The use of nitric oxide (NO) as a specific pulmonary vasodilator offers several clinical advantages over vasodilators like tolazoline, prostacyclin and nitroprusside. In the treatment of conditions associated with pulmonary hypertension and ventilation-perfusion mismatch [1,2], it causes selective pulmonary vasodilation with no systemic effect. Thus, ventilation-perfusion mismatch and cardiac instability  are reduced. Nitric oxide use as a pulmonary vasodilator has important clinical potential. However, as it is not a licensed drug, no standardized procedure for its use exists; an issue of importance considering its potential toxicity. The conversion of NO to nitrogen dioxide (NO2) and hence formation of nitric and nitrous acids, is an additional concern with its use, severe pulmonary oedema, acid pneumonitis and death being associated risks . Nitrite, formed when NO dissolves in aqueous solutions, can also be hazardous. As well as acting systemically, it has been proposed that high concentrations of these anions could contribute to neural toxicity and potential carcinogenesis [5,6]. One of the objectives of this study therefore was to collect water samples from the respiratory circuit to be assayed for nitric acid. One study has investigated this , nitric acid was detected following administration of 80 ppm NO in 90% oxygen, humidified at 37°C.
Many studies have attempted to determine the behaviour of NO and devise an optimal delivery system [8-10]. Although NO equipped ventilators are now available commercially, there is concern that commonly used delivery systems produce widely variable NO concentrations , particularly when combining continual infusion and intermittent flow ventilators [11-14]. These factors tend to produce a 'bolus' effect. As the inspiratory limb fills with NO during the expiratory phase , the patient may receive a gas mixture during the subsequent breath which has both a reduced oxygen content and a significant peak NO level . Nevertheless, continuous infusion is considered acceptable for neonatal ventilator systems where the constant flow exceeds the patient's minute ventilation . Identification of the optimal infusion site in NO delivery systems is fundamental in securing the future of NO as a clinical drug. Studies in this field are currently inconclusive. Many suggest that NO should be administered close to the patient in order to minimize the potential reaction time with oxygen . Conversely, others have suggested that NO should be administered close to the ventilator to ensure stable, low peak, concentrations [12, 15]. To address this issue, infusion sites were compared using a 'model lung' to identify the 'optimal' infusion site for delivery of stable, predictable and low levels of NO with as low concentrations as possible of NO2[4,16,17]. The effect of minute ventilation on the pattern and concentration of NO delivered was also observed.
The effect of a humidifier on NO delivery is largely unknown . Current opinion is that NO should be administered post humidifier, so that its transit and thus oxidation times in the inspiratory limb are reduced . This study investigated the effect of a humidifier on NO delivery.
There is also conflicting evidence concerning the optimal sampling point. Several studies conclude monitoring should take place as close to the patient as possible [1,12]. Cuthbertson and co-workers  conclude NO and NO2 concentrations will be higher at the alveoli than elsewhere in the delivery system. Many studies have sampled at the trachea . However, sampling with slow response analysers has been viewed with scepticism as NO concentrations delivered to the alveoli may be underestimated. This study aimed to compare NO sampling at alveolar and tracheal ports to determine the optimal point for NO monitoring.
Materials and methods
A model lung was built to simulate those of an infant (Fig. 1). Nitric oxide was infused at three sites on the inspiratory limb and sampled at the 'alveolar and tracheal' ports in the model. The minute ventilation was changed from 2 to 3, 4 and 5 L min−1, increasing tidal volumes from 57.1 mL to 95.9 mL, 114.3 mL and 142.9 mL respectively. At each sampling port, each combination of infusion site and minute ventilation were repeated six times in random order, to assure repeatability.
Materials - model lung
A 1 litre anaesthetic rebreathing bag was used to simulate the mechanical properties of the infant lung (Fig. 1b) during ventilation. Two hinged aluminium plates 3 mm thick were constructed to hang on either side of the rebreathing bag. To improve the modelling of elastic recoil, an elastic band circled the base of the two plates. Attached to one plate was a cantilevered and weighted arm serving to collapse the lung to its functional residual capacity on expiration, i.e. 250 mL for a 8-kg infant . This model lung was suspended in a sealed glass bell jar representing the thorax. A subatmospheric pressure of approximately 20 mm water was created in the thorax by withdrawing air via a 1 litre gas syringe and monitored using a simple graduated water manometer connected to the bell jar. Tubing modelled the tracheal compartment with an anatomical deadspace of 60 mL. A standard ventilator circuit (Fig. 1a) connected the model to a time-cycled, pressure-regulated, flow generator ventilator (Engström system ER300, LBK Medical AB, Bromma, Sweden). A calibrated pneumotachograph (Gould Godart BV, Gould Medical Products Division, Bilthoven, Netherlands) was connected to the inspiratory limb near to the ventilator to monitor flow rates and tidal volumes. The ventilator was set at a frequency of 30 breaths per min, with an end-expiratory pressure of zero. Distilled water levels in the humidifier were checked before each run.
