1. What does the PEER Diabetes Medication Decision Aid do?

The PEER Diabetes Medication Decision Aid is an interactive decision support tool to assess individualized prognosis and potential impact of medication for patients with type 2 diabetes. It is not intended for use in people under the age of 18, adults with type 1 diabetes, or persons with gestational diabetes. The PEER Diabetes Medication Decision Aid is divided into 4 steps designed to emulate the process of shared decision-making at the point of care: (1) risk prediction, (2) preference elicitation, (3) discussion of options and their pros/cons, and (4) debrief and documentation.

Step 1: The clinician enters patient information into a validated clinical prediction model (called RECODe) to estimate the patient's risk of diabetes-related outcomes. The evidence behind this model is summarized and referenced on this page.

Step 2: The clinician asks the patient to consider what matters most to them in terms of diabetes-related outcomes. This will focus the initial discussion of treatment options in Step 3 by displaying a graph with this outcome by default (this can be changed later).

Step 3: Medication options are selected to manage the risks associated with type 2 diabetes. Each selected option will display the estimated impact on diabetes-related outcomes. Potential side-effects and other relevant considerations are displayed in the dropdown menu. These options have all been evaluated in at least one high-quality randomized controlled trial; this evidence is summarized below. A bar graph displays the absolute risk (%) of diabetes-related outcomes with and without therapy.

Step 4: A comparative risk summary (with and without treatment) is generated based on final medication decisions. A link for future reference can also be generated which will contain the patient’s personalized risks and treatment choices. These personalized risks and choices are also documented onto a displayable EMR note which can be copied into an electronic medical record. 

2. How do I use the PEER Diabetes Medication Decision Aid?

Step 1

  • Input patient data for sex, age, race, current medications, medical history, and clinical information such as blood pressure and labs.
  • This information is automatically entered into the model and instantly updates the risk estimates in Step 3.

Step 2

  • Select an area of focus based on which outcome matters most to the patient. Step 3 will default to displaying this outcome.

Step 3

  • Select from the available drug therapy options. Any drugs that the patient is already taking as selected in Step 1 will be highlighted in green and will not be selectable again in Step 3. Medications with no proven benefit are also not selectable and are greyed out.
  • Clicking the dropdown arrow will display a treatment's possible side-effects and other relevant considerations.
  • The patient's current risk (based on the inputs in Step 1) as well as the patient's risk on selected medications is displayed here. By default, only the area of focus from Step 2 will be displayed, but there is an option to display all outcomes as well.
  • If multiple treatment options are selected, the cumulative relative benefit will be shown based on the assumption that the benefits of the different treatments are additive.

Step 4

  • The medications selected in Steps 1 and 3 are displayed here. From these options, select the intended strategy going forward (e.g. start a new medication, stop previous medication, take more time to consider).
  • The patient's current risk (based on inputs in Step 1) compared to their risk on therapies being either started or continued (based on Steps 3 and 4) is displayed numerically.
  • Click “Link to Save/Share” to generate a link to this personalized assessment of patient risk and treatment choices for future reference.
  • Click “Generate Note for EMR” to create a note detailing the patient’s risk and treatment choices that can be copied into an electronic medical record.

3. Why was the RECODe tool chosen to estimate risk of diabetes-related outcomes in Step 1?

A variety of clinical prediction models (also known as \"risk scores\" or \"risk calculators\") have been developed to predict outcomes in people with type 2 diabetes, each with their own strengths and limitations. To identify the best risk scores to use in our decision aid, we performed a targeted search of existing diabetes guidelines ( and PubMed (to December 2021) for systematic reviews and validation studies of clinical prediction models in patients with type 2 diabetes. Specifically, we considered clinical prediction models based on the following factors:

1. Evaluated diabetes-related outcomes, including death, cardiovascular complications (e.g. myocardial infarction, stroke, heart failure), and microvascular complications (nephropathy, neuropathy, and retinopathy);

2. Incorporated variables that are readily available in clinical practice;

3. Had been externally validated in at least 1 study; and

4. Demonstrated good predictive power based on discrimination and calibration.

RECODe was found to have the best predictive power among clinical prediction models for cardiovascular and kidney outcomes in patients with type 2 diabetes in a 2021 systematic review and meta-analysis.

RECODe (development & validationexternal validation)

OutcomesDefinitionDiscrimination: C-statisticCalibrationNote
Heart attack/strokeFatal or non-fatal myocardial infarction or stroke0.73-0.77Good/underestimated at lower riskOutperformed UKPDS risk calculator
Heart failureSymptomatic heart failure (regardless of ejection fraction)0.73-0.80Good/over-predicted at lower riskOutperformed UKPDS risk calculator
Kidney failureNeed for dialysis, or serum creatinine concentration >290 umol/L0.78-0.91(Internal validation only) Risk estimates too moderate, but no under/over-prediction
Severe vision loss<20/200 visual acuity by Snellen chart(Internal validation only): 0.62 (0.60-0.64)(Internal validation only) Good
NeuropathyPressure sensation loss0.69 (0.63-0.74)


4. Why is Step 1 producing unintuitive results?

For example, selecting "Anticoagulant" in Step 1 will raise the risk of death, heart attack/stroke, heart failure, and kidney failure. This might seem like an error, since anticoagulants would be expected to lower the risk of some of these events (e.g. heart attack/stroke). However, these risk variables in Step 1 measure more than the effects of anticoagulation.

Rather, the impact on outcomes applied from everything selected in Step 1 is based on their predictive effect in the RECODe clinical prediction model. In the case of anticoagulants, this is not the effect of putting people on these drugs, but the fact that patients already on these drugs are put on them because of certain other factors (e.g. atrial fibrillation, venous thromboembolism) not included in the risk calculator. 

For similar reasons, selecting a statin in Step 1 as a risk variable will not produce the same result as selecting a statin in Step 3 as a treatment option. This is because statin therapy in Step 1 represents also includes other factors associated with statin prescribing (e.g. increased cardiovascular risk). Whereas in Step 3, the predicted risk reduction is causally isolated to the effects of statin therapy. For the most accurate results - use Step 1 if the patient is already on a statin and Step 3 if a statin is being added.

5. How were options in Step 3 chosen and where do the estimates of benefits and harms come from?

We selected pharmacological interventions for inclusion in the PEER Diabetes Medication Decision Aid based on a comprehensive review of guidelines, reviews, as well as consultation with content experts. Treatments are sub-categorized as (1) glucose-lowering medications (2) lipid lowering medications (3) other medications and (4) blood pressure targeting. 

The estimates of benefits and side-effects come from randomized controlled trials (RCTs) and meta-analyses of RCTs. Efficacy is reported as a relative risk (RR), rounded to the nearest 0.05. A patient’s individualized benefit is estimated by applying the cumulative relative risk reduction of all treatments selected in Step 3 to the estimated risk calculated using RECODe in Step 1. 

For side-effects, adverse events that were statistically significantly higher with therapy in RCTs are reported as absolute risk increases.

Relative risks (RR) for diabetes-related outcomes by treatment

TreatmentDeathASCVDHeart failureKidney failureSevere vision lossNeuropathyReferences
ACE inhibitors or ARBs0.75(1)0.75(1)0.80(1)No Albuminuric CKD:

Albuminuric CKD:
NANA1 HOPE-diabetes. Lancet 2000;355:253-9
2 Cochrane Database Syst Rev. 2006;2006:CD006257
Aspirin1.000.90NANANANAASCEND. NEJM 2018;379:1529-39
Blood pressure lowering0.98 per 1 mmHg lowered(1)0.99 per 1 mmHg lowered(1,2)0.98 per 1 mmHg lowered(1,2)1.00(1)NANALancet 2016;387:957-67
Lancet Diabetes Endocrinol 2022;10:645-54

