PURPOSE: With the mean lung dose (MLD) as an estimator for the normal tissue complication probability (NTCP) of the lung, we assessed whether the probability of tumor control of lung tumors might be increased by dose escalation in combination with a reduction of field sizes, thus increasing target dose inhomogeneity while maintaining a constant MLD. METHODS AND MATERIALS: An 8-MV AP-PA irradiation of a lung tumor, located in a cylindrically symmetric lung-equivalent phantom, was modeled using numerical simulation. Movement of the clinical target volume (CTV) due to patient breathing and setup errors was simulated. The probability of tumor control, expressed as the equivalent uniform dose (EUD) of the CTV, was assessed as a function of field size, under the constraint of a constant MLD. The approach was tested for a treatment of a non-small cell lung cancer (NSCLC) patient using the beam directions of the clinically applied treatment plan. RESULTS: In the phantom simulation it was shown that by choosing field sizes that ensured a minimum dose of 95% in the CTV ("conventional" plan) taking into account setup errors and tumor motion, an EUD of the CTV of 43.8 Gy can be obtained for a prescribed dose of 44.2 Gy. By reducing the field size and thus shifting the 95% isodose surface inwards, the EUD increases to a maximum of 68.3 Gy with a minimum dose in the CTV of 55.2 Gy. This increase in EUD is caused by the fact that field size reduction enables escalation of the prescribed dose while maintaining a constant MLD. Further reduction of the field size results in decrease of the EUD because the minimum dose in the CTV becomes so low that it has a predominant effect on the EUD, despite further escalation of the prescribed dose. For the NSCLC patient, the EUD could be increased from an initial 62.2 Gy for the conventional plan, to 83.2 Gy at maximum. In this maximum, the prescribed dose is 88.1 Gy, and the minimum dose in the CTV is 67.4 Gy. In this case, the 95% isodose surface is conformed closely to the "static" CTV during treatment planning. CONCLUSIONS: Iso-NTCP escalation of the probability of tumor control is possible for lung tumors by reducing field sizes and allowing a larger dose inhomogeneity in the CTV. Optimum field sizes can be derived, having the highest EUD and highest minimum dose in the CTV under condition of a constant NTCP of the lungs. We conclude that the concept of homogeneous dose in the target volume is not the best approach to reach the highest probability of tumor control for lung tumors.