We appreciate the interest of Leese and Morris (1) in our article and enjoyed reading their publication (2), which was published 1 month before our own and hence was not cited by us (3). We wish to point out that their characterization of our population as having been “recruited from a hospital diabetes clinic” is not correct, as we clearly describe that our subjects were recruited from a general internal medicine clinic (3). With regard to the generalizability of our findings, they should be applicable to the ∼25 million U.S. veterans and other males with similar clinical and demographic characteristics enrolled in a primary care setting. Although Leese and Morris write that they have “already addressed” foot ulcer prediction in their publication, they do not cite our publication from 1999, which presented a prediction model for diabetic foot ulcer and anticipated several of the findings of our recent article (4).
Leese and Morris describe their findings as useful for “all-comers” in a general community setting and in specialized clinics; they also suggest that their findings are valid for male and female subjects and for both types of diabetes (1). However, their data and analysis do not provide strong support for these statements for several reasons. First, although they used a population-sampling strategy that targeted 8,923 subjects with diabetes, only 3,526 (40%) underwent a clinical foot risk assessment. Thus, most subjects in the sampling frame were not included in the analysis, which certainly raises concerns regarding the validity of findings and limits the degree of confidence with which one can recommend these results for “all-comers.” Second, although both sexes and diabetes types were included in their study, the appropriate analysis of interaction to determine whether the results apply similarly in these subgroups of interest was not performed.
Leese and Morris bring up the issue of competing risks, in that individuals at higher risk for diabetic foot ulcer are also at higher risk for death. They refer to our “suspicion” that the death rate in people at high risk for ulcer is increased, but this would not be our preferred wording, as we feel certain about this association, having published data reporting this finding in 1996 (5). Comparison of cumulative risks, as in their report, may thus lead to biased results due to differential follow-up times by degree of risk. Fortunately, analytic methods that compare failure times and employ censoring, which we used in our recent publication, can address this problem.
Hopefully these recent efforts to develop foot ulcer prediction models and others will lead to more accurate identification of individuals at higher risk of foot ulcer and stimulate the development of better preventive strategies to reduce morbidity and mortality from these complications.
