OBJECTIVE: The goal was to reduce adverse pulmonary adverse outcomes, oxidative stress, and inflammation in neonates of 24 to 28 weeks of gestation initially resuscitated with fractions of inspired oxygen of 30% or 90%.
METHODS: Randomized assignment to receive 30% (N = 37) or 90% (N = 41) oxygen was performed. Targeted oxygen saturation values were 75% at 5 minutes and 85% at 10 minutes. Blood oxidized glutathione (GSSG)/reduced glutathione ratio and urinary o-tyrosine, 8-oxo-dihydroxyguanosine, and isoprostane levels, isofuran elimination, and plasma interleukin 8 and tumor necrosis factor α levels were determined.
RESULTS: The low-oxygen group needed fewer days of oxygen supplementation (6 vs 22 days; P < .01) and fewer days of mechanical ventilation (13 vs 27 days; P < .01) and had a lower incidence of bronchopulmonary dysplasia at discharge (15.4% vs 31.7%; P < .05). GSSG/reduced glutathione × 100 ratios at day 1 and 3 were significantly higher in the high-oxygen group (day 1: high-oxygen group: 13.36 ± 5.25; low-oxygen group: 8.46 ± 3.87; P < .01; day 3: high-oxygen group: 8.87 ± 4.40; low-oxygen group: 6.97 ± 3.11; P < .05). Urinary markers of oxidative stress were increased significantly in the high-oxygen group, compared with the low-oxygen group, in the first week after birth. GSSG levels on day 3 and urinary isofuran, o-tyrosine, and 8-hydroxy-2′-deoxyguanosine levels on day 7 were correlated significantly with development of chronic lung disease.
CONCLUSIONS: Resuscitation of preterm neonates with 30% oxygen causes less oxidative stress, inflammation, need for oxygen, and risk of bronchopulmonary dysplasia.
Comments
How to plan, perform and critically appraise clinical trials: methodology matters.
We read with great interest the report of Dr. Vento and colleagues on the impressive effect of low versus high FiO2 during the initial minutes of resuscitation of very preterm infants (1). Based on our joint responsibility for education of younger colleagues, several methodological inconsistencies in this report need to be addressed: 1) “The main clinical outcome was neonatal death (death at < 28 days) plus the incidence of BPD“, however, the sample size calculation was only based on an estimated incidence of BPD of 30% in the control group. 2) A study that aims to show a reduction in the incidence of BPD from 30% to 20% (under the standard assumption of α=0.05, β=0.20, 1:1 allocation, and a two-sided test) requires about 290 patients in each group and not 35 as stated by the authors. In contrast, a study with only 35 patients per group has only about 30% power; i.e., the probability to miss an existing beneficial effect of the experimental intervention would be 70%. The authors were extremely fortunate to show the beneficial effect of their intervention against these odds. 3) It is confusing that the authors report that “[…] of 106 eligible neonates, 9 in the low-oxygen group and 10 in the high-oxygen group were lost to randomization because parents declined to participate“, because one would expect that randomization occurred after parental consent had been obtained. On the other side, a rate of withdrawal of consent after initial approval to participate of 18% appears to be high when the intervention lasts only a few minutes. We are concerned that young scientists in clinical medicine should not only be taught that randomized controlled trials are the gold standard for the evaluation of experimental interventions, but should also be taught how to plan and perform such trials correctly according to GCP guidelines and how to critically appraise trial reports.
1 Vento M, Moro M, Escrig R, et al. Preterm Resuscitation With Low Oxygen Causes Less Oxidative Stress, Inflammation, and Chronic Lung Disease. Pediatrics 2009;124:e439-e449
Conflict of Interest:
None declared
Preterm resuscitation with low oxygen: a case for developing countries
We read the article by Vento et al with great interest as the harmful effects of oxygen are becoming more obvious now.(1) Newborn babies in developing countries are often resuscitated in conditions with no availability of oxygen-air blender and pulse-oxymeter. In such situations clinicians are often left with the option of using high oxygen concentrations with short-term and long-term adverse consequences. This can be decreased by avoiding attaching the reservoir to the ambu-bag which provides around 40% oxygen without the reservoir. If the neonate does not improve with this then attaching the reservoir will provide around 90% oxygen. This may be a simple way of avoiding oxygen toxicity where sophisticated equipment is not available. Reference: Maximo Vento, Manuel Moro, Raquel Escrig, Luis Arruza, Gema Villar, Isabel Izquierdo, L. Jackson Roberts, II, Alessandro Arduini, Justo Javier Escobar, Juan Sastre, and Miguel A. Asensi
Preterm Resuscitation With Low Oxygen Causes Less Oxidative Stress, Inflammation, and Chronic Lung Disease
Pediatrics 2009; 124: e439-e449
Conflict of Interest:
None declared
Preterm resuscitation with low oxygen: problems with study design and analysis
I read with interest, Dr Vento’s paper on preterm resuscitation published in the September issue of Pediatrics (1). I have a number of observations about the methodology of this paper.
