1. Kris McGill, Jon Godwin, Catherine Sackley, Marian C Brady
Efficiency of recruitment to stroke rehabilitation RCTs Secondary analysis of recruitment data
Kris McGill, Jon Godwin, Catherine Sackley, Marian C Brady
CSO funded research department
Don’t mention supervisors

2. Stroke Incidence Rates
Over 152,000 people suffer from a stroke each year
Over 1.2 million stroke survivors in the UK
Survivors experience varying degrees of impairments and rehabilitation needs
Stroke Association. State of a Nation Stroke Statistics. 2016

3. Stroke Impairment Impairment % Upper limb weakness 77% Aphasia 33%
Lower limb weakness
72%
Slurred speech
50%
Visual problems
60%
Depression
Facial weakness
54%
Bowel control
Bladder control
Dementia
30%
Swallowing
These % are from soon after stroke. But many impairments continue to have an affect indefinitely.
Stroke Association. State of a Nation Stroke Statistics. 2016

4. Stroke Rehabilitation
Rehabilitation is key for recovery and return to life after stroke (Stroke unit Trialists’ Collaboration, 1997)
RCTs are used to test the effectiveness of rehabilitation interventions
RCTs adopt formal methods that reduce the chances of bias
RCTs have high internal validity
We all know that RCTs…
Stroke unit Trialists’ Collaboration. (1997). BMJ, 314(7088).

5. Recruitment
Successful recruitment to RCTs is essential for effective evaluation of treatment effects
Power: Adequate sample sizes essential for confidence in statistics
Research waste: inadequate recruitment key contributor
Poor estimations: Informed estimates of recruitment efficiency lead to more cost efficient trials

6. Haidich et al. BMC medical research methodology, 2001.
Foy et al. Family Practice, (1). McDonald et al. Trials, (1).
Treweek et al. BMJ open, (2).

7. Recruitment speed for stroke trials
Elkins et al (2006) recruitment of acute (up to 1 month after stroke) stroke survivors for pharmacological trials
Recruitment speed: 0.79 participants per site per month
Trials conducted in one country were more efficient than international trials
Elkins, J.S. Stroke, (1).

8. Research aims
Extent of the problem with recruitment rate, speed and dropout in stroke rehabilitation trials
Effects of trial features potentially associated with recruitment
CONSORT reporting standards

9. Methods Stroke RCTs (2005 to 2015) No language restrictions
Cochrane Stroke Trials Register
Inclusion criteria:
Stroke survivors only
Non-pharmacological
Rehabilitation interventions
- Reflective of clinical practice
- Administered by stroke rehabilitation team
- Targeting stroke related impairment
Cochrane stroke trials register comprehensive. Includes 35 electronic databases, most trial register, hand searching of grey literature

10. Data Extraction Country of recruitment Funding support
Year of publication
Ethical approval
Trial size
Type of intervention
Stroke survivor’s living context
Targeted impairment
Control condition
Recruitment strategy
Stage of stroke rehabilitation
Profession of the recruiter
Number of recruiting staff

11. Recruitment Efficiency
Rate of recruitment = 𝑛 𝑅𝑎𝑛𝑑𝑜𝑚𝑖𝑠𝑒𝑑 𝑛 𝑆𝑐𝑟𝑒𝑒𝑛𝑒𝑑
Speed of recruitment per site= ( 𝑛 𝑅𝑎𝑛𝑑𝑜𝑚𝑖𝑠𝑒𝑑 𝑅𝑒𝑐𝑟𝑢𝑖𝑡𝑚𝑒𝑛𝑡 𝑡𝑖𝑚𝑒 (𝑀𝑜𝑛𝑡ℎ𝑠) ) 𝑛 𝑆𝑖𝑡𝑒𝑠
Dropout rate = 𝑛 𝐷𝑟𝑜𝑝𝑜𝑢𝑡 𝑛 𝑅𝑎𝑛𝑑𝑜𝑚𝑖𝑠𝑒𝑑

12. Secondary Outcomes CONSORT diagram adherence
Coded with a traffic light system
Not reported
Partially reported
Reported

13. PRISMA Diagram

14. Countries Included (N=47)

15. CONSORT diagram reporting

16. Recruitment efficiency - participants, rate, speed, dropouts
RCTs =512
Mean (n) SD
min-max
Stroke survivors
Screened
321
320.88
536.87
Randomised
512
56.26
86.02
Mean
Recruitment
Time (months)
305
21.87
17.8
Rate
.40
.28
Speed
242
3.03
4.93
Dropout
414
.09
.11
Sites
363
2.55
5.69
1 - 71

