Materials and Methods

Sample population—Data obtained for 18 neonatal foals (9 males and 9 females) between 1 and 6 days old and weighing 32 to 61 kg were published in 2 studies.^4,5 Those data were collated and used in the study reported here. Birth of all foals was uneventful at the end of a full-term gestation. Foals were considered healthy on the basis of results of a complete physical examination and assessment of adequate transfer of passive immunity, which was confirmed by measurement of plasma IgG concentration by use of a commercial immunoassay.^a The 2 studies^4,5 were approved by the Institutional Animal Care and Use Committee of the University of Florida.

Anesthesia and instrumentation—Several procedures were performed during completion of the 2 studies.^4,5 The anesthetic procedure performed during completion of the 2 studies included insertion of a 16-gauge, 2- or 12-inch catheter in the right jugular vein for subsequent induction of general anesthesia. In the first study,⁴ 10 foals were sedated by IV injection of xylazine hydrochloride^b (1.0 mg/kg), whereas in the second study,⁵ 8 foals were sedated by IV injection of diazepam^c (0.05 mg/kg). In both studies, anesthesia was induced by IV injection of ketamine hydrochloride^d (2.0 mg/kg).

For both studies,^4,5 foals were positioned in left lateral recumbency. An orotracheal tube was inserted for maintenance of anesthesia by administration of isoflurane^e in oxygen by use of a rebreathing circuit system. Oxygen flow rate was 2 to 3 L/min, and end-tidal isoflurane concentration was monitored by use of an infrared gas analyzer^f calibrated before each experiment with a standardized calibration gas mixture^g designed for the analyzer.

Foals were mechanically ventilated at a rate of 8 to 10 breaths/min and a tidal volume of 10 to 15 mL/kg to maintain the ETCO₂ concentration between 40 and 50 mm Hg. Because the primary objectives differed between the 2 studies, ETCO₂ concentrations differed for foals in each study. In foals of the first study,⁴ the ETCO₂ concentration was maintained constant within the desired range. However, the second study,⁵ the ETCO₂ concentration was maintained constant for only the first 30 minutes of anesthesia; subsequently, the respiratory rate was kept constant but with no attempt to correct the ETCO₂ concentration thereafter.

A 20-gauge, 1.88-inch catheter was inserted in the right metatarsal artery for measurements of direct blood pressure and CO. An electronic pressure transducer^f was positioned and calibrated to 0 at the level of the sternal manubrium for continuous monitoring of SAP, DAP, and MAP. Heart rate and ECG were also monitored throughout the experiments. Normovolemia and normoglycemia were maintained throughout the experiments by administration of a balanced electrolyte solution^h (5 mL/kg/h, IV). Body temperature was monitored electronically and maintained between 37.5° and 39°C by use of a blanket.

Measurement of CO by LiDCO—A lithium dilution cardiac computerⁱ was used to determine CO during both studies.^4,5 The hemoglobin concentration and serum sodium concentration required for this method were measured by use of a blood gas analyzer^j immediately before CO determination. The sensor^k for measurement of lithium chloride in blood was placed between the side port of a 3-way valve and connected to the catheter inserted in the metatarsal artery. Extension tubing that passed through a peristaltic pump directed blood to a collection bag at a flow rate of 4 mL/min across the sensor. The dose of lithium chloride^l (0.16 to 0.28 mmol) was placed in an extension set attached to the catheter inserted in the jugular vein; the dose was administered, and the extension set was flushed with 8 to 12 mL of heparinized saline (0.9% NaCl) solution 8 seconds after starting the LiDCO computer. Measurements of CO were obtained in duplicate or triplicate to verify that variation between measurements was < 20%. An interval of at least 2 minutes was allowed between subsequent measurements.

