The TensorTip™ MTX is a non-invasive device designed to determine several physiological parameters with additional analysis of haemoglobin, haematocrit and blood gas analysis by interpreting blood diffusion colour of the finger skin based on spectral analysis. The aim of our study was to investigate the accuracy and precision of the TensorTip MTX in a clinical setting in comparison with routine analysis of blood samples.
Forty-six patients, scheduled for elective surgery, were enrolled in this study. Placement of an arterial catheter had to be part of the standard of care. Measurements were performed during the perioperative period. The measurements obtained with the TensorTip MTX were compared with the results of routine analysis of the blood samples as a reference using correlation, Bland-Altman analysis and mountain plots.
No significant correlation was present in the measurements. Measurement of haemoglobin with the TensorTip MTX had a mean bias of 0.4 mmol/L, haematocrit’s bias was 3.0 %. Bias of partial pressure of carbon dioxide and oxygen was 3.6 and 66.6 mmHg, respectively. Calculated percentage errors were 48.2 , 48.9, 39.9 and 109.0 %. Proportional bias was present in all Bland-Altman analyses. Less than 95 % of the differences fell within the pre-set limits of allowable error.
Non-invasive blood content analysis with the TensorTip MTX device is not equivalent to and did not correlate sufficiently with conventional laboratory analysis. None of the parameters measured showed results within the limits of allowable error. Therefore, the use of the TensorTip MTX is not recommended for perioperative care.
Point-of-care testing (POCT) of blood content is playing an increasingly important role in current care . Particularly in the last decade a substantial increase in different forms and possibilities of POCT has been observed [1, 2]. Initially, these methods were all invasive, but also non-invasive methods have recently become available, which have an obvious advantage of being patient friendly . POCT can be of benefit to patients because it enables measurements at locations where previously no analysis could be performed, such as analysis at remote locations or analysis during patient transportation . Secondly, the immediate bed-side availability of test results enables early commencement of treatment .
The testing possibilities with POCT are various . Devices are usually limited to few parameters. A wide variety of testing options may come at the expense of accuracy, portability or quality . It is clearly described in the ISO standard which technical requirements a POCT device must meet . However, there are no uniform statistical requirements that a new POCT device must meet .
The TensorTip MTX (MTX, Cnoga Medical Ltd., Caesarea, Israel) is a non-invasive POCT device that is designed to measure haemoglobin (Hb), haematocrit (Ht) and blood gas analysis (BGA), including carbon dioxide partial pressure (pCO2), oxygen partial pressure (pO2) and capillary pH. The device complies with applicable European Union legislation (CE marking). In addition, the device is claimed to be able to measure blood pressure, oxygen saturation and cardiac output as well. Precision and accuracy of these latter functions are described elsewhere . The device is placed on the patient’s finger in such a manner that it seals off all environmental sources of light. Four monochromatic light sources radiate light with different wavelengths through the finger’s capillary tissue [9, 10]. This light is projected on a colour image sensor, used as a 3D spectrometer and colour distributor. The colour information relative to place and time, allows the calculation of the concentration of several blood constituents through the use of a specifically designed algorithm. To our knowledge, findings of the studies thus far published, have not been independently verified in surgical patients , , , . The aim of this study was to evaluate the analytical precision and accuracy of Hb, Ht and BGA using the TensorTip MTX compared with measurements obtained through conventional laboratory testing of blood samples drawn from an arterial catheter during peri-operative care.
Materials and methods
This study was conducted according to the declaration of Helsinki and in accordance with the ICH guidelines for Good Clinical Practice. Ethical approval for this study was provided by the Medical Ethical Committee Arnhem-Nijmegen, the Netherlands on 18 June 2020 (Ethical Committee N° 2020-6660, Chairman Prof. Dr. P.N.R. Dekhuijzen). This study was registered at www.trialregister.nl (national trial registry number NL9164). Data was obtained between November 2020 and August 2021.
Study design and subjects
This prospective observational study was conducted by the department of anaesthesia at the Radboud University Medical Centre, Nijmegen, The Netherlands. All patients enrolled in the analysis provided written informed consent. Patients aged 18 years or older, who received an arterial catheter as part of their elective procedure, were eligible for inclusion. Exclusion criteria were patients with anatomical abnormalities preventing application of the TensorTip MTX and patients suspected of Covid-19 infection. Blood values were compared only if they were determined for routine care.
