Girardis M, Busani S, Damiani E, Donati A, Rinaldi L, Marudi A, et al. Effect of Conservative vs Conventional Oxygen Therapy on Mortality Among Patients in an Intensive Care Unit: The Oxygen-ICU Randomized Clinical Trial. JAMA. 2016 Oct 18;316(15):1583-1589.
What’s the question we’re looking at? Why do we care in CCM?
Does a liberal versus conservative oxygenation goal in ICU patients have an effect on mortality?
Historically, oxygen delivery was considered one of the main goals of critical care, so more was presumed to equal better.
Over time, data gradually accumulated showing harm from hyperoxia in various conditions (MI, stroke, COPD), leading some to lean towards a “Goldilocks” approach rather than “more is better.” This is perhaps largely due to free radicals generated by oxygen. Prolonged use of high FiO2 can easily be shown to cause pulmonary edema/fibrosis/atelectasis and systemic effects like encephalopathy and impaired immune response, even in healthy patients.
However, we really didn’t know the right oxygenation goal for routine ICU purposes.
What type of study was this?
Who was the study population? What were the inclusion/exclusion criteria?
All ICU patients over a 2-year period with…
- Age 18+
- Expected stay 72+ hours
- ICU readmission
- Comfort care
- Immunosuppressed or neutropenic
- Enrolled in another study
- COPD or ARDS (P/F ratio <150), due to separate protocols for those populations
How was randomization done? Were patients blinded? Providers? Evaluators/analysts?
Upon admission and enrollment, a random-number generator enrolled them on a 1:1 ratio by sequential opaque envelopes. No blinding after that.
What was the intervention? What was the control?
Patients received FiO2 of >.4 (presumably this means they were usually not titrated below this) to maintain…
– PaO2 “up to” 150
– SpO2 97–100%
“Lowest possible FiO2” (presumably this means titrating down to room air) to maintain…
– PaO2 70–100
– SpO2 94–98%
100% FiO2 still used for intubation, suction, transfers, but otherwise supplemental oxygen only used for SpO2 <94%.
In both groups, daily ABGs and other respiratory care (vent management, etc) as usual.
What is the primary outcome? Is it clinically significant and patient oriented?
ICU mortality. Good, although not as good as a more patient-oriented goal like 30- or 90-day mortality or survival to discharge
What were the secondary outcomes?
- Respiratory failure
- Cardiovascular failure
- Liver failure
- Renal failure
- (All defined as SOFA score of ≥3 for that organ), occurring 48+ hours after ICU admission
- Need for reoperation (in surgical patients)
- Infections (bloodstream, respiratory, surgical site)
- Vent-free hours during ICU stay and hospital mortality were performed as post-hoc analyses
Were the groups initially similar?
Sort of (Table 1). Control group was slightly older and had slightly more organ failure (shock, liver failure, renal failure, infection, etc).
Other than the therapy under investigation, is there any reason why patients might have been treated differently in the two groups?
Unblinded study (probably very difficult to blind this).
Were there any losses/failures after enrollment, and if so, were they analyzed using intention-to-treat?
46 lost: 2 for withdrawing consent, 9 for inadequate data, and 35 for an ICU stay less than 72 hours.
There was both a full intent-to-treat analysis including all patients except those who withdrew consent, and also a “modified” intent to treat analysis, which included everyone who stayed at least 72 hours and had at least 1 ABG per day.
How long was study follow-up, and was this adequate?
Not clear, but seemed to be through hospital discharge.
What are the results? Are they statistically significant? Clinically meaningful? Are they the same primary/secondary outcomes initially described? Are they plausible?
Planned for 660 patients.
Actually enrolled 480 — stopped early after an earthquake damaged the hospital and cut down admissions. An unplanned interim analysis was performed and the study was ended early as it was considered impractical to complete the full enrollment.
Intervention group did have lower FiO2 (median 36% vs 39%) and lower PaO2 (87 mmHg vs 102 mmHg).
Secondary outcomes: Hospital mortality (a post-hoc outcome, not pre-specified) was less in the treatment group (24.2% vs 33.9%, RRR 71%, ARR 9.9%, p=.03).
Treatment group had:
- Less new shock (RRP 35%, ARR 6.8%, p=.006)
- Less liver failure (RRR 29%, ARR 4.6%, p=.02)
- More hours free off the vent (median difference 24 hours; p=.02), a post-hoc analysis
- No significant difference in new respiratory failure or renal failure
- No significant difference in new infections overall, but treatment group had less bloodstream infection (RRR 50%, ARR 5%, p=.049)
Similar results in both intention-to-treat and “modified” intention-to-treat
As a post-hoc subgroup analysis, ICU mortality seemed to be reduced specifically in the patients with respiratory failure already mechanically ventilated at time of enrollment or who had a length of stay less than the overall median. (This seems to me like a meaningless result born from overanalyzing the data.)
Interpret these results
There was a dramatic decrease in mortality and improvements in several other outcomes in the lower oxygen group.
Some or all of this effect may be attributable to two factors: the baseline difference in groups (control group was sicker) and lack of blinding (perhaps more attention was paid to the intervention group).
Does the early termination compromise the results?
No! They reached significance regardless. A larger study would have had greater numbers, but they already achieved their prespecified threshold for significance.
Studies are “powered” to gather enough patients to demonstrate the effect size they predict to find. It’s only a guess to help them design their trial. Underenrolling their goal doesn’t affect the analysis, although it is a bummer.
On the other hand, due to the modest numbers, the “fragility index” of this study is fairly low (only about 3 for mortality), meaning if only a few patients had a different outcome it could have changed the results. A larger n would have boosted this number.
Also, they were unable to do meaningful subgroup analysis with these numbers, which could have been useful (it is plausible that some patients/diseases have heterogeneous responses to hyperoxia).
