CRITICAL CARE

Kevin Fallon, Ph.D.

Sohrab Mansouri, Ph.D.

The GEM PremierTM 3000

with Intelligent Quality
Management (iQMTM):
Features, Technical Description
& Statistical Validation

Kevin Fallon, Ph.D.

Sohrab Mansouri, Ph.D.

The GEM PremierTM 3000

with Intelligent Quality
Management (iQMTM):
Features, Technical Description
& Statistical Validation

Table of Contents

I. Features

Page

II. Technical Description

Page

10

III. Statistical Validation

Page

17

IV. Glossary of Acronyms

Page

21

Appendix

Page

22

I. Features

What It Is
The new Intelligent Quality Management system (iQMTM) from Instrumentation
Laboratory (IL), is a fully automated Quality Assurance system for use on ILs
GEM Premier 3000 analyzer. iQM replaces the need for conventional external
Quality Control (QC) with its combination of software, Process Control (PC)
Solutions, and Calibration Validaton Product (CVP) components.
Designed to ensure at least the same level of quality as traditional QC methods,
iQM significantly reduces time and costs of performing QC, while helping
to assure regulatory compliance and improving the quality of test results.

What It Does
iQM monitors performance of all critical components of the GEM Premier 3000.
The "intelligent" quality management system:
Validates cartridge integrity
Identifies changes that affect analytical performance and potential
failure patterns
Automatically performs corrective actions
Automatically documents the failure and corrective actions taken

How It Works
Unlike external forms of QC, iQM monitors performance real-time, conducting:
System checks
Sensor checks
Stability checks
Pattern checks

System checks include fluidics checks to test sample integrity, reagent

solutions, and the peristaltic pump. Additionally, system checks test
mechanical components of the cartridge, such as the valve position and
sample arm position, as well as the instrument heater block and electronics.
All system checks are performed on an ongoing basis.

If any system failure is found, iQM confirms the nature of the failure, then
either rejects the cartridge or halts instrument operation.

Sensor checks are performed using three different solutions, each at different
intervals. PC Solution B is measured after every patient sample, and in
the event the instrument is unused, every half hour. Sensor outputs are
monitored in PC Solution B every thirty seconds. During these checks, the
drift is analyzed to assess the difference between a measured value and
the expected value.
To further test for drift and slope, iQM uses fresh PC Solution A every four
hours, and fresh PC Solution C every twenty-four hours. iQM confirms any
sensor failure and either suppresses the result of a failed parameter, or
disables any sensor that shows persistent failure.

Process stability checks are also performed in order to verify the process
control solution stability during the life of a cartridge. In the event of process
stability failure, the cartridge is rejected. (Solutions tend to be very stable;
the feature simply further ensures accuracy and precision in test results.)
Last, iQM performs Failure Pattern Recognition (FPR) Checks. Common
sources of failure, such as micro-clots on sensors, any sensor malfunction,
or sensor interference, leave "fingerprints" on the sensors. These fingerprints
are read as patterns that iQM can detect and attempt to correct. For example,
in the event of a micro-clot, iQM will automatically perform a special rinse. If it
cannot correct the problem, the sensor will be disabled. Additionally, iQM
can use FPR to warn of the presence of interference in a sample. See
"Technical Explanation" and "Statistical Validation" for detail and supporting
data.

Why It Works: Compatibility with the GEM Premier 3000

Designed specifically to assure accurate and precise test results, iQM
exemplifies the simplicity and ease-of-use features associated with the
GEM Premier 3000 system. The following unique elements make iQM
possible on the GEM Premier 3000:
All analytical components (e.g., sensors, solutions, tubing, sampler)
of the GEM Premier 3000 system are included in a self-contained,
multi-use, disposable cartridge (PAK).

The GEM PAK is a completely closed system. The user cannot introduce
changes to the analytical system during the three-week use-life of the PAK.
The GEM PAK constitutes a "run," per the National Committee for Clinical
Laboratory Standards (NCCLS) definition (the period of time during which
an analytical system is expected to be stable).
The GEM PAK solution values are traceable to NIST standards and
stated in each PAK barcode.
Each sensor and each lot of solutions manufactured by IL are functionally
tested prior to PAK assembly. In addition, a representative sample of
each PAK batch is functionally tested. Therefore, prior to release of each
batch of PAKs, IL is able to test the complete analytical system, including
the actual combination of analytical elements (e.g., sensors, solutions,
tubing, sampler) that will eventually be used by the operator.
The GEM Premier 3000 analyzer records comprehensive information
about all used PAKs, including all calibrations and millivolt (mV) sensor
readings. This has allowed IL to study and understand the patterns
generated by any PAK failure.
Immediately after insertion of the iQM PAK into the GEM Premier 3000,
and before the system will report any results, an external CVP is run
to validate the integrity of the PC Solutions and overall performance
of the analytical system.

Why It's Better Than Conventional QC Methods

In order to assure quality results from a blood gas analyzer, hospital staff
routinely perform daily liquid QC. This process, as well as monitoring,
interpreting and documenting the results, places a significant burden
on hospital staff. Traditional QC methods pose many challenges, which
may even compromise the quality and accuracy of test results.
For example, with traditional QC:
Manual process requires skilled labor
Aqueous QC ampoules must be run immediately due to the volatility
of O2 and CO2
Corrective action involves potentially complex processes and skills
Manual documentation is needed for regulatory compliance
Discrete process may leave errors undetected until a QC or a series
of QCs are run (after hours or days)
The alternative QC methods advocated by various manufacturers of critical
care analyzers, namely on-board automated quality control or electronic
QC (EQC), are still inadequate in addressing all of the issues previously
mentioned.
On-board automated quality control systems are not capable of continually
monitoring the system functionality and do not address the manual corrective
action and documentation issues. EQC only confirms the operation of the
instruments measurement circuitry and its computational components.
However, EQC fails to provide information about the chemical measurement
process, which includes the sensors and reagents, and it is a manual,
discrete process.
iQM has been developed to address all of these issues. iQM is an active
quality process control program designed to help ensure that the GEM
Premier 3000 analyzer provides reliable results at all times. The quality
control process is an integral part of system operation, and reduces time
to error detection from hours to minutes.

iQM continuously monitors operation of the entire testing process including,

sensors, fluidics and electronics, and automatically performs and documents
corrective actions upon detection of an error. Running traditional liquid
quality controls are no longer needed.
iQM
Continuous control
Automatic corrective
action
Automatic
documentation
Automated
operation
Integrated in GEM
Premier 3000 PAKs
Total Analytical
System QC

Traditional QC
Intermittent control
Manual corrective
action
Manual
documentation
Labor-intensive
operation
Additional QC
materials required
May leave errors
undetected

Electronic QC
Intermittent control
Manual corrective
action
Manual
documentation
Manual operation

Auto QC
Intermittent control
Manual corrective
action
Manual
documentation
Manual maintenance
of external modules
Additional QC
Additional QC and/or
simulators required
hardware required
Electronic component May not check the
check only
full fluidic pathway

Calibrators and Controls

iQM cartridges utilize PC Solutions A, B and C to perform multiple functions:
To indicate if the system has drifted from the reference points established
after cartridge insertion and verification with CVP
As a reference, to eliminate natural sensor drift (calibrate)
To identify abnormal sensor behavior drift and trigger FPR software
to determine the cause of the failure, and initiate corrective action
To determine the status of the analytical system, rather than a single
analysis of an external quality control sample. This is achieved by
analyzing the collective values of the three PC Solutions, tested hundreds
of times during the life of a PAK, along with FPR.
The perception that the same solution cannot be used for two different
functions most often originates from experience with other types of diagnostic
equipment, such as clinical chemistry systems. In the past, calibrators and
controls required manual reconstitution. Reagents were also reconstituted
and changed, sometimes within the course of a day. Tubing changes and
other maintenance functions that may affect system performance, were
also performed routinely. A separate control material was a check that the
calibrator was reconstituted correctly, or had not expired, and that other
changes to the system did not alter calibration. The GEM Premier 3000, as
previously discussed, is a closed system. The operator cannot change any
of the internal cartridge components, and the cartridge will not be accepted
by the system if the expiration date has past. Therefore, prior concerns that
resulted in separate solutions being used for calibrators and controls do not
exist with the GEM Premier 3000 iQM cartridge.
PC Solutions play important roles within the iQM cartridge. Therefore, a
Process Stability check is performed on the solutions, in the very unlikely
event that all PC Solutions deteriorated at the same rate. The Process
Stability check is a method to verify PC Solution stability throughout the
cartridge use-life. The measured oxygen in PC Solution A during the use-life
is compared to the very first measured PC Solution A oxygen value during
warm-up. The delta must be within allowable limits. The PO2 in PC Solution
A is used for the Process Stability check because:

Oxygen is considered the most sensitive parameter for detecting

deterioration in PC Solutions, since they have no oxygen-buffering
capacity.
The process of measuring oxygen in PC Solution A utilizes all three PC
Solutions. Therefore, deterioration in any of the PC Solutions will affect
the measured oxygen in the PC Solution A.
In the event of Process Stability failure, the cartridge will be rejected.
There are PC Solutions used to perform additional sensor checks that do
not perform a secondary function for a particular sensor within the cartridge.
However, only pH, PCO2 and PO2 sensors require the additional pattern
check for identifying malfunctions that are very rare and very slow in
progression. Therefore, a PC Solution not used in the initial cartridge
verification process is used to further evaluate the performance of these
sensors. PC Solution C serves this purpose for pH and PCO2, and PC
Solution A provides an additional measurement for PO2.
The GEM Premier 3000 with iQM initiates a new paradigm in quality
management. It is the first clinical instrument to apply the Process Control
approach in the laboratory setting: determine the quality requirements,
identify the potential sources of error, and design a method of control for
each potential failure.
Statistical analysis of iQM compared to traditional QC has shown that iQM
is better than traditional controls at detecting failures that could result in
clinically significant errors. iQM can detect more errors, and detect them
sooner.

iQM Benefits
iQM provides laboratories and point-of-care testing locations with the
following key benefits:

Quality Assurance
Active, continuous, real-time quality process
Detects and attempts to correct errors immediately
Detects errors sooner than external QC, reducing time to error detection
from hours to minutes
Ensures that the optimal quality control protocol is followed at all times,
regardless of the time of day or operator training

Patient Safety
Helps assure the quality of each patient result
Helps assure that the right medical decision is made for the patient
Prevents reporting sample results when there is a risk of reporting an
incorrect result

Cost Reduction
Eliminates all manual processes associated with traditional quality control
Performs automatically, with no human intervention
Documents failures and corrective actions automatically, with no human
intervention
Eliminates the labor required to maintain a quality control protocol, train
operators, run routine external quality control material (or the maintenance
and refilling of mechanical ampoule beakers), document quality control
results, troubleshoot when a failure is identified, and document the failure
and corrective action taken
Produces reports required by regulatory agencies
Reduces inventory management and procurement procedures

10

II. Technical Description

Understanding GEM Premier 3000 Operation

This section provides a detailed description of the GEM Premier 3000
operation, a prerequisite in understanding how iQM functions.

Sampling Position

Figure 1. Internal components of the GEM Premier 3000 cartridge

System description
The GEM Premier 3000 system includes two components: the instrument
and a disposable cartridge. The cartridge measures pH, PCO2, PO2, Na+,
K+, Ca++, Glucose, Lactate and Hematocrit.
The following provides an overview of the cartridge:
All required components for sample analysis are included in the cartridge,
including sensors, sampler, pump tubing, distribution valve, waste
bag and PC Solutions. The cartridge components and fluidic path
are schematically illustrated in Figure 1.
The sensor card contains all of the sensors in a gas-tight chamber.
The sensors are monitored with three PC Solutions: A, B and C. These
solutions are pre-tonometered and contain known quantities of the analytes
tested using known reference standards. The solutions are sealed in gas
impermeable bags with no headspace, allowing their use over a wide
range of ambient temperatures and atmospheric pressures. PC Solution
B is also used for rinsing.
There is a fourth bag called "Reference Electrode Solution" that contains
silver ion. This solution is pumped into the reference channel in the
sensor card to form the Ag/Ag+ reference electrode.
The sensor card resides in a thermal block, which maintains the
temperature at 37C and provides an electrical interface to the sensors.
The peristaltic pump moves various fluids (sample, PC Solutions and
reference solutions) into the sensor card and eventually to the waste bag.

11

Cartridge Operation
When the cartridge is inserted in the instrument, all information associated
with the cartridge is read from the cartridge bar code and recorded. This
information includes cartridge type (specifying menu, sample capacity and
use-life), expiration date and PC Solution values. Once the cartridge bar
code is validated (new cartridge with valid expiration date), the sensors are
hydrated and calibrated within 30 minutes before the analyzer becomes
ready for use.

Cartridge Quality Assurance

Cartridge quality assurance is achieved primarily by two methods:
1. Testing critical components of the cartridge before assembly
2. Downloading all cartridge data for investigation

Testing Cartridge Components

The critical components of the cartridge, sensor card and PC Solutions, are
designed for testing before cartridge assembly. Unlike sensors in other
cartridge-based systems, the GEM Premier 3000 sensor card can tolerate
many cycles of hydration and drying without any deterioration in sensor
performance. Furthermore, the fluid opening valve in the reagent bag is
configured for sampling without affecting its solution composition.
All manufactured sensor cards are tested in an elaborate and automated
computer-controlled testing system. The test utilizes PC Solutions similar
to those in the cartridge, plus an additional test solution for performance
evaluation. Only the sensor cards that pass stringent criteria based on
sensitivity (slope), accuracy, and drift are assembled in cartridges.
A statistically significant number of reagent bags from a production lot are
tested for quality assurance and for assigning analyte values. The value
assignment process is traceable to National Institute of Standards and
Technology (NIST) standards. The assigned concentration values are
included in the cartridge bar-code.

Recording Cartridge Data

All system and cartridge operations are recorded on the instrument hard
drive. The recorded data is very comprehensive. For example, every sensor
mV, every sample or PC Solution measurement, and every check or alarm is
recorded. This information can be downloaded onto a floppy disk or directly
e-mailed from the instrument through the GEMweb feature. In case of
cartridge malfunction, the user can send the cartridge data to the complaint
investigation department at IL for analysis and failure determination.

Understanding iQM Operation

iQM operation is achieved by adding additional layers of checks to the
GEM Premier 3000 operation. iQM uses FPR to identify cartridge malfunctions
that are currently undetected by internal system and sensor checks.
iQM operation can be summarized as follows:
Monitors performance of the system in real-time
Identifies potential failure patterns
Automatically performs corrective actions
Automatically documents the failure and action taken

12

iQM operation begins following cartridge verification after warm-up.

External CVP is used to independently verify cartridge calibration.
Four ampoules of CVP (two levels for pH, blood gases, electrolytes and
metabolites, and two levels for hematocrit) must pass before the cartridge
can be used for sample reporting. If one or more of the analytes fails CVP
testing, those channels will not be available for sample analysis.
Once cartridge calibration is verified, iQM monitors the status of the calibration during cartridge use-life. Upon detecting a problem, the analyzer automatically performs corrective actions.
The following are examples of corrective actions:
Automatically performing a special rinse cycle, then subsequently verifing
cartridge function if micro-clots are detected
Permanently disabling a failed sensor if its functionality cannot be recovered
Rejecting a cartridge for process stability failure
Alerting the operator if an interfering substance in the sample is detected
During cartridge operation, the instrument automatically and continuously
performs various checks that can be categorized into four groups:
1. System checks
2. Sensor checks
3. FPR checks
4. Process Stability checks

System Checks
System checks include basic functionality of the instrument and the cartridge
Examples of these checks include:
Cartridge fluidic checks to assess sample integrity, detect the presence
of PC Solutions, and verify peristaltic pump functionality
Cartridge mechanical checks, such as proper operation of the distribution
valve and sampler arm
Instrument heater-block checks, such as heater-block temperature and
mV-output threshold from sensors (e.g., outputs at rail)
Instrument electronic checks, such as analog/digital electronic calibration
verification and communication between processors
Any failure in the system checks will lead to a corrective action. The corrective
action will include verification of the failure followed by one of the following
steps:
Cartridge rejected, in case of cartridge-related failure
Instrument operation halted, in case of instrument-related system failure

Sensor Checks
Sensor checks address functionality of the sensors. PC Solutions A, B and C
are automatically brought into the sensor card at various intervals to verify
sensor operation. The solution that is residing in the sensor card is measured
and the drift is determined. (Drift is the delta between the measured and
expected value.) The expected values are obtained from either the cartridge
bar-code or from prior PC Solution measurements.
Sensor checks are performed with the following frequencies:
PC Solution B is measured most frequently. Fresh B is measured, at
minimum, every 30 minutes or after every sample. Furthermore, PC
solution B is measured every 30 seconds while residing in the sensor
card in between fresh B measurements.

