Inhaled nitric oxide (NO) is used as an investigational treatment for intraoperative pulmonary hypertension and hypoxemia (1–3). With adult critical care ventilators (4), NO delivery is simplified because fresh gas flow (FGF) is delivered during inspiration only; the circuit is open and gas rebreathing does not occur. In neonatal ventilators (4), FGF is continuous throughout the respiratory cycle; again, the circuit is open and rebreathing does not occur. With anesthesia machines, NO delivery is more complicated because FGF is continuous and the breathing circuit semi-open; thus, rebreathing can occur. During inspiration, the patient receives a variable combination of fresh and exhaled gas. During exhalation, the circuit may be filled by FGF without NO, FGF with a set NO concentration ([NO]), or exhaled gas with a variable [NO]. The breathing circuit is further complicated by the presence of a gas reservoir (bag or ventilator bellows) and a carbon dioxide (CO2) absorber. Ultimately, inspired [NO] is the result of several variables, including the [NO] added to the system, FGF rate, ratio of inspiratory-to-expiratory (I:E) time, breathing circuit size, patient minute ventilation (V̇E), dead space (VD), and NO uptake.
Most studies of intraoperative NO inhalation have not described their delivery systems in detail, relying instead on NO analyzers to gauge the NO dose (1–3). The analyzers used, however, had a slow response times (several seconds), which underestimate variations of [NO] (5). Increased accuracy is achieved with fast response analyzers, but these have practical disadvantages: their response times must be verified with appropriate techniques (6), they do not always analyze nitrogen dioxide (NO2), and they are more expensive and delicate than slow response analyzers (4).
We studied the accuracy of two methods of intraoperative NO administration. With the first method, NO instead of nitrous oxide (N2O) was delivered via the N2O flowmeter of the anesthesia machine. With the second, NO was delivered by the INOvent (Datex-Ohmeda, Madison, WI). Our principal hypothesis was that NO can be administered accurately only when the FGF rate is higher than the patient’s V̇E. At lower FGF, NO delivery would be less predictable because of the complexity of gas rebreathing.
The experimental setup is shown in Figure 1. The two chambers of a lung model (Training Test Lung model 1600; MI Instruments, Grand Rapids, MI) were connected by a lift bar, so that one lung was inflated with FGF by positive pressure and lifted the other passively. Lung compliance was set at 30 mL/cm H2O to allow full exhalation. A standard 100-cm anesthesia breathing circuit (Sims Medical Systems, Ft. Myers, FL) was connected to the lung model through a “Y” composed of three 15-cm corrugated tubes with inner diameters of 22 mm. The lung model VD was 180 mL.
One lung received FGF from the anesthesia machine via a one-way valve (BE 142–50; Instrumentation Industries, Pittsburgh PA) and exhaled into the room. The second lung filled passively with room air during inspiration via another one-way valve (Airlife 001504; Baxter Healthcare, Deerfield, IL) and exhaled into the anesthesia circuit. This set up (5,7) allowed independent control of inspired and expired gases and simulation of various degrees of NO uptake. We set three NO uptake ratios during 10 ppm NO inhalation by delivering various exhaled [NO] to the second lung: 0 ppm for 100% uptake, 4 ppm for 60% uptake, and 7 ppm for 30% uptake. This was done by connecting in a T-fashion the one-way valve of the passive lung to a corrugated tube filled with the appropriate mixture of NO and air, obtained with standard flowmeters.
We tested three anesthesia machines: Ohmeda Modulus II ([OMII]; Datex-Ohmeda), Ohmeda Excel 210 (OE210; Datex-Ohmeda), and Dräger Narkomed II ([DNII], North American Dräger, Telford, PA). Two 2.5-lb soda lime packs (Sodasorb; WR Grace, Atlanta, GA) were lodged in the CO2 canister. No part of the machine was modified from the usual clinical setup.