The lack of a flow diagram and lack of denominators for each outcome has created confusion about the number of subjects involved. There were 106 eligible neonates, of which 19 were excluded post-randomisation because parents refused to participate and another 7 died, leaving 37 and 41 subjects included in the low oxygen and high oxygen groups, respectively. That makes it a total of 104 (=19 + 7 + 37 + 41), leaving 2 patients unaccounted for.
The authors state that if an infant was randomized, but not resuscitated, then the data was not included in the study. They have not clarified how many such subjects were there in the study. Table 1 shows that only 32 and 34 randomised subjects respectively received supplementary oxygen for resuscitation. Was the remainder (5 and 7 subjects respectively) not included in the study, and is this the same as the randomized but non-resuscitated group the authors allude to in the text? If so, the denominators for all outcome measures would be 32 and 34, instead of 37 and 41.
There is also some lack of clarity about the deaths. The main clinical outcome was a composite of “neonatal death plus the incidence of BPD”. Though this was the primary outcome, the sample size was calculated for only BPD as the outcome. If death was an outcome, it is not clear why 4 deaths in the low oxygen group and 3 in the high oxygen group were excluded post-randomization because they did not complete the study. Yet, death was analyzed as an outcome in table 2. Are these the same deaths that were supposedly excluded? If so, the issue of the denominators becomes even more puzzling.
I do not understand how the authors arrived at a sample size of only 35 patients per group. They do not mention what alpha and beta errors they assumed for the calculation; but supposing they were (as per convention) 5% and 20% respectively with an effect size of 10% and baseline risk of 30%, the sample size works out to be 313 per group, using EpiInfo v3.5.1 (http://www.cdc.gov/EpiInfo/). This is almost a 10-fold difference.
The p-value quoted for the comparison of incidence of BPD is incorrect. The authors claim it is significant at <0.05 (Table 2), but both by EpiInfo and EBM calculator v1.2 (http://www.cebm.utoronto.ca/palm/ebmcalc/), the p-value of the 2x2 Yates corrected Chi square test for BPD works out to be 0.18. Similarly, the p- value quoted for the comparison of breathing 21% oxygen at arrival in NICU is incorrect. The authors claim it is significant at <0.05 (Table 1), but it is actually 0.15.
The comparison of the main outcome (a composite of death and BPD) finds no mention in the paper.
The authors say that stratified randomisation was done, but there is no mention of blocking. Without blocking, stratified randomisation would not balance the potential confounder, because within each stratum the confounder would remain unbalanced.
In numerous instances in the study, repeated measurements of parameters have been performed. Wherever two groups have been compared, the authors have opted to perform multiple statistical tests with p-values of Student’s t or Mann Whitney shown for comparisons at each individual point of time. This is not the appropriate approach because it inflates the alpha error substantially and increases the possibility of getting significant differences purely by random chance. For instance, Figure 1 alone entails 32 comparisons. The correct method would have been to perform two-way repeated measures ANOVA or two-way Friedman's test.
All p values in the paper have been expressed as <0.05 or <0.01, without giving exact p-values or 95% confidence intervals. It is off course evident that an intention-to-treat analysis was not performed.
With all these problems in design and analysis, I am skeptical about the results of the investigators.
References
1. Vento M, Moro M, Escrig R, Arruza L, Villar G, Izquierdo I, Roberts LJ, Arduini A, Escobar JJ, Sastre J and Asensi MA. Preterm resuscitation with low oxygen causes less oxidative stress, inflammation and chronic lung disease. Pediatrics 2009; 124: e439-e449
Conflict of Interest:
None declared