DDP-4 Inhibitors1.00(1)1.00(2-5)1.00(1,2) (Exception: Increase in HF with saxagliptin, included in other considerations)1.00(1)NANA1 BMJ 2023;381:e074068
Can J Hosp Pharm. 2019;72:385-7
3 CARMELINA. JAMA 2019;321:69-79
4 CARMELINA renal subgroup. Diabetes Care 2020;43:1803-12
5 SAVOR. NEJM 2013;369:1317-26
Ezetimibe1.00(1,2)0.90(1)NANANANA1 IMPROVE-IT. NEJM 2015;372:2387-97
2 EWTOPIA 75. Circulation 2019 ;140(12):992-1003. (also reported ASCVD, but not used here given open-label design)
0.90(1, RR 1.00 if eGFR ≥60ml/min/1.73^2 and ACR ≤30mg/mmol)1.00(2)0.80(1, RR 1.00 if eGFR ≥60ml/min/1.73^2 and ACR ≤30mg/mmol)
0.85(1, RR 1.00 if eGFR ≥60ml/min/1.73^2 and ACR ≤30mg/mmol)
NANA1 BMJ. 2023 Apr 6;381:e074068.
2 Eur Heart J. 2022 Feb 10;43(6):474-84.
GLP-1 receptor agonists0.90(1)0.90(2)1.00(3)0.85(1)NA1.00(3)1 BMJ. 2023 Apr 6;381:e074068.
2 Lancet Diabetes Endocrinol. 2019 Oct;7(10):776-85.
 BMJ. 2021 Jan 13;372:m4573.

Insulin1.00 (1)1.00(2,3)1.00(1)1.00(1)1.00(2)NABMJ. 2023 Apr 6;381:e074068.
2 Lancet. 1998 Sep 12;352(9131):837-53.
N Engl J Med. 2012 Jul 26;367(4):319-28.
Metformin1.00(1)0.70 (2; note, no RR available for ASCVD composite, using approximate RR from myocardial infarction [0.67] & stroke [0.80])1.00(1)1.00(1)1.00(3)NA1 BMJ. 2023 Apr 6;381:e074068.
 UKPDS 80. N Engl J Med. 2008 Oct 9;359(15):1577-89.
3 UKPDS 34. Lancet. 1998 Sep 12;352(9131):854-65.
SGLT2 inhibitors0.90(1)0.90(2)0.65(1)No Albuminuric CKD: 0.70(3)

Albuminuric CKD: 0.65(2,4-7)
NA1.00(2)1 BMJ. 2023 Apr 6;381:e074068.
2 JAMA Cardiol. 2021 Feb 1;6(2):148-58.
 BMJ 2021;372:m4573.
4 CREDENCE. N Engl J Med. 2019 Jun 13;380(24):2295-306.
5 DAPA-CKD. N Engl J Med. 2020 Oct 8;383(15):1436-46.
6 Diabetes Res Clin Pract. 2021 Oct;180:109033.
7 Lancet. 2022 Nov 19;400(10365):1788-801.
Statins0.900.80NANANANALancet. 2008 Jan 12;371(9607):117-25.
Sulfonylureas1.00(1)1.00(2,3)1.00(1)1.00(1)1.00(2)NABMJ. 2023 Apr 6;381:e074068.
 Lancet. 1998 Sep 12;352(9131):837-53.
3 Tools for Practice #202

Adverse effects and other considerations by treatment


Adverse effects (absolute % increase)




Caution & Others

ACE inhibitors or ARBs

Cough (ACEI only; +5%)

Hyperkalemia (+2%)

Dizziness (+1%)

Angioedema (ACEI only; +0.2%)

1 HOPE-diabetes. Lancet. 2000 Jan 22;355(9200):253-9.

2 Am J Cardiol. 2012 Aug 1;110(3):383-91.

3 Clin Nephrol. 2005 Mar;63(3):181-7.

About $25 for 90 days ($100/year)

One pill once a day


Major bleed (+1%)

ASCEND. N Engl J Med. 2018 Oct 18;379(16):1529-1539.