Measurements were performed in the perioperative setting. After induction of general anaesthesia an arterial catheter was placed in the radial artery (Becton Dickinson arterial cannula 20 G/1.10 × 45 mm, Becton Dickinson infusion therapy Systems Inc. Utah, USA) and calibrated accordingly. Measurements were obtained with the patient in supine position. The arms were positioned in 80° extension or next to the body. Body temperature was measured in every patient. Normothermia was aimed for in every patient.
A mandatory self-test of the TensorTip MTX was performed before use. The device was applied to the middle finger or the index- or ring finger if the middle finger did not show any results.
Blood samples drawn from the arterial catheter were sent to the central laboratory for analysis. Hb/Ht were analysed with Sysmex (Sysmex Corporation, Kobe, Japan) and pO2/pCO2 with RAPIDPoint 500 (Siemens Healthcare, Sudbury, UK). Available performance characteristics of the TensorTip MTX, Sysmex and RAPIDPoint are shown in the Supplementary Material, Appendix A.
Within a five-minute window after the blood was drawn, the TensorTip analysis was performed three times. The average of the three measurements with the TensorTip was used in the statistical analysis. The measurements were automatically recorded and saved in a software application provided by the manufacturer (Singular, CnogaCare, Cnoga Medical Ltd., Caesarea, Israel).
Statistical analyses were performed using IBM SPSS statistics 25 software (SPSS Inc., Chicago, IL, USA) and GraphPad Prism 5.03 (GraphPad software, Inc, La Jolla, CA).
A calculation of sample size was not possible, because there are no published margins of error for the TensorTip MTX. The Clinical Laboratory Standards Institute (CLSI EP9-A2 protocol) describes in their protocol that a minimum of 40 samples is required when comparing a POCT device with standard laboratory values . Considering possible technical problems with the measurements and the fact that both BGA and Hb would not be measured in every patient, the sample size was increased to 50 patients. Patient characteristics were assessed using descriptive analysis and reported as median [25–75 % percentile]. Data were checked for normal distribution using the Shapiro-Wilk test and by visual interpretations of the quantile-quantile plot. The analytic variation of the TensorTip per variable was calculated using the triple measurements. Correlation analysis was performed using Pearson correlation coefficient. Bland-Altman analysis was used for calculation of bias, limits of agreement (LoA) and percentage error (PE), along with their respective 95 % confidence intervals (CI). The PE was calculated as follows:
The Bland-Altman figures are presented as a difference plot, meaning that the measured values of the reference method are used for the x-axis. This is preferable when the reference method is considered to have a small measurement error . This deviates from the classical method of Bland and Altman where the mean of the reference and test method are used on the x-axis . Regression analysis was performed to assess proportional bias .
The criteria that a new measurement method must meet, formulated as the total allowable error, was described in the clinical laboratory improvement amendments (CLIA) in 1988 . This guideline is endorsed by the American Association of Bioanalysts (ABB). For Hb and Ht a total error of ±4 % was considered acceptable . The 4 % error was calculated out of the average Hb/Ht found using the reference method. The allowable total error for pCO2 was ±5 mmHg and for pO2 ±15 mmHg . In addition to the Bland-Altman analysis, a mountain plot was constructed, as advised by the Clinical and laboratory standards institute . The allowable total errors were plotted as a line in the mountain plot. For a new method to replace the reference method, 95 % of the measured differences must lie within the limits of the allowable total error. The calculated PE does not apply as a reference value for laboratory measurements. We have used a liberal precision of 10 % of the reference method for the calculation of the acceptable PE. The test method should have the same precision, leading to a PE of 14 % (√(102+102)) . Outcomes below this percentage were considered acceptable.
