Search Results

Creatine Kinase is an in vitro test for the quantitative determination of creatine kinase (CK) in human serum and plasma on Roche/Hitachi cobas c systems. The determination of CK and CK isoenzyme activities is utilized in the diagnosis and monitoring of myocardial infarction and myopathies such as the progressive Duchenne muscular dystrophy.

Measurements of creatine phosphokinase and its isoenzymes are used in the diagnosis and treatment of myocardial infarction and muscle diseases such as progressive, Duchenne-type muscular dystrophy.

Device Description

The Creatine Kinase assay is a two reagent assay for the quantitative determination of creatine kinase (CK) in human serum and plasma on automated clinical chemistry analyzers. Photometrically measured NAPDP formation is directly proportional to CK activity in a human sample.

AI/ML Overview

Here's a breakdown of the acceptance criteria and study information based on the provided document:

Creatine Kinase Device Performance Study Summary

1. Acceptance Criteria and Reported Device Performance

Feature / Metric	Acceptance Criteria (Claim in Document)	Reported Device Performance
Limit of Blank (LoB)	Claim: 7 U/L	0.3 U/L
Limit of Detection (LoD)	Claim: 7 U/L	1.0 U/L
Limit of Quantitation (LoQ)	Claim: 7 U/L	3.3 U/L
Precision (Repeatability)	Achieved with Human Serum samples and PreciControl ClinChem Multi controls. Specific acceptance limits for CV/SD are not explicitly stated as "acceptance criteria" but implied by the presentation of successful results. For example: Human Serum 1 (18.7 U/L) CV <= 3.0%, Human Serum 3 (477 U/L) CV <= 0.6%, PreciControl ClinChem Multi 2 (301 U/L) CV <= 0.4%.	Human Serum 1: Mean 18.7 U/L, SD 0.6 U/L, CV 3.0%Human Serum 2: Mean 137 U/L, SD 0.8 U/L, CV 0.6%Human Serum 3: Mean 477 U/L, SD 3.0 U/L, CV 0.6%Human Serum 4: Mean 946 U/L, SD 5.3 U/L, CV 0.6%Human Serum 5: Mean 1816 U/L, SD 9.4 U/L, CV 0.5%PreciControl ClinChem Multi 1: Mean 154 U/L, SD 0.9 U/L, CV 0.6%PreciControl ClinChem Multi 2: Mean 301 U/L, SD 1.3 U/L, CV 0.4%
Precision (Intermediate)	Similar to repeatability, specific acceptance limits for CV/SD are not explicitly stated, but implied by successful results. For example: Human Serum 1 (18.7 U/L) CV <= 3.2%, Human Serum 3 (477 U/L) CV <= 0.6%, PreciControl ClinChem Multi 2 (301 U/L) CV <= 0.9%.	Human Serum 1: Mean 18.7 U/L, SD 0.6 U/L, CV 3.2%Human Serum 2: Mean 137 U/L, SD 1.1 U/L, CV 0.8%Human Serum 3: Mean 477 U/L, SD 3.1 U/L, CV 0.6%Human Serum 4: Mean 946 U/L, SD 5.8 U/L, CV 0.6%Human Serum 5: Mean 1816 U/L, SD 10 U/L, CV 0.6%PreciControl ClinChem Multi 1: Mean 154 U/L, SD 1.7 U/L, CV 1.1%PreciControl ClinChem Multi 2: Mean 301 U/L, SD 2.6 U/L, CV 0.9%
Linearity (Serum)	Pearson correlation coefficient (R) of 0.9999	y=1.001x-0.646, Pearson correlation coefficient (R)=0.9999
Linearity (Plasma)	Pearson correlation coefficient (R) of 0.9999	y=1.002x-1.205, Pearson correlation coefficient (R)=0.9999
Matrix Comparison (Anticoags)	Good correlation (r >= 0.998) and slope close to 1, intercept close to 0 implied for substantial equivalence.	Serum vs. Serum Gel Separation: y = 0.998x + 0.010, r = 0.999Serum vs. Li-heparin: y = 1.00x - 1.994, r = 0.998Serum vs. K2-EDTA: y = 0.993x - 2.016, r = 0.998Serum vs. K3-EDTA: y = 0.981x - 2.671, r = 0.999
Interference (Hemolysis)	No significant interference up to an H index of 100 (approx. 62.1 µmol/L or 100 mg/dL).	No interference up to Level 1: 103 H Index, Level 2: 130 H Index.
Interference (Lipemia)	No significant interference up to an L index of 1000.	No interference up to Level 1: 1356 L Index, Level 2: 1143 L Index.
Interference (Unconjugated Bilirubin)	No significant interference up to an I index of 60 for conjugated and unconjugated bilirubin (approx. 1026 µmol/L or 60 mg/dL).	No interference up to Level 1: 67 I Index, Level 2: 67 I Index.
Interference (Conjugated Bilirubin)	No significant interference up to an I index of 60 for conjugated and unconjugated bilirubin (approx. 1026 µmol/L or 60 mg/dL).	No interference up to Level 1: 68 I Index, Level 2: 76 I Index.
Interference (Drugs)	No interference at therapeutic concentrations using common drug panels. (Exception noted for Cyanokit).	No interference found at therapeutic concentrations using common drug panels. (Cyanokit (Hydroxocobalamin) at therapeutic concentrations interferes with the test).
Method Comparison (Predicate)	The objective is for results to demonstrate substantial equivalence to the predicate device (COBAS INTEGRA Creatine Kinase, K951595) via Passing Bablok Regression where the slope is close to 1 and the intercept close to 0, and a high correlation coefficient (r).	y = 1.021x + 5.88 U/L, r = 0.999

