(301 days)
The Sysmex® CS-5100 is a fully automated blood coagulation analyzer intended for in vitro diagnostic use using plasma collected from venous blood samples in 3.2% sodium citrate tubes to analyze clotting, chromogenic and immunoassay methods in the clinical laboratory.
For determination of:
- Prothrombin Time (PT) seconds and PT INR with Dade® Innovin®
- Activated Partial Thromboplastin Time (APTT) with Dade® Actin® FSL
- Fibrinogen (Fbg) with Dade® Thrombin Reagent
- Antithrombin (AT) with INNOVANCE® Antithrombin
- D-dimer with INNOVANCE® D-Dimer
The performance of this device has not been established in neonate and pediatric patient populations.
The Sysmex CS-5100 is an automated blood coagulation instrument which can analyze samples using clotting, chromogenic and immunoassay methods. Analysis results are displayed on the Information Processing Unit (IPU) screen. They can be printed on external printers or transmitted to a host computer. Sold separately from the instrument are the associated Reagents, Controls, Calibrators, and Consumable materials. The subject of this 510(k) notification are reagent applications which perform the coagulation tests Prothrombin Time (PT) seconds and PT INR with Dade® Innovin®; Activated Partial Thromboplastin Time (APTT) with Dade® Actin® FSL; Fibrinogen (Fbg) with Dade® Thrombin Reagent; Antithrombin (AT) with INNOVANCE® Antithrombin; and D-dimer with INNOVANCE® D-Dimer.
Here's an analysis of the acceptance criteria and the studies performed for the Sysmex CS-5100 device, based on the provided document:
1. Table of Acceptance Criteria and Reported Device Performance
The document does not explicitly state pre-defined acceptance criteria values for most of the performance metrics within the tables. Instead, it states that "Results from each application met the pre-established acceptance criteria" for method comparison, and "The data for all tested reagents met the pre-determined acceptance criteria" for detection capability. For linearity, it states "All reagents met the pre-determined acceptance criteria."
However, we can infer some implied criteria based on the reported values alongside statements of acceptance, especially for Method Comparison, Reproducibility, Detection Capability, and the D-Dimer PE Exclusion Study. For the D-Dimer study, specific lower bounds for 95% LCL are provided for sensitivity and NPV, which act as de facto acceptance criteria.
Inferred Acceptance Criteria and Reported Performance for Sysmex CS-5100:
| Study Category | Performance Metric | Inferred Acceptance Criteria | Reported Device Performance |
|---|---|---|---|
| Method Comparison | Passing-Bablok Regression (y = a*x + b, r) | Slope (a) close to 1, Intercept (b) close to 0, Correlation coefficient (r) close to 1 (indicating strong agreement with predicate). The document states "Results from each application met the pre-established acceptance criteria." | Prothrombin Time (seconds) with Dade® Innovin®: n=469 (combined), y = 1.000x + 0.100, r = 0.998 |
| Prothrombin Time (INR) with Dade® Innovin®: n=465 (combined), y = 1.047x - 0.047, r = 0.999 | |||
| Activated Partial Thromboplastin Time (APTT) with Dade® Actin® FSL: n=466 (combined), y = 1.026x - 1.315, r = 0.996 | |||
| Fibrinogen quantitation with Dade® Thrombin Reagent: n=368 (combined), y = 1.018x + 4.633, r = 0.995 | |||
| Antithrombin quantitation with INNOVANCE® Antithrombin: n=381 (combined), y = 0.980x + 0.222, r = 0.996 | |||
| D-dimer quantitation with INNOVANCE® D-Dimer: n=361 (combined), y = 1.021x - 0.007, r = 0.996 | |||
| Reproducibility | Within Run %CV, Between Run %CV, Between Day %CV, Site-to-Site %CV, Total CV (Within Site) | %CVs within acceptable limits for laboratory assays (e.g., typically 93.0%; NPV 95% LCL > 99.0% (for non-standardized); NPV* 95% LCL > 97.7% (for standardized to 15% prevalence). | US and OUS Combined Data (n=1467): |
| Sensitivity % = 98.0 (95% LCL= 93.0) | |||
| Specificity % = 54.5 (95% LCL= 51.9) | |||
| NPV % = 99.7 (95% LCL= 99.0) | |||
| NPV % (standardized to 15% prevalence) = 99.4* (95% LCL= 97.7) | |||
| PPV % = 13.8 (95% LCL= 11.4) | |||
| PPV % (standardized to 15% prevalence) = 27.6* (95% LCL= 23.5) |
2. Sample Size Used for the Test Set and Data Provenance
-
Method Comparison:
- Sample Sizes:
- PT (seconds): n=469 (combined from 3 sites: n=125, n=209, n=135)
- PT (INR): n=465 (combined from 3 sites: n=122, n=208, n=135)
- APTT: n=466 (combined from 3 sites: n=126, n=210, n=130)
- Fibrinogen: n=368 (combined from 3 sites: n=145, n=91, n=132)
- Antithrombin: n=381 (combined from 3 sites: n=135, n=120, n=126)
- D-dimer: n=361 (combined from 3 sites: n=137, n=108, n=116)
- Data Provenance: Conducted at three external sites in the United States (US). The samples were patient samples.
- Sample Sizes:
-
Reproducibility Studies:
- Sample Sizes: Not explicitly stated as 'n=' count for each sample tested, but refers to "Sample Range (mean of all sites)" for each application. These studies involved "two runs per day, with two replicates per run, at each of the three sites". Given the "20-day precision studies", this implies a significant number of measurements for each sample type.
