Measuring lifetime alcohol exposure: A scoping review and implications for translational research and alcohol‐related harm

Abstract

Background and aims

Understanding lifetime alcohol exposure is fundamental to appreciating the risks of alcohol dependence and the multiplicity of alcohol‐related harms, such as cirrhosis. However, most research relies on self‐reported alcohol use measures that rarely extend beyond recent consumption. This represents a critical limitation in assessing the relationship between alcohol exposure and biological or psychological outcomes. Concerns about recall bias may be overstated and risk obstructing translational research into long‐term effects of alcohol, including treatment engagement. This scoping review examined the evidence for the reliability and validity of self‐reported measures of lifetime alcohol consumption.

Methods

We searched the MEDLINE, EMBASE and PsycINFO databases in June 2024 for English‐language articles published since 1970. We included studies that performed at least one of the following psychometric assessments on a lifetime alcohol exposure measure: test–retest reliability, internal consistency, concurrent validity or construct validity.

Results

The search identified 1607 unique records. After title and abstract screening, 24 studies underwent full‐text review, with nine meeting the inclusion criteria. Most studies were conducted in North America (78%). The most frequently used instrument was the Lifetime Drinking History (LDH) (67%). Across all studies, there were 6010 participants (3478 male, 2532 female), with sample sizes ranging from 49 to 3255. In 78% of studies, alcohol exposure data were collected via face‐to‐face interview. Eight studies assessed test–retest reliability, with retest intervals ranging from 14 days to 13 years; none assessed internal consistency. Three studies (33%) formally assessed validity. Where assessed, test–retest reliability and concurrent and construct validity were moderate to strong. Test–retest reliability ranged from 0.67 to 0.92, and concurrent and construct validity with external reference measures ranged from 0.40 to 0.80, indicating generally acceptable performance.

Conclusions

Existing instruments for measuring lifetime alcohol exposure appear to be valid and reliable and can overcome concerns regarding biases; however, these instruments vary in structure, lack standardisation and may fail to capture binge or episodic drinking, highlighting important gaps for refinement.

INTRODUCTION

A foundational principle in alcohol research is ensuring that the temporal assessment of an individual's alcohol use is done with valid and reliable methods [1]. Accurate assessment is essential in monitoring alcohol consumption patterns in populations over time, identifying hazardous drinking behaviour in the community and for individuals with alcohol use disorder (AUD), to assess the relative effectiveness of interventions and treatment [2]. Optimising the precision of estimated alcohol consumption is fundamental to investigating biological mechanisms of alcohol's effects and outcomes of harm [3].

Most extant research relies on drinkers recalling their alcohol use and reporting the average number of standard drinks/g of ethanol consumed per day. This measure is often used to assign individuals to specific treatment approaches in clinical practice, as well as in clinical trials, to predict alcohol cessation outcomes, and as a criterion for research study inclusion. Importantly, exposure over time is used to help predict risk for physical, biological and/or histological outcomes [4].

Alcohol consumption is a complex behaviour. The volume and pattern of consumption, even for an individual, can be heterogeneous over time, making retrospective recall problematic. Refining assessments of alcohol exposure is vital to better define differences in risk factors for an individual's current and historical drinking patterns and to investigate associated morbidity and mortality. The ‘pack‐year’ variable, which is a measure of cigarette tobacco exposure over time, predicts smoking‐related complications such as obstructive airways disease and lung cancer and exemplifies the utility of quantifying cumulative exposure over the lifetime [5]. There is no equivalent alcohol exposure variable. Importantly, a lack of research on the validity and reliability of self‐reported measures of cumulative alcohol exposure undermines our ability to assess the relationship of alcohol exposure to associated pathology, morbidity and mortality [3]. Despite its importance, few instruments are specifically designed to measure lifetime alcohol intake. The lifetime drinking history (LDH), cognitive LDH and Concordia LDH are the only published instruments measuring lifetime alcohol exposure [6, 7, 8].

Valid and reliable measurement of lifetime alcohol exposure is essential for understanding alcohol's role in disease development, particularly in translational and cohort‐based research [1, 2, 3, 9]. In liver disease, frameworks like metabolic dysfunction‐alcoholic liver disease (MET‐ALD) depend on historical community safe‐alcohol thresholds to distinguish between metabolic and alcohol‐predominant cases [10, 11].

