Choose a diabetes app you will keep using, that shares data your care team can read, protects your privacy and is cleared if it suggests insulin doses. Here is what trials, regulators and privacy studies show, and the red flags to avoid.
To choose a diabetes app, start with the job you want it to do, then look at four things: the evidence behind it, whether a regulator has cleared it for what it claims, what it does with your data, and whether it shares readings in a form your care team can use. In meta-analyses of randomized trials, mostly in type 2 diabetes, people given an app lowered their A1c by about 0.4 to 0.5 percentage points more than people who were not 3, 4.
The American Diabetes Association (ADA) notes that some apps have been fully reviewed and approved by the US Food and Drug Administration (FDA), while others come from people with relatively little knowledge of diabetes treatment 8. A joint report by the ADA and the European Association for the Study of Diabetes (EASD) found that evidence on the safety and effectiveness of diabetes apps “remains limited” 11.
This article describes kinds of apps rather than naming or rating any of them.
Most diabetes apps fall into five groups: the app that comes with your glucose sensor, meter or pump; logging and tracking apps; insulin dose calculator apps; coaching and education programs; and food and carbohydrate apps. Some apps do several of these jobs.
| Kind of app | What it usually does | How regulators usually treat it |
|---|---|---|
| The app that comes with a CGM, meter, pump or AID system | Shows readings and trends, and can share them with family or your care team 12, 13 | Part of a regulated medical device; most CGMs are FDA-cleared class II devices 11, 8 |
| Logging and tracking app | Keeps readings, food, medicines and activity in one record 14 | Not a medical device when it only records and organizes information 1, 2 |
| Insulin dose calculator app | Suggests an insulin dose from planned carbohydrate and measured glucose 7 | Subject to FDA clearance, according to the ADA 8 |
| Coaching or education program | Lessons, prompts, feedback or coaching 8, 5 | In the US, coaching for conditions such as diabetes is treated as low risk, and the FDA does not intend to enforce device rules for it 1 |
| Food and carbohydrate app | Estimates carbohydrate from a database, label, barcode or photo | Stand-alone health apps are largely unregulated unless they meet the definition of a medical device 11 |
A continuous glucose monitor (CGM) measures glucose every few minutes with a sensor inserted under the skin and provides “real-time blood sugar information through smartphone apps” 15. Most CGMs send readings wirelessly to a smartphone app, and some can also send them to a parent, partner or caregiver 12. Readings from many blood sugar testing devices can also be sent to a computer or smart device 14. An automated insulin delivery (AID) system links a CGM to an insulin pump 16, and its dosing algorithm “can be hosted in the pump body, in an insulin pod, or on a phone app” 8.
Coaching programs pair tracking with lessons or contact with a coach, and the ADA recommends considering combining technology with online or virtual licensed coaching 8. Food apps are covered in our guides to counting carbs and carb counting with photos, labels and barcodes, and dose calculations in insulin on board.
In the US, apps that control an insulin pump, analyze glucose readings to raise alerts, or give personal treatment recommendations are regulated as medical devices, while apps that only record and organize your information are not 1.
The FDA’s September 2022 guidance on device software functions and mobile medical apps lists software that is not a medical device, including functions that record (collect and log) data such as blood glucose to share with a health care professional, and tools that organize health information “without providing recommendations to alter or change a previously prescribed treatment or therapy” 1. Coaching software for conditions such as diabetes is low risk, so the FDA “does not intend to enforce requirements” for it 1. Oversight focuses on apps that control insulin delivery on a pump, software that analyzes readings from a blood glucose monitor and alerts the user to readings outside a set range, and software that gives patient-specific diagnosis or treatment recommendations to patients or caregivers 1.
Insulin dose calculators fall on the regulated side. The ADA says they “are subject to FDA clearance to ensure safety and efficacy in terms of algorithms used and subsequent dosing recommendations,” and that people should be encouraged to use ones that are FDA approved 8.
