Exploring Phosphatidylethanol Cutoffs for Self-Reported Unhealthy Alcohol Use: An International Multi-Site Analysis

Background: Unhealthy alcohol use is a preventable cause of morbidity and mortality, yet screening is hampered by inaccurate reporting. Phosphatidylethanol (PEth) is a biomarker that quantifies total drinking over the past 2–4 weeks, but PEth cutoffs for unhealthy drinking have not been well-examined. Methods: We pooled data from 22 studies (11,088 persons globally) that previously collected PEth and self-reported alcohol use. Within a 90% training set, we calculated PEth cutoffs per Youden's J in 1000 bootstrapped samples and explored differences by region, age, sex, race/ethnicity, body mass index (BMI), HIV status, hemoglobin level, and an indirect serum marker of liver fibrosis, FIB-4. For each cutoff, we estimated sensitivity, specificity, and positive and negative predictive values in a 10% validation dataset. We used two definitions for self-reported unhealthy drinking per Alcohol Use Disorders Identification Test Consumption (AUDIT-C) and National Institute on Alcohol Abuse and Alcoholism (NIAAA). Results: Optimal PEth cutoffs using self-reported alcohol use as the reference standard differed substantially by region. The cutoff for AUDIT-C-measured unhealthy alcohol use in studies from the United States (US) was 14.0 ng/mL (95% CI: 12.3–18.6) with 73.0% sensitivity (95% CI: 67.3–79.1) and 77.4% specificity (95% CI: 73.2–81.4); and was 65.7 ng/mL (95% CI 19.3–90.7) in studies from Africa, with 71.7% sensitivity (95% CI: 64.4–78.7) and 65.2% specificity (95% CI: 58.1–72.7). Cutoffs for AUDIT-C did not differ between subgroups in the US, but within Africa, cutoffs were higher for men and lower for those with BMI ≥ 25 kg/m². Cutoffs for NIAAA defined unhealthy alcohol use were similar to those using the AUDIT-C. Conclusions: Using self-report as the reference standard, PEth cutoffs differed substantially by region and by some other characteristics, which may be attributable to differences in PEth formation, elimination and/or reporting bias. Further work using objective gold-standard measures of alcohol consumption is needed for more definitive conclusions.