Taking slow vent weans by the collar

Our ICUs are crowded by patients that are slow to liberate from mechanical ventilation. In North America, such patients are often transferred to long term mechanical ventilation facilities (LTAC) – where they are weaned to liberation. There are many strategies for weaning tracheostomised  patients from mechanical ventilation – a progressive reduction in pressure support, intermittent tracheal mask (trach collar) trials, external CPAP etc.

Martin Tobin has, for many years, questioned the now conventional wisdom of progressive pressure support weans, and in this weeks JAMA his group have published a paper comparing Volume Assist Control Ventilation with intermittent tracheal mask trials to progressive pressure support weaning. The patients in the pressure support group (PSG) started, if they were able to tolerate it, pressure support ventilation at 14cmH2O. Tolerance to wean was checked every 6 hours, and 2cmH2O decrements of PS were applied up to a maximum of 6cmH2O per day. Once the pressure support reached 6cmH2O, a 5 day vent liberation process began. In the tracheal mask group (TMG), patients were put on tracheal mask  for up to 12 hours per day for 2 days (then back on assist control) and on day 3 started on the 5 day vent liberation programme.

The primary outcome was weaning duration, defined from the first day of randomization to the day the patient was successfully weaned. Weaning was considered successful when patients breathed without ventilator assistance for at least 5 days. If the patients had not liberated by day 45, this was considered “failure to wean.”

Three hundred and twelve patients were randomized, of which one third died while in the study (equal numbers each group). Among the entire group of randomized patients (n = 312), median weaning time was shorter in TMG versus PSG: 15 days (IQR, 8-25) vs 19 days (IQR, 12-31), P = .004.. Among patients who completed the study (n = 194), median weaning time was shorter in TMG vs PSG: 13 days (IQR, 8-30) vs 19 days (12-43), P = .006.

Weaning time had no effect on survival at 6 and 12 months.

Impression: this article suggests that weaning patients using decrements of pressure support is not aggressive enough and that weaning in more likely to be successful with unassisted tracheal mask. Although, unsurprisingly, this had no effect on survival, a 6 day reduction in ventilator times translates into considerable resource savings – particularly with ICU/HDU beds are very scarce (as in our hospitals). Whether or not patients should be weaned from assist-control or high levels of pressure support appear moot: I doubt that it makes a difference. I continue to see any value for using external CPAP through a t-piece as a vent liberation process: there is not a shred of supportive evidence.

High Frequency Oscillation – Shaken by Bad News!

oscillatorSummary: Two Papers Published Online in the NEJM, OSCILLATE and OSCAR, have failed to demonstrate that HFOV benefits patients with ARDS. In the OSCILLATE study there was an 11% increase in 28 day mortality (NNH 9). HFOV should not be used in routine management of patients with ARDS.

For the past two decades many intensivists have used an “open lung” approach to managing hypoxaemia in ARDS. This approach involves using high levels of PEEP, inverse ratio ventilation (IRV), airway pressure release ventilation (APRV) or high frequency oscillation ventilation (HFOV) – to keep severely injured lungs in a state of inflation during most of the respiratory cycle. “Open Lung” can be achieved by using high PEEP and small tidal volumes, or moderate levels of PEEP and long inspiratory times. In either case mean airway pressure increases. The most sophisticated method of using PEEP involves the construction of pressure volume curves and setting the PEEP above the lower inflection point. To date, the high PEEP approach has neither been proven to be better or worse than the incremental PEEP approach based on fiO2. However a 2010 meta-analysis has suggested that high PEEP may be beneficial in ARDS. With long inspiratory times oxygenation occurs principally during inspiration rather than expiration; functionally ventilation occurs on the expiratory limb of the volume-pressure curve.
There is good physiologic reasoning behind the open lung approach. Severe ARDs is characterized by massive atelectasis, increased lung water and pleural effusions; functional residual capacity is obliterated. As most gas exchange occurs in expiration, and end expiratory lung volumes are lost, open lung approaches use the inspiratory reserve volume. Further, it is widely believed that phasic opening and closing of injured lung units and uninjured adjacent units results in ventilator induced lung injury (atelectrauma). Open lung approaches prevent atelectrauma. The majority of ICUs utilize conventional ventilator strategies such as pressure support, pressure control (or bilevel) or volume assist-control in early ARDs. However, in patients with severe and refractory hypoxic respiratory failure, units differ between using high PEEP and so called “advanced” modes to maintain oxygenation. Despite the conceptual attractiveness of IRV, APRV and HFOV there has always been an evidence gap in support of these modes: nobody knows how much the lung stretches. There is no easy way to measure end inspiratory lung volumes and hence to evaluate the risk of volutrauma. Proponents of “advanced” modes will argue that absence of evidence does not mean absence of efficacy: clearly open lung modes of ventilation improve oxygenation – so this must be good. However, we have decades of data suggesting that prone positioning improves oxygenation, but not outcomes.

