Patients with nasopharyngeal carcinoma (NPC) treated with Helical Tomotherapy (HT) are commonly positioned in an extended neck (EN) or a flexed neck (FN) position. The two set-up positions may have different dosimetric influences to the HT plans and to the reproducibility of HT treatments. The study was to investigate possible dosimetric benefits of EN and FN set-up positions, especially in intracranial structures and to investigate the their reproducibilities in the neck regions of these two set-up positions throughout the treatment course. Twenty NPC patients with various disease stages from each set-up group, treated with HT from 2014 to 2016, were conveniently recruited. All organs-at-risk (OARs) were re-contoured and the nasopharyngeal target volumes were standardized. Forty HT plans were re-computed using the same target dose prescription and OARs constraints. Conformity Index (CI), Homogenity Index (HI), maximum, minimum, mean doses of the targets and the maximum and mean doses of OARs were compared. Apart from HT re-planning, past daily megavoltage computed tomography (MVCT) images of each patient were retrieved and re-registered with the matching focus on the clivus. Set-up errors on the cervical level of C4 and C7 were obtained and compared between two groups. The CI of PTV-NP in FN group was significantly higher than that of EN group(p<0.05), while the CIs of both neck targets in FN group were significantly smaller than those of EN group (p <0.05). All target volumes of FN group had significantly higher HI values than those of EN group with all p < 0.05. For OARs in FN group, both cerebellum and hippocampus received significantly lower maximum and mean doses and the temporal lobes received significant lower mean doses. In contrast, all optic structures, brainstem, spinal cord and parotid glands received higher doses in FN group. For set-up errors calculation, the systematic errors in FN group at both C4 and C7 level were larger in both superoinferior (SI) and anteroposterior (AP) directions than those in EN group. FN position was beneficial in reducing the received radiation doses to intracranial structures including cerebellum, temporal lobes and hippocampus and thus minimize chronic neurocognitive impairments as side effects of HT treatment of NPC. On the contrary, FN position at the same time might increase radiation dose to optic structures and was more prone to greater deviations in SI and AP directions. Hence, reducing target margin near optic structures, applying planning organ-at-risk volume to the optic structures and modifying the FN set-up devices would improve the performance of FN position in dosimetric outcome and reproducibility. Otherwise, extra efforts in suppressing radiation dose to cerebellum, hippocampus and temporal lobes should be considered in an EN setting to minimize late complications to these structures.