NO Delivery via the N2O Flowmeter
A tank containing 800 ppm NO in nitrogen (N2) (INO Therapeutics, Liberty Corner, NJ) was connected to the N2O inlet of the anesthesia machine by a specially fitted pressure hose. The gas pressure was 50 psi. The [NO] delivered was calculated by a standard formula (4):MATH
NO Delivery via the INOvent
The INOvent is a commercial delivery system that injects NO proportionally to the inspiratory flow. Its monitoring unit consists of slow response electrochemical NO and NO2 analyzers. The INOvent injection module was placed just distal to the inspiratory valve and the sampling line just proximal to the Y-piece, as recommended by the manufacturer (4,7).
Measurements and Calibrations
Inspired [NO] was measured continuously by a chemiluminescence analyzer (Sievers model NOA 280; Sievers Instruments, Boulder, CO) with a response time of 170 ms (5). The analyzer was calibrated with 0 and 25 ppm NO. The [NO] signal was digitized at 100 Hz and recorded by wave form analysis software (Windaq; Dataq Instruments, Akron OH). Gas was sampled for analysis at four different locations along the anesthesia breathing circuit: just distal to the inspiratory valve, at the Y-piece, distal to the exhalation valve, and proximal to the ventilator bellows. A screen pneumotachometer (Hans-Rudolph model 3700A, Kansas City, MO) was inserted at the Y-piece. Its pressure differential was measured by a pressure transducer (Validyne 45–142-871, ± 2 cm H2O, Northridge, CA), converted into flow, amplified (Hewlett-Packard amplifier 8805C, Waltham, MA), digitized at 100 Hz, and recorded. The pneumotachometer was calibrated at 60 L/min with air delivered by a precision rotameter (Brooks Instruments, Hatfield, PA). The tidal volume (VT) was obtained by digital integration of the flow signal. Pressure at the Y-piece was measured with a pressure transducer (Validyne 45–32-871, ± 100 cm H2O), amplified (Hewlett-Packard amplifier 8805), anddigitized at 100 Hz and recorded.
The accuracy of flowmeters delivering N2O and NO in N2 was tested in several ways. First, we measured gas volume at the FGF outlet of a DNII machine by setting known flows of N2O and O2, collecting the gas in a 6-L anesthesia bag for precisely 1 min, and measuring its volume with a calibrated 1-L syringe (Hamilton, Reno NV). We also measured O2 volume delivered by the hospital O2 supply at the wall by a standard O2 flowmeter (Precision Medical, model 2MFA1001, Northampton, PA). Second, we measured [NO] at the FGF outlet after setting the O2 and N2O flowmeters of an OMII machine to deliver 10 ppm NO at 2, 4, 6, 8, and 12 L/min FGF. An OMII was used because low flow rates can be read more easily on its N2O flowmeter. Four additional experiments were performed with an OMII, OE210, and DNII (twice), set to deliver multiple [NO] between 6.5 and 75 ppm. In three experiments, [NO] also was measured immediately distal to the CO2 absorber. Third, we used a bubble flowmeter (Digital Flowmeter 650, Fisher Scientific, Pittsburgh, PA) to measure various flow rates of N2O or N2 by the N2O flowmeter of an OMII machine. All volume and concentration measurements were repeated in triplicate.
NO was administered by an OMII machine to the lung model. Ventilator settings in all experiments were: volume-limited ventilation, 5–6 L/min V̇E, 700 mL exhaled VT, and rate of 8/min. FGF rates of 2, 4, and 6 L/min were tested at I:E ratios of 1:3 and 1:1. The VT was kept constant. The N2O flowmeter was set to deliver 10 and 20 ppm NO. The INOvent was set at 10 ppm NO. Three NO uptake ratios, 100%, 60%, and 30%, were tested. Measurements were performed after 5 min of equilibration. The effect of soda lime on [NO] delivered was tested by comparing a fresh pair of soda lime packs with a fully used pair during NO rebreathing induced by decreasing NO uptake to 60% and 30%. Both delivery systems were also tested during manual and spontaneous ventilation (simulated by manually lifting the lungs) (5), varying the VT and rate within physiological values. Continuous measurements were obtained during investigational protocols from two individuals breathing NO through an anesthesia machine.