After a long wait, we have now seen the publication of two studies on HFOV – the multi-national OSCILLATE study, and UK based OSCAR trial. The results are not good.


Background and Methods: This was a multi center trial of 38 hospitals in Canada, the United States, Saudi Arabia, Chile, and India from July 2009 thru August 2012. Patients were eligible for inclusion if they had had an onset of pulmonary symptoms within the previous 2 weeks (i.e. this was an acute illness), had been intubated, had hypoxemia defined as a PaO2/FiO2 (PF) ratio of ≤200, with an FiO2 of ≥0.5), and had bilateral infiltrates on CXR. These criteria must have been met within the previous 72 hours. The authors were careful – patients who met criteria were put on pressure control ventilation, TV 600ml, FiO2 0.6 and PEEP 10cmH2O. If, after 30 minutes, the PF ratio was still below 200, the patients were enrolled in the trial. Patients were randomized to HFOV or conventional ventilation (CV). The patients were remarkably well balanced at baseline – and very sick – with mean Apache II scores in both groups of 29.

Technique: a recruitment maneuver was performed initially (40cmH2O for 40 seconds) then HFOV was commenced at a mean airway pressure of 30 cmH2O, adjusted to keep the PaO2 between 55 to 80 mm Hg (7.3 to 10.5 kPa). The frequency was adjusted to keep the pH above 7.2. Once mean airway pressure (mPaw) dipped below 24 cmH2O conversion to CV was considered; if mPaw went below 20cmH2O – conversion was mandatory. The CV group received a recruitment maneuver, then pressure controlled ventilation with TV 6ml/Kg and PEEP adjusted according to protocol. Initial PEEP was 20cmH2O. Patients could also receive volume assist control or pressure support ventilation. For patients with good compliance and gas exchange (presumably during the weaning phase), the investigators did not set limits for tidal volumes. A weaning protocol governed both limbs.

Results: the primary outcome was in-hospital mortality. The study was stopped early because of increased risk in the study group. At the time of termination, 571 patients had been enrolled, of whom 548 had undergone randomization: 275 to the HFOV group and 273 to the control-ventilation group. A total of 129 patients (47%) in the HFOV group, as compared with 96 patients (35%) in the control group, died in the hospital (relative risk of death with HFOV, 1.33; 95% confidence interval, 1.09 to 1.64; P=0.005) – an absolute risk INCREASE of 12%! This was independent of any other risk factor other than HFOV and was consistent at 28 days. Indeed 28 day mortality was 40% in HFOV group (which is not particularly bad) versus 29% (which is really excellent) in the conventional ventilation group – absolute risk increase 11% – number needed to injure 9. The duration of hospital stay for survivors was, on average, 5 days fewer in the CV group (although statistical significance is not reported) 25 versus 30 days.

Did HFOV have a physiological effect? Yes – patients almost immediately needed more vasopressors, more sedation and more neuromuscular blockade. There were fewer cases of refractory hypoxaemia but this had no statistically significant impact on outcomes.

My Impression

So, what to make of this study? Firstly, let us be clear – this was an excellent study (that we were all awareards of) performed by experts in the field of mechanical ventilation, who have a long track record of publication and interest in HFOV. They were not expecting the published results. This is a significant setback to our understanding of ARDS and interventions to rescue patients from severe hypoxaemia. Among my critical care colleagues, there are those of us who are proponents of ARPV and those of us who use HFOV. The published results ask as many questions about APRV as they do about oscillation. Can the worse outcomes be explained by increased doses of vasopressors and more sedation? No, I don’t believe it to be so: it is inconceivable that increased drug utilization could explain an 11% risk increase at 28 days. It is more likely that HFOV exacerbates ventilator induced lung injury (VILI).


The OSCAR study took place in the UK and included nearly 800 patients that had PF ratio <200 and wereOscar randomized to HFOV or CV. There was no significant between-group difference in the primary outcome, death at 28 days, which occurred in 166 of 398 patients (41.7%) in the HFOV group and 163 of 397 patients (41.1%) in the conventional-ventilation group (P=0.85). In other words – HFOV did not benefit these patients. Patients enrolled in this study had an average Apache II of 21 – lower than Oscillate – but had equally bad gas exchange when recruited. Although there was an increase in the use of muscle relaxants, there was no evidence of increased vasopressor requirements.

