Background and purpose: The aim of this study is to evaluate and delineate the deficiencies in conventional two-dimensional (2-D) radiotherapy planning of nasopharyngeal carcinoma (NPC) treatment and to explore the means for improvement of the existing treatment technique aiming at enhancing local tumor control and reducing treatment complications.Methods and materials: Ten patients with NPC sparing the skull base and without intracranial extension or cranial nerve(s) palsy were chosen in the present study. Two sets of CT images for Phases I and II of the radiotherapy treatment were taken with patient immobilized in the flexed-head and the extended-head positions, respectively. Based on the CT images and endoscopic findings, the gross tumor volume (GTV) was defined. The clinical target volume (CTV) circumscribing the GTV was defined according to Ho's (Halnan, K.E. (ed.) Treatment of Cancer. London: Chapman and Hall, 1982. pp. 249–268) description of the organs at risk of tumor infiltration. The planning target volume (PTV) was defined by adding a margin to the CTV which catered for geometrical inaccuracies. The field borders and shields were set at standard distances from certain bony landmarks and were drawn on the simulator radiograph. Data on the beams and shield arrangements were then transferred to the planning computer via a digitizer. By applying 3-D volumetric dose calculation using a commercial three-dimensional (3D) treatment planning computer, the dose–volume-histograms (DVHs) of GTV, CTV, PTV and critical normal organs were generated for both phases of Ho's treatment technique. The same patients were re-planned using a modified Ho's technique which used 3-D beams-eye-view (BEV) in placing the shielding blocks and the same set of DVHs were generated and compared with those obtained from Ho's technique.Results: The median volumes of GTV, CTV and PTV covered by the 95% isodose in Ho's phase I treatment were around 60%. The dose coverage was unsatisfactory in the superior and inferior and the posterolateral regions. In phase II treatment, the median volume of GTV, CTV and PTV covered by the 95% isodose were 99, 96 and 72%, respectively. Even though the dose coverage of the PTV in both phases of treatment were unsatisfactory, radiotherapy with the original Ho's technique had consistently produced good local control for NPC. However, there is potential room for enhancing the local control further because after modifying Ho's technique by using 3-D BEV customization of the treatment portals, the median volume of the target covered by the 95% isodose was defined as V95. The V95 of the PTV during the Phase II treatment was improved by 13%. The 90% of the volume of temporo-mandibular joints and parotid glands were both irradiated to 53 Gy and 43.6 Gy of the total prescribed dose of 66 Gy, respectively, in phase I and II treatments. With the addition of a hypothalamus-pituitary shield to Ho's technique, 50% of the volume of optic chiasma and temporal lobes received, respectively, 19.3 Gy and 4.5 Gy. However, small volume of the temporal lobes received a maximum dose (Dmax) of 62.8 Gy (95.2% of 66Gy). Most of the brainstem was shielded from the lateral portals but 5% of its volume received a dose ranging from 25.4 to 50.4Gy. The spinal cord (at C1/C2 level) received a Dmax of 40.8 Gy in phase I and of 4.8 Gy in phase II. After modifying Ho's technique by 3-D BEV customization of the treatment portals, the Dmax to the brainstem, the optic chiasma and the temporal lobes could be reduced by 8, 12 and 5%, respectively.Conclusions: Our study indicated that the dose-coverage of the PTV in Ho's radiotherapy technique for the early T-stage NPC was less than satisfactory in the superior and inferior and the posterolateral regions. However, in view of the excellent historical local tumor control with Ho's technique, we have to postulate that the present definition of CTV (and hence the PTV after adding margins to the CTV) lacks clinical significance and can be improved. It appears that the inclusion of the entire sphenoid sinus floor and both medial and lateral pterygoid muscles in the CTV is not necessary for maximal tumor control in the absence of clinical/radiological evidence of tumor infiltration of these organs. Ho's technique can be improved by using 3-D BEV to customize the treatment portals with multileaf collimators or blocks.