Questions remain about the nature of exposure over time. It remains unclear whether recent or cumulative exposure is the stronger predictor of harm [12]. A short duration of very heavy drinking, versus a longer period of moderately heavy drinking, may both give a similar ‘average’ alcohol exposure, but clearly these exposures carry different sorts of risk. Similarly, heavy alcohol exposure may have different impacts at different ages. Younger people may have more physiological reserve, while comorbidity is associated with aging. This phenomenon also underlies the importance of lifetime measures to provide context for understanding abstainer bias, where current abstinence may be misinterpreted without accounting for past harmful use, therefore, understating prior exposure‐harm relationships, which can be confounded as a protective alcohol effect [13, 14]. Recall bias, underestimation and stigma all challenge accurate reporting, and no universally accepted gold standard currently exists [15].

Assessment of the reliability and validity of instruments used to measure lifetime alcohol exposure is a research and clinical gap. This article presents a scoping review of the literature on instruments used to measure lifetime alcohol exposure, with a specific focus on assessing their reliability and validity.

METHODS

Protocol

This review followed the Preferred Reporting Items for Systematic Reviews and Meta‐Analyses for Scoping Reviews (PRISMA‐ScR) guideline, and the protocol adhered to the PRISMA‐ScR protocol [16]. The review protocol was initially submitted to PROSPERO in June 2024. However, it was deemed ineligible because it was a scoping review. The protocol was subsequently registered with the Open Science Framework in September 2024 with the digital object identifier https://doi.org/10.17605/OSF.IO/AC2U9.

Included studies

We included studies that assessed at least one psychometric property, either reliability (e.g. test–retest reliability) or validity (e.g. construct validity or concurrent validity), of a lifetime alcohol exposure instrument. Studies were not required to report both reliability and validity for inclusion.

Data sources and searches

We searched MEDLINE, Embase and PsycINFO from 1970 to June 2024 using MeSH and keyword terms. The 1970 start point reflects the earliest time point when structured instruments for assessing lifetime alcohol exposure first appeared in the literature. The search for relevant citations was performed using subject headings (MeSH, Thesaurus and EMTREE) and synonym terms: (((alcohol) ADJ4 (consumption OR ‘use’)).tw. OR drinking.tw. OR exp “Alcohol Drinking”/) AND (lifetime.tw. OR life‐event*.tw.) AND (measurement*.tw. OR self‐report*.tw. OR indices.tw.) OR (Lifetime Drinking History.tw. OR Lifetime Drinking Questionnaire.tw.) AND (exp “Sensitivity and Specificity”/ OR exp “Diagnostic Errors”/ OR sensitive.mp. OR sensitivity.mp. OR specificity.mp. OR accurate.mp. OR accuracy.mp. OR “golden standard” OR “gold standard” OR (reference.mp. AND [test.mp. OR standard.mp.]) OR “index test” OR validity.mp. OR validation.mp. OR validate*.mp. OR valid.ti. OR “validation studies”.pt. OR verif*.ti. OR “evaluation studies”.pt. OR evaluat*.ti. OR (false.mp. AND [positive.mp. OR negative.mp.]) OR pretest.mp. OR pre‐test.mp. OR posttest.mp. OR post‐test.mp. OR “predictive value” OR predict*.ti. OR roc.mp. OR likelyhood.mp. OR likelihood.mp. OR value*.ti. OR exp “reference values”/ OR cutoff.mp. OR cut‐off.mp. OR exp “quality control”/ OR exp “reproducibility of results”/OR repeatability.mp. OR reproducibility.mp. OR efficacy.mp. OR reliability.mp. OR “comparative study”.pt. OR odds.mp. OR error*.mp. OR suitability.mp. OR utility.mp.).

Study selection

After duplicate removal in Covidence, four reviewers (A.J.P., J.Y., K.R. and A.H.) screened titles and abstracts against inclusion criteria. Two reviewers (A.J.P., J.C.) assessed the full texts and documented the reasons for exclusion. Disagreements were resolved by a third reviewer (P.J.C.).

