To the Editor.
—I read with interest the article by Dr Grayson and colleagues.1 The authors appropriately offered a critique of prior studies, citing retrospective design, small numbers of patients, and lack of diagnostic criteria. However, they failed to discuss the limitations of their study. In an ideal study of contiguous osteomyelitis complicating pedal ulcers, each patient would have undergone both the index test (probing to bone) and the "gold" standard test (histologic diagnosis). The authors referred for histologic diagnoses only 46 (92%) of 50 patients with a positive probe test and only seven (27%) of 26 patients with a negative probe test. This represents test referral bias2 whereby patients with a positive index test (probing to bone) are more likely to receive the criterion standard test (histologic diagnosis). As a result, both the true positives and false positives are overrepresented in the study sample. If the clinical