Enrollment, Retention, and Adherence in a Fully Remote Pilot Randomized Controlled Trial of Alcohol Abstinence

The U.S. Dietary Guidelines for Americans recommend that adults of legal drinking age should consume no more than moderate amounts (1 and 2 drinks/day for women and men, respectively).¹ However, the effects of moderate alcohol consumption on cardiovascular disease remain unclear.² High-quality evidence from long-term randomized controlled trials (RCTs) comparing clinical outcomes from abstention versus moderate drinking is lacking.²

The costs and practical challenges of recruitment from brick-and-mortar facilities, as well as assuring participants’ retention and compliance, are impediments for conducting RCTs of alcohol abstinence. Volunteer recruitment from digital cohorts may be an efficient way to enroll participants. We established the Health eHeart study,³ a fully remote cohort comprised of volunteers interested in contributing to research regarding heart disease. Here, we aimed to assess a) the willingness to participate in an RCT of alcohol abstinence; b) participant retention throughout a 1-month trial period; and c) compliance with instructions.

The study was approved by the University of California, San Francisco Institutional Review Board. Health eHeart study participants were invited via email to indicate their interest in participating in a future long-term trial randomizing individuals to consume alcohol within current U.S. Dietary Guidelines limits or abstain from alcohol. Those expressing interest in participating in a one-month pilot RCT received a survey to identify participants who were ≥ 21 years of age, used a smartphone, drank at least 5 days of the week, and had an Alcohol Use Disorders Identification Test (AUDIT) score <20. The first 100 eligible participants were remotely consented and randomized 1:1 to either consume alcohol as they normally would (not to exceed 2 drinks in 24 hours) or abstain from all alcohol for 28 days.

Participants received baseline and weekly surveys about recent alcohol consumption (number of drinks per week). They were mailed 5 urine Ethyl Glucuronide (EtG) dip cards (Confirm BioSciences, San Diego, CA). On 5 randomly selected days, participants were texted instructions to complete an EtG card and send back a digital picture of the result within 24 hours. The trial outcome was adherence to alcohol consumption instructions assessed by self-report and EtG testing. To assess the agreement between self-report and urine EtG testing, assessments were dichotomized as any alcohol consumption or none, and as positive or negative, respectively. Descriptive statistics were employed. Two-tailed p-values <0.05 were considered statistically significant. The data are available from the corresponding author upon reasonable request.

Of 100,264 Health eHeart Study participants that opened an invitation email, 10,448 (10%) completed the willingness survey, and, of those, 6,113 (58.5%) were willing to participate in a years-long randomized trial of moderate alcohol consumption versus abstinence. Of those, 1,719 (28.1%) were interested in and eligible to participate in the pilot trial, and 100 (mean age 57.1 years, 51% females, 91% white) from 27 states (44% from California) were randomized.

Most participants completed and returned all 5 urine EtG tests (mean completion ± standard deviation percentage 82±27%), with no difference between arms. There was good compliance with the instructions to consume or abstain from alcohol as assessed by the presence of EtG in urine (Figure, Panel A), with more negative cards (89±3% vs 32±4%; p<0.001) and less positive cards (8±3% vs 65±5%; p<0.001) in the Abstain arm.

Figure. Alcohol consumption assessed by the presence of Ethyl Glucuronide in urine (Panel A) and by self-report (Panel B).

Panel A: Each row represents a participant. The results are categorized as technical error (if the participants did not receive the card or did not receive the alert to complete the urine test), incomplete (if the participant received the card and the alert, but did not complete the test), invalid (if the control line did not appear on the card), negative, or positive.

Panel B: Comparison of self-reported number of drinks per week between the abstain and consume arms at baseline and weeks 1 to 4, using the Mann-Whitney U test.

Completion of weekly surveys declined over time, from 100% at baseline to 91% at week 1, 91% at week 2, 82% at week 3, and 78% at week 4, but was similar between the two arms. Abstain arm participants reported significantly less consumption than those in the Consume arm throughout the study (Figure, Panel B). There was moderate agreement between urine EtG tests and self-reports of alcohol consumption (kappa 0.62, 95% confidence interval 0.50–0.69).

The large number of remote respondents exhibiting interest in participating in the proposed trial and the reasonably consistent adherence to randomization assignment instructions based on objective assessments in the pilot suggest that conducting a remote-based randomized trial of important lifestyle exposures may be feasible.

Because our baseline cohort was comprised of individuals already engaged in internet-based research, our findings should not be extrapolated to the general population. Although positive results of an EtG test are almost certainly true, a negative test does not exclude the possibility that a given person consumed some alcohol, especially at the limited doses studied here. EtG testing also only reflects alcohol consumption in the past 3 days.⁴ The moderate agreement between self-reported alcohol consumption and EtG testing is worth noting, recognizing the reference test is specific but not sensitive.

In conclusion, a large proportion of individuals participating in a digital cohort reported willingness to participate in a years-long RCT of moderate alcohol consumption versus abstinence. A 28-day pilot trial demonstrated the feasibility of remote monitoring and suggested good adherence to randomly assigned drinking assignments.

Availability of data and material:

The data are available upon reasonable request.