Randomised Clinical Trials. David Machin
the numbers receiving pharmacologically inactive systemic treatment, and to increase the likelihood of identifying infrequent adverse events.
Key features include the following:
Design: Single‐centre, randomised double‐blind, placebo‐controlled, randomised 2 : 1 allocation ratio using minimisation,
Endpoint: SASSAD,
Size: 63 patients with moderate‐to‐severe atopic eczema,
Analysis: Comparison of mean group regression slopes over a 12‐week period,
Conclusion: Azathioprine produces a clinically relevant improvement.
Example 1.8 Cross‐over trial – known or suspected hypertension
Kerley, Dolan, James and Cormican (2018) describe a randomised placebo (P) controlled, two‐period cross‐over trial of dietary nitrate (N) in 20 patients with known or suspected hypertension. The P and N interventions were delivered in beetroot juice in a nitrogen‐depleted or nitrogen‐enriched form, respectively. Thirteen of the individuals were randomised to receive the sequence NP, that is N in Period I of the trial followed by P in Period II, and the other seven were allocated the sequence PN. Amongst the many endpoints, plasma nitrate and ambulatory blood pressure were recorded prior to randomisation, then 7 days later following the start of treatment in Period I and a further 7 days following Period II. The authors concluded:
Our results support … an anti‐hypertensive effect of dietary nitrate … .
Key features include the following:
Design: Single‐centre, randomised placebo‐controlled, two‐period cross‐over trial,
Size: 20 patients with known or suspected hypertension, unequal numbers assigned to the sequences,
Washout: None included,
Endpoints: Plasma nitrate and ambulatory blood pressure,
Analysis: Complex methodology described. All statistical tests were conducted at the two‐sided 0.05 significance level,
Conclusion: Nitrogen‐enriched has an anti‐hypertensive effect.
Example 1.9 Paired design – glaucoma
Glaucoma Laser Trial Research Group (1995) recruited 271 subjects with newly diagnosed primary open‐angle glaucoma, and one eye of each patient was randomly assigned as initial treatment by argon laser (L) trabeculoplasty followed by Stepped (S) medication (LS). The other eye then received the treatments in reverse order, SL. They reported on the 261 eyes and found that measures of visual field status for eyes treated by the sequence LS were slightly better than those treated by SL. The authors’ state:
Statistical significance was attained for only some of the differences, and the clinical implications of such small differences are not known.
Key features include the following:
Design: Multicentre, paired design, compares alternative schedules for administering two procedures – the schedule was randomised to one eye with the other eye receiving the alternative,
Endpoint: Visual field status,
Size: 271 patients with primary open‐angle glaucoma,
Analysis: Comparison of means at particular time points following initiation of treatment using the paired t‐test,
Conclusion: Eyes treated with laser trabeculoplasty first were slightly better than those eyes treated with topical medication first.
Example 1.10 Split‐mouth design – implants for edentulous sites
Pozzi, Agliardi, Tallarico and Barlattani (2012) conducted a trial in 34 partially edentate patients who required at least two single implant‐supported crowns. A split‐mouth design was used in which one of two different prosthetic interfaces and configurations: internal conical connection with back‐tapered collar and platform shifting (CC) or external‐hexagon implants with flat‐to‐flat implant‐abutment interface (EH), were randomly allocated at each edentulous site. From a total of 88 implants included in the trial, the authors concluded that both implants performed similarly in terms of failure rates.
Key features include the following:
Design: Single‐centre, split‐mouth, random allocation,
Endpoint: Failure rates and marginal bone loss,
Size: 34 patients with 88 edentulous sites,
Analysis: Comparing implants using paired t‐tests at several intervals postrandomisation,
Conclusion: Lower marginal bone loss with CC when compared to EH.
Example 1.11 Cluster trial – hip protectors for the elderly
Meyer, Warnke, Bender and Mülhauser (2003) conducted a trial involving 942 residents from 49 nursing homes. In this cluster trial design, the nursing homes contain ‘clusters’ of residents and the homes (not the individual residents) were randomised, with 25 homes, comprising a total of 459 residents, assigned to the intervention group and 24, with 483 residents, to the usual care (control) group. The intervention comprised a single education session for nursing staff, who then educated residents, and the provision of three hip protectors per resident. The control clusters administered usual care optimised by brief information to nursing staff about hip protectors and the provision of two hip protectors per cluster for demonstration purposes. The main outcome measure was the incidence of hip fractures. There were 21 hip fractures in 21 (4.6%) residents in the intervention group and 42 in 39 (8.1%) residents in the control group – a difference of 3.5% (95% CI 0.3–7.3%, p‐value = 0.072). The authors concluded:
The introduction of a structured education programme and the provision of free hip protectors in nursing homes may reduce the number of hip fractures.
Key features include the following:
Design: Multicluster, randomised,
Size: 49 nursing homes comprising 942 residents with high risk of falling,
Endpoint: Hip fractures,
Analysis: Chi‐squared test adjusted for cluster randomisation,
Conclusion: A structured education programme and the provision of hip protectors may reduce the number of hip fractures.
Example 1.12 Single arm – discrete de novo lesions in a coronary artery
Erbel, Di Mario, Bartunek, et al. (2007) describe a non‐randomised multicentre trial involving 8 centres in which 63 patients were enrolled with single de novo lesions in a native coronary artery. In these patients, a total of 71 biodegradable magnesium stents were successfully implanted. The (composite) primary endpoint was the rate of major adverse cardiac events (MACE) defined as any one of: cardiac death, Q‐wave myocardial infarction or target lesion revascularisation at 4 months poststent implant. This was to be compared with an anticipated rate of 30%. They reported a rate of MACE at 4 months of 15/63 (23.8%); all of which were attributed to target lesion revascularisation (there were no deaths or Q‐wave myocardial infarctions) and concluded: