• Showcases
  • >
  • Real-world studies of SGLT2 inhibition and amputation risk

Real-world studies of SGLT2 inhibition and amputation risk

The results of the CANVAS program showed that patients with type 2 diabetes taking canagliflozin had an increased risk for amputation of the toes, feet, or legs, with a rate per 1000 patient–years of 6.3 versus 3.4 in the placebo group.

Of note, however, the amputation rate in the CANVAS placebo group was around half of that in the placebo and empagliflozin groups in the EMPA-REG OUTCOME trial, in which there was no evidence for an increased amputation risk. There was also no effect on amputation risk with dapagliflozin in the recently published DECLARE-TIMI 58, which is the largest of the sodium-glucose cotransporter (SGLT)2 inhibitor cardiovascular outcome trials, involving more than 17,000 participants.

Since the CANVAS findings, four large observational studies have been published, attempting to reproduce or refute the link between SGLT2 inhibition and amputation risk. The results, inevitably, conflict both with each other and with the SGLT2 inhibitor cardiovascular outcome trials; three of the real-world studies show a numerically higher rate of amputations with SGLT2 inhibitor use, which did not attain statistical significance and was not restricted to canagliflozin.

Complicating the interpretation are the variable cohort characteristics. For example, the average age of two of the cohorts was nearly 10 years younger than that of the CANVAS participants, and the studies vary as to which comorbidities they reported and how they were classed, and whether or not they included patients with previous amputations. Also, the average time on treatment was considerably shorter in these real-world studies than in the cardiovascular outcome trials.

Worth noting, however, is that the absolute risk for amputation remains very small, and all the cardiovascular outcome trials have shown clear benefits of the medication class, especially for heart failure.

Ryan PB, et al. Diabetes Obes Metab 2018; 20: 2585–2597

Patient population: Age: Observation time:
142,800 taking canagliflozin
110,897 taking other SGLT2 inhibitors
460,885 taking other antidiabetic medications
Median of 50 to 70 years Median 61 to 104 days

Published in June 2018, the OBSERVE-4D study looked at amputation rates specifically among new users of canagliflozin identified in four US administrative claims databases.

The incidence of below-knee amputation in the on-treatment analysis ranged from 1.0 to 5.0 events per 1000 person–years across the four databases. This did not vary significantly relative to the rates in propensity-matched users of other SGLT2 inhibitors and of all other antidiabetic medications, at hazard ratios of 1.14 and 0.75, respectively.


Adimadhyam S, et al. Diabetes Obes Metab 2018; 20: 2792–2799

Patient population: Age: Observation time:
30,216 taking SGLT2 inhibitors
30,216 taking dipeptidyl peptidase (DPP)-4 inhibitors
Mean of 55 years Median of 0.6 years

This study, published in July 2018, also used US insurance claims data to identify SGLT2 inhibitor users, compared this time specifically with propensity-matched users of DPP-4 inhibitors – “a clinically relevant therapeutic alternative” to SGLT2 inhibitors.

The researchers reported a numerically increased rate of amputations among the SGLT2 inhibitor users versus the DPP-4 inhibitor users, at 1.62 versus 1.15 per 1000 person–years. The 95% confidence interval for the nonsignificant hazard ratio (of 1.38) encompassed a 17% risk reduction and a 131% risk increase.

They found the possible increased risk to be driven by events in patients at overall low risk for amputations, with corresponding rates of 0.73 and 0.36 per 1000 person–years, although this difference was again nonsignificant. The rates among patients at high amputation risk were 3.50 and 3.78 per 1000 person–years.

Although again nonsignificant, the risk for amputation was higher among patients taking dapagliflozin or empagliflozin than among those taking canagliflozin, implying that the adverse effect, if real, was not restricted to canagliflozin.


Chang, H-Y, et al. JAMA Intern Med 2018; 178: 1190–1198

Patient population: Age: Observation time:
39,869 taking SGLT2 inhibitors
105,023 taking DPP-4 inhibitors
39,120 taking glucagon-like peptide (GLP)-1 receptor agonists
769,894 taking older antidiabetic medications
Mean of 52 years Median of 99 to 127 days

This was another study using US insurance claims data, which was published in August 2018. Rather than creating matched patient cohorts, as in the preceding studies, the researchers adjusted their findings according to a propensity score.

Crude amputation rates per 10,000 person–years ranged from 4.90 to 10.53 for new users of older diabetes medications and SGLT2 inhibitors, respectively. After adjustment, SGLT2 inhibitor users had nonsignificant risk increases for amputation of 1.50- and 1.47-fold relative to users of DPP-4 inhibitors and GLP-1 receptor agonists, respectively. And they had a significant 2.12-fold increased risk relative to users of older medications.


Ueda P, et al. BMJ 2018; 363: k4365

Patient population: Age: Observation time:
17,213 taking SGLT2 inhibitors
17,213 taking GLP-1 receptor agonists
Mean of 61 years Median of 274 days

This study, published in November 2018, compared new users of SGLT2 inhibitors with matched new users of GLP-1 receptor agonists. It was the first study to not be based on US insurance claims data, with patients instead identified in Swedish and Danish national registries.

The researchers reported amputation rates of 2.7 per 1000 person–years in the SGLT2 inhibitor users, compared with 1.1 per 1000 person–years in the GLP-1 receptor agonist users, equating to a significant 2.32-fold increased relative risk. However, the GLP-1 receptor agonist cardiovascular outcome trials have shown that at least three medications in this class protect against atherosclerotic vascular disease, which could in turn have a positive effect on amputation risk.

Of note, just 1% of the patients taking SGLT2 inhibitors were taking canagliflozin, again suggesting a class effect, despite the lack of evidence in the EMPA-REG OUTCOME and DECLARE-TIMI 58 trials.

By Eleanor McDermid

Don’t fall behind!

Get the latest headlines – in your specialties of choice – delivered straight to your inbox.



Eleanor McDermid

medwireNews Deputy Bureau Chief

Related Showcases