Data extraction

Data extraction was performed independently (A.J.P., J.C.) in duplicate after full‐text screening. Data extracted included author information, study design, publication year, whether the population was treatment seeking and/or had an AUD diagnosis, age, sex (as all included studies preceded contemporary distinctions between sex and gender identity, the term ‘sex’ is used consistently throughout this review to reflect available demographic data), where available more detailed population description (e.g. type of employment, marital status), method used to gather drinking history (e.g. face‐to‐face interview, questionnaire), outcomes [including but not limited to quantity, frequency (QF), QF index (QFI) (quantity multiplied by frequency), total alcohol consumed, phases of drinking across the lifetime, times intoxicated and age first consumed alcohol] and assessment of at least one of reliability (test–retest and/or internal reliability) and validity (concurrent validity and/or construct validity). We interpreted Pearson or Spearman correlation coefficients as follows: <0.30 = weak, 0.30–0.59 = moderate and ≥0.60 = strong [17]. Where available in the extracted manuscripts' original data, acceptable thresholds for test–retest reliability were 0.7, concurrent criterion‐related validity 0.7 [18], content (assessed qualitatively on criteria: relevance, comprehensiveness and comprehensibility) [19] and construct validity (if applying confirmatory factor analyses, confirmatory fit indices >0.9; root mean square error of approximation, 0.05–0.08) [20].

RESULTS

We identified 1964 records, with 357 duplicates removed. After screening 1582 titles/abstracts, 25 full texts were reviewed. Sixteen were excluded (no lifetime measure = 10; reviews = 3; conference abstract = 1; insufficient validity/reliability data = 1; duplicate = 1) [21]. Citation searching yielded no additional studies. Nine studies met the inclusion criteria (Figure 1) [6, 7, 8, 22, 23, 24, 25, 26, 27].

FIGURE 1.

Preferred reporting items for systematic review and meta‐analysis (PRISMA) diagram for the systematic review.

All instruments aimed to quantify cumulative alcohol exposure, but differed in terms of recall anchors, time frame segmentation and detail. Table 1 summarises features such as period definition (fixed vs. flexible), data fields (e.g. quantity, beverage type), use of visual/cognitive aids and ability to assess heavy episodic drinking (HED).

TABLE 1.

Comparison of lifetime alcohol exposure instruments: structural features, data captured and anchoring methods.

Instrument	Timeframe structure	Data collected	Captures heavy episodic drinking	Quantifies lifetime volume	Captures change over time	Administration mode	Anchoring method
LDH.	Fixed decade blocks or user‐defined phases anchored to life events	Usual quantity, frequency, max quantity per phase, beverage type and abstinence periods	Partial‐ maximum quantity per phase gives an upper bound but lacks the frequency of heavy use	Yes: total volume estimated from quantity/frequency years in each phase	Yes: maps distinct drinking periods with linked reasons for change	Interview or self‐administered	Verbally prompted to recall distinct drinking phases anchored to significant life events such as relationships, employment or health changes. No visual aids or timelines were used; anchoring relied on interviewer‐facilitated reflection and recall.
Cognitive LDH.	Fixed decades or cognitively anchored floating intervals	Drink size, beverage‐specific quantity/frequency, no. of intoxication episodes and contextual anchors	Strong proxy: explicitly captures the no. of intoxication episodes per interval	Yes: cumulative grams of alcohol derived across intervals	Yes: designed to capture shifts in pattern using cognitive cues	Structured interview	Drinking phases were anchored using autobiographical and contextual memory prompts, such as personal milestones or life transitions. Although the interviewer cognitively scaffolded recall, no visual timeline was used. The method relied on respondent‐defined drinking intervals structured through guided narrative reconstruction.
Concordia LDH.	User‐defined phases are mapped to 5 life periods	Average quantity, frequency, beverage type and periods of abstinence	Weak: binge episodes are not consistently captured, relying instead on average consumption	Yes: total intake estimated, although less granular than LDH	Partially: assumes life‐stage drinking patterns but may miss nuance	Self‐administered questionnaire	Participants were asked to reflect on their life course and drinking patterns before the interview. During the session, the interviewer constructed an illustrated timeline divided into age‐based life periods based on participant responses. This visual tool was then used to support recall and organise data collection across the structured questionnaire.

Study characteristics

All studies were cross‐sectional with three nested in cohorts [24, 25, 26]. Seven came from North America, one each from Australia and the Netherlands. Five examined the LDH [6, 22, 24, 25, 26], one each the cognitive LDH [8], a modified LDH [23] and Concordia LDH [7]. One excluded participants with possible alcohol dependence [27], and two sampled treatment‐seeking patients [6, 22].

