Multi Frequency Phase Fluorimetry (MFPF) for Oxygen Partial Pressure Measurement: Ex Vivo Validation by Polarographic Clark-Type Electrode

Background Measurement of partial pressure of oxygen (PO2) at high temporal resolution remains a technological challenge. This study introduces a novel PO2 sensing technology based on Multi-Frequency Phase Fluorimetry (MFPF). The aim was to validate MFPF against polarographic Clark-type electrode (CTE) PO2 measurements. Methodology/Principal Findings MFPF technology was first investigated in N = 8 anaesthetised pigs at FIO2 of 0.21, 0.4, 0.6, 0.8 and 1.0. At each FIO2 level, blood samples were withdrawn and PO2 was measured in vitro with MFPF using two FOXY-AL300 probes immediately followed by CTE measurement. Secondly, MFPF-PO2 readings were compared to CTE in an artificial circulatory setup (human packed red blood cells, haematocrit of 30%). The impacts of temperature (20, 30, 40°C) and blood flow (0.8, 1.6, 2.4, 3.2, 4.0 L min−1) on MFPF-PO2 measurements were assessed. MFPF response time in the gas- and blood-phase was determined. Porcine MFPF-PO2 ranged from 63 to 749 mmHg; the corresponding CTE samples from 43 to 712 mmHg. Linear regression: CTE = 15.59+1.18*MFPF (R2 = 0.93; P<0.0001). Bland Altman analysis: meandiff 69.2 mmHg, rangediff -50.1/215.6 mmHg, 1.96-SD limits -56.3/194.8 mmHg. In artificial circulatory setup, MFPF-PO2 ranged from 20 to 567 mmHg and CTE samples from 11 to 575 mmHg. Linear regression: CTE = −8.73+1.05*MFPF (R2 = 0.99; P<0.0001). Bland-Altman analysis: meandiff 6.6 mmHg, rangediff -9.7/20.5 mmHg, 1.96-SD limits -12.7/25.8 mmHg. Differences between MFPF and CTE-PO2 due to variations of temperature were less than 6 mmHg (range 0–140 mmHg) and less than 35 mmHg (range 140–750 mmHg); differences due to variations in blood flow were less than 15 mmHg (all P-values>0.05). MFPF response-time (monoexponential) was 1.48±0.26 s for the gas-phase and 1.51±0.20 s for the blood-phase. Conclusions/Significance MFPF-derived PO2 readings were reproducible and showed excellent correlation and good agreement with Clark-type electrode-based PO2 measurements. There was no relevant impact of temperature and blood flow upon MFPF-PO2 measurements. The response time of the MFPF FOXY-AL300 probe was adequate for real-time sensing in the blood phase.


Introduction
The polarographic Clark-type electrode remains the gold standard for oxygen partial pressure (P O2 ) measurement to date, and is the underlying electrochemical principle of conventional blood gas P O2 analysis [1,2,3,4]. Moreover, polarographic P O2 probes, such as the Licox probe, are used for the continuous monitoring of oxygen. These probes provide favourable properties, including high accuracy and stability; however, they lack high temporal resolution [5]. Several publications reveal that oxygen levels may largely fluctuate over time in the systemic (e.g., thoracic aorta) or capillary circulation (e.g. vasomotion, brain resting state) [6,7,8,9,10] due to oxygen consumption and delivery. Therefore, novel, fast oxygen-sensing technologies are required for the investigation of these physiologic or pathophysiologic phenomena [11,12].
Recently, a novel technology for fast oxygen sensing based on Multi Frequency Phase Fluorimetry (MFPF), termed NeoFoxH (OceanOptics, Dunedin, USA; www.oceanoptics.com), was released. This technology is suitable for 10 Hz oxygen measurements using fluorescent dye sensors such as the FOXY-AL300 (Ocea-nOptics, Dunedin, USA). This indwelling, uncoated ruthenium tipped fibre optic probe allows for rapid measurement of oxygen within blood or tissue. The technology, however, has not yet been compared to conventional polarographic Clark-type electrode (CTE)-based P O2 measurements [13,14].
This study aimed to validate MFPF-P O2 measurements versus CTE-based measurements of oxygen partial pressure. Firstly, in a porcine lung model, MFPF-P O2 was measured in vitro without the interference of blood flow using two FOXY-AL300 probes (OceanOptics, Dunedin, USA) and was then compared to CTE blood gas P O2 analysis. Secondly, for further validation, we developed an artificial extracorporeal circulation setup primed with a mixture of crystalloid, colloid and human packed red cells to imitate in vivo conditions. In this setup, MFPF-P O2 readings were assessed and then compared to conventionally measured P O2 by CTE. Furthermore, in the artificial circulation setup, temperature and blood flow were randomly varied in order to evaluate their impact on MFPF-P O2 measurements. In addition, MFPF FOXY-AL300 (OceanOptics, Dunedin, USA) probe response times were determined in the gas-and blood-phases to further assess this novel technology. In summary, the overall aim could be achieved.

