Use of Mastectomy for Overdiagnosed Breast Cancer in the United States: Analysis of the SEER 9 Cancer Registries

Figure 2

Sensitivity analysis showing the robustness of the study findings to omitted variable bias. We investigated the sensitivity of our main analysis to the existence of omitted variables that are associated with use of mastectomy. The sensitivity analysis is governed by a sensitivity parameter, which is the largest odds ratio (OR) by which the omitted variables can change probabilities of mastectomy from the values estimated in the main analysis. (In the main analysis, probabilities of mastectomy were estimated using 33 patient and case variables. However, some relevant variables were unavailable.) The figure presents results for several sensitivity parameter values, ranging from OR=1 to OR=25. Results for OR=1 are equivalent to the main analysis. On the other hand, OR=5 is equivalent to the omission of a very important determinant of mastectomy; among all the variables in Table 3, the largest OR was 4.9. Accordingly, the results in the figure show that even omitted variables with large ORs produce little change from the main analysis. Therefore, the study findings are largely robust to omitted variables. Appendix S1 includes a full description of the sensitivity analysis and definition of the sensitivity parameter.