Health Canada draws a similar line. Software is not a medical device only if it meets four criteria, including that it is “not intended to replace the clinical judgement of a health care professional”; devices are ranked from Class I, the lowest risk, to Class IV, the highest 17. In Health Canada’s examples, apps that give people with chronic diseases such as diabetes simple tools to track their information are outside the regulations because they “do not provide alerts or recommendations to alter or change a previously prescribed treatment or therapy” 2. Software that analyzes glucose monitor data to detect a diabetic emergency is Class III, because “an erroneous result would lead to immediate danger” 2.
Across the US and Europe, the ADA and EASD report found that health and wellness apps are “largely unregulated unless they meet the definition of medical devices for therapeutic and/or diagnostic purposes” 11.
In plain terms: An app that stores what you type, or draws a chart of it, is usually not treated as a medical device. An app that tells you how much insulin to take, or analyzes your readings to warn you, can cause harm if it gets the answer wrong, so regulators treat it as one.
On average, yes, by a modest amount: in meta-analyses of randomized trials, mostly in adults with type 2 diabetes, people given an app lowered their A1c by about 0.4 to 0.5 percentage points more than comparison groups 3, 4.
A1c is a blood test that shows your average blood sugar over the past two to three months 16. A 2016 review in Diabetes Care pooled 14 randomized trials with 1,360 participants 3. A1c fell by a mean of 0.49% more with an app than in control groups, with a 95% confidence interval (the range the true effect probably lies in) of 0.30 to 0.68, and the evidence was rated moderate quality. Younger participants were more likely to benefit, the effect was larger with feedback from a health care professional, and there were too few data to judge apps in type 1 diabetes 3.
A 2023 review in The Lancet Digital Health pooled 26 randomized trials (4,546 adults with type 2 diabetes) of digital health programs 4. Overall, A1c fell by 0.30 percentage points more than in control groups. The difference was significant for smartphone apps (0.42 points) and text messages (0.37 points) but not for websites (0.09 points). Results varied a lot between studies, and the evidence was rated moderate 4.
Diabetes Canada cites a meta-analysis in type 2 diabetes in which computer-based self-management programs had a small effect on A1c (−0.2%), larger in the mobile phone subgroup (−0.5%), and no evidence of benefit for other outcomes 5. It says technologies including mobile apps “may be used to support self-management in order to improve glycemic control,” with stronger evidence in type 2 diabetes (Grade A, Level 1A) than in type 1 (Grade B, Level 2) 5.
These averages cannot predict what one app will do for you. The ADA says there are insufficient data to recommend specific apps without randomized trials and validation, unless they are FDA cleared 8. See A1c, GMI and average glucose for how A1c relates to daily readings.
An app only helps while you use it, and many people stop: in 17 studies of app-based programs for chronic diseases, a pooled 43% of participants dropped out 6.
Dropout was 49% in observational studies and 40% in randomized trials, and the reasons included social, demographic and behavioural factors that the authors say could be addressed 6. Diabetes Canada notes that most studies report Internet and web use declining over time 5.
Research on which features keep people logging is thinner than research on A1c, but some patterns recur. Effective programs had a higher dose intensity, such as more frequent app use 4. The A1c effect was larger when a health care professional gave feedback 3. In mobile programs, especially text messaging, Diabetes Canada reports mixed acceptability, including participants asking to stop messages early and low acceptability for challenging interfaces or people inexperienced with the mobile web 5.
That points toward apps in which entering a reading, a meal or a dose is quick. Mayo Clinic’s list of what a record should hold, the date, time, results, medicines and doses, and diet and exercise 14, is a fair test of whether an app captures what you and your care team need (see keeping a blood sugar log).
Many health apps share user data with other companies, and their privacy policies don’t always say so: in one BMJ study, 79% of top-rated medicines apps shared user data, and in a larger one, fewer than half of the data transmissions observed complied with the app’s own privacy policy 9, 10.
The first study tested 24 top-rated medicines-related apps from app stores in the UK, the US, Canada and Australia 9. Data went to 55 companies and services, and 67% of these offered services for collecting and analyzing user data, including analytics or advertising. The authors concluded that sharing of user data “is routine, yet far from transparent” 9.
