Diets · Metabolic & Cardiometabolic
ketogenic diet with SGLT2 inhibitor causes euglycemic ketoacidosis
In plain terms: Can keto plus a diabetes drug cause dangerous ketoacidosis?
Strong support Diets 💰 Industry COI noted
RefutedContestedStrong support
consensus score 1.00
Yes, combining a very low-carb diet with an SGLT2-inhibitor drug is a genuine, well-documented risk even when blood sugar looks normal.
📅 Last reviewed: 2026-07-15 ⓘ
Evidence ladder
How far up the ladder this claim has climbed. A high consensus on a low rung means "consistent so far," not "proven in people."
Top evidence so far: Population patterns (Observational)
MechanismIn-vitroAnimalObservationalRCTMeta-analysis
How the studies fall
9 support 0 contradict 0 tested null 0 mixed · 9 sources, 9 independent groups
What the evidence shows
Combining a ketogenic diet with an SGLT2-inhibitor drug raises the risk of euglycemic diabetic ketoacidosis - a genuine, dangerous interaction to flag for any diabetic family member.
The evidence (18)
| Source | Grade | Stance | Quality | Finding |
|---|---|---|---|---|
| Mistry S, Eschler DC 2021 · AACE Clin Case Rep. 2021;7(1):17-19 | observational | supports | low | Two-patient case series (with an appended narrative review), AACE Clin Case Rep 2021: 'We describe 2 patients with T2DM who developed euDKA caused by SGLT2 inhibitor use while on a ketogenic diet and provide a review of the literature.' Both carry the exact co-exposure this claim names. Case 1, 47F, T2DM dx 2 months earlier, on a low-carbohydrate diet for 2 months, presented the day after 'her first and only dose of empagliflozin': 'plasma glucose level was 187 mg/dL, bicarbonate level was 11 mmol/L, anion gap level was 22 mmol/L, β-hydroxybutyrate level was 6.78 mmol/L', pH 7.24, 'hemoglobin (Hb) A1C level was 13.6%'; 5 days on an insulin drip. Case 2, 34M, canagliflozin 100 mg daily for ~60 days, 'had started a ketogenic diet a week before presentation': 'serum glucose level of 251 mg/dL, bicarbonate level of 12 mmol/L, anion gap level of 24 mmol/L, β-hydroxybutyrate level of 5 mmol/L, HbA1C level of 8.2%', pH 7.27; 3 days on the drip. The authors' recommendation is unambiguous: 'Patients taking SGLT2 inhibitors should be advised to avoid a ketogenic diet.' Three reasons this stays LOW quality despite full-text reading: n=2 with no comparator and no denominator; the authors concede the diabetes type was never established - 'glutamic acid decarboxylase antibodies and islet antigen-2 antibodies were not assayed in either case. It is possible that these patients could have had a missed diagnosis of T1DM or latent autoimmune diabetes' - which matters most for Case 1 at HbA1c 13.6%, where insulinopenia alone could explain the event after a single drug dose; and Case 2 fails the euglycemic definition the paper itself cites, 'the glucose level was 251 mg/dL, which was near the cutoff level for a diagnosis of euDKA'. The literature-review half tabulates 5 prior cases (SGLT2i exposure 1-365 days, low-carbohydrate diet 7-3960 days) and re-states Earle 2020, already counted separately as s32426668. No funding declared; 'The authors have no multiplicity of interest to disclose.' |
| Koutentakis M et al 2023 · Journal of Cardiovascular Development and Disease | observational | supports | low | Narrative review (no search strategy, no pooled estimate) with a dedicated section on combining SGLT-2 inhibitors with a ketogenic diet: 'one major concern is the increased risk of euglycemic diabetic ketoacidosis (euDKA, DKA), a potentially life-threatening emergency'. Combination-specific support is an explicit inference — 'Because euglycemic ketosis can be brought on by either a ketogenic diet or SGLT-2 inhibitors alone, there is probably a greater possibility that it will occur when both therapies are combined' — plus three cited case-report patients (a 73-year-old woman with perioperative euDKA on an SGLT2i and a strict ketogenic diet; 'two patients in which euDKA occurred after just one dose of an SGLT2 inhibitor while on a ketogenic diet'). Supports the claim; quality low because no measurement of combined exposure and the vote is borrowed from refs 96/101. |
