The Role of High-Frequency Oscillatory Ventilation in the Management of Acute Respiratory Distress Syndrome A Critical Review

The Role of High-Frequency Oscillatory Ventilation in the Management of Acute Respiratory Distress Syndrome A Critical Review Dr Stephen Wimbush Prepared and submitted for the Intercollegiate Diploma in Intensive Care Medicine April 2005

Contents Glossary of Terms... 3 Abstract... 4 Introduction and Background... 7 Objectives... 9 Methodology... 10 Mechanisms of High-Frequency Oscillation... 11 Neonatal Data... 13 Effect of High-Frequency Oscillation on Oxygenation... 15 Effect of High-Frequency Oscillation on Mortality... 18 Safety of High-Frequency Oscillation... 21 Other Roles for High-Frequency Oscillation... 25 Discussion... 27 Conclusions... 29 Recommendations... 30 Appendix... 31 References... 35 2

Glossary of Terms P Amplitude or driving pressure of the oscillator ANOVA Analysis of Variance APACHE II Acute Physiology and Chronic Health Evaluation score (1985) ARDS Acute Respiratory Distress Syndrome BPD Bronchopulmonary Dysplasia CO Cardiac Output CV Conventional Ventilation in Pressure Control mode CVP Central Venous Pressure ETT Endotracheal Tube Freq. Frequency of oscillation in Hertz (Hz) HFFI High-Frequency Flow Interruption HFJV High-Frequency Jet Ventilation HFOV High-Frequency Oscillatory Ventilation HFPPV High-Frequency Positive Pressure Ventilation HFPV High-Frequency Percussive Ventilation HR Heart Rate ino Inhaled Nitric Oxide LIS Lung Injury Score MAP Mean Arterial Pressure MeSH Medical Subject Heading mpaw Mean Airway Pressure OI Oxygenation Index PAOP Pulmonary Artery Occlusion Pressure PCV Pressure Controlled Ventilation PEEP Positive End-Expiratory Pressure PRISM Paediatric Risk of Mortality score (1988) SAPS II Simplified Acute Physiology score (1993) VALI Ventilator-Associated Lung Injury V t Tidal Volume 3

Abstract Background: Acute respiratory distress syndrome (ARDS) is a clinical syndrome of respiratory failure characterised by non-cardiogenic pulmonary oedema with radiological pulmonary infiltrates and hypoxaemia which is often refractory. Mechanical ventilation and treatment of the underlying cause form the mainstay of treatment in patients with ARDS. Mortality in recent years has reduced to between 30% and 40% from a previously static rate in the region of 50%. This has coincided with changes in ventilatory strategies such as the use of smaller tidal volumes, which has been shown to improve outcome in patients with ARDS. Over the same time period, but on a smaller scale, interest has developed in the use of high-frequency oscillatory ventilation for ARDS. High frequency oscillation is already widely used in neonatal intensive care in the management of respiratory disease of prematurity. Its characteristics of very low tidal volume ventilation and the ability to achieve a high mean airway pressure while maintaining an acceptable plateau pressure, combine to offer theoretical advantages of both a reduction in conventional ventilation-associated barotrauma and improved oxygenation. Objectives: The primary objective of this dissertation is to review the current literature with respect to the use of high frequency oscillatory ventilation in patients with ARDS. In addition the mechanisms of gas exchange in oscillated patients will be presented, as well as the theoretical basis for the use of this mode of ventilation in ARDS. As part of this process, the following questions will be addressed: Does high frequency oscillation result in improved oxygenation in patients with ARDS? Do the theoretical benefits of high frequency oscillation translate into improved outcome in patients with ARDS? Is it safe to ventilate patients using high frequency oscillation? Are there any roles for high frequency oscillation other than in patients with respiratory disease of prematurity or isolated ARDS? Outline and Methodology: References were primarily obtained by conducting a MEDLINE search using the PubMed database between 1966 and January 2005. Initial searching used the search term: (High Frequency Ventilation (MeSH)). 4

References were further sought and classified using the subset search terms: (High Frequency Oscillation) OR (HFO). Specific references relating to high frequency oscillation in ARDS were obtained by using the following search terms: (High Frequency Ventilation (MeSH)) AND (Adult Respiratory Distress Syndrome (MeSH)). All references obtained were cross-checked with results obtained by conducting the same search using the DialogStar database. In addition the reference lists of all relevant papers and reviews were hand-searched for any additional papers. Detailed analysis of the literature was performed with greater emphasis being placed on evidence obtained from large randomized controlled trials compared to that from smaller studies, case series and case reports. Evidence was assessed and graded from A to E based on a modified version of the grading system originally described by Sackett in 1989. Conclusions: The widespread use of high frequency oscillation in the treatment of respiratory disease of prematurity may be justified on the basis of efficacy as well as safety. Most studies confirm an early improvement in oxygenation although there is limited evidence of improved survival or reduced incidence of bronchopulmonary dysplasia. Of the published studies on the use of high frequency oscillatory ventilation in the management of ARDS, only two have been conducted on a prospective randomized controlled basis. One of these was in a paediatric population, although there have been more prospective studies and retrospective case series in adults. There is evidence of an early improvement in oxygenation in patients with ARDS treated with high frequency oscillatory ventilation compared to conventional ventilation, although these benefits are lost by 24 hours. There is a non-significant trend towards improved survival in these patients although changes in contemporary ventilation strategies since these studies were designed prevent any firm conclusions being drawn. It is safe to ventilate using high frequency oscillation with no evidence to suggest increased adverse effects compared to conventional ventilation. The use of high frequency oscillation in patients other than those with respiratory disease of 5

prematurity or ARDS is anecdotal and there is no evidence to support its use in these circumstances. Recommendations: There is insufficient evidence to recommend the use of high frequency oscillatory ventilation over conventional ventilation in the management of patients with ARDS. Results of its use so far are encouraging and further studies are required to determine its role in these patients compared to currently accepted optimal conventional ventilation strategies. 6

