Helical tomotherapy is an increasingly common form of intensity modulated radiation therapy that allows for image guided adaptive radiotherapy. Its treatment planning system (TPS) uses a convolution superposition algorithm for dose distribution calculations. The accuracy of this algorithm in the presence of heterogeneities was evaluated against Monte Carlo (MC) calculations and measurements. This work performed BEAMnrc-and DOSXYZnrc-based MC dose calculations of tomotherapy deliveries to a CIRS anthropomorphic heterogeneous phantom with typical clinical inverse planning and delivery settings. Point measurements with A1SL ion chambers and relative measurements with Kodak EDR2 film were carried out in the phantom. The experimental results were used to evaluate both the TPS and MC dose calculations. Furthermore, the dose distribution for a clinical head-and-neck cancer plan was calculated on the TPS and MC systems. The results support this MC system as a viable option for the accurate simulation of the tomotherapy process in the presence of heterogeneities where direct measurement may not be practical. Ion chamber measurements in the CIRS phantom suggested the TPS has an average relative difference of 2.3%, with the largest difference being -4.1% in one of the organs at risk. The MC system accurately predicted the dose to these measurement points within statistical uncertainty. The film measurements in the CIRS phantom demonstrated 90.7% (of pixels) agreed with the MC system using a +/-3%/3 mm acceptance criteria, where only 50.3% agreed with the TPS. In the clinical head-and-neck cancer plan evaluation where MC served as a reference against which to compare the TPS result, an average of 92.7% of the voxels within volumes of interest passed a 3%/3 mm criteria. The PTV54 showed the worst agreement with 85.4% of the volume passing the 3% /3 mm criteria. In general, the +/-3%/3 mm criterion was found to be a challenge for the TPS in the presence of lung inhomogeneity.