression curves, a DVH template indicating critical borders of V10-V40 was generated for the ipsilateral as well as for the total lung.

Conclusion: Our data indicate that it is reasonable to disperse the dose outside the target volume over large areas in order to reduce the volumes of lung receiving > 40 Gy. Reducing the high-dose volume reduces the pneumonitis rate more than a corresponding reduction in the low-dose regions of the DVH. Landmarks for DVH optimization as defined in this analysis may serve as a basis for DVH contrains in IMRT planning. Separate organ analysis produced more reliable results and should be preferred to whole-organ analysis, if techniques mainly involving one side of the lung are applied. Further validation of these constraints is necessary prior to general recommendation.