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

Detection Limits (LoB, LoD):
- LoB: One analyte-free sample, measured with three lots, 10-fold determination in 6 runs (total 60 measurements per lot).
- LoD: Five low-analyte concentration samples, measured with three lots, two-fold determination in 6 runs (total 60 measurements per lot).
Limit of Quantitation (LoQ):
- Five human serum samples, tested in 5 replicates per sample on 5 days.
Precision (Repeatability & Intermediate):
- Not explicitly stated as a single number, but experiments conducted with multiple human serum samples (5) and control materials (2), with two aliquots per run, two runs per day for ≥ 21 days on the same analyzer using 3 lots of reagent.
Linearity (Serum & Plasma):
- Two separate dilution series (serum and plasma), each with 14 concentrations, measured in triplicate.
Matrix Comparison - Anticoagulants:
- For each of the four tube types (Serum Gel Separation, Li-heparin, K2-EDTA, K3-EDTA), 30 tubes were filled completely. This implies 30 samples for each comparison.
Interferences - H, L, I Indices:
- Pooled human serum samples spiked with varying levels of interferent. The resulting sample series (10 dilution steps per sample) were tested in triplicate.
Interferences - Drugs:
- Two sample pools (low and high CK concentrations), divided into aliquots. Each spiked aliquot measured in triplicate.
Method Comparison to Predicate:
- 132 human serum samples (9 of which were spiked with human recombinant CK MB). Samples tested in singlicate.

Data Provenance: The document explicitly mentions "human serum samples" and "pooled human serum samples." The manufacturer is Roche Diagnostics Operations, located in Indianapolis, IN, USA. Based on the context of FDA submission for a device to be marketed in the USA, it is highly probable the data is retrospective and likely collected in a clinical laboratory setting (possibly in the USA or aligned with international standards like CLSI), but the specific country of origin for the patient samples is not stated.

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

N/A. This is a submission for an in vitro diagnostic device that quantitatively measures a biomarker (Creatine Kinase). The "ground truth" for the test set is established by the reference measurement procedure or the known concentration of the analyte in calibrators/controls, not by expert interpretation of images or clinical data. Therefore, human experts for ground truth adjudication are not applicable in this context.

4. Adjudication Method for the Test Set

N/A. See explanation above. The measurements are quantitative chemical analyses against instrument-derived values and established reference methods/materials, not subjective interpretations.

5. Multi-Reader Multi-Case (MRMC) Comparative Effectiveness Study

No. This is an in vitro diagnostic device for quantitative measurement of a biomarker in biological samples. MRMC studies are typically performed for imaging devices or other diagnostic tools where human interpretation plays a significant role, to evaluate how AI assistance impacts human reader performance. This is not applicable here.

6. Standalone (Algorithm Only) Performance Study

Yes, the entire submission is a standalone performance study of the Creatine Kinase assay on the Roche/Hitachi cobas c systems. The device itself is an "algorithm only" in the sense that it is an automated analytical system (reagent + instrument) that produces quantitative results without human intervention in the measurement process itself, beyond sample loading and system maintenance. The studies described (Detection Limits, Precision, Linearity, Interference, Method Comparison) directly assess the performance of this standalone analytical system.

7. Type of Ground Truth Used

The ground truth for the performance studies relies on:

Reference Materials/Methods: For calibration and verification of accuracy, the method is stated to be traceable to IFCC (International Federation of Clinical Chemistry and Laboratory Medicine) Method for Creatine Kinase.
Known Concentrations: For studies like LoB, LoD, LoQ, linearity, and interference, samples are either analyte-free, spiked with known concentrations of analyte or interferents, or diluted from samples with established reference values. This essentially uses calibrated reference standards and materials as the ground truth.
Predicate Device Measurements: For method comparison, the predicate device's results (COBAS INTEGRA Creatine Kinase cleared in K951595 on the COBAS INTEGRA analyzer) serve as a comparative ground truth to establish substantial equivalence.

8. Sample Size for the Training Set

N/A. This is not a machine learning or AI-based diagnostic device in the modern sense that requires a "training set" of data to learn from. It is a traditional in vitro diagnostic assay based on a defined enzymatic chemical reaction and photometric measurement. The development of such assays involves extensive research and development to optimize reagents and instrument parameters, but not in the framework of a "training set" as understood in AI/ML.

