1. Randomised Controlled Trials: a workshop Ngaire Kerse Professor and General Practitioner and asker of questions.

2. Key issues - randomised trials There must be uncertainty The question must be answerable and feasible –Outcome reasonably common Examples of randomised trials in GP –Green Script, Health promotion Is it ethical to continue non-randomised studies that are evaluating interventions where there is uncertainty about effectiveness?

3. The beginning Uncontrolled trials –18th century Oranges and lemons, sailors and scurvy Smallpox, Queen Caroline, Newgate Prison and ‘charity children’ Controlled trials Agricultural trials (1920’s) Archie Cochrane - Salonica (1941) Vit C vs yeast for oedema Streptomycin for TB (1946)

4. Who’s afraid of the randomised trial and why? Unethical Explanatory (efficacy) vs pragmatic trials (effectiveness) Reluctance to pay – do it anyway

5. Stepping into the unknown Common sense interventions still need to be tested Kerse caused Falls JAGS 2004. Social work interventions –1930s Cambridge-Somerville Youth Study –1971 Blenker, 164 older persons

6. Why randomise? To decide whether an observed event is attributable to the meaningless play of chance or to causation (Silverman, 1980) Reduces bias –researchers unable to consciously/unconsciously load the intervention group with ‘good’ GPs, good communicators, high SES patient groups Evenly distribute known/unknown factors associated with certain outcomes

7. Where to randomise? Patient vs GP Individual GP vs Practice? Practice vs PHO? DHB vs Local Government area?

8. Analysis of findings Analysis should occur at the level of randomisation Are traditional analysis techniques appropriate? Randomisation by cluster accompanied by an analysis appropriate to randomisation by individual is and exercise in self-deception (Cornfield, 1978) Adjust for the design effect, increase sample size

9. Where to randomise? Individual Cluster 0 0.2 0.3 2.5 2.0 1.0 1.5 Inflation Factor Total contamination

10. Measuring - more than numbers? Use of validated instruments - SIP, EPNDS, SF36 Use of specific questions –open-ended, likert scales, categorical Self-report vs observed practice Observed practice - patient outcomes, simulated patient

11. Randomisation Randomly selected: 8 rest-homes, 4 private hospitals, 2 large complexes All residents 560 mean age 85 years Outcome Evaluation >2,000 falls Falls and injury Prevention pilot – FIPPS Control Group usual care Intervention Group Lo intensity intervention Manual, risk assessment and suggestions. Process Evaluation focus groups with staff

12. Randomisation (no stratification) 41 Rest-homes in Christchurch and Auckland Falls, function, QOL, 682, mean age 87 years. Outcome evaluation No Impr QOL Function (on average, signif subgroup) No increase in falls Promoting Independence in Residential Care Social Group 2 visits Activity Group PIRC, goal set, functional assessment, PIP to caregiver falls surveillance

13. Thinking differently RCT design does not preclude ‘evaluation’ process and impact ‘Qualitative’ methods can be incorporated into trials trial development - getting the intervention right making sure the intervention is working measuring the outcomes of the intervention

14. The big question….. Is it ethical to continue non-randomised studies in the general practice setting that are testing/evaluating interventions where there is a genuine uncertainty about the effectiveness?

15. PICOT BMC Med Res Methodol. 2010; 10: 11. Population, Intervention, Comparator, Outcome, Time-frame Everyone in the trial gets the same research process except for the intervention Is the research process greater than the intervention