Aim:

To determine whether there was an explanation as to why the effects of the sodium‐glucose co‐transporter‐2 (SGLT2) inhibitor canagliflozin on amputation risk vary between the CANVAS program and the CREDENCE trial.
Methods:

We performed a pooled analysis of patient‐level data from the CANVAS program and the CREDENCE trial. Patient characteristics associated with amputation risk were assessed in univariable and multivariable regression models and compared between studies. Effects of canagliflozin on amputation risk were determined from Cox proportional hazards models and compared between studies, subgroups and for a range of amputation outcomes. Effects over time were explored by cumulative event curves.
Results:

In the CANVAS program (n = 10 142; median follow‐up 2.4 years) and CREDENCE trial (n = 4401; median follow‐up 2.5 years), 2.3% and 5.3% of participants, respectively, reported baseline amputation history. Key differences at baseline were the proportions with nephropathy (CREDENCE higher, 100% vs. 17.5%) and cardiovascular disease (CANVAS higher, 66% vs. 50%). There were 133 amputations in CREDENCE (3.0% annual event rate) and 187 amputations in CANVAS (1.8% annual event rate), with prior amputation being the strongest predictor of future amputation in both groups. Effects of canagliflozin on amputation risk were significantly different between trials (P heterogeneity .02, I2 = 82%), but this was not explained by participant or trial differences. There was no evidence that foot disease management protocols instituted during CREDENCE ameliorated amputation risk.
Conclusions:

We identified no explanation for the difference in amputation risk between the CREDENCE trial and the CANVAS program. In the context of null effects of SGLT2 inhibition on amputation in CREDENCE and all other large trials, there is a possibility that the finding in CANVAS was the result of chance.