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

N/A. As explained above, there is no "training set" in the context of an AI/ML device for this product. The development of the assay involves chemical and enzymatic principles, optimized through laboratory experimentation and knowledge of biochemistry, rather than learning from a dataset via a specified ground truth.

Ask a Question

Ask a specific question about this device

K Number

K143691

Device Name

LDLC3 LDL-Cholesterol Gen.3

Manufacturer

Roche Diagnostics Operations (RDO)

Date Cleared

2015-01-28

(35 days)

Product Code

Regulation Number

Type

Panel

Reference & Predicate Devices

k012287,k011658,k102016,k974733

Predicate For

N/A

Intended Use

The LDL-Cholesterol Gen. 3 assay is an in-vitro test for the quantitative determination of LDL-cholesterol in human serum and plasma on Roche/Hitachi cobas c systems. Lipoprotein measurements are used in the diagnosis and treatment of lipid disorders (such as diabetes mellitus), atherosclerosis, and various liver and renal diseases.

Device Description

The LDL-Cholesterol Gen. 3 assay is a homogeneous enzyme colorimetric assay which provides the quantitative measurement of LDL-cholesterol in human serum and plasma. Reagents are packaged in a cassette labeled with their instrument positioning R1 (Reagent 1) and R2 (Reagent 2).

R1 contains Bis-trisb) buffer: 20.1 mmol/L, pH 7.0; 4-aminoantipyrine:0.98 mmol/L; ascorbic oxidase (AOD, Acremonium spec.): ≥ 66.7 µkat/L; peroxidase (recombinant from Basidiomycetes): ≥ 166.7 µkat/L; BSA: 4.0 g/L; preservative R2 contains MOPSC) buffer: 20.1 mmol/L, pH 7.0; EMSE: 2.16 mmol/L, cholesterol esterase (Pseudomonas spec.): ≥ 33.3 µkat/L; cholesterol oxidase (recombinant from E.coli)): ≥ 31.7 µkat/L; peroxidase (recombinant from Basidiomycetes): ≥ 333.3 µkat/L; BSA: 4.0 g/L; detergents; preservative

AI/ML Overview

Here's a breakdown of the acceptance criteria and study information for the LDLC3 LDL-Cholesterol Gen.3 device, organized as requested:

Acceptance Criteria and Device Performance Summary

Performance Metric	Acceptance Criteria	Reported Device Performance
Limit of Blank (LoB)	Claim: 3.87 mg/dL	Result: 0.406 mg/dL (Meets acceptance criteria as it's below the claim)
Limit of Detection (LoD)	Claim: 3.87 mg/dL	Result: 0.99 mg/dL (Meets acceptance criteria as it's below the claim)
Limit of Quantitation (LoQ)	Claim: 3.87 mg/dL	Result: 2.28 mg/dL (Meets acceptance criteria as it's below the claim)
Drug Interference	Difference in recovery to the reference sample: ≤ ± 10%	All data passed the acceptance criteria for various common drugs, Simvastatin, Bezafibrate, and Nicotinic Acid. Specific highest concentrations shown not to interfere were reported for each drug (e.g., Acetylcysteine: 553 mg/L, Simvastatin: 16 mg/L).
Interference from VLDL, HDL, Chylomicrons	≤ ± 10% in recovery for VLDL-Cholesterol: ≤ 140 mg/dL, HDL-Cholesterol: ≤ 75 mg/dL, Chylomicrons: ≤ 2000 mg/dL triglycerides	All data passed the acceptance criteria for VLDL, HDL, and Chylomicrons within their specified concentration limits. The testing methodology confirmed the device's ability to selectively measure LDL-cholesterol.
Endogenous Substances Interference	≤ 10%	No significant interference was observed up to a Lipemia L index of 1000, Hemolysis H index of 1000, and Bilirubin I index of 60 (both conjugated and unconjugated). All data passed the ≤ 10% acceptance criteria.
Matrix Comparison	Comparisons with plasma vs. serum passed specification (details on specific regression equations and correlation coefficients are provided in the document).	Serum vs. Gel Separation P/B: y = 1.004x + 0.091, r = 1.000; Serum vs. Li-heparin P/B: y = 0.99x - 1.50, r = 0.999; Serum vs. K2-EDTA P/B: y = 0.98x - 0.248, r = 1.000; Serum vs. K3-EDTA P/B: y = 0.95x - 0.246, r = 0.999. All passed specification.
Linearity	3.87 mg/dL - 549 mg/dL: ≤ ± 10%	For both plasma and serum: Range tested: Plasma 3.66 - 584 mg/dL, Serum 3.53 - 565 mg/dL. Range found: Plasma 3.66 - 584 mg/dL, Serum 3.53 - 565 mg/dL. Recommended measuring range: 3.87 - 549 mg/dL. Linear regression equations and r-squared values indicate good linearity (e.g., Plasma: y = 1x + 0, r2 = 0.9995). Data passed the ≤ ± 10% acceptance criteria within the recommended range.
Precision	Not explicitly stated as a single acceptance criterion value in the provided text, but the data indicates typical precision study results, which are generally evaluated based on CV% limits for various concentrations. The reported CVs for both repeatability and intermediate precision are low (mostly <3%), indicating good precision.	Precinorm L: Repeatability CV 0.7%, Intermediate Precision CV 2.3%; Precipath HDL/LDL-C: Repeatability CV 0.7%, Intermediate Precision CV 2.1%; Human Serum pools (various concentrations): Repeatability CVs 0.7% - 1.2%, Intermediate Precision CVs 1.9% - 2.5%.
Method Comparison	Not explicitly stated as a single acceptance criterion in the provided text, but determined by the correlation and agreement with the predicate device.	Passing/Bablok regression: y = 0.984x = 0.786 mg/dL, r = 0.999. This indicates very strong correlation and close agreement with the predicate device.

