Implications of enrolment eligibility criteria in alcohol treatment outcome research: Generalisability and potential bias in 1-and 6-year outcomes

Introduction and Aims. It has been acknowledged that participants in clinical trials differ from real-world service users, primarily due to the extensive use of research eligibility criteria (EC). Generalisability and outcome bias become pressing issues when evidence-based treatment guidelines, crystallised from outcome research, influence treatment provision. This study reports on the effects of EC on generalisability and short- and long-term outcomes among real-world treatment-seekers. Design and Methods. Ten of the most commonly used EC were operationalised and applied to a large representative service user sample (n = 1125) from Stockholm County, Sweden, to determine the percentage of real-world problem alcohol users that would have been excluded by each EC and the extent to which EC bias the 1 and 6-year alcohol outcomes. Results. Individual EC excluded between 5% and 80% of real-world service users and 96% would have been excluded by at least one EC. Most of the EC introduced a positive/upwards bias in 1- and 6-year outcomes. Most notably, the removal of the unmotivated/non-compliant service users caused an upwards bias that would considerably boost estimates of treatment effectiveness. Other bias effects were smaller. Six-year effects were generally higher than for 1 year. Discussion and Conclusions. Outcome studies that exclude complex and non-compliant cases are not representative of real-world service users, and thus effectiveness estimates from clinical trials are biased by several commonly used EC. EC should be used judiciously and be taken into account in practice guidelines. This burgeoning research area should be further developed.