Reviews/SGLT2 Inhibitors/Canagliflozin is associated with increased risk of lower limb amputation
· SG-SF-02 · v2
The strongest evidence here is a large randomised trial programme, CANVAS, which found roughly twice as many lower-limb amputations in people taking canagliflozin as in those taking placebo [1][6]. Several real-world studies of patient records pointed the same way, finding a similar near two-fold signal [9][12][13]. But the picture is not uniform: a second randomised trial, CREDENCE, found no amputation effect [5], and when the two trials were pooled the difference between them was large enough that chance could not be ruled out as the explanation for the CANVAS result [7]. A very large claims-database analysis also found no excess amputations compared with other diabetes drugs [10]. The panel concluded that an association was clearly seen in CANVAS and echoed elsewhere, but that the genuine counter-evidence stops it being treated as settled.
The pivotal CANVAS programme reported amputation rates of 6.3 versus 3.4 per 1000 patient-years, HR 1.97 (1.41-2.75), mainly minor toe or metatarsal amputations [1][6]. The signal was consistent across albuminuria subgroups within that trial (P heterogeneity=0.66) [2]. Independent observational cohorts corroborated a roughly two-fold magnitude: Fralick found HR 1.66 (1.33-2.07) in adults ≥65 [9], Udell's EASEL cohort HR 1.99 (1.12-3.51) for below-knee amputation [13], and Chang HR 2.12 (1.19-3.77) for the SGLT-2 class versus older agents [12]. No single number is supplied outside these abstracts. The contested point: The division is over whether the CANVAS signal is a durable, generalizable drug effect or a chance finding confined to one trial and one population. CREDENCE showed no amputation effect [5], the pooled CANVAS+CREDENCE analysis found significant between-trial heterogeneity (Pheterogeneity .02, I2=82%) and raised chance explicitly [7], and OBSERVE-4D was null versus non-SGLT2i on an intent-to-treat basis, HR 1.01 (0.93-1.10) [10].
All 16 papers behind this claim with their DOIs and what each does to it, the panel’s arbitration, the evidence gap map and the open research questions. Free, and sent to your inbox.
We’ll also send you other claims from SGLT2 Inhibitors, one at a time, and tell you when a release or a pivotal trial changes a grade. Unsubscribe by replying.
One reviewer's only job is to attack the finding. The challenges below were weighed before the grade was set.
The association is inconsistent across the pivotal trials: CREDENCE [5] found no amputation effect and the pooled CANVAS+CREDENCE analysis [7] reported significant between-trial heterogeneity (I2=82%), explicitly raising chance as the explanation for the CANVAS signal.
The largest rigorously designed real-world cohort, OBSERVE-4D [10], found HR 1.01 (0.93-1.10) ITT versus non-SGLT2i, i.e. no excess against the most relevant comparator, undercutting the claim's generalizability beyond CANVAS.
Where it remains uncertain. CREDENCE found no effect and the pooled analysis reported I2=82% heterogeneity raising chance [5][7], while OBSERVE-4D found no excess versus non-SGLT2i, HR 1.01 (0.93-1.10) [10].
Aleatory
The pooled analysis found I2=82% between-trial heterogeneity not explained by trial or patient differences, and explicitly raised chance as an explanation for CANVAS [7].
Epistemic
Matthews identified risk factors but no specific aetiological mechanism or at-risk subgroup for canagliflozin [6], so no biological account constrains interpretation.
Inferential
OBSERVE-4D found no excess versus non-SGLT2i in routine data, HR 1.01 (0.93-1.10) ITT [10], so extrapolating the CANVAS effect to real-world prescribing is uncertain.
The load-bearing papers from a corpus of 16, each classified by what it does to the claim. Every DOI is taken from the corpus the claim was graded on.
Pivotal RCT evidence of a significant amputation association.
Panel auditInternal validity: high · scope of inference: Type 2 diabetes patients at high cardiovascular risk, mean age 63, treated with canagliflozin.Key limitations: amputation was a secondary safety finding · high-cardiovascular-risk population only
Shows the CANVAS amputation signal is consistent across albuminuria strata.
“including increased risks for amputation across albuminuria subgroups ( P heterogeneity=0.66)”
Panel auditInternal validity: medium · scope of inference: Same CANVAS population, analysed by baseline albuminuria level.Key limitations: post-hoc subgroup analysis · no effect size for amputation reported in abstract
Primary counter-evidence: CREDENCE found no amputation effect.
“Canagliflozin's lack of effect on serious adverse events, amputations, and fractures appeared consistent among eGFR subgroups.”
Panel auditInternal validity: high · scope of inference: Type 2 diabetes with chronic kidney disease and substantial albuminuria.Key limitations: no amputation effect size in abstract · nephropathy-enriched population
Before the EMA warning, SGLT2i initiation carried significantly higher amputation risk in patients without PAD, though the difference disappeared after the warning.
Panel auditInternal validity: medium · scope of inference: German type 2 diabetes patients initiating SGLT2i, with and without peripheral arterial disease.Key limitations: class-level not canagliflozin-specific · effect attenuated after EMA warning · observational
4 of 16 papers shown. The full review lists all of them, with the panel’s arbitration →
Claims are graded at each quarterly release, and re-graded between releases only when a pivotal trial lands. Every change is a new version of the claim.
An association was robustly observed in CANVAS and echoed by independent datasets, but heterogeneity and null real-world and CREDENCE findings warrant a Likely rather than Established verdict.
Release 2026-q4 re-graded every claim from its corpus under protocol v0.19. An earlier release's grade is shown for the record; it came from an earlier protocol and is not a movement.
Assumptions that carry weight but are not themselves established.
The panel records the gap that would most move the claim, and the study that would close it.
Reconcile CANVAS versus CREDENCE divergence. The verdict turns on whether the CANVAS signal is real or chance, and the pooled analysis could not explain the I2=82% heterogeneity [7]. The panel rates this the highest-importance open gap for the claim, and names the study design that would close it: A prospectively pooled or individual-participant meta-analysis across all canagliflozin RCTs with prespecified, adjudicated amputation endpoints and tests of effect modification.
Open question: Is the CANVAS amputation excess a true canagliflozin effect or a chance finding? Why it is still open: CREDENCE showed no effect [5] and the pooled analysis reported I2=82% heterogeneity unexplained by trial or patient differences, raising chance [7].
Priors is a sceptical first reader, not peer review, and not medical advice.
Release 2026-q4 · graded 23 Oct 2026 · protocol v0.19