Two Prediction Models Evaluate CA125 and Ovarian Cancer Risk
Posted: Monday, January 27, 2020
Two prediction models, one linear and one dichotomous, have been developed to help identify healthy women at risk of ovarian cancer based on elevated levels of the biomarker CA125. These models were validated in a study that sought to include the variation of personal characteristics in the use of circulating CA125 as a screening biomarker, according to research published in the Journal of Ovarian Research.
“Our dichotomous model could be used to identify healthy women who may have CA125 levels greater than the current clinical cutoff, which may contribute to reducing false-positive tests using CA125 as [a] screening biomarker,” concluded Naoko Sasamoto, MD, of Harvard Medical School, and colleagues.
Both the linear and dichotomous models were developed using data from women without ovarian cancer who had participated in one of the following studies: PLCO (n = 26,981), EPIC (n = 861), NHS/NHSII (n = 81), and the New England Case Control Study (NEC, n = 923). The models were developed based on stepwise regression from the PLCO data and validated using the data from EPIC, NHS/NHSII, and NEC.
According to the study authors, the linear model was successful in accounting for a small fraction (5%) of CA125 variance. It included age, race, body mass index, history of smoking, history of hysterectomy, age at menopause initiation, duration of prior hormone therapy, and history of childbirth. The dichotomous model, however, resulted in an improved performance, with an area under the receiver operating characteristic curve of 0.64 in data from PLCO and 0.80 in the validation set. This model included the same factors as the linear version, with the exception of parity. The predicted and measured CA125 values were similar in both the PLCO (r = 0.18) and validation (r = 0.14) data sets.
Disclosure: The study authors reported no conflicts of interest.