Objective: To evaluate the reproducibility of the clinical judgement of gynaecologists, gynaecologists in training and gynaecologic oncologists and to compare the predictive performance of the offhand assessment with the predictive performance of existing mathematical models for the pre-operative assessment of the adnexal mass.
Design: Questionnaire with paper cases of women operated on for an adnexal mass.
Setting: Gynaecological unit in a teaching hospital in the South of The Netherlands.
Population: Women who underwent surgery for adnexal mass between January 1991 and December 1998.
Methods: We offered 45 gynaecologists five different sets of 34 cases, with data on female age and menopausal status, a written description of the sonography, Doppler flow measurement and serum CA125 measurement. Nine observers for every set were asked to estimate the probability of malignancy.
Main outcome measures: The reproducibility of the risk estimates as made by the participants was expressed with an intraclass correlation coefficients. The accuracy of the judgement of the clinicians and the result of mathematical models in the prediction of malignancy were expressed with sensitivity, specificity, and receiver-operating characteristic curves.
Results: Neither clinically relevant nor statistically significant differences could be found between the accuracy of the risk assessments made by the clinicians and the accuracy of the risk assessments made by prediction models.
Conclusion: This study demonstrates that at this moment there is no need to introduce complicated predictive scoring systems such as neural networks or logistic regression models for the pre-operative assessment of the adnexal masses.