The background for this thesis is constituted by two cornerstones. First, the primary endpoint of interventions for treatment-seeking patients with alcohol use disorders (AUD) is alcohol consumption measured by self-report. Although self-report is considered to be an acceptable outcome measure, previous research suggests that there is a need for adding biomarkers of alcohol consumption as an outcome measure, since several studies have indicated that self-reports of alcohol consumption alone may not be able to identify all alcohol consumption.
Elderly patients are more vulnerable to alcohol exposure, have more comorbidities, and are more likely to suffer adverse effects of polypharmacy compared to younger patients, and it is therefore important to have accurate information on the elderly patients’ alcohol consumption in both clinical and research settings.
To date, very little is known about the extent of misidentification of alcohol consumption by means of self-reported alcohol consumption (SRAC) in elderly patients treated for AUD. Due to the above-mentioned reasons, it is pertinent to identify the extent of misidentification of alcohol consumption and to examine potential predictors of under-reporting.
This Phd thesis is based on three studies. The first study aimed to systematically review the empirical literature addressing the consistency between SRAC and biomarkers among treatment-seeking patients with AUD (and no illicit drug use). The results of the systematical literature review informed the design of studies 2 and 3. Study 2 aimed to assess the extent of under-reporting of alcohol consumption by means of hEtG in elderly patients after completion of treatment for AUD (and no illicit drug use). The aim of study 3 was to examine potential predictors of under-reporting of alcohol consumption in the same sample of elderly patients upon completion of AUD treatment.
Study 1: The systematic review was pre-registered in the International Prospective Register of Systematic Reviews (Prospero) (CRD42018105308) and conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement. Electronic databases i.e., the online version to the Medical Literature Analyses and Retrieval System (MEDLINE), the Psychological Information Database (PsycINFO), Excerpta Medica Database (EMBASE), the Cochrane Database of Systematic Reviews (CDSR), and the Cochrane Central Register of Controlled Trials (CENTRAL) were used to identify all original studies that aimed to examine the validity of SRAC by means of a biomarker in treatment-seeking patients with AUD. Outcomes in the studies that fulfilled the eligible criteria were included in a qualitative synthesis. The included studies were quality rated by means of the Quality Assessment Tool for Observational Cohort and Cross-sectional Studies, developed by the National Heart, Lung, and Blood Institute.
Studies 2 and 3: Data for both studies stem from the Elderly Study, a multisite, randomized controlled trial conducted in Germany, Denmark, and the United States during the period 2014 – 2016. Data for studies 2 and 3 were collected at the time of enrollment to treatment and at the follow-up interview (26 weeks after baseline). SRAC was assessed at both time points by means of the Form 90, and hair samples were collected at the follow-up interview. A total of 603 patients participated in the Elderly Study, and 544 were interviewed at follow-up. Of the 544 patients, 389 agreed to provide a hair sample.
The baseline characteristics of patients who provided a hair sample were compared to those who declined to provide a hair sample by means of the Mann-Whitney test, chi-square test, or gamma test. Based on their SRAC level and corresponding hEtG concentration, patients were grouped into one of five categories: (1) abstinence, (2) low consumption, (3) moderate consumption, (4) high consumption, and (5) excessive consumption. Rates of under- and over-reporting were calculated, and interrater reliability was reported as Cohen’s kappa. Associations between categories of hEtG and corresponding SRAC and effect modifications regarding natural hair color and body mass index (BMI) were examined by conditional logistic regression. Hair samples that 1) had been exposed to treatment, 2) were not of an optimal length or sufficient amount, or 3) were provided by patients with renal disease or patients who reported cannabis use, were excluded. Overall, 190 samples were included in the analyses in study 2. In study 3, only patients with hEtG > 4.9 pg/mg were included (n = 142) and predictors of under-reporting were examined by logistic regression. In both studies, p-values were corrected for multi-testing using the Bonferroni procedure.
Study 1: A total of 7672 hits were identified by the database search, and 11 articles based on 13 eligible studies were included in the systematic review. Inconsistencies between SRAC and biomarkers were found in all the included studies, the most common type being under-reporting. Regarding short-term biomarkers, seven studies (n = 15-585) identified under-reporting in a range from 5.5% to 56.0% of the patients, and two studies (n = 34-65) identified over-reporting in a range from 5.9% to 74.1%. Regarding intermediate-term biomarkers, two studies (n = 18-54) identified under-reporting in a range from 5.0% to 50% of the patients. Concerning long-term biomarkers, three studies (n = 73–1580) identified under-reporting in a range from 0.1% to 40.0% of the patients, and two studies (n=15-1580) found over-reporting in a range from 13.0% to 70.6%. Correlations between SRAC and intermediate-term biomarkers ranged from moderate to strong, and for short- and long-term biomarkers, correlations with SRAC were primarily weak. The quality assessment mostly suggested a fair quality of the included studies.
Study 2: The baseline characteristics of patients who provided a hair sample did not differ significantly from those who declined to provide a hair sample. Of the 34 patients who reported abstinence at follow-up, under-reporting was seen in 16 (47.1%). Of the 122 patients who reported consuming alcohol in a range from 0.1 to 59.9 g/day at follow-up, under-reporting was seen in 80 (65.6%). Kappa values varied between 0.19 and 0.36, indicating no or minimal level of agreement between hEtG and SRAC. Any amount of alcohol consumption was significantly more likely to be identified by means of hEtG than by self-report (OR 5.4 – 12.1, all p-values 0.05). The results indicated no significant effect modification of either natural hair color (defined as gray/white or non-gray/white hair) or BMI on the association between hEtG and exposure to alcohol (p ≥0.02).
Study 3: The findings showed that the average number of drinks per day reported during the 90 days prior to baseline was significantly negatively associated with under-reporting. For every unit of alcohol reported, the risk of under-reporting was reduced by 16% (adjusted OR 0.84, 95% CI 0.75 – 0.94, p-value 0.002). Also, there were indications that the risk of under-reporting of alcohol consumption increases with older age and number of drinks per drinking day reported during the 90 days prior to baseline, but the findings narrowly missed statistical significance.
Study 1: Inconsistencies between SRAC and biomarkers often occurred, and under-reporting was the most common type of inconsistency. The review also substantiated that the most optimal biomarkers are those that are non-oxidative compounds of alcohol metabolism, easy to collect, and cover the same timespan as the self-reported data.
Studies 2 and 3: In line with the results from the systematic review, study 2 found under-reporting to be the most frequently occurring type of inconsistency between SRAC and hEtG among elderly patients. The results underpin the importance of the abilities of hEtG in identifying alcohol consumption. Study 3 identified that an increasing average number of drinks per day 90 days prior to baseline reduces the risk of under-reporting.
Overall, the findings of this PhD thesis highlight the important role of clinicians and researchers in being aware of the potential risk of misidentifying the extent of the patients’ alcohol consumption as well as the need for using different outcome measures when assessing alcohol consumption.