Of interest, patients in OSCAR had similar (although worse) 28 day mortality rates to those in the HFOV arm of the OSCILLATE study. It is worth looking at the LOV and EXPRESS trials for comparison. The EXPRESS trial had slightly older patients (60y both groups) with better gas exchange – PF ratio 143 and 144. Twenty eight day mortality was 31.2% and 27.8% (lower in high PEEP approach but not significant). The LOV study had patients in the same age range as both of these studies, but the PF ratios were significantly higher (144 in both groups); Apache II scores were 24 and 25 (patients were sicker than OSCAR). Mortality rates for all cause in hospital were 36.4% and 40.4% (again not significant). The lower end outcome is similar to the conventional group (and indeed had similar vent strategy to that group) in OSCILLATE. This suggests that the better outcomes in the conventional group in OSCILLATE accurately reflect true outcomes, and that the 28 day mortality in the LOV trial over-estimates outcomes at hospital discharge. But, can the difference in gas exchange really explain why the outcomes in OSCAR, in the conventional group at least, were 10% worse than EXPRESS? I don’t know – but comparing OSCILLATE to OSCAR – a 29% 28 day mortality (similar to LOV) is substantially lower than the 41% in the UK trial. These data suggest that the conventional group in that study did worse than expected. And, if this were the case, then HFOV was harmful in the OSCAR study too: the mortality rate for HFOV was essentially the same in both studies. Ok that might be conjecture – but take my point: a 41% mortality rate for ARDS is very high in the setting of a randomized controlled trial. For example, in the original ARMA trial the mortality rate in the control group (high TV) was 39.8% versus 31%; PF ratios 138 and 134. In the ALVEOLI trial the mortality rates were 24.9% (high PEEP) and 27.5%; PF ratios were 165 and 154. I have long argued that many of the patients in the NIH studies probably didn’t have ARDS – but the outcomes in OSCAR are the worst published in a multi center trial since the control group in Aries. Whatever way you look at it – this is all BAD NEWS for HFOV.

HFOV – is it safe?

One striking thing about these studies is how much the conflict with a previous meta-analysis of previously published papers. In that paper eight randomized controlled trials (n=419 patients) were included; the majority of patients had ARDS. Patients on HFOV had better oxygenation and reduced mortality (risk ratio 0.77, 95% confidence interval 0.61 to 0.98, P=0.03; six trials, 365 patients, 160 deaths). These results conflict directly with the outcomes of OSCILLATE and OSCAR: it is scary how meta-analysis of small studies (publication bias) can be misleading.

So, is it time to wrap up your oscillator and consign it to the ventilator graveyard? For ARDS – for routine practice – I think so.

1-forward-2-backFor rescue therapy or as a bridge to ECMO – HFOV may still have a role. Oscillation remains a useful therapy for broncho-pleural fistula and perhaps for severe lung contusion. But in other settings – I believe – use with extreme care. Is it time for a complete re-evaluation of open lung approaches in ARDS? Perhaps so – certainly before we put patients on extreme IRV modes (such as APRV) or HFO we should ask – “have I maxed out my options with conventional ventilation?”
One Step Forward – Two Steps Back.

Beachchair and Blood Pressure

BeachChairMAPIt is now more than 5 years since the Anesthesia Patient Safety Foundation highlighted the risk of central nervous system injury following anesthesia for shoulder surgery (SS) in the Beachchair position (BCP) (click here). Although we can never be certain, it appears likely that such injuries – principally devastating stroke, results from hypoperfusion and watershed ischaemia. Most anesthetists agree that, in the seated position, the mean arterial pressure as measured by a brachial cuff, under-estimates the pressure at the circle of Willis by 15mmHg or more. The safe lower range of MAP with regard to cerebral auto regulation remains unclear, but it is certainly not below 50mmHg (click here). Regardless, if we are to believe in the “Waterfall” effect of blood pressure, then a MAP of 65mHg from an arm BP cuff is marginal. If an arterial line is place – a procedure rarely performed in orthopedic shoulder patients – the transducer should be sited at the external auditory meatus. Although there are proponents for cerebral oximetry as a monitor in this setting – I do not believe that data are sufficient to demonstrate sensitivity and specificity of this device (here). The majority of experts in the field agree that the best option is to keep BP as close to baseline as possible if general anesthesia is administered for SS in BCP (here here and here). Clearly – “induced hypotension” is a bad idea in this setting. Also, the placement of the BP cuff on the leg or ankle (so that it does not cut of the iv line periodically) would appear most unwise. In one paper from Korea (click here), ankle blood pressure was substantially higher (up to 30mmHg for systolic pressure, 20mmHg mean pressure) compared with brachial pressure in the Beachchair position. In other words – if you are using an ankle cuff, the measured blood pressure (MAP) may be 35mmHg or more higher than the pressure seen at the circle of Willis.

