Canagliflozin and Cardiovascular and Renal Events in Type 2 Diabetes

n engl j med 377;21 nejm.org November 23, 2017 2097 sumptions add noise to the measurement. Scoring with item-response theory or Rasch analysis results in precision of measures that is up to 10 times as high as that of summary scoring.3 We suggest that the authors reanalyze their data using psychometric methods, such as Rasch analysis, that are based on item-response theory; such methods significantly reduce measurement noise and produce interval-level measurement.

The authors reply: The VR-12 is a well-established instrument for the assessment of healthrelated quality of life 1 and has been used in many settings, including the Medicare Advantage Program, quasi-experimental studies, and randomized, controlled trials. 1,2 The VR-12 scoring algorithms involve complex methods with weights that control for the effects of regression toward the mean and for imputation of missing values. 3 Collectively, the literature indicates that the physical and mental summary scores are sensitive to discriminating between clinically relevant groups and are responsive to change in pre-post prospective intervention studies and in random-ized clinical trials, and the ceiling effects for the VR-12 have not been found to be substantial. 4 In addition, a previous study that was conducted to develop links between scores on the Patient Reported Outcomes Measurement Information System (PROMIS) 29 and those on the VR-12 indicated that the psychometric properties of the VR-12 performed well in comparison with the PROMIS 29, a measure that is based on itemresponse theory. 5 Although a Rasch analysis could prove interesting, it is unlikely that it would change our conclusions.

Canagliflozin and Cardiovascular and Renal Events in Type 2 Diabetes
To the Editor: In the trials of the Canagliflozin Cardiovascular Assessment Study (CANVAS) Program (Aug. 17 issue), 1  No potential conflict of interest relevant to this letter was reported.

DOI: 10.1056/NEJMc1712572
To the Editor: Because of the transparency policy of the Journal, readers have access to exhaustive documentation of published trials. In the CANVAS trials, there are some surprising inconsistencies between the statistical analysis section of the trial protocols and the final article, affecting important points concerning the design of the two trials. The primary hypothesis, that "rela-tive to placebo plus standard of care, canagliflozin plus standard of care reduces CV [cardiovascular] risk," is stated as "a test of noninferiority, with the use of a margin of 1.3 for the hazard ratio for the primary outcome with canagliflozin as compared with placebo" in the published article, whereas it is described as just exploratory in the CANVAS-Renal (CANVAS-R) trial. Sample-size calculations in the original protocol are related to changes in the glycated hemoglobin level and not to cardiovascular outcomes and undergo several transformations before they settle down. Methodologic inconsistencies in both design and analysis items that should be clearly prespecified at the beginning of a trial arouse important concerns for the critical reader trying to ascertain the validity of a trial. No potential conflict of interest relevant to this letter was reported.

DOI: 10.1056/NEJMc1712572
To the Editor: In the CANVAS trials, the use of canagliflozin was associated with an increased risk of amputation. The majority of the amputations were minor and occurred distally. Although the mechanisms are still unclear, the fact that the preponderance of events occurred in patients with a history of amputation and peripheral arterial disease may offer some clues. In multivariate analysis, the most important predictor of amputation was previous amputation (hazard ratio, 20.9; 95% CI, 4.2 to 30.8). This raises the possibility that patients in the trial who had critical limb ischemia may have been vulnerable to hemoconcentration and alterations in viscosity, which are well known to occur with sodium-glucose cotransporter 2 (SGLT-2) inhibitors. 1 Patients with critical limb ischemia have seriously impaired perfusion, and even small changes in viscosity or perfusion may have had a deleterious effect on limb outcomes. Although the prevalence of peripheral-artery disease in the CANVAS trials (21% of all patients) was similar to that in the EMPA-REG OUTCOME trial, the reason why this was not observed in the EMPA-REG OUTCOME trial may have been related to the fact that data regarding events that were related to peripheralartery disease, such as amputations, were not collected. 2 Sanjay Rajagopalan, M.D.

University of Michigan Ann Arbor, MI
No potential conflict of interest relevant to this letter was reported.