Pigs of several breeds are established models for translational research and might frequently require sedation or general anesthesia.1–3 In addition, the caseload of pet pigs requiring veterinary medical care, or those belonging to rescue organizations, is also on the rise in several regions. A cursory search of the electronic medical records in the authors’ institution revealed that the annual caseload for pigs referred to the Anesthesiology and Pain Medicine service increased from 96 in 2019 to 287 in 2023. Parallel with the increase in caseload, the complexity of medical and surgical procedures, length of hospitalization, and, consequently, the challenges for anesthesia providers have increased. One of such challenges during anesthetic management is the measurement of arterial blood pressure (ABP). Direct measurement using a catheter placed in an artery, largely considered the gold-standard method, is technically demanding, particularly in sedated animals that may not remain completely immobile. Indirect methods offer a more user-friendly solution but have limitations, namely their accuracy and/or precision.4–6 The oscillometric method is widely used to assess ABP in several veterinary species, but information on its performance in pigs is scarce. Indirect measurements of the mean arterial pressure (MAP) can be particularly useful as it represents the driving pressure that determines, at least in part, tissue perfusion, and it is typically used as a target outcome during treatment of hemodynamic instability.7,8 This parameter is also likely the most commonly used during clinical veterinary anesthesia for assessing arterial blood pressure.
The goals of the current study were to (1) assess the accuracy of a single oscillometric monitor in anesthetized pigs subjected to changes in ABP, (2) test the ability of the device to track directional changes in ABP, and (3) assess the ability of the device to detect arterial hypotension and hypertension in reference to direct measurements.
Methods
This study was performed under IACUC approval (protocol no. 2021-0066), and data were collected as part of a larger project investigating the effects of various ventilation strategies in anesthetized pigs subjected to changes in cardiac function, as well as changes in preload and afterload. As such, a range of ABP measurements were expected in each animal. Nine (3 males and 6 females) Yorkshire pigs, approximately 6 months old and weighing 35 to 55 kg, were included. All pigs were considered healthy and were fasted from solid food overnight before anesthesia. A combination of 0.2 mg/kg midazolam (25 mg/mL; Hospira), 10 mg/kg ketamine (100 mg/mL; Dectra), and 0.1 mg/kg detomidine (10 mg/mL; Zoetis) was administered IM. The pigs were then transferred to a surgical table and provided supplemental oxygen via a loose-fitting face mask. A catheter was placed in an auricular vein, and 2 mg/kg propofol (10 mg/mL; Sagent) IV was administered for induction of general anesthesia. The trachea was then intubated, and anesthesia was maintained with constant rates of infusion of propofol (0.1 to 0.2 mg/kg/min) and ketamine (10 to 20 µg/kg/min). Then, 0.1 mg/kg morphine (10 mg/mL; Hikma) was given IV before further catheterization. All pigs were placed in dorsal recumbency, and the lungs were ventilated with 100% oxygen. Mode of ventilation and respiratory variables (tidal volume, frequency, and airway pressures) were altered as part of the primary investigation. Monitoring under anesthesia included pulse oximetry, capnography, spirometry, ECG, transpulmonary cardiac output, and rectal temperature (Cardell Touch [Midmark Corporation], PULSION Medical Systems SE, and Respironics NM3 [Phillips]).
Blood pressure was measured directly from a femoral artery and indirectly with an oscillometric cuff placed around the metatarsus on the contralateral limb. For invasive ABP, the catheter was connected via noncompliant tubing filled with heparinized saline (2 U/mL) to a pressure transducer (Deltran Disposable Pressure Transducer; Utah Medical Products Inc) located at the expected level of the heart, using the shoulder as the reference point, and connected to a multiparametric monitor (Cardell Touch; Midmark Corporation), where systolic, diastolic, and mean pressures were displayed continuously. A square wave test was performed and assessed visually for the presence of 2 to 3 observable oscillations. The arterial catheter was flushed periodically, or if the quality of the waveform appeared to change (eg, if the waveforms became suddenly of smaller magnitude or lost definition of the dicrotic notch without interventions). An oscillometric cuff, with a width of approximately 40% of the circumference of the metatarsus, was connected to the same multiparametric monitor and programmed to cycle every 2 minutes. The accuracy of the monitor was confirmed before these experiments with 5-point calibrations with simulators (BP-28 Pressure Transducer Simulator [Fogg System Company] and AccuPulse Benchtop NIBP [Clinical Dynamics]). Values for systolic, diastolic, and mean blood pressure measured invasively and noninvasively were recorded in the monitor and downloaded after the procedure was completed. Data collection continued until the end of the primary experiment.
