The purpose of our study was to describe the relationship between settings for initial treatment and outcomes from breast cancer on the basis of available data.1 In our paper we acknowledged the limitations of these data. Nevertheless, we believe that it is important to publish such results to promote discussion. Improvement and accountability in our health care system are contingent on access to such information.
Robert Myers, T.J. Muckle, Peter Willard and Brian Higgins would prefer that our results be attributed to differences in patient characteristics. We adjusted for the differences between the patient populations using statistical analysis, which increased the difference in survival from an overall relative risk of 0.67 (95% confidence interval [CI] 0.53–1.03) to an adjusted relative risk of 0.47 (95% CI 0.23–0.96) for women with tumours less than or equal to 20 mm in diameter. Unfortunately, the manner in which the tumour was detected was not routinely recorded by clinicians.
As Myers notes, information about tumour grade was more likely to be missing for women seen in community hospitals. These women experienced poorer survival than women with moderate-grade tumours. Conversely, information about estrogen receptor status was more likely to be missing for women seen at teaching hospitals, and these women experienced better survival than those with positive tumours. This could be viewed as a source of misclassification of patients or as an indicator of differences in the process of care. In either case, we believe it is important to examine the relationship with outcomes, as well as reasons for such potential differences in processes of care.
The study by Golledge and colleagues was a single-hospital study that looked at outcomes before and after introduction of specialization.2 It did not, and could not by design, comment on impact of teaching status. Other British studies that have defined specialization in terms of teaching hospital status,3 surgeon's workload4 and local perception5 have also found differences in survival. It is perhaps premature to conclude which aspect of specialization contributes to differences in outcome.
Robert Fingerote takes exception to wording in the introduction of the article and suggests that the study is biased. In fact, we were very careful to maintain a balanced approach in discussing possible interpretations of our results. It is our view that our article's wording is far more balanced than that of our correspondents.
Breast cancer treatment occurs within a complex system involving radiologists, surgeons, radiation and medical oncologists, pathologists and nurses, among others. We considered the initial treatment setting (a system of care) rather than the skills of individual clinicians. If the difference in survival that we observed can be attributed to differences in the process of care, we need to determine which element of the care provided at teaching hospitals is responsible for the differences and whether it can be applied to the community setting, particularly since the majority of women are initially seen at community hospitals.