Measurement of CO by NICO—A commercially available monitor^m was used to measure CO during both studies.^4,5 Measurements were obtained in accordance with the manufacturer's recommendations. Hemoglobin concentration and arterial pressures of carbon dioxide and oxygen required for NICO were determined before CO measurement. A carbon dioxide and volume sensor was connected between the orotracheal tube and anesthetic circuit system. A pulse oximeter was placed on the tongue or lip. Values for CO were determined at 3-minute intervals. This included 60 seconds without rebreathing for baseline determinations of the volume of carbon dioxide eliminated over time, PaCO₂, and ^ETCO₂ concentration; 35 seconds of rebreathing during which the volume of carbon dioxide eliminated was decreased, PaCO₂ and ETCO₂ were increased, and mixed-venous carbon dioxide concentration remained unchanged; and 85 seconds of stabilization and no rebreathing during which all variables returned to baseline values. The rebreathing phase forced each horse to temporarily inhale only a portion of the exhaled gases, and the monitor calculated CO by use of detected differences in carbon dioxide elimination and ETCO₂ between the normal and rebreathing state for the following modified Fick equation:

View Expanded

where ΔVCO₂ is the difference in carbon dioxide elimination and ETCO₂ concentration between the normal state and rebreathing state, and $CaCO₂ is the difference in arterial carbon dioxide content between the normal state and rebreathing state.

Experimental design—Designs for the 2 studies^4,5 have been described elsewhere. In brief, low, intermediate, and high CO values were achieved in the first study by use of an end-tidal isoflurane concentration of 1.3% to 2.1%, an end-tidal isoflurane concentration of 0.85% to 1.4% and a constant rate IV infusion of dobutamineⁿ (1 to 3 μg/kg/min), and an end-tidal isoflurane concentration of 0.83% to 1% and a constant rate IV infusion of dobutamine (3 to 6 μg/kg/min), respectively. Four foals also received intermittent IV injections of phenylephrine^o (total cumulative dose, 1.1 to 1.7 mg) for the highest CO value; phenylephrine was administered after data from the dobutamine phase were recorded. For each CO value, a stable hemodynamic plane of at least 10 minutes was maintained before data collection. At each anesthetic protocol, LiDCO and concurrent NICO measurements as well as SAP, DAP, MAP, and heart rate were recorded.⁴

Foals in the second study⁵ were maintained at a steady deep plane of anesthesia by use of an end-tidal isoflurane concentration of 2.0% to 2.5% that causes a hypotensive hemodynamic state. Each foal received 3 treatments in accordance with an orthogonal factorial design; there was at least 24 hours between treatments. Treatment 1 consisted of dobutamine administered at rates of 4 and 8 μg/kg/min, treatment 2 consisted of norepinephrine^p administered at rates of 0.3 and 1.0 μg/kg/min, and treatment 3 consisted of vasopressin^q administered at rates of 0.3 and 1.0 mU/kg/min. Each drug was always administered at the lower infusion rate first. All treatments were prepared in equal volumes of infusate by dilution in 5% dextrose solution to create appropriate concentrations (dobutamine, 2 mg/mL; norepinephrine, 0.2 mg/mL; and vasopressin, 0.2 U/mL) and were administered by use of a syringe infusion pump.^r After a period of at least 15 minutes, LiDCO and NICO measurements as well as blood pressures and heart rate were recorded in the hypotensive state; the vasopressor drug was then administered at the lower infusion rate for a period of at least 15 minutes or until a stable hemodynamic plane was achieved, and all cardiovascular variables were recorded again. Immediately thereafter, the higher infusion rate was initiated in a manner similar to that for the lower infusion rate and cardiovascular variables were recorded for a third time.⁵

All values for CO were transformed to CI by use of the following equation:

View Expanded

where BW is body weight in kilograms.