What are the results in terms of NNT?
NNT of 12 to avoid 1 ICU death
Who sponsored the study? Any reason to suspect bias? Conflicts of interest among the authors?
Funded by the National Fund for Scientific Research of the University of Modena and Reggio Emilia. The disclosures do not seem pertinent (no money to be made from this).
- Result is biologically plausible and consistent with some prior data, although the effect size seems unreasonably large
- A multi-center Australasia trial (ICU-ROX) on the same subject is currently underway, may provide a more definitive answer
- Probably no downside to adopting this approach, which has no real negatives except a risk of hypoxia — not mentioned here.
- My recommendations: nobody should sit with a SpO2 of 100% and pO2 >100. Do not tolerate this. Once recovered from intubation, wean them immediately. Target an SpO2 at least <100%. Write your orders for a SpO2 range somewhere in the 90s. For neuro patients, maybe target high 90s (maybe 97-99%) to give a buffer, but still do not need 100%.
- You do not need an ABG to titrate FiO2 as long as the SpO2 is reliable; use the sat.
- You do not need to titrate FiO2 in tiny steps. Respiratory/nursing may be inclined to “wean” by 5-10% per day or less. You can make reasonable reductions every few minutes (waiting just long enough to see the effects); changes in oxygenation concentration will be rapidly reflected in SpO2.
Supplemental oxygen = larger infarct size after MI.
Supplemental oxygen perioperatively does not affect wound healing, atelectasis, pneumonia
Barr J, Fraser GL, Puntillo K, Ely EW, Gélinas C, Dasta JF, et al. Clinical practice guidelines for the management of pain, agitation, and delirium in adult patients in the intensive care unit. Crit Care Med. 2013 Jan;41(1):263-306.
Who developed these?
Official product of the American College of Critical Care Medicine, a branch of the Society of Critical Care Medicine that publishes guidelines and does a few other things (like recognizing fellows). Also produced by the American Society of Health-System Pharmacists (ASHP).
Endorsed by the American College of Chest Physicians, American Association for Respiratory Care, and “reviewed” by the New Zealand Intensive Care Society.
Any conflicts of interest?
Several authors received consulting fees and honoraria (money) from various pharmaceutical companies, such as Hospira, patent-holder of Precedex.
1.b.ii: Behavioral Pain Scale (BPS) and the Critical-Care Pain Observation Tool (CPOT) are the most valid and reliable behavioral pain scales for monitoring pain in medical, postoperative, or trauma (except for brain injury) adult ICU patients who are unable to self-report and in whom motor function is intact and behaviors are observable
1.b.iii: We do not suggest that vital signs (or observational pain scales that include vital signs) be used alone for pain assessment in adult ICU patients….
1.b.iv: … we suggest that vital signs may be used as a cue to begin further assessment of pain in these patients, however.
1.c.v: We suggest that nonopioid analgesics be considered to decrease the amount of opioids administered (or to eliminate the need for IV opioids altogether) and to decrease opioid-related side effects
1.c.x: We suggest that thoracic epidural analgesia be considered for patients with traumatic rib fractures
2.a.i: Maintaining light levels of sedation in adult ICU patients is associated with improved clinical outcomes (e.g., shorter duration of mechanical ventilation and a shorter ICU length of stay [LOS]).
2.a.iv: We recommend that sedative medications be titrated to maintain a light rather than a deep level of sedation in adult ICU patients, unless clinically contraindicated.
2.b.i: The Richmond Agitation-Sedation Scale (RASS) and Sedation-Agitation Scale (SAS) are the most valid and reliable sedation assessment tools for measuring quality and depth of sedation in adult ICU patients (B).
2.b.iv: We recommend that EEG monitoring be used to monitor nonconvulsive seizure activity in adult ICU patients with either known or suspected seizures, or to titrate electrosuppressive medication to achieve burst suppression in adult ICU patients with elevated intracranial pressure.
i. Delirium is associated with increased mortality in adult ICU patients
ii. Delirium is associated with prolonged ICU and hospital LOS in adult ICU patients
iii. Delirium is associated with the development of post-ICU cognitive impairment in adult ICU patients
i. We recommend routine monitoring of delirium in adult ICU patients
ii. The Confusion Assessment Method for the ICU (CAM-ICU) and the Intensive Care Delirium Screening Checklist (ICDSC) are the most valid and reliable delirium monitoring tools in adult ICU patients
iii. Routine monitoring of delirium in adult ICU patients is feasible in clinical practice
iii. Conflicting data surround the relationship between opioid use and the development of delirium in adult ICU patients
iv. Benzodiazepine use may be a risk factor for the development of delirium in adult ICU patients
v. There are insufficient data to determine the relationship between propofol use and the development of delirium in adult ICU patients
vi. In mechanically ventilated adult ICU patients at risk of developing delirium, dexmedetomidine infusions administered for sedation may be associated with a lower prevalence of delirium compared to benzodiazepine infusions
3.d.i: We recommend performing early mobilization of adult ICU patients whenever feasible to reduce the incidence and duration of delirium
3.e.v: We suggest that in adult ICU patients with delirium unrelated to alcohol or benzodiazepine withdrawal, continuous IV infusions of dexmedetomidine rather than benzodiazepine infusions be administered for sedation to reduce the duration of delirium in these patients
a. We recommend either daily sedation interruption or a light target level of sedation be routinely used in mechanically ventilated adult ICU patients
b. We suggest that analgesia-first sedation be used in mechanically ventilated adult ICU patients
c. We recommend promoting sleep in adult ICU patients by optimizing patients’ environments, using strategies to control light and noise, clustering patient care activities, and decreasing stimuli at night to protect patients’ sleep cycles