Search

Aims/hypothesis The primary analysis of the Canagliflozin cardioVascular Assessment Study (CANVAS) Program showed canagliflozin to have a beneficial effect on cardiovascular and renal outcomes in people with type 2 diabetes at high cardiovascular risk, but also an unexpected increased risk of major or minor lower extremity amputation. These secondary analyses explore this finding in more detail. Methods The effect of canagliflozin on amputation risk in the CANVAS Program was calculated for amputations of different types and proximate aetiologies and different canagliflozin doses. Univariate and multivariate associations of baseline characteristics with amputation risk were determined and proportional and absolute effects of canagliflozin were compared across subgroups. Results There were 187 (1.8%) participants with atraumatic lower extremity amputations (minor 71%, major 29%); as previously published, rates were 6.30 vs 3.37 per 1000 participant-years with canagliflozin vs placebo (HR 1.97 [95% CI 1.41, 2.75]). Risk was similar for ischaemic and infective aetiologies and for 100 mg and 300 mg doses. Overall amputation risk was strongly associated with baseline history of prior amputation (major or minor) (HR 21.31 [95% CI 15.40, 29.49]) and other established risk factors. No interactions between randomised treatment and participant characteristics explained the effect of canagliflozin on amputation risk. For every clinical subgroup studied, numbers of amputation events projected were smaller than numbers of major adverse cardiovascular events averted. Conclusions/interpretation The CANVAS Program demonstrated that canagliflozin increased the risk of amputation (mainly minor) in this study population. Anticipated risk factors for amputation were identified, such as prior history of amputation, peripheral vascular disease and neuropathy, but no specific aetiological mechanism or at-risk subgroup for canagliflozin was identified. Electronic supplementary material The online version of this article (10.1007/s00125-019-4839-8) contains peer-reviewed but unedited supplementary material, which is available to authorised users.

Several risk stratification systems have been proposed for predicting development of diabetic foot ulcer. However, little has been published that assesses their similarities and disparities, diagnostic accuracy and evidence level. Consequently, we conducted a systematic review of the existing stratification systems. We searched the MEDLINE database for studies (published until April 2010) describing the creation and validation of risk stratification systems for prediction of diabetic foot ulcer development. We included 13 studies describing or evaluating the following different risk degree stratification systems: University of Texas; International Working Group on Diabetic Foot; Scottish Intercollegiate Guideline Network (SIGN); American Diabetes Association; and Boyko and colleagues. We confirmed that five variables were included in almost all the systems: diabetic neuropathy, peripheral vascular disease, foot deformity, and previous foot ulcer and amputation. The number of variables included ranged from four to eight and the number of risk groups from two to six. Only four studies reported or allowed the calculation of diagnostic accuracy measures. The SIGN system showed some higher diagnostic accuracy values, particularly positive likelihood ratio, while predictive ability was confirmed through external validation only in the system of Boyko et al. Foot ulcer risk stratification systems are a much needed tool for screening patients with diabetes. The core variables of various systems are very similar, but the number of included variables in each model and risk groups varied greatly. Overall, the quality of evidence for these systems is low, as little validation of their predictive ability has been done.

In 2006 a risk stratification model was developed by Boyko et al. to predict foot ulceration in patients with diabetes, using seven commonly available clinical variables. We sought to validate and optimise this clinical prediction rule in a different setting. A retrospective cohort study was conducted on all patients with diabetes attending the podiatry section of a diabetic foot clinic at a tertiary hospital in Portugal (n = 360). Assessment at baseline included variables evaluated in the previous study and other relevant variables. Type 2 diabetes was present in 98% of patients, 45% were men and (at baseline) the median age was 65 years. Median follow-up was 25 months (range 3–86), during which 94 patients (26%) developed a foot ulcer. Boyko’s model had an area under the receiver operating curve of 0.83 (95% CI 0.78–0.88). The corresponding value for the optimised model, which included the footwear risk variable, was 0.88 (95% CI 0.84–0.91). Both models had high classification accuracy for prediction of foot ulceration. However, the optimised model tended to produce higher specificity and positive likelihood ratio values at all levels. This study confirmed that Boyko’s proposed model has a high capacity to predict foot ulceration in diabetes patients of both sexes. Our results suggest that the inclusion of a further footwear variable could improve the model. Nevertheless, prospective validation in a larger population is still necessary.