We read Dr. Hart’s (1) response to our article with great interest. The issue of racial differences in the presence of variant hemoglobins that may affect HbA1c (A1C) test results is certainly an important one. Ours (2) was a retrospective analysis using electronic medical record data that did not contain information on either the presence of sickle hemoglobin or the results of patient self-monitoring of blood glucose (SMBG) testing. However, because we found persistent differences in A1C lab values by race, even when controlling for individual-level A1C at baseline in our multivariate analyses, we do not believe the presence of sickle hemoglobin in 8% of our population would eliminate the racial disparities we observed. Still, the issue of measurement raised by Dr. Hart is worthy of discussion. Because of possible variations in the calculation of A1C over time, we ran several diagnostic tests on our A1C measures to test for systematic differences in measurement over time by race. While we did not identify shifts in A1C by race, we did find a shift in A1C values for the entire cohort midway through our study period due to a change in the calculation of A1C by an external vendor. As stated in our article (2), we adjusted for this change using statistical techniques and found no race-based differences in the effect of this adjustment.
We agree with Dr. Hart that a combination of patient SMBG and A1C results represents a better standard for assessing actual control. Unfortunately, rates of SMBG testing in this population were below optimal and were particularly low for black patients. Furthermore, information from patient SMBG is not consistently recorded in the medical record. For this reason, we are now exploring strategies for increasing SMBG among all diabetic patients, especially black patients. We are also exploring interventions that would incorporate patient data from both lab A1C testing and SMBG values in clinical decisions.