Introduction and Background In 1915 it was speculated that there may easily be a gaseous exchange sufficient to support life even when the tidal volume is considerably less than the dead space 1 and by 1954 it had been observed that during small inspirations, inspired air penetrates the dead space into alveoli without washing all the dead space gas into the alveoli. 2 These observations led to the realisation that ventilation using sub-dead space tidal volumes was possible and, in association with the high frequencies required, may be of benefit in certain circumstances. High-frequency ventilation is defined as ventilation at a frequency greater than four times normal respiratory rate. 3 Several categories exist including high-frequency positive pressure ventilation (HFPPV), high-frequency percussive ventilation (HFPV), high-frequency jet ventilation (HFJV), high-frequency flow interruption (HFFI) and high-frequency oscillatory ventilation (HFOV). They are differentiated by the equipment used to facilitate high-frequency ventilation and by the ventilatory frequencies attainable with each device. They have all been used as alternatives to conventional ventilation during surgery 4,5 and in respiratory failure of various aetiologies. 6,7,8,9 While the principles supporting the use of each are similar, only issues surrounding high-frequency oscillatory ventilation will be addressed. Lunkenheimer first described HFOV as we know it today in 1972 10 and by 1981 it had been proposed as an alternative ventilatory strategy for neonates with respiratory distress syndrome. 11 Since then its use in neonatal intensive care has become widespread and interest has developed in its role in older children and adults with respiratory failure, particularly acute respiratory distress syndrome (ARDS). Conventional ventilation (CV) has been shown to potentially exacerbate existing lung injury in patients with ARDS due to both its effect on inflammatory mediators 12 and direct barotrauma. 13 Strategies to reduce the risk of ventilator-induced barotrauma while maintaining acceptable gas exchange include the limitation of alveolar overdistension and prevention of alveolar collapse and re-expansion. 14 This may be achieved by limiting inspiratory plateau pressures (<30cm H 2 O) after a period of initial lung recruitment, followed by the maintenance of relatively high levels of 7

positive end-expiratory pressure (PEEP) and the use of low tidal volumes (6ml/kg). 15 This protective ventilatory strategy has been shown to reduce mortality in patients with ARDS. 16,17 High-frequency oscillation has been proposed as an alternative ventilatory strategy in ARDS 18. Its characteristics of very low tidal volume ventilation and the ability to achieve a high mean airway pressure while maintaining an acceptable plateau pressure combine to offer theoretical advantages of both a reduction in conventional ventilation-associated barotrauma and improved oxygenation. 8

Objectives This dissertation is a review of the current literature with respect to the use of HFOV in patients with ARDS. The mechanisms of gas exchange in oscillated patients are presented, as well as the theoretical basis for the use of this mode of ventilation in ARDS. A brief summary of the evidence for HFOV in the neonatal literature is presented to demonstrate the evolution of HFOV in clinical practice. As part of this review, the following questions are addressed: Does high frequency oscillation result in improved oxygenation in patients with ARDS? Do the theoretical benefits of high frequency oscillation translate into improved outcome in patients with ARDS? Is it safe to ventilate patients using high frequency oscillation? Are there any roles for high frequency oscillation other than in patients with respiratory disease of prematurity or isolated ARDS? Answers to these questions are discussed in detail and recommendations for the role of high-frequency oscillatory ventilation are made. 9

Methodology References were primarily obtained by conducting a MEDLINE search using the PubMed database between 1966 and October 2004. Initial searching used the search term: (High Frequency Ventilation (MeSH)). References were further sought and classified using the subset search terms: (High Frequency Oscillation) OR (HFO). Specific references relating to high frequency oscillation in ARDS were obtained by using the following search terms: (High Frequency Ventilation (MeSH)) AND (Adult Respiratory Distress Syndrome (MeSH)). All references obtained were cross-checked with results obtained by conducting the same search using the DialogStar database. In addition the reference lists of all relevant papers and reviews were hand-searched for any additional papers. Following a review of the literature the evidence was graded according to criteria used in 1997 at a consensus conference on the use of pulmonary artery flotation catheters. 19 These criteria (detailed in Appendix, Table 1) are modifications of levels of evidence first described by Sackett in 1989 20 and later expanded in 1995. 21 Although further modifications in 1998 22 reduced the complexity of the grading system and provided additional information with respect to risk/benefit analysis, the selected criteria were deemed most appropriate to answer the questions outlined above. All studies evaluating HFOV in patients with ARDS were considered for evaluation, including studies from the non-neonatal paediatric population. 10