Study Details

This submission describes the performance evaluation of the LDLC3 LDL-Cholesterol Gen.3 assay.

Sample sizes used for the test set and the data provenance:
- LoB Protocol: N=60 determinations (analyte-free sample, tested in five-fold, on two analyzers, over three days).
- LoD Protocol: Five low-analyte samples, measured in singlicate on two analyzers over three days.
- LoQ Protocol: Five low-level samples, tested in five replicates per sample on five days with three reagent lots, one run per day on one analyzer.
- Precision: 5 human serum sample pools and two control samples (4 aliquots per run, 1 run per day, 21 days).
- Drug Interferences: Two sample pools (low and high LDL concentration) for each drug, tested in triplicate. Total number of samples not explicitly stated for all drugs combined, but involves at least 2 pools x 3 replicates x 19 drugs = 114 measurements.
- Interference from VLDL, HDL, Chylomicrons: HDL: Two sample pools (low and high LDL), each divided into two aliquots, tested in triplicate. VLDL: Two sample pools (low and high LDL), spiked with increasing VLDL, measured in duplicate. Chylomicrons: Four sample pools (various chylomicron concentrations, including some ≥ 2000 mg/dL triglycerides), spiked with chylomicrons, measured in duplicate.
- Endogenous Interferences: Two human serum pools (one spiked, one unspiked) mixed in 10 dilution steps; samples tested in triplicate.
- Matrix Comparison: 59 samples for Gel Separation, 59 for Li-heparin, 57 for K2-EDTA, 59 for K3-EDTA. Samples ranged from 12.3 to 495 mg/dL. (Retrospective/Prospective not explicitly stated, but typically clinical lab studies like these are prospective collections or de-identified banked samples).
- Linearity: Two dilution series (serum and plasma), each with 14 concentrations, measured in triplicate.
- Method Comparison: 100 human serum samples (including 5 spiked and 2 diluted).
- Data Provenance: The document does not explicitly state the country of origin for the clinical samples. It refers to "human serum samples" and "human serum and plasma." This type of in-vitro diagnostic device testing would typically use samples collected under IRB-approved protocols, but the specific origins are not detailed. The studies are described in a manner consistent with prospective analytical validation.
Number of experts used to establish the ground truth for the test set and the qualifications of those experts:
- This is an in-vitro diagnostic device (IVD) for quantitative determination of a biomarker (LDL-cholesterol). The "ground truth" for calibrating and evaluating such devices is typically established through reference methods and certified reference materials, not through expert human interpretation of images or clinical cases. The device is being compared against a predicate device (LDL-Cholesterol plus 2nd generation).
- The document states: "This method has been standardized against the beta quantification method as defined in the recommendations in the LDL Cholesterol Method Certification Protocol for Manufacturers." This beta quantification method serves as the analytical ground truth/reference method for LDL-C measurement.
Adjudication method (e.g., 2+1, 3+1, none) for the test set:
- This concept of "adjudication" is not applicable to quantitative IVD studies. Adjudication methods (like 2+1, 3+1) are used in diagnostic imaging studies where multiple readers interpret cases, and discrepancies are resolved by a super-reader or consensus. For this device, direct quantitative measurements are being performed and compared to either a reference method or a predicate device.
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 a standalone quantitative assay for measuring LDL-cholesterol, not an AI-powered diagnostic imaging tool that assists human readers. Therefore, the concept of human reader improvement with/without AI assistance is not relevant here.
If a standalone (i.e., algorithm only without human-in-the-loop performance) was done:
- Yes, the performance studies described are for the device (assay on the cobas c system) operating in a standalone manner. This is a fully automated quantitative measurement system, where the result is generated directly by the instrument, without human-in-the-loop interpretative steps that would influence the quantitative reading.
The type of ground truth used (expert consensus, pathology, outcomes data, etc.):
- The primary ground truth for the analytical validation is the beta quantification method as defined in the LDL Cholesterol Method Certification Protocol for Manufacturers. This is an accepted standardized reference method for truly measuring LDL-cholesterol.
- For the method comparison study, the predicate device (LDL-Cholesterol plus 2nd generation reagent) served as the comparative reference.
The sample size for the training set:
- This document describes an analytical validation for a quantitative in-vitro diagnostic reagent. There is no "training set" in the context of machine learning for this type of device. The assay relies on established chemical and enzymatic reactions, not machine learning algorithms trained on large datasets. The formulation of the reagent itself is based on chemical and biological principles.
How the ground truth for the training set was established:
- As explained above, there is no "training set" for this type of device. Therefore, the concept of establishing ground truth for a training set is not applicable. The assay's "truth" is rooted in its chemical-enzymatic mechanism and its standardization against a reference method (beta quantification).