My own practice is to use a brachial cuff and administer a phenylephrine infusion to keep the MAP above 75mmHg (or at the normal awake range for the patient). This appears to be the best approach based on my reading of the literature and available technologies. I would urge orthopedic anesthetists in the West to read the articles referenced here and come up with their own protocols. Comments?

Transfusion Strategy – Think Restrictive

nejm_transf_1.2013A half generation ago, the TRICC trial (here) suggested that routine blood transfusion in critically ill patients did not confer benefit if the haemoglobin level was above 7g/dl. This resulted in a evidence based paradigm for lower transfusion triggers. The problem was – how do you deal with the bleeding patients?
A recent study in NEJM (here) looked at liberal versus restrictive transfusion practices on patients admitted with gastrointestinal bleeding.

921 patients with severe acute upper gastrointestinal bleeding were included in the study: and 461 of them were randomly assigned to a restrictive strategy (transfusion when the hemoglobin level fell below 7 g per deciliter) and 460 to a liberal strategy (transfusion when the hemoglobin fell below 9 g per deciliter). Randomisation was stratified according to the presence or absence of hepatic cirrhosis.

Substantially more patients in the liberal strategy group received transfusion: 395 (85%)  versus 236  (49%) liberal versus restrictive (P<0.001). The conventional wisdom would hold that greater oxygen carrying capacity in the liberal group would result in better outcomes. The null hypothesis would be that there was no difference. However, the patients in the restrictive group had BETTER OUTCOMES. The hazard ratio for 6 week mortality was 0.55; 95% confidence interval [CI], 0.33 to 0.92; P=0.02) [with HR a number of <1 reflects benefit, >1 reflects injury]. The absolute risk reduction was 4% (5% restrictive, 9% liberal p = 0.02; NNT 25). In addition, patients in the liberal strategy group had a 6% absolute increase (number needed to injure 16; (P=0.01) ) in the risk of further bleeding.
There was an absolute risk reduction of adverse events of 8% in the restrictive group (NNT 12; p = 0.02). Restrictive transfusion also resulted in better survival in patients with peptic (gastric or duodenal) ulcers, and those with mild to moderate cirrhosis.

Why would bleeding patients do better if transfusions are witheld? There are many potential reasons: 1. The concept of damage control resuscitation: teleologically we have evolved to handle hypovolaemia and can survive considerable blood loss. Blood transfusion without source control may cause clots to destabilise and further bleeding to occur. 2. Blood is immunosuppressive: patients who are transfused are at elevated risk of infectious complications. 3. In this particular study patients in the liberal transfusion group had higher portal pressures and were more likely to rebleed (but so too were patients with peptic ulcers). 4. Transfusion may result in volume overload, abdominal compartment syndrome, myocardial ischaemia and transfusion related lung injury.

What are the implications of this study. Approach with caution! This study does not license clinicians to withhold blood from ex-sanguinating patients. Nor does it prove anything about transfusion in the setting of non gastrointestinal blood loss. However, it does provide us with further information about the safety and implications of blood transfusion in a specific setting. Allied with accumulating data detailing the hazards of colloid transfusion, adverse outcomes associated with crystalloid over-resuscitation, and the ongoing controversy regarding albumin – one has to wonder where we are with fluid resuscitation. Remembering that red cell transfusion is a key component of the Rivers’ surviving sepsis protocol, one wonders if this is the first real nail in the coffin for that approach.

Comments are welcome here.

Troponin Leak Postop – what does it mean?

vision studyTwenty years ago perioperative myocardial ischaemia was a relatively easy thing to diagnose – we checked ECG looking for ST segment and T wave changes, and looked for an MB-CK rise. Then troponin arrived, and suddenly the proportion of patients with perioperative ischaemia increased drastically. For many of us, the report of a “postoperative troponin leak” results in a shoulder shrug: we don’t know what it means, we don’t really know the long term implications.