| Guirguis H, Beroukhim Afrahimi S, Pham C 2022 · Clin Med Insights Case Rep | observational | supports | low | Single case report, n=1, no comparator. 58-year-old man with type 2 DM on metformin, empagliflozin, sitagliptin and repaglinide 'endorsed having started a ketogenic diet 1 month ago along with intermittent fasting for weight loss and had stopped metformin and repaglinide but continued with the rest of his medications.' He presented obtunded and hypotensive (MAP 40-50 mmHg, temp 34.4 C) in severe ketoacidosis: 'glucose of 373 mg/dL... anion gap of 22 mEq/L... beta hydroxybutyrate level of 9.8 mmol/L (Ref. range 0.0-0.3 mmol/L), with profound acidemia on venous blood gas with pH 6.9'. HbA1c was 6.7%. Infectious workup, ECG, chest radiograph, abdominal CT and brain imaging were all unrevealing, and the authors concluded by exclusion that 'SGLT-2 inhibitor use combined with his ketogenic diet and intermittent fasting was the most likely etiology for his DKA.' Three limits to how far this vote can carry. (1) The presentation was HYPERGLYCEMIC, not euglycemic - glucose 373 mg/dL - so it evidences the keto+SGLT-2i ketoacidosis interaction but not the 'euglycemic' qualifier in the claim's object; what it does show is that severe acidosis is possible with well-controlled diabetes ('a patient does not have to have severely uncontrolled diabetes to present in DKA and profound acidosis in such a setting'). (2) The keto exposure is fully confounded with concurrent intermittent fasting and with self-discontinuation of metformin and repaglinide. (3) Attribution is diagnosis by elimination in a single patient, and the paper's own cited pharmacovigilance context puts SGLT-2i DKA incidence in type 2 DM at 'less than 0.2%' with a median onset 2 weeks after drug initiation - this patient's drug start date is never given, so the diet, not the drug, is the dated exposure. No funding; authors declare no conflicts. |
| Needham N et al 2023 · medRxiv | observational | supports | low | Single-arm open-label ketogenic-diet feasibility pilot in euthymic bipolar disorder (27 recruited, 26 commenced, 20 completed, 6-8 weeks; ISRCTN61613198): 'One serious adverse event was reported (euglycemic ketoacidosis in a participant taking SGLT2 -inhibitor medication).' Case-grade signal - no comparator, no reported denominator of SGLT2i-taking participants, no causality adjudication in the available text; medRxiv preprint, abstract only. |
| Peters 2020 · Diabetes Care | RCT | supports | high | OFF-SCOPE for the diet x drug combination, though a strong study of the drug half. Pooled prespecified safety analysis of inTandem1 (NCT02384941) + inTandem2 (NCT02421510): two identically designed 52-week double-blind placebo-controlled phase 3 trials, 1,575 adults with type 1 diabetes randomized 1:1:1 to placebo / sotagliflozin 200 mg / 400 mg adjunctive to insulin, all ketosis events centrally adjudicated. DKA rose clearly on drug: 'Of these, 37 events (36 patients) were adjudicated as DKA, with an exposure-adjusted incidence rate of 0.2, 3.1, and 4.2 events per 100 patient-years for placebo, sotagliflozin 200 mg, and sotagliflozin 400 mg, respectively' (EAIR difference vs placebo 2.9 [95% CI 1.3-4.5] and 4.0 [2.1-5.9]; 43% of events severe by pH<7.0 or bicarbonate<10; 34/37 hospitalized; no deaths). Ketone distribution shifted too: BHB >0.6 mmol/L in 19.8% / 46.9% / 45.5% and >1.5 mmol/L in 2.1% / 6.3% / 8.2%. But DIET WAS NEVER MEASURED, and the authors say the interaction could not be tested: 'because data regarding patients’ dietary carbohydrate intake were not collected prospectively, it was not possible to evaluate whether the combination of a low-carbohydrate diet and SGLT inhibitor use affected DKA risk.' They further note the program 'did not include ... consensus recommendations on avoiding low-carbohydrate diets.' Identified triggers were insulin reduction/pump failure (19) and intercurrent illness (14), not diet. The euglycemic object is only loosely matched as well: 'one-third of DKA events were associated with blood glucose <250 mg/dL' but 'No event was associated with a blood glucose value <150 mg/dL.' Subject also mismatches on class - sotagliflozin is 'a dual inhibitor of SGLT1 and SGLT2'. Sponsored and conducted by Lexicon Pharmaceuticals with five employee co-authors; the harm estimate runs against sponsor interest and is credible, but the abstract's reassurance that the risk 'can be managed with patient education' rests on a non-randomized calendar before/after whose numbers appear only in a supplement absent from the cached text. |
