« Previous | Contents | Next »
Listen
APPENDIX: Power calculation to inform field based study (extracted from protocol)
Design and Power Calculation
This is a cohort study comparing two geographically distinct areas with similar population demographics. The design has been drafted in association with advice from our statistician Dr G Taylor. The project will comprise two groups of participants , the study group (Dundee) who will receive the paraphernalia with safer injecting information and the control group (Aberdeen) who will receive standard needle exchange care. As users of both services are similar it is envisaged that the composition of the two groups (age, gender, length of time injecting) will be similar. Ideally people will be in the study for 6 months although it is anticipated that duration may be shorter for some. This will be accounted for in the statistical analysis.
Dr G Taylor's (medical statistician) assisted with the design and power calculations. As no previous work has been done to look at paraphernalia outcomes, a basis for guiding statistical calculation had to be constructed. The basis chosen was the incidence of abscess and/or soft tissue infections reported in IDUs in the literature, since it is hypothesised that the supply of sterile paraphernalia, accompanied with appropriate advice not to share, will reduce this incidence (more common and more studied than vascular problems). The prevalence of skin/soft tissue infection in IDUs in community based studies (comparable to the samples for this study), is reported as 33% [Spijkerman et al (1996). J Clin Epidemiology. 49(10):1149-54], 32% [Binswanger et al (2000) Clin Infect Dis. 30(3): 579-81] and 29% [Bassetti and Battegay. (2004). Infection. 32:163-169].
An approximate prediction of the expected outcome is needed to inform the power calculation. This is difficult due to lack of previous work. The closest comparison can be drawn from the Swiss heroin trials that supplied paraphernalia with pharmaceutical heroin (not street heroin as used by needle exchange clients). Conrad et al. (2000). Schweiz Rundsch Med Prax.89(46):1899-906 reports an incidence of 7% (cellulitis) and 5% (abscess), 18 months into treatment. An estimate from UK pharmaceutical heroin prescriber Dr Tom Carnwarth was given as 10% prevalence (Carnwarth,T. personal communication with J Scott, 2004). Pharmaceutical heroin will be 'cleaner' than street heroin which will be more bacteriologically contaminated so these may be underestimates of likely outcome. Therefore bearing these factors in mind, suggesting a 'best case' outcome of reduction in prevalence to approx. 10%, a more likely outcome of approx. 20% is estimated.
Based on 80% power and 95% confidence intervals:
A change from 33% prevalence to 10% prevalence would require 49 people per group.
A change from 33% prevalence to 20% prevalence would require 180 people per group.
To allow for drop outs this is rounded up to a target of 200 per group.
« Previous | Contents | Next »