MFPF Technological Principle
The MFPF platform (NeoFoxH, OceanOptics, Dundin, USA) was mounted with the indwelling FOXY-AL300 (OceanOptics, Dunedin, USA) aluminium-jacketed optical fibre probe (500 mm outer and 300 mm core diameter) with an uncoated ruthenium complex at the tip to assess luminescence lifetime in the region surrounding the fluorescent dye, which was roughly equal to the fibre-optic tip surface area (about 0.3 mm 2 ). MFPF assesses luminescence lifetime by modulation of light excitation frequency where the oxygen dependent phase shift of the emission is detected. Lifetime is related to phase shift by the equa- Calibration of the FOXY-AL300 Oxygen Probe P O2 probes were calibrated at different oxygen levels and temperatures in a closed thermostat controlled gas-tight chamber according to the manufacturer's instructions. The chamber was purged with pure nitrogen (P O2 = 0 mmHg), compressed air (P O2 = 158 mmHg) and pure oxygen (P O2 = 750 mmHg), respectively, in consideration of atmospheric pressure with a fixed laminar flow of 2 L min 21 at 20, 30 and 40uC. After stabilisation, the corresponding lifetimes were used to create a multi-temperature calibration table.

Ethics statement
This study was carried out in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institute of Health. The protocol was approved by the Animal State Care and Use Committee of the Rhineland Palatinate, Germany (Permit Number: G07-10-013). All surgery was performed under deep anaesthesia, and all efforts were made to minimise suffering [15]. Animal experiments were performed at the Department of Anesthesiology, Medical Center of the Johannes Gutenberg-University, Mainz, Germany and laboratory experiments at the Department of Anesthesiology and Pain Therapy, Inselspital, Bern University Hospital, and University of Bern, Switzerland.

MFPF Validation in vitro using a porcine lung model
Eight piglets, Sus scrofa domestica (2062 kg), were investigated. After the induction of anaesthesia with a bolus of fentanyl 5 mg kg 21 i.v., propofol 2 mg kg 21 i.v. and single shot atracurium 0.5 mg kg 21 i.v. via an ear vein, the pigs were orotreachally intubated (ID 8.0 mm cuffed tubing) and ventilated in pressurecontrolled mode (AVEA, Viasis Healthcare, Höchberg, Germany). Vascular access was achieved by surgical cut down for the placement of an arterial blood pressure line in the femoral artery. Normocapnia (P aCO2 35-45 mmHg) was maintained with a tidal volume of 6-8 ml kg 21 BW 21 . Anaesthesia was maintained by continuous infusion of fentanyl (5 mg 21 kg 21 h 21 i.v.) and propofol (5-10 mg 21 kg 21 h 21 i.v.). During the experiment, F IO2 was adjusted (0.21, 0.4, 0.6, 0.8 and 1.0) to obtain P O2 readings over the whole normobaric measurement range. Core temperature by rectal probe was maintained within the physiologic range (37.0-38.5uC) using a heating blanket.
Blood samples (10 ml each) were withdrawn at each F IO2 level into heparinised gas tight syringes covered with opaque tape. For each sample, the tips of two FOXY-AL300 probes (OceanOptics, Dunedin, USA) were placed into the syringe outlet sealed with a rubber port to avoid air contact. After reaching steady state values, both MFPF-P O2 readings were simultaneously recorded for 2 min using the corresponding animal core temperature for the measurement. Then, the same blood sample was analysed by conventional polarographic Clark-type electrodes (RapidLab 415, BayerHealthcare, Leverkusen) immediately afterwards.

