Of course you can have too much of a good thing, but you should also try and stay above the minimum.
The COT trial investigators, led by Christian Poets, have published a secondary analysis of the results, examining in particular data from the study oximeters. Despite the relatively low fidelity of the oximeter data (1 sample each 10 seconds) they were able to analyze the number and duration of hypoxic episodes among the 1020 babies who were randomized and who survived to 36 weeks PMA.
An episode of hypoxemia was diagnosed when the saturation was below 80%, and of bradycardia when below 80. There was a clear correlation between a longer time with a saturation under 80 and poorer outcomes at 18 months PMA. Cognitive and Language scores below 85 on the Bayley version3, motor disability and severe retinopathy were all associated with longer durations of hypoxia.
The effects were mostly associated with longer episodes, so when the authors looked at hypoxic episodes where 6 or more consecutive readings (taken as you recall every 10 seconds) were hypoxic, the effects were substantial, whereas shorter episodes of 5 readings or fewer, were less important.
What causes these hypoxic episodes? Most are probably due to apnea of prematurity, although those occurring during assisted ventilation may additionally have other mechanisms, including forced expiration and loss of lung volume.
I have published 2 articles (here and here) trying to determine if there is a relationship between apnea and long term outcomes, both seemed to show a relationship. The improved outcomes in the CAP trial among preterm babies who were treated with caffeine, could also have been due to the babies having fewer hypoxic episodes, although there were no recordings of oximetry to confirm that.
I also wonder if the problem with prolonged episodes may be related to post-desaturation hyperoxia, I and others have shown that apneas are often followed by hyperoxia, possibly leading to re-oxygenation injury, and free radical mediated CNS injury. In my 2 studies that I published only as abstracts, I analyzed recordings of babies who usually had saturations below 95%, and showed that after apneas, they were often over 95%. I thought that the FiO2 of the babies was probably being increased by the bedside staff in response to the apneic hypoxia, but I couldn’t be sure as we didn’t have FiO2 recorded. A much better study was published last year, van Zanten HA, et al. The risk for hyperoxaemia after apnoea, bradycardia and hypoxaemia in preterm infants. Archives of Disease in Childhood – Fetal and Neonatal Edition. 2014. in that study babies on nCPAP who had apneas often had hyperoxia after the apnea, and the hyperoxia was much longer than the hypoxia! The hypoxia (less than 80%) lasted on average 2 minutes, and the post-apneic hyperoxia lasted 13 minutes.
The always perceptive Lex Doyle has written an editorial to accompany the article, he points out that the association may not be causative, and that we really have only a limited number of ways to prove whether it is. A therapy proven to reduce the number and duration of hypoxic episodes, and then prospective randomized trials to demonstrate whether outcomes are improved. We are already part way there with the CAP trial, but most of the babies in this new trial will have been getting caffeine, for substantial periods of time, and they were still having lots of hypoxic episodes.
Higher doses of caffeine have been touted, but the concerning results of a pilot trial which showed an increase in cerebellar injury should make us pause. Other agents such as doxapram are suggested by Christian Poets in the discussion section of the article, which sort of brings me full circle, as I made my start in clinical research studying doxapram. But while it is effective in reducing apneas, long term effects are uncertain, infants who received doxapram have been reported in observational studies to have worse outcomes. Which again of course, does not mean causation, but does mean that any studies should be performed with adequate safety protections, and long-term outcome evaluations.