Under a previously approved institutional review board protocol for prostate cancer patients, implanted metal-oxide semiconductor field-effect transistor dosimeters (dose verification system, Sicel Technologies) were used for measurement of the in vivo delivered daily dose. This dosimetric information provided the ability to adapt the plan if the measured doses did not match the dose expected from the planning system. Due to the inherent uncertainty in the dosimeters, the decision to adapt the treatment plan was made only if readings differed by more than 7 % for three consecutive days. To validate this method, we acquired daily cone beam computed tomography images for five patients, and the dose delivered to the dosimeters was calculated by use of (1) an automated procedure (MIM Maestro, MIM Software) and (2) the treatment planning system (XIO, Elekta). 72 % of the doses calculated automatically fell within 1 % of the doses calculated in the planning system, and 99 % agreed within 2 %. When compared to the calculated dose, 53 % of the in vivo measurements fell within 3 % of the calculated dose, 80 % fell within 5 %, and 9.8 % were greater than 7 %, but never on three consecutive days. The measured doses agreed reasonably well with the calculated doses, supporting the decision to adapt the plan only if there were discrepancies of more than 7 % over three consecutive days. Even with the inherent uncertainty in the dosimeters, this adaptive planning method can detect delivery inaccuracies that would not otherwise be caught with the use of only daily image guidance or other dose calculation surrogates.