Primary care risk tool could help to identify ovarian cancer earlier
MedWire News: UK researchers have devised a risk algorithm based on simple clinical variables that they say could be used in primary care to identify those women who are most at risk for having as yet undiagnosed ovarian cancer.
The risk tool could be integrated into clinical computer systems and "potentially be used to identify those at highest risk of ovarian cancer to facilitate early referral and investigation," report Julia Hippisley-Cox and Carol Coupland, from University of Nottingham, in the BMJ.
It has been estimated that 10% of deaths from ovarian cancers might be avoidable, the researchers note. Among efforts to improve early diagnosis, tools to help assess absolute risk for ovarian cancer are needed, they add, to help ensure the right patients are investigated and the optimal use of scarce resources including abdominal and transvaginal ultrasonography, computed tomography, or magnetic resonance imaging.
To develop the model, Hippisley-Cox and Coupland analyzed data for 1,158,723 women aged 30-84 years registered at 375 general practices in QResearch (a large UK primary care database). They focused on established predictor variables that are likely to be recorded in the patient's electronic record and/or to be known to the patient, and the diagnosis of incident ovarian cancer over 2 years.
A total of 976 incident cases of ovarian cancer were recorded (either by the general practitioner or in a linked death record) during the observation period, giving a crude rate of 48 per 100,000 person-years.
Nine clinical factors independently predicted ovarian cancer in multivariate analysis and so were included in the final model. These were age, family history of ovarian cancer, anemia, abdominal pain, abdominal distension, rectal bleeding, postmenopausal bleeding, and appetite and weight loss.
Subsequent validation of the risk model, in a further 608,862 women aged 30-84 years from another 189 QResearch general practices, indicated that the algorithm showed good discrimination, explaining 57.6% of the variation in the time to diagnosis of ovarian cancer and having a receiver operating characteristic curve statistic of 0.84, and a D statistic of 2.38.
Furthermore, the model was well-calibrated, the team reports, showing close correspondence between observed and predicted 2-year risks across each 10th of predicted risk.
Finally, Hippisley-Cox and Coupland report that 63% of all new cancers diagnosed in the validation cohort occurred among women in the top 10% of predicted risk.
They say that software implementing the algorithm could be used to calculate risk from information already recorded in the electronic health record, and patients at highest risk for an existing but as yet undiagnosed ovarian cancer thereby identified for clinical assessment.
Nevertheless, they conclude: "Further research is needed to assess how best to implement the algorithm, its cost effectiveness, and whether, on implementation, it has any impact on the stage of ovarian cancer at diagnosis and subsequent survival."
By Caroline Price