Diabetes and chronic disease
Software that analyzes glucose, insulin and laboratory data to recommend insulin doses, coach patients between visits or estimate the risk that chronic kidney disease will progress.
Category overviewHide category overview
These tools analyze glucose, insulin and laboratory data to recommend insulin doses, coach patients between visits or estimate the risk of complications. FDA created the device type insulin pump therapy adjustment calculator for healthcare professionals when it granted DreaMed's De Novo request in June 2018, and granted a De Novo request for KidneyIntelX, a machine learning prognostic test for diabetic kidney disease, in June 2023. Prescription coaching apps such as BlueStar reached the market through 510(k) clearance.
Randomized evidence is thin. In the ADVICE4U trial, algorithm-guided insulin pump adjustments were noninferior to intensive titration by academic specialists in 108 youths, and a pragmatic trial of the BlueStar app found no difference in HbA1c. A European Diabetes Forum working group has noted concerns that regulatory processes may not be adequate for AI systems that adapt to new data, and an American Diabetes Association scientific statement said diabetes technologies can be integrated into primary care with standardized workflows, training and access to specialty expertise.
DreaMed Advisor Pro
Decision-support software that analyzes glucose and insulin data and recommends insulin dose adjustments for clinicians to review.
Full reviewHide full review: DreaMed Advisor Pro
Advisor Pro analyzes continuous glucose monitor, meter and insulin pump or injection data and recommends new settings, such as basal rate, carbohydrate ratio and correction factor, for the clinician to review. FDA's decision summary states that it does not replace clinical judgment.
Why this grade. The ADVICE4U randomized trial in 108 youths using insulin pumps met its primary outcome of noninferiority to specialist physicians for time in glucose range. Noninferiority shows the software matched physicians rather than improved care, and no trial measured clinician time saved, so the grade is not an A.
The evidence
In ADVICE4U, a six-month trial at academic centers, participants aged 10 to 21 received insulin dose adjustments every three weeks guided by the software or by physicians. Time in range was 50.2% with the software and 51.6% with physicians, and time below 54 mg/dL was 1.3% against 1.0%. Three severe diabetes-related adverse events occurred in the physician arm and none in the software arm.
In a survey study, 17 physicians from 11 countries agreed with each other on the direction of pump adjustments about half the time, and the software's agreement with them was similar. A single-arm study of 16 young adults on injections who received recommendations directly reported HbA1c falling from 7.5% to 7.1%; DreaMed employees were coauthors of both.
- Used by
- Academic diabetes centers in the U.S., Israel, Germany and Slovenia, including Joslin and the Barbara Davis Center, took part in its randomized trial, which DreaMed supported as a collaborator.
- FDA
- FDA granted the De Novo request on June 12, 2018 (DEN170043), creating the device type insulin pump therapy adjustment calculator for healthcare professionals. K201476 (Aug. 28, 2020) extended use to patients over 65, and K210561 (Sept. 28, 2021) added injection users and type 2 diabetes. Later 510(k) clearances extended use to patients over 65 (K201476, 2020) and added injection users and type 2 diabetes (K210561, 2021).
- Randomized trial
- Yes. ADVICE4U, a six-month randomized noninferiority trial in 108 youths using insulin pumps (Nature Medicine, 2020).
- Key study
- Nature Medicine, 2020, ADVICE4U randomized noninferiority trial, 108 youths
- Limits
- The trial compared the software with intensive titration at specialist centers, not with usual care, and had 108 participants. FDA's decision summary said the small size, short duration and design of its supporting studies precluded definitive conclusions about glycemic benefit. Most studies involved the same Israeli research group and company staff.
- Cost
- No published price found. checked
- Specialties
- Pediatrics, Endocrinology
KidneyIntelX
Machine learning blood test that sorts adults with type 2 diabetes and kidney disease into risks of progression.
Full reviewHide full review: KidneyIntelX
KidneyIntelX combines three plasma biomarkers (TNFR-1, TNFR-2 and KIM-1) with blood urea nitrogen, HbA1c and urine albumin-to-creatinine ratio in a random forest model and reports a low, moderate or high risk of a sustained 40% fall in eGFR within five years, according to FDA's decision summary. It is not for screening or stand-alone diagnosis.
Why this grade. Observational studies in a health system program of 5,348 patients and two medical centers found more SGLT2 inhibitor prescribing and referrals among higher-risk patients after testing. No randomized trial has tested whether using the result improves kidney outcomes, so the grade is not an A.
