Ethical considerations
Even if the device was developed in Switzerland the trial was conducted in a reference laboratory accepted by the FDA, the Hypoxia Research Laboratory at UCSF. The study was reviewed and approved by the UCSF Ethical Committee on Human Research. The Approval number is 10–00437, expiring on March 8th, 2021. The approval letter is on file at UCSF. The laboratory conforms to Good Clinical Practice Standards for the involvement of human subjects and handling of test data. Written informed consent was obtained from each participant.
System description
The SmartCardia 7 L device is a wireless patch with a low-cost disposable component and a re-usable electronic unit (Fig. 1). The patch acquires ECG and measures heart rate (HR) and SpO2. The data are transmitted by Bluetooth to a mobile phone or router. The measured signals and parameters are also stored on the device. The device for SpO2 recording is placed on the left arm of the subject (Fig. 1). The size of the patch is 55 × 130 mm. The patch-based device offers up to 14-days monitoring and data storage and 7-days real-time connectivity with the cloud through a smartphone. The ability to receive, store, and interpret a broad range of signals offers the opportunity to go far beyond monitoring individual parameters. If the patient’s vital measurements reach a pathological value, the system gives an alert on the cloud and the physician can see the real-time parameters and ECG signals.
Protocol
The protocol required brief stable arterial oxygen desaturation in healthy volunteers and sampling arterial blood when a stable level of hypoxia has been attained. The blood sample was analyzed for oxygen saturation with a gold standard bench CO-oximeter, currently a Radiometer ABL-90 multi-wavelength oximeter (Hemoximeter, Radiometer, Copenhagen (Denmark), serial 1393-090R0359N0002). This instrument contains factory certified calibration standards and quality control algorithms. Twenty to twenty-five arterial blood samples from each subject can be analyzed following a protocol aligned with current ISO and FDA guidance documents for pulse oximeter testing.
A radial arterial cannula was placed in either the left or right wrist of each subject for arterial blood sampling and blood pressure monitoring.
Our approach to obtaining stable, safe, and controlled hypoxia was breath-by-breath respiratory gas analysis. A computer program permits the inspired gas mixture to be adjusted to achieve a level of lung alveolar gas that will achieve the desired degree of hypoxia. This was obtained with the use of a nonerebreathing circuit with CO2 removal. Typically, saturation is determined once with air breathing and then at each of the following levels, e.g., 93, 90, 87, 85, 82, 80, 77, 75 and 70% saturation for about 30–60 s at each level. An arterial blood sample is obtained from the catheter at the end of each hypoxic plateau. The operator changes the inspired oxygen concentration at the end of each blood sampling to attain the next desired steady-state hypoxic condition. A run takes 10–15 min, and each run is terminated by a breath of 100% O2 followed by room air. Two runs together enable obtaining a total of 20–25 blood samples, two samples at each different plateau. Saturation of each arterial blood sample is determined by direct oximetry using the Radiometer ABL-90 multi- wavelength oximeter.
Sites for affixing pulse oximeter probes was the fingers for the Masimo device, the ears for the Nellcor, and the arm for the SmartCardia. In order to avoid inaccurate readings due to head or fingers movements with these 2 devices, the arm of the subjects was fixed during the measurements period. In addition, the participants were asked not to move their heads, as much as possible for the duration of the measurements.
Statistics
Pulse oximeter data was taken as 5 s averages corresponding to the point of SaO2 analysis. Individual data points may be missed or excluded for dropped signals or failure of the oximeter signal to achieve an appropriate plateau. Agreement in SpO2 and arterial SaO2 were analyzed using Bland Altman analysis [14, 15]. Bland Altman curve gives a graphical representation of the agreement between SpO2 and arterial SaO2 value. The average of the SpO2 and SaO2 is plotted on the x-axis while the difference between the two values is plotted on the y-axis. The more the graph points toward a zero difference with narrow dispersion of the ‘body’, the better the agreement.
Tables of mean, standard deviation, standard error, minimum, maximum, 95% CI, count and root mean square are provided for each oximeter’s bias, and all oximeters combined in the following ranges of SaO2: 70–100%, 60–70%, 70–80%, 80–90%, and 90–100%. LoA was obtained between methods of measurement with multiple observations per individual [16]. Root mean square error (ARMS) was calculated as it represents the best way to compare SaO2 measurements. It includes both values of SaO2 and its stability. The following formula was used for calculation:
$$\mathrm{ARMS}=\sqrt{\frac{\sum {\left({\mathrm{SpO}}_2-{\mathrm{SpO}}_2\right)}^2}{\mathrm{n}}}$$
Data were managed on MS excel spreadsheet and analysed using stata 9.0 software (Stata Corp, College Station, TX). On Bland Altman curves linear regression is shown for all subjects combined and the equation with R2 is shown on the plot. Mean bias is displayed as a solid horizontal line, and the upper and lower limits of agreement (mean bias ±1.96•SD*) are shown by dashed horizontal lines. For “pooled” plots, different markers are used for each pulse oximeter. Continuous variables were compared using ‘t’ test. A p value < 0.05 was considered as significant.