13

PC Solution A is measured, at minimum, every 4 hours. All sensor slope

values are also measured and checked. Slope, which is an indicator
of sensor sensitivity, must be within allowable limits.
PC Solution C is measured, at minimum, every 24 hours. PC Solution C is
primarily used for measuring low-level oxygen and for verifying performance
of the pH and PCO2 sensors. It is also used for conditioning the glucose
and lactate inner membranes, which provide rejection of interference
compounds present in blood.
It should be noted that the PC Solution B and PC Solution A measurements
are transparent to the operator. The measurement can be interrupted at any
time to run a sample (except the A measurement within the first 4 hours of
cartridge life, which is uninterruptible). The PC Solution C measurement is
uninterruptible, but the operator has the option of selecting the exact time
of day for performing PC Solution C.
The process of verifying sensor operation by measuring PC Solutions is very
similar to the process of sample measurement. As indicated in Figure 2, the
sample path and the PC Solution path into the sensor card is identical. In
addition, the algorithms for calculating the PC Solution values are similar to
those for sample calculations using prior calibration values. After performing
sensor checks, the analyzer conducts a number of actions. If all measured
values are within allowable limits, the sensors will be calibrated and, as a
result, the drifts will be set to zero.
If any measurement or slope value is outside the allowable limits, the
following corrective actions will take place:
Parameter result in subsequent sample report will be suppressed
If the failure fits with a known pattern, specific corrective actions will
follow (see FPR checks). If not, fresh solution will be brought into the
sensor card for up to two more times and measured
If the failure persists in two consecutive C measurements, in three
consecutive A measurements, or 15 consecutive B measurements,
the failed parameter will be permanently disabled

Home Position

Figure 2. Similarities between sample path and PC Solution path

14

FPR Checks
FPR checks were developed through years of investigating field complaints.
As indicted in the previous section, GEM analyzers (GEM Premier 3000 as
well as the GEM Premier) provide comprehensive information about used
cartridges. This information, plus in-house investigations about the cause
of failures, provided the background for the FPR checks.
Two distinct failure patterns were identified: micro-clot-related failures and
certain sensor malfunctions that are not well detected by other internal
checks or by external QC. Furthermore, certain interference patterns were
also specified.

Micro-Clot Patterns
Micro-clots are small blood clots or fibrin strands that adhere to a sensor and
induce a change in sensor characteristics, such as sluggish response or
sensitivity change. Micro-clot patterns are distinct for various sensors. Sensorcheck failures (drift errors) are used to identify the presence of micro-clots.
In the event of micro-clot pattern detection, iQM automatically initiates a special
rinse cycle, using PC Solution C (PC Solution B is used for normal rinse). PC
Solution C has added surfactants for better clot removal. Upon completion of
the rinse cycle, iQM checks for a clot pattern on the affected sensor. If a clot
pattern still remains, the affected sensor is disabled. If a clot pattern is not
detected, the sensor status becomes green (ready for measurement).
Figure 3 illustrates an example of micro-clot formation on the oxygen sensor
and the pursuing corrective actions. In this example, the detected pattern
60

Rinse Ready
B

PO2 Output (mV)

40

B B

50

B B

Sample # 17

30

Sample # 19

Sample # 18

20

Sample # 21
Sample # 20

Clot Removal Verified

Clot Detected
10

Clot Verified

0
48.5

49.5

50.5

Cartridge Use Life (hr)

200

Measured B

PO2 (mmHg)

180

B drift

160

Micro-Clot Pattern Detected

Corrected

- Negative B drift error after sample

- Positive A drift error

140

Measured A

A drift

120

100
100

110

120

130

A & B Sequence

Figure 3. Example of Micro-Clot Pattern for PO2 Sensor

140

15

was a negative drift error in the PC Solution B oxygen check after sample
#19, followed by a positive drift error in the oxygen PC Solution A check
which was performed automatically and immediately after the PC Solution B
check failure. After confirming the presence of a micro-clot (by the PC
Solution A check), the special rinse cycle was initiated. Following the rinse
cycle, the removal of the clot was verified by re-running PC Solution B and A
checks. In this example, the entire process of detection, clot removal and
verification took about 10 minutes.

Sensor Malfunction Patterns

Only pH, PCO2 and PO2 sensors need an additional pattern check for
identifying malfunctions. Existing sensor checks are found adequate for
detecting any malfunction in other sensors. It should be noted that normal
sensor malfunctions for pH and PCO2 sensors are identified with the existing
sensor checks. The specific malfunctions for which iQM is checking in these
sensors are very rare and very slow in progression. Therefore, the daily PC
Solution C check is adequate in detecting these malfunctions.
In case of a sensor malfunction pattern, the affected sensor is permanently
disabled by iQM.

Interference Patterns
Two interference patterns are checked, one for positively charged lipophilic
compounds, and one for negatively charged lipophilic compounds. These
compounds can cause false readings for a few of the ion-selective electrodes.
Benzalkonium Chloride is a good example of a positively charged lipophilic
compound that can cause false-elevated readings for sodium and ionized
calcium.
Sodium Thiopental is an example of a negatively charged lipophilic compound.
For an interference pattern to be positively identified, all conditions must be
met. In the event of interference pattern detection in a sample, iQM notifies
the user.
Figure 4 illustrates an example of Benzalkonium interference on the ionized
calcium sensor. The detected pattern was a positive drift error in the ionized
calcium PC Solution B check, along with certain drift movements in the
sodium and potassium sensors. Following the interference detection, PC
Solution B is performed at a higher than normal frequency (every 3 minutes
versus every half-hour).
50

Sample with Benzalkonium

Sensor Output (mV)

45

Sample
40

35

B drift error

30
21

21.5

22

22.5

23

23.5

24

Cartridge Life (hours)

Figure 4. Example of Interference Caused by a Positively Charged Lipophilic

Compound

16

Process Stability Check

Process Stability check is a method of verifying the PC Solution stability
throughout the cartridge use-life. The measured oxygen in PC Solution A
during the use-life is compared to the very first measured PC Solution A
during warm-up. The delta has to be within allowable limits.
The PO2 in PC Solution A is used for the process stability check because:
Oxygen is considered the most sensitive parameter for detecting
deterioration in PC Solutions, since they have no oxygen-buffering
capacity
The process of measuring oxygen in PC Solution A utilizes all three PC
Solutions, so that deterioration in any of the PC Solutions will affect
the measured oxygen in PC Solution A.
In the event of process stability failure, the cartridge will be rejected.

