(306 days)
The GEM Premier 5000 is a portable critical care system for use by health care professionals to rapidly analyze heparinized whole blood samples at the point of health care delivery in a clinical setting and in a central laboratory. The instrument provides quantitative measurements of Hematocrit and Total Hemoglobin from venous and arterial heparinized whole blood, as well as quantitative measurements of O2Hb. COHb. MetHb. HHb. sO2 from venous, arterial and capillary heparinized whole blood. These parameters, along with derived parameters, aid in the diagnosis of a patient's oxygen delivery capacity.
- Hematocrit (Hct) measurements in whole blood of the packed red cell volume of a blood sample are used to distinguish normal from abnormal states, such as anemia and erythrocytosis (an increase in the number of red cells).
- Total Hemoglobin (tHb): Total hemoglobin measure the hemoglobin content of whole blood for the detection of anemia.
- · COHb: Carboxyhemoglobin measurements are used to determine the carboxyhemoglobin content of human blood as an aid in the diagnosis of carbon monoxide poisoning.
- · MetHb: Methemoglobin measurements are used to determine different conditions of methemoglobinemia.
· HHb: Deoxyhemoglobin, as a fraction of total hemoglobin, is used in combination with oxyhemoglobin to measure oxygenation status.
· O2Hb: Oxyhemoglobin, as a fraction of total hemoglobin, is used in combination with deoxyhemoglobin to measure oxygenation status.
· sO2: Oxygen saturation, more specifically the ratio between the concentration of oxyhemoglobin and oxyhemoglobin plus deoxyhemoglobin, is used to measure oxygenation status.
The GEM Premier 5000 system provides health care professionals in central laboratory or point-of-care clinical settings with fast, accurate, quantitative measurements of Hematocrit and Total Hemoglobin from venous and arterial heparinized whole blood, as well as quantitative measurements of O₂Hb, COHb, MetHb, HHb, sO2 from venous, arterial and capillary heparinized whole blood.
Key Components:
Analyzer: Employs a unique color touch screen and a simple set of menus and buttons for user interaction. The analyzer guides operators through the sampling process with simple, clear messages and prompts.
GEM Premier 5000 PAK (disposable, multi-use GEM PAK): Houses all required components necessary to operate the instrument once the cartridge is validated. These components include the sensors, CO-Ox optical cell, Process Control (PC) Solutions, sampler, pump tubing, distribution valve and waste bag. The GEM PAK has flexible menus and test volume options to assist facilities in maximizing efficiency. NOTE: The EEPROM on the GEM PAK includes all solution values and controls the analyte menu and number of tests.
Intelligent Quality Management 2 (iQM2): iQM2 is an active quality process control program designed to provide continuous monitoring of the analytical process before, during and after sample measurement with real-time, automatic error detection, automatic correction of the system and automatic documentation of all corrective actions. iQM2 is a statistical process control system that performs 5 types of continuous, quality checks to monitor the performance of the GEM PAK, sensors, CO-Ox, and reagents. These checks include System.
The provided text describes the GEM Premier 5000, a portable critical care system for analyzing heparinized whole blood samples. It focuses on the device's technical specifications and performance studies to demonstrate substantial equivalence to a predicate device (GEM Premier 4000) for Hematocrit and CO-Oximetry measurements.
1. A table of acceptance criteria and the reported device performance
The document does not explicitly state "acceptance criteria" for precision or method comparison studies in a single table with target values. Instead, it presents the "Within Analyzer SD/CV" and "Total SD/CV" or "Bias" for various analytes and compares them to specifications which are indirectly implied to be the acceptance criteria. For the purpose of this response, I will interpret "SD/CV Spec" as the acceptance criteria for reproducibility and "Medical Decision Levels" with a calculated "Total Error Observed" compared to an unstated "GEM Premier 5000 Total Error Specifications" as acceptance criteria for accuracy.