Statistical analysis
For simplicity, and because this is the variable most commonly used in clinical settings, only the invasive and noninvasive mean arterial pressures (iMAP and NiMAP) were used for analysis. Values for iMAP and NiMAP were recorded every 2 minutes. The effect of method (invasive or oscillometric) on MAP was evaluated with a mixed-effect model, using method, time, and their interaction as the fixed effects and animal as the random effect. The histogram of the residuals was observed to confirm normal distribution. Significance was set at P < .05.
Paired values for iMAP and NiMAP were plotted, and the coefficient of determination was calculated. A Bland-Altman graph was created by plotting the difference between iMAP-NiMAP over the average of the paired values, and the bias and limits of agreement (LOA) were calculated.9 The percentage error was calculated as (2 X SD of bias)/mean MAP. Next, the average of 5 consecutive values for iMAP and NiMAP were calculated. This average was calculated in an overlapping fashion as previously described,10 where the first average was composed of measurements 1 to 5, the second average was composed of values 2 to 6, and so on. The paired averaged measurements were used to reconstruct the regression and Bland-Altman plots.
A 4-quadrant graph was created by plotting the change in NiMAP over the change in iMAP.11 To capture large differences, changes in MAP were measured over 20-minute intervals (over 10 consecutive measurements), for example, MAP at 20 minutes − MAP at minute 0. The concordance was calculated as the number of paired values in which the change in MAP occurred in the same direction by both methods over the total number of pairs. An exclusion zone was imposed whereby pairs in which the change in MAP was less than 10 mm Hg by both methods were not considered.12
Finally, a receiver operating characteristics (ROC) curve was created to test the ability of the oscillometric method to detect hypotension, which was defined as an iMAP less than 70 mm Hg.13 The area under the ROC curve was calculated, and the value providing the highest combined sensitivity and 1-specificity (true and false positives) was identified. The same methodology was used to produce a ROC curve for hypertension, defined as an iMAP greater than or equal to 120 mm Hg. All calculations were performed with statistical software (JMP Pro 16).
Results
Data were obtained from 9 animals. A total of 1447 pairs of values were recorded. Significant effects of method (NiMAP vs iMAP) and time were observed in the mixed-effect model (both P < .001), with iMAP being higher than NiMAP [least square means (SD error)] 97.1 (4.7) versus 88.5 (4.7) mmHg. The coefficient of determination using all values and overlapping averaged values are shown (Figure 1). Predictably, the number of values analyzed was reduced when the averages were used; however, the coefficient of determination increased.
Invasive and noninvasive mean arterial pressure values (iMAP and NiMAP) measured simultaneously in 9 anesthetized pigs. Individual pairs were recorded every 2 minutes (A), and the average of 5 consecutive values (B) are shown. The sample size and coefficient of determination are shown.
Citation: American Journal of Veterinary Research 85, 12; 10.2460/ajvr.24.05.0145
Bland-Altman graphs using all pairs and averaged values are shown (Figure 2). NiMAP showed a negative bias of −8.59 mm Hg (SD 11.6) when consecutive pairs were used and −8.85 mm Hg (SD 8.4) when averaged pairs were plotted. The LOAs and percentage errors were reduced when averages were plotted instead of individual values (percentage error, 19% and 26%, respectively).
Bland-Altman plots of the difference of iMAP and NiMAP over their average obtained from 9 anesthetized pigs. Each animal is identified by a different color. Individual pairs were recorded every 2 minutes (A), and the average of 5 consecutive values (B) are shown. The bias, limits of agreement, and percentage error (PE) for each plot are indicated.
