Introduction
Source of heterogeneity | Explanatory trial | Pragmatic trial |
---|---|---|
Patients and setting of included centres | May try to limit enrolment to academic centres and centres with a more homogeneous patient case mix. Usually have a large number of selection criteria, some of which may be designed to include a homogeneous group of patients more likely to respond to treatment, adhere to treatment and complete follow-up. | Deliberately include a variety of centres, possibly with a heterogeneous patient case mix. Usually have fewer selection criteria so as to deliberately include a diversity of patients who will qualify for the treatment in clinical practice. |
Intervention and control | Treatments are highly protocolised; enforce strict adherence to the protocol in each arm; co-interventions are not allowed or are limited and specified in detail; blind patients and providers to eliminate performance bias and subjectivity in assessment of outcomes. | Permit some tailoring while keeping core interventions common across all sites and for all participants; do not enforce adherence to protocol; permit co-interventions that would be used in target sites and settings after the intervention is shown to be effective; avoid using placebo; avoid blinding patients or providers. |
Outcome | Standardised outcome assessment; use central adjudication; blind outcome assessors to eliminate subjectivity in assessment of outcomes. | Favour objective outcomes, relevant for both patients and physicians; standardisation, adjudication and blinding of outcome assessors should be discouraged except if there is a risk of biasing the trial result. |
Regulatory and ethical issues | Requirement for written, informed consent from all participants; vulnerable participants (e.g. patients with co-morbidities or lacking decision-making capacity) commonly excluded. | In some cases, “clinical-style” or integrated consent may be permitted. Vulnerable participants generally included. |
Trial design | Trial conduct | Trial analysis |
---|---|---|
Sources: Centres and patients | ||
• Ensure that centres where results are intended to be applied are clearly defined. • Attempt to recruit centres typical of target population (rather than convenience sample). • Increase number of centres (possibly at cost of decreasing sample size per centre). • Ensure that patient populations to whom results are intended to apply are clearly defined and reflected in selection criteria, which should not be too restrictive. • Stratify by centre in randomisation, whenever possible. • Consider stratifying randomisation by other important prognostic factors. • Sample size may need to be increased to reflect pragmatic features, including potentially attenuated intervention effect and larger variance estimate. • In cluster randomised trials, ensure that intracluster correlation and cluster size variation are taken into account. | • Inclusion criteria should be applied consistently across all centres and investigators. | • Include stratification factors in analysis. • Limit number of planned subgroup analyses to those relevant to usual clinical decision-making. |
Sources: Intervention and control | ||
• Implement the trial such that, aside from the intervention, the trial has the lowest possible impact on patients and caregivers so as not to distort usual care conditions. • Consider likely extent of non-adherence, contamination, and co-interventions when specifying effect size for sample size calculation. | • Allow tailoring but maintain core intervention features that define the intervention being assessed. • Avoid interventions to monitor and promote compliance that may change patient behaviour unless they can be incorporated into future scale-up of intervention. • Assessment of compliance (by centres, providers, and patients) should be incorporated into intervention or assessed as a secondary outcome but measured in an unobtrusive way that does not interfere with compliance that would be expected outside the trial. | • Conduct analysis of superiority trials by intention-to-treat. • Ancillary process analyses can shed light on understanding mechanisms of observed treatment effects as long as data to facilitate such analyses are collected unobtrusively. |
Sources: Outcomes | ||
• Ideally, outcomes should be selected to be relevant to patients or trial stakeholders and assessed in usual care. • Avoid blinding outcome assessors except if there is a high risk of detection bias; select an outcome as objective as possible. • If needed, standardise how the outcome is assessed and adjudicate it. | • Sensitise data monitoring committees to the pragmatic nature of the trial to prevent undue intrusion through excessive data collection. | |
Sources: Regulatory and ethical issues | ||
