Centre for Clinical Epidemiology & Biostatistics, School of Medicine & Public Health, University of Newcastle, Newcastle, Australia.
Deakin Biostatistics Unit, Faculty of Health, Deakin University, Melbourne, Australia.
J Clin Epidemiol. 2020 Feb;118:55-59. doi: 10.1016/j.jclinepi.2019.09.019. Epub 2019 Nov 7.
Nonresponse can bias prevalence estimates in population surveys. Effects of selective participation in behavior change intervention trials have been little studied. We tested hypotheses that trial participants who are hard to recruit are (1) more likely to be lost-to-follow-up and (2) less responsive to intervention.
We undertook a two-stage individual participant data meta-analysis of four alcohol intervention trials involving 9,251 university students in Australia, New Zealand, and Sweden, comparing participants who enrolled "late" (after the final invitation to participate) vs. "early" (before that). Outcomes were whether participants completed assessments at each trial's primary endpoint (late/early) and number of drinks consumed per week (intervention/control) among late enrolees vs. early enrolees.
Late enrolees were more likely to be lost-to-follow-up than early enrolees (OR 2.3, 95% CI: 1.7, 2.9). Intervention effect estimates were smaller for late vs. early enrolees, but not significantly so (RR = 0.93; 95% CI: 0.79, 1.08).
Greater effort to recruit trial participants was associated with higher attrition, but there was no clear evidence of bias in effect estimation. The possibility that intervention effect estimates do not generalize beyond a relatively compliant minority of trial participants may warrant further study.
无应答可能会使人群调查中的患病率估计产生偏差。行为改变干预试验中选择性参与的影响研究甚少。我们检验了以下假设：难以招募的试验参与者（1）更有可能失访，（2）对干预的反应性较低。
我们对涉及澳大利亚、新西兰和瑞典的四所大学的 9251 名学生的四项酒精干预试验进行了两阶段个体参与者数据荟萃分析，比较了“晚期”（最后一次参加试验的邀请后）和“早期”（在此之前）参与者。结果是在每个试验的主要终点（晚期/早期）中，参与者是否完成了评估，以及晚期入组者与早期入组者相比每周饮酒量（干预/对照）。
晚期入组者比早期入组者更有可能失访（OR 2.3，95%CI：1.7, 2.9）。晚期入组者与早期入组者相比，干预效果估计值较小，但差异无统计学意义（RR=0.93；95%CI：0.79, 1.08）。
更大的努力招募试验参与者与更高的脱落率相关，但在效果估计中没有明显的偏差证据。干预效果估计值是否不能推广到试验参与者中相对顺从的少数群体之外，这一可能性可能需要进一步研究。