III. Statistical Validation

17

Statistical Evaluation of Drift Limits2

As indicted in the previous sections, drift limits are used as a trigger for
sensor and FPR checks, and for subsequent iQM corrective actions.
Therefore, drift limits should be optimized for high probability of error detection
and low probability of false rejection. This section explains how statistical
control methods are used for evaluation of drift limits.

Statistical Method2
Drift limits on PC Solutions A and B can be characterized as a single
measurement of a control material. Statistical control methods are then
used to develop probabilities for error detection and false rejection. This
approach allows a comparison of performance expected for iQM with
performance of traditional QC procedures.
The method is as follows:
Define the quality requirement in terms of total allowable error (TEa)
- pH, PCO2, Na+, and K+ = CLIA 88 limits
- PO2 = 10%
- Ca++= 9%
- Glucose = 12% or 12 mg/dL
- Lactate = 15% or 0.4 mmol/L
- Hematocrit = 2%
Define method performance
- Mean and SD values are obtained from the data collected from
24 cartridges representing a wide variety of uses from customers
in the field and in-house tests
Predict QC performance
- Calculate Method Sigma = TEa/SD
- Calculate Control Limit = Drift Limit/SD
- Determine probability of false rejection (Pfr) from normal probability
distribution (from tables of areas under normal curve, or z charts)
Pfr = Prob(z Control Limit)
- Determine probability of error detection (Ped) with 95% confidence
from normal probability distribution
Ped = 1 Prob(z (Method Sigma Control Limit 1.65))
- Calculate Average Run Length for rejectable quality
ARLr = 1/Ped
- Determine average detection time (unit of time for detecting error that
can be compared to traditional QC)
Average detection time = ARLr x sampling time
As specified previously, sampling time for Solution A is between 1 to 4 hours
and for Solution B is between 3 and 30 minutes (3 minutes when there is a
sample in between B measurements). For a given TEa, the drift limits must
provide a high probability of error detection (Ped 1) and a low probability
of false rejection (Pfr 0). Tables 1 and 2 provide a summary of the analysis
of the drift limits for PC Solutions A and B.
Results of the drift limit analysis indicate that the probability of false rejection is
close to zero for all parameters in PC Solutions A and B. PC Solution B is the
primary means for error detection because of high measurement frequency.
Probability of error detection is high in the PC Solution B. Even for sodium
and glucose with low Ped values, the average error detection time is within
10 minutes.

18

Table 1. Statistical analysis of drift limits for PC Solution A

Mean
SD
TEa
Method Sigma
Drift Limit
Control Limit
Pfr
z for Ped
Ped
ARLr
ATDr (4 hr)

pH
6.90
0.006
0.04
6.4
0.03
4.8
0.000
-0.06
0.476
2.10
8.4

PCO2
64
1.25
5
4.0
4
3.2
0.001
-0.85
0.198
5.05
20.2

PO2
119
1.65
11.89
7.2
6
3.6
0.000
1.92
0.973
1.03
4.1

Na +
102
0.73
4.0
5.5
3.0
4.1
0.000
-0.27
0.393
2.54
10.2

K+
6.7
0.043
0.50
11.5
0.30
6.9
0.000
2.96
0.998
1.00
4.0

Ca++
2.63
0.034
0.25
7.4
0.15
4.4
0.000
1.29
0.902
1.11
4.4

Glu
141
3.71
17.0
4.6
14
3.8
0.000
-0.84
0.200
4.99
20.0

Lac
3.1
0.09
0.5
5.6
0.30
3.3
0.000
0.58
0.718
1.39
5.6

Hct
25
0.05
2
39.2
1
19.6
0.000
17.96
1.000
1.00
4.0

Lac
mmol/L
0.0
0.05
0.4
9.7
0.3
6.4
0.000
0.942
1.06
3

Hct
%
11
0.09
2
23.1
1
11.6
0.000
1.000
1.00
3

Table 2. Statistical analysis of drift limits for PC Solution B

Mean
SD
TEa
Method Sigma
Drift Limit
Control Limit
Pfr
Ped
ARLr
Average Detection Time (min)*

pH
unit
7.40
0.002
0.04
20.0
0.03
15.0
0.000
1.000
1.00
3

PCO2
mmHg
32
0.41
5
12.3
3
7.3
0.000
1.000
1.00
3

PO2
mmHg
176
1.37
18
13.6
10
7.3
0.000
1.000
1.00
3

Na +
mmol/L
144
0.90
4
4.5
3
3.3
0.001
0.298
3.36
10

K+
mmol/L
3.6
0.011
0.5
45.9
0.3
27.5
0.000
1.000
1.00
3

Ca++
mmol/L
1.16
0.009
0.10
11.1
0.06
6.8
0.000
1.000
1.00
3

Glu
mg/dL
0
1.4
12
8.9
10
7.3
0.000
0.425
2.35
7

*Compare with a typical QC program where QC is performed only every 8 hours

Evaluation of Cartridge Malfunction Detection

The effectiveness and efficacy of iQM error detection was investigated using
the existing field complaint data files for the GEM Premier 3000 with Phase II
cartridge configuration. (Phase II cartridge configuration has the capability
to include glucose and lactate). It should be noted that the FPR checks were
originally developed from field and in-house data using the GEM Premier
3000 and Phase I cartridge configurations.
The investigation focused on all reported field QC failures from US accounts
beginning with the release date of Phase II cartridges in June 2001, to the end
of March 2002. (Only QC failures were selected because iQM is expected to
detect cartridge malfunctions currently identified by running external quality
controls.) Approximately 7000 cartridges were shipped to U.S. GEM Premier
3000 accounts during this period.
Summary data is outlined below.
A total of 164 QC-related cartridge failures were reported to IL
From 164 reported QC failures, IL obtained 144 data disks
From 144 obtained data disks, IL confirmed 117 real QC failures
(others either had no QC failure or had failures other than QC)
From 117 confirmed QC failures, 79 passed initial QC out of warm-up,
but failed QC later

