Background: Research findings based on patient-important outcomes (PIO) provide more clinically-relevant conclusions than those from surrogate outcomes. It is unclear to what extent PIOs are represented in randomized controlled trials (RCTs) in Neonatology.
Objectives: We determined the proportion of PIOs in the neonatal RCTs in Cochrane neonatal reviews.
Methods: We analysed RCTs included in the published Cochrane neonatal reviews up to January 2016 by extracting up to five outcomes per study. Two reviewers independently determined whether each outcome was a PIO. A Neonatologist acted as an arbiter for unresolved cases and randomly cross-checked 5% of the selection for accuracy. We defined PIOs as outcomes that matter to patient care, such as clinical events, carer perception or certain physiological parameters that were widely incorporated in the guidelines as key treatment indicators. We reported descriptive statistics and performed ordinal regression using the number of PIOs (0 to 5) as the dependent variable and year of publication as a covariate (SPSS 22, Chicago, IL, USA).
Results: We extracted 6441 outcomes in 1770 RCTs published between 1952 and 2015 in 275 Cochrane neonatal reviews. A total of 4990 (77.5%) outcomes were considered PIOs. Among the studies, 426 (24.1%) included five or more PIOs, 247 (14.1%) included four PIOs, while 957 (54.0%) included one to three PIOs and 140 (7.9%) did not include any PIO. There were more dichotomous than continuous PIOs (risk ratio (RR) 1.36, 95% confidence interval (CI) 1.33 to 1.40), and slightly more subjective than objective PIOs (RR 1.12, 95% CI 1.03 to 1.22). A significant association between the year of publication and its likelihood of including more PIOs was observed (adjusted OR 1.03, 95% CI 1.02 to 1.05).
Conclusions: The large and increasing representation of PIOs over the years suggests an improving awareness by the trialists on the need for such outcomes in neonatal trials to justify the efforts and resources. There remains a concern that a small proportion of trials had no PIOs included.