Mechanisms of High-Frequency Oscillation The ventilator used in most trials of high-frequency oscillation in ARDS is the Sensormedics 3100B (Sensormedics, Yorba Linda, CA, USA). 23 This is a modification of the 3100A used in neonates and small children and is capable of the higher flow rates and greater amplitudes of oscillation required in adults. 7 It consists of an electromagnetically driven diaphragm that oscillates around a mean pressure determined by the bias gas flow settings. Oscillatory frequencies can be as high as 40Hz and these ventilators differ from other forms of high-frequency ventilators in that both inspiration and expiration are active processes. 3 When transferring a patient from conventional ventilation to HFOV the following variables are set by the clinician: mean airway pressure (MAP), usually set to 5cm H 2 O above that on conventional ventilation; amplitude of oscillation ( P), titrated to achieve visible chest vibration; I:E ratio, usually 1:2 and frequency of oscillation, usually 5-10 Hz in adults and 10-15 Hz in neonates. Gas transport in conventional ventilation is primarily a result of the interaction between bulk convection (predominantly in the upper airways) and molecular diffusion (predominantly in the lower airways). 24 Although the precise mechanisms for gas transport in HFOV are not completely understood, it is believed that there are five components involved. Bulk convection can still contribute significantly to alveolar ventilation (particularly in more proximal alveoli) as long as tidal volumes are at least 80ml. 25 As in conventional ventilation, diffusion due to thermal oscillation of gas molecules is likely to be the dominant form of gas transport approaching the alveolocapillary membrane. 24 Three other mechanisms are thought to play a role in gas transport, the most important of which is gas mixing by high-frequency pendelluft. 24 The concept that different lung units have different time constants which are each products of local resistance and compliance has been described. 26 This results in fast lung units (with small time constants) emptying into slower lung units (with greater time constants) at the end of rapid inspiration and vice versa at the end of rapid expiration. This results in a sloshing movement of gas between neighbouring units during cyclical ventilation. This is pendelluft and is greatly enhanced in high-frequency ventilation. Fourthly the 11

velocity profiles of gas in the proximal airways differ in inspiration and expiration. 27 In laminar flow, during inspiration, gases in the centre of the airway travel forward faster than gases near the bronchial wall, resulting in a parabolic gas front. During expiration, however, the gases at both points travel backwards at similar velocities, resulting in net distal propagation of gases near the centre of each airway branch, and movement towards the mouth of gases nearer the bronchial walls. This process is otherwise known as convective dispersion and relies on laminar flow. 24 Finally, a process known as Taylor dispersion may be important in CO 2 elimination from the proximal airways. When convective flow is superimposed on molecular diffusion, turbulent eddies result with mixing of fresh gas and alveolar gas. 3 This is most prominent in proximal airways where turbulent flow is most likely to be present. The relative contributions of convective and Taylor dispersion depend on the point at which transition from laminar to turbulent flow occurs in the airways. This is very difficult to determine, although the transition during high-frequency ventilation takes place at a higher Reynolds number, 24 suggesting a greater contribution of convective dispersion. The linearity of the CO 2 dissociation curve means that the clearance of CO 2 in highfrequency oscillation remains, as in conventional ventilation, a product of both tidal volume (V t ) and frequency (f), with V t being the greater contributor (Vco 2 = V 2 t.f) 28,29 When adjusting the oscillator, CO 2 reduction is best achieved by increasing the P (amplitude of oscillation) and, paradoxically, reducing the frequency. Oxygenation in high-frequency oscillation has been shown in rabbits to be directly related to lung volume. 30 Optimal lung volume is achieved by setting a higher mean airway pressure than in conventional ventilation, thereby facilitating alveolar recruitment and preventing derecruitment. 29 This is possible by eliminating the need for the additional volume excursions which are demanded by traditional ventilation for the elimination of CO 2. High frequency oscillation induces a reflex inhibition of spontaneous breathing which leads to apnoea at normocarbia in both animals 31 and humans. 32 12

Neonatal Data Much of the early animal work exploring the use of HFOV was in the context of its use in respiratory disease of prematurity. While some inconsistencies exist in the early trials, high-frequency oscillation has been shown to both enhance CO 2 clearance 33 and improve oxygenation in this setting. 34 Up to one third of neonates ventilated for respiratory disease of prematurity go on to develop bronchopulmonary dysplasia (BPD) due to a combination of the underlying hyaline membrane disease as well as the ventilation and possibly oxygen they receive. 35 Most trials evaluating the role of HFOV have focussed on its role in reducing this incidence. Of the nine studies comparing the incidence of BPD in premature neonates with hyaline membrane disease treated with HFOV and conventional ventilation, three (34.2% of all patients 36 ) demonstrated a significant reduction in BPD in the HFOV groups. 37,38,39 No study showed an increase in BPD in the HFOV groups. It is interesting to note that in all three of the studies demonstrating benefit, the ventilators used in the HFOV groups were exclusively Sensormedics 3100A ventilators. In all but one 40 of the other six studies either different ventilators or a combination of ventilators were used. 41,42,43,44,45 While some high-frequency oscillatory ventilators have been validated against each other 46, others have different characteristics and their clinical effects may vary. 47 In the HIFI study an increased incidence of significant intraventricular haemorrhages was noted in the oscillation group. 48 Although this increase was again seen in the HIFO study 49, a later meta-analysis showed that if the results of the HIFI study 48 are excluded, high-frequency oscillation is not associated with increased rates of intraventricular haemorrhage. 50 This was further substantiated in two recent large randomized controlled trials comparing high-frequency oscillation and conventional ventilation. 38,41 Despite theoretical benefits of high-frequency oscillation, no study has demonstrated any mortality benefit when compared to conventional ventilation. Effect on 13

oxygenation is difficult to establish as no study has compared changes in oxygen requirement in the hours or days following enrolment. Oxygen requirement after 28-30 days was compared in five studies with no difference demonstrated between patients who had received HFOV and those who had received conventional ventilation. 42,51,40,44,45 The majority of patients in the United Kingdom ventilated using high-frequency oscillation are neonates with respiratory disease of prematurity. The evidence to support its use is inconsistent although early studies suggesting worse neurological outcomes have not been substantiated in later studies. 14

