METHODS

Search strategy

PubMed Medline (1966–2002) was searched using the terms indomethacin and infant and (prophylactic or prophylaxis or prevention) and the Cochrane Controlled Trials Register using indomethacin and infant. Abstracts of the Society for Pediatric Research and the European Society for Paediatric Research were searched from 1996–2001, and full text articles sought using Medline searches of the authors’ names. Previous reviews were cross referenced, and personal files searched for additional references. No language restrictions were applied.

Inclusion criteria

Both authors assessed all published articles and abstracts identified as potentially relevant by the literature search for inclusion in the review. In order to be included, trials had to meet all four of the following criteria:

• Study design: randomised controlled trials

• Participants: preterm infants

• Intervention: intravenous indomethacin given within 24 hours of birth

• Outcome measures: included any of the following—death, IVH, PDA, or long term neurodevelopmental outcome

Quality assessment and data abstraction

Both authors assessed each article according to the following criteria: blinding of randomisation, blinding of intervention, completeness of follow up, and blinding of outcome assessment. Both authors extracted the data from each trial independently, then compared results and resolved differences.

Data analysis

Data measuring similar outcomes were combined in a meta-analysis where appropriate. For categorical outcomes, treatment effect was analysed using relative risk (RR), risk difference, and number needed to treat with their 95% confidence intervals (CI). A fixed effects model was used. Evaluation of heterogeneity of results was performed for all outcomes, and p < 0.05 on γ² test was considered to represent significant heterogeneity.

RESULTS

Quality assessment

Nineteen randomised trials comprising 2872 infants fulfilled the inclusion criteria. Four of these reported long term outcomes (1862 infants). The exact method of randomisation was specified in 12 of the studies.^3–14 Methods included telephone randomisation, sealed envelopes, and coded drug vials. In the remaining seven trials it was not possible to determine how well the process of randomisation was blinded.^15–21 In three studies it was not possible to determine whether the caregivers and those assessing the outcomes of interest were blinded to the intervention group.^17,19,20 In the remainder there was adequate description of methods used to ensure blinding including description of the placebo.

Follow up rates for short term outcomes were adequate—that is, > 90%—for all included studies. Long term follow up rates were less complete, ranging from 75% to 100%.

Trial characteristics

Table 1 shows clinical details of included studies. Some clinical practices changed over time—for example, surfactant was given either as prophylaxis or rescue therapy in seven trials.^4,10–14,16

Quantitative data synthesis

Table 2 shows the pooled results of the trials. There was no difference in mortality to latest follow up between the treatment and placebo groups (RR 0.96 (95% CI 0.81 to 1.12)). The reduction in cranial ultrasound abnormalities (all IVH (RR 0.88 (95% CI 0.80 to 0.96)) and severe (grades 3 and 4) IVH (RR 0.66 (95% CI 0.53 to 0.82)) seen in the treatment group did not translate into improved long term outcomes. Rates of mortality and severe neurosensory impairment (blindness, deafness, cerebral palsy, or developmental quotient more than two standard deviations below the mean) were high in this group of infants and were not decreased by prophylactic indomethacin (RR 1.02 (95% CI 0.90 to 1.15)).

There were some short term advantages observed in infants given indomethacin. The rate of PDA (RR 0.44 (95% CI 0.38 to 0.50)) and particularly the rate of surgical ligation for this condition (RR 0.51 (95% CI 0.37 to 0.71) were reduced in the treatment group. Twenty infants would need to be treated with prophylactic indomethacin to prevent one surgical ligation. There were no differences in other short term outcomes including necrotising enterocolitis and bronchopulmonary dysplasia. Although oliguria was more often observed in infants receiving indomethacin, there were no differences in proportions of babies developing high serum creatinine levels.

Heterogeneity of results was found only for the outcome “all IVH” (p = 0.011). The remaining pooled outcomes had p values > 0.2 for heterogeneity, indicating that variability between studies may be explained by chance alone.

DISCUSSION

There is now a substantial body of literature available to evaluate the role of prophylactic indomethacin in preterm infants. The quality of trials included in this systematic review is good but there is variation in enrolment criteria, indomethacin dosage regimens, and some of the outcome definitions.

This review confirms the usefulness of prophylactic indomethacin in the prevention of symptomatic PDA. Some clinicians would find the reduction in the need for surgical ligation of the ductus, combined with the lack of evidence for short or long term harm, justification for providing this treatment. Other factors influencing such a decision would be the background rate of ligations and the availability of cardiology and cardiac surgery services.

In the past, the presence of a PDA was thought to increase the risk of developing both pulmonary haemorrhage and bronchopulmonary dysplasia. Reduction in the rate of PDA without a significant change in rates of bronchopulmonary dysplasia or pulmonary haemorrhage challenges traditional assumptions about the pathophysiology of these conditions.

This review confirms the significant reduction in rates of severe intraventricular haemorrhage in infants given indomethacin. This result is not accompanied by any of the adverse outcomes that were possible given the vasoconstrictive nature of the drug—that is, important renal side effects, gastrointestinal perforation, and necrotising enterocolitis. However, the improvement in rates of IVH did not translate to improvement in rates of neurosensory impairment. One clear message of this systematic review is that traditional surrogate outcomes used in evaluating interventions in preterm infants may not be sufficient to guide changes in treatment. Put another way, when new interventions are considered in this population, particularly when they are being given prophylactically (and therefore to some infants who are not expected to benefit from them), long term outcomes should be assessed.