Ovarian cancer prediction: development of a scoring system for primary care.

OBJECTIVE

Recent studies have identified specific symptoms of ovarian cancer at all stages, raising the hope of reducing diagnostic delays. We aimed to devise a scoring system for symptoms of ovarian cancer in primary care.

DESIGN

Secondary analysis of data from a case-control study.

SETTING

Thirty-nine general practices in Exeter, mid-Devon and east Devon.

POPULATION: Two hundred and twelve women with ovarian cancer and 1060 age-, sex- and practice-matched controls.

METHODS

Conditional logistic regression was used to produce an additive scoring system and its receiver operator characteristic (ROC) curve. Several different cut-offs were then tested using a simple costs model.

MAIN OUTCOME MEASURES: The ROC curve value.

RESULTS

Each woman was assigned a score based on her symptoms in the year before diagnosis: we added a score for women aged ≥ 50 years, reflecting their increased incidence of ovarian cancer. The area under the ROC curve was 0.883 (95% confidence interval 0.853-0.912). The chosen cut-off had a sensitivity of 72.6% and a specificity of 91.3%.

CONCLUSION

This scoring system could potentially direct general practitioners to appropriate investigations for ovarian cancer on the basis of symptoms and save a substantial number of unnecessary ultrasound scans being requested.