Abstract The study investigated the dosimetric impact of an iterative metal artifact reduction (iMAR) tool on carbon ion therapy for pelvic cancer patients with hip prostheses. An anthropomorphic pelvic phantom with unilateral and bilateral hip prostheses was used to simulate pelvic cancer patients with metal implants. The raw data obtained from phantom CT scanning were reconstructed with a regular filtered back projection (FBP) algorithm and then corrected with iMAR. The phantom without hip prosthesis was also scanned and used as a reference ground truth (GT). The CT images of three prostate and four sarcoma patients with unilateral hip prosthesis were also reconstructed by FBP and iMAR algorithm and compared. iMAR algorithm reduced the metal artifacts and the maximum WEPL deviation in phantom images from −19.1 to −0.4 mm. However, the CT numbers cannot be retrieved using iMAR for periprosthetic bone materials, eventually leading to a WEPL deviation of −3.6 mm. The use of iMAR improved large discrepancies in DVHs of PTVs and the gamma index between FBP and GT images but increased the difference in the bladder DVH for bilateral hip prostheses due to newly introduced artifacts. In the patient study, the discrepancies of dose distribution were small on iMAR images when compared with FBP images for most cases, except for two sarcoma cases where gamma analysis failed and dose coverage in 98% of the PTV maximally reduced due to large volume of dark metal artifacts. iMAR reduced the metal artifacts and improved dose distribution accuracy in carbon ion radiotherapy for pelvic cancer. However, the residual and newly introduced artifacts, especially with bilateral hip prostheses, may potentially increase WEPL inaccuracy and dose uncertainty. The use of iMAR has the potential to improve carbon ion treatment planning of pelvic cancer but should be used with caution.