Is earlier better when it comes giving caffeine to preterm infants or are we risking unnecessary treatment and serious harm?

The first clinical study on using caffeine to treat apnoeic spells in preterm infants was conducted by Aranda et al four decades ago.1 However, using caffeine only became widespread after the results of the Caffeine for Apnoea of Prematurity (CAP) trial were published in 2006.2 Between 1999 and 2004, this placebo-controlled trial enrolled 2006 infants born below 1250 grams in Canada, Australia, Europe, the United States and Israel. Caffeine started during the first 10 days of life, and the median postnatal age at randomisation was 3 days. When caffeine treatment was compared with placebo, it reduced the composite outcome of death or neurodevelopmental disability from 40% to 46%, with an adjusted odds ratio (aOR) of 0.77 and 95% confidence interval (95% CI) of 0.64-0.93 (P = .008).2 Other key results were a reduction in cerebral palsy (4% vs 7%, aOR 0.58, 95% CI 0.39-0.87, P = .009) and cognitive delay at a corrected age of 18 months, defined as a mental development index score of less than 85 (34% vs 38%, aOR 0.81, 95% CI 0.66-99, P = .04). It is noteworthy that clinicians in the CAP trial could enrol preterm infants for three different clinical indications: to prevent or treat apnoea or to facilitate the removal of an endotracheal tube. Post hoc subgroup analyses on the same population showed that respiratory improvement was greater when caffeine administration started within the first 2 days of life.3 However, the same authors urged caution when interpreting this subanalysis because the infants who started caffeine early were probably seen as more ready to wean from respiratory support than infants who were enrolled later.3 The overall benefits, and the lack of relevant harm, led to more widespread general use of caffeine in very preterm infants and to earlier treatment. Starting caffeine in the first day of life increased from 21% in 2000 to 67% in 2014.4 It has been suggested that all extremely preterm infants should receive caffeine in the very first minute of life to minimise the risk of intubation, in association with surfactant administration during non-invasive respiratory support.5 But is earlier caffeine treatment better in very preterm infants? In this issue of Acta Paediatrica, Belkhatir et al6 report the results of a two-centre retrospective cohort study on stabilisation practices and caffeine therapy in very low birthweight infants. Their aim was to assess initial preterm stabilisation practices, with a particular focus on the timing of caffeine therapy and respiratory outcomes, in two tertiary neonatal units: the John Radcliffe Hospital, Oxford, UK, and the Charité University Medical Centre, Campus Mitte, Berlin, Germany. Caffeine was discontinued at the same corrected gestational age of 34.7 weeks, in line with other studies such as the Caffeine for Apnoea of Prematurity trial.7 However, there was an important difference in the timing of caffeine initiation, with medians and interquartile ranges of zero days (0-2.5) in the UK and 2 days (1.5-4) in Germany (P < .001).6 The authors state that the earlier use of caffeine in the UK was not associated with reduced rates of bronchopulmonary dysplasia. However, due to substantial differences in baseline characteristics between the two neonatal units, we doubt that it was possible to interpret these data and draw conclusions with confidence. For example, there were important differences between the UK and German units in: the use of antenatal steroids (97% versus 79%, P < .001), Caesarean sections (56% vs 96%, P < .001) and infants born at tertiary neonatal units (85% vs 94%, P = .02). Each of these three factors might have substantially influenced the direction of the results and, importantly, could have limited the interpretation of potential associations between the timing of caffeine administration and respiratory morbidities, such as bronchopulmonary dysplasia. Moreover, immediate postnatal management was dramatically different, as the intubation rate at birth was 76% and 33% in the UK and German units, respectively (P < .001). It is likely that many other procedures might have differed, such as respiratory support settings, intubation and extubation criteria, pharmacological strategies and social factors in the areas covered by the two units. Also, the median duration of mechanical ventilation was 6 days vs 1 day in the UK and German units (P < .001). Was this large difference due, to some extent, to fewer Caesarean sections and inborn infants in the UK or to the higher intubation rate at birth? And what was the impact of the timing of caffeine initiation on these outcomes? These questions cannot be answered. Other large observational studies have attempted to address the optimal timing for caffeine administration. A retrospective analysis of 62 056 very low birthweight infants reported that initiating caffeine on the third day of life was associated with a lower rate of bronchopulmonary dysplasia and shorter mechanical ventilation than later administration.8 Similarly, another observational study by Lodha et al9 reported that early caffeine was associated with reduced bronchopulmonary dysplasia, severe neurological injuries and neurodevelopmental impairment in 2108 infants. However, comparing early to late caffeine administration in observational studies is affected by several biases that risk distorting the results. These include confounding and immortal time bias.10 Since inclusion in the observational studies on the optimal timing of caffeine administration was based on clinical indications for caffeine therapy, it was inevitable that early and late groups differed with regard to important prognostic factors. Of note, most observational studies made no attempt to adjust for clinical indications or respiratory status before starting caffeine therapy. Even if statistical adjustments had been made, we would still not know whether the results had been distorted by residual confounding. Immortal time bias arises when part of the follow-up period includes time when the participants, by definition, could not have died. This creates a selection bias of living, stronger babies that favours late intervention groups.10 In other words, the late group will only comprise infants who survived until they could receive caffeine, while the early group will include some infants who died in the period between when they received caffeine and when inclusion for the late group started. Some published trials have had a low risk of bias. For example, Amaro et al compared the effects of early vs late caffeine on time with extubation in ventilated preterm infants born at 23-30 weeks of gestation.11 The early caffeine group received a loading dose of caffeine followed by maintenance therapy, while the control group received saline boluses. Just before extubation, the control group received a blinded loading dose of caffeine, while the early caffeine group received saline. However, the trial was stopped at 75% enrolment because the interim analysis showed that the early caffeine group had a trend towards higher mortality. The subsequent unblinded analysis revealed that there was no statistically significant difference in death before discharge between the two groups: 9/41 (22%) in the early caffeine group vs 5/42 (12%) in the control group (P = .22). Because the trial ended early, it was underpowered to draw firm conclusions. Another trial by McPherson et al12 randomised very preterm infants to either high or standard doses of caffeine in the first few hours of life and showed higher rates of cerebellar haemorrhage in the high dose group. Although this trial compared the dose, not the timing of, caffeine administration, it should be cautiously considered that administering caffeine at an early stage might cause potential harm. In conclusion, we doubt that it is possible to entirely avoid residual confounding and immortal time bias in observational studies that compare early versus late caffeine administration. Randomised controlled trials are needed to address whether earlier caffeine treatment is better for preterm infants or whether we risk unnecessary treatment and serious harm? The authors have no conflicts of interest to declare. Minna Johansson Karsten Juhl Jørgensen Matteo Bruschettini

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.