All data were recorded continuously and stored for analysis. To identify [NO] at exact times during the breathing cycle in the recorded traces, the [NO] signal was adjusted for transport delay (approximately 1 s) based on the flow trace. Peak [NO], mean inspired [NO], and time-averaged total inspired NO amounts were measured electronically. When multiple measurements were obtained, data were expressed as mean ± SD. Error between calculated and measured [NO] was expressed in percent as bias, precision, and limits of agreement (8).
The volumes of O2 plus N2O, and of O2 plus NO in N2 delivered at 3 and 6 L/min FGF were slightly higher (maximum, +7.5%) than the volumes set. Volumes of N2O and of NO in N2 delivered by the N2O flowmeter were equal. Variability among triplicate measurements was trivial. For the sake of brevity, we did not report these results.
The N2O and N2 flow rates delivered by the N2O flowmeter of an OMII machine measured at the FGF outlet by a bubble flowmeter are shown in Figure 2. At low flow, 0.02–0.300 L/min, N2 delivery was slightly lower (−3 to −12%) than N2O delivery. At a higher flow, 1 L/min, N2 delivery was slightly higher (3.5%) than N2O delivery.
The [NO] delivered by the N2O flowmeter of an OMII machine set to administer 10 ppm at 2, 4, 6, 8, and 12 L/min FGF was 11.1 ± 0.2 ppm, 10.8% (bias) ± 1.7% (precision) higher than the [NO] set. Various [NO] delivered by all the machines tested are shown in Figure 3. The OMII delivered an [NO] 18% ± 4.4% higher than set, the OE210 delivered an [NO] 7.4% ± 5.3% lower than set, and the DNII delivered [NO] 8.1 ± 5% lower than set. The [NO] measured at the CO2 canister was identical to the [NO] at the FGF outlet.
NO Delivery to the Lung Model via the N2O Flowmeter
The N2O flowmeter of the OMII was set to deliver 10 and 20 ppm NO during positive pressure ventilation. Because no difference was found in the delivery of these two [NO], we present only the 10 ppm data (Figure 4).
Inspired [NO] measured at the inspiratory valve and the Y-piece were equal, indicating adequate gas mixing. With 100% NO uptake, FGF ≥ V̇E (6 L/min) and 1:3 I:E ratio, inspired [NO] was 11.7 ppm, with a stable inspiratory plateau (Figure 5A). With 4 and 2 L/min FGF and 1:3 I:E ratio, [NO] decreased to 8.8 and 5.2 ppm, respectively, and the [NO] wave form was far less stable (Figure 5B). A 1:1 I:E ratio, not shown in the figure, resulted in slightly lower [NO]: 11.3 ppm at 6 L/min FGF, 8.5 ppm at 4 L/min, and 5.1 ppm at 2 L/min. With lower NO uptake (60% and 30%) and FGF ≥ V̇E, inspired [NO] did not change. At FGF < V̇E, however, inspired [NO] increased: at 4 L/min it was 10% and 18% higher with 60% and 30% uptake, respectively, and at 2 L/min, it was 33% and 69% higher with 60% and 30% uptake, respectively. Despite the increase caused by reduced NO uptake, inspired [NO] always decreased with low FGF. The [NO] at the expiratory valve and ventilator bellows was approximately 20% of the inspired [NO] with 100% NO uptake, representing VD ventilation, and increased when NO uptake decreased (Figure 4).
Inspired [NO] delivered via the N2O flowmeter was similar when either new or exhausted soda lime was used. At 4 L/min FGF, [NO] was 11.0 ppm and 11.3 ppm at 60% and 30% NO uptake, respectively, with both new and exhausted soda lime. At 2 L/min FGF, [NO] was 9.1 and 9.8 ppm at 60% and 30% NO uptake, respectively, with new soda lime and was 8.9 and 9.6 ppm at 60% and 30% NO uptake, respectively, with exhausted soda lime.
NO Delivery to the Lung Model via the INOvent
The inspired NO wave form showed a characteristic “hump” (Figure 5D). The amount of NO composing the hump, quantified by digital analysis, was 0.71 ± 0.15 μL and did not change with different [NO] and I:E ratios. This fixed NO volume was more evident at low inspired [NO] (Figure 5E) and short I:E. It accounted for 3.3% of the inspired NO at 20 ppm, 7.1% at 10 ppm, and 25.4% at 2 ppm.
