In intraoperative high-dose-rate (IOHDR) brachytherapy, a 2-dimensional (2D) geometry is typically used for treatment planning. The assumption of planar geometry may cause serious errors in dose delivery for target surfaces that are, in reality, curved. A study to evaluate the magnitude of these errors in clinical practice was undertaken. Cylindrical phantoms with 6 radii (range: 1.35-12.5 cm) were used to simulate curved treatment geometries. Treatment plans were developed for various planar geometries and were delivered to the cylindrical phantoms using catheters inserted into Freiburg applicators of varying dimension. Dose distributions were measured using radiographic film. In comparison to the treatment plan (for a planar geometry), the doses delivered to prescription points were higher on the concave side of the geometry, up to 15% for the phantom with the smallest radius. On the convex side of the applicator, delivered doses were up to 10% lower for small treated areas (<or= 5 catheters) but, interestingly, the dose error was negligible for large treated areas (>5 catheters). Our measurements have shown inaccuracy in dose delivery when the original planar treatment plan is delivered with a curved applicator. Dose delivery errors arising from the use of planar treatment plans with curved applicators may be significant.