(147 days)
The Extended Lipid Panel Assay is an in vitro diagnostic test for quantitative determination of Total Cholesterol, High Density Lipoprotein Cholesterol, and Triglycerides in human serum and Apolipoprotein B in human serum. Values for Total Cholesterol, High Density Lipoprotein Cholesterol, Triglycerides and Apolipoprotein B are calculated by the Vantera® Clinical Analyzer.
· Total Cholesterol measurements are used in the diagnosis and treatment of disorders involving excess cholesterol in the blood, lipid and lipoprotein metabolism disorders.
· High Density Lipoprotein Cholesterol measurements are used in the diagnosis and treatment of lipid disorders (such as diabetes mellitus), atherosclerosis, and various liver and renal diseases.
· Triglyceride measurements are used in the diagnosis and treatment of patients with diabetes mellitus, nephrosis, liver obstruction, other diseases involving lipid metabolism, or various endocrine disorders.
· Apolipoprotein B measurements are used in the diagnosis and treatment of lipid disorders and atherosclerosis.
The Extended Lipid Panel Assay involves the acquisition of a 400 MHz proton NMR spectrum of serum or plasma, passing the spectral information through a Partial Least Squares (PLS) regression model, and deriving analyte concentrations from the spectrum based on the trained PLS model. The proton NMR spectrum of serum and plasma is replete with information from the lipids packaged in lipoproteins. The spectrum consists of multiple proton signals emanating from the TG, cholesteryl esters and free cholesterol present in chylomicrons, VLDL, LDL and HDL, out of which the methylene and methyl proton signals are the most abundant. NMR spectra were recorded for several hundred to several thousand representative serum specimens for which the TG, TC, HDL-C and ApoB were chemically measured. Using a PLS regression routine, the spectral information in the combined methylene and methyl region (0.56 - 1.40 ppm) was trained against the chemical measurements where the information is connected through latent variables. Cross-validation was performed with PRESS statistics to optimize the regression model with an appropriate number of latent variables. Once trained with sufficient number of specimens, for any test specimen spectrum, the spectral information is then converted into lipid or ApoB concentrations through the optimum number of 24 to 27 latent variables for which the regression coefficients were known from the predictor matrix.
The Extended Lipid Panel Assay is an in vitro diagnostic test for the quantitative determination of Total Cholesterol (TC), High Density Lipoprotein Cholesterol (HDL-C), Triglycerides (TG), and Apolipoprotein B (ApoB) in human serum and plasma. The device uses Nuclear Magnetic Resonance (NMR) technology to derive analyte concentrations.
Here's an analysis of the acceptance criteria and the study that proves the device meets them:
1. Table of Acceptance Criteria and Reported Device Performance
The acceptance criteria for this device are demonstrated through various analytical performance studies. The document does not explicitly state pre-defined acceptance criteria in a dedicated table with pass/fail results against specific thresholds for each performance metric, but rather presents the results of these studies. Based on the provided performance data, the implicit acceptance criteria would be for the device to show good analytical sensitivity, precision, linearity, and minimal interference, and to demonstrate substantial equivalence to predicate devices through method comparison studies.
Below is a table summarizing the reported device performance, which implicitly demonstrates it meets the necessary criteria for "substantial equivalence" based on the FDA's assessment in the 510(k) clearance.
| Performance Metric | Analyte | Reported Device Performance | Implicit Acceptance Standard (Inferred from predicates/regulatory guidance for clinical chemistry tests) |
|---|---|---|---|
| Analytical Sensitivity (LoQ) | TC | 24 mg/dL | Lowest concentration measurable with acceptable precision and accuracy (implicitly deemed acceptable by FDA) |
| TG | 15 mg/dL | Lowest concentration measurable with acceptable precision and accuracy | |
| HDL-C | 13 mg/dL | Lowest concentration measurable with acceptable precision and accuracy | |
| ApoB | 18 mg/dL | Lowest concentration measurable with acceptable precision and accuracy | |
| Within-run Precision (%CV) (n=20) | TC | 0.9 - 1.6% | Low %CV indicating high within-run precision (typically 0.975 typically sought) and Deming regression slope/intercept close to 1 and 0, respectively. |
| TG | r = 0.9805 (implied from y= -3.858 + 0.9805x and visual) | High correlation | |
| HDL-C | r = 0.985 | High correlation | |
| ApoB | r = 0.980 | High correlation |
2. Sample sizes used for the test set and the data provenance
- Analytical Sensitivity (LoQ): Not specified how "LoQ" was determined by sample size, but indicates "lowest concentration measurable with acceptable precision and accuracy."
- Assay Precision (Within-run and Within-laboratory): 20 replicates for within-run and 80 replicates (for n=80, it means 40 total runs over 20 days with 2 replicates per run on 1 instrument) for within-laboratory precision, using three patient serum pools (low, medium, high). The provenance is "patient serum pools."