A total of fifty-three patients were included in the study. In seven patients the TensorTip MTX did not show blood chemistry results due to a technical error. Therefore, statistical analyses were performed on data of forty-six patients. Patient characteristics are presented in Table 1.
|Patient characteristics||Median [25–75 % percentile]|
|Patients included, n||46|
|Age, years||63.5 [51.5–75.5]|
|Height, m||1.71 [1.65–1.77]|
|Weight, kg||73.0 [63.6–82.5]|
|BMI, kg m−2||24.7 [21.8–27.6]|
|Gender, male, n (%)||24 (52.2 %)|
|ASA 1||2 (4.4 %)|
|ASA 2||14 (31.1 %)|
|ASA 3||29 (62.2 %)|
|ASA 4||1 (2.2 %)|
|Type of surgery, n (%)|
|16 (34.8 %)
15 (32.6 %)
10 (21.7 %)
2 (4.3 %)
1 (2.2 %)
1 (2.2 %)
1 (2.2 %)
|Temperature of patient, °C||36.1 [35.6–36.6]|
|Situation at time of measurement|
|Start of surgery
End of surgery
Post anaesthetic care unit
|11 (23.9 %)
10 (21.7 %)
25 (54.3 %)
Data are presented as median with interquartile range or absolute numbers (%). BMI, body mass index; ASA, American Society of Anesthesiologists.
The mean reference haemoglobin concentration was 6.8 (±1.0) mmol/L (10.9 g/dL) with a range of 5.0–9.3 mmol/L. Hb measured with the TensorTip MTX had a range of 3.5–9.3 mmol/L with a mean of 6.4 (±1.4) mmol/L. The TensorTip had an analytic variation of 8.7 % (0–26.0 %). No linear correlation between the two methods could be observed (r0.07) (Figure 1A). Bias, as shown in the Bland-Altman graph (Figure 1B) was 0.4 mmol/L (95%-CI: −0.1 to 0.8). LoA were ±3.2 mmol/L (95%-CI: 2.4 to 4.0). The calculated PE was 48.2 % (95%-CI: 35.9 to 60.5). Proportional bias was present. Less than 95 % (6.5 %) of the measured differences in the mountain plot (Figure 1C) were within the pre-set limits.
Haematocrit measurements had a mean of 33.8 % (±4.9) as measured by the reference method, with a range from 25.0 to 47.0 %. Measurements with the TensorTip MTX had a range of 18.3–45.0 %, with a mean of 30.8 % (±6.6). The analytic variation of the TensorTip was 8.7 % (1.3–21.0 %) No linear correlation was present between the two methods (Figure 2A). The Bland-Altman plot is depicted in Figure 2B which reveals a Bias of 3.0 % (95%-CI: 0.6 to 5.3), with LoA ±15.8 % (95%-CI: 11.8 to 19.8). Calculations showed a PE of 48.9 % (95%-CI: 36.4 to 61.3). Proportional bias was present. Figure 2C shows the mountain plot, with less than 95 % (2.2 %) of the differences within the allowable error.
Carbon dioxide partial pressure
The mean reference pCO2 was 40.5 (±5.3) mmHg with a range of 31.5–54.0 mmHg. Mean measurements with the TensorTip MTX were 36.9 mmHg (±6.9), with a range of 31.3–72.3 mmHg. The calculated analytic variation was 5.2 % (0–20.2 %). Between the two methods no linear correlation was present as showed in Figure 3A. Figure 3B shows the Bland-Altman figure of pCO2 analysis. The bias was 3.6 mmHg (95%-CI: 1.4 to 5.9). The LoA was ±15.4 mmHg (95%-CI: 11.5 to 19.3). PE was calculated to be 39.9 % (95%-CI: 29.7 to 50.1). Evaluation showed presence of proportional bias. The mountain plot (Figure 3C) shows less than 95 % (43.5 %) of the differences within the limits.
Oxygen partial pressure
The mean of pO2 of the reference method was 165.5 (±73.0) mmHg with a range from 60.8 to 430.5 mmHg. The tested method had a range of 25.3–111.0 mmHg, with a mean of 98.9 mmHg (±14.5). The TensorTip had an analytic variation of 4.5 % (0–38.7 %). Figure 4A indicates no linear correlation between the two methods. Figure 4B shows the Bland-Altman analysis. Bias was 66.6 mmHg (95%-CI: 45.3 to 87.8), with LoA ±144.1 mmHg (95%-CI: 107.3 to 180.9). Calculated PE was 109.0 % (95%-CI: 81.2 to 137.0). Again, proportional bias was present. Less than 95 % (13.0 %) of the measured differences on the mountain plot (Figure 4C) are within the set limits.
