Improving the accuracy of a photographic assessment system for breast cosmesis

Abstract

To determine how the accuracy of a photographic system for cosmetic assessment of the breast after conserving treatment can be maximised. Photographs of 40 women undergoing breast-conserving treatment for breast cancer were taken. Upward and lateral retraction was measured using a digitiser to enter the locations of reference marks on the photographs into a computer programme. In the readings produced, reduction in three potential sources of variation was considered by determining how few repeated photographs or measures were necessary to achieve an estimate within +/-2 mm of a mean value obtained from a large series. These included variation between assessments of the same photograph, between different photographs of the same patient and between different observers. We also sought to determine whether the attachment of white nipple markers (WNM) at the time of photography would reduce variation. A minimum of five readings was required from each photograph in order to obtain a result that was stable (defined as within 1 mm of that achieved by 20 readings). Using similar criteria, we found that at least four photographs of each patient were required, and that each should be assessed by at least four observers. The addition of WNM reduced the standard deviations by 36-45%. The use of WNM, which is simple, quick and cheap, is effective in reducing variation and can be recommended for use in clinical trials where overhead costs need to be contained.