Extremely preterm infants (<28weeks’ gestational age) have higher ratesof adverseneurodevelopmental outcomes inearly childhood than children born at term.1 In this issue of JAMA, Poets et al2 present a post hoc analysis of data from 1019 extremelypreterminfants in the multicenter Canadian Oxygen Trial (COT),3 analyzing the relationships between episodesofhypoxemia (pulseoximetryoxygen saturation levels <80%) andbradycardia (pulse rate <80/min)measured every 10 seconds for a mean of 68 days and the primary outcomeofdeathordisability (cognitiveor languagedelay,motor impairment, deafness, or blindness) at 18 months’ corrected age. Secondary outcomes included cognitive or language delay, motor impairment, and severe retinopathy of prematurity. The investigators found that episodes of hypoxemiawere more important thanepisodesofbradycardia.Hypoxemicepisodeswere associatedwith death or disability in 56.5% in the highest decile of exposure vs 36.9% in the lowest decile (modeled relative risk, 1.53; 95% CI, 1.21-1.94). The strengthsof the study include thecompletenessof the collectionandanalysis of thepulseoximetrydata, thehigh follow-up rates and blinded assessment of important endpoints at 18months of age, and thedetailed statistical analyses. That the association between hypoxemia and adverse outcomes could be causative is supported by an apparent doseresponse relationship,with theassociationonly significant for episodes lasting more than 1 minute. Two questions follow from their observations. First, are the findings valid? Second,whatdo the findingsmean for current neonatal care of extremely preterm infants? The observed associations could be confounded by other perinatal events related to both hypoxemic episodes and adverse neurodevelopmental outcomes. For the findings to be valid, confounders, particularly those thatmight predate the exposure to hypoxemic episodes, should have been adjusted for. It would not be appropriate to adjust for all known perinatal variables related to adverse neurodevelopmental outcomes, particularly those that occur after the hypoxemic episodes, because theymay just be a consequence of hypoxemia and be included in the pathway from hypoxemia to adverse neurodevelopmental outcome. The variables adjusted for includedgestationalage, sex,primarycaregiver’seducation level, useof antenatal corticosteroids,multiplebirth, andstudycenter. Variables that were not adjusted for include sepsis/ necrotizing enterocolitis, early brain abnormalities such as grade IV intraventricular hemorrhage, caffeine treatment, or othermeasures of early disease severity, such as ventilator or oxygen requirements. Sepsis, including necrotizing enterocolitis, is associated with more severe brain white matter abnormalities at termequivalent age and later adverse neurodevelopmental outcomes4 and often presents clinically with episodes of hypoxemia.Caffeine treatment reducesapnea,5oneof themechanisms causing hypoxemia. Caffeine administration improvedneurodevelopmental outcomes at 18months of age in preterm infants6 and also was associated with improvedmotor outcome at age 5 years7 but was not associated with significantly improved rates of survivalwithout disability at that age.8 Someof these confoundersmight havepredated the occurrence of the hypoxemic episodes, and, therefore, the resultswouldhavebeenstrengthenedwithadjustment for these variables, at least in sensitivity analyses, to determine if the association between hypoxemia and adverse neurodevelopmental outcomes was attenuated. Theassociationbetweenhypoxemiaandadverseneurodevelopmental outcome is consistent with the known associationbetweenapnea,onecauseofhypoxemia, andadverseneurodevelopmental outcomes.9 Apnea, in turn, is related to abnormalbrain stemfunction,withprolongedbrain stemconduction times occurring in preterm infants with vs without apnea.10Whetherhypoxemiaperse is relatedtoabnormalbrain stem conduction time remains to be determined. What do the findings mean for neonatal care of all extremely preterm infants? It is not clear to whom the results shouldapply, for several reasons. First, as is true formany randomized clinical trials, participants may not be representative. Less than 50%of screened infants born at 23 to 27weeks’ gestation entered theoriginal COT study,3 and theremayhave been additional unscreened infants in the participating centers during the recruitment phase. Also, the prognosis for infants entered into randomized clinical trials may differ systematically from those not enrolled.11 Second, although the authors limited the study to infantswho survived to 36weeks and outcomes to late death or disability, the hypoxemic episodesweredocumentedpredominantly before 36weeks’ gestational age. Infantswhodiedbefore36weeks, comprising 14% (166/1201) of those enrolled into theCOT study,mayhavehad more or less time with prolonged hypoxemic episodes than those who survived to 36 weeks, and inclusion of their data in the primary outcomemay have altered the observed association with prolonged hypoxemia. The other challenge in translating the results into clinical practice relates to the importance of the major contributor to Related article page 595 Opinion