Recovery from anesthesia—In both studies,^4,5 all instrumentation was removed once the experiments were concluded each day. For the second study, the catheter was allowed to remain in place in the jugular vein with a protective bandage applied to the neck area to avoid repetitive catheterization on subsequent days. Foals were assisted throughout the recovery period and returned to their stalls once fully awake. Foals were observed for evidence of typical suckling activity and behavior during both studies.^4,5

Data analysis—The 2 methods were compared by use of a conventional Bland-Altman analysis.⁶ Bias was defined as the mean value of differences between LiDCO and NICO. Precision (ie, limits of agreement) was defined as 1.96 X SD of the differences and included 95% confidence intervals. For a specific foal with a particular physiologic condition caused by administration of drugs, CO was measured 2 or 3 times concurrently by both the LiDCO and NICO methods. In the case of replicated measurements, additional analyses have been described⁷ to determine the limits of agreement and account for possible causes of variation. This additional step of repeated, concurrent measurements allows estimation of repeatability of each method and an alternative statistical analysis to determine the limits of agreement. By use of Bland-Altman procedures, the repeatability coefficient for technique x was estimated as follows: s_x X (1.96 X √2), where s_x is the square root of the residual mean square calculated from an ANOVA^s for all individual data points of a particular method (ie, LiDCO or NICO).

In addition to estimation of repeatability, the alternative analysis allows investigators to use all data gathered during investigations.⁷ Because each foal had the CO value determined 2 or 3 times by each method, an analysis to estimate bias and limits of agreement for an equal number of replicate measurements was used, wherein each foal for a specific physiologic condition was considered a subject. In this case, it is assumed that variances of the means for a given set of individual values are independent of the magnitude of subject means. Therefore, the SD of these measurements for a specific foal was plotted against the mean value of these measurements to ascertain whether the assumption was true. Next, a second assumption of independence that the difference between the means for this data set did not depend on the magnitude of average mean values for a given observation in a particular subject was verified. Therefore, the difference between means was plotted against the mean value of average values.

Once these 2 assumptions were confirmed to be true, the average of the difference in means that provided an estimate of bias between the methods was calculated. To calculate the limits of agreement by use of the entire data set, the following equation⁷ for an equal number of replicate measurements was used:

View Expanded

where σ^ˆ2_d is the variance of the difference between LiDCO and NICO means; s²/_d is the observed variance of the differences between within-subject means; m_x and m_y are the number of observations on each subject with methods x (LiDCO) or y (NICO), respectively; and s²_xw and s²_yw are residual mean square values for the LiDCO and NICO data, respectively, which were calculated for each group separately by use of an ANOVA.^t The SEM was calculated as √σ^ˆ2_d.

The confidence errors for these limits of agreement were determined by use of the following equation:

View Expanded

where n is number of subjects; s⁴/_d, s⁴_xw, and s⁴_yw correspond to the variables s²/_d, s²_xw, and s²_yw, respectively, for the preceding equation that follow a S² distribution.

Regression analyses were used to determine correlations and examine the linear relationship between LiDCO and NICO and between LiDCO and heart rate, SAP, DAP, and MAP by use of the recorded value of each variable at the time of a single CO determination. Values of P < 0.05 were considered significant.

Results

Repeatability coefficients for the LiDCO and NICO methods were 32.4 and 40.7 mL/kg/min, respectively. Variances in the mean appeared to be independent of mean values for a specific method (Figure 1). Mean difference (bias) was −17.3 mL/kg/min, and the 95% limits of agreement were −131.3 to 96.7 mL/kg/min (mean ± SD, 114.0 ± 4.2 mL/kg/min; Figure 2).

View Full Size

Plots of within-subject SD versus the subject mean for CI measured by use of LiDCO (A) or NICO (B) methods in 18 foals in 2 studies.^4,5 A subject was defined as a specific foal for a particular physiologic state. Notice the 1 extreme outlier for NICO, which was considered to be a valid observation.

Citation: American Journal of Veterinary Research 68, 2; 10.2460/ajvr.68.2.141

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Bland-Altman plot of subject bias and precision (ie, limits of agreement; 1.96 X SD) determined by use of replicated measurements for 18 foals in 2 studies.^4,5 The solid horizontal line indicates the mean bias of the measurements, the dotted horizontal lines depict the limits of agreement, and the dashed lines around each dotted line represent the 95% confidence intervals of the limits of agreement. See Figure 1 for remainder of key.