| Juneja D, Nasa P, Jain R, Singh O 2023 · World J Diabetes. 2023;14(8):1314-1322 | observational | supports | low | Meta-summary pooling 108 case reports + 17 series (125 studies, 169 unique SGLT2i-EDKA patients); contributes NO new patients - 'the included studies were only case reports and case series which had no control arm' and 'had a high risk of bias and missing data', so it is a review-layer pooling of the same case literature the vault already holds as primaries. On the claim's own subject arm the paper is supportive: Table 1 lists 'Reduced carbohydrate/ketogenic diet 14 (8.3)' as an identified trigger among the 78.7% with a trigger (commonest infection 37.9%, perioperative 24.3%), and the discussion states 'Other common factors which may predispose patients to develop EDKA include prolonged fasting, low carbohydrate or ketogenic diet, excessive alcohol intake, dehydration and reduction in insulin dosage', with the mechanism that 'High protein and low carbohydrate diets may increase serum glucagon and reduce serum insulin levels.' Median presenting glucose 184.5 mg/dL, median pH 7.14, mortality 2.4%. Corrects the previous extract, which wrongly stated that diet was not reported as a trigger category. |
| Kamath SD, Kumar U, Shrivastava V 2024 · Cureus | observational | supports | low | Single-patient case report (n=1, no comparator, no denominator), Cureus 2024. Carries the DRUG half of this claim's subject and its exact euglycemic outcome, but NOT the diet half. 'We report a case of a 58-year-old lady with a history of T2DM on multiple anti-diabetes medications including dapagliflozin for one year, who during intercurrent illness developed EDKA.' Presentation: 'A 58-year-old lady was admitted for multiple episodes of vomiting, abdominal pain with distension of three days duration associated with generalized weakness, and fever on and off for four days.' Labs confirm the euglycemic signature - 'Her blood sugar on admission was 203 mg/dL, and arterial blood gas showed high anion-gap metabolic acidosis (HAGMA) with ketonuria and ketonemia (blood beta-hydroxybutyric (BOHB) acid level: 5.4 mmol/L).', 'a pH of 7.142', 'Thus, ABG showed HAGMA with a base deficit of 16.6 mmol/L.', urine ketones 4+, HbA1c 9.5%. Attribution is the authors' clinical inference: 'As her blood sugar on presentation was 203 mg/dL, diagnosis of EDKA induced by dapagliflozin and precipitated by an infection was made.' Resolution took 'about 90 hours', discharged day 6. WHY OFF-SCOPE: no ketogenic diet exists anywhere in this case - the token 'ketogenic' occurs once in 22,581 characters of full text, in a Discussion mechanism list, and the patient's carbohydrate deficit was involuntary: 'Fever and decreased carbohydrate intake due to vomiting might have precipitated EDKA in our patient.' The abstract's 'Low carbohydrate intake, dehydration resulting from repeated vomiting, and skipping the previous two days’ dose of insulin could have precipitated this condition.' names starvation, dehydration and insulinopenia, each independently sufficient to cause ketoacidosis, so the case cannot isolate any dietary contribution. Two further limits: sepsis was treated concurrently ('Broad spectrum antibiotic piperacillin-tazobactum 4.5 g intravenously was started in view of sepsis (as demonstrated by high procalcitonin and fever), though we could not demonstrate the exact cause of sepsis.'), and the Discussion's supporting numbers (canagliflozin HR 3.58, incidence 0.16-0.76 events per 1000 patient-years, median onset two weeks) are citations to Goldenberg and Wang, not this paper's data. One genuinely novel datum: 'However, our case was unique in the sense that the patient was on dapagliflozin for up to one year.' - a counter-example to the early-onset window the paper itself cites. Grade left at observational only because the vocabulary has no case-report tier. 'The authors have declared that no competing interests exist.' Full PMC XML read 2026-08-19 (PMC11095996). |