The second study examined 20,991 medical and health and fitness apps 10. In all, 88.0% contained code that could potentially collect user data and 28.1% had no privacy policy. Of the user data transmissions observed, 23.0% used insecure communication methods and only 47.0% complied with the app’s privacy policy 10. Neither study looked at diabetes apps alone.
The ADA calls online privacy and security “an area of particular importance” and advises reading an app’s policy on data privacy and sharing before entering data 8. A clear policy tells you whether data go only to the developer or also to analytics and advertising services 9, and whether you can choose to share more or less, such as joining a registry or data repository or not 8.
Many can: most CGMs send readings to a phone app, and the ADA recommends that standardized reports of at least one page be available for every CGM, insulin pump, AID system and connected insulin device, with daily and weekly reports and raw data as options 12, 8.
Ideally, the ADA says, people with diabetes and their care teams can both access and analyze the data, between and at clinic visits 8. CGM data can be downloaded to a computer to look at trends and shared with your health care provider 13. Our guide to reading an AGP report walks through one standardized CGM summary.
Links between products are less tidy. The ADA and EASD report lists poor interoperability, meaning devices and apps that don’t easily exchange data, among the main problems with diabetes apps 11. Whether a logging app can read your meter, your CGM or your phone’s built-in health record depends on both products.
Glucose is reported in two units: 10.0 mmol/L is the same as 180 mg/dL 8. A reading in the wrong unit can make a normal value look dangerous or the reverse, so an app should let you choose the unit your meter and care team use (see mmol/L and mg/dL). In the dose calculator assessment, 48% of apps used ambiguous terminology 7.
If you don’t use a smartphone, some CGMs offer a handheld device instead 13, and a written record still works 14. Larger text, screen-reader support, voice entry and menus in your preferred language can all make an app easier to keep using.
The main red flags are a claim to measure glucose without a sensor, insulin dose suggestions without regulatory clearance, big promises without trial evidence, and a missing privacy policy.
Whatever app you use, keep your device’s own alerts working as your care team advises. The NIDDK says that, for safety, “it is important to act quickly if a CGM alarm sounds when your glucose level is too low or too high” 12. Our guide to low blood sugar covers what to do.
No single app is best for everyone. The ADA says there are insufficient data to recommend specific apps without randomized trials and validation, unless they are FDA cleared 8.
In randomized trials, mostly in type 2 diabetes, A1c fell by a mean of 0.49% more with apps than in control groups in one review 3, and by 0.42 percentage points with smartphone apps in another 4. There are too few data to judge apps in type 1 diabetes 3.
Only some. Apps that control an insulin pump, analyze glucose readings to raise alerts, or give personal treatment recommendations are regulated as medical devices, while apps that only record and organize information are not 1, 2. The ADA says insulin dose calculators are subject to FDA clearance 8.
No. Readings from many blood sugar testing devices can be sent to a computer or smart device, and a written or typed record also works 14. A CGM adds an estimate every few minutes, sent wirelessly to an app on your phone 12.
Many health apps do. In a BMJ study, 19 of 24 top-rated medicines apps (79%) shared user data 9. The ADA advises reading an app’s privacy and data-sharing policy before entering any data 8.
Glyve is a logging app, one of the categories above. It keeps glucose, food, insulin, exercise, sleep and stress on one timeline, and you can log by typing, by voice or with a photo of a meal, a nutrition label or a barcode; every estimate is a draft you check before saving. Glucose comes from Apple Health, from Dexcom and FreeStyle Libre through your own Dexcom Share or LibreLinkUp follower account (readings can arrive later than in your CGM’s own app), or from meter readings you type in, so a CGM is not required. You set your own target range, choose mmol/L or mg/dL, and can work offline. When you choose, you can export a PDF summary, a CSV, an AGP report or a glucose graph with logged moments to share with your care team. Glyve is not a medical device: it gives no medical advice, recommends no insulin doses and sends no glucose alarms, so it doesn’t replace your CGM’s own alerts or your care team.
How we write about diabetes. Spotted a mistake? Email info@glyve.ca and we’ll check it against the source.
Glyve keeps glucose, food, insulin, exercise, sleep and stress on one timeline.
For iPhone, iPad and Apple Watch.