The purpose of this study was to analyse the occurrence of non-visualisation during preoperative lymphoscintigraphy for sentinel node identification in breast cancer. Preoperative lymphoscintigraphy was performed in 495 clinically node-negative breast cancer patients (501 sentinel node procedures) after injection of technetium-99m nanocolloid. Anterior and prone lateral (hanging breast) planar images were obtained a few minutes and 4 h after injection. The sentinel node was intraoperatively identified with the aid of patent blue dye and a gamma-ray detection probe. A sentinel node was visualised on the 4-h images in 449 of 501 procedures (90%). This visualisation rate improved from 76% to 94% during the study period. Delayed imaging (5-23 h) in 19 patients whose sentinel nodes failed to show, resulted in visualisation in four of them. A repeat injection of radiocolloid in 11 patients revealed a sentinel node in six. In the end, the visualisation rate was 92%. The sentinel node was surgically retrieved in 24 of the remaining 42 patients with non-visualisation (57%). Sentinel nodes that were visualised were tumour-positive in 38% and non-visualised sentinel nodes were involved in 50% (chi2, P=0.17). In a multivariate regression analysis, scintigraphic non-visualisation was independently associated with increased patient age (P<0.001), decreased tracer dose (P<0.001) and increased number of tumour-positive lymph nodes (P=0.013). The use of a sufficient amount of radioactivity (at least 100 MBq) is recommended for lymphatic mapping in breast cancer, especially in elderly women. Delayed imaging and re-injection of the radioactive tracer increase the visualisation rate. The non-visualised sentinel node can be identified intraoperatively in more than half of the patients.