Purpose: To examine the concordance of various measures to Bayesian Confidence Propagation Neural Network (BCPNN) measure used by the WHO Collaborating Centre for International Drug Monitoring based on the database of Guangdong province in China.

Methods: The Reporting Odds Ratio (ROR)--1.96 standard errors (SE), Proportional Reporting Ratio (PRR)--21.96 SE, combination chi (2)-PRR measure used by the Medicines and Healthcare Products Regulatory Agency (MHRA) were compared with the IC--2 standard deviations (SDs) with the tested database collected from 1 January 2002 to 30 June 2007 in Guangdong. Additionally, the concordance of the various measures, in respect to the number of reports per combination, was examined.

Results: Sensitivity, specificity, positive predictive value and negative predictive value were high in respect to the BCPNN measure when a combination of point and interval estimate was used based on four types of signal detection for the ROR and PRR measures, and the four evaluation indexes increased when A (the number of reports per combination) increased. However the MHRA measure had decreasing sensitivity and increasing specificity with A increasing.

Conclusions: The ROR, PRR and BCPNN measures used are broadly comparable when three or more cases per combination have been collected. The MHRA measure is comparable to the BCPNN when seven or less case per combination have been collected for 'drug-ADR', 'drug category-ADR', 'drug-ADR category' combinations based on the Guangdong database, and the MHRA is unsuitable for detecting signals from 'drug category-ADR category' combinations based on the Guangdong database because of its low sensitivity.