Article type
Year
Abstract
Background: In a recent Cochrane review, we wished to quantify the effect of substitution treatment for opioid-dependent injecting drug users on the risk of HIV transmission. The extent of quantification was limited due to diversity in studies and the need to focus on before-after, rather than group, comparisons.
Objectives: To outline some of the methodological issues encountered in this review.
Methods: Eligible studies included opioid dependent injecting drug users; involved the oral administration of opioid drugs; and considered HIV risk behaviours or HIV infections in relation to substitution treatment. Outcome measures included injecting drug use, sharing of injecting equipment, frequency of condom use, multiple sexual partners, and providing sex in exchange for money or drugs.
Results: Included studies reported three types of data: 1) HIV risk behaviour prior to treatment entry, and at follow-up, after a period of substitution treatment; 2) HIV risk behaviour for participants receiving substitution treatment at the time of assessment compared to participants receiving no or limited substitution treatment; and 3) exposure to substitution treatment for cohorts who HIV seroconverted or remained HIV seronegative over a defined period. The statistical methods of RevMan were inappropriate for analysis of continuous measures of risk behaviours at baseline and follow-up as these methods assume independent samples. Studies varied in a number of aspects, including the interval between baseline and follow-up interviews; the proportion of participants injecting at baseline; the reporting period for assessment of HIV risk behaviours; and the means of reporting frequency data. This diversity, and the statistical difficulties in combining before-after data, prevented the calculation of combined measures of effect. Our approach was to list the findings of individual studies and consider the consistency of outcome.
Conclusions: Comparison of group data from randomised controlled trials is not always possible. Analysis of before-after data on behavioural change is complex. It would be helpful for the Cochrane Collaboration to provide greater guidance on possible approaches to outcome data that is based on before-after comparisons and non-randomised studies.
Objectives: To outline some of the methodological issues encountered in this review.
Methods: Eligible studies included opioid dependent injecting drug users; involved the oral administration of opioid drugs; and considered HIV risk behaviours or HIV infections in relation to substitution treatment. Outcome measures included injecting drug use, sharing of injecting equipment, frequency of condom use, multiple sexual partners, and providing sex in exchange for money or drugs.
Results: Included studies reported three types of data: 1) HIV risk behaviour prior to treatment entry, and at follow-up, after a period of substitution treatment; 2) HIV risk behaviour for participants receiving substitution treatment at the time of assessment compared to participants receiving no or limited substitution treatment; and 3) exposure to substitution treatment for cohorts who HIV seroconverted or remained HIV seronegative over a defined period. The statistical methods of RevMan were inappropriate for analysis of continuous measures of risk behaviours at baseline and follow-up as these methods assume independent samples. Studies varied in a number of aspects, including the interval between baseline and follow-up interviews; the proportion of participants injecting at baseline; the reporting period for assessment of HIV risk behaviours; and the means of reporting frequency data. This diversity, and the statistical difficulties in combining before-after data, prevented the calculation of combined measures of effect. Our approach was to list the findings of individual studies and consider the consistency of outcome.
Conclusions: Comparison of group data from randomised controlled trials is not always possible. Analysis of before-after data on behavioural change is complex. It would be helpful for the Cochrane Collaboration to provide greater guidance on possible approaches to outcome data that is based on before-after comparisons and non-randomised studies.