Here's a table summarizing the reported device performance based on the precision and method comparison studies:
Table 1: Summary of Device Performance against Implicit Criteria
| Analyte | Test Type | Performance Metric | Reported Performance | Implicit Acceptance Criteria (based on provided data) |
|---|---|---|---|---|
| All Analytes (tHb, O2Hb, COHb, MetHb, HHb, sO2, Hct) | Internal Precision (Aqueous Controls) | Within Analyzer %CV / SD | All results were within specification. (Specific values are provided in the table on page 10) | Not explicitly stated, but the values reported meet the internal specifications. |
| All Analytes (tHb, O2Hb, COHb, MetHb, HHb, sO2, Hct) | Internal Precision (GEM PAK PCS) | Within Analyzer %CV / SD | All results were within specification. (Specific values are provided in the table on page 11) | Not explicitly stated, but the values reported meet the internal specifications. |
| All Analytes (Hct, tHb, O2Hb, COHb, MetHb, HHb, sO2) | Internal Precision (Whole Blood) | Within Run %CV / SD, Total %CV / SD | All results were within specification. (Specific values are provided in the tables on pages 12-14) | Not explicitly stated, but the values reported meet the internal specifications. |
| Hct (%) | Reproducibility (Aqueous Controls - POC) | Reproducibility SD / %CV | Level 1: 0.0 SD / 0.0% | |
| Level 2: 0.0 SD / 0.0% | ||||
| Level 3: 0.6 SD / 0.9% | SD/CV Spec: 2 | |||
| tHb (g/dL) | Reproducibility (Aqueous Controls - POC) | Reproducibility SD / %CV | Level 1: 0.16 SD / 0.8% | |
| Level 2: 0.13 SD / 0.9% | ||||
| Level 3: 0.10 SD / 1.4% | SD/CV Spec: 0.5 (Level 1), 0.35 (Level 2), 0.35 (Level 3) | |||
| O2Hb (%) | Reproducibility (Aqueous Controls - POC) | Reproducibility SD / %CV | Level 1: 0.01 SD / 0.0% | |
| Level 2: 0.05 SD / 0.1% | ||||
| Level 3: 0.04 SD / 0.0% | SD/CV Spec: 1.5 | |||
| COHb (%) | Reproducibility (Aqueous Controls - POC) | Reproducibility SD / %CV | Level 1: 0.05 SD / 0.2% | |
| Level 2: 0.04 SD / 0.3% | ||||
| Level 3: 0.07 SD / 2.1% | SD/CV Spec: 1.0 | |||
| MetHb (%) | Reproducibility (Aqueous Controls - POC) | Reproducibility SD / %CV | Level 1: 0.06 SD / 0.7% | |
| Level 2: 0.06 SD / 2.4% | SD/CV Spec: 1.0 | |||
| HHb (%) | Reproducibility (Aqueous Controls - POC) | Reproducibility SD / %CV | Level 1: 0.06 SD / 0.2% | |
| Level 2: 0.06 SD / 0.8% | ||||
| Level 3: 0.08 SD / 2.3% | SD/CV Spec: 1.5 | |||
| sO2 (%) | Reproducibility (Aqueous Controls - POC) | Reproducibility SD / %CV | Level 1: 0.06 SD / 0.1% | |
| Level 2: 0.07 SD / 0.1% | ||||
| Level 3: 0.08 SD / 0.1% | SD/CV Spec: 1.5 | |||
| Hct (%) | Total Error (Medical Decision Levels) | Total Error Observed | LDL: 1.1 | |
| MDL: 0.9 | ||||
| HDL: 1.3 | "GEM Premier 5000 Total Error Specifications" - Not explicitly provided. | |||
| tHb (g/dL) | Total Error (Medical Decision Levels) | Total Error Observed | LDL: 0.22 | |
| MDL: 0.36 | ||||
| HDL: 0.70 | "GEM Premier 5000 Total Error Specifications" - Not explicitly provided. | |||
| O2Hb (%) | Total Error (Medical Decision Levels) | Total Error Observed | MDL: 0.96 | TEa (Acceptable Total Error) = ± 3.0 (from Clinical Testing for Capillary) |
| COHb (%) | Total Error (Medical Decision Levels) | Total Error Observed | LDL: 0.64 | |
| MDL: 0.72 | TEa (Acceptable Total Error) = ± 2.0 (from Clinical Testing for Capillary) | |||
| MetHb (%) | Total Error (Medical Decision Levels) | Total Error Observed | LDL: 0.54 | |
| MDL: 0.58 | TEa (Acceptable Total Error) = ± 3.0 (from Clinical Testing for Capillary) | |||
| HHb (%) | Total Error (Medical Decision Levels) | Total Error Observed | MDL: 0.89 | TEa (Acceptable Total Error) = ± 3.0 (from Clinical Testing for Capillary) |
| sO2 (%) | Total Error (Medical Decision Levels) | Total Error Observed | MDL: 0.79 | TEa (Acceptable Total Error) = ± 3.0 (from Clinical Testing for Capillary) |
| O2Hb (%) | Clinical Testing (Capillary Samples) | Bias at MDL / 95% CI | Bias at 90.0: 1.19 / (0.48 to 1.89) | TEa: ± 3.0 |
| COHb (%) | Clinical Testing (Capillary Samples) | Bias at MDL / 95% CI | Bias at 3.0: -0.31 / (-0.49 to -0.12) | TEa: ± 2.0 |
| HHb (%) | Clinical Testing (Capillary Samples) | Bias at MDL / 95% CI | Bias at 6.0: -0.56 / (-1.07 to -0.04) | TEa: ± 3.0 |
| sO2 (%) | Clinical Testing (Capillary Samples) | Bias at MDL / 95% CI | Bias at 90.0: 1.20 / (-0.06 to 2.45) | TEa: ± 3.0 |
Note: For the Total Error Observed, the document states these were compared to "GEM Premier 5000 Total Error Specifications," but these specifications are not explicitly provided in the text for all analytes. However, for capillary clinical testing, specific TEa values are given for some analytes, which are used here as proxies for acceptance criteria where available.