Ask a Question

Ask a specific question about this device

K Number

K131544

Device Name

COBAS C BILIRUBIN TOTAL GEN.3

Manufacturer

Roche Diagnostics

Date Cleared

2013-07-17

(49 days)

Product Code

Regulation Number

Type

Panel

Reference & Predicate Devices

k981632,k101456,k042389,k102016,k063543

Predicate For

N/A

Intended Use

cobas c Bilirubin Total Gen.3 is an in vitro test for the quantitative determination of total bilirubin in serum and plasma of adults and neonates on Roche/Hitachi cobas c systems. Measurement of the levels of bilirubin, an organic compound formed during the normal and abnormal destruction of red blood cells, is used in the diagnosis and treatment of liver, hemolytic, hematological, and metabolic disorders, including hepatitis and gall bladder block.

Device Description

cobas c Bilirubin Total Gen.3 reagent provides quantitative measurement of the total bilirubin that is present in serum and plasma of adults and neonates. Reagents are packaged in a cassette with two bottles labeled with their instrument positioning, R1 (Reagent 1) and R2 (Reagent 2). R1 contains detergent, buffer, and stabilizers at pH 1.0. R2 is a 3,5-dichlorophenyl diazonium salt: ≥ 1.35 mmol/L.

AI/ML Overview

The provided text describes the 510(k) summary for the cobas c Bilirubin Total Gen.3 device, a quantitative colorimetric method for determining total bilirubin in serum and plasma. The acceptance criteria and supporting studies are detailed for various performance characteristics.

Here's a breakdown of the requested information:

1. Table of Acceptance Criteria and Reported Device Performance

Performance Characteristic	Acceptance Criteria	Reported Device Performance
Precision	Not explicitly stated as a single criterion, but implied by the detailed imprecision (SD & CV%) requirements generally found in CLSI EP5-A2 studies.	Repeatability (Within Run Imprecision): - PCCC1: 0.02 mg/dL (2.1% CV) - PCCC2: 0.02 mg/dL (0.6% CV) - Human Serum 1 (0.51 mg/dL): 0.01 mg/dL (2.9% CV) - Human Serum 2 (17.7 mg/dL): 0.10 mg/dL (0.6% CV) - Human Serum 3 (31.8 mg/dL): 0.14 mg/dL (0.4% CV) Intermediate Precision (Total Imprecision): - PCCC1: 0.02 mg/dL (2.1% CV) - PCCC2: 0.03 mg/dL (0.8% CV) - Human Serum 1 (0.51 mg/dL): 0.02 mg/dL (3.3% CV) - Human Serum 2 (17.7 mg/dL): 0.14 mg/dL (0.8% CV) - Human Serum 3 (31.8 mg/dL): 0.18 mg/dL (0.6% CV)
Linearity/Measuring Range	For both serum and plasma, the first-order (linear) regression must be significant.	Serum: Range tested and found: 0.12-38.9 mg/dL. Recommended measuring range: 0.15-35.1 mg/dL. Linear Regression: y=1.0021x-0.0317, r² = 0.999881 (Significant). Plasma: Range tested and found: 0.12-39.0 mg/dL. Recommended measuring range: 0.15-35.1 mg/dL. Linear Regression: y = 1.0014x - 0.0232, r² = 0.999954 (Significant).
Detection Limit (LoB, LoD, LoQ)	Not explicitly stated in terms of acceptance criteria values, but the reported claims represent the specifications. The LoQ is determined based on precision at 20% CV.	LoB claim: 0.10 mg/dL LoD claim: 0.15 mg/dL LoQ claim: 0.15 mg/dL
Analytical Specificity (Endogenous Substances)	Lipemia: ≤± 0.10 mg/dL for samples ≤ 1 mg/dL or ≤± 10% for samples > 1 mg/dL Hemolysis HbA: ≤±0.20 mg/dL for samples ≤ 2 mg/dL or ≤± 10% for samples > 2 mg/dL Hemolysis HbF: ≤± 0.10 mg/dL for samples ≤ 1 mg/dL or ≤ ± 10% for samples > 1 mg/dL Indican: ≤± 0.10 mg/dL for samples ≤ 1 mg/dL or ≤± 10% for samples > 1 mg/dL	Lipemia: No significant interference up to an L index of 1000. (Tested up to 1196-1217 L index) Hemolysis HbA: No significant interference up to an H index of 800. (Tested up to 946-951 H index) Hemolysis HbF: No significant interference up to an H index of 1000. (Tested up to 1047-1053 H index) Indican: No significant interference from indican up to 3 mg/dL. (Tested up to 3.75 mg/dL)
Analytical Specificity (Common Drugs)	Difference in recovery to the reference sample: ≤± 10%	All tested drugs (Acetylcystein, Ampicillin - Na, Ascorbic acid, Phenylbutazone, Cyclosporine A, Cefoxitin, Levodopa, Methyldopa + 1.5, Metronidazole, Doxycyclin, Acetylsalycilic acid, Rifampicin, Acetaminophen, Ibubrofen, Theophylline) passed the acceptance criteria at their respective highest concentrations.
Matrix Comparison (Anticoagulants)	For sample concentrations ≤ 0.99 mg/dL, the deviation must be ≤ ± 0.10 mg/dL. For sample concentrations > 0.99 mg/dL, the deviation must be ≤± 10%.	All data passed the criteria. - Li-Heparin (full & half), K2-EDTA (full & half), and Gel Separation Tube showed acceptable recovery within the tested ranges (e.g., Li-Heparin full: 0.35 - 34.52 mg/dL). - Serum vs. Li-heparin: y = 1.000x + 0.000, r = 0.9998
Adult Method Comparison with Predicate Device	Not explicitly stated with a numerical criterion, but the strong correlation (r=0.9997) and the regression equation (y = 0.959x + 0.091 mg/dL) demonstrate substantial agreement.	Equation: y = 0.959x + 0.091 mg/dL Correlation coefficient: r = 0.9997