In this observational study we have determined the accuracy and precision of the TensorTip MTX with regard to the non-invasive measurement of Hb, Ht, pCO2 and pO2 using conventional central laboratory analysis of blood samples as a reference. The study was performed during or shortly after surgery.
The measurements of Hb and Ht show substantial bias indicating the inaccuracy of the method. In addition, inadequate precision was found. For both parameters, the calculated PE’s were much higher than the pre-proposed 14 % (48.2 and 48.9 % respectively). In addition proportional bias was found, indicating an overestimation of Hb and Ht at low values, and an underestimation at high values. Analysis using mountain plots showed a wide distribution of differences between the reference method and the TensorTip. Only a small percentage of these differences were located between the limits of the allowable error. Moreover, the peak of the mountain plot was located outside the limits of allowable error in three out of four parameters (Hb, Ht, pO2). With these findings the mountain plots confirmed poor accuracy and precision .
Previously Hb and Ht measurement with the TensorTip was validated . This post marketing study compared Hb and Ht in 276 healthy subjects at a blood bank. To compare the results of this study with our results we converted Hb from g/dL to mmol/L (factor 0.6206). The results from this study revealed a smaller bias (respectively 0.05 mmol/L for Hb and 1.13 % for Ht) and smaller LoA’s (−1.56 to 1.67 mmol/L for Hb and −7.13 to 9.39 % for Ht), which was not in line with our results. Although not mentioned by the authors, a PE of 18.3 % can be calculated from the available data, when assuming that the male/female ratio in the study is equal. An explanation for the found differences may be the choice of study population, as the Hb and Ht values were lower in our patients compared with the values found in their healthy volunteers (Hb: 6.8 mmol/L [5.0–9.3] vs 8.9 mmol/L [6.8–11.8] and Ht: 33.8 % [25.0–47.0] vs 43.1 % [32–54] respectively). While it is understandable that a new device is first tested at normal values, a medical measurement device must for other than screening in healthy populations also be accurate and precise at the extremes in patients, especially when considering to use the device in a clinical setting. It would be desirable to use this type of non-invasive devices for monitoring of anaemia at home  and for that purpose, it is crucial to reliably distinguish between normal and aberrant values. Compared to four other non-invasive devices recently described in a meta-analysis, the TensorTip MTX performs less well, as all four devices show smaller bias and lower LoA on average .
In addition, our study shows that measurements of pCO2 and pO2 are inaccurate and imprecise, represented by the larger than acceptable bias, LoA and PE. Proportional bias was observed in both pCO2 and pO2, varying from accurate at low to normal range to an underestimation at increasing pCO2 and pO2 values. The former is particularly relevant as it is crucial to reliably detect hypercapnia during clinical use of the device. Finally, the mountain plots of both pCO2 and pO2 show, as with Hb and Ht, that the majority of the differences lie outside the allowable error limits. To our knowledge, no literature exists comparing the TensorTip MTX with a reference device measuring pCO2 and pO2. The manual indicates that a mean square error of <10 % has been determined for both values . Based on our study we are not able to confirm their findings. More devices exist that measure pCO2 in a non-invasive transcutaneous way, however, to date the method has proven insufficient . In general, transcutaneous ear measurements were to be found more reliable than measurement at the arm , questioning the finger as the optimal location for the measurement. The iSTAT (Abbott Point of Care, Princeton, NJ, USA) is a well-known POCT device that shows good accuracy in ABG measurements [24, 25]. However, a blood sample must always be taken to perform a measurement and therefore is not non-invasive.
In addition to not being able to measure in certain patients, there might be influence on the results in case measurement results were obtained. In cold fingers for example, less blood is present due to vasoconstriction, theoretically, this might lead to lower Hb values, than the average Hb in the whole body and this might explain differences when compared to the Sysmex Hb measurement which is obtained by venapunction from a larger vessel in the patients arm.
