We investigated the use of PCR (the Polymerase Chain Reaction) to detect the presence of infection in a group of patients undergoing revision arthroplasty for loose TJR (total joint replacement), compared to internationally agreed criteria used as the ‘gold standard’ for infection.

We prospectively tested samples taken from 108 patients undergoing revision arthroplasty (76 hips, 32 knees). Antibiotics were omitted prior to obtaining samples. DNA was extracted by 2 methods – a previously published technique (reference) and a commercial extraction kit (Qiagen^®). PCR involved amplification of an 882 base pair segment of the universal bacterial 16S RNA gene. During revision arthroplasty multiple specimens were taken from around the joint for microbiological and histological examination and the presence or absence of pus was noted. The patient was deemed to be infected if one of the following criteria was found: presence of a sinus pre-operatively; 2 or more intra-operative cultures positive for the same organism; an acute inflammatory response on histology; pus in the joint at revision.

Using the published DNA extraction technique PCR had a sensitivity of 50%, specificity of 93%, positive predictive value of 67% and negative predictive value of 88%. Using commercial extraction the sensitivity improved to 60%, specificity to 98%, positive predictive value to 90% and negative predictive value to 90%.

The previous report stated that PCR had a high sensitivity but a low specificity for detecting low grade infection. However, when using the published technique we found the opposite results – a moderate sensitivity and a high specificity. Introduction of a new DNA extraction technique improved the sensitivity. The refined PCR technique had a high accuracy, but further work is needed to improve sensitivity before we would recommend this method for routine clinical use.