FDA 510(k) Search - By Innolitics (SaMD and AI Experts)

The provided 510(k) summary for "My Insulin Doser/IDS" offers limited information regarding specific acceptance criteria and detailed study results that would typically be found in a comprehensive clinical or performance study report. The document primarily focuses on establishing substantial equivalence to a predicate device.

Based on the provided text, here's an attempt to answer your questions, highlighting where information is missing or inferred:

1. A table of acceptance criteria and the reported device performance

The document does not explicitly state quantitative acceptance criteria (e.g., specific accuracy thresholds for insulin dosage calculation) for the "My Insulin Doser/IDS." It broadly states: "Performance testing was performed and all tests showed satisfactory results." and "My Insulin Doser/IDS uses the same fundamental technology features as the Intelligent Dosing System and delivers the same level of effectiveness."

Since the device's function is a "dosing calculator," the implied acceptance criterion would be accuracy of insulin dose calculation. However, no specific metrics or reported performance values (e.g., percentage accuracy, mean absolute error) are provided for this.

The "Labeling Comprehension study" does mention a design target of "6th grade reading comprehension level," which could be considered an acceptance criterion for usability, but no specific performance metric (e.g., percentage of users achieving comprehension) is given for this.

2. Sample sized used for the test set and the data provenance (e.g. country of origin of the data, retrospective or prospective)

Labeling Comprehension Study:
- Sample Size: 55 patient users.
- Data Provenance: Not explicitly stated, but based on the overall context of an FDA submission in the US, it's likely US-based. It was a usability study, so inherently prospective.
Clinical Performance Data:
- Sample Size: "nursing home residents who were maintained on two or more OADs." The exact number is not specified.
- Data Provenance: Not explicitly stated, but likely US-based given the FDA submission. The study is described as a "Phase IV" study, which by definition is typically prospective, after a drug/device has been broadly marketed. The context suggests a prospective study where insulin titration was employed.

3. Number of experts used to establish the ground truth for the test set and the qualifications of those experts (e.g. radiologist with 10 years of experience)

This information is not provided for either the labeling comprehension study or the clinical performance data. For the "Labeling Comprehension Study," the "ground truth" would be related to users' understanding, likely assessed by researchers or observers, but their qualifications are not mentioned. For the "Clinical Performance Data," the "insulin replacement algorithm" was "developed by Dimensional Dosing Systems, Inc.," which implies internal expertise, but no external experts are mentioned for establishing ground truth for individual patient cases in the context of the study.

4. Adjudication method (e.g. 2+1, 3+1, none) for the test set

This information is not provided.

5. If a multi reader multi case (MRMC) comparative effectiveness study was done, If so, what was the effect size of how much human readers improve with AI vs without AI assistance

No, an MRMC comparative effectiveness study was not explicitly mentioned for the "My Insulin Doser/IDS."
This device is a "dosing calculator," not an imaging or diagnostic AI, so the concept of "human readers" and "AI assistance" in the typical diagnostic sense does not directly apply. The device is the "AI" (algorithm) assisting the patient or healthcare professional.
The comparison is to the predicate device, which uses the "same algorithm." The effectiveness is derived from the algorithm itself, which calculates doses. The benefit primarily comes from providing a calculation tool.

6. If a standalone (i.e. algorithm only without human-in-the loop performance) was done

Yes, the device itself is a standalone algorithm, but intended for use by patients (human-in-the-loop). The "Performance testing" mentioned for both the proposed device and the predicate device would likely involve testing the algorithm's calculations against known correct answers or simulated scenarios. However, the details of such standalone testing (e.g., specific test cases, expected outcomes) are not described. For the "My Insulin Doser/IDS," the patient enters values, and the device calculates a dose; thus, the human is intrinsically involved in data input and eventual administration based on the calculation.

7. The type of ground truth used (expert concensus, pathology, outcomes data, etc)

For the "Labeling Comprehension study," the ground truth would be the correct understanding of the instructions as defined by the device manufacturer, assessed through survey or observational methods.
For the "Clinical Performance Data," the study aimed to "help stabilize blood glucose following conversion from OAD therapy." The clinical outcome (stabilized blood glucose) would be the ultimate measure, and perhaps specific blood glucose targets or ranges would define "ground truth" for successful insulin titration per protocol. No specific method for establishing a "ground truth" for individual insulin doses (e.g., expert-derived ideal dose for each patient day) is described. The "ground truth" for the algorithm's accuracy in calculation is not specified beyond being based on the same algorithm as the predicate.

8. The sample size for the training set

The document does not mention a training set specific to the "My Insulin Doser/IDS" algorithm. It states, "Uses the same algorithm" as the predicate device (TRxF Intelligent Dosing System, K011571). This implies the core algorithm was developed and validated prior to this submission. Therefore, if there was a "training set," it would have been for the predicate device's algorithm development, which is not described here.

9. How the ground truth for the training set was established

As no training set for the "My Insulin Doser/IDS" is mentioned (due to using an existing algorithm), this information is not provided. For the predicate device's algorithm, the ground truth for developing its dosing logic would typically involve medical expertise, pharmacological models, clinical guidelines, and potentially retrospective patient data, but these details are not present in this document.

Summary

Regulation Number and Section

§ 868.1890 Predictive pulmonary-function value calculator.

(a)
Identification. A predictive pulmonary-function value calculator is a device used to calculate normal pulmonary-function values based on empirical equations.(b)
Classification. Class II (performance standards).