Combine Evidence, Keep Provenance

 Quick Analysis Plan Copied

 Long Analysis Plan

Core Evidence Gap: Hair EtG Accuracy Data Are Absent

The supplied evidence records do not contain hair-EtG sensitivity, specificity, or cutoff-based 2x2 data. The 2001 EtG review reports only that hair EtG was detectable in some alcohol misusers and social drinkers (n=14 hair samples) and notes that hair interpretation remains controversial with strong methodological heterogeneity, without any cutoff or accuracy estimates EtG biomarker review (Addiction Biology). The hair-methods study validates washing/extraction chemistry (DCM wash + methanol extraction, 30 min sonication; water best for EtG recovery) but is analytical validation, not diagnostic accuracy Hair EtG/FAEE washing and extraction validation.

The only cutoff-based diagnostic accuracy data supplied concern a different matrix: postmortem vitreous humor EtG immunoassay at 0.3 mg/L gave 92% sensitivity and 91% specificity (n=58), outperforming serum EtG (84%/88%) and BAC (68%/100%) for detecting ante-mortem alcohol use Vitreous humor EtG immunoassay validation. This is not a substitute for hair data and should not be pooled with it.

Caveat: this figure contrasts non-hair matrices only; it visualizes the diagnostic-accuracy format the hair literature lacks. Units are %, from one postmortem cohort, with reference standard based on case records (misclassification possible) VH-EtG study limitations.

Analysis Plan for the Hair-EtG Meta-Analysis

• Search & indexing: register PROSPERO protocol; search PubMed/Embase/Web of Science for hair EtG accuracy studies; extract cutoffs (e.g., pg/mg thresholds), 2x2 tables, reference standard, population (abstinence monitoring vs abusers), hair length/segment.
• Quality assessment: QUADAS-2 (patient selection, index test, reference standard, flow/timing); flag incorporation bias and partial verification.
• Statistical model: bivariate random-effects (Reitsma) or hierarchical HSROC in Python via statsmodels/pymc or R mada/metaDTA; pool sensitivity/specificity, fit SROC, compute AUC and likelihood ratios per cutoff.
• Threshold effect: explicit cutoff-stratified subgroups (low vs high pg/mg cutoffs); if threshold heterogeneity dominates, report per-cutoff estimates rather than one pooled pair.
• Pre-specified sensitivity analyses: LC-MS/MS vs immunoassay index test; chronic-drinker vs abstinence-monitoring populations; washing protocol (DCM pre-wash reduces external contamination) using the extraction findings above as covariates Pre-analytical protocol effects on hair EtG.
• Bias controls: funnel asymmetry tests for publication bias; leave-one-out influence; prediction intervals; report certainty via GRAAD-style downgrading for non-blinded, non-randomized designs.

What Would Change the Conclusion

If primary hair-EtG studies reporting 2x2 tables at stated pg/mg cutoffs (e.g., the widely used 7–30 pg/mg thresholds) become available and pass QUADAS-2, pooled estimates can replace this plan. Disconfirming evidence for hair-EtG validity would be: cutoff-dependent instability across washing protocols, or false positives from external contamination persisting after harmonized DCM pre-washes. Known unknowns: inter-individual UGT activity effects on EtG formation (UGT1A1/UGT2B7 mediate formation, ~1.6% of ethanol clearance) could drive false negatives, but no supplied record quantifies this in hair UGT-mediated EtG formation kinetics.