Search Results

The MeMed BV test is an automated semi-quantitative immunoassay that measures three non-microbial (host) proteins (TRAIL, IP-10, and CRP) in adult and pediatic serum and venous whole blood samples and is intended for use in conjunction with clinical assessments and other laboratory findings as an aid to differentiate bacterial from viral infection. MeMed BV is indicated for use in patients presenting to the emergency department or urgent care center and with samples collected at hospital admission from patients with suspected acute bacterial or viral infection, who have had symptoms for less than seven days. The MeMed BV test generates a numeric score that falls within discrete interpretation bins based on the increasing likelihood of bacterial infection.

Device Description

The MeMed BV® ("BV test" or the "test") is an In-Vitro-Diagnostic device that measures in parallel the blood concentrations of TRAIL, IP-10 and CRP. The test consists of an automated analyzer with built-in hardware and software that conduct chemiluminescence based analyte measurements of patient serum and venous whole blood samples and their computational integration (MeMed Key®), and a disposable cartridge that contains the reagents and controls needed to detect the analytes of interest (MeMed BV® cartridge). The test generates an answer to each sample, with a test run time of approximately 15 minutes.

AI/ML Overview

Here's a breakdown of the acceptance criteria and the study proving the device meets them, based on the provided FDA 510(k) summary for MeMed BV:

The MeMed BV test is intended for use in conjunction with clinical assessments and other laboratory findings as an aid to differentiate bacterial from viral infection in patients presenting to the emergency department or urgent care center, or with samples collected at hospital admission, who have had symptoms for less than seven days. The device generates a numeric score that falls within discrete interpretation bins based on the increasing likelihood of bacterial infection.

1. Table of Acceptance Criteria and Reported Device Performance

The 510(k) summary details various analytical performance studies and a clinical study to support the expanded indications for use. Key acceptance criteria and reported performance include:

Test Category	Acceptance Criteria	Reported Device Performance
Analytical Performance
Limit of Quantitation (LoQ)	Total Error (TE): TRAIL < 30%, IP-10 < 40%, CRP < 30%	Serum Test Script: - TRAIL: Max TE at LLOQ (15 pg/mL) was 21%. (One sample at X0.8 showed 51% TE for TRAIL, but the defined LLOQ concentration level (X1.0) met criteria).- IP-10: Max TE at LLOQ (100 pg/mL) was 15%.- CRP: Max TE at LLOQ (1 mg/L) was 9%.Whole Blood (WB) Test Script: - TRAIL: Max TE at LLOQ (15 pg/mL) was 10%.- IP-10: Max TE at LLOQ (100 pg/mL) was 14%.- CRP: Max TE at LLOQ (1 mg/L) was 10%. All defined LLOQ concentrations for both serum and WB met the acceptance criteria.
Reproducibility/Precision	Measurands (TRAIL, IP-10, CRP): CV ≤ 15% (for concentrations above LoQ).MeMed BV® Test Score: SD < 12.5 score units.	Serum Samples: All reported repeatability, intermediate precision, and reproducibility CVs for TRAIL, IP-10, and CRP were ≤ 11.3%. All reported SDs for the MeMed BV Score were ≤ 6.6 score units. WB Samples (Precision): All reported CVs for TRAIL, IP-10, and CRP were ≤ 12.0%. All reported SDs for the MeMed BV Score were ≤ 3 score units. All reported values met the pre-established acceptance criteria.
Lot-to-Lot Reproducibility	Measurands (TRAIL, IP-10, CRP): CV ≤ 15% (for concentrations above LoQ).MeMed BV® Test Score: SD < 12.5 score units.	All reported between lot CVs for TRAIL, IP-10, and CRP were ≤ 10.7%. All reported between lot SDs for the MeMed BV Score were ≤ 2.3 score units. All reported values met the pre-established acceptance criteria.
Linearity	Allowable deviation from linearity (ADL) < 15% or 10 mg/L for CRP; 15% or 10 pg/mL for TRAIL; 20% or 50 pg/mL for IP-10.	Serum Samples: Max observed % deviation from linearity was 6.8% (TRAIL).Whole Blood Samples: Max observed % deviation from linearity was 8.6% (TRAIL). All results were within acceptance criteria.
Hook Effect	No hook effect observed up to tested concentrations (TRAIL – 1,000 pg/mL, IP-10 – 10,000 pg/mL, CRP – 500 mg/L).	All concentrations up to TRAIL – 1,000 pg/mL, IP-10 – 10,000 pg/mL, and CRP – 500 mg/L showed higher signal than the ULOQ sample. No hook effect observed.
Carry Over	WB Samples: Difference between average score of high score sample run after low score sample and high score sample baseline average score of ≤ 12.5 score units. Difference between average score of low score sample run after high score sample and low score sample baseline average score of ≤ 12.5 score units.	Maximal difference in score obtained for high score sample was 1.4 score unit difference. No carry-over occurred with the MeMed BV test.
Interference/Cross Reactivity	95% Confidence Interval for bias within +/- 12.5 score units for all interferents and cross-reactants.	Previously submitted data (K222332) demonstrated this. The recovery of TRAIL, IP-10 and CRP were within the predetermined +/- 10% of the sample nominal concentration. Assays are tolerant to high HAMA concentrations and no interference/cross-reactivity from tested compounds.
Correlation to Reference Standard (New Calibration Scheme vs. Legacy)	1. <5% of samples have MCC scores deviating from legacy calibration scores by an amount placing them in non-adjacent bins.2. Pearson correlation > 0.95.3. Absolute bias < 12.5 units at bin cutoff points (10, 35, 65, 90).	1. The study successfully met the clinically relevant criterion (no paired samples assigned to nonadjacent bins).2. Pearson correlation was 1 (Deming Regression slope=1.00, 95% CI 0.99-1.00; intercept 0.00-0.06).3. Estimated bias at cutoff points ranged from -0.57 to -0.55, with 95% CIs well within +/- 12.5 units. New MCC is equivalent to legacy calibration.
Sample In-Use Stability (WB)	Allowable handling conditions demonstrated from blood draw to sample input.	The minimal acceptable period of time was approximately 140 minutes for TRAIL viral sample 1. Formal in-use stability of WB sample type established at 120 minutes prior to testing on analyzer.
Clinical Studies
Matrix Equivalency (WB vs. Serum)	Passing & Bablok Regression: Slope in range of 0.9-1.1; Intercept in range of (-5) to 5.	Slope: 1.00 (95% CI 0.99-1.00). Intercept: 0.00-0.06. Both predefined acceptance criteria for analytical equivalency were fulfilled.
Bin Impact Analysis (WB vs. Serum)	<5% of paired samples demonstrating a score deviation that causes a patient to be assigned to a nonadjacent bin.	No paired samples demonstrated a score deviation that caused the patient to be assigned to a nonadjacent bin. This strengthens the conclusion of analytical equivalency.
Diagnostic Accuracy (Simulated WB against Adjudication)	Cochran-Armitage (CA) Test for trend: Reject null hypothesis (no trend of increasing probability of bacterial infection with higher test score) for ≥ 95% of simulations.Likelihood Ratio (LR): 95% CI should exclude 1 for some bins (preferably Bins 1,2,4,5) for ≥ 95% of simulations.	All-inclusive cohort: CA p<0.001 for 100% of 100K simulations. CI of LR for exactly 4 bins (1,2,4,5) excluded 1 in 100% of simulations.Suspected cohort: CA p<0.001 for 100% of 100K simulations. CI of LR for exactly 4 bins (1,2,4,5) excluded 1 in 99.98% of simulations (0.02% had all 5 bins exclude 1). Both acceptance criteria passed, validating diagnostic accuracy of simulated WB samples.

2. Sample Sizes Used for the Test Set and Data Provenance

Analytical Performance Test Sets:
- LoQ: Samples used per test script (serum and whole blood) and two cartridge lots. Specific number of unique samples not detailed, but each was tested three times on three non-consecutive days with results across 4 concentration levels.
- Reproducibility/Precision & Lot-to-Lot Reproducibility: Panel of 4 scores (representing various infection statuses) for serum, and 3 scores for WB. Serum study involved 90 replicates per panel member across 3 labs. WB study involved runs on 5 different analyzers. Lot-to-lot used 18 replicates per panel member.
- Linearity: Five replicates of eleven dilutions for each measurand.
- Hook Effect: 4 samples with varying high concentrations.
- Carry Over: Two whole blood samples (one high score, one low score) run in sequences.
- Correlation to Reference Standard (Calibration Scheme Comparison): 100 serum specimens.
Clinical Study Test Set (Perseverance Study):
- Sample Size: 216 prospectively recruited subjects.
- Data Provenance: Multi-center study from 5 medical centers (2 in the US, 3 in Israel).
- Retrospective/Prospective: Prospective.

