International Consensus on DKA risk management with SGLTs in type 1 published in Diabetes Care – February 11, 2019

Executive Highlights

  • The International Consensus on DKA risk management with SGLT inhibitors in type 1 diabetes was just published in Diabetes Care, housing guidelines developed by 20+ thought leaders following a June 2018 ATTD meeting. While these guidelines reflect growing consensus around DKA risk mitigation and treatment strategies – the publication has already been frequently cited as evidence of such –  real-world effectiveness of these measures still must be measured and documented.

  • The authors present an educational framework on which future risk mitigation protocols might be based. The paper emphasizes patient and clinician education, particularly with regards to ketone monitoring, patient selection, and treatment strategies (such as STICH). Importantly, the consensus does not endorse any one particular ketone testing regimen, as little evidence for any particular method exists – instead, testing frequency should be commensurate with the patient’s lifestyle and risk factors. However, authors implored researchers to delineate the effectiveness of ketone monitoring in preventing DKA.

  • Blood and urine ketone levels were stratified into ketonemia, impending DKA, and probable DKA, alongside recommended patient/clinical treatment response. As a reminder, the point at which patients and providers should become concerned about DKA was heavily debated at the original ATTD discussion, underscoring the need for personalized DKA risk management. Data of course should enable even stronger consensus in the future.

  • Authors arranged 16 DKA risk factors by importance, many of which are not inherent patient characteristics but rather acute events that can increase DKA risk (such as reduced basal insulin or pump failure). Notably, low BMI is listed as a minimal/low risk factor for DKA in the consensus statement, though EMA’s CHMP decided on a BMI cutoff (≥27 kg/m2) as its primary selection criteria. To be fair, it’s entirely possible that sub-analysis of pooled data from the DEPICT program may have uncovered a reduced DKA signal in the BMI ≥27 kg/m2 population. We see this as a very practical approach as well as an appropriately conservative one.

The “International Consensus on Risk Management of Diabetic Ketoacidosis in Patients with Type 1 Diabetes Treated with Sodium-Glucose Cotransporter (SGLT) Inhibitors” was just published in Diabetes Care, offering the first formal guidelines on the topic. The Consensus was developed by 20+ thought leaders following an ATTD-sponsored meeting in June 2018. To briefly summarize, the paper includes the following sections:

  • Background on putative mechanism of DKA in regard to SGLT use in type 1 patients, along with DKA diagnosis;

  • Approaches to DKA prevention, including patient selection, insulin dose adjustments, initiation and dosing of SGLT inhibitors, ketone monitoring, and clinician/patient selection; and

  • Research questions in the area that still must be addressed.

Notably, Lexicon – in preparation for its FDA Advisory committee on SGLT-1/2 dual inhibitor sotagliflozin in type 1 adults – has frequently cited the publication of these guidelines in support of coalescing agreement amongst the academic community on how to best mitigate DKA risk with this adjunct therapy. Nevertheless, the real-world effectiveness of these measures still needs to be documented. There is a real and substantial risk that bringing SGLTs officially to market for type 1s would scale DKA risk to a larger population than ever before. On the other hand, there is opportunity with a major educational campaign to reduce DKA risk within type 1 diabetes broadly speaking. T1D Exchange data shows us that ~3% of participants have experienced DKA within the previous three months. To this end, we hope this consensus statement encourages discussion on DKA risk mitigation strategies, regardless of SGLT-inhibitor use.

For context, most recently in the SGLT for type 1 landscape, EMA’s CHMP recommended approval of AZ’s Forxiga (dapagliflozin) for specialist-treated type 1s with BMI ≥27 kg/m2. This opinion energized further the movement to bring SGLT inhibitors to people with type 1 diabetes, following a complex FDA Ad Comm meeting on Sanofi/Lexicon’s sotagliflozin – capped by a split 8-8 committee vote. Depending on sotagliflozin’s fate (FDA’s PDUFA date is March 22) and EMA’s final rule, either drug could be the first-to-market oral adjunct therapy for type 1 diabetes in the US or Europe.

DKA Diagnosis

  • Ketonemia, impending DKA, and probable DKA were stratified by blood ketone and urine ketone levels alongside patient/clinical response. As a reminder, the point at which patients and providers should become “concerned” was heavily debated at the original ATTD discussion, underscoring the need for personalized DKA risk management.

Cut-points for ketosis/DKA and corresponding remedial actions

Blood ketone (BHB) Level

Urine Ketone*

Remedial Actions

<0.6 mmol/L (Normal)


No Action Needed

0.6 – 1.5 mmol/L (Ketonemia)

Trace or Small (+)

Treat as follows or per clinician instructions:

  • Ingest 15-30 g rapidly absorbed carbohydrate and maintain fluid consumption (300-500 mL) hourly

  • Administer rapid-acting insulin based on carbohydrate intake (hourly)

  • Check blood/urine ketones (every 3-4 hours) until resolution

  • Check blood glucose frequently to avoid hyperglycemia and hypoglycemia

Seek medical attention if levels persist and symptoms present.