- Data Provenance: Two external sites in Germany and one external site in the United States.
-
Detection Capability Studies:
- Sample Sizes: Not explicitly stated, but performed for specific reagents: Fibrinogen, Antithrombin, and D-dimer.
- Data Provenance: Not explicitly stated, but typically in vitro studies.
-
Linearity & Measuring Range Studies:
- Sample Sizes: Not explicitly stated, but performed for Fibrinogen, Antithrombin, and D-dimer.
- Data Provenance: Not explicitly stated, but typically in vitro studies.
-
Reference Interval Studies:
- Sample Sizes: Not explicitly stated as a number, but indicates the study population "did not include neonate and pediatric sample populations".
- Data Provenance: Conducted at three clinical study sites in the United States.
-
D-Dimer PE Exclusion Validation Study:
- Initial Patients: 1930 consecutive outpatients.
- Excluded Patients: 96.
- Patients for Final Analysis: n=1467 (US: n=1424, OUS: n=43).
- Data Provenance: Multi-center study using frozen specimens collected prospectively from outpatients presenting to emergency or ambulatory departments with suspected PE. Data from both the US and "OUS" (Outside US, specifically European population for prevalence) were included.
3. Number of Experts Used to Establish the Ground Truth for the Test Set and Qualifications of those Experts
-
For Method Comparison, Reproducibility, Detection Capability, Linearity, and Reference Intervals: The ground truth for these analytical performance studies is established by the methods themselves, often comparing the device to a reference method or evaluating inherent precision characteristics against statistical benchmarks. No human experts are described for establishing ground truth in these sections.
-
For D-Dimer PE Exclusion Validation Study:
- Ground Truth Establishment: The ground truth for PE diagnosis was established through imaging methods (e.g., spiral CT and/or VO scan) and 3-month follow-up for patients with initial negative diagnosis.
- Number/Qualifications of Experts: The document mentions "physician's discretion" for imaging and that "All potentially eligible patients were evaluated using the Wells' rules...". This implies that medical professionals (e.g., emergency physicians, radiologists, other clinicians) were involved in the diagnosis and follow-up, but the specific number and detailed qualifications of these individual experts are not provided.
4. Adjudication Method for the Test Set
-
For Method Comparison, Reproducibility, Detection Capability, Linearity, and Reference Intervals: No adjudication method as typically understood for human interpretation of data (e.g., 2+1 rule) is described. These are analytical performance studies where results are quantitative values compared against other instruments or statistical metrics.
-
For D-Dimer PE Exclusion Validation Study: The ground truth involved a combination of imaging and 3-month follow-up. While a final diagnosis would likely involve a consensus among the treating clinicians or specialists (e.g., radiologists, internists), no explicit "adjudication method" (like multiple readers reaching a consensus on an image) is described for the final PE diagnosis. The process relies on standard clinical diagnostic pathways and follow-up.
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 involving human readers with/without AI assistance was not done. The Sysmex CS-5100 is an automated blood coagulation analyzer, not an AI-powered diagnostic imaging tool that assists human readers. The studies focus on the analytical performance of the instrument itself and its diagnostic utility (e.g., D-dimer for PE exclusion).
6. If a standalone (i.e., algorithm only without human-in-the-loop performance) was done
- Yes, the studies described are standalone performance evaluations of the device (or algorithm if you consider the reagent-instrument combination an 'algorithm') in an automated setting. The device operates automatically, and the performance data (method comparison, reproducibility, detection capability, linearity, D-dimer PE exclusion) reflect its direct analytical output without direct human interpretation influencing the measurement itself at the point of testing. Human input is involved in sample collection, entering patient data, and clinical interpretation of the results, but not in the "reading" of the output in a way that an AI for image analysis would be 'standalone'.
7. The type of ground truth used (expert consensus, pathology, outcomes data, etc.)
- For Method Comparison: The ground truth is the result from the predicate device (Sysmex CA-1500), which is a legally marketed device for the same intended use.
- For D-Dimer PE Exclusion Validation Study: The ground truth for the presence or absence of PE was based on a combination of imaging methods (e.g., spiral CT and/or VO scan) and 3-month clinical outcomes data (follow-up) to confirm the absence of PE in initially negative cases.
8. The sample size for the training set
The document describes premarket notification for a medical device (instrument), not an AI/ML algorithm that typically requires a distinct training set for model development and a test set for performance evaluation. For this type of device, the "training" analogous to an AI model would be the internal development and calibration of the instrument by the manufacturer using their own data, which is not typically detailed in a 510(k) summary in terms of "training set sample size." The studies presented are all performance validations of the final device. Therefore, a specific "training set sample size" is not applicable/provided in this context.
9. How the ground truth for the training set was established
As there is no "training set" described in the context of an AI/ML algorithm development, this question is not applicable to this device submission. The device's internal calibration and development would have used reference materials and established laboratory methods, but these are not referred to as a "training set" with ground truth established in the same manner as for an AI model.
§ 864.5425 Multipurpose system for in vitro coagulation studies.
(a)
Identification. A multipurpose system for in vitro coagulation studies is a device consisting of one automated or semiautomated instrument and its associated reagents and controls. The system is used to perform a series of coagulation studies and coagulation factor assays.(b)
Classification. Class II (special controls). A control intended for use with a multipurpose system for in vitro coagulation studies is exempt from the premarket notification procedures in subpart E of part 807 of this chapter subject to the limitations in § 864.9.