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

Precision:
- Human Sera Samples: 3 samples (0.51, 17.7, and 31.8 mg/dL)
- Control Samples: 2 serum-based control samples (PCCC1, PCCC2)
- Each sample/control run in two aliquots per run, two runs per day for 21 days.
- Data Provenance: Not explicitly stated, but implied to be laboratory-generated (not from real patient populations with specific countries of origin). Retrospective or prospective is not specified, but the study design suggests prospective lab testing.
Linearity/Assay Reportable Range:
- Serum dilution series: 14 levels
- Plasma dilution series: 13 levels
- Data Provenance: Laboratory-generated, with human serum/plasma pool spiked with unconjugated bilirubin. Not specified for country of origin or retrospective/prospective.
Detection Limits (LoB, LoD, LoQ):
- LoB: One blank sample
- LoD: Five low-analyte samples
- LoQ: A low-level sample set of nine
- Data Provenance: Laboratory-generated.
Analytical Specificity (Endogenous Substances):
- Interferents: Hemoglobin, lipids, indican.
- Two pools of human serum used (one spiked, one unspiked) to create dilution series.
- Interference tested at two levels of bilirubin.
- Data Provenance: Laboratory-generated using human serum.
Analytical Specificity (Common Drugs):
- 15 commonly used drugs.
- Serum sample pools at two target concentrations of total bilirubin (~1.0 mg/dL and ~14.0 mg/dL).
- Data Provenance: Laboratory-generated using serum.
Adult Method Comparison with Predicate Device:
- Sample Size: n=131 human sera adult samples.
- Data Provenance: Not explicitly stated for country of origin or retrospective/prospective, but implies de-identified human serum samples.
Matrix Comparison (Anticoagulants):
- Sample Size: 35 tubes collected per anticoagulant type (Li-heparin, K2-EDTA, Gel Separation Tube).
- Data Provenance: Not explicitly stated for country of origin or retrospective/prospective, but implies human plasma/serum samples.

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

This device is an in vitro diagnostic (IVD) for quantitative measurement of total bilirubin. Expert consensus is not typically used to establish ground truth for this type of quantitative biochemical assay. The ground truth is generally established by:

Reference Methods: For this device, the "ground truth" or reference method for traceability is explicitly stated as "Standardized against the Doumas manual reference method."
Predicate Device: For method comparison, the predicate device (Total Bilirubin reagent on the cobas c 501) serves as the comparator.

Therefore, the concept of "experts" in the context of clinical interpretation for ground truth is not applicable here.

4. Adjudication Method for the Test Set

Adjudication methods (e.g., 2+1, 3+1) are typically used in studies where human readers provide subjective assessments (e.g., image interpretation). This is a quantitative chemical assay, where measurements are objective. Therefore, no adjudication method was used or is relevant.

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 MRMC comparative effectiveness study was done. This device is a fully automated in vitro diagnostic test for measuring bilirubin levels. It does not involve human readers for interpretation, nor does it incorporate AI (Artificial Intelligence) in a way that would assist human readers.

6. If a Standalone (i.e., Algorithm Only Without Human-in-the-Loop Performance) Was Done

Yes, a standalone study was done. The entire performance evaluation (precision, linearity, detection limits, interference, method comparison) described in the document is for the device operating as a standalone quantitative assay without human intervention in the measurement process. The "algorithm" here refers to the chemical reaction principles and photometric measurement methodology.