MFPF validation ex vivo using an artificial circulatory setup
In order to provide standardised conditions and to simulate in vivo conditions, an artificial circulatory system was set up. The apparatus was built of: a) two oxygenators (Jostra Quadrox D, Maquet Cardiopulmonary AG, Hirrlingen, Germany), b) 3/8 inch silicon tubing (Fumedica, Muri, Switzerland) with 3/8-3/8 inch connecting pieces (Luerlock, Fumedica, Muri, Switzerland), c) two multi-flow roller pumps (Stöckert, Munich, Germany), d) a heat exchanger (HCU-30, Maquet Cardiopulmonary AG, Hirrlingen, Germany) and e) an ambient light protected measurement chamber ( Figure 1). A switch valve between two independent closed circuits enabled the induction of a step change between highly oxygenated blood (F IO2 1.0, circuit 1) to oxygen free blood (F IO2 0.0, circuit 2) in order to assess the time constants of the FOXY-AL300 probes in blood-phase. The artificial circulatory system was primed with a mixture of crystalloid, colloid and human packed red cells to achieve a haematocrit of 30% and electrolyte values within the physiological limits. This setup allowed for control and variation of blood flow (by variation of rpm at the rollerpumps), temperature (by variation at the membrane equilibrator) and oxygen content (by P O2 variation at the blender).
In order to get measurements over the normobaric measurement range for the validation of MFPF against Clark-type electrode P O2 , F IO2 was set at random according to a computer generated list [16] from 0.0 to 1.0. To achieve an F IO2 of less than 0.21, a mixture of oxygen and nitrogen was used. For all measurements, duplicate MFPF P O2 readings (2 FOXY-AL300 probes) were recorded for 1 min duration alongside blood withdrawal for conventional polarographic Clark-type electrode P O2 analysis (ABL 700, Radiometer, Copenhagen, Denmark).
Measurements were performed according to the following protocols: 1) To assess the agreement of MFPF vs. Clark-type electrode P O2 at physiological body temperature conditions: at a temperature of 37uC, blood flow of 1.6 L min 21 and a sweep gas flow of 1 L min 21 ; 55 measurements were performed at varying P O2 levels. 2) To assess the impact of temperature: at a blood flow of 0.5 L min 21 and sweep gas flow 0.5 L min 21 , temperature was set at random according to a computer generated list [16] to 20, 30 or 40uC. For each temperature level, 15 measurements were performed at varying P O2 levels.
3) To assess the impact of blood flow: at a temperature of 37uC and a sweep gas flow of 2 L min 21 , blood flow was set at random according to a computer generated list [16] to 0.8, 1.6, 2.4, 3.2 and 4.0 L min 21 . For each blood flow level, 10 measurements were performed at varying P O2 levels.

Statistical Analysis
Descriptive and statistical data analysis was performed using SPSSH V.19 (IBM Inc., New York, USA). For metric variables, statistical measures such as mean and standard deviation (mean6SD) were calculated. Linear regression analysis upon P O2 values measured by MFPF and CTE were performed. Bias and precision values were calculated by the Bland-Altman method [17,18]. The comparability was presented in explorative manner by description of mean differences, range of differences and 1.96-SD limits (lower and upper limit of agreement) [19]. In addition, linear regression analysis of difference versus mean was performed and then tested, if the slope is significantly different from zero. Furthermore, intra-class correlation between the two MFPF probes was assessed. To test the influence of blood flow and temperature variation, a multiple linear regression model was fitted: magnitude of P O2 measurements [(CTE+MFPF)/2], temperature and blood flow as independent variables; differences of MFPF and CTE as dependent variable. The categories of temperature and blood flow were handled with dummy coding. The analyses of data were done in explorative manner and the outcome of a statistical test with a P-value,0.05 was interpreted as statistically significant.

MFPF validation ex vivo in artificial circulatory setup
In artificial circulatory setup, MFPF-P O2 measurements ranged from 20 to 567 mmHg and the corresponding CTE samples from 11 to 575 mmHg. Linear regression yielded the equation: CTE = 28.73+1.05*MFPF (R 2 = 0.99; P,0.0001) ( Figure 3A). Bland-Altman analysis calculated a mean difference (mean diff ) between CTE and MFPF P O2 measurements of 6.669.8 mmHg. Differences ranged from 29.7 to 20.5 mmHg within the 1.96-SD limits (lower limit -12.7, upper limit 25.8 mmHg) ( Figure 3B). Linear regression of the difference versus mean showed CTE-MFPF = 28.62+0.05*[(CTE+MFPF)/2] (R 2 = 0.78, slope significantly different from zero: P,0.0001). Linear regression in a subset of P O2 between 0 and 140 mmHg yielded CTE = 27.84+1.13*MFPF (R 2 = 0.99; P,0.0001) ( Figure 4A). Bland-Altman analysis of this sub-analysis showed a mean difference of 1.764.9 mmHg ( Figure 4B). The differences ranged Figure 1. Artificial Circulatory Setup. Artificial circulatory setup with two independent circuits filled with human packed red blood cells (haematocrit of 30%). A switching valve between the circuits enabled a step-change between highly oxygenated (circuit 1: purged with pure oxygen) and oxygen free blood (circuit 2: purged with nitrogen). Black arrows represent direction of blood flow. Adapting the settings of the rollerpumps and the heating-cooling device (heat exchanger) allowed blood-flow and temperature to be controlled. Via the O 2 /N 2 blenders, oxygen content could be adapted at fixed sweep gas flow over the oxygenators. Measurement chamber contained 1.) ports for insertion of MFPF probes (Foxy-AL 300); 2.) a temperature probe and 3.) a sampling port for Clark-typed based (CTE) P O2 analysis (ABL 700  In total, reproducibility of R 2 = 0.99 between the two MFPF probes (intra-class correlation) was depicted.

