RCT evidence should drive clinical practice: A day without randomisation is a day without progress

‘We must change community culture so that one of the first questions asked by pregnant women and their families is, “Are there any randomised trials we can join?”’ Melinda Cruz, parent.

Doctors should never be obliged to invite patients to join trials. Instead, we need a self-improving health system that embeds randomised trials in routine care. This has already been achieved with remarkable success in treating acute lymphoblastic leukaemia. Fifty years ago, about 90% of affected children died within a year. Following a steady succession of trials in which about 90% of eligible children were enrolled, about 90% now survive. When randomised care replaces random, opinion-based care, incremental progress can be anticipated in any speciality (Djulbegovic et al., Nature 2013;500:395–6).

There is now abundant evidence that Phase II and Phase III RCTs represent no net risk for patients. In retrospective analyses, RCTs of new versus established treatments posed no greater hazard than did established treatment outside RCTs (Fernandes et al., CMAJ 2014;186:E596–E609, Foglia et al., JAMA 2015;313:2377-9). In prospective analyses of 860 Phase III trials in around 350 000 patients across various clinical fields, on average, newer treatments improved primary outcomes and survival just over half the time (Djulbegovic et al., Cochrane Database Syst Rev 2012;10:MR000024). Thus, patients included in a Phase III RCT had a better chance of superior treatment than those receiving usual care. In a commentary describing six consecutive series of Phase III controlled trials, it was noted that when randomised care replaces random, opinion-based care, incremental progress can be anticipated in any speciality (Djulbegovic et al., Nature 2013;500:395-6). We suggest that hospital websites and research consent forms should summarise evidence like this when describing the risks and benefits of research.

Many treatment benefits are moderate, requiring thousands of patients to demonstrate them reliably (Tarnow-Mordi et al., Semin Fetal Neonatal Med 2015;20:389–402). In the CHIPS trial of 987 pregnant women, less-tight versus tight control of hypertension had little effect on the odds of pregnancy loss or high-level neonatal care (odds ratio 1.02 [95% CI, 0.77–1.35]), but this wide confidence interval does not rule out a 23% lower or 35% higher odds of this outcome (Tarnow-Mordi et al.). In many neonatal RCTs the risk ratios of the primary outcomes also have wide CIs. This imprecision means that trials conducted at substantial cost, time and burden upon families and clinicians ultimately cannot reliably exclude clinically relevant benefit or harm. Trials must be easier and less costly and burdensome, especially if large sample sizes are needed.

How can larger, simpler, more inexpensive and efficient comparative effectiveness trials be integrated into routine care? We suggest six strategies: (i) developing networks of local staff to support enrollment and conduct of trials; (ii) closer partnership with parents and clinicians in all aspects of trials; (iii) routine electronic point-of-care patient data of high quality for patient identification, randomisation and measurement of outcomes, reducing the need for new databases for each trial; (iv) information sheets of two pages or less to minimise ‘injurious misconception’, where patients refuse because of exaggerated perceptions of harm; (v) explicit mention of possible inclusion benefit in information sheets; (vi) opt-out consent with enrollment as default (Gale et al. Arch Dis Child Fetal Neonatal E 2016; DOI: 10.1136/archdischild-2016-310935).

The newly conceived ALPHA Collaboration will advance large, efficient perinatal trials for health outcomes assessment through international partnership. Ultimately, it aims to catalyse a succession of low-cost perinatal mega-trials enrolling 5000–50 000 patients or more, to guide practice. Enquiries will be warmly welcomed by the authors.