Thankfully, a landmark study, VISION (click here), has provided us with quality epidemiologic data. This was a cohort study of 15000 patients >45 years that underwent non cardiac surgery and had troponin T (TnT) measured in the first 3 postoperative days. All patients had to have procedures that required overnight stay in hospital. The main outcome measure was 30 day mortality.

After 30 days 1.9% of patients had died. Patients were more likely to die if their peak TnT level was 0.02 ng/ml (versus reference range of <0.01 ng/ml). This occurred in 11.6% of patients. The greater the TnT level, the more likely the patient was to die. They were able to stratify risk depending on TnT levels. Patients with a peak TnT value of 0.01 ng/mL or less, 0.02, 0.03-0.29, and 0.30 or greater had 30-day mortality rates of 1.0%, 4.0%, 9.3%, and 16.9%, respectively (figure above).

Risk was expressed in terms of Hazard Ratio (HR): greater HR = more likely adverse outcome with 1 being equivalent to no additional risk, <1 lower risk, >1 higher risk. Peak TnT of 0.02 ng/mL (adjusted hazard ratio [aHR], 2.41; 95% CI, 1.33-3.77); 0.03 to 0.29 ng/mL (aHR, 5.00; 95% CI, 3.72-6.76); and 0.30 ng/mL or greater (aHR, 10.48; 95% CI, 6.25-16.62).

Who was at increased risk? The older the patient the higher the risk. Emergency surgery, general surgery, neurosurgery were associated with increased risk. Vascular surgery was not, although the presence of peripheral vascular disease, COPD, previous stroke, coronary arterial disease and cancer did predict adverse outcome. Diabetes, obesity, afib, OSA, hypertension, orthopaedic/thoracic urology surgery – did not predict adverse events.

Conclusions: these data demonstrate the efficacy of TnT measurement in determining perioperative prognosis. 1in 25 patients with a peak TnT measurement of 0.02ng/mL,1 in11patients with a peak TnT measurement of 0.03 to 0.29ng/mL, and 1 in 6 patients with a peak TnT measurement of at least 0.30ng/mL will die within 30 days of surgery. Two questions arise from this study: 1. should we be routinely measuring TnT postoperatively in surgical inpatients >45 years; 2. If the patient has a troponin leak – what should be do then: PCI, aspirin, clopidogrel, statins, betablockers, all of the above, none of them? Will routine measurement of TnT result in a dramatic increase in cardiology consultations with little evidence that there are interventions that may improve outcomes in this setting?

The Obesity Paradox – Weight there’s more!

chandler-weightThe media are constantly harping on about the obesity epidemic – “two in three of us are fat – this is going to lead to an explosion of obesity related morbidity – heart disease, cancer, cerebrovascular disease, crumbling joints etc.” It never seems to occur to the same talking heads on television and in print media that life expectancy continues to improve worldwide, without any great new medical advances over the past 2 decades (during which the obesity epidemic emerged). In perioperative medicine and critical care we have data that an “obesity paradox” exists – that individuals with a BMI between 25 and 35 (overweight and grade 1 obesity) have lower mortality rates in perioperative medicine (references 1. Here, 2. Here) and critical illness (references 1. Here; 2 Here; 3 Here; 4. Here). The reasoning why overweight and mild obesity (being chubby) may confer benefit is an increase in physiology reserve, delivered in part by an increase in lean body mass. As patients become heavier (body mass index; BMI>35, grade 2 and 3 obesity), they start manifesting mass related injury and metabolic disease. Metabolic syndrome unquestionably increases long term risk (here).
But what of the general population? Life insurance companies continue to penalize chubby folks based on actuarial figures from the 1960s – it that fair? It turns out that it is not. Hold your breath – a 2.88 million patient meta analysis performed by a group of US researchers and published in last week’s JAMA (here), has completely moved the goalposts for desirable BMI. In terms of Hazard Ratio (where HR of 1 = BMI 20-25), BMI of 25-30 had a HR for all cause mortality of 0.94 (lower; CI 0.91-0.96). BMI  30-35 had HR 0.95 (lower; but CI 0.88 – 1.01 NS). BMI 35-40 HR 1.29 (higher CI 1.18-1.41). In other words – overweight patients had LOWER all cause community mortality than BMI 20-25. Being obese up to BMI 25 did NOT increase mortality risk, and only individuals with BMI >35 had increased risk.

People – you need to face up to it: OVERWEIGHT IS THE NEW “NORMAL.” Perhaps “Chubby Chandler (above)” was the healthy one.

More comments to follow.