In this single centre observational study only a limited number of subjects was included. In addition, only one measurement was taken per subject. This made it impossible to examine the device’s trending ability. This is desirable in a method comparison study . It may be possible that a method with insufficient accuracy is able to follow the trend properly, for example in blood glucose monitoring. It would be valuable to further investigate this for the TensorTip MTX.
In our study, the presence of proportional bias is shown as a dotted line. If proportional bias is present, the Bland-Altman figure should be adjusted accordingly, because there will also be proportionality in the LoA . In fact, the bias and LoA may be non-linear . This, for example, can be observed in our findings of pO2, where there is strong proportional bias present (Figure 4). Because of the limited number of subjects in our study and the exploratory nature we have deliberately chosen to present our results in the conventional Bland-Altman figure. In the case of few measurements, a single deviating measurement can have a large effect on the degree of proportionality (Figure 4). In the case of a suspicion of proportionality, a new study with larger inclusion numbers should confirm this.
The position of the patient in our study deviated from the advised positioning by the manufacturer. Measurements should be performed in a sitting position with the elbow resting and the device should hang on the finger downwards below the heart . Furthermore, measurements should not be taken shortly after exercise. We suspect that these positioning errors mainly affect the measurements of the vital parameters, but we cannot exclude an influence on the measurements reported. Moreover, the study population used by us (perioperative patients) is not the primary group for which this device was developed, which may have resulted in more error messages than usual. Testing this device in an environment for which it was not intended may have adversely affected the results. However, Segman et al. performed, for their validation studies, similar measurements on the recovery in post cardiac surgery patients [9, 12]. Positive results were described in these studies. Because the positioning creates ambiguity, It would be useful and interesting to investigate this device in its intended environment: General practitioners office, out-patient care and emergency room . The non-invasive nature of the device allows it to potentially be used as a screening tool, provided it is validated at the extremes in patients.
Performing blood content analysis non-invasively has a great advantage and could be clinically relevant. The TensorTip MTX is able to measure Hb, Ht, pCO2 and pO2 non-invasively. However, our study demonstrates that non-invasive measurements of Hb, Ht, pCO2 and pO2, measured using the TensorTip MTX, are not equivalent to conventional laboratory testing with regard to accuracy and precision in the peri-operative setting in surgical patients. Therefore the perioperative use of the TensorTip MTX is not recommended. Although the use of the device still could be useful in another setting, further validation of the device in patients is mandatory.
Cnoga Medical Ltd provided the TensorTip MTX free of charge.
Research funding: None declared.
Author contribution: All authors have accepted responsibility for the entire content of this manuscript and approved its submission.
Competing interests: Authors state no conflict of interest.
Informed consent: Informed consent was obtained from all individuals included in this study.
Ethical approval: Research involving human subjects complied with all relevant national regulations, institutional policies and is in accordance with the tenets of the Helsinki Declaration (as revised in 2013), and has been approved by the authors’ Institutional Medical Ethical Committee. (Ethical Committee N° 2020-6660) on 18 June 2020.
Data availability: The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.
2. Shu, T, Hunter, H, Zhou, Z, Sun, Y, Cheng, X, Ma, J, et al.. Portable point-of-care diagnostic devices: an updated review. Anal Methods 2021;13:5418–35. https://doi.org/10.1039/d1ay01643a.Search in Google Scholar PubMed
3. Young, MF, Raines, K, Jameel, F, Sidi, M, Oliveira-Streiff, S, Nwajei, P, et al.. Non-invasive hemoglobin measurement devices require refinement to match diagnostic performance with their high level of usability and acceptability. PLoS One 2021;16:e0254629. https://doi.org/10.1371/journal.pone.0254629.Search in Google Scholar PubMed PubMed Central