Three infusion sites along the inspiratory limb were chosen (Fig. 1a). Site 1 proximal to the humidifier, site 2 distal to the humidifier and site 3 near the Y-piece thus representing infusion at the patient end. Plastic cannulae with 5 mm internal bores and Luer fittings were inserted into the silicon sleeves of the tubing ensuring gas delivery to the central bore. Medical grade NO (1000 ppm), balance nitrogen (BOC Special gases, BOC Ltd, Guildford, UK), was infused at a constant rate. A calibrated Rotameter (0-250 mL min−1) was used to monitor the flow rates being delivered. The gas was delivered for at least 4 min or until a stable reading at the sampling port was obtained.
Gases were sampled from the alveolar and tracheal compartments through separate sampling ports. Nitric oxide and NO2 were measured using a Noxbox inhaled NO therapy monitor (Bedfont Scientific Ltd, Upchurch, Kent, UK). Sample flow rates were controlled at 100 mL min−1 and 250 mL min−1, respectively, by a calibrated Rotameter.
The ventilator and gas analyser were scavenged via tubing to outside air. A sealed water trap was inserted into the inspiratory limb for acid analysis. At the end of each experimental run, any water found in the water trap, water sampled from the humidifier and any condensed water from the scavenging tube were collected to establish the presence of acid. A further complete run of the experiment with NO was also carried out bypassing the humidifier.
The data distribution was assessed from normal probability plots and parametric statistics employed when homogeneity was shown. Data were analysed descriptively and results presented as mean ± SEM. Further analysis of data used a 2 by 3 by 4 multivariate analysis of variance (MANOVA). Polynomial regression analysis was used to assess the relation between minute ventilation and NO. Probability values of P < 0.05 were taken to be significant.
Sampling port, infusion site and minute ventilation all had significant effects (P < 0.05) on the concentration of NO measured. A Wilks' Lambda test confirmed all the interactions between these variables were significant (P < 0.05). (Sampling port and minute ventilation, sampling port and infusion site, minute ventilation and infusion site and the combined interaction between all three variables).
Effect of minute ventilation
Figure 2 shows the predicted NO concentrations which would be expected given complete mixing between the delivered NO and the ventilated volumes at the 4 min volumes used. Figure 2 also shows the plot of the steady-state NO concentrations measured at the tracheal port when NO was delivered at site 1. The concentrations are consistently and significantly (P < 0.05) lower than predicted.
Effect of infusion site
Similar curved relations were observed at the tracheal port at each infusion site. Furthermore, the alveolar NO concentrations followed a similar pattern when NO was infused away from the model trachea (sites 1 & 2), but not when infused close to it (site 3). As these data were repeatable, significant (P < 0.01) second order polynomial exponential relation between NO concentrations and minute ventilation were found for each infusion site. It follows that the NO concentrations are predictable, but the relation for each combination of infusion site, infusion rate, minute ventilation and sampling port would have to be found (i.e. in practice, for each patient).
The steady-state NO concentrations sampled at the tracheal port did not differ (Fig. 3) when the infusion was delivered at sites 1 & 2. When using site 3, the concentrations were significantly higher with a minute ventilation of 2 L min−1 and significantly lower with 4 and 5 L min−1 minute ventilation (P < 0.05 in all cases).
The NO concentrations sampled at the alveolar port were similar to those measured at the tracheal port when the infusion was given at sites 1 & 2 but were significantly higher (P < 0.05) when the infusion was close to the model trachea, site 3.
NO concentrations in the compartments
Nitric oxide concentrations found when infusing at sites 1 and 2 were very similar between the two sampling ports (P > 0.05, Fig. 3). However, when site 3 was employed, the NO concentrations at the alveolar port were significantly and markedly raised (P < 0.05, Fig. 3) compared with those at the tracheal port.
When the humidifier was omitted from the circuit, a 2 by 2 analysis of variance and Scheffé post hoc analysis test identified a non-significant (P > 0.05) effect on NO delivery observed for differing infusion sites, sampling ports and minute ventilations.
All fluid samples taken from the humidifier were found to be slightly acidic (pH 6). Samples taken from the sealed water trap in the inspiratory limb post humidifier were also found to be slightly acidic (pH 6). When the humidifier was bypassed the water trap acidity was increased (pH 5). Fluid collected from the ventilator exhaust vent was also acidic, pH 5. Assuming that the hydrogen ion resulted from nitric acid accumulation, equivalent to 0.1 mL of nitric acid in 1 litre of water, given a molarity of 0.002.