Citation: American Journal of Veterinary Research 85, 12; 10.2460/ajvr.24.05.0145
The concordance for detecting directional changes in NiMAP was 82% (Figure 3). The area under the ROC curve for detecting hypotension with NiMAP was 0.936 (Figure 4). The NiMAP that maximized the combination of sensitivity (true positives) and 1-specificity (false positives) to diagnose hypotension (iMAP < 70 mm Hg) was 63 mm Hg. The area under the ROC curve for diagnosing hypertension was 0.940, with a NiMAP value of 101 mm Hg representing the best combination of sensitivity and 1-specificity.
Change in mean arterial over 20 minutes pressure measured iMAP and NiMAP in 9 anesthetized pigs. Values in the upper right and lower left quadrants indicate changes in the same direction by both methods (concordant values). Changes of < 10 mm Hg with both methods were excluded. The sample size, coefficient of determination, concordance rate, and number of excluded values are indicated.
Citation: American Journal of Veterinary Research 85, 12; 10.2460/ajvr.24.05.0145
Receiver operating characteristics curves for the detection of hypotension (A) defined as an iMAP < 70 mm Hg and hypertension (B) defined as an iMAP > 120 mm Hg in 9 anesthetized pigs. The curves are created by plotting the true positives over the false positives, as measured with an oscillometric monitor (NiMAP), in 9 anesthetized pigs. The values with the best combination of true and false positives are identified in red (63 mm Hg for hypotension and 101 mm Hg for hypertension).
Citation: American Journal of Veterinary Research 85, 12; 10.2460/ajvr.24.05.0145
Discussion
The purpose of this study was to assess the performance of an in oscillometric blood pressure monitor in anesthetized pigs under a wide range of pressures and to assess its ability to track changes in MAP and to detect hypotension and hypertension. Whether agreement is sufficient for one method (in this case oscillometric) to be an acceptable alternative to the gold standard (arterial catheterization) is a clinical question, rather than a statistical one.14 To aid in this decision, we performed a number of tests that provide information on the level of agreement. Guidelines for the evaluation of oscillometric monitors have been reported by the American College of Internal Medicine and have been used when evaluating this technology in anesthetized animals.4,15,16 However, those guidelines were created for the diagnosis of arterial hypertension in dogs and cats.17 Since abnormally low pressures are also often encountered during anesthesia, and the detection of arterial hypotension is paramount during monitoring of anesthetized animals, we used different criteria to compare both technologies. Moreover, we proposed a novel approach that consists of observing several measurements rather than one.10 The advantage of this approach is that the impact of any isolated highly inaccurate NiMAP measurement can be minimized by evaluating a group of consecutive measurements. Such isolated errors may arise not just from the intrinsic error of the device but also from movement or inadvertent external pressure applied over the cuff. This strategy of observing more than one value before a diagnosis is made or an intervention triggered is possibly something that many veterinarians already instinctively do. Finally, we assessed the ability of the oscillometric monitor to track directional changes in MAP, that is, to faithfully detect whether MAP increases or decreases. Tracking changes (or trends) during monitoring might provide clinically important information, for example, if ABP is decreasing during hemorrhage or improving with treatment.
We first used a mixed-effect model, which showed a significant effect of method and time but not of the interaction. The difference between methods was small (8.6 mm Hg) and likely of limited clinical impact.
While regression is generally considered an outdated method that is insufficient by itself to compare technologies, the regression plots allow an initial cursory visual inspection of the data. It also allows an initial evaluation of the effect of grouping pairs for comparison, which in our case resulted in an increase in the coefficient of determination.12
Bland-Altman plots show the level of agreement between methods over the range of MAP measured and allow measurement of the LOA. While the bias did not change substantially when we used averages instead of individual values, the LOA decreased. This observation supports our contention that averaging a group of consecutive values will reduce the impact of isolated substantial disagreements between the methods. We did not evaluate the number of values that would maximize this effect. We selected 5 consecutive values because we considered this interval (10 minutes) to be practical during clinical monitoring, but it is likely that the results would differ slightly if averages were constructed using a different number of values.
