Both overtreatment and undertreatment of DCIS are pervasive problems given that no useful approaches for identifying which DCIS patients are likely or highly unlikely to go on to develop invasive breast cancer are currently available. The identification of useful markers could have considerable clinical impact in guiding both treatment and follow-up decision making.
Diagnosis of IBC in either breast at least 6 months after the diagnosis of DCIS.
A. Preliminary Phase
Candidate markers will be selected in the preliminary phase if they have good performance on their own (defined as TPR>0.3, corresponding to FPR= 0.2) or if they contribute substantially to a combination score (defined as increasing the TPR corresponding to FPR=0.2 by 0.05). TPR corresponding to FPR= 0.2 will be derived from time dependent ROC curves that will be calculated using event rates and minimal clinical data for subjects in the underlying cohort and biomarker data for subjects in the nested case-control subset. Time dependent ROC curves compare marker distributions in biomarker controls (defined as those alive and without IBC 7 years after diagnosis with DCIS) with marker distributions in cases (subjects who develop IBC within 5 years of DCIS diagnosis). Markers will be combined into a score using relative risk regression with cross validation to correct for over fitting. Time dependent ROC curves will then be calculated for the combination score. We will proceed with the evaluation phase only if data from the preliminary phase are sufficiently promising, i.e. a marker or marker combination is found that appears to have the target performance specified in the sample size calculations.
Although the preliminary phase concerns primarily selection of markers for evaluation, we will also use the preliminary phase data to test some hypotheses concerning markers recently reported by Kerlikowske, et al. for women treated with lumpectomy alone, namely p16, COX-2, and Ki-67,24 as well as any other promising candidates reported in the literature. Using the marker cut-offs defined by Kerlikowske and colleagues, we will compare sensitivities and specificities in our study with theirs. They also defined risk categories based on markers and mode of detection. We will test if rates of IBC in those categories are comparable estimated rates observed in our study for women treated with lumpectomy alone. Regardless of these results we will also perform analyses more extensive than those reported by Kerlikowske, including time dependent ROC analyses for individual markers and for marker combinations.
B. Evaluation Phase
In the evaluation phase more extensive analyses will be conducted for the preliminary phase marker combination and for individual markers selected in the preliminary phase.
Relative risks beyond those conferred by other predictors will be examined using relative risk regression models. Other predictors will include demographic factors (age, race/ethnicity, etc.), established breast cancer risk factors (e.g., first degree family history of breast cancer, use of menopausal hormone therapy, body mass index, etc.), DCIS tumor characteristics (e.g., size, histology, grade, etc.), DCIS treatments, and mode of DCIS detection.
To evaluate discrimination ROC curves will be calculated. ROC curves will be compared for different biomarkers. We will especially focus on the marker co
There are currently no biomarkers annotated for this protocol.
No datasets are currently associated with this protocol.
Thank you to everyone who made the 29th EDRN Steering Committee Meeting a success. The orientation for new and continuing EDRN PIs will be held October 15-16, 2015 on the NCI campus. More information will be available later this summer.