2. Sample sized used for the test set and the data provenance (e.g. country of origin of the data, retrospective or prospective)
- Internal Precision Study (Aqueous Controls):
- Sample Size: 120 per analyte/level (3 analyzers, 20 days, 2 runs/day, 1 replicate/run).
- Data Provenance: Internal study (Instrumentation Laboratory Co.), prospective.
- Internal Precision Study (GEM PAK Process Control Solutions):
- Sample Size: 120 per analyte/level (3 analyzers, 20 days, 2 runs/day, 1 replicate/run).
- Data Provenance: Internal study (Instrumentation Laboratory Co.), prospective.
- Internal Precision Study (Whole Blood):
- Sample Size: 120 per analyte/sample mode (3 analyzers, 5 days, 1 run/day, 8 replicates/run).
- Data Provenance: Internal study (Instrumentation Laboratory Co.), prospective. Origin of whole blood samples likely internal or procured, not specified by country.
- Reproducibility Study (Aqueous Controls – Point-of-Care (POC) Setting):
- Sample Size: 90 per analyte/level (3 external POC sites, 3 different GEM Premier 5000 instruments, triplicate measurements, twice a day, total 30 replicates per level, pooled).
- Data Provenance: External clinical POC settings, prospective. Country of origin not specified, but likely USA based on the FDA submission.
- External Precision (Whole Blood - Central Lab and POC settings):
- Sample Size: Varies by analyte and site (e.g., Hct Normal Mode POC-All N=126, CSL N=36, Lab1 N=30, Lab2 N=30). The overall study involved multiple whole blood specimens (at least two per day) analyzed in triplicate daily for 5 days.
- Data Provenance: 2 external central laboratories, 1 internal Customer Simulation Laboratory (CSL), and 3 external POC locations. Prospective, "contrived whole blood specimens were analyzed in addition to native specimens" in the CSL. Country of origin not specified, but likely USA based on the FDA submission.
- Analytical Specificity (Interference Study):
- Sample Size: Not explicitly stated as a single "test set" size for all substances, but experiments were conducted with various concentrations of interfering substances.
- Data Provenance: Internal study (Instrumentation Laboratory Co.), prospective.
- Internal Method Comparison (Clinical Samples):
- Sample Size: Varies by analyte (N=376 for Hct, N=373 for O2Hb, etc.).
- Data Provenance: Clinical samples (heparinized whole blood) altered as needed to cover medical decision levels. Internal study, retrospective/prospective hybrid (samples altered to cover ranges).
- Clinical Testing (Point-of-Care Setting - Normal Mode):
- Sample Size: Varies by analyte (e.g., N=490 for Hct, N=485 for COHb).
- Data Provenance: Three (3) external point-of-care (POC) sites and one (1) internal Customer Simulation Laboratory (CSL) at IL. Heparinized whole blood patient samples from the intended use population. Samples spiked in CSL to cover reportable ranges. Provenance likely USA, prospective.
- Clinical Testing (Capillary Samples):
- Sample Size: Native capillary samples: N=52 (external POC), N=100 (IL CSL) for a total of 152 native samples for O2Hb, HHb, sO2. Total N becomes 180-182 when pooled with additional internally prepared contrived capillary samples for specific analytes (pages 29-30).
- Data Provenance: One external POC site and the IL internal Customer Simulation Laboratory (CSL). Native finger-stick samples and contrived capillary samples. Provenance likely USA, prospective.
3. Number of experts used to establish the ground truth for the test set and the qualifications of those experts (e.g. radiologist with 10 years of experience)
The document relates to an in-vitro diagnostic device measuring blood parameters. Ground truth in this context is typically established by comparative measurements against a recognized reference method or a predicate device, not by expert consensus in the same way an imaging AI would use expert radiologist reads.
For the Internal Method Comparison and Clinical Testing studies, the GEM Premier 5000 was compared to the predicate device, the GEM Premier 4000. The predicate device itself serves as the "reference standard" or "ground truth" for demonstrating substantial equivalence. The document does not specify human experts establishing ground truth for the test set; rather, the predicate device and established analytical methods (e.g., CNMetHb procedure - CLSI H15-A3 for tHb) are used as comparators.