Citation: American Journal of Veterinary Research 68, 2; 10.2460/ajvr.68.2.141

• Download Figure
• Download figure as PowerPoint slide

The collated data yielded 217 comparisons. There was good correlation between LiDCO and NICO (r = 0.77; Figure 3). Mean ± SD LiDCO and NICO values from 217 measurements were 138 ± 62 mL/kg/min (range, 40 to 381 mL/kg/min) and 154 ± 55 mL/kg/min (range, 54 to 358 mL/kg/min), respectively. Mean of the differences of LiDCO minus NICO was 4.37 + (0.87 X NICO).

View Full Size

Linear regression for a wide range of CI values measured concurrently by LiDCO and NICO methods in 18 anesthetized foals in 2 studies.^4,5 The solid line indicates the best fit, and dashed lines indicate 95% confidence intervals. There was good correlation (r = 0.77) between the methods. Equation for the slope of the line was as follows: y = 4.37 + 0.87x, which indicates that LiDCO = 4.37 + [0.87 X NICO]).

Citation: American Journal of Veterinary Research 68, 2; 10.2460/ajvr.68.2.141

• Download Figure
• Download figure as PowerPoint slide

In the first study,⁴ 107 comparisons were made between the 2 methods. Eight comparisons were excluded because variation was > 20% between the LiDCO measurements or there were technical errors at the time of determination. For all CO determinations, there was good correlation (r = 0.63) between the 2 methods. Mean ± SD LiDCO and NICO values from 99 measurements were 133 ± 43 mL/kg/min (range, 64 to 237 mL/kg/min) and 146 ± 34 mL/kg/min (57 to 209 mL/kg/min), respectively. Mean of the differences for LiDCO minus NICO was −17.81 + (0.78 X NICO). Magnitude of the limits of agreement of the differences between LiDCO and NICO was 66.5 mL/kg/min (−80.1 to 52.8 mL/kg/min), with a mean difference (bias) of −13.7 mL/kg/min.

In the second study,⁵ 2 foals had cardiac arrest after collection of baseline data in the hypotensive state. They were successfully resuscitated but were excluded from the remainder of the study. Initial data points for both foals were included because they were collected before the incident.

For the second study,⁵ 120 comparisons were made between the 2 methods. Two comparisons were excluded because they equaled or exceeded the maximum value on the NICO monitor (19.9 L/min). For all CO determinations, there was good correlation (r = 0.81) between the 2 methods. Mean ± SD LiDCO and NICO values from 118 measurements were 143 ± 74 mL/kg/min (range, 40 to 381 mL/kg/min) and 160 ± 68 mL/kg/min (range, 54 to 358 mL/kg/min), respectively. Mean of the differences for LiDCO minus NICO was −0.06 + (0.89 X NICO). Magnitude of the limits of agreement of the differences between LIDCO and NICO was 87.5 mL/kg/min (−105.2 to 69.7 mL/kg/min), with a mean difference (bias) of −17.8 mL/kg/min.

The strongest correlation between CI determined by LiDCO and the other cardiovascular variables measured simultaneously was for heart rate (r = 0.62). There were weak correlations between CI determined by LiDCO and SAP (r = 0.41), DAP (r = 0.25), and MAP (r = 0.34).

