Length time bias is a form of selection bias that lead to the perception that screened patients have better outcome as more indolent tumors are diagnosed during screening. However, tumors diagnosed in the interval between mammographys or detected by symptom onset are likely more aggressive. The aim was to analyze using a genomic platform if unscreened tumors were more agresive than screened ones in a homogeneous cohort not affected by stage or subtype.
Since 2014 BC pts with T1-T2 N0-N1mic tumors and/or high ki67 are selected for genomic platform-based risk assessment in order to guide adjuvant treatment. We performed an exploratory retrospective cohort study in a single institution between 2014 and 2018 in operated stage I-IIA BC pts with ER and/or PR + who underwent an Oncotype risk assessment before deciding adjuvant therapy. Results of the Recurrence score (RS) were compared according to the type of diagnosis of breast cancer as 1) Screened: Diagnosis during screening (when diagnosed occurred during a foreseen mammography visit) and 2) Unscreened: Diagnosis occurred outside screening (in an interval between mammographys or by symptom onset).
105 pts were included. Median age was 56.4 y (45.2-74.6 y). All patients were ER + (range 50-100%), HER2- and grade was I (12.4%) II (82,9%) or III (4.8%). Median tumor size was 13.7 mm (4-45). 89.5% were N0 and 10.5% N1mic. 68 pts (64.8%) were in the screened and 37 (35.3%) in the unscreened group. Foreseen adjuvant treatment was changed according to Oncotype results in 24.8% patients. Median RS was 18.2 (range 3-46). According to RS risk categories 81% were classified as low , 9.5% as intermediate and 9.5% as high risk. Median RS was 17.6 in the screened vs 19.2 in the unscreened group, these differences were not significant (p = 0.34). Differences by RS categories were also not significant (Chi square p = 0.67 for two categories low vs intermediate/high risk with a RR 0.92 (0.62-1.38) and p = 0.2 for three categories).
No risk differences according to RS was seen between screened vs unscreened patients. These suggest that length time bias in a cohort not affected by stage or subtype might have mínimum impact on screening outcomes.
Clinical trial identification
Legal entity responsible for the study
Hospital Clinico Universitario de Valencia. INCLIVA.
Has not received any funding.
All authors have declared no conflicts of interest.