| Garg R et al 2025 · J Gen Intern Med | observational # was mechanism — review of human case reports, nothing mechanistic measured (reread #355, precedent k) | mixed | moderate | Narrative review (J Gen Intern Med; PubMed pub-type Review) - no new data. Literature search 'identified five articles with six cases of euglycemic ketoacidosis in non-diabetic patients who were on SGLT-2 inhibitors for heart failure with reduced ejection fraction.' Mechanism as stated: 'a general state of starvation or relative insulin deficiency, which exacerbates the mild baseline ketonemia caused by this class of medications while normoglycemia is maintained.' A ketogenic DIET was an exposure in none of the six cases: 'The common risk factor in five out of the six cases was decreased oral intake due to acute illness, fasting, or a perioperative state.' Supports the shared mechanism, does not test this claim's diet arm; zero-weighted per CONVENTIONS section 5 (a review is not evidence). |
| Somagutta 2022 2022 · Korean J Fam Med | observational | supports | low | Systematic review of published case reports — it contributes no patients of its own, so it is zero-weighted as evidence_role: review (precedent (u)). Direction as asserted is supportive and squarely on-claim: 'their use in people with diabetes taking SGLT2i is associated with EDKA', with the proposed mechanism that 'Ketogenic diets, VLCD, and SGLT2i increase glucagon secretion while limiting serum glucose levels, thereby synergistically amplifying the risk of EDKA.' The pool: 'Thirteen case reports identifying 14 patients on a ketogenic diet and SGLT2i diagnosed with EDKA were thoroughly reviewed' (12 T2DM, 2 T1DM; 1,089 records retrieved, 762 screened; inclusion limited to 'case reports published in the last 5 years between 2016 and 2021' in English). Presentation was the euglycemic signature — 'At presentation, average blood glucose was 167.50±41.80 mg/dL, pH 7.10±0.10', mean HCO3 8.1 mmol/L, mean anion gap 23.6 mmol/L, mean HbA1c 10% — with diet exposure of 1-90 days and over 90% hospitalised within 4 weeks of starting the diet. DOUBLE-COUNT: four of the fourteen pooled patients are already held as independent sources on this claim, matched by title and author in the reference list — ref [5] Earle/Ault/Bonney = s32426668 (Table 1: 31F, canagliflozin, 2 wk), ref [1] Mistry & Eschler = s33851013 supplying TWO patients (47F empagliflozin 25 mg; M canagliflozin 100 mg), and ref [6] Latif/Gastelum/Sood/Reddy, the Drug Ther Bull reprint (doi 10.1136/dtb.2021.235117rep) of the BMJ Case Rep article held as s32513767 (Table 1: 43F, empagliflozin 25 mg, pH 7.01). Steinmetz-Wood (s32855828), Kapila (s34527480) and Guirguis (s35418794) are NOT in the pool. GRADE GUARD: the Discussion's phrase 'This systematic review and meta-analysis included 13 case reports' is a mislabel — no pooled estimate, CI, or heterogeneity statistic exists anywhere, and 'All studies were case reports, as no observational studies or clinical trials were conducted on this novel topic.' Quality low is earned: three of the thirteen included reports are conference abstracts only (Leys 2019, Richstein 2020, Castellanos-Diaz 2020), there is no protocol registration despite the PRISMA framing, and the authors concede 'The study was limited by the risk of bias assessment to determine the methodological quality of the evidence from the case reports' and that 'high-quality studies, such as observational studies and randomized clinical trials, need to investigate this topic in depth.' No denominator, comparator or unexposed arm, so the design cannot carry the claim's causal predicate by itself. 'No potential conflict of interest relevant to this article was reported.' Full PMC XML read 2026-08-19 (reread #349). |