NO delivery with the INOvent and the effects of NO rebreathing are summarized in Figure 6. Only values obtained with 1:3 I:E ratio are reported because they were equivalent to the values at 1:1 I:E ratio. With 100% NO uptake and FGF ≥ V̇E (6 L/min), inspired [NO] (i.e., plateau, excluding the hump) at the Y-piece was slightly lower (<0.8 ppm) than the [NO] set. At lower FGF, inspired [NO] increased to 10.9 ppm at 4 L/min and 12.3 ppm at 2 L/min. With 60% and 30% NO uptake and FGF ≥ V̇E (6 L/min), [NO] proximal to the injection site remained negligible and rebreathing did not occur. At FGF < V̇E rebreathing occurred. At 4 L/min, inspired [NO] increased to 12 and 12.3 ppm with 60% and 30% uptake, respectively; at 2 L/min, inspired [NO] increased to 13.1 and 16 ppm with 60% and 30% uptake, respectively. The [NO] at the expiratory valve and ventilator bellows was approximately 30% of the inspired [NO] with 100% NO uptake, representing VD ventilation, and increased equally when NO uptake decreased (Figure 6).
Manual and Spontaneous Ventilation
During manual and simulated spontaneous ventilation, delivery of 10 ppm NO via the N2O flowmeter or the INOvent at 6 and 4 L/min FGF followed the same patterns observed during mechanical ventilation. During spontaneous ventilation with the INOvent, no hump was present.
Patients’ Wave Forms
Inspired [NO] wave forms were collected from two individuals during 10 ppm NO inhalation via the N2O flowmeter of the OMII and via the INOvent. Representative breaths, shown in Figure 5, C and F, are comparable to those obtained with the lung model.
This lung model study demonstrates that inhaled NO can be delivered accurately with an anesthesia machine using either its N2O flowmeter or an INOvent. Inspired [NO], however, can differ from the set [NO] under a number of circumstances. We defined sources of error and provided guidelines for intraoperative NO delivery.
NO Delivery via the N2O Flowmeter or an Anesthesia Machine
Flowmeters are calibrated for a specific gas. If a new gas is used, its flow may differ from what is set, depending mainly on its physical characteristics, i.e., density at turbulent flow and viscosity at laminar flow (9). In our case, N2 (the amount of NO in the NO/N2 mixture is negligible) has a lower density than N2O (1.25 and 1.97 g/L respectively) and a higher viscosity than N2O (1.74 × 10−5 and 1.42 × 10−5 PQ · s, respectively). Thus, when NO in N2 is delivered by an N2O flowmeter, one can expect a higher [NO] than set by at turbulent flows and a lower [NO] at laminar flows. At flow rates sufficient to generate turbulence, e.g., >1 L/min, the effect of a new gas’s density can be calculated (10). At the low flow rates (20–100 mL/min) used to deliver NO, however, flow is likely to be laminar, and the effect of a new gas’s viscosity has not been examined systematically. Furthermore, gas flow through an anesthesia flowmeter is also affected by the accuracy and physical condition of the flowmeter and the shape and weight of the float (11). The combination of these factors complicates attempts to construct a simple model to predict the [NO] delivered by an N2O flowmeter.
Our results suggest that a flowmeter calibrated for N2O delivers a slightly lower (maximum, −3.4%) N2 flow than set at low flow, and a slightly higher (maximum, 10.6%) N2 flow than set at high flow, (Figure 2) because of the higher viscosity and lower density of N2. The N2O flow rate itself (Figure 2), however, was consistently measured higher than set, (maximum, 13.5%) in agreement with the existing literature (12,13). Thus, the −8.1% to 18% bias of the N2O flowmeters tested to deliver NO (Figure 3) was caused at least in part by the reduced accuracy of N2O flowmeters at very low flow rates, regardless of which gas was delivered. More importantly from a practical standpoint, the N2O flowmeters displayed a high precision, i.e., only a 4.1%–5.4% variability of the measured [NO] during repeated settings.