Impact of temperature and blood flow on MFPF P O2 measurements
The influence of the temperature upon the differences of CTE and MFPF P O2 measurements reached from 20.6 mmHg (temperature = 20uC) to 31.8 mmHg (temperature = 40uC) ( Table 1). The influence of blood flow on difference in P O2 of the two methods reached from 0.6 mmHg (blood flow = 1.6 L min 21 ) to 14.8 mmHg (blood flow = 0.8 L min 21 ) ( Table 1). The linear regression model with the magnitude of P O2 measurements [(CTE+MFPF)/2], temperature and blood flow as independent variables showed a significant influence upon the magnitude of P O2 measurements and a non-significant influence of temperature or blood flow upon the differences of P O2 measurements (dependent variable) ( Table 2).

MFPF FOXY-AL300 Response Time
MFPF in combination with the uncoated FOXY-AL300 probes showed a mean response time of 1.4860.26 s in gas-phase and 1.5160.20 s in blood-phase after a step-change from pure oxygen (P O2 = 749 mmHg) to pure nitrogen (P O2 = 0 mmHg) at a digital sampling rate of 10 Hz ( Figure 5). The monoexponential response time by Boltzmann fitting represents the time it takes the system to reach 63.2% of its final asymptotic value.

Discussion
Findings demonstrate MFPF technologys capacity for detecting changes in P O2 compared to Clark-type electrode (CTE) based P O2 measurements. For discrete samples, in vitro porcine blood  Validation of MFPF for Fast P O2 Measurements PLOS ONE | www.plosone.org phase measurements (R 2 = 0.93) did not give very tight limits of agreement. However, this study was not designed to validate MFPF as replacement for traditional in vitro blood gas analysis. For such a setup, we would have needed a configuration where the sensor and sample are part of a heated block with very tight control of both probe and blood sample temperature. The authors want to point out that the real application of MFPF is for flowing blood when the probe is placed in vivo -which was investigated by ex vivo measurements in an artificial circulatory setup filled with human blood. These findings demonstrate that MFPF technology provides high accuracy for P O2 measurements compared to CTE based P O2 measurements (R 2 = 0.99). The ex vivo measurements showed that MFPF technology accurately compensates for variation in temperature and that P O2 measurements are not affected by variation in blood flow. MFPF response time in combination with the indwelling FOXY-AL300 probes was 1.4860.26 s in the gas-phase and 1.5160.20 s in the blood-phase (step-change from pure oxygen to pure nitrogen).
In summary, MFPF technology allows for fast, accurate and valid P O2 readings over the normobaric P O2 range (0-749 mmHg) for ex vivo experiments, as MFPF P O2 measurements ranged between 11 and 749 mmHg. This fact is the most important, as novel oxygen sensing technologies should be validated over the whole measurement range.
The in vitro animal model (porcine blood) showed good correlation (R 2 = 0.93), although the data was determined in clusters of oxygen concentration (F IO2 0.21, 0.4, 0.6, 0.8, 1.0). The observed high bias with positive drift as well as the enlarged  standard deviation can be most likely explained by the systemic experimental error to use the animals' core temperature for the in vitro measurements. The main intention of this first series was to obtain readings without the influence of blood flow. Therefore, we focused on how to avoid air contact to the withdrawn blood and decided not to insert the temperature probe into the syringes. However, we did not expect such an impact of temperature, which presumably seems even more evident at high P O2 -levels ( Figure 2B). This finding can be explained by the underlying MFPF technology, as in fluorescent quenching, signal intensity increases at lower P O2 values. To fully understand the observed discrepancy of the in vitro (porcine blood) and ex vivo (human blood) results, we evaluated the difference of animal core temperature and blood temperature in the syringes two minutes after withdrawal post hoc in one further animal and found an error range of 0.9 to 1.6uC in ten repeated measures. Nevertheless, in the interesting physiologically hypoxic and normoxic (0-140 mmHg) P O2 range, porcine results show that MFPF technology provides good agreement ( Figure 2B) for discrete samples, although in our setup the technique was less accurate for high P O2 values. Due to the underlying technological principle precision of MFPF, P O2 values should increase within this physiologically interesting range (0-140 mmHg), which is supported by the present data. Thus, it seems unlikely that MFPF technology characteristics were responsible for this discrepancy, as intra-class correlation between the MFPF probes was extremely high (R 2 = 0.99) in porcine setup and MFPF measurements in the artificial circulatory setup (human blood) showed excellent agreement to CTE, even at high P O2 levels. From our point of view, the porcine results nicely demonstrate the high temperature sensitivity of MFPF measurements and further highlight the importance of correct temperature acquisition. Instead of repeating this series, we decided to develop a new artificial circulatory system for further validation, where we could also investigate the influence of temperature and blood flow by standardised variation of these physiological values.
