Effect of Postextubation High-Flow Nasal Cannula vs Noninvasive Ventilation on Reintubation and Postextubation Respiratory Failure in High-Risk Patients: A Randomized Clinical Trial, by Gonzalo Hernandez et al. 
Why does everyone seems so happy using high-flow nasal cannula? At least when googling it they do. How come? It’s like having two alien-air-blower-fingers in your nostrils. Internet can make everything beautiful, even when you are using a noninvasive ventilation mask and look like Bane, it’s just lovely. But that’s not the point, the thing is, something around 2 out of 10 patients we extubate will see our heads upside down holding a laryngoscope again, doesn’t matter how good your team is. Reintubation sucks! We increase the risk of procedural complications, mortality rates, ICU length of stay, hospital lengh of stay, risk of infections and costs. When I say “we”, it’s not actually “we”, well, sometimes might be, but, it’s the circumstances, the disease, the comorbidities… So, for a long time we are trying to find ways to prevent our patients from falling off the reintubation cliff. And that’s what this is all about!
Last year, our flamenco dancing friends from Spain published in JAMA a study evaluating if High-Flow Nasal Cannula (HFNC) is non-inferior than Noninvasive Ventilation (NIV) on preventing reintubation and postextubation respiratory failure in high-risk patients . We know that HFNC improves patients comfort and oxygenation along with other speculative benefits, also we do know the benefits of NIV. When comparing both methods I’m gonna stick with 1 advantage of HFNC: patient comfort. Although in the article the authors talk about lower costs, this is not true here in Brazil, at least for now.
How they did it
They screened adult patients with >12h of mechanical ventilation, excluding patients with tracheostomies, unplanned extubation or hypercapnia during the spontaneous breathing trial. Now, it might not be that easy to define what “high-risk” extubation is, the authors defined it as: >65 y/o, intubated due heart failure, COPD (mod-sev), APACHE II >12, BMI >30, airway potency problems, >1 weaning failure, ≥2 comorbidities, MV >7d. In fact, 50% of patients receiving mechanical ventilation >12h were considered high-risk in the present study! Patients were accessed daily for weaning using a protocol with too much specificity for my taste. Why? One might argue that hemoglobin >8g/dl, temperature <38°C, no need for sedatives (this is relative) and no vasoactive drugs (low dose of dopamine only) is too much. If patients met these criteria they underwent a spontaneous breathing trial (SBT) with either T-tube or pressure support (7cmH2O) for 30 to 120 minutes. I cannot imagine my patients with a T-tube snorkel for 2 hours, c’mon! 2 hours? We know for two decades that 30 minutes and 2 hours are equivalent! Focus! If the patients tolerated the SBT and had no other contraindication to extubation they were randomized.
Behind the scenes
The interventions were:
-HFNC group: Set at 10L/min and titrated as needed, FiO2 targeting a SpO2 >92% for 24h.
-NIV group: BiPAP. PEEP, inspiratory pressure and FiO2 aiming a respiratory rate of 25/min, SpO2 >92% and pH>7,35 for 24h.
Just a reminder: it’s impossible someone stays 24h on BiPAP! The primary outcomes were reintubation and postextubation respiratory failure, both <72h after the extubation. They had well-defined criteria for reintubation and the definition of postextubation respiratory failure included the usual criteria and two things I kinda disagree: agitation and decrease >1 point in GCS (both things can be due other process than respiratory failure). With baseline reintubation rate of 20-25%, 80% power and noninferiority margin of 10% (based on clinical relevance and previous data) a sample size of 300/arm were needed, already accounting for 15% patient loss. Of course the ICU team was not blinded, but the investigators were excluded from clinical decisions.
What they found
After randomization they had 314 patients in NIV group and 290 in HFNC group. The baseline characteristics are summarized below:
Patients were old, sick and ventilated for 4 days (median) before the extubation. Few differences between groups, NIV group had more patients with prolonged MV, intubated due heart failure and >65y/o; the HFNC had more surgical patients and more patients with “other respiratory disease” (88 vs 0). Hence, some unbalances that might influence the primary endpoint, but what strikes me are those other respiratory diseases (interstitial lung disease, pulmonary embolism, pulmonary hypertension, asthma and smoking habit), how can it be so unbalanced? Very strange my Dear Watson! Probably someone forgot to type the data for the NIV group. As we predicted, the NIV group had a median time under NIV of 14h; the HFNC group did the intervention for the whole 24h.
For the primary outcomes, HFNC was noninferior to NIV, within the non inferiority margin of 10%; Reintubation: 66 in HFNC vs 60 in NIV (risk difference of −3.7%, and marginal odds ratio similar to the odds ratio conditioned to covariables after the multivariable analysis); postextubation respiratory failure: 78 in HFNC vs 125 in NIV. I think this last difference might be explained, or not, by the secondary outcomes, when the authors subdivided the causes of postextubation respiratory failure. Patients in the NIV group had more respiratory acidosis, hypoxia and inability to clear secretions, by the way, the last one was not listed in the definition of postextubation respiratory failure. Touché. And of course there is no plausible biological explanation, at least that I’m aware of, why HFNC would increase the ability to clear secretions. The inability to clear secretions was the cause for intubation in 20 NIV-patients vs 13-HFNC patients. It just doesn’t seem right. Also there was a difference in adverse events, only the NIV group had them (non tolerance to NIV was the adverse effect).
Things get cloudy when you are not able to run a completely blinded study, and in this case, it was impossible. Here, ascertainment bias might have played a role. Another point, I do not agree with the 10% noninferiority margin (10% is A LOT) and you’ll read on the next paragraph what I think about NIV. The authors didn’t seem very confident in choosing it as active control, and one might even argue that NIV is worse or non-inferior than standard of care, for instance, just oxygen. But this is a sluggish terrain to walk through, basically due one thing called Biocreep. I’m gonna give you a pretty shitty example: Imagine you have 100 bucks, and within a noninferiority (NI) margin of 10% you are left with 95 bucks, and since it’s within the NI margin you assume that 95=100(noninferior). Now 95 bucks is your standard money, and again, you do a noninferiority trial and show that 95=88, all within your NI margin. Here lies the problem: 88 is not equal 100, even within a NI margin of 10%. You guys got me? Although I don’t take this as a huge issue in this trial, I thought worth mentioning it.
The choice of NIV as a active control group was smart, of course we do not have that herculean-data, but I do think NIV might have a role for preventing reintubation. But don’t be a smart ass, use it preemptively, choose your patient wisely and reassess the situation around the clock! Forget about postextubation respiratory failure, at least in this study, I won’t even waste our time. The real deal is the reintubation rate. If you’re the “I just read the abstract” kinda guy, case closed! But if you have at least 10 ICU-synapses (1 for getting coffee, 1 for central lines, 1 for vasoactive drugs, 1 for sedation, 1 for antibiotics, 1 for respecting the nurses, 1 for intubation, 1 for Lasix and 1 for palliative care), please, use the last one here and take these results with a grain of salt. I agree HFNC offers advantages compared to NIV, but I don’t think preventing reintubation is one of them.
1. Hernández G, Vaquero C, Colinas L, et al. Effect of Postextubation High-Flow Nasal Cannula vs Noninvasive Ventilation on Reintubation and Postextubation Respiratory Failure in High-Risk Patients: A Randomized Clinical Trial. JAMA. 2016;316(15):1565-1574.