The Linearity study also mentions comparing results to "reference analyzers or standard reference procedures (i.e. CNMetHb procedure - CLSI H15-A3 for tHb)".
4. Adjudication method (e.g. 2+1, 3+1, none) for the test set
Not applicable for this type of in-vitro diagnostic device. "Adjudication" typically refers to the process of resolving discrepancies between human readers or between human readers and an AI output, especially in imaging studies where subjective interpretation is involved. For this device, ground truth is established through quantitative comparisons against established reference methods or a predicate device.
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. This is an in-vitro diagnostic device that provides quantitative measurements of blood parameters, not an AI-based imaging or diagnostic aid that assists human readers/clinicians in interpretation. Therefore, no MRMC study or assessment of human reader improvement with/without AI assistance was conducted or would be relevant.
6. If a standalone (i.e. algorithm only without human-in-the-loop performance) was done
Yes, the studies presented evaluate the performance of the device itself in a standalone capacity. The GEM Premier 5000 provides quantitative measurements directly. Operators interact with the device to load samples and retrieve results, but the analytical process and result generation are entirely performed by the instrument's internal "algorithm" (i.e., its measurement and data processing capabilities).
The device description on page 5 highlights its autonomy:
- "The analyzer guides operators through the sampling process with simple, clear messages and prompts."
- "After inserting the GEM PAK, the instrument will perform an automated PAK warm-up..."
- "Auto PAK Validation (APV) process is automatically completed..."
- "iQM2 manages the quality control process, replacing external quality controls."
This indicates that the device operates autonomously for its core measurements, with human intervention for sample loading and result review, but not for interpreting the raw measurement signals.
7. The type of ground truth used (expert consensus, pathology, outcomes data, etc)
The ground truth for the performance studies was established through:
- Comparison to a Legally Marketed Predicate Device: The GEM Premier 4000 served as the primary comparator for method comparison studies, demonstrating substantial equivalence.
- Standard Reference Procedures/Analyzers: For linearity studies and potentially for other analytical performance aspects, comparison to "reference analyzers or standard reference procedures (i.e. CNMetHb procedure - CLSI H15-A3 for tHb)" was used.
- Certified Control Materials: Aqueous controls and GEM PAK Process Control Solutions that are "traceable to NIST standards, CLSI procedures or internal standards" were used for precision and reproducibility studies. These materials have known, certified values which serve as ground truth for assessing device accuracy and precision.
8. The sample size for the training set
The document does not explicitly mention a "training set" in the context of machine learning or AI algorithms as the primary focus is on proving the analytical performance of an in-vitro diagnostic instrument through traditional analytical validation studies (precision, linearity, method comparison, etc.).
However, the device incorporates "Intelligent Quality Management 2 (iQM2)" which is described as "an active quality process control program designed to provide continuous monitoring of the analytical process... with real-time, automatic error detection, automatic correction of the system and automatic documentation of all corrective actions." While iQM2 itself would have been developed and "trained" or calibrated using extensive internal data during the device's R&D phase, the public 510(k) summary does not provide details on a specific training set size for the iQM2 component. The studies presented are for the validation of the final device's performance, not the internal development of iQM2.
The "internal precision study" and "internal method comparison" are validation studies of the final device. The data shown in these tables is for testing the device's performance, not training it.
9. How the ground truth for the training set was established
As noted in point 8, a "training set" in the context of typical AI/ML models is not explicitly described or relevant for the approval justification in this 510(k) summary. The document validates the measurement accuracy and precision of the device against predicate devices and reference methods, not an AI model that learns from data in the field.
If "training set" refers to the data used to initially calibrate the instrument's sensors and algorithms during development, this ground truth would have been established through:
- Primary Reference Methods: Employing highly accurate and precise laboratory methods (e.g., gas chromatography, certified spectrophotometric methods, gravimetric/volumetric methods) to determine the true concentration of analytes in reference materials and clinical samples.
- Reference Materials/Standards: Calibrating the device using NIST-traceable standards and other certified reference materials with known analyte concentrations.
- Extensive Internal Testing and Optimization: Using a large array of characterized samples (e.g., blood samples with varying analyte concentrations, interfering substances) to optimize the device's measurement algorithms and ensure accurate performance across its claimed reportable range.
§ 864.5600 Automated hematocrit instrument.
(a)
Identification. An automated hematocrit instrument is a fully automated or semi-automated device which may or may not be part of a larger system. This device measures the packed red cell volume of a blood sample to distinguish normal from abnormal states, such as anemia and erythrocytosis (an increase in the number of red cells).(b)
Classification. Class II (performance standards).