Effect of High-Frequency Oscillation on Oxygenation Does high frequency oscillation result in improved oxygenation in patients with ARDS? Answer: Yes Grade: C The effect of HFOV on oxygenation has been studied in numerous animal models with inconsistent results. 29 Later studies in animals with induced lung injury have demonstrated clear improvements in oxygenation when compared to CV strategies, but only when the HFOV is preceded by a period of lung inflation to increase lung volume. 52,53 This recruitment of atelectatic lung units and maintenance of open alveoli has also been demonstrated to minimise lung injury during HFOV, 54,55 while low-pressure ventilation (below the lower inflection point on an inspiratory pressurevolume curve) augments lung injury in rats. 56 CV has also been shown to increase the levels of both pulmonary and systemic inflammatory mediators 57,58,12 which have been implicated in the pathogenesis of ventilator-induced lung injury. 59 HFOV results in lower levels of these inflammatory mediators and may therefore minimise further lung injury in ARDS. 58,60 Reducing ventilator-induced lung injury and maintaining large lung volumes at higher mean airway pressures than CV combine to offer theoretical benefit of HFOV in improving oxygenation in ARDS. 30 Oxygenation index (OI = mpaw x FiO 2 x 100/PaO 2 (mmhg)) is a marker of oxygenation relative to lung compliance and has traditionally been used in the neonatal and paediatric literature for this purpose. It was introduced into the adult literature by Fort et al. in 1997 because it was seen as a better marker of severity of ARDS than pure oxygenation indices alone. 61 Its use has become widespread in studies of HFOV in ARDS. Hypoxaemia, reflected as PaO 2 /FiO 2 200mmHg forms part of the definition of ARDS. 62 PaO 2 /FiO 2 (or PaO 2 /PAO 63 2 ) and OI were used in all studies of HFOV to assess its effect on oxygenation. With the exception of two retrospective reviews, 64,65 protocols were in place for managing variations in gas exchange. Once established on HFOV, mpaw and FiO 2 were titrated to oxygen saturations of 88% 93% and P and 15

frequency titrated to acceptable PaCO 2 (see Appendix, Tables 2, 3 & 4). If acceptable oxygenation was not achieved on HFOV, patients were converted back to CV. Arnold et al. 63 and Derdak et al. 66 both permitted crossover from HFOV to CV and from CV to HFOV if ventilatory failure limits were reached. It is important to note that patients studied by Arnold et al. are from a paediatric intensive care population. Arnold et al. 63 demonstrated no difference in PaO 2 /PAO 2 over 72 hours between HFOV and CV groups. Although they report a significant association between decreasing OI and time in the HFOV group, not seen in the CV group, OI at 72 hours was no different between the groups. 19 of 29 patients initially randomized to receive CV were crossed over to HFOV and 12, of the 16 for whom data is available, demonstrated 20% improvement in PaO 2 /PAO 2 compared to 1 of the 11 patients converted from HFOV to CV (p=0.08). These differences were lost by 72 hours. Of the 19 converted from CV to HFOV, 7 were for intractable hypoxaemia, while 2 of the 11 converted from HFOV to CV were for hypoxaemia. There was no difference between these proportions. Derdak et al. 66 demonstrated similar results with a significantly higher PaO 2 /FiO 2 in the HFOV group at both 8 hours and 16 hours following commencement of each ventilatory strategy (178 versus 131 mmhg at 8 hours and 205 versus 143 mmhg at 16 hours). These differences were lost by 24 hours and no further differences in PaO 2 /FiO 2 were demonstrated after this point. OI decreased in both groups over 72 hours but with no difference between the HFOV and CV groups. Crossover to the alternate ventilator strategy was permitted at the physicians discretion when adjunctive therapies were felt to be potentially life threatening. 4 of the 75 patients randomized to HFOV received adjunctive CV, while 9 of the 73 patients randomized to CV received adjunctive HFOV. Details of how these adjunctive therapies affected oxygenation were not recorded. All three prospective studies demonstrate a significant increase in PaO 2 /FiO 2 over time following institution of HFOV compared to levels immediately prior to its commencement, and a corresponding reduction in OI over the same period. 67,61,7 The improvements are early except in the pilot study by Fort et al. 61 where the reduction in OI only reaches significance at 12 hours. The improvements were consistently 16