Discussion

Methods for CO measurement that are minimally invasive, such as NICO, are optimal for monitoring neonatal foals, but they have not been validated for use in this species. The LiDCO method is accurate and can be used as an alternative to more traditional methods, including thermodilution, in foals and other species.^3,8-10

Bias and limits of agreement were similar for analysis conducted by use of nonreplicated and replicated data. Replicates were 2 or more measurements on the same patient obtained in identical conditions in quick succession, with an expected mean difference between them of 0.⁷ When 1 method has poor repeatability so that there is considerable variation in the replicated measurements, the agreement between the 2 methods will be poor. Given the relative similarity of values for the limits of agreement, both methods appeared to be sufficiently precise to allow repeated data that are valid. The limits of agreement were approximately 2- to 3-fold greater than the repeatability coefficients. When the limits of agreement are dependent only on the repeatability of measurements, then the limits of agreement should be equal to the repeatability coefficients. Because the limits of agreement were wider than the repeatability, then there must have been some other factor that decreased agreement between the methods. For example, it is conceivable that the NICO device was designed such that it would require a correction factor for intrapulmonary shunt flow, a value that may cause overestimation or underestimation of CO when the correction factor varies from true intrapulmonary shunt flow.¹¹ In another study,¹² investigators determined that the LiDCO measurement increased with sequential lithium administration for determination of CO as a result of increases in lithium concentrations.

In addition to inaccuracy attributable to these factors, other unrecognized errors could have contributed to the gap in repeatability values and limits of agreement. Variations between the 2 methods were also possible sources of error because in the first study,⁴ but not in the second study,⁵ ETCO₂ concentrations were manipulated to remain within a specific range, which may have affected the accuracy of NICO.

Objectives differed for the 2 studies that provided the data, but each provided useful scenarios that resemble aspects of clinical situations encountered in anesthetized foals. These include depressed and improved cardiovascular performance in response to anesthetics and inotropic or vasopressor drugs. Investigators in the first study⁴ manipulated anesthetic depth (0.69 to 1.75 MAC)^u in combination with administration of inotropic drugs and vasopressors to induce various degrees of cardiovascular function. This is in contrast with the second study⁵ in which researchers investigated the ability of inotropic drugs and vasopressors to counteract induced cardiodepressive effects for constant administration of a high end-tidal concentration of isoflurane (2.4 to 3.0 MAC).

The range of CI induced in the 2 studies^4,5 included values (40 to 381 mL/kg/min) for an age group (ie, neonatal foals) in which reported¹ CI values for clinically normal conscious animals are 197 to 234 mL/kg/min. Thus, we were able to compare LiDCO and NICO methods over a wide range of values. In the first study,⁴ data for 30 averaged pairs of LiDCO-NICO comparisons were reported; however, in the study reported here, those measurements for each method were used individually and not as replicated averaged pairs. Other measurements recorded at other time intervals in that study⁴ were also included, which resulted in 99 comparisons. In addition, it was determined in the first study⁴ that CO influenced NICO measurements; therefore, those data were reported for low, intermediate, and high CO values. In the study reported here, overall bias and precision of the 99 individual comparisons were calculated, in addition to the 118 individual comparisons of the second study.⁵

The comparison between LiDCO and NICO in each of those 2 studies^4,5 revealed mean differences of −14 to −18 mL/kg/min and limits of agreement of −105 to 70 mL/kg/min between LiDCO and NICO. The negative value for bias indicated that use of NICO would overestimate the LiDCO values. Analysis of data from the first study⁴ revealed that the negative value for bias, and therefore overestimation for NICO, tended to decrease progressively as CI increased toward more physiologic values because maximum CI values achieved by use of the LiDCO method in that study were 237 mL/kg/min. Studies in pigs¹³ and sheep¹⁴ have also revealed overestimation at low CI values and underestimation at high CI values, whereas in dogs, NICO slightly underestimates LiDCO at all CI values, although the difference is larger at higher CI values.¹⁵

It was suggested in 1 study¹⁴ that the increase in mixed-venous PaCO₂ is a result of recirculation during hyperdynamic states (ie, high CI), which overestimates theETCO₂ concentration and affects the assumption for NICO that the difference in partial pressures between end-capillary carbon dioxide and ETCO₂ is constant during rebreathing and nonrebreathing states. The underestimation of CI by NICO is more evident as CI increases.¹⁴ Conversely, during hypodynamic states for NICO, the time during the rebreathing states needed to achieve equilibrium between the partial pressures of end-capillary carbon dioxide and ETCO₂ is insufficient, which leads to underestimation of the partial pressure of end-capillary carbon dioxide and a smaller difference between end-capillary carbon dioxide and ETCO₂ partial pressures.¹⁴

