[en] Purpose: To review the physical aspects of high dose rate (HDR) brachytherapy, including commissioning and quality assurance, source calibration and dose distribution measurements, and treatment planning methods. Following the introduction of afterloading in brachytherapy, development efforts to make it 'remote' culminated in 1964 with the near-simultaneous appearance of remote afterloaders in five major medical centers. Four of these machines were 'high dose rate', three employing 60Co and one (the GammaMed) using a single, cable-mounted 192Ir source. Stepping-motor source control was added to the GammaMed in 1974, making it the precursor of modern remote afterloaders, which are now suitable for interstitial as well as intracavitary brachytherapy by virtue of small source-diameter and indexer-accessed multiple channels. Because the 192Ir sources currently used in HDR remote afterloaders are supplied at a nominal air-kerma strength of 11.4 cGy cm2 s-1 (10 Ci), are not collimated in clinical use, and emit a significant fraction (15%) of photons at energies greater than 600 keV, shielding and facility design must be undertaken as carefully and thoroughly as for external beam installations. Licensing requirements of regulatory agencies must be met with respect both to maximum permissible dose limits and to the existence and functionality of safety devices (door interlocks, radiation monitors, etc.). Commissioning and quality assurance procedures that must be documented for HDR remote afterloading relate to (1) machine, applicator, guide-tube, and facility functionality checks, (2) source calibration, (3) emergency response readiness, (4) planning software evaluation, and (5) independent checks of clinical dose calculations. Source calibration checks must be performed locally, either by in-air measurement of air kerma strength or with a well ionization chamber calibrated (by an accredited standards laboratory) against an in-air measurement of air kerma strength for the same type of source. A separate planning procedure is required for each class of applicator to be used clinically, e.g., vaginal cylinder, endobronchial or esophageal catheter, cervix applicator, interstitial needles, and tumor-bed mold. Time constraints are such that, whenever all the dose-prescription points are at fixed locations relative to the applicator (even a flexible one), the use of planning atlases should be considered. Because source dwell time may be considered a continuous variable, analytic (as opposed to iterative) least-squares approaches to optimization are often feasible, with the advantage of a global (rather than local) minimum. Applicator-surface points or inter-catheter points may be represented in the objective function, to achieve greater dose uniformity throughout an implant. Other approaches include 'geometric' optimization, in which dwell time is made proportional to the sum of the inverse square of the distances from all other sources, and simulated annealing, an iterative process in which a global minimum is approximated more closely by accepting poorer solutions part of the time