4. Herd, GCE, Musaad, SMA. Point-of-care testing in rural and remote settings to improve access and improve outcomes: a snapshot of the New Zealand experience. Arch Pathol Lab Med 2021;145:327–35. https://doi.org/10.5858/arpa.2020-0104-ra.Search in Google Scholar
5. Kankaanpaa, M, Raitakari, M, Muukkonen, L, Gustafsson, S, Heitto, M, Palomaki, A, et al.. Use of point-of-care testing and early assessment model reduces length of stay for ambulatory patients in an emergency department. Scand J Trauma Resuscitation Emerg Med 2016;24:125. https://doi.org/10.1186/s13049-016-0319-z.Search in Google Scholar PubMed PubMed Central
6. International Organisation for Standardisation (ISO). 2016. Point-of-care testing (POCT) — Requirements for quality and competence (ISO standard No. 22870.2016). Available from: https://www.iso.org/standard/71119.html.Search in Google Scholar
8. Servaas, S, van Eijk, LT, de Vreede, S, Malagon, I, Slagt, C. Perioperative validation of the TensorTip™ MTX device for noninvasive arterial pressure measurement: a method comparison study. Cardiol Res 2022;13:372–9. https://doi.org/10.14740/cr1438.Search in Google Scholar PubMed PubMed Central
10. Segman, YJ, Sheiman, E. Post marketing study of hemodynamic and hematological noninvasive readings in a blood bank. SAGE Open Med 2018;6:1–8. https://doi.org/10.1177/2050312118796065.Search in Google Scholar PubMed PubMed Central
13. Clinical Laboratory Standards Institute. Clinical Laboratory Standards Institute (CLSI) guidelines. Method comparison and bias estimation using patient samples; approved guideline – third edition. Volume 22 nr 19(document EP9-A2); 2002.Search in Google Scholar
15. Bland, JM, Altman, DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986;1:307–10. https://doi.org/10.1016/s0140-6736(86)90837-8.Search in Google Scholar
16. Montenij, LJ, Buhre, WF, Jansen, JR, Kruitwagen, CL, de Waal, EE. Methodology of method comparison studies evaluating the validity of cardiac output monitors: a stepwise approach and checklist. Br J Anaesth 2016;116:750–8. https://doi.org/10.1093/bja/aew094.Search in Google Scholar PubMed
19. Clinical Laboratory Standards Institute. Estimation of total analytical error for clinical laboratory methods; approved guideline. Volume 23 nr20(document EP21-A). The National Committee for Clinical Laboratory Standards; 2003.Search in Google Scholar
20. Cecconi, M, Rhodes, A, Poloniecki, J, Della Rocca, G, Grounds, RM. Bench-to-bedside review: the importance of the precision of the reference technique in method comparison studies--with specific reference to the measurement of cardiac output. Crit Care 2009;13:201. https://doi.org/10.1186/cc7129.Search in Google Scholar PubMed PubMed Central
22. Dietzel, F, Dieterich, P, Dorries, F, Gehring, H, Wegerich, P. Invasive and non-invasive point-of-care testing and point-of-care monitoring of the hemoglobin concentration in human blood – how accurate are the data? Biomed Tech 2019;64:495–506. https://doi.org/10.1515/bmt-2018-0066.Search in Google Scholar PubMed
23. Conway, A, Tipton, E, Liu, WH, Conway, Z, Soalheira, K, Sutherland, J, et al.. Accuracy and precision of transcutaneous carbon dioxide monitoring: a systematic review and meta-analysis. Thorax 2019;74:157–63. https://doi.org/10.1136/thoraxjnl-2017-211466.Search in Google Scholar PubMed
24. Sediame, S, Zerah-Lancner, F, D’Ortho, S, Adnot, A, Harf, A. Accuracy of the i-STAT bedside blood gas analyser. Eur Respir J 1999;14:214–7. https://doi.org/10.1034/j.1399-3003.1999.14a36.x.Search in Google Scholar PubMed
25. Thomas, FO, Hoffman, TL, Handrahan, DL, Crapo, RO, Snow, G. The measure of treatment agreement between portable and laboratory blood gas measurements in guiding protocol-driven ventilator management. J Trauma 2009;67:303–13. https://doi.org/10.1097/ta.0b013e3181a5f055.Search in Google Scholar PubMed
26. Critchley, LA, Lee, A, Ho, AM. A critical review of the ability of continuous cardiac output monitors to measure trends in cardiac output. Anesth Analg 2010;111:1180–92. https://doi.org/10.1213/ane.0b013e3181f08a5b.Search in Google Scholar
This article contains supplementary material (https://doi.org/10.1515/cclm-2023-0113).
© 2023 the author(s), published by De Gruyter, Berlin/Boston
This work is licensed under the Creative Commons Attribution 4.0 International License.