NO delivery via the N2O flowmeter during positive pressure ventilation was accurate at FGF ≥ V̇E but not at FGF < V̇E (Figures 4 and 5). At low FGF, the inspired [NO] plateau was not stable, because an increasing proportion of the VT was made up by rebreathed gas with a lower [NO]. Under these circumstances, the inspired [NO] wave form showed a slow ascent, did not reach a stable plateau, and continued to rise during early exhalation, as a result of the rebreathing of NO-rich dead space gas (Figure 5B). This pattern of gas delivery is unpredictable and unsuitable for clinical applications.
NO delivery via the INOvent has been proved accurate with critical care ventilators (4,7), but delivery with anesthesia machines has not been investigated. The inspired [NO] wave form of the INOvent showed a distinctive “hump” (Figure 5D) representing a fixed volume of NO unrelated to the [NO] set, the airway pressure, or the FGF. At inspired [NO] ≥ 10 ppm, this extra volume was of little practical relevance, (<10% of total NO) but at lower [NO] it represented a significant portion of the delivered NO. We speculate that the inspiratory pressure within the breathing circuit compresses gas within the INOvent injection tubing, which then decompresses at the onset of exhalation, delivering a small bolus of NO with the next breath. Accordingly, the hump moved toward the beginning of inspiration when we injected NO closer to the Y-piece, decreased in size when we reduced the length of the injection tubing, and was undetectable during simulated spontaneous breathing. While this hump has not been previously described, it is evident in Figure 2 of Kirmse et al.’s article (7). It is likely that the volume of NO confined to the hump accounted for the slightly lower [NO] that we reproducibly measured at the inspiratory plateau in our experiments.
Ten ppm NO was delivered accurately by the INOvent at FGF ≥ V̇E. At lower FGF, rebreathing caused the inspired [NO] to increase. This contrasts with delivery by the N2O flowmeter, where low FGF decreased inspired [NO]. The different effects of NO rebreathing on delivery by the two systems are explained by the different sites of NO injection: distal to the inspiratory valve with the INOvent and proximal to the circuit and inspiratory valve with the N2O flowmeter. Because the INOvent adds a set amount of NO to the mixture of fresh and exhaled gas, the rebreathing of exhaled NO increases inspired [NO]. With the N2O flowmeter, the [NO] set is a portion of the FGF. When FGF enters the circuit, it is diluted by rebreathed gas, which decreases the inspired [NO]. The physical state of the soda lime did not affect inspired [NO] during rebreathing, in agreement with a previous study (14).
Our study has several limitations. The complexity of pulmonary physiology cannot be adequately reproduced with a lung model. We were able, however, to simulate the clinical setting more closely than in previous studies (5,7) by varying NO uptake. The wave forms produced by the model closely mimicked those recorded during NO inhalation by patients (Figure 5, C and F). We tested only three anesthesia machines, and other machines could provide different results. Flowmeters, however, are simple instruments, and it seems reasonable to predict that the performance of an appropriately serviced flowmeter will be within the range that we observed (11). We did not systematically analyze NO2 production, but at the low inspired [NO] and the high FGF used intraoperatively, NO2 production should be minimal.
Both systems perform appropriately at FGF ≥ V̇E.
With the N2O flowmeter, the delivered [NO] may vary from the [NO] set and always should be confirmed with an analyzer. Rebreathing of NO greatly decreases the accuracy of NO delivery, but the inspired [NO] cannot be greater than the [NO] delivered at high FGF rates. Inspired [NO] < 10 ppm can be delivered accurately if an appropriately high FGF rate is used. Delivering NO via the N2O flowmeter prevents the use of N2O.
With the INOvent, NO rebreathing increases inspired [NO]. Delivery of <10 ppm NO is less accurate, because of the addition of a small bolus of NO to each breath.
With either system, it is prudent to monitor the inspired [NO] during clinical applications.
We are grateful to Ricardo Mills for his outstanding technical help and to Robert M. Kacmarek for supporting our lung model studies.