1.6 – 3.0 mmol/L (Impending DKA)

Moderate (++)

Follow treatment recommendations listed above.

Consider seeking immediate medical attention.

>3.0 mmol/L (Probable DKA)

Large to Very Large (+++/++++)

Seek immediate medical attention.

*Urine ketone concentrations are dependent on hydration and other factors; these values do not closely correlate with blood BHB levels.

DKA Risk Factors and Patient Selection

  • Authors arranged 16 risk factors for DKA by severity (see table below), many of which are not inherent patient characteristics but rather acute events that can increase DKA risk (such as reducing basal insulin by too much or experiencing a pump failure). Notably, low BMI is listed as a minimal/low risk factor for DKA in the consensus statement, while the EMA’s CHMP decided on a BMI cutoff as its primary patient eligibility criterion. To be sure, it’s entirely possible that sub-analysis of pooled data from the DEPICT program may have uncovered a reduced DKA signal in the BMI ≥27 kg/m2 population.

Risk factors for DKA associated with SGLT-inhibitor therapy

Risk level for DKA



  • Reduced basal insulin by more than 10 to 20%

  • Insulin pump or infusion site failure

  • Reduced or inconsistent carbohydrate intake

  • Excessive alcohol use

  • Use of illicit drugs

  • Volume depletion/dehydration

  • Acute illness of any sort (viral or bacterial)

  • Vomiting


  • Vigorous or prolonged exercise

  • Reduced prandial insulin dose by more than 10 to 20%

  • Travel with disruption in usual schedule/insulin regimen

  • Insulin pump use


  • Low BMI (<25 kg/m2)

  • Inconsistent caloric intake

  • Moderate alcohol use*

  • Female sex

*If ketone levels increase from baseline

  • Patient criteria for SGLT-inhibitor therapy in type 1 diabetes was agreed as:

    • >18 years of age

    • Adherent to prescribed diabetes regimen

    • Willing/able to perform all prescribed diabetes self-management tasks

    • Performs blood glucose monitoring or uses CGM as prescribed

    • Willing/able to perform ketone testing as prescribed

    • Has received education/training in ketone testing and interpreting/acting upon test results

    • Has access to ketone testing materials

    • Has immediate access to a clinician if blood or urine ketone levels are elevated

    • No or moderate use of alcohol; no use of illicit drugs

    • Unimpaired cognition

    • Not pregnant or wanting to become pregnant

DKA Prevention Protocol

  • In patients with an A1c <7.5%, a 10-20% reduction in insulin dose alongside CGM or frequent BGM and HCP input is recommended; for those with an A1c ≥7.5% only slight reductions in prandial and basal insulin may be needed. Insulin dose adjustments must be individualized to the patient and specific SGLT-inhibitor, as dose reductions observed in clinical trials for SGLT-1/2 dual inhibitor sotagliflozin were largely in mealtime insulin, while reductions in both basal and bolus insulins were made with SGLT-2 inhibitors. Blood ketone levels should be below 0.6 mmol/L prior to initiating an SGLT-inhibitor, and it is recommended that patients begin with the lowest dose available (or even split tablets). As seen in the phase 3 EASE program for Lilly/BI’s empagliflozin, a lower dose (2.5 mg) may confer significant glycemic benefits without an increased risk of DKA.

  • The consensus does not endorse any one particular ketone testing regimen, as little evidence for any exists and testing frequency should be commensurate with the patient’s lifestyle and risk factors (see table above). For all patients, ketones should be measured in the case of DKA symptoms: malaise, fatigue, nausea, vomiting, diet change, infection, dehydration, surgery, injury, pump failure, or stress. However, this invokes many of the questions surrounding the utility of ketone monitoring brought up during the sotagliflozin advisory committee. Specifically, what happens when someone uneducated in the finer points of diabetes management takes an SGLT inhibitor and:

    • Does not have a meter and has to use urine ketone strips (~20% of registry members in the newest T1D Exchange database) – there still seems to be some debate over to what degree (if at all) this is a problem, particularly among certain patients;

    • Has a meter but doesn’t know how to use it – i.e., how to assess what is a baseline level and what are inappropriate levels;

    • Doesn’t have a blood ketone monitor and doesn’t know whether and how to use urine ketone strips (or whether they are “better” for his or her cohort);

    • Has a meter and has run out of strips and has no urine ketone strips (this seems fixable);

    • Has strips but not multiple strips for when approaching DKA and needs to test frequently (within a matter of hours);

    • Doesn’t understand the protocol of what to do when approaching or in DKA and blood glucose is normal (“take insulin” may be counter-intuitive to some, though followed by eating carbs should be reassuring – though if glucose is low, this may particularly hard to follow);

    • Can’t afford additional strips in months when more are used (we recently saw a receipt for 60 strips that cost nearly $600 – some months far more strips than others may be needed and for some low-risk patients, virtually no strips may be needed, and this may be hard to determine).