3. Number of Experts Used to Establish the Ground Truth for the Test Set and Qualifications of Those Experts

The document states that the Apollo study (NCT04690569), which provided the basis for the original serum MeMed BV clearance (K210254), used a rigorous reference standard based on etiological adjudication by experts provided with comprehensive patient data.

Number of Experts: Not explicitly stated how many experts for the adjudication, but plural "experts" is used.
Qualifications of Experts: Not explicitly stated but they were responsible for "etiological adjudication" which implies medical professionals with expertise in differential diagnosis of infections (e.g., infectious disease specialists, clinical microbiologists, relevant clinical physicians). The term "comprehensive patient data" suggests they had access to clinical, laboratory, and other relevant information.

4. Adjudication Method for the Test Set

The ground truth for the clinical utility (diagnostic accuracy) of the MeMed BV was based on an adjudication-based reference standard from the Apollo study. The document doesn't specify the exact adjudication method (e.g., 2+1, 3+1), but implies a consensus or majority decision by the experts based on comprehensive patient data.

5. If a Multi-Reader Multi-Case (MRMC) Comparative Effectiveness Study Was Done

No, an MRMC comparative effectiveness study was not mentioned as part of this 510(k) submission. This submission primarily focuses on analytical equivalency of whole blood samples to serum samples and simulated diagnostic accuracy, not human reader improvement with AI assistance. The MeMed BV is an in vitro diagnostic device that provides a numeric score, not an imaging AI algorithm designed to assist human readers.

6. If a Standalone (i.e., algorithm only without human-in-the-loop performance) Was Done

The device's performance, represented by its numeric score and bin assignment, is inherently standalone in terms of its output. The analytical performance evaluations (LoQ, precision, linearity, etc.) and the clinical study's simulation of diagnostic accuracy (comparison of MeMed BV score to adjudicated ground truth) assess the algorithm's performance directly. The device is intended to be used "in conjunction with clinical assessments and other laboratory findings," meaning its score is an aid, but its performance itself is measured as a standalone diagnostic aid.

7. The Type of Ground Truth Used

For Analytical Performance: Ground truth is established by the known concentrations of analytes in controls/calibrators, established reference methods, or through robust statistical measurements.
For Clinical Performance (Diagnostic Accuracy Simulation): The ground truth was an etiological adjudication by experts provided with comprehensive patient data from the Apollo study. This is a form of expert consensus based on extensive clinical information.

8. The Sample Size for the Training Set

The document does not specify a separate "training set" sample size for the MeMed BV algorithm because it is an in vitro diagnostic device that measures specific biomarkers and computationally integrates them. The algorithm's "training" or development would have occurred prior to the studies presented for this 510(k) (which are validation studies). The previous 510(k) (K222332) for the serum-only version would have covered the initial development and validation, and this submission focuses on extending the indication to whole blood samples.

The phrase "new Master Calibration Curve (MCC)" suggests a change to the underlying measurement calculation, which might involve a new calibration dataset for the algorithm, but this is an analytical validation, not a "training set" in the sense of machine learning model development. For the calibration scheme comparison, 100 serum specimens were used as a test set.

9. How the Ground Truth for the Training Set Was Established

As noted above, a distinct "training set" for the MeMed BV algorithm itself (as a machine learning model might have) is not described in this regulatory submission. For the analytical validations:

Calibration: Established against reference standards and through controlled experiments to define the relationship between measured signals (RLU) and analyte concentrations. The "Master Calibration Curve" is related to how the instrument translates raw signals into quantitative measurements and ultimately into the BV score.
Clinical Utility (Reference Standard): If earlier development involved classification algorithms, their "ground truth" would have similarly been derived from expert adjudication or robust clinical diagnoses in previous studies (like the Apollo study mentioned, which provided the original basis for serum clearance). The current "clinical studies" section primarily describes validation of the expanded indication.

Ask a Question

Ask a specific question about this device

Page 1 of 1