17. RATE: Trial features that have had a significant effect
Kruskal- Wallis p
RCTs
Rate
Stroke survivor living context
X2(3)= 10.11
p= .018
239
Recruitment strategy
X2(2)= 10.34
p= .006
167
Number of recruiters
X2(2)= 6.06
p= .048
133
Stage of rehab
X2(5)= 16.82
p= .005
300

18. RATE: Trial features that have not had a significant results
Kruskal-Wallis p
RCT
RATE
Publication date
X2(1)= .1
p= .318
321
Continent of recruitment
X2(2)= 5.06
p= .08
283
Trial size
X2(4)= 5.5
p= .240
Profession of recruiter
X2(3)= 2.02
p= .568
171
Recruiters per site
X2(3)= 1.78
p= .620
116
Type of intervention
X2(1)= 2.19
p= .139
315
Funding support
X2(6)= 6.42
p= .378
257
Ethical approval
X2(2)= 1
p= .606
248
Targeted impairment
X2(3)= 5.62
p= .132
Control condition
X2(2)= .98
p= .612

20. (p= .035)
(p= .001)

21. (p= .025)

22. (p= .001)
(p= .003)
(p= .03)

23. SPEED: Trial features that have had a significant effect
Kruskal- Wallis p
RCTs
Speed
Trial size
X2(4)= 15.07
p= .005
242
Continent of recruitment
X2(2)= 24.21
p= > .001
283
Recruiters per site
X2(3)= 15.97
p= .001
122
Targeted impairment
X2(3)= 14.97
p= .002
241

24. SPEED: Trial features that have not had a significant results
Kruskal-Wallis p
RCT
SPEED
Publication date
X2(1)= .208
p= .648
242
Living context
X2(2)= 1.495
p= .474
221
Recruitment strategy
X2(2)= .49
p= .782
157
Profession of recruiter
X2(3)= 2.02
p= .232
151
Recruiters per site
X2(2)= 2.92
122
Type of intervention
X2(1)= .01
p= .917
241
Funding support
X2(6)= 9.97
p= .126
206
Ethical approval
193
Stage of stroke rehab
X2(5)= 9.98
p= .076
226
Control condition
X2(2)= 2.9
p= .235

25. (p= .008)
(p= >.001)
(p= .019)
(p= .008)

26. (Both p= >.001)

27. (p= .001)
(p= .008)
(p= >.001)

28. Dropout: Trial features that have had a significant effect
Dropout: Trial features that have had a significant effect
Trial feature
Kruskal- Wallis p
RCTs
Dropout
Publication date
X2(1)= 10.73
p= .001
414
Trial size
X2(4)= 25.38
Continent of recruitment
X2(2)= 11.91
p= .003
354
Recruitment Strategy
X2(2)= 6.09
p= .048
205

29. Dropout: Trial features that have not had a significant results
Kruskal-Wallis p
RCT
Dropout
Living context
X2(2)= .237
p= .888
344
Profession of recruiter
X2(3)= 3.62
p= .306
198
Number of recruiters
X2(2)= 1.34
p= .511
149
Recruiters per site
X2(3)= .98
p= .805
128
Type of intervention
X2(1)= .96
p= .327
395
Funding support
X2(6)= 6.14
p= .407
310
Ethical approval
X2(2)= 1.24
p= .537
309
Targeted impairment
X2(3)= 3.94
p= .268
396
Stage of stroke rehab
X2(5)= 9.41
p= .094
373
Control condition
X2(2)= 1.67
p= .433

30. Make sure and say published not conducted.

32. (p= .002)
(p= .005)

33. (p= .015)

34. RATE of recruitment
Higher recruitment rate for studies recruiting stroke survivors living at home
Higher recruitment rate when using 2 or 3 recruiters
Higher recruitment rate when screening chronic (6 months after stroke and above) stroke survivors

35. SPEED of recruitment Small trials recruit slower
One recruiter covering multiple sites appears much slower recruitment
Trials in Asia recruit faster

36. Dropout Trials published more recently experience higher dropout
Smaller trials retain participants more effectively
Less dropout in trials conducted in Asia
Higher dropout for studies screening admissions

37. Study strengths Extensive inclusion of stroke rehabilitation RCTs
Large sample sizes for analysis
Two authors screened and data extracted independently
High degree of trialists participation (>40%)

38. Limitations Reliant on authors descriptions of trial feature
Missing data due to lack of reporting of recruitment details
Despite limitations, this study will hopefully improve recruitment through more accurate estimations of recruitment Rate, Speed, and Dropout

39. Kris McGill Email: kris.mcgill@gcu.ac.uk Twitter: @krismcgil