A point of controversy when Bland-Altman plots are created is whether values obtained from the same animal should be considered replicates, in which case modifications to the plot are suggested. Replicate values were defined by Bland and Altman as “two or more measurements on the same individual taken in identical conditions.”18 While this may be applicable to 2 consecutive MAP measurements under very stable conditions in an anesthetized animal, it clearly does not apply when MAP is changing, particularly in an experiment in which interventions to deliberately alter the MAP are imposed. It is likely that our data are composed of a mixture of replicate values, and nonreplicates when MAP is changing. We opted to use all pairs, rather than transforming the data for replicates. It is expected that while the bias would not be substantially affected one way or the other, the LOA may be narrower when all the pairs are computed. In an attempt to provide more information about the interindividual differences, we identified each animal in the plots. Animals in which the agreement between methods appears to deviate from the general population can be identified (Figure 2).
The 4-quadrant plot, which shows the change in MAP by each method, is commonly used to assess the tracking performance of a device, that is, its ability to determine whether the variable increased or decreased.11,12 Concordant pairs, those in which MAP changed in the same direction by both monitors, are shown in the upper right and lower left quadrants, whereas the opposite quadrants show the discordant values (those that changed in opposite directions by each monitor). An “exclusion zone” is commonly imposed, which represents changes of small magnitude that might represent the intrinsic error of the devices.12 Since we do not know the intrinsic error of this device, we arbitrarily excluded values in which the change in MAP was less than 10 mm Hg, as we think that the magnitude of change in MAP is of little clinical consequence. Concordance plots are commonly constructed by computing the change in the variable from one stage of a study to the next one, for example, the change between baseline and hemorrhage. Our design did not allow for this, as interventions were imposed that resulted in the progressive change of MAP in most pigs. Since MAP was recorded every 2 minutes and it changed progressively, only small changes would be measured between consecutive values. This would result in a graph with most pairs centered within the exclusion zone. Moreover, the test would be focused on how the NiMAP tracked relatively small changes in blood pressure. Therefore, we computed the change over 10 measurements, that is, after 20 minutes. This way, larger changes in MAP are evaluated. We do not propose that changes should be evaluated over that interval in a clinical setting. Simply, we imposed this interval to capture larger changes in MAP, which we consider of clinical importance. Even with this imposition, 52% of the data points were excluded. The concordance rate for the remaining pairs was 82%. Although concordance rates have not been widely used for evaluations of oscillometric monitors, rates greater than 90% have been recommended when cardiac output monitors are compared.19
The area under the ROC curve for diagnosing hypotension was high, suggesting that NiMAP values less than 63 mm Hg can be used reliably to detect cases of hypotension. In our case, this was defined as an iMAP less than 70 mm Hg. This is to say that the diagnosis of hypotension using that value (63 mm Hg by NiMAP) is expected to be highly sensitive and specific. Detection of hypertension was also reliable. A NiMAP of 101 mm Hg was identified as the highest combination of sensitivity and 1-specificity for this purpose. In this study, we defined hypertension when iMAP greater than or equal to 120 mm Hg. Both thresholds were defined arbitrarily but represent values at which the authors will often consider intervention.
Our results should be interpreted within the constraints of our project. We evaluated a specific monitor placed only at one location. Results might be different when other devices are used or if the cuff is placed in a forelimb or the tail, as the tissues surrounding the artery may affect how the oscillations are detected.20 The oscillometric cuff was situated close to the level of the shoulder, where the transducer for direct measurements was zeroed. However, any small differences in height were not accounted for and small errors might be introduced. We enrolled a very small number of pigs, with body weights within a relatively narrow range. Therefore, we cannot speculate on the performance of an oscillometric device on larger or smaller animals. Unfortunately, we did not record when the oscillometric method failed to produce a measurement and hence cannot report the failure rate of this device. The threshold for defining hypotension and hypertension was arbitrarily set. These values represent thresholds at which we will commonly consider intervention in anesthetized animals. It is likely that some results will differ if other cutoff points are used.
In summary, our data suggest that averaging several consecutive values improves the accuracy of NiMAP measurements. The device can correctly track changes in MAP approximately 80% of the time and appears reliable for diagnosing arterial hypotension.
Acknowledgments
None reported.
Disclosures
Dr. Martin-Flores is a member of the AJVR Scientific Review Board but was not involved in the editorial evaluation of or decision to accept this article for publication.
No AI-assisted technologies were used in the generation of this manuscript.
Funding
The authors have nothing to disclose.
