Does NIRs serve as a good predictor of outcome in ECPR?

In this issue of the journal Bertini and colleagues (JCVA-D-22-00988)) conducted a meta-analysis of rSO₂ values in patients who received extracorporeal cardiopulmonary resuscitation (ECPR) either in hospital or out of hospital. The authors were only able to identify 3 trials (245 participants for inclusion) for analyses. But the results of their analysis suggest that low rSO₂ may predict mortality. The authors found that a low pre-cannulation rSPO₂ was associated with an increased risk of mortality and overall worse neurological outcomes in patients who underwent ECPR. However, the interpretation of this meta-analysis raises significant concerns about the clinical validity. Firstly, the small sample size of the meta-analysis, and secondly, the included studies used the initial measurement of NIRs values, after a significant period of CPR for out of hospital arrest and limited CPR for in hospital arrest while many of the other CPR trials look at changes in NIRs during CPR. More importantly, while the mortality in the low rSO₂ group is high, it is not 100% suggesting that failing to proceed with ECPR would result in additional deaths, Finally, one of the trials was published in 2011 and so may not reflect contemporary practice. In the article, low rSO₂ was defined as a pre ECPR value of less than 16%, the lowest value detectable by the INVOS system (used in all the trials) with mortalities in the two groups of 64.9% in the high group and 92.5% in the low group. Likewise, the authors identified an association with neurologic outcome, with patients in the high rSO₂ group having good neurologic outcome (27.8%) versus those in the low rSO₂ group (2.1%). However, interestingly if you look at survivors those in the high group had a similar rate of good neurologic outcome as those in the low rSO2 group (42/151 high, versus 2/7 in the low group), suggesting that poor neurologic outcome may not be as easy to predict using NIRs.

The use of NIRs in CPR has had mixed results, with some trials suggesting that NIRs can predict return of spontaneous circulation, a lower bar than mortality, while others suggest no such ability^1-3 It should be noted that many of these trials looked at change in baseline rSO₂ with CPR, but that the identified cut off values were not consistent between these studies and those presented in the paper by Bertini et al.

An interesting paper⁴ by Nomoto-Nemoto, showed that brain dead patients had NIRs values of 60-65, suggesting blood flow but not extraction, which may make neurologic recovery difficult to determine of the basis of NIRs measurements alone.

Recently, two randomized controlled trials were conducted and published to answer a clinically significant question about the relationship between blood oxygen saturation at the time of cardiac arrest and patient prognosis^{5, 6}. In one paper, 789 patients who experienced out-of-hospital cardiac arrest were randomized into two groups⁵. One group was targeted to have a PaO₂ level of 68 to 75 mmHg, while the other group was targeted to have a level of 98 to 105 mmHg. The study results showed no significant difference in 90-day mortality or in the rates of severe disability or coma. A separate study⁶ conducted by Bernard et al. in 2022 investigated the impact of oxygen saturation on survival after out-of-hospital cardiac arrest. One thousand seven hundred nineteen patients who underwent resuscitation were randomly assigned to two groups: a high oxygen group with a saturation level of 94% and a low oxygen group with a level of 90%. The results showed no significant difference in the rate of survival to hospital discharge between the two groups. These findings suggest that the blood oxygen concentration and hemoglobin saturation (SPO₂) may not play a significant role in survival after cardiac arrest. Hirakawa et al. ⁷ examined pre-ECMO pO₂ and its relationship with ischemic stroke and found no association.. However, the incidence of stroke was associated with lower pre-ECMO pH and higher pO₂ value measured at 24 hours after ECMO was established. It is important to keep in mind that, in a retrospective study, only correlations can be demonstrated, not causation. The more factors there are, the more likely some will come up as correlated. Further thorough investigation and well-designed randomized controlled trials are necessary to establish causation.

It's important to note that none of the papers analyzed in the current meta-analysis included long-term follow-up or reported on any vital organ dysfunctions. Without this information, caution must be exercised when making clinical decisions about ECMO, as it is a costly and resource-intensive procedure. The decision to initiate and discontinue ECMO should be based on multiple factors, and it would be reckless to make decisions solely based on pre-cannulation rSPO₂ values without considering all relevant data.

Also, of concern is the small sample size of this meta-analysis. For example, in the current meta-analysis, there are around 100 patients in each of the high and low rSO₂ groups. Using such a small number of patients to make inferences is not reliable and raises questions about the generalizability of the conclusions drawn. Additionally, the finding that patients with pre-cannulation rSO₂ > 60% had a worse prognosis than 40% < rSO₂ < 60% is based on a sample size that is smaller than 10, which further underscores the importance of caution when interpreting the results.

The authors note that all the studies analyzed lacked information on the timing of rSO₂ measurement, oxygen administration, and CPR. It is important to note that oxygen administration can significantly impact rSO₂ readings, and this should be considered when interpreting the results. Looking closely at three papers analyzed in the meta-analysis, Joo et al. investigated 121 patients using the large Japanese database whose data was taken from 15 different centers. This study is retrospective and not blinded. The timing of rSO₂ measurements is not specified in the paper. Ito et al.⁸ looked at 27 consecutive patients undergoing ECPR. The results, from 2011, were only ever published as an abstract and have not undergone peer review. This should raise questions about the quality of the published data, and should have resulted in a clear statement within the trial that one of the studies was in abstract form only and a subsequent sensitivity analysis looking at the impact on the results of including or excluding the trial. Wiest et al. published a short paper⁹ describing the study investigating 97 patients undergoing ECPR retrospectively. In this study, rSO₂ was measured upon arrival of the resuscitation team. It is crucial to carefully evaluate the quality and representativeness of the studies included in the meta-analysis and to take these limitations and potential sources of bias into account when interpreting the results. This will ensure that the results of the meta-analysis are accurate and reliable and provide a more comprehensive understanding of the topic under investigation.

Generating data in trails of CPR is challenging and results in many small trials, or retrospective observational studies, making definitive conclusions difficult. In addition, the ability of any test to predict, not only good but bad outcomes, is inherently related to the pretest probability of the event occurring. Analysis of a tests ability to predict outcome should rely on the tests positive and negative predictive values and not solely on positive association, as was shown by Bertini et al. As such deciding what cut offs to employ for appropriate sensitivity and specificity is clinically challenging, especially when the outcome is so irreversible. The use of NIRs as a predictor for outcome in both CPR and ECPR remains unclear and further work needs to be done to assess its clinical usefulness.