Breast cancer patient surveillance following curative-intent treatment: Variation among US census regions

Background: There are about 3 million breast cancer survivors in the US, with about 230,000 new cases of breast cancer diagnosed annually. Most patients receive curative-intent treatment. Post-treatment surveillance is commonly done and we have previously documented dramatic variation in surveillance intensity among ASCO experts. The only surveillance modalities endorsed by ASCO for asymptomatic patients are office visit and mammogram. ASCO guidelines specify that all other modalities are not recommended. We sought to determine the effect of the physician's practice location on the intensity of surveillance by comparing US Census Regions. Methods: The 3245 ASCO members who had indicated that breast cancer was a major clinical focus of their practice were surveyed regarding their surveillance practices. Members were asked to consider 4 idealized clinical vignettes and indicate their surveillance plan for each. A menu of 12 testing modalities was offered. Repeated-measures ANOVA was employed to compare practice patterns by US Census Region. Statistical significance was set at p<0.05. Results: There were 1012 responses; 915 were evaluable. The modalities most frequently recommended were office visit, CBC, LFTs, and mammogram. We analyzed the 4 US Census Regions plus a category called Foreign. Significant variation (p<0.05) in recommended frequency of use was present for all frequently recommended surveillance modalities. For example, in year 1 mammogram was recommended 1.5 ± 0.8 (mean ±SD) times in the Midwest Census Region and 1.9 ± 1.8 times in the South Census Region. Every modality discouraged by ASCO was recommended by some respondents. Conclusions: Evidence-based clinical guidelines have proven effective in providing high quality care and decreasing variation in practice. ASCO members often recommend surveillance modalities that are not recommended by the evidence-based ASCO clinical guidelines, suggesting overuse. Statistically significant variation among the US Census Regions was present. However, this variation was clinically quite small. Variation among US Census Regions cannot explain the known marked variation in clinical practice we have previously documented.