Under ex vivo conditions in the artificial circulatory setup (human blood), we found excellent correlation (R 2 = 0.99) of MFPF and CTE P O2 with measurements equally distributed over the entire normobaric range (Figure 3). In general, MFPF values were slightly lower in comparison to CTE values. These findings were unexpected, as the MFPF technology, unlike polarography, does not consume any oxygen. These results might be explained by the fact that the blood-gas-analysers used allow a potential bias of 5% in calibration drift and it is well known that these are not really linear over the entire range. This underestimation decreases with lowered P O2 values and is therefore of limited clinical importance (Figure 2, 3).
In general and as demonstrated by porcine results, temperature is obviously one major confounder on P O2 measurement [20,21]. However, the present results of the artificial circulatory setup show that the MFPF technology accurately compensates for alterations in temperature if assessed strictly at the measurement site. This fact highlights the efficiency of multi-temperature compensation performed by MFPF software between 20 and 40uC. In this context, it would be desirable to have sensors available with an  integrated temperature thermistor to prevent errors such as temperature probe misplacement. In addition to temperature, blood flow may influence P O2 measurements of an indwelling sensor [22]. The present ex vivo results confirm that MFPF technology correctly measures P O2 , even under variations of blood flow within the range of 0.8 to 4.0 L min 21 . One major advantage of MFPF P O2 readings, next to high agreement and reproducibility, is high temporal resolution. We measured the monoexponential time constant for a step change in P O2 at 1.5 s in blood. In contrast, the polagraphic Licox probe showed response times up to several min [5,23]. Recently, faster P O2 probes, such as the Neurovent PTO probe based on the fluorescent quenching of oxygen, have been presented. Although these probes reported higher response times (30 -120 s), none of the commercially available technologies come near the temporal resolution of the uncoated ruthenium probe used in the current study [23,24,25,26,27,28].
The MFPF system is easy to set up and needs no complex calibration. In this validation study, the FOXY-AL300 probes were individually calibrated to minimise for potential errors, but in less demanding use, a calibration file supplied by the manufacturer for each probe can be used with acceptable accuracy. Furthermore, phase shift is not influenced by the medium the oxygen sensor is placed in, allowing simplified calibration in gas phase. Moreover, phase shift measurement is not influenced by the probe indicator dye concentration, photo bleaching of dye, or excitation source intensity. This results in stable signals with high signal to noise ratio [29].
Unfortunately, we could not obtain readings in a clinical setting in patients, as the indwelling probes available up to now (e.g. the FOXY AL-300) are not licensed for human application. Therefore, we investigated the MFPF technology ex vivo in a self-built artificial circulatory system filled with human blood. Although ruthenium-based dyes are commonly used in biological monitoring applications and are thought to be non-toxic, for future application in humans, the metallic surface of the probe and its tip containing ruthenium need to be coated with silicone or TeflonH.
In contrast to a silicone probe coating, TeflonH would allow for more rapid diffusion of oxygen and thereby maintain ultrafast probe response time. The challenge for the sensor engineering community will be to avoid direct contact of the fluorescent dye to blood components. Also, it is necessary to avoid protein absorption and blood clotting to maintain signal stability and accuracy. Therefore, an intensive research effort is still necessary, but if engineering is to advance from laboratory-based research to commercial production levels, the MFPF technique could provide clinicians with an effective tool.
A limitation of the study is that we did not explicitly investigate the measurement stability over time. However, the present findings showed no calibration drift of the probes.
In summary we conclude that ex vivo MFPF-P O2 readings are reproducible, and show excellent correlation and high agreement with the gold standard Clark-type electrode (CTE)-based P O2 analysis. This novel technology adequately compensates for changes in temperature and allows for accurate measurements at various blood flow states. This feasible, accurate and easy to calibrate method has the potential to dynamically follow changes in blood oxygenation under various physiologic and pathophysiologic conditions with a high temporal resolution.