ORCID
M. Martin-Flores https://orcid.org/0000-0003-2014-9040
References
- 1.↑
Swindle MM. Swine as replacements for dogs in the surgical teaching and research laboratory. Lab Anim Sci. 1984;34(4):383–385.
- 2.
Lunney JK, Van Goor A, Walker KE, Hailstock T, Franklin J, Dai C. Importance of the pig as a human biomedical model. Sci Transl Med. 2021;13(621):eabd5758. doi:10.1126/scitranslmed.abd5758
- 3.↑
Marchant JN, Herskin M. Pigs as laboratory animals. In: Marek Š, ed. Advances in Pig Welfare. Woodhead Publishing; 2018:445–475.
- 4.↑
Reed R, Barletta M, Grimes J, et al. Accuracy of an oscillometric blood pressure monitor in anesthetized pigs. Lab Anim. 2018;52(4):490–496. doi:10.1177/0023677218763686
- 5.
Chow PK, Ng TH, Heng D, Mack PO. A simple method of blood pressure measurement in the pig using a neonatal cuff. Ann Acad Med Singap. 1999;28(1):15–19.
- 6.↑
Musk GC, Costa RS, Tuke J. Doppler blood pressure measurement in pigs during anaesthesia. Res Vet Sci. 2014;97(1):129–131. doi:10.1016/j.rvsc.2014.04.011
- 7.↑
Tibby SM, Murdoch IA. Measurement of cardiac output and tissue perfusion. Curr Opin Pediatr. 2002;14(3):303–309. doi:10.1097/00008480-200206000-00004
- 8.↑
Evans L, Rhodes A, Alhazzani W, et al. Surviving sepsis campaign: international guidelines for management of sepsis and septic shock 2021. Intensive Care Med. 2021;47(11):1181–1247. doi:10.1007/s00134-021-06506-y
- 9.↑
Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet. 1986;1(8476):307–310.
- 10.↑
Martin-Flores M. Can we improve our assessment and interpretation of oscillometric blood pressure measurements? Res Vet Sci. 2021;142:12–14.
- 11.↑
Perrino AC Jr, O’Connor T, Luther M. Transtracheal doppler cardiac output monitoring: comparison to thermodilution during noncardiac surgery. Anesth Analg. 1994;78(6):1060–1066.
- 12.↑
Saugel B, Grothe O, Wagner JY. Tracking changes in cardiac output: statistical considerations on the 4-quadrant plot and the polar plot methodology. Anesth Analg. 2015;121(2):514–524.
- 13.↑
Mandrekar JN. Receiver operating characteristic curve in diagnostic test assessment. J Thorac Oncol. 2010;5(9):1315–1316. doi:10.1097/JTO.0b013e3181ec173d
- 14.↑
Hartnack S. Issues and pitfalls in method comparison studies. Vet Anaesth Analg. 2014;41(3):227–232.
- 15.↑
Cremer J, da Cunha A, Aulakh K, Liu CC, Acierno MJ. Validation of the oscillometric blood pressure monitor Vet20 SunTech in anesthetized healthy cats. Vet Anaesth Analg. 2020;47(3):309–314. doi:10.1016/j.vaa.2019.12.007
- 16.↑
Felisberto R, Gonzalez AG, Flaherty D, Casoria V, Auckburally A. Validation of oscillometric blood pressure measurement using a Datex S/5 Compact multiparameter monitor in anaesthetized adult dogs. Vet Anaesth Analg. 2022;49(2):156–164. doi:10.1016/j.vaa.2021.12.001
- 17.↑
Acierno MJ, Brown S, Coleman AE, et al. ACVIM consensus statement: guidelines for the identification, evaluation, and management of systemic hypertension in dogs and cats. J Vet Intern Med. 2018;32(6):1803–1822. doi:10.1111/jvim.15331
- 18.↑
Bland JM, Altman DG. Measuring agreement in method comparison studies. Stat Methods Med Res. 1999;8(2):135–160.
- 19.↑
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(5):1180–1192.
- 20.↑
Bodey AR, Young LE, Bartram DH, Diamond MJ, Michell AR. A comparison of direct and indirect (oscillometric) measurements of arterial blood pressure in anaesthetised dogs, using tail and limb cuffs. Res Vet Sci. 1994;57(3):265–269. doi:10.1016/0034-5288(94)90116-3