About $40/year

One pill once a day

Consider taking with food to minimize stomach upset

Blood pressure lowering

Varies based on which medications are used

Varies based on which medications are used

Varies based on which medications are used

DPP-4 inhibitor

Heart failure hospitalization (Saxagliptin only; +0.7%)

"Gallbladder attack" (cholecystitis) (+0.15%)

BMJ. 2022 Jun 28;377:e068882. 

SAVOR. Circulation. 2014 Oct 28;130(18):1579-88.

About $335 for 90 days ($1400/year)

One pill once a day

Does not increase the risk of severe hypoglycemia



IMPROVE-IT. N Engl J Med. 2015 Jun 18;372(25):2387-97.

About $30 for 90 days ($120/year)

One pill once a day


Hyperkalemia (+6%)

Eur Heart J. 2022 Feb 10;43(6):474-84. 

Not available in Canada

One pill once a day

GLP-1 receptor agonist

Vomiting and/or diarrhea) (+6%)

BMJ. 2021 Jan 13;372:m4573.

About $500-900 for 90 days ($1800-3600/year)

Semaglutide: Available as a tablet (one table once a day) or injection (one injection once a week).

Others: Injection once a day or once a week.

Does not increase the risk of severe hypoglycemia. Causes weight loss (average ~2 kg).


Symptomatic low blood sugar (+32%)

Severe low blood sugar (+4%)

N Engl J Med. 2012 Jul 26;367(4):319-28.

Varies based on dose

Varies from one injection once a day to several times per day

Causes weight gain (average ~2 kg)

Make sure to follow good sick-day management: Does not increase the risk of severe hypoglycemia


Nausea, vomiting, flatulence, and/or diarrhea (+16%)

N Engl J Med. 2006 Dec 7;355(23):2427-43.

About $25 for 90 days ($100/year). May be combined with other diabetes medications in a combination pill.

One or two pills twice a day

Does not increase the risk of severe hypoglycemia

Make sure to follow good sick-day management:

SGLT2 inhibitor

Genital yeast infection (+14%)

Ketoacidosis (+0.1%)

BMJ. 2021 Jan 13;372:m4573.

About $300 for 90 days (or $1200/year). Can prescribe empagliflozin 12.5 mg daily (splitting 25-mg tablets) to cut the cost to ~$150 for 90 days (or $600/year).

One pill once a day

Does not increase the risk of severe hypoglycemia

Make sure to follow good sick-day management:


Muscle pain (+1%)

Elevated liver enzymes (+1%)

Rhabdomyolysis (+0.01%)

Lancet. 2022 Sep 10;400(10355):832-45.

Eur J Prev Cardiol. 2014 Apr;21(4):464-74.

N Engl J Med. 2016 May 26;374(21):2021-31.

About $30 for 90 days ($120/year)

One pill once a day


Severe low blood sugar (+0.6%)

N Engl J Med. 2006 Dec 7;355(23):2427-43.

About $25-35 for 90 days ($100-140/year)

One pill once or twice a day

Causes weight gain (~2kg)

Make sure to follow good sick-day management:

6. Why are only some medications listed in both Step 1 and Step 3?

Medications that are included in Step 1 are used directly by the risk calculator to estimate current risk. On the other hand, the effect of medications listed in Step 3 is based on applying the relative risk reduction derived from randomized controlled trials to the current risk. As a result, for example, the effect of receiving a statin at baseline can be different from the estimated effect of adding a statin "during the visit".

