This slide shows a comparison of the 95% confidence limits with the limits of the sensitivity analysis based on external adjustment.  In the external adjustment shown, we used the most extreme assumptions as follows: The lower sensitivity limit was based on the high risk subgroup’s relative risk of breast cancer = 3.0, proportion of population in high risk subgroup = 25%, and odds ratio of reserpine use by high risk subgroup = 2.0.  The upper sensitivity limit was based on the low risk subgroup’s relative risk of breast cancer = 0.33, proportion of population in low risk subgroup = 50%, and odds ratio of reserpine use by low risk subgroup = 2.0.  Note that the uncertainty due to chance is greater than the uncertainty due to a high degree of confounding.  We have often found this true except for very commonly used drugs where large numbers of exposed cases can yield narrow 95% confidence intervals.