• Plans for recruitment and informed consent must adhere to internationally accepted ethical principles even though this may affect pragmatism and bias. • Inclusion of vulnerable participants is appropriate provided plans are in place to identify and protect those who cannot provide informed consent and who may be at risk because of co-morbidities. • “Clinical-style” or integrated consent may be appropriate for recruitment in clinical settings when interventions involve usual care. • Waiver of consent may be requested from a research ethics committee when a cluster-level intervention poses only minimal risk. |
Patients and setting of included centres
Trial planning: select typical centres
Patients: Adults receiving invasive mechanical ventilation and vasoactive drugs for shock Centres: French intensive care units Intervention: Enteral nutrition Control: Parenteral nutrition Outcome: Day 28 all-cause mortality Design: Two parallel-groups, individually randomised trial | |
Centre selection | Both community and university care hospitals were recruited: “44 French ICUs, including 28 (64%) in university hospitals” |
Randomisation | “Randomisation was stratified by centre using permutation blocks of variable sizes” |
Participants: Women aged over age 16 Centres: General practices Intervention: Practice-based training sessions and pop-up template in electronic medical record Control: Usual care Outcome: Number of referrals Design: Two parallel-group, cluster randomised trial, clusters being general practices | |
Centre selection | “To ensure inclusion of practices with a range of characteristics, we stratified them by four characteristics (proportion of whole time equivalent female doctors, general practice postgraduate training status, number of patients registered with the practice, and percentage of the practice population on low incomes defined by the low income scheme index), then ordered them randomly within strata and invited them to participate in the trial sequentially within each strata by email or letter.” |
Variability in cluster (centre) size | “With 24 intervention practices and 24 control practices, with the assumption of an identification rate of 1% in control practices (a conservative estimate based on our survey of 12 east London practices) and an intracluster correlation coefficient of 0·03, we would be able to detect a difference of 5·2% in the identification rate with a power of 80% at a significance level of 0·05. This calculation assumed an average of 1600 women in the relevant age group in every practice, and took account of variation in cluster size.” |
Randomisation | “To ensure inclusion of practices with a range of characteristics, we stratified them by four characteristics (…).” “Within every primary care trust area we randomised practices with a computer minimisation programme, with a random component (Minim Version 1.3), maintaining allocation concealment. JR ran the minimisation programme for every practice after they were recruited and then informed the research associates of the allocation. The minimisation variables were the same as the stratification variables.” |
Statistical analysis | “Analysis was done for all practices for which we obtained baseline data, adjusted for minimisation factors (…).” |
Trial planning: relax patient selection criteria
Patients: Adults with thrice-weekly in-centre haemodialysis (see below) Centres: Dialysis facilities Intervention: Haemodialysis session duration of ≥ 4.25 h Control: Usual care Outcome: Mortality Design: Two parallel-group cluster randomised trial, clusters being haemodialysis facilities | |
Patient selection criteria | “Inclusion criteria for patients were (1) age ≥18 years, (2) treatment with thrice-weekly in-center hemodialysis, and (3) initiation of dialysis within the previous 120 days. Exclusion criteria for patients were (1) use of a health care proxy to provide consent for dialysis treatment and (2) unwillingness to have clinical data included in the trial dataset.” |
Randomisation | “Dialysis facilities were randomised 1:1 to the intervention or the usual care group, using a permuted block randomisation procedure with stratification by dialysis provider organization, and by factors known to be associated with mortality: racial composition (≤50% or >50% black patients) and use of central venous catheters for hemodialysis vascular access (≤20% or >20% of patients).” |
Compliance | “Participant follow-up ended on January 31, 2017 on the basis of the recommendation by the DSMB to terminate the trial because of a lower than anticipated difference in session duration between the intervention and usual care groups (…).” “For the primary analysis population, the estimated mean prescribed session duration was 219 (95% confidence interval [95% CI], 217 to 222) minutes in the intervention group and 210 (95% CI, 209 to 213) minutes in the usual care group.” “Discussions with facility staff and medical directors during the course of the trial indicated that the major reasons for poor uptake of the intervention were unwillingness by patients to have longer dialysis treatments, perception by the treating nephrologists that longer dialysis was not needed because of adequate solute clearance, and perception by the treating nephrologists that longer session durations were not in the best interest of a patient because of older age and/or frailty.” |