7. The Type of Ground Truth Used (Expert Consensus, Pathology, Outcomes Data, Etc.)

The ground truth for the quantitative measurement of total bilirubin is generally established by:

Reference Methods: The device is standardized against the Doumas manual reference method (as stated under "Traceability"). This is the gold standard for bilirubin measurement.
Comparator Methods: In the adult method comparison study, the predicate device (Total Bilirubin reagent) values served as the comparator for assessing agreement.

8. The Sample Size for the Training Set

The provided document describes a 510(k) submission for a diagnostic test. Unlike AI/ML-based diagnostic devices, this type of device does not typically involve "training sets" in the machine learning sense. The "training" in developing such a device involves refining chemical reagents and optimizing instrument parameters, which is a different process than training an algorithm on a dataset. The studies described are performance validation studies.

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

As explained above, there isn't a "training set" in the context of an AI/ML algorithm for this type of IVD device. The development process would involve optimizing the reagent formulation and assay conditions against an established reference method (like the Doumas method) to ensure accurate and precise measurements.

Ask a Question

Ask a specific question about this device

K Number

K123965

Device Name

COBAS INTEGRA BILIRUBIN DIRECT GEN.2

Manufacturer

Roche Diagnostics

Date Cleared

2013-02-28

(64 days)

Product Code

Regulation Number

Type

Panel

Reference & Predicate Devices

k102016,k063543

Predicate For

N/A

Intended Use

COBAS INTEGRA Bilirubin Direct Gen.2 is an in vitro test for the quantitative determination of direct bilirubin in human serum and plasma on COBAS INTEGRA systems. Measurement of the levels of bilirubin, an organic compound formed during the normal and abnormal destruction of red blood cells, is used in the diagnosis and treatment of liver, hemolytic, hematological, and metabolic disorders, including hepatitis and gall bladder block.

Device Description

COBAS INTEGRA Bilirubin Direct Gen.2 reagent provides quantitative measurement of the direct bilirubin that is present in a human serum or human plasma sample. Reagents are packaged in a cassette with two bottles labeled with their instrument positioning, R1 and SR. R1, or Reagent 1, contains Phosphoric acid 85 mmol/L, NaCl 50 mmol/L, and HEDTA 4.0 mmol/L at pH 1.9. SR, or Start Reagent, is a 3,5-dichlorophenyl diazonium salt at 1.5 mmol/L in acid buffer, pH 1.3.

AI/ML Overview

Here's a summary of the acceptance criteria and study information for the COBAS INTEGRA Bilirubin Direct Gen.2 reagent, based on the provided text:

Acceptance Criteria and Device Performance

Acceptance Criteria Category	Specific Acceptance Criteria	Reported Device Performance (COBAS INTEGRA Bilirubin Direct Gen.2)
Precision/Reproducibility	Based on CLSI EP5-A2 guidelines.	Repeatability:- Human Serum 1 (0.12 mg/dL): SD 0.01 mg/dL, CV 7.4%- Human Serum 2 (3.8 mg/dL): SD 0.01 mg/dL, CV 0.4%- Human Serum 3 (13.2 mg/dL): SD 0.04 mg/dL, CV 0.3%Intermediate Precision:- Human Serum 1 (0.12 mg/dL): SD 0.01 mg/dL, CV 7.7%- Human Serum 2 (3.8 mg/dL): SD 0.04 mg/dL, CV 1.0%- Human Serum 3 (13.2 mg/dL): SD 0.05 mg/dL, CV 0.4%
Measuring Range (Linearity)	Based on CLSI EP6-A guidelines.	Plasma: Range tested 0.01 - 19.5 mg/dL, Range found 0.01 - 19.5 mg/dL, Recommended measuring range 0.07 - 13.8 mg/dLSerum: Range tested 0.02 - 19.4 mg/dL, Range found 0.02 - 17.4 mg/dL, Recommended measuring range 0.07 - 13.8 mg/dLBoth showed a significant quadratic model, with linear regression for serum y = 1.0000x + 0.0000 (r² = 0.9944) and for plasma y = 1.0000x - 0.0000 (r² = 0.9977).
Detection Limits (LoB, LoD, LoQ)	Based on CLSI EP17-A2 guidelines.	LoB claim: 0.05 mg/dLLoD claim: 0.07 mg/dLLoQ claim: 0.07 mg/dL (based on 20% CV)
Analytical Specificity (Endogenous Substances)	"No significant interference"	Lipemia: No significant interference up to an L index of 750 (reported lowest L index for no interference was 1098).Hemolysis: No significant interference up to an H index of 25 (reported lowest H index for no interference was 25).
Analytical Specificity (Common Drugs)	"No interference" from specific drugs.	Phenylbutazone causes falsely low bilirubin results (stated in labeling). The other 17 tested drugs (e.g., Acetylcystein (150 mg/L), Ampicillin - Na (1000 mg/L), Ascorbic acid (300 mg/L), Heparin - Na (5000 U)) produce no interference.
Method Comparison with Predicate Device	Substantial equivalence to predicate device (COBAS INTEGRA Bilirubin Direct).	Passing/Bablok regression with predicate device showed: y = 1.0490x + 0.0699 mg/dL with R² = 0.9979.
Matrix Comparison (Anticoagulants)	Median recovery: 90 to 110%Median absolute deviation: < 0.20 mg/dL	Li-Heparin: 102% recovery (+0.02 to -0.05 mg/dL MAD)K2-EDTA: 101% recovery (+0.00 to -0.02 mg/dL MAD)K3-EDTA: 100% recovery (-0.00 to -0.03 mg/dL MAD)Gel Separation Tube: 104% recovery (+0.02 mg/dL MAD)Regression analysis: Serum vs. Li-heparin (y = 0.01 + 1.0179x, r = 0.9988), Serum vs. K2-EDTA (y = -0.01 + 1.0120x, r = 0.9988), Serum vs. K3-EDTA (y = -0.03 + 1.0095x, r = 0.9988).
Expected Values/Reference Range	To be established or confirmed.	≤ 0.20 mg/dL