| Roberts 2025 · Diabetes Obes Metab | observational | supports | moderate | Retrospective real-world cohort carrying its own data - NOT a review; the previous one-line extract was an abstract-derived stub (it paraphrased the abstract's Aims sentence about risks being 'not well-documented' and reported none of the paper's findings) and is replaced. This is the only source on this claim with a denominator, an internal comparator and formally adjudicated events. Virta Health telemedicine VLCKD clinic, de-identified records 2015-2023: 'In 72 751 patient‐years of follow‐up, there were 148 unique flagged health events involving higher ketones, giving an incidence rate of 2.03 per 1000 person‐year.' Of those, 86 met the illness-associated-with-ketosis definition (BHB >=3 mmol/L plus seeking in-person care) and 51 were adjudicated DKA against the 2024 criteria; there were no deaths. ON THE CLAIM'S EXACT CO-EXPOSURE ARM - the whole cohort is ketogenic, so the SGLT2-inhibitor contrast is internal with diet held constant: 'The incidence rates of DKA in PWD per 1000 person‐years is 1.01, 2.90 among those ever on an SGLT2‐inhibitor, and 0.37 among those never on an SGLT2‐inhibitor. The use of SGLT2‐inhibitors in PWD increased the risk of IAK by 4.24‐fold ... 2.11‐fold for non‐DKA IAK (not significant) and 7.18‐fold for DKA' (p < 0.001). Absolute harm: 'When an SGLT2‐inhibitor is added to a VLCKD, the NNH for DKA is 260, and when an SGLT2‐inhibitor is added or used before transitioning to VLCKD, it is 223.' THE EUGLYCEMIC OBJECT is separately evidenced: 'Of the 51 people with DKA, 26 (54%) had blood glucose ≥200 mg/dL, and 22 (46%) were normoglycemic.' against the authors' own benchmark - 'The incidence of glucose concentration ≤ 200 mg/dL among patients with adjudicated DKA in this population is higher than the reported incidence of about 10%.' - which they attribute to the diet itself: 'A possible explanation is that the very low carbohydrate nutrition plan prevents higher blood glucose while not preventing ketoacidosis.' COUNTERWEIGHT, stated as loudly as the support: the authors' conclusion runs the other way on absolute risk - 'Very low carbohydrate ketogenic diets are generally safe with low rates of IAK, including DKA, in people with and without type 2 diabetes.' - and they argue the drug alone may account for the whole effect: 'The 2.90 per 1000 person‐year incidence rate of DKA in PWD on VLCKD and SGLT2‐inhibitors in this analysis is comparable to published incidence rates of 0.5 to 4.9 per 1000 person‐years in real world data analyses of SGLT2‐inhibitor exposure in PWD not specifically following VLCKD.' That is an unmatched external/historical comparison, not an internal test of whether the diet adds anything on top of the drug, so it qualifies the vote rather than reversing it; the diet-adds-to-drug question is logged separately as a tested-null candidate. WHY QUALITY STAYS MODERATE AND IS NOT RAISED despite 72,751 patient-years: (1) retrospective and self-selected - 'the potential for bias introduced by the self‐selection of individuals with unique characteristics into this dietary pattern.'; (2) the clinic actively suppresses the very exposure under study - 'the safety protocol is to stop SGLT2‐inhibitors a few days before diet changes in people without cardiorenal indications.' - alongside automated alerts at BHB >=3 mmol/L and hold-the-drug guidance, so these rates describe a monitored, deprescribed setting and understate unsupervised risk, which the authors concede ('The impact of routine BHB monitoring ... on IAK and DKA incidence rates is unknown'); (3) laboratory data were missing for 20-40% of events and under-detection is conceded. Note the euglycemic/hyperglycemic split is NOT stratified by SGLT2-inhibitor use, so that half of the support sits at the diet level rather than the combination level. Conflicts run AGAINST the reported harm, which strengthens the finding while discounting the reassuring framing: 'C.G.P.R. and S.J.A. are employees of Virta Health and have been offered stock options.' 'No funding was received for this work.' Full PMC JATS XML read 2026-08-19 (PMC11964994). |
| Earle, Ault, Bonney 2020 · Clin Pract Cases Emerg Med | observational | supports | low | Single-patient case report (n=1, no comparator, no denominator), 31-year-old woman with T2DM and no prior DKA, on the exact co-exposure this claim names: 'The patient had been attempting to control her T2DM by following a very low-carbohydrate, ketogenic diet for about two weeks, during which she had restricted her carbohydrate intake to 10-15 grams per day' while 'her T2DM management regime consisted solely of canagliflozin and dietary modifications' (insulin detemir stopped when the diet started). She presented in severe euglycemic ketoacidosis: 'a pH on a venous blood gas of 7.056 with a bicarbonate of 8.0 milliequivalents per liter (mEq/L), blood glucose of 139 mg/dL, blood ketones of 80 mg/dL, lactate of 1.4 millimole per liter (mmol/L), and an anion gap of 29'; ICU admission, recovered on insulin infusion with glucose-containing fluids. The authors claim no more than plausibility - 'a synergistic effect increasing the risk for EDKA is plausible' and 'This association will require further study for validation'. No industry funding; 'The authors disclosed none.' |