- Reproducibility: Five levels of serum panels were tested for 5 days, 6 runs per day, 2 replicates per run at 3 sites. This results in 5 (levels) * 5 (days) * 6 (runs/day) * 2 (replicates/run) * 3 (sites) = 900 measurements per analyte. The provenance is from "serum panels," presumably patient-derived.
- Linearity: "Reference serum pools were prepared from patient specimens with low to high values" and "mean values from analysis of four replicates of each pool" were used. The exact number of 'pools' created for each analyte is not explicitly stated but implies multiple points across the range. The provenance is "patient specimens."
- Interfering Substances: "samples with spiked concentrations of interferent" were used. The number of samples is not specified, but it covered "Eight endogenous agents and thirty drugs."
- Method Comparison:
- TC: n=281 pooled serum samples
- TG: n=270 pooled serum samples
- HDL-C: n=15575 (This number seems exceptionally high compared to other n values; it might be a typo or represent cumulative data points over various experiments) pooled serum samples
- ApoB: n=266 pooled serum samples
The data provenance for method comparison is "pooled serum samples across the reportable range" of the device. All studies appear to be retrospective using banked or prepared samples, as typical for analytical performance studies of this nature. The country of origin is not specified but is presumably the US given the FDA submission.
3. Number of experts used to establish the ground truth for the test set and the qualifications of those experts
This device is a clinical chemistry assay for quantifying specific analytes (TC, TG, HDL-C, ApoB) rather than an imaging or diagnostic AI device that interprets qualitative information and requires human expert consensus for ground truth.
- Ground Truth for Method Comparison: The "ground truth" for the method comparison studies was established by predicate clinical chemistry devices (biochemical assays like enzymatic colorimetric assay or nephelometric immunoassay). These are established methods that serve as the comparative standard. No human experts were involved in establishing the ground truth for the values themselves, as it's a quantitative measurement.
- Qualifications of Experts: Not applicable in the context of establishing ground truth for quantitative chemical measurements. However, the development, validation, and regulatory submission would have been overseen by qualified scientists and regulatory affairs professionals.
4. Adjudication method for the test set
Not applicable. Adjudication methods (like 2+1, 3+1) are typically used for qualitative or semi-quantitative diagnostic devices where human interpretation might differ, and an expert panel resolves discrepancies to establish ground truth. For quantitative clinical chemistry assays, the reference method (predicate device in this case) provides the "adjudicated" ground truth.
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
Not applicable. MRMC studies are designed for evaluating diagnostic devices that involve human interpretation of images or other qualitative data, often with AI assistance, to assess the impact of AI on reader performance. This device is a standalone clinical chemistry analyzer that provides quantitative measurements; it does not involve human "readers" or AI assistance in a diagnostic interpretation workflow.
6. If a standalone (i.e., algorithm only without human-in-the-loop performance) was done
Yes, this device is inherently a standalone algorithm. The Extended Lipid Panel Assay is described as utilizing "passing the spectral information through a Partial Least Squares (PLS) regression model, and deriving analyte concentrations from the spectrum based on the trained PLS model." This process is entirely automated and does not involve human intervention in the result generation once the sample is run on the Vantera® Clinical Analyzer. The performance studies (precision, linearity, method comparison) directly evaluate this standalone algorithmic performance.
7. The type of ground truth used (expert consensus, pathology, outcomes data, etc.)
The ground truth for evaluating the Extended Lipid Panel Assay was based on measurements obtained from legally marketed predicate clinical chemistry devices. These predicate devices are established methods for quantifying the specific lipid and lipoprotein analytes (TC, TG, HDL-C, ApoB).
8. The sample size for the training set
For the Partial Least Squares (PLS) regression model: "NMR spectra were recorded for several hundred to several thousand representative serum specimens for which the TG, TC, HDL-C and ApoB were chemically measured." The exact number is not precisely stated but indicates a substantial training dataset.
9. How the ground truth for the training set was established
The ground truth for the training set was established by chemical measurements (presumably using standard, often wet-chemistry or enzymatic reference methods) for TG, TC, HDL-C, and ApoB on the "several hundred to several thousand representative serum specimens." These chemical measurements served as the target values for the PLS regression model to learn the relationship between the NMR spectra and the analyte concentrations.
§ 862.1175 Cholesterol (total) test system.
(a)
Identification. A cholesterol (total) test system is a device intended to measure cholesterol in plasma and serum. Cholesterol measurements are used in the diagnosis and treatment of disorders involving excess cholesterol in the blood and lipid and lipoprotein metabolism disorders.(b)
Classification. Class I (general controls). The device is exempt from the premarket notification procedures in subpart E of part 807 of this chapter subject to § 862.9.