Study Information

Sample size used for the test set and the data provenance:
- Precision/Reproducibility: Not explicitly stated, but samples were "human sera samples (0.12, 3.76, and 13.2 mg/dL)" and "two serum-based control samples." Samples were tested for 21 days with 2 aliquots per run, 2 runs per day.
- Linearity: Not explicitly stated, but "two separate dilution series differing by sample type (serum and plasma) were prepared with thirteen levels each." "High analyte native samples" were spiked with ditaurobilirubin.
- Detection Limits (LoB): One blank sample (n=5) tested on two analyzers with three reagent batches for two runs per day across three days.
- Detection Limits (LoD): Five low-analyte samples measured in singlicate on two analyzers with three reagent batches for two runs per day across three days.
- Detection Limits (LoQ): A low-level sample set of nine measured in singlicate, using three reagent batches on two analyzers for two runs per day across three days.
- Analytical Specificity (Endogenous Substances): One pool of human serum spiked with interferent, a second pool without, and then mixed in different ratios to create a dilution series (0 to 10 concentrations).
- Analytical Specificity (Common Drugs): Eighteen commonly used drugs added to native patient samples. Serum sample pools were at two target concentrations (~1.8 mg/dL and ~4.9 mg/dL).
- Method Comparison: n=71 human sera samples.
- Matrix Comparison: 32 tubes collected per anticoagulant (Li-heparin, K2-EDTA, K3-EDTA).
- Data Provenance: Not explicitly stated, but implies clinical lab settings ("human sera samples," "native patient samples"). The text does not specify country of origin or whether the studies were retrospective or prospective, though typical analytical validation studies are prospective.
Number of experts used to establish the ground truth for the test set and the qualifications of those experts:
- This is an in vitro diagnostic (IVD) reagent for quantitative determination of a chemical substance. The "ground truth" for such devices is typically established through reference methods or highly accurate laboratory technologies, not expert human readers/adjudicators as would be common for imaging or pathology devices.
- The predicate device's traceability is stated as "Standardized against the Doumas manual reference method." The candidate device claims the same traceability. The Doumas method serves as the ground truth reference.
Adjudication method (e.g., 2+1, 3+1, none) for the test set:
- Not applicable as this is a quantitative chemical assay, not an interpretative task requiring human adjudication of results.
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:
- Not applicable. This is a laboratory diagnostic reagent, not an AI-assisted diagnostic imaging or pathology device that would involve human readers.
If a standalone (i.e. algorithm only without human-in-the-loop performance) was done:
- Yes, this is implicitly a standalone device. The COBAS INTEGRA Bilirubin Direct Gen.2 Reagent, when run on the COBAS INTEGRA system, performs the measurement automatically without human interpretative input for the actual bilirubin value. The studies presented (precision, linearity, detection limits, interference, method comparison, matrix comparison) are all "standalone" performance evaluations of the reagent/analyzer system.
The type of ground truth used (expert consensus, pathology, outcomes data, etc.):
- The ground truth for the analytical performance characteristics is established by highly controlled laboratory measurements using validated reference methods (e.g., the Doumas method for bilirubin), highly characterized samples, and adherence to CLSI (Clinical and Laboratory Standards Institute) guidelines (EP5-A2, EP6-A, EP17-A2), which are standard for IVD device validation.
The sample size for the training set:
- Not applicable. This is a chemical reagent, not a machine learning model that requires a "training set" in the conventional sense. The "training" here would be the development and optimization of the reagent formulation and assay parameters based on extensive chemical and analytical research and development.
How the ground truth for the training set was established:
- Not applicable (see point 7). The "ground truth" during the development phase would involve using highly accurate and precise analytical techniques to characterize various bilirubin concentrations and interfering substances to optimize the reagent's performance against these known values.

Ask a Question

Ask a specific question about this device

Page 1 of 1