  • Moreover, an increased risk of DKA was witnessed in the inTandem program for sotagliflozin despite all participants being given a ketone monitor for the trial – an ideal unlikely to occur unless manufacturers or medical device companies partnered to provide ketone monitoring equipment with SGLT prescription. In the case that one cannot afford a blood ketone meter, authors suggested use of urine strips which, while effective when used properly, present a whole host of challenges in their own right, including: (i) Urine strips test an average of ketone concentration since the last bathroom use; (ii) urine output is frequently low in patients with DKA due to dehydration, potentially delaying treatment; and (iii) urine strips measure acetoacetate, not beta-hydroxybutyrate (BHB), which can rise even if BHB levels are dropping. Altogether, many questions remain on the effectiveness and generalizability of ketone monitoring suggestions, though we are ultimately pleased to see at least some progress and guidance on this front.

  • Given the very real risk of DKA, any approval would likely come with a strict mitigation strategy incorporating educational and patient selection components; to this end, authors provided an educational framework on which such a protocol might be based (see below). For context, Sanofi/Lexicon presented a valuable start to risk mitigation – but only a start. For one, there were questions of whether risk mitigation protocols actually “work” – an incredibly valid concern that should be backed by evidence before exporting any strategies to patients, as well as for more meaningful patient and provider education in addition to written materials and letters. Yet collecting information about what is “not” working would also be hard to get, given the design itself of such a trial may be fairly complicated. That said, Sanofi/Lexicon’s plan touched on many of the below – patient education through an in-product card (hopefully incorporating the STICH protocol), medication guide, and website (we’d love a video) and PCP education through product labelling and a letter sent within 60 days of approval (target of over one million HCPs). Glaringly missing was any mention of patient selection (unlike the CHMP recommendation of Forxiga approval for BMI>27 kg/m2), and beyond surface-depth details on ketone monitoring education (e.g., blood vs. urine, baseline values, etc.). With little data to suggest that DKA risk wouldn’t increase outside of the highly-monitored, RCT environment, FDA was also quick to point out (a little bit wryly) that certain trial components, such as complementary ketone monitors, were not included in the sponsor’s proposal.

Educational components of a risk mitigation strategy when introducing SGLT-inhibitors for type 1 diabetes

Patient Education

All patients initiating SGLT-inhibitor therapy should receive thorough training/education in the following areas:

  • DKA causes and symptoms

  • Euglycemic ketoacidosis

  • Importance of ketone monitoring

  • Use of ketone monitoring – training in testing procedure, proactive monitoring, situations when monitoring is indicated

  • Treatment protocol for addressing ketosis

  • Guidance in when to seek medical attention

Clinician Education

All prescribing clinicians should acquire full understanding of the safe use and risks associated with SGLT-inhibitor therapy:

  • Criteria for patient selection – baseline ketone level, demographic/behavioral considerations

  • Training/educational needs of patients – detection (ketone levels, symptoms), prevention strategies, treatment

  • Potential for missed DKA, euglycemic DKA

  • Treatment strategies – STICH protocol recommended (see below)

Risk Communication

  • Product labelling, website

  • Healthcare professional education

  • Medication guide, patient alert card (STICH)

Phase 3 Program Adjudication Differences

  • Authors made a point of recognizing the substantial differences in DKA adjudication between phase 3 SGLT-inhibitor programs conducted thus far, recommending a standardized approach for future trials (see below). On a grander scale, these disparities exemplify the many questions that remain surrounding DKA risk, monitoring, and prevention, and we see an agreed upon, consistent process for future trials as important progress.

Differences in DKA Adjudication Definition

Tandem Program (sotagliflozin)

DEPICT Program (dapagliflozin)

EASE Program (empagliflozin)

Adjudicated DKA characterized as:

Adjudicated DKA characterized as:

Adjudicated DKA characterized as:

Yes, with certainty



Yes, probably



No, unlikely


Ketosis (not DKA)

No, with certainty

Only “definite” was adjudicated positively as a DKA event




Insufficient data

“Yes, with certainty” and “yes, probably” were both adjudicated positively as a DKA event

Close Concerns Questions

  • How will efficacy and risk change in real-world settings compared with RCTs?

  • How can we develop a better understanding of the dose-dependent effect on both efficacy and DKA risk with SGLT-inhibitors?

  • Should those with A1c levels >10% receive an SGLT-inhibitor?

  • What is the ideal insulin reduction algorithm, and how much should it vary from person-to-person, product-to-product, and dose-to-dose? Should insulin be reduced at all?

  • How frequently should users measure ketones? To what extent can this be safely “personalized”? Does ketone monitoring really decrease population-level risk?

  • If ketone levels are consistently low, would patients be less likely to adhere to strict monitoring regimens?

  • How can patient and clinician education programs and materials be evaluated before or after approval? How can patient and clinician knowledge be evaluated?

  • What can be done to make ketone testing supplies more accessible and affordable?


--by Peter Rentzepis, Martin Kurian, Ann Carracher, and Kelly Close