| Latif 2020 · BMJ Case Rep | observational | supports | low | euDKA in T2DM after 2wk keto on empagliflozin; glucose only 169 mg/dL |
| Woronow 2024 · Endocr Pract | observational | supports | low | FDA FAERS case series, 29 uncontrolled spontaneous reports of prolonged/relapsed ketoacidosis in adults with T2DM on SGLT2 inhibitors: 'Associated factors for prolonged or relapsed KA included surgery, decreased caloric intake, and ketogenic/carbohydrate restricted diet.' No denominator, no comparator, no count of how many cases involved the diet; 62% were on 3+ glycemic medications. Severity real (median pH 7.1; 4 patients at pH <=7.0; KA treated 3-20 days). Names the exposure but cannot measure risk, and never documents euglycemia. |
| Dynka D, et al. 2026 · Ann Med | observational | supports | low | Narrative review (no new data) asserting that 'SGLT2 inhibitors also carry a risk for diabetic ketoacidosis when combined with the KD.' Its actual evidence is hedged and secondhand: 'Thirdly, some reports point to a possibly greater risk of euglycemic diabetic ketoacidosis (euDKA) when combining KD with SGLT-2 inhibitors', supported by two cited case reports - 'One patient following a carbohydrate-restricted diet developed euDKA after just the first dose of empagliflozin' - plus an American Society of Clinical Endocrinology recommendation to stop these drugs before starting a KD. Funded by the Rodzen Brothers Foundation, whose two co-founders are co-authors. |
| Steinmetz-Wood, Gilbert, Menson 2020 · Case Rep Endocrinol | observational | supports | low | n=1 case report. T2DM man, 47, on empagliflozin + self-reported Atkins/ketogenic diet. The EUGLYCEMIC episode is the first one: 'an anion gap of 21, a bicarbonate level of 13 mmol/L, a pH level of 7.22, and 3+ urinary ketones with a glucose level of 7 mmol/L (127 mg/dl)' - it resolved on a carbohydrate-increase instruction alone while empagliflozin continued ('His repeat labs one week later did result in closure of his anion gap to 10'). He relapsed 24 days later with severe but HYPERGLYCEMIC DKA (pH 6.94, BHB 8.9 mmol/L, glucose 269 mg/dl) amid pharyngitis, aspiration pneumonitis, recurrent UTI and vomiting - so the second episode is confounded and only the first carries the euglycemic object. Drug was missed on the ED med list both times. Authors: 'physicians should ask whether patients are following low-carbohydrate diets as this likely puts them at higher risk for DKA.' Same paper quotes trial incidence of 0.13-0.76 DKA events per 1000 patient years, so the interaction is real but the base rate is small. |
| Patel K et al 2022 · Cureus | observational | supports | low | Non-systematic literature review (Cureus): names the ketogenic diet as a euDKA precipitant in its own abstract - 'Additional factors like the ketogenic diet, illness, surgery, and pregnancy contribute to precipitating these episodes' - and gives the mechanism, 'Stress, illness, surgery, fasting, and ketogenic diets result in increased levels of counterregulatory hormones like cortisol and catecholamines, which ultimately causes an increased glucagon/insulin ratio'. Its only keto-specific datum is one cited case: 'Alhemeiri et al. described a case where euglycemic DKA occurred even after the patient self-stopped the drug within 48 hours of his presentation because the patient had started following a ketogenic diet 72 hours before his presentation'. Quality low: the methods are a keyword search of PubMed/Google Scholar and 'We have included case reports, case series, literature reviews, and systematic reviews from 2009 to 2022' with no risk-of-bias appraisal or pooling, in a venue that states 'Cureus is not responsible for the scientific accuracy or reliability of data or conclusions published herein.' |