sustained for the study period. Significant improvements in PaO 2 /FiO 2 and OI are mirrored in the three retrospective reviews, although the reduction in OI demonstrated by Anderson et al. 64 fails to achieve significance by 72 hours. Assessing the impact of HFOV on oxygenation in prospective studies and retrospective reviews is difficult due to the absence of controls. It is impossible to be certain that the improvements demonstrated in both PaO 2 /FiO 2 and OI would not have occurred had the patients remained on CV. The effect of the natural progression of each patient s ARDS cannot be quantified, meaning that although changes in oxygenation appear impressive following institution on HFOV, a causal relationship cannot be demonstrated. The early improvements in oxygenation demonstrated by Arnold et al. and Derdak et al. in their randomized controlled trials are more significant. It is important to note that, as part of the protocols in both studies, the mpaw was significantly higher in the HFOV groups at all intervals when compared to the CV groups. It has been demonstrated in animal models of ARDS, as well as human subjects, that during CV mean airway pressure is an important determinant of oxygenation. 68,69 Would this early improvement in oxygenation relative to the CV groups have occurred if the mpaw was the same in each group? In a study of newborn piglets, HFOV and CV using a lung-protective strategy, with matching mpaw, resulted in similar oxygenation. 70 In pigs with oleic-acid induced lung injury however, a significantly higher mpaw was required in the HFOV group to return oxygenation to pre-injury levels compared to those receiving CV. 71 A weakness of this study is the absence of lung recruitment prior to commencing HFOV and this is likely to have resulted in the higher pressures required to achieve matching oxygenation with the CV group. The trials demonstrate early improvements in oxygenation using HFOV compared to CV strategies that are no longer regarded as optimal in the ventilatory management of ARDS. These improvements are not sustained beyond 24 hours following initiation of HFOV. There is no data comparing HFOV to CV using the current open-lung approach, which limits the interpretation of the above studies. Because of this the evidence for improved oxygenation using HFOV from both randomized controlled trials achieves Level II, and from the prospective and retrospective studies, Level V. 17

Effect of High-Frequency Oscillation on Mortality Do the theoretical benefits of high frequency oscillation translate into improved outcome in patients with ARDS? Answer: No Grade: C Mortality in patients with ARDS has changed over the past decade. A review of 101 papers including 3264 patients with ARDS demonstrated constant mortality rates between 1967 and 1994 in the region of 50%. 72 A change in mortality was noted in the mid-1990s with rates falling to about 40%, 73,74 although these patients were not matched for severity of illness. The quoted mortality for ARDS in recent years is between 30% and 40%. 75 This may have been in part due to changes in ventilatory strategies in ARDS including the use of small tidal volumes. 16,17 This is the current standard against which outcomes from the use of HFOV in ARDS must be judged. Both Arnold et al. 63 and Derdak et al. 66 reported 30-day mortality as a primary outcome measure, while only Derdak et al. reported 6-month mortality. Based on initial randomization, Arnold et al. report 30-day mortality of 41% in the CV group and 34% in the HFOV group. This difference is not significant (RR 0.83, 95% CI 0.43 1.62). 66% of the CV group were crossed over to HFOV due to meeting failure criteria and 38% of the HFOV group were crossed over to CV, making attribution of any trend in improved mortality to HFOV difficult. When comparing survivors and non-survivors in each group, age was identified as an independent pre-treatment predictor of survival. This suggests that mortality data from this study cannot be extrapolated to the adult population. Derdak et al. demonstrated 30-day mortality rates of 37% in the HFOV group and 52% in the CV group based on initial randomization, with fewer crossovers than in the Arnold study. This difference failed to achieve significance (RR 0.72, 95% CI 0.50 to 1.03). 6-month mortality was 47% in the HFOV group versus 59% in the CV group (RR 0.79, 95% CI 0.58 1.08). Patients were matched for pre-intervention APACHE II scores. Although mortality in the study group is in keeping with 18

currently quoted figures, the higher mortality in the control group may be partially explained by the fact that the study protocol was designed prior to evidence demonstrating the impact on mortality of lower tidal volume ventilation in ARDS. 16 In addition the tidal volumes used in the CV group were based on actual, not ideal body weight. This trend in mortality benefit may, therefore, be less than it appears. Fort et al. 61 report a 30-day mortality of 53% in the context of a mean pre-hfov APACHE II score of 23.3 which would in itself be consistent with a predicted mortality of 40 60%. 76 The changes in outcome from ARDS since the publication of their supporting paper must partially invalidate this comparison. Mehta et al. 7 report a 53% ICU mortality and a 66% hospital mortality in their cohort of 24 patients with a mean APACHE score of 21.5. These relatively high rates may be contributed to by the 100% mortality seen in the subgroup of 5 patients with severe burns. Cartotto et al. 77 in their review of 25 burn patients report only 32% hospital mortality and although pre-hfov APACHE II scores are not apparent, mean intensive care admission scores are 16 which may have impacted on their mortality. David et al. 67 report a 30-day mortality of 43% and hospital mortality of 52% although their cohort had a higher median pre-hfov APACHE II score of 28. The relatively low mortality reported by Andersen et al. 64 of 31% at 3 months is lower than might be expected for their mean APACHE II score of 26.6. Two of the prospective studies found the duration of CV prior to the institution of HFOV to be an independent predictor of survival, with non-survivors having mean pre-hfov ventilation days of 7.22 61 and 7.8 7 days compared to 2.5 and 1.6 days respectively. The mean pre-hfov ventilation period of 7.2 days reported by Andersen et al. compares favourably with these figures. While this illustrates the pitfalls of predicting mortality based on historical controls, the low mortality may be attributed to the small sample size not being representative of the severe ARDS intensive care population. Mehta et al., 65 in their review of HFOV in Toronto, found a relatively high 30-day mortality of 68.7% in 156 patients with mean pre-hfov APACHE II scores of 23.9. This may in part be attributable to the 31% of immunocompromised patients as well as burn patients whose severity of illness may have been underestimated by their APACHE II scores. Both studies by Arnold et al. and Derdak et al. are relatively small studies with moderate to high risk of false-positive or false-negative results. They do, however, 19

agree that there is no significant mortality benefit from the use of HFOV in ARDS and both studies achieve Level II evidence in this respect. All prospective and retrospective studies achieve Level V evidence with respect to mortality benefit. 20