7. How were costs estimated? (last updated 28 Feb 2020)

Cost estimates are based on Canadian dollars (CAD), updated annually or more frequently as needed, and estimated using the following websites:

8. Quality checklist: International Patient Decision Aid Standards instrument (IPDASi) and Readability

INFORMATION: Providing information about options in sufficient detail for making a specific decision8/8
1. Describes the health condition or problem for which the index decision is requiredYes
2. Describes the decision that needs to be consideredYes
3. Describes the options available for the decisionYes
4. Describes the natural course of the health condition or problem if no action is takenYes
5. Describes the positive features (benefits or advantages) of each optionYes
6. Describes negative features (harms, side-effects or disadvantages) of each optionYes
7. Makes it possible to compare the positive and negative features of the available optionsYes
8. Shows the negative and positive features of options with equal detailYes
PROBABILITIES: Presenting outcome probabilities7/8
1. Provides information about outcome probabilities associated with the optionsYes
2. Specifies the defined group of patients for which the outcome probabilities applyYes
3. Specifies the event rates for the outcome probabilities (in natural frequencies)Yes
4. Specifies the time period over which the outcome probabilities applyYes
5. Allows the user to compare outcome probabilities across options using the same denominator and time periodYes
6. Provides information about the levels of uncertainty around event or outcome probabilitiesNo
7. Provides more than one way of viewing the probabilities (e.g. words, numbers & diagrams)Yes
8. Provides balanced information about event or outcome probabilities to limit framing biasesYes
VALUES: Clarifying and expressing values4/4
1. Describes the features of options to help patients imagine what it is like to experience the physical effectsYes
2. Describes the features of options to help patients imagine what it is like to experience the psychological effectsYes
3. Describes the features of options to help what it is like to experience the social effectsYes
4. Asks patients to think about which positive and negative features of the options matter most to themYes
DECISION GUIDANCE: Structured guidance in deliberation and communication2/2
1. Provides a step-by-step way to make a decisionYes
2. Includes tools like worksheets or lists of questions to use when discussing options with a practitionerYes
DEVELOPMENT: Using a systematic development process4/6
1. The development process includes finding out what patients need to prepare them to discuss a specific decisionYes
2. The development process included finding out what health professionals need to prepare them to discuss a specific decision with patientsYes
3. The development process included expert review by patients not involved in producing the decision support technology*Pending*
4. The development process included expert review by health professionals not involved in producing the decision support technologyYes
5. Field tested with patients who were facing the decision*Pending*
6. Field tested with practitioners who counsel patients who face the decisionYes
1. Provides citations to the studies selectedYes
2. Describes how research evidence was selected or synthesizedYes
3. Provides a production or publication dateYes
4. Provides information about the proposed update policyYes
1. Provides information about the funding used for developmentYes
2. Includes author/developer credentials or qualificationsYes
1. Reports readability levelsYes
1. Evidence that the decision aid improves the match between the features that matter most to the informed patient and the option that is chosenNo
2. Evidence that the patient decision aid helps patients improve their knowledge about options’ featuresNo

Flesch-Kincaid Grade Level: 10.0

Developed by:

This tool was developed by the PEER Diabetes Medication Decision Aid Development Panel:

Ricky D. Turgeon BSc(Pharm), ACPR, PharmD (
Assistant Professor – Greg Moore Professor in Clinical and Community Cardiovascular Pharmacy, Faculty of Pharmaceutical Sciences, University of British Columbia, Vancouver, BC
Clinical Pharmacy Specialist - PHARM-HF, St. Paul's Hospital, Vancouver, BC

James McCormack BSc, BSc(Pharm), PharmD
Professor, Faculty of Pharmaceutical Sciences, University of British Columbia, Vancouver, BC

Blair MacDonald BA, PharmD
Research Coordinator, Faculty of Pharmaceutical Sciences, University of British Columbia, Vancouver, BC

Funding: None.

Conflict of interest: Drs. Turgeon, McCormack, and MacDonald declared no conflict of interest.

Translated by:

Nicolas Dugré PharmD, MSc, BCACP

Pharmacist at the CIUSSS du Nord-de-l’Île-de-Montréal and Clinical Associate Professor in the Faculty of Pharmacy at the University of Montreal in Quebec

Version history

Version 1.0 (last update: February 8 2023 | last evidence review December 2022)

Update policy: Update at least annually in January, or more frequently as needed with availability of new evidence.