LogoFDA 510(k) Search
Search Filters

Search Results

Found 1 results

510(k) Data Aggregation

    K Number
    DEN200052

    Validate with FDA (Live)

    Device Name
    KidneyIntelX.dkd
    Manufacturer
    Renalytix AI, Inc.
    Date Cleared
    2023-06-29

    (1037 days)

    Product Code
    QWZ
    Regulation Number
    862.1223
    Type
    Direct
    Panel
    Clinical Chemistry
    Age Range
    All
    Reference & Predicate Devices
    N/A
    Predicate For
    N/A
    AI/MLSaMDIVD (In Vitro Diagnostic)TherapeuticPediatricDiagnosticis PCCP AuthorizedThirdpartyExpeditedreview
    Intended Use

    KidneyIntelX.dkd is for in-vitro diagnostic use for the determination of a KidneyIntelX.dkd Level using an algorithm to combine clinical variables (blood urea nitrogen (BUN), hemoglobin A 1c (HbAlc) and urine albumin creatinine ratio (uACR)) and the quantitative measurements of tumor necrosis factor receptor-1 (TNFR-1), tumor necrosis factor receptor-2 (TNFR-2) and kidney injury molecule-1 (KIM-1) in human plasma employing a Meso Sector S 600 electrochemiluminescence immunoassay. It is indicated for use as an aid in assessment of the risk of progressive decline in kidney function (sustained decrease in eGFR greater than or equal to 40% lasting more than 3 months) within a period of up to 5 years following KidneyIntelX.dkd Level measurement in adult patients with type 2 diabetes and existing chronic kidney disease (defined for the purposes of this device as patients with an estimated glomerular filtration rate of 30-59 ml/min/1.73m2 or eGFR ≥ 60 ml/min/1.73m2 with albuminuria (uACR ≥ 30 mg/g)).

    KidneyIntelX.dkd is not intended for screening or as a stand-alone diagnostic test.

    Device Description

    KidneyIntelX.dkd provides the KidneyIntelX.dkd level, generated by the KidneyIntelX.dkd algorithm. The algorithm includes inputs from clinical data collected on the patient as well as component analytes measured directly by KidneyIntelX.dkd.

    The three component analytes measured by KidneyIntelX.dkd directly are quantitated via a multiplex electrochemiluminescence technology for the in vitro quantitative determination the concentration of three circulating biomarkers in human EDTA plasma:

    • i) soluble Tumor Necrosis Factor Receptor 1 (TNFR-1),
    • ii) soluble Tumor Necrosis Factor Receptor 2 (TNFR-2),
    • iii) and Kidney Injury Molecule (KIM)-1.

    The three measured components of the KidneyIntelX.dkd level are measured using the MESO SECTOR® S 600 instrument, by trained laboratory personnel using the assay components that includes the KidneyIntelX 96-well plate, the detection antibody, calibrator, and controls along with the Meso Scale Diagnostics diluents, blocker, wash buffer and read buffer.

    A sandwich immunoassay is used to measure KIM-1, TNFR-1, and TNFR-2. The plate wells are precoated with goat polyclonal antibodies to KIM-1, and mouse monoclonal antibodies specific to TNFR-1 and TNFR-2. After treating the wells with a blocking solution, the specimens and controls are incubated in the wells then washed. The analytes are quantified via electrochemiluminescence, with secondary detection antibodies (goat polyclonal antibodies to KIM-1, and mouse monoclonal antibodies specific to TNFR-1 and TNFR-2) covalently linked to electrochemiluminescent labels. The MESO SECTOR® S 600A instrument measures the light emitted via charged-coupled device (CCD) camera.

    A software with an artificial intelligence derived algorithm provides a risk assessment score by combining the biomarker results from the assay (TNFR-1, TNFR-2, and KIM-1) and a set of clinical data, which are provided by a patient's physician via a test requisition form (urine albumin creatinine ratio [uACR], hemoglobin A1c [HbA1c], and Blood Urea Nitrogen [BUN]). The algorithm generates a risk level (Low, Moderate, or High) for progressive decline in kidney function, defined as sustained decrease (<40% of baseline value) in eGFR. The KidneyIntelX.dkd algorithm was modeled using machine learning (Random Forest) to identify and analyze risk factors for kidney disease progression.

    The software is installed and maintained in the KidneyIntelX Portal, a cloud-based system. The biomarker results and clinical data set are manually inputted into the KidneyIntelX Portal determining the KidneyIntelX.dkd level. A KidneyIntelX.dkd Test Report containing the KidneyIntelX.dkd level is provided to the ordering physician. The KidneyIntelX.dkd level is to be used in conjunction with clinical information as an aid in assessing progressive decline in kidney function in T2D patients with DKD.

    AI/ML Overview

    Here's a breakdown of the acceptance criteria and the study proving the device meets them, based on the provided text:

    Acceptance Criteria and Device Performance

    Acceptance CriteriaReported Device PerformanceComments
    Clinical Performance: Ability to predict progressive decline in kidney function (sustained decrease in eGFR ≥ 40% lasting > 3 months) or reaching end stage kidney disease (sustained eGFR <15 ml/min/1.73 m2) within up to 5 years.Shows significant differentiation in risk of progressive decline in kidney function across Low, Moderate, and High KidneyIntelX.dkd Levels.
    - Low LevelEstimated 5-year event rate: 5.9% (95% CI: 3.6-9.4%)
    - Moderate LevelEstimated 5-year event rate: 21.5% (95% CI: 15.9-28.6%)
    - High LevelEstimated 5-year event rate: 66.9% (95% CI: 49.3-83.5%)
    Analytical Precision/Reproducibility: Classification consistency despite cumulative imprecision of all 6 inputs.Majority of patients remained in the measured level across 100 simulations for each patient. For KidneyIntelX.dkd Level: - Low: 89% remained Low - Moderate: 90.4% remained Moderate - High: 85.5% remained High No patients changed from "low" to "high" level.Worst-case scenario imprecision used for analytes; publicly available imprecise results for BUN, HbA1c, uACR. Criteria: <95% match original category considered affected by imprecision.
    Analytical Bias: Classification consistency despite systematic biases in input features.Majority of patients remained in the measured level across 100 simulations for each patient. For KidneyIntelX.dkd Level: - Low: 96.0% remained Low - Moderate: 94.3% remained Moderate - High: 96.4% remained High No patients changed from "low" to "high" level.Worst-case scenario bias used; higher bias than observed in labeled performance of assays.
    Analytical Linearity: KIM-1, TNFR-1, and TNFR-2 measurements.Deviation from linearity < ±10% within claimed measuring intervals.Conforms to CLSI EP06-Ed2.
    Analytical Specificity/Interference: Robustness against interfering substances.Majority of patients remained in the measured level across 100 simulations for each patient. For KidneyIntelX.dkd Level: - Low: 84.2% remained Low - Moderate: 68.9% remained Moderate - High: 80.0% remained High No patients changed from "low" level to "high" level. Some specific interferences identified with > ±10% change for KIM-1, TNFR-1, TNFR-2 (Total Protein, Rheumatoid Factor).Worst-case scenario interference (+/- 10%) used for simulations. Specific limitations added regarding high RF or total protein. Criteria: >=5% of patient simulations not matching original category considered affected.
    Assay Reportable Range: Established measuring intervals for KIM-1, TNFR-1, TNFR-2.KIM-1: 12 pg/mL to 3,915 pg/mL TNFR-1: 1,057 pg/mL to 14,322 pg/mL TNFR-2: 4,270 pg/mL to 43,291 pg/mLData supports these intervals considering LoB/LoD/LoQ, linearity, and precision.
    Traceability: KIM-1, TNFR-1, TNFR-2 methodologies.Traceable to internal standards.
    Stability: KIM-1, TNFR-1, TNFR-2 in samples for shipping.Stable for 3 days at up to 37℃.Consistent with handling instructions for shipping to the lab.
    Detection Limit (LoB, LoD, LoQ): For KIM-1, TNFR-1, TNFR-2.LoB, LoD determined per CLSI EP17-A2. LoQ defined as lowest concentration with total within-laboratory precision of %CV<20%. Specific values are listed but not explicitly included in the table above as they are part of meeting the reportable range and precision.
    Hook Effect: No hook effect for individual components.No hook effect expected at tested levels.
    Subgroup Performance: Performance similar across major subgroups (e.g., eGFR, race).Demonstrated similar performance across subgroups for progressive decline in kidney function (e.g., eGFR 30-59.9 vs >60, Black/African American vs Non-Black/African American).Shown via Kaplan-Meier curves and estimated 5-year rates.
    Lack of Diagnosis/Staging: Not intended for diagnosing or staging DKD.Labeling explicitly states this.Part of the "Special Conditions for Use Statement(s)".
    Rx - For Prescription Use Only: Restriction on use.Labeling explicitly states this.Part of the "Special Conditions for Use Statement(s)".

    Study Information:

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

    2. Sample sized used for the test set and the data provenance:

      • Sample Size: N = 657 patients.
      • Data Provenance: Retrospective analysis of biobanked samples. Sourced from a single biorepository site that collected from 10 clinical sites in the United States (implied by references to US population and demographics like "American Indian/Alaskan Native", "Black/African American", "White", "Hispanic"). The study included up to 5 years of data per sample.

    3. Number of experts used to establish the ground truth for the test set and the qualifications of those experts:

      • The document does not specify the number or qualifications of experts used to establish the ground truth for the test set. The ground truth ("progressive decline in kidney function" and "end-stage kidney disease") is defined by clinical outcomes (e.g., sustained decrease in eGFR, eGFR <15 ml/min/1.73 m2), rather than expert interpretation of images or other subjective data.

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

      • The document does not describe an adjudication method. Since the ground truth is defined by objective clinical measurements (eGFR decline, ESKD), there would typically be no need for expert adjudication.

    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 done. This device is an in-vitro diagnostic test generating a risk score, not an AI-assisted diagnostic imaging tool that would typically involve human readers.

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

      • Yes, a standalone performance study was done. The clinical study evaluates the KidneyIntelX.dkd algorithm's ability to predict outcomes in a cohort of patients without human intervention in its scoring process. The algorithm generates a risk level (Low, Moderate, High) which is then correlated with observed clinical outcomes. The outputs are intended to be used in conjunction with other clinical findings, but the generation of the score itself is standalone.

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

      • Outcomes data. The ground truth was defined by time to the first occurrence of a composite endpoint:
        • Sustained decrease in eGFR greater than or equal to 40% lasting more than 3 months.
        • Reaching end-stage kidney disease (sustained eGFR <15 ml/min/1.73 m2).

    8. The sample size for the training set:

      • The document does not explicitly state the sample size for the training set. It mentions that the "KidneyIntelX.dkd algorithm was modeled using machine learning (Random Forest) to identify and analyze risk factors for kidney disease progression," but details on the training set used for this modeling are absent from the provided text. The N=657 is described as the "clinical validation" set.

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

      • The document does not describe how the ground truth for the training set was established. Given the nature of the clinical validation ground truth (outcomes data), it is highly probable that similar outcomes-based ground truth was used for training, but this is not explicitly stated.
    Ask a Question

    Ask a specific question about this device

    Page 1 of 1