Know Labs, Inc. announced further interim results from a clinical research study that assessed the accuracy of Know Labs? proprietary non-invasive radiofrequency (RF) dielectric sensor in measuring blood glucose. Participants with prediabetes and Type 2 diabetes were studied and venous blood was used as a comparative reference.

The study found that the accuracy of Know Labs? proprietary sensor in estimating blood glucose values remained statistically stable, with an expanded dataset and a new machine learning (ML) model. This study reflects the latest results in Know Labs?

first clinical research protocol involving people with diabetes and using venous blood as a comparative reference. In March of this year, Know Labs presented early interim results from the same clinical research protocol at the 17th International Conference on Advanced Technologies & Treatments for Diabetes (ATTD), in which its non-invasive blood glucose monitor and ML model trained on data collected in a lab setting achieved a MARD of 11.1%. Compared to the previous study, which was based on 10 participants and 650 paired RF and reference blood glucose values, this study involved more than twice as much data collected from 22 participants yielding 1,430 paired values.

Study Design: The proprietary RF sensor employed in the study measures glucose levels using dielectric spectroscopy by rapidly scanning a large range of RF frequencies and recording voltage values detected at each frequency to quantify, with trade-secret ML algorithms, real-time continuous blood glucose levels. The sensor continuously scanned participants' forearmsfor up to three hours during a 75g Oral Glucose Tolerance Test. From the 22 participants, 1,430 venous blood samples were collected using a peripheral intravenous catheter and analyzed using an FDA-cleared blood glucose hospital meter as a reference device.

Data was preprocessed using smoothing techniques and an 80/20 split was performed to create model training and test datasets, respectively. Know Labs trained a ML model to estimate reference venous blood glucose values on 80% of the data consisting of 1,143 paired RF device and venous blood glucose values randomly selected from the total dataset and then tested on the remaining, held-out 20% of data (287 paired values). Results: On the held-out test dataset, blood glucose was estimated with a MARD of 11.8% ± 1.5% relative to venous blood.

It performed similarly on normoglycemic (12.1% ± 1.8%) and hyperglycemic (11.0% ± 2.3%) ranges. Compared to the previous results presented in March, the MARD of 11.8% is not statistically significantly different, nor were the results in the hyperglycemic range (>180 mg/dL) and normoglycemic range (70 to 180 mg/dL), indicating stability in accuracy. These interim results are part of a larger, now completed clinical study with over 30 participants, conducted September 2023 through February 2024.

As the Company continues on the path toward FDA clearance, Know Labs will deploy the recently announced KnowU wearable non-invasive continuous glucose monitor in ongoing clinical and bench studies. The wearable form factor of the KnowU device allows these studies to evaluate the technology?s performance throughout continuous wear, in ?real-world? environments outside of the lab where new elements of interference are likely, and within more extreme glycemic ranges (below 70 mg/dL and above 350 mg/dL).

As new data is collected in these areas and additional variables come into play, the Company will make necessary refinements to the device and accompanying algorithms.