Safety of High-Frequency Oscillation Is it safe to ventilate patients using high frequency oscillation? Answer: Grade: Yes C The effects of high-frequency oscillation on other organ systems have been examined in animals. Cardiac output is reduced in HFOV when compared to spontaneously breathing dogs, 78 but not when compared to conventional ventilation at the same mean airway pressure. 79 Reduction of cardiac output and the associated increase in pulmonary vascular resistance appear to be linearly related to mean airway pressure. 80 This may be clinically significant at the mean airway pressures required for acceptable gas exchange in ARDS. One study in newborn lambs has assessed the impact of HFOV on intracranial pressure with no difference found when compared to CV 81 but the interpretation of this is limited by the absence of data to compare intracranial pressure rises at mean airway pressures matched for equivalent gas exchange. Another study in preterm baboons has reported an increased incidence of fatty liver change in animals receiving HFOV. 82 This is speculated to be a result of hepatic and splenic hypoperfusion secondary to reduced cardiac output and increased hepatic vascular resistance due to increased airway pressures in HFOV. These findings have not been reported in humans. Assessing safety of HFOV in clinical trials has centred on the relative incidence of clinical events such as persistent hypotension, mucus plugging and air-leak or pneumothorax, as well as episodes of equipment malfunction. Andersen et al. 64 and David et al. 67 used safety of HFOV as a primary end-point, although both studies are small and neither is controlled. Three studies used safety as a secondary end-point; the two prospective randomized controlled trials 63,66 and a prospective study of 24 patients. 7 The remaining studies were again uncontrolled and recorded data on complications that occurred and compared these to complication rates in the literature. Although Anderson et al. 64 declare safety of HFOV as a primary outcome measure, it is only the haemodynamic variables of mean arterial pressure (MAP), heart rate (HR), central venous pressure (CVP), cardiac output (CO) and pulmonary artery occlusion 21

pressure (PAOP) that are analysed statistically. Values at 4, 12, 24 and 72 hours after commencement of HFOV show no difference in any variable compared to pre-hfov levels. Comparison is longitudinal and uses repeated measure analysis of variance (ANOVA) which assumes the data is parametric. The data is incomplete with none of the variables being recorded in all 16 patients, with CO and PAOP only recorded in three patients. One patient developed bilateral pneumothoraces after 5 days of HFOV and one patient developed bilateral pneumothoraces on CV prior to the commencement of HFOV. David et al. 67 recorded their data prospectively resulting in no lost data. Specific HFOV-related complications recorded as a primary outcome measure were: the occurrence of mucus obstruction, pulmonary air leak and arterial hypotension. One patient developed a unilateral pneumothorax during HFOV which, although requiring the insertion of an intercostal drain, did not impair oxygenation or ventilation. There were no cases of endotracheal tube obstruction secondary to mucus plugging or unresponsive arterial hypotension (defined as MAP less than 60mmHg for 2 hours). When comparing MAP at 1, 6, 12, 24 hours and at the conclusion of HFOV to pre- HFOV MAP levels, there is no difference demonstrated, although, unlike the previous study, this data is treated non-parametrically. There were no cases of tracheal injury and the four patients with intracranial pressure monitoring showed no deterioration during HFOV. Interpretation of the incidence of complications in patients receiving HFOV is difficult due to the low number of patients in these two studies. Arnold et al. 63 focus on the effect of HFOV on neurological outcome based on its potential effect on cerebral blood flow. 83 This concern may also have related to the suggestion that neonates with respiratory disease of prematurity had worse neurological outcomes after HFOV, 48,84 although the sample populations are different. They conclude that there was no evidence of clinically important neurologic injury attributable to the mode of ventilation in the high-frequency oscillatory ventilation group. 63 No details of the criteria used to draw this conclusion are provided, making interpretation of this statement difficult. Although both intractable shock and progression of air leak were criteria for failure of a particular ventilation strategy, no details are provided of the relative occurrences of these events. 22