In the study reported here, values generally became scattered as CI increased, which was most obvious when CI exceeded 200 mL/kg/min. Similar results have been reported¹⁵ for LiDCO and NICO comparisons in dogs at CI values that exceed 110 mL/kg/min. Typical values for CI in conscious foals and dogs are close to the values in which scattering was initially detected in the study reported here and another study.¹⁵ It is likely that the hyperdynamic state induced by inotropic drugs, which results in a higher degree of pulmonary shunting and lower difference between venous PCO₂ and Paco₂, decreased the accuracy for NICO, similar to results in other reports.^13,14,16

Cardiovascular depression induced by inhalant anesthetics reduces CI and may affect pulmonary blood flow. Use of inotropic and vasopressors to counteract depressive effects may also affect pulmonary blood flow. The acute nature of changes induced by low blood flow and hyperdynamic circulatory states for the methods used in the 2 studies^4,5 as well as in another study may have contributed to decreased accuracy of NICO. The correlation between NICO and aortic blood flow determined by use of a periaortic transit-time flow probe in pigs¹³ was similar to the correlation obtained between NICO and LiDCO (r = 0.77) for the study reported here but was less than the correlation between those 2 methods for a similar study¹⁵ in dogs (r = 0.88).

The NICO method should be limited to patients that can tolerate increases in PaCO₂ that result from the forced partial rebreathing needed for this method. In addition, mechanical ventilation increases accuracy and aids in maintaining ETCO₂ concentrations. There were differences in the methods for the studies because ETCO₂ pressures were maintained between 40 and 50 mm Hg throughout the first study⁴ but not during the second study.⁵ In the second study, investigators assessed the effects of several inotropic drugs or vasopressors on gastric carbon dioxide tonometry during hypotension; therefore, no attempts were made to alter ETCO₂ values. However, mean ETCO₂ values did not exceed 65 mm Hg in the second study. Concentrations > 65 mm Hg measured during the baseline cycle or > 80 mm Hg during the rebreathing cycle of NICO can interfere with accuracy of the monitor.¹⁷ In addition, tidal volume (10 to 15 mL/kg) and minute ventilation (rate of 8 to 10 breaths/min) were maintained constant in all foals by means of mechanical ventilation to avoid negative effects on accuracy of NICO attributable to low tidal volumes, reduced minute ventilation, and spontaneous respiration.^17,18

Methods for determination of CO are important when evaluating cardiovascular performance in awake and anesthetized patients. Measurement of CO is less commonly performed, compared with the frequency for other variables (such as measurement of blood pressure), primarily because of the more invasive nature and the need for instrumentation of patients to obtain CO determinations. Blood pressure is considered a reliable indicator of cardiovascular function, and blood pressure is typically monitored by direct or indirect methods during anesthesia. In the study reported here, CI had a poor correlation with blood pressure values but a better correlation with heart rate. Similar results have been reported¹⁵ in dogs. Cardiac performance should be interpreted with caution when based on blood pressure alone because systemic vascular resistance determines blood pressure to a great extent, but it adversely affects CI. These effects have been reported^19,20 in adult horses ventilated during halothane- or isoflurane-induced anesthesia in which CI had an inverse relationship with MAP, which was attributed to changes in systemic vascular resistance.

The NICO method is an ideal noninvasive technique for determination of CO in anesthetized foals that are mechanically ventilated. The method is limited by the need to ventilate patients, the necessity for tolerance to a higher ETCO₂ concentration during the rebreathing cycle, and the fact a 3-minute cycle necessary to obtain a new measurement may delay the detection of a substantial change in CO. However, because it is easy to use, is minimally invasive, and performed well in the study reported here, NICO compares well with LiDCO, although overestimation for was evident at low values of CI.