The evidence
In the Mount Sinai program, new prescriptions for an SGLT2 inhibitor or GLP-1 receptor agonist, or specialist referral, followed in 19%, 33% and 43% of low, intermediate and high-risk patients, and the eGFR slope in high-risk patients improved from minus 7.08 to minus 4.27 mL/min/1.73 m² a year. Renalytix staff were authors.
In a derivation and validation study of 1,146 biobank patients, the area under the curve was 0.77, against 0.61 for a clinical model. In validation cohorts of 657 and 1,197 patients, progression occurred in 66.9% and 59.6% of high-risk patients and 5.9% and 6.7% of low-risk patients. In banked samples from the CANVAS trial, canagliflozin's effect on eGFR slope was numerically largest in high-risk patients.
- Used by
- A health system population health program described by Mount Sinai and Renalytix authors tested 5,348 patients from November 2020 to April 2023.
- FDA
- FDA granted the De Novo request for kidneyintelX.dkd on June 29, 2023 (DEN200052). The decision summary indicates it for adults with type 2 diabetes and an eGFR of 30 to 59 mL/min/1.73 m², or 60 or higher with a urine albumin-to-creatinine ratio of at least 30 mg/g.
- Randomized trial
- None published
- Key study
- Journal of Primary Care and Community Health, 2024, observational before-and-after study in a health system program, 5,348 patients
- Limits
- The care studies are before-and-after comparisons without control groups, run with the company, and changes in prescribing may reflect the accompanying care pathway rather than the test. The validation studies are retrospective, on banked samples, and many authors were Renalytix employees.
- Cost
- Renalytix said in June 2024 that the established Medicare price for kidneyintelX.dkd is $950 per test. checked
- Specialties
- Family medicine, Internal medicine, Endocrinology, Nephrology
Welldoc BlueStar
Prescription diabetes app that logs glucose and medications, sends automated coaching messages and reports to clinicians.
Full reviewHide full review: Welldoc BlueStar
BlueStar captures glucose readings, medications, carbohydrates and activity, sends real-time motivational, behavioral and educational messages based on glucose values and trends, and offers workflow and decision support to clinicians, according to its FDA summary; the Rx version adds an insulin dose calculator. Company researchers say the coaching messages are generated with artificial intelligence.
Why this grade. Randomized trials disagree. A 2008 trial of 30 patients found HbA1c fell 2.03 points with WellDoc software against 0.68 without it, but a pragmatic trial of 223 patients at three Ontario diabetes programs found no significant difference in HbA1c at three months.
The evidence
In the 2008 trial, at three community practices over three months, 84% of intervention patients had medications changed by their clinicians against 23% of controls; the P value was one-tailed.
In the Ontario trial, the baseline-adjusted HbA1c difference at three months was minus 0.42 points (95% CI minus 1.05 to 0.21), and self-care, quality of life and health care use did not differ. Use was low: 46.4% of intervention patients logged in on 10 days or fewer over 26 weeks. In a company study of 52 users of BlueStar with a Dexcom sensor, time in range rose 15 percentage points among those with mean glucose above 180 mg/dL.
- Used by
- Participants in a company study were current or former enrollees in an employer-sponsored health plan.
- FDA
- FDA cleared DiabetesManager through K100066 on July 15, 2010, BlueStar and BlueStar Rx through K162532 on Jan. 12, 2017, for adults 21 and older with type 2 diabetes, and the BlueStar CGM insulin dose calculator through K222888 on Aug. 11, 2023.
- Randomized trial
- Yes. A 2008 trial of 30 patients at three community practices (Diabetes Technology and Therapeutics, 2008) and a pragmatic trial of 223 patients at three Ontario diabetes programs (JMIR mHealth and uHealth, 2019).
- Key study
- JMIR mHealth and uHealth, 2019, pragmatic randomized trial at three Ontario diabetes programs, 223 patients
- Limits
- The positive trial was small, three months long and analyzed one-tailed; the larger pragmatic trial was underpowered for small differences and lost 34.5% of participants. The newer data with continuous glucose monitors come from a small, uncontrolled company study.
- Cost
- MobiHealthNews, citing an August 2012 New York Times report, wrote that two unnamed insurers had agreed to pay more than $100 a month for WellDoc's earlier product, DiabetesManager; Welldoc's site lists no BlueStar price. checked
- Specialties
- Family medicine, Internal medicine, Endocrinology
Reviews of the published evidence, not medical advice or an endorsement of any product.