To estimate the impact of dose calculation approaches adopted in different treatment planning systems (TPSs) on proton therapy dose delivered with pencil beam scanning (PBS). Treatment plans for six regular volumes in water and 15 clinical cases were optimized with Syngo-VC13 and exported for forward recalculation with Raystation-V7.0 pencil beam (RS-PBA) and Monte Carlo (RS-MC) algorithms and with the independent Fluka-MC engine. To verify clinical consistency between the two TPS dosimetric outcomes, the average percentage variations of clinical target volume (CTV) D 98%, D 50% and D 2%, adopted for plan prescription and evaluation, were considered. Ionization chamber measurements served as a further reference for comparison in homogeneous conditions. CTV dose volume histogram (DVH) analysis and gamma evaluation with 3 mm-3% agreement criteria quantified the dose deviation of TPS calculation algorithms, in heterogeneous conditions, against the Fluka-MC code. CTV D 50%, representing the plan dose prescription goal, was higher on average over H&N cases of (3.9 ± 0.9)% and (2.3 ± 0.6)% as calculated with RS-PBA and RS-MC, respectively, compared to Syngo. For tumors located in the pelvis district, average D 50% variations of (1.6 ± 0.7)% and (1.2 ± 0.7)% were found. Syngo underestimated target near maximum doses with respect to all computation systems. Calculation accuracy in heterogeneous conditions of RS-PBA H&N plans resulted poor when a range shifter was required. Target DVH and γ-analysis showed excellent agreement between RS-MC and Fluka-MC, with γ-pass rates >98% for all patient groups. Different TPS dose calculation approaches mainly affected dose delivered in H&N proton treatments, while minor deviations were found for pelvic tumors. RS-MC proved to be the most accurate TPS dose calculation algorithm when compared to an independent MC simulation code.