Trial planning: account for pragmatic features in sample size calculation
Trial planning: stratify randomisation
Patients: Both adults and children with symptoms of influenza-like illness Centres: Medical practices that were part of primary care research networks Intervention: Oseltamivir plus usual primary care Control: Usual primary care Outcome: Time to recovery Design: Two parallel-group individually randomised trial | |
Randomisation | “Stratified block randomisation was implemented, with random blocks of two, four, and six participants and stratification by age (< 12, 12–< 65, and ≥ 65 years), overall severity of influenza like illness (rated by the responsible clinician as mild, moderate, or severe), any relevant comorbidity (yes or no for heart disease, diabetes, chronic respiratory condition, hepatic, haematological, neurological, or neurodevelopmental condition, stroke or transient ischaemic attack, or overnight hospital stay in previous year), and previous duration of symptoms since onset (≤48 h or >48–72 h, based on recommendations that oseltamivir should be started within 48 h of symptom onset).” |
Blinding | “This was an open-label study, so no placebo was used and drugs were not masked.” “Some might consider the absence of a placebo control as a limitation. We deliberately chose to do an open-label trial in the context of everyday practice, because effect sizes identified by placebo-controlled, efficacy studies with tight inclusion criteria might not be reproduced in routine care. We also wished to estimate time to patient reported recovery from the addition of an antiviral agent to usual care rather than benefit from oseltamivir treatment compared with placebo. This pragmatic, open trial design makes our findings likely to reflect real world effects in primary care, because knowledge of what medication one is taking could affect subsequent help seeking and health behaviour and use of symptomatic medications. However, the design did not allow us to be sure of mechanisms or how much of the observed effect can be attributed to specific oseltamivir or other possible effects, and the relative contribution of such possible effects which might differ for the various subgroups.” |
Trial analysis: adjust on stratifying variables, notably centres (e.g. IRIS trial, Table 4)
Trial analysis: limit subgroup analyses to those that inform decision-making
Patients: Fetuses from women expected to deliver before 30 weeks of gestation Centres: 25 centres in seven countries Intervention: Delayed cord clamping Control: Immediate cord clamping Outcome: Composite outcome of death or major morbidity Design: Two parallel-group individually randomised trial | |
Sample size calculation: non-adherence | “The original sample was 1600 infants, yielding 90% power (two-sided P = 0.05) to detect an absolute difference in the incidence of the primary outcome of 8 percentage points between the two groups (30% in the immediate-clamping group vs. 22% in the delayed-clamping group; relative difference, 27%), with the assumption of 10% nonadherence. If the rate of nonadherence to the Intervention and loss to follow-up reached 20%, there was more than 80% power to detect this difference.” |
Subgroup analysis | “Tests for interaction were used to detect heterogeneity for the primary outcome in three prespecified subgroups: gestational age (<27 weeks vs. ≥27 weeks), sex, and method of delivery (cesarean section vs. vaginal delivery).” |
Intervention and control groups
Trial planning: permit some tailoring of the intervention
Patients: Care home residents aged ≥ 65 years Centres: Care homes from Coventry and Warwickshire and northeast London Intervention: Depression awareness programme delivered by physiotherapists, plus physical activity programme (see below) Control: Depression awareness programme for care home staff Outcome: Prevalence of depression Design: Two parallel-group cluster randomised trial, clusters being residential care homes | |
Intervention | “On the basis of the assessment the physiotherapist determined a plan of action for the intervention programme elements. The first was a bespoke physical activity programme tailored to each resident and aimed at increasing the level of habitual physical activity, developed in co-operation with the physical activity champion/senior carers. This included the provision of mobility aids, advice on footwear, and manual handling tips to enable mobility. The second was to determine the appropriate level of exercise activities for the group exercise programme.” |