Nitrogen dioxide production
The concentrations of NO2 measured at the two sampling ports were significantly correlated with NO concentration (r = 0.93, P < 0.001).
The NO concentrations measured by the analyser are low compared with the calculated level. There are well documented problems of continuous flow infusion into an intermittent flow ventilator [10-14]. As the NO was infused continuously, it presumably is delivered in relatively concentrated 'boluses' with a peak and trough pattern.
The lower measured concentrations may be due to the analyser output being unrepresentative of an appropriate averaged measure. The response time of the NO meter is 10 s, slow compared the duration of the 'bolus' and the intervening period given a respiratory rate of 35 breaths per min. Whatever the cause there are important clinical implications. A delivery system to minimize the 'bolus' effect has been reported (Young's 'universal delivery system', 1994).
The pattern of NO delivery to the alveolar compartment varied depending on where it was infused, at what minute ventilation the ventilator was operating and from where the NO was being sampled. Moreover, an exponential relation was found between NO concentrations observed in the model compartments and the ventilation rate used. If shown in the clinical situation, then predictive equations will be needed for each combination of infusion rate, minute ventilation, respiratory rate and possibly lung compliance.
Effect of infusion site
It has been shown that the administration of NO proximal or distal to the humidifier (sites 1 and 2), results in stable and predictable concentrations of NO delivery . No significant difference between infusion sites 1 and 2 was found when sampling at the model trachea. An exception was found at the alveolar port at lower minute ventilations, slightly higher NO concentrations resulted from infusion at site 2. When the infusion was close to the model lung (site 3), the steady-state concentrations of NO were different. At the tracheal port with low minute ventilation, the NO concentration was raised. The opposite occurred with higher minute ventilation. In these circumstances the NO concentrations measured at the alveolar port were all markedly higher. Tracheal sampling has been suggested previously to underestimate alveolar NO concentration . The cause is unknown but it could be a feature of 'bolus' build-up: the more concentrated NO being delivered to the alveolar compartment during early inhalation, although sample withdrawal from this compartment in the model may have been a contributory factor. So too may have been the large dead-space (60 mL) in the model.
The possibility that alveolar concentrations are higher than those sampled in the trachea raises grave concerns if replicated in the clinical situation and requires assessment of alveolar levels in practice. This may be difficult in critically ill patients . Infusing NO furthest from the model lung reduced the NO concentration reaching the alveolar compartment, but provides equivalent data whether the sample was taken from the tracheal or the alveolar port. The variation, measured by calculating the SEMs, was lowest with the more distal infusions, the reverse of that suggested by Young and Dyar (1996). These differences emphasize the need for further information on the optimal monitoring point for NO and NO2. It has even been suggested that tracheal sampling may be 'useless' at determining the concentration of NO or NO2 in inhaled gas . Certainly enhanced speed of NO measuring instruments capable of providing a breath-by-breath profile in the trachea for 'alveolar like' measures during exhalation are needed. Without rapid NO measurements, the model supports the suggestion [12, 15] that infusion of NO close to the patient should be avoided but it raises concerns that the delivery of NO to and resulting concentrations in the alveolar compartment are not known with certainty.
A correlation between NO and NO2 concentrations has been reported previously . Although air was used in our model producing low oxygen tensions, a significant, positive correlation was found as expected.
The pattern of NO observed across the sampling ports, infusion sites and minute ventilations was comparable both with and without the humidifier in the circuit. This negligible effect on the concentrations of NO measured confirms the prediction that 'the position of the humidifier is not critical'  and does not serve to improve gas mixing. Nitrogen dioxide concentrations were no higher when the humidifier was bypassed. This observation supported the prediction that the humidifier may delay the passage of gas through the circuit but does not appear to lead to a greater NO2 concentration .
The finding of acid in the water trap and humidifier has serious clinical and financial implications. In comparison with the length of time NO may be administered clinically [21,22], our model lung operated for short periods but nonetheless sufficient acid was produced.
Greater acidity was found in the water trap following experimentation when the humidifier was bypassed. Contrary to recommendations by Young and Dyar , this suggests NO infusion should occur proximal to the humidifier to attenuate acidification. In any case, the adverse effects of severe pulmonary oedema and acid pneumonitis caused by nitric acid  are potentially an issue.
This model employed a constant NO infusion into an intermittent ventilator circuit. NO and NO2 were measured with a slow response meter. Therefore the model replicated a typical clinical situation. Although there are now devices available for intermittent NO infusion , many clinicians will still be denied this option. In conclusion, the model suggests that NO should be infused close to the ventilator, prehumidifier (site 1) and that the resulting NO concentrations should be measured in the trachea. The reasons are: (i) under these circumstances the NO concentrations measured at the tracheal port were stable and mirrored those found at the alveolar port; and (ii) the finding of acid in the water-trap can be taken as proof that acid would be deposited in the airways. The model indicated that greater attention needs to be paid to this latter feature of NO therapy.
The model shows that, whilst calculation and prediction of NO and NO2 concentrations is possible, the measured concentration will depend on the location of the infusion site, site of measurement, infusion rate, minute volume, respiratory rate and PO2. In practice these will vary, so too will patients. Calculation of delivered concentration will reveal the mathematically averaged dose to which the lungs may or may not have been exposed. It is possible that information from the above monitoring system may in practice be no more accurate than that calculated, and that the measured value displayed by the monitor is really a product of all of the above uncertainties and the response time of the analyser itself. In spite of this, it is necessary to measure NO routinely and consistently (to prevent accidental overdose or disconnection). Measuring at the trachea, whilst essential to monitor treatment, is still only an approximation of the concentration delivered to the alveoli.
We thank Dr Rhys D Evans, Consultant Anaesthetist, Nuffield Department of Anaesthetics, Radcliffe Infirmary, Oxford, for helpful comments on this manuscript.
1 Miller OI, Celermajer DS, Deanfield JE, Macrae DJ. Guidelines for the safe administration of inhaled nitric oxide. Arch Dis Childhood
2 Glenfield Hospital Trust. Guidelines for the use of Nitric Oxide.
1996. Department of Anaesthesia, Glenfield General Hospital, Leicester.
3 Brett SJ, Evans TW. Inhaled vasodilator therapy in acute lung injury: first, do NO harm? Thorax
4 Frostell C, Fratacci M-D, Wain JC, Jones R, Zapol WM. Inhaled nitric oxide: a selective pulmonary vasodilator reversing hypoxic pulmonary vasoconstriction. Circulation
5 Warren JB, Higenbottam T. Caution with use of inhaled nitric oxide. Lancet
6 Tamir S, Tannenbaum SR. The role of nitric oxide (NO) in the carcinogenic process. Biochimica Biophysica Acta - Rev Cancer
7 Putensen C, Rasanen J, Thomson MS, Braman RS. Method of delivering constant nitric oxide concentrations during full and during full and partial ventilatory support. J Clin Monitoring
8 Young JD. A universal nitric oxide delivery system. Br J Anaesthesia
9 Tibballs J, Hochmann M, Carter B, Osborne A. An appraisal of techniques for administration of gaseous nitric oxide. Anaesth Intensive Care
10 Imanaka MD, Hess D, Kirmse M et al.
Inaccuracies of nitric oxide delivery systems during adult mechanical ventilation. Anesthesiology
11 Foubert L, Mareels K, Fredholm M, Lundin S, Stenqvist O. A study of mixing conditions during nitric oxide administration using simultaneous fast response chemiluminescence and capnography. Br J Anaesthesia
12 Cuthbertson BH, Dellinger P, Dyar OJ et al.
UK guidlines for the use of inhaled nitric oxide therapy in adult ICUs. American-European Consensus Conference on ALI/ARDS. Intensive Care Med
13 Moors AH, Pickett JA, Mahmood N, Latimer RD, Oduro A. Nitric oxide administration. Anaesth Intensive Care
14 Sydow M, Bristow F, Zinserling J, Allen SJ. Variation of nitric oxide concentration during inspiration. Crit Care Medicine
15 Valley RD, Zar H, Boysen PG. Nitric Oxide infusion during controlled mechanical ventilation. Crit Care Medicine
16 Westfelt UN, Lundin S, Stenqvuist O. Nitric oxide administration after the ventilator: Evaluation of mixing conditions. Acta Anaesthesiol Scand
17 Young JD, Dyar OJ. Delivery and monitoring of inhaled nitric oxide. Intensive Care Med
18 Cotes JE. Lung Function,
4th Edn. Oxford: Blackwell Scientific Publications, 1979.
19 Cuthbertson BH, Stott S, Webster NR. Inhaled nitric oxide. Lancet
20 Kelly PKP, Deehan SC, Grant ZS. Nitric oxide delivery system. Br J Anaesthesia
21 Pearl RG. Inhaled nitric oxide - The past, the present and the future. Anesthesiology
22 Rossaint R, Falke KJ, Lopez F, Slama K, Pison U, Zapol WM. Inhaled nitric oxide for the adult respiratory distress syndrome. New Engl J Med
23 Young JD, Roberts M, Gale LB. Laboratory evaluation of the I-NOvent nitric oxide delivery device. Br J Anaesthesia