Inhomogeneous target-dose distributions: A dimension more for optimization?

Purpose: To evaluate if the use of inhomogeneous target-dose distributions, obtained by 3D conformal radiotherapy plans with or without beam intensity modulation, offers the possibility to decrease indices of toxicity to norm al tissues and/or increase indices of tumor control stage III non-small cel l lung cancer (NSCLC). Methods and Materials: Ten patients with stage III NSCLC were planned using a conventional 3D technique and a technique involving noncoplanar beam int ensity modulation (BIM). Two planning target volumes (PTVs) were defined: P TV1 included macroscopic tumor volume and PTV2 included macroscopic and mic roscopic tumor volume. Virtual simulation defined the beam shapes and incid ences as well as the wedge orientations (3D) and segment outlines (BIM). We ights of wedged beams, unwedged beams, and segments were determined by opti mization using an objective function with a biological and a physical compo nent. The biological component included tumor control probability (TCP) for PTV1 (TCP1), PTV2 (TCP2), and normal tissue complication probability (NTCP ) for lung, spinal cord, acid heart. The physical component included the ma ximum and minimum dose as well as the standard deviation of the dose at PTV 1. The most inhomogeneous target-dose distributions were obtained by using only the biological component of the objective function (biological optimiz ation). By enabling the physical component in addition to the biological co mponent, PTV1 inhomogeneity was reduced (biophysical optimization). As indi ces for toxicity to normal tissues, NTCP-values as well as maximum doses or dose levels to relevant fractions of the organ's volume were used. As indi ces for tumor control, TCP-values as well as minimum doses to the PTVs were used. Results: When optimization was performed with the biophysical as compared t o the biological objective function, the PTV1 inhomogeneity decreased from 13 (8-23)% to 4 (2-9) % for the 3D-(p = 0.00009) and from 44 (33-56)% to 20 (9-34)% for the BIM plans (p < 0.00001). Minimum PTV1 doses (expressed as the lowest voxel-dose) were similar for both objective functions. The mean and maximum target doses were significantly higher with biological optimiza tion for 3D as well as for BIM (al p values < 0.001). Tumor control probabi lity (estimated by TCP1 x TCP2) was 4.7% (3D) and 6.2% (BIM) higher for bio logical optimization (p = 0.01 and p = 0.00002 respectively). NTCPlung as w ell as the percentage of lung volume exceeding 20 Gy was higher with the us e of the biophysical objective function. NTCPheart was also higher with the use of the biophysical objective function. The percentage of heart volume exceeding 40 Gy tended to be higher but the difference was not significant. For spinal cord, the maximum dose as well as NTCPcord were similar for 3D plans (D-max: p = 0.04; NTCP: p = 0.2) but were significantly lower for BIM (D-max: p = 0.002; NTCP: p = 0.008) if the biophysical objective function was used. Conclusions: When using conventional 3D techniques, inhomogeneous dose dist ributions offer the potential to further increase the probability of uncomp licated local control. When using techniques as BIM that would lead to larg e escalation of the median and maximum target doses, it seems indicated to limit target-dose inhomogeneity to avoid dose levels that are so high that the safety becomes questionable. (C) 1999 Elsevier Science Inc.