Process analysis | “Alongside the main study we carried out a process evaluation and long-term follow-up using both qualitative and quantitative methodologies to explore the process of implementing the study in a care home setting to develop a set of transferable principles regarding both the OPERA depression awareness training and the OPERA ‘whole-home’ exercise intervention to inform its implementation on a wider scale. We did independent observations of the process of obtaining consent from participants. We did focus groups and interviews with key informants about the process of consent in care home studies.” |
Trial planning: ensure that the control intervention reflects usual care
Trial planning: consider the impact of compliance on sample size
Study conduct: do not enforce compliance
Study conduct: allow co-interventions
Trial analysis: apply the intent-to-treat principle
Patients: Adults with depression Centres: General medical practices Interventions: Acupuncture and counselling Control: Usual care Outcome: Depression prevalence assessed with the Patient Health Questionnaire 9 at 3 months Design: Three parallel-group individually randomised trial | |
Statistical analysis | “Multiple imputation by chained regression was used for missing data using treatment group, baseline measures (PHQ-9, BDI-II, SF-36, EQ-5D Anxiety/Depression), and demographics (age and gender). The primary analysis was based on the imputed rather than raw data in order to take account of the profile of non-responders.” |
Trial analysis: make sure ancillary studies will not interfere with not imposing specific constraints on patients or physicians
Outcome
Trial planning: select a routinely collected outcome regarded as important by clinicians and patients
Patients: Adults with ST-segment elevation myocardial infarction Centres: 29 Swedish centres and 1 Icelandic coronary intervention centre Intervention: Thrombus aspiration followed by percutaneous coronary intervention Control: Percutaneous coronary intervention Outcome: All-cause mortality at 30 days (see below) Design: Two parallel-group individually randomised trial within cohort | |
Outcome | “Data on mortality obtained from the national population registry”. “The concept of a trial design using a national registry as the basis for continuous enrolment and randomisation of all-comers is potentially limited by the lack of formal central adjudication of clinical events. Therefore, we have chosen all-cause mortality from the national complete mortality registry as the primary end point of the trial.” |
Trial planning: avoid standardisation, blinding and adjudication as much as possible
Patients: Adults surviving spontaneous intracerebral haemorrhage Centres: 122 hospitals in the UK Intervention: Antiplatelet therapy Control: Usual care Outcome: Recurrence of symptomatic intracerebral haemorrhage (see below) Design: Two parallel-group individually randomised trial | |
Outcome | “Although we did not mask the assigned treatment to participants and physicians, the outcomes were objective and adjudicated masked to treatment allocation, which minimises bias.” |
Trial conduct: sensitise data-monitoring committee to the pragmatic nature of the trial
Ethical and regulatory issues
Patients: Women treated for breast cancer, with a low to moderate risk of recurrence Centres: 2 UK centres Intervention: Telephone follow-up Control: Traditional hospital follow-up Outcome: Psychological morbidity assessed notably by the mean state-trait score Design: Two parallel-group individually randomised trial | |
Difference between patients who agreed or refused to be included | “Those who refused to take part differed from participants in study site, social class, and follow-up status. Patients at the specialist breast unit (71%) were more likely to want to participate than those at the district general hospital (61%, χ2 = 5.01, df = 1, P = 0.025), participants from higher social classes (professional occupations) were more likely to want to participate than those from lower social classes (χ2 = 15.77, df = 8, P = 0.046), and participants with three to 12 months between visits (67.7%, 70.6%) were more likely to participate than those on six monthly follow-up (58.1%, χ2 = 7.66, df = 2, P = 0.022). Time from diagnosis did not differ significantly for those who did or did not take part (t = − 0.26, P = 0.80); those who refused to take part were a median of 21 months from diagnosis.” |
Patients: Patients with an implanted medical device Centres: 28 Canadian centres Intervention: Incremental periprocedural antibiotics, i.e. pre-procedural cefazolin plus vancomycin, intraprocedural bacitracin pocket wash, and 2-day post-procedural oral cephalexin Control: Conventional periprocedural antibiotics, i.e. pre-procedural cefazolin infusion Outcome: 1-year hospitalisation for device infection Design: Cluster randomised cross-over trial | |
Ethical requirements | “All centers’ ethics boards approved the trial with waiver of consent for treatment. Ten centers required patient consent for data collection, which was generally obtained during follow-up.” |