The study by Derdak et al. 66 was powered to detect a 20% difference in the incidence of adverse clinical events such as new or progressive airleak, mucus plugging requiring ETT change or intractable hypotension. There was no difference in the incidence of any of these measures between the groups receiving HFOV and those receiving CV. Longitudinal haemodynamic responses were no different between the two groups over the 72-hour study period except for PAOP (pulmonary artery catheters were present in 56% of the HFOV group and 51% of the CV group) which was slightly higher in the HFOV group throughout the study period. CVP and PAOP were also higher at 2 hours compared to baseline in those receiving HFOV. All other studies report general haemodynamic stability on HFOV with no sustained changes in MAP although Fort et al. 61 report three patients who needed to be transferred back to CV due to hypotension. In two of these cases the hypotension was attributed to HFOV. Mean CVP and PAOP (only recorded in those patients with pulmonary artery catheters in situ) rose early after commencement of HFOV but tended to fall in tandem with mpaw after 12 hours. 7,61 CO reduced significantly after commencement of HFOV although always remained within the normal range and the reduction was not associated with any other adverse haemodynamic events. 7,65 A late reduction in heart rate (after 24 hours) was noted in two studies, 7,65 although the patients from the prospective study 7 were later included in the retrospective review. 65 A pneumothorax rate of 21.8% reported by Mehta et al. 65 is higher than quoted in other studies and the reasons for this are unclear. In addition 26% of patients in this study did not tolerate HFOV due to difficulties with oxygenation, ventilation or haemodynamics. Retrospective sub-group analysis of these patients to identify any predisposing factors such as hypovolaemia or cardiac dysfunction is not possible. Three patients reported by Cartotto et al. 77 developed hypercapnoea, two of whom required conversion to CV. Significant mucus plugging was only reported in one patient in these studies. 7 Equipment malfunction occurred in five patients and is unlikely to be recurrent. In three of these cases the cause of the malfunction has been addressed by the manufacturers. 7 The only study that allows conclusions to be drawn regarding the safety of HFOV is by Derdak et al. 66 It is randomized and powered specifically to detect a 20% 23

difference in adverse event rates and achieves Level II evidence. No difference is demonstrated in any of the identified adverse events although there is no evidence that HFOV is safer than CV in ARDS as no study has attempted to demonstrate this. Arnold et al. 63 do not provide sufficient data to draw safety conclusions. Evidence of safety provided by David et al. 67 achieves Level V, as despite the fact that safety was a primary outcome measure, there are no controls. Similarly, the other prospective studies and retrospective reviews achieve Level V evidence. 24

Other Roles for High-Frequency Oscillation Are there any uses of high frequency oscillation other than in patients with respiratory disease of prematurity or isolated ARDS? Answer: Uncertain Grade: E The use of HFOV in circumstances other than respiratory disease of prematurity or ARDS alone is minimal and evidence in the literature is limited to two case reports of its use in the management of significant air leak syndromes with ARDS. 85,86 Both patients experienced prolonged periods of ventilation with unresolving air leaks; one due to bilateral spontaneous pneumothoraces during conventional ventilation for ARDS 85 and the other due to a high-output bronchopleural fistula following thoracotomy and pleural decortication for empyema. 86 Both patients were initially failing CV and therefore transferred to HFOV. They received 16 and 28 days of HFOV respectively, following which they were transferred back to, and successfully weaned from CV. The dilemma for clinicians faced with a patient with significant air leak in the context of ARDS is how to allow low enough airway pressures to minimise the air leak while achieving high enough pressures to oxygenate the remaining lung. Previously proposed solutions such as differential lung ventilation are logistically complex and not without hazard. 87 High-frequency ventilation 88 and more specifically oscillation may offer an alternative strategy. Arnold et al. 63 closely monitored the progress of patients with air leak, while Derdak et al. 66 excluded patients with significant air leak (more than one chest tube per hemithorax with a persistent air leak of more than 120 hours) from their study, presumably because of concerns that the higher mean airway pressures in the HFOV groups had the potential to worsen air leaks. There are reasons to believe that air leaks may in fact be better treated with a high-frequency ventilatory strategy rather than a conventional strategy. While higher mean airway pressures may be used in HFOV, peak pressures are usually lower than those required for similar gas exchange in CV. In a lung injury model, peak pressures were demonstrated to be inversely 25

related to frequency of ventilation in rabbits with bilateral pneumothoraces. 89 The use of HFOV in a piglet model of lung injury with pneumothorax demonstrated a similar inverse relationship between frequency of oscillation and gas flow through the chest tube. 90 The authors postulate that the short inspiratory times and low tidal volumes prevent enlargement of the pleural defects, and may in fact promote their closure. The high frequencies used in this study may be translated into paediatric practice and HFOV has been used successfully in the management of post-pneumonic tension pneumatocoele in a 3-year-old. 91 In adults oscillatory frequencies tend to be nearer 5Hz than 10-15Hz used in the animal and paediatric reports, although the principles should still apply at the lower frequencies when compared to CV. This may explain the apparent benefit of HFOV in the cases reported. No published reports could be identified of the use of HFOV in circumstances other than those already described. The reports of HFOV in the management of respiratory failure complicated by air leak syndromes are anecdotal and although encouraging, with the absence of any controls, the evidence can only achieve Level V. 26

Discussion With only two randomized controlled trials comparing HFOV to conventional ventilatory strategies in ARDS it is difficult to draw firm conclusions regarding its role. The prospective studies and retrospective case reviews are important in determining safety and demonstrating ventilatory possibilities but have no role in declaring whether or not HFOV is a better strategy than CV. The two randomized controlled trials are similar with respect to their trial design. Both studies used the same type of ventilator (Sensormedics 3100) and employed lung recruitment manoeuvres prior to the commencement of HFOV. While patients could be crossed over to the alternate strategy in both studies, this was based on protocols by Arnold et al. and at individual physicians discretion in Derdak s study. Randomisation was appropriate in both studies while blinding was possible in neither due to the physical differences between the respective ventilators. In both studies the study and control groups were acceptably matched at randomization with no differences in demographics or cardio-respiratory status. Important differences in the studies prevent the results from being pooled for metaanalysis. The studies included mutually exclusive groups of patients (children under 35kg versus adults over 35kg). This would not necessarily affect the interpretation of the results except that only 55% of the patients in the paediatric study had ARDS by currently accepted criteria; the others had respiratory failure due to other aetiologies. These patients are not differentiated in the results. In addition, younger age was shown to be an independent predictor of improved outcome, making extrapolation of results into the adult population impossible. It is also not clear whether the natural history of paediatric ARDS resembles that of adults, making comparisons of outcome data potentially unreliable. The authors do point out that although infants were included in their study, no child with chronic lung disease following respiratory disease of prematurity could be included, thus minimising a significant source of bias. Final data analysis in the paediatric study included 58 patients while 70 had initially been randomized. This high dropout rate (17%) had the potential to significantly affect the outcome results although a subsequent meta-analysis has shown that even if data from all randomized patients had been analysed, the results would not have been affected. 92 The high crossover rate in the paediatric study makes it difficult to tease 27

out the impact of HFOV alone, although this group did have significantly better ranked outcomes (survival without severe lung disease, survival with severe lung disease and death) than those receiving CV alone. Similarly, the ranked outcomes were significantly better in those crossing over from CV to HFOV than those crossing over from HFOV to CV, and in those ending the study on HFOV compared to those ending on CV. Over the past decade the change in ventilatory strategies for severe ARDS has been associated with a reduction in mortality. This demands a re-evaluation of the potential benefit that HFOV can offer. The influence of ventilator-associated lung injury (VALI) on outcomes following severe ARDS has been a focus of attention since a lung-protective ventilatory strategy was shown to improve outcome. 16 The lower tidal volumes and high PEEP used in these strategies splint the lungs open thereby maintaining oxygenation while preventing the derecruitment of recruited alveoli and minimising the volutrauma and barotrauma associated with higher tidal volume strategies. This achieves much the same as HFOV sets out to achieve. In addition, the current lung-protective approach to ventilation has been shown to result in lower levels of cytokines and other inflammatory mediators such as IL-6. 12,16 This potential reduction in the biotrauma associated with CV has also been demonstrated in HFOV. 58,60 When considering this, it is important to question just how different HFOV and current low-tidal volume, high PEEP ventilatory strategies are. It has been demonstrated in sheep that at the frequencies used in adults receiving HFOV (4-5Hz), tidal volumes can be as high as 5.6ml/kg. 93 It could therefore be argued that HFOV is just an extreme version of the currently accepted gold-standard ventilation. HFOV and pressure controlled ventilation (PCV), both using recruitment manoeuvres and aimed to maintain lung volume with low tidal volumes in lung injury models have been shown to result in similar lung mechanics and gas exchange in rats 94 and rabbits. 95 A similar study in sheep found no difference in lung mechanics or gas exchange although the histologic data suggesting greater lung damage in the PCV group may be attributable to the higher tidal volumes (8.9ml/kg) in that group. 96 The only way to determine whether HFOV really does have a role in the ventilatory management of patients with ARDS is to conduct an appropriately powered prospective randomized controlled trial comparing HFOV and the currently used lung-protective strategy. 28

Conclusions The widespread use of high frequency oscillation in the treatment of respiratory disease of prematurity may be justified on the basis of efficacy as well as safety. Most studies confirm an early improvement in oxygenation although there is limited evidence of improved survival or reduced incidence of bronchopulmonary dysplasia. Of the published studies on the use of HFOV in the management of ARDS, only two have been conducted on a prospective randomized controlled basis, although there have been more prospective studies and retrospective case series. There is evidence of an early improvement in oxygenation in patients with ARDS treated with high frequency oscillatory ventilation compared to conventional ventilation, although these benefits are lost by 24 hours. There is a non-significant trend towards improved survival in patients with ARDS treated with high-frequency oscillation. The fact that the conventional ventilation strategies to which HFOV was compared are no longer considered optimal prevents any firm conclusions from being drawn. It is safe to ventilate using high frequency oscillation with no evidence to suggest increased adverse effects compared to conventional ventilation. The use of HFOV in patients other than those with respiratory disease of prematurity or ARDS is anecdotal and there is no evidence to support its use in these circumstances. 29

Recommendations There is insufficient evidence to recommend the use of high frequency oscillatory ventilation over conventional ventilation in the management of patients with ARDS. Results of its use so far are encouraging and further studies are required to determine its role in these patients compared to currently accepted optimal conventional ventilation strategies. Any future studies should also address the timing of initiation of HFOV. 30

Appendix Table 1 Grading of Responses to Questions A Supported by at least two level I investigations B Supported by only one level I investigation C Supported by level II investigations only D Supported by at least one level III investigation E Supported by level IV or level V evidence Levels of Evidence I Large, randomized trials with clear-cut results: low risk of falsepositive (α) error or false-negative (β) error II Small, randomized trials with uncertain results; moderate to high risk of false-positive (α) error and/or false-negative (β) error III Non-randomized, contemporaneous controls IV Non-randomized, historical controls and expert opinion V Case series, uncontrolled studies, and expert opinion Grading of responses to questions and levels of evidence 19 31