19

Data from the 79 GEM Premier 3000 cartridges that had confirmed QC
failures during cartridge use-life was extensively analyzed. The FPR checks
were applied to the data to determine if iQM could detect any malfunction.
The results are summarized below.
68 of the data files exhibited micro-clot failure patterns.
- PO2: 46
- Hematocrit: 16
- Ca++: 2
- pH: 1
- PO2 and Hematocrit: 2
- PO2, Na+ and Ca++: 1
6 showed a sensor failure pattern.
- PCO2: 4
- pH: 2
3 demonstrated one QC level at the limit of the acceptable range
with a subsequent QC failure.
- PO2: 2
- Glucose: 1
2 had no identifiable failure pattern; both QC failures were marginal
and occurred at only one level.
- Na+: 1
- Ca++: 1
This analysis indicates that FPR was able to detect malfunction in 77 out
of 79 reported QC failures. It is interesting to note that in most cases, FPR
flagged the failure immediately after the sample that caused the malfunction;
while the operator typically became aware of the problem only later when
the QC was run and failed. In some cases, QC had been run several hours
after occurrence of the malfunction. Furthermore, no false-positive flag was
generated by FPR in all 79 investigated cases.
No obvious malfunction was detected in the data from the two cartridges
that FPR did not flag but had reported QC failure. All system parameters
were found within specifications. Blood results for the parameter with reported QC failure were examined and found reasonable. In both cases, the
reported QC failures were marginal and at one level only. It was concluded
that the reported QC failures in those two cartridges could not represent
a serious cartridge malfunction. The failures could be considered as a
false-positive QC failure.

iQM Error Detection During Limited Distribution

During a limited distribution phase, iQM was implemented in 20 hospital sites.
These selected sites represented high volume locations, many with a
history of pre-analytical issues tha may affect patient results. The issues
included samples containing micro-clots and samples contaminated with
Benzalkonium.
Data from 42 cartridges from the various clinical sites, using various cartridge
types, were analyzed at IL. Cartridge files with reported iQM failures were
of particular interest in the investigation. Two-thirds of the cartridge data
collected showed no cartridge issue, while one-third had at least one
reported problem during cartridge use-life. The iQM Corrective Action
Report was used to identify any cartridge issue. Additionally, the sensor
output data files, the sensor calibration data files, and the system logs
were examined to confirm the validity of iQM reported errors.

20

iQM correctly flagged the presence of micro-clots in 11 cartridges. Five of

the micro-clots were successfully rinsed out, while 6 could not be removed,
and the analyzer automatically disabled the affected sensors. There were
5 flags for Benzalkonium interference on the sodium and ionized calcium
measurements. The sodium and ionized calcium values in those flagged
samples were abnormally high. In all cases, iQM flagged the problem within
minutes of the occurrence.
iQM was very effective in identifying cartridge issues promptly. The automatic
corrective action function of iQM was effective in removing about one-half
of the detected micro-clots.

Summary
iQM is an active, real-time quality process control program that allows for
immediate error detection and correction, thus further enhancing quality
assurance in the GEM Premier 3000 analyzer.

IV. Glossary of Acronyms

ARLr:

Average Run Length for Rejectable Quality

CAR:

Corrective Action Report

CVP:

Calibration Validation Product

EQC:

Electronic Quality Control

FPR:

Failure Pattern Recognition

iQM:

Intelligent Quality Management

IL:

Instrumentation Laboratory

NCCLS: National Committee for Clinical Laboratory Standards

NIST:

National Institute of Standards and Technology

PC:

Process Control

Ped:

Probability of Error Detection

Pfr:

Probability of False Rejection

QC:

Quality Control

SD:

Standard Deviation

TEa:

Total Allowable Error

21

22

Appendix
Note: the following information is based on data obtained during the development of
iQM cartridges; please refer to the complete GEM Premier 3000 Operators Manual
(Revison 3) for complete specifications on all available GEM Premier 3000 cartridges.

AI. Interferences
The following substances can potentially interfere with sample analysis:
Benzalkonium Chloride and Benzalkonium Heparin: Arterial lines and
sampling devices coated with these substances may interfere with Na+
and Ca++ determinations, causing falsely elevated Na+ and Ca++ readings.
Following sample analysis, and analysis of PC Solution B, if Benzalkonium
Chloride or Benzalkonium Heparin patterns are detected, the following
message will be displayed on the analyzer, and will persist until acknowledged
by the operator:

Sensor Interference Detected for Na and iCa

on last sample likely due to Benzalkonium
The GEM Premier 3000 offers the operator the ability to enable flagging of patient
results if an interference pattern is detected. In addition, this option, when enabled,
delays the reporting of results until PC Solution B is evaluated following sample
analysis. If flagging of patient results is enabled, the following message (plus
progress bar) will be presented while the post-analysis PC Solution B check is
underway:

Checking for presence of interference and micro clots

This message will remain displayed until the analysis of PC Solution B, is
complete. If an interfering substance pattern is detected, the affected blood
result(s) will be flagged. In addition, the analyzer will beep three times to
alert the operator. The following message disappears only after operator
acknowledgment:

Sensor Interference Detected for Na and iCa

on last sample likely due to Benzalkonium
Thiopental sodium may interfere with the Na+, K+, PCO2 and Ca++ readings.
Thiopental Sodium is also known by other names, including: Thiomebumal
Sodium, Penthiobarbital Sodium, Thiopentone Sodium, Thionembutatal,
Pentothal Sodium, Nesdonal Sodium, Intraval Sodium, Traoanal, and
Thiothal Sodium.
Following sample analysis and analysis of PC Solution B, if the associated
pattern is detected in PC Solution B, the following message will be displayed
on the analyzer, and will persist until operator acknowledgment:

Sensor Interference Detected for xxxxx on last sample

(where xxxx is the analyte or analytes affected)
When flagging of patient results for an interference is enabled, the following
message (plus progress bar) will be presented while the post analysis PC
Solution B check is underway:

Checking for presence of interference and micro clots

This message will remain displayed until the PC Solution B is complete. At
that point, processing will continue until results are displayed. If Thiopental
Sodium is detected, the affected blood result(s) will be flagged. In addition,
the analyzer will beep three times to alert the operator. The following
message disappears only after operator acknowledgment.

Sensor Interference Detected for xxxxx on last sample

(where xxxx is the analyte or analytes affected)

23

AII. Performance Characteristics Summary

GEM CVP Precision
Precision data were generated at IL using GEM CVP: 2 levels for pH, blood
gases, electrolytes and metabolites and 2 levels for Hematocrit.
Based on NCCLS guidelines, the verification material levels were run in
singlet once a day for 14 days (twice on day 1) for a total of 15 replicates
on each of 9 different GEM Premier 3000 instruments (N=135). The table
below lists the combined results of the 9 instruments.
Note: SD is used for pH since differences are so small that %CV would be misleading.

GEM CVP Level 1:

Parameter
pH
PCO2 (mmHg)
PO2 (mmHg)
Na+ (mmol/L)
K+ (mmol/L)
Ca++ (mmol/L)
Glucose (mg/dL)
Lactate (mmol/L)

Mean
7.200
70.8
54.5
129.3
2.90
1.493
46.1
0.93

Day-to-Day %CV (or SD)

0.005 (SD)
1.39
4.97
0.46
0.25
0.95
2.23
4.73

Total %CV (or SD)

0.007 (SD)
1.63
5.16
0.55
0.70
1.26
2.99
4.87

Mean
7.640
29.9
148.2
158.7
6.46
0.486
192.8
5.54

Day-to-Day %CV (or SD)

0.002 (SD)
1.78
1.33
0.44
0.75
1.15
1.67
1.85

Total %CV (or SD)

0.003 (SD)
1.91
1.93
0.56
0.98
2.06
1.78
2.19

Mean
23.4

Day-to-Day %CV
2.14

Total %CV
2.11

Mean
43.8

Day-to-Day %CV
1.21

Total %CV
1.23

GEM CVP Level 2:

Parameter
pH
PCO2 (mmHg)
PO2 (mmHg)
Na+ (mmol/L)
K+ (mmol/L)
Ca++ (mmol/L)
Glucose (mg/dL)
Lactate (mmol/L)

GEM CVP Level 3:

Parameter
Hematocrit (%)

GEM CVP Level 4:

Parameter
Hematocrit (%)

24

AIII. Method Comparison

Arterial, venous, heart bypass and liver transplant blood samples were
obtained from patients using heparinized syringes. The table and graphs
below demonstrate that the GEM Premier 3000 using an iQM cartridge is
statistically similar in performance to a reference analyzer.
Analyte
pH
PCO2 (mmHg)
PO2 (mmHg)
Na+ (mmol/L)
K+ (mmol/L)
Ca++ (mmol/L)
Glucose (mg/dL)
Lactate (mmol/L)
Hct (%)

N
281
282
282
271
271
264
283
279
284

Slope
1.0802
1.0674
0.9715
0.9801
0.9743
0.9620
1.0111
0.9382
0.9983

Intercept
-0.581
-2.380
6.990
2.926
-0.061
0.0396
8.868
0.163
-0.800

r
0.9917
0.9839
0.9988
0.9584
0.9871
0.9923
0.9860
0.9957
0.9600

Sample Range
7.129-7.559
25.3-87.5
26-489
119-148
3.2-7.4
0.86-1.56
66-389
0.49-15.07
17-56