| Colacci M, Fralick J, Odutayo A, Fralick M 2022 · Can J Diabetes. 2022 Feb;46(1):10-15.e2 | meta-analysis | supports | high | Systematic review + random-effects meta-analysis (HKSJ), search inception to April 15, 2019: 7 RCTs (42,375 participants) and 5 cohort studies (318,636 participants) in adults with type 2 diabetes. Inclusion contrast, verbatim: 'Study selection consisted of randomized controlled trials and observational studies that quantified the rate of DKA with an SGLT2 inhibitor in comparison to other diabetes medications or placebo.' Result: 'Four randomized trials were included in the meta-analysis and, compared with placebo or comparator medication, SGLT2 inhibitors had a 2.5-fold higher risk of DKA (relative risk [RR], 2.46; 95% confidence interval [CI], 1.16 to 5.21]; I2=0%; p=0.54).' Observational pooled RR 1.74 (1.07-2.83, I2=45%); absolute rates 0.6-2.2 (RCT) and 0.6-4.9 (cohort) events per 1,000 person-years. The exposure is the DRUG ALONE: the terms ketogenic, diet, carbohydrate, low-carb, fasting and euglycemic do not occur once in the readable text, no arm or stratum is defined by diet, and DKA is not disaggregated into euglycemic vs hyperglycemic. OFF-SCOPE for a diet-plus-drug co-exposure claim; on-scope, and strong, for a drug-only 'SGLT2 inhibitors increase DKA risk' claim, which the vault does not yet hold. |
| Kapila, Topf 2021 · Cureus | observational | supports | low | Single-patient case report (n=1, no comparator, no denominator), Cureus 2021. Carries both halves of this claim's subject and its exact outcome: 'This report describes a case of EDKA one day after bariatric surgery in a 51-year-old female with type 2 diabetes mellitus managed with the SGLT2i, canagliflozin. She was following a ketogenic diet for three weeks prior to surgery.' Labs on presentation: 'Laboratory evaluation revealed modest hyperglycemia with serum levels ranging from 150 to 180 mg/dL and lactic acid of 1 mmol/L', with 'low serum bicarbonate of 8 mmol/L with an elevated anion gap of 37 mmol/L', arterial pH 7.21, ketonuria and glucosuria; potassium fell to 2.6 mmol/L and bicarbonate to 4 mmol/L four hours later. The authors conclude in the claim's direction - 'The risk of EDKA from SGLT2i may be increased by a low carbohydrate diet or postoperative status.' Three qualifiers keep this at the floor of the evidence base. First, the drug was NOT concurrent: 'To prepare for surgery, canagliflozin, an SGLT2 inhibitor, was discontinued two days prior to surgery', and the authors compute that only 'between 3.12% and 6.25% remained in the patient's system at the time of surgery' - so what is tested is residual drug plus ketosis. Second, the 'ketogenic diet' of the abstract is, in the case text, 'a diet consisting of three protein shakes per day for three weeks prior to surgery' - a pre-bariatric very-low-calorie liquid regimen, ketosis-inducing by restriction rather than a high-fat ketogenic diet. Third, a third precipitant is present and the authors explicitly decline to attribute: 'It is uncertain if the patient would have experienced EKDA if she did not follow a ketogenic diet or undergo surgery', and with no preoperative labs ('Since we do not have preoperative labs on our patient') even the onset timing is unestablished. Full PMC text read (PMC8432437). Venue caveat: Cureus prints its own disclaimer that it 'is not responsible for the scientific accuracy or reliability of data or conclusions published herein.' 'The authors have declared that no competing interests exist.' Role set to primary because the appraised datum is the authors' own case; the paper's literature-review half separately restates Steinmetz-Wood 2020, already counted as s32855828. |
Disagree, or know a study we missed?
We grade by evidence, not opinions. The way to weigh in is to point us to a study we haven't cited (check the evidence table above first), or to flag a problem with one we have. Every submission is reviewed; if it holds up, the grade updates and shows in Science Changes Its Mind.
Opens a short form. You'll sign in with Google so submissions are tied to a real account — we don't display your identity, and we only accept a link we can verify (PubMed, DOI, ClinicalTrials.gov).
Educational only, not medical advice. Grades and scores reflect published evidence weighted by study design and quality; see the methodology.