pH Method Comparison
7.60

pH Units iQM GEM

7.50

7.40

7.30
7.20

y = 1.0802x - 0.581
r = 0.9917
n = 281

7.10
7.200

7.100

7.300

7.400

7.500

7.600

pH Units Reference

PCO2 Method Comparison

100
90
mmHg iQM GEM

80
70
60
50
40

y = 1.674x - 2.38
r = 0.9839
n = 282

30
20
20.0

30.0

40.0

50.0

60.0
mmHg Reference

70.0

80.0

90.0

100.0

25

PO2 Method Comparison

600

mmHg iQM GEM

500
400
300
200
y = 0.9715x - 6.99
r = 0.9988
n = 282

100
0
100

200

300

400

500

600

mmHg Reference

Na+ Method Comparison

155
150

mmol/L iQM GEM

145
140
135
130
125
y = 0.9801x - 2.93
r = 0.9584
n = 271

120
115
115

120

125

130

135

140

145

150

15 5

mmol/L Reference

K Method Comparison

mmol/L iQM GEM

7.5
6.5
5.5
4.5
y = 0.9743x - 0.06
r = 0.9871
n = 271

3.5
2.5
2.5

3.5

4.5

5.5
mmol/L Reference

6.5

7.5

26

Ca++ Method Comparison

1.7
1.6

mmol/L iQM GEM

1.5
1.4
1.3
1.2
1.1
1
0.9

y = 0.962x - 0.0396
r = 0.9923
n = 264

0.8
0.7
0.9

0.7

1.1
1.3
mmol/L Reference

1.5

1.7

Glucose Method Comparison

450
400

mg/dL iQM GEM

350
300
250
200
150
y = 1.0111x - 8.87
r = 0.9860
n = 283

100
50
50

150

250

350

450

mg/dL Reference

Lactate Method Comparison

16.0
14.0
mmol/L iQM GEM

12.0
10.0
8.0
6.0
4.0
y = 0.9382x - 0.1633
r = 0.9957
n = 279

2.0
0.0
0.00

5.00

10.00
mmol/L Reference

15.00

27

Hematocrit Method Comparison

60
55
50
% iQM GEM

45
40
35
30
25

y = 0.9983x - 0.7999
r = 0.9600
n = 284

20
15
15

25

35
% Reference

45

55

28

AIV. iQM Reports

iQM Delta Charts
An iQM Delta Chart is generated for each Process Control Solution and
analyte combination. A Delta Chart, depicted in Figure 5, includes:
1. Analyzer (GEM Premier 3000)
2. Analyte name
3. Analyzer serial number and name, if the analyzer is given a unique
name by the facility
4. Nominal target value for the analyte
5. Report month and year
6. PC Solution
7. The number of times the PC Solution was run each day (If the number
of points exceeds 99, 99 will be displayed)
8. The maximum delta between the Process Control Solution target value
and the actual measured value for the selected analyte each day,
represented by a short horizontal line
9. The mean delta between the Process Control Solution target value
and the actual measured values for the selected analyte each day,
depicted by a round, bolded dot
10. The maximum delta between the Process Control Solution target value
and the actual measured value for the selected analyte each day,
represented by a short horizontal line
11. A vertical line, which connects the maximum, mean and minimum
delta points
12. The maximum allowable delta for the analyte
13. The minimum allowable delta for the analyte
14. The date of cartridge insertion (indicated by an arrow), along with the
cartridge lot number

1
2

3
4

7
12
11

8
9
10

13
14
15

Figure 5. iQM Delta Chart: Components.

5
6

29

Delta points outside the designated limits will not be shown on the Delta
Chart, but will be included in the iQM Corrective Action Report. This is
because the GEM Premier 3000 requires the use of static graphs (graphs
with ranges that do not change).
Please refer to Figure 6 for graphical representations of the following situations:
1. It is possible to have the minimum and maximum delta values coincide
with one another and the mean delta result. This happens when:
The same delta value is obtained for an analyte each time a PC
Solution is run.
Only one delta value is obtained for an analyte through the course of
a day, which is expected for PC Solution C.
2. A minimum or maximum delta result (represented by a horizontal line) may
coincide with a daily mean delta result. An example of this occurring is:
K+ is measured in PC Solution B 90 times in one day
89 x the delta value = 0
1 x the delta value = +0.01
In this case, the mean and minimum delta values will be the same (0).
The horizontal line representing the minimum delta value will intersect the
round, bolded dot representing the mean value. Furthermore, since the
minimum delta value is 0, the horizontal line representing the minimum delta
result will coincide with the 0 horizontal line on the graph.
3. If the maximum and/or minimum delta obtained during a day coincides
with the maximum or minimum allowable delta limits, then the horizontal
line will blend with the upper/lower graph line limits.

Figure 6. iQM Delta Chart: Specific Delta Value Representations.

30

iQM Corrective Action Report (CAR)

The Corrective Action Report contains information for significant events that
occur during the cartridge on-board use-life. The CAR, shown in Figure 7
includes:
1. Analyzer, analyzer serial number and name
2. Month and year
3. Date and time event occurred
4. Cartridge Lot Number
5. Detected failure description
6. Operator ID, if entered (if the function included operator interaction)
7. Corrective action description
8. Corrective action result
9. A Cartridge Removal entry that lists the total number of Process Control
solution adjustments made during the cartridge use-life. The adjustments
represent the total number of minor drift errors that are corrected by
reanalyzing Process Control solution B. Events, such as interfering
substance detection, micro-clot detection, and fatal errors, will be listed
and described as individual events on the log.

1
3

5
6
7

Figure 7. iQM Corrective Action Report-Components.

Calibration Validation Product Reports

Monthly CVP reports can be printed. A CVP report contains results for all
CVP ampoules run within the specified month.

31

References
1. GEM Premier 3000 Operating Software, Volume 2, Cartridge Internal
Operations, Software Version 5.2, Revision 2.7.
2. Westgard JO, Fallon KD, Mansouri S. Validation of iQM Active Process
Control Technology Point of Care, 2003; Vol 2 (1): 1-7.
3. Thiopental Interference in GEM Blood Gas Systems by H. Hanford,
Technical Note #112-833.
4. Tietz Fundamental of Clinical Chemistry by C.A. Burtis and E.R.
Ashwood, Forth Edition (1996), page 828.

Biographies of Authors
Kevin Fallon, Ph.D.
Dr. Fallon was a member of Instrumentation Laboratorys scientific staff
for 25 years, most recently serving as the Director of Scientific Affairs.
Additionally, he served on the faculty of the University of Texas Medical
School and was Director of the STAT Labs at Hermann Hospital in Houston.
Dr. Fallon continues to offer his expertise to IL on a consulting basis.

Sohrab Mansouri, Ph.D.

Dr. Mansouri has over 20 years of experience in the design, development and
implementation of chemical and biochemical sensors for medical applications.
He has spent the past 13 years advancing the GEM technology and is
responsible for the development of the Premier and Premier 3000 cartridges.

 2003 Instrumentation Laboratory - Printed in Italy - Speed 2000 - 06/03

Part. No 9808619