The included studies mostly included middle‐age white adults. Cohort size ranged from 49 to 3255. Three studies enrolled only males [7, 25, 26] and one female participants [23]. Of the five mixed‐sex studies, four had more males [6, 8, 22, 24] and one more females [27]. Seven used face‐to‐face interviews, one self‐administered and one phone/face‐to‐face. Two likely used overlapping cohorts, with one focusing on reliability [25] and the other on validity [26].

The test–retest period varied from 14 days to 13 years. Six of the studies had a test–retest period greater than 12 months [7, 22, 23, 24, 25, 26]. Eight of the nine studies reported on test–retest reliability [6, 7, 8, 22, 23, 24, 25, 27]. None assessed internal consistency. Three studies made formal assessments of validity [7, 24, 26]. Two assessed concurrent validity [7, 26]. All three assessed construct validity [7, 24, 26]. Two additional papers provided indirect support for validity, although they did not formally assess it [6, 25].

In summary, participants across all included studies represented samples from the general community of treatment‐seeking individuals with alcohol dependence and clinical cohorts. Most studies included both male and female participants, however, few studies stratified psychometric results by sex. No studies reported subgroup performance by age, race/ethnicity or gender identity. Consequently, although the instruments appeared broadly applicable across settings, their performance across demographic subgroups remains underexplored. Table 2 summarises the characteristics of the included studies.

TABLE 2.

Characteristics of the included studies.

Reference	Measure	Country	Population	Treatment seeking	Setting	Sample size	Age	Sex (% female)	Participant description	Time between tests	Interview type
Jacob et al. [25]	LDH	US	AD	No	Community	49	40.8 (SD = 6.8)	0	Employed in skilled or semi‐skilled occupations and had a mean education attainment of 12.4 years (SD = 2.1). All participants were married, with an average marriage length of 16.3 years (SD = 5.6) and a mean of 2.6 children (SD = 1.4)	5‐years	Face‐to‐face
Russel et al. [8]	CLDH	US	MI patients and controls (licensed drivers, no MI), all having consumed >12 drinks in a 12‐month period.	No	University	81 (MI); 138 (controls)	MI: 56.44 (9.04, 35–70). Control: 60.75, (7.33, 33–70)	15 post MI and 43 controls	Post‐MI participants were younger, predominantly male and had a slightly lower proportion of non‐Hispanic White individuals than controls, with no significant differences in education or marital status	At least 1 week, with some participants tested after more than 2 weeks	Computer‐assisted face‐to‐face
Skinner et al. [6]	LDH	Canada	Participants had alcohol‐related problems as assessed using the MAST (mean score = 24.5)	Yes	Outpatient addiction medicine clinic	83	33 (SD = 11, range = 16–58)	12	Mean of 11.7 years of education (SD = 2.7), with 35.2% having post‐secondary education. Marital status included 45.8% single, 37.3% married/de facto and 16.8% separated, divorced or widowed. Employment varied, with 45.7% in unskilled/skilled jobs, 19% in managerial/professional roles and 16.9% in clerical positions	Mean of 4.8 months (±SD = 2.6 months).	Face‐to‐face
Melo et al. [7]	Modified CLD	Australia	Wine drinkers without a history of alcohol problems.	No	Community‐ recruited through a newspaper advert	51	58.1 (SD = 12.4)	54.9	Participants were 21+ years old, drank wine weekly and agreed to 2 interviews. Those with potential alcohol dependency were excluded via CAGE screening	4‐weeks	Face‐to‐face
Sobell et al. [22]	LDH	Canada	AD	Yes	Specialist addiction unit: both inpatient and outpatient	69	41.6 (SD = 9.5)	10.1	94% White, with a mean of 11.3 years of education (SD = 2.8). 30% were married, 71% employed and 78% worked in blue‐collar jobs. The mean duration of AD was 10.9 years (SD = 9.4)	16.8 ± 3.4 day retest period (median = 15, range = 13–29 days).	Face‐to‐face
Friesema et al. [24]	LDH	Holland	Dutch Lifestyle and Health Study	No	34 general practices throughout Holland	3255	Age at baseline (45–71 years, born 1925–1951)	46	The study sampled 3255 individuals from an initial cohort of 16 211 Dutch adults (≥45 years), drawn from the south‐eastern and western Netherlands to account for regional differences in drinking habits. Follow‐up response rate: 77.2%	1‐year	Self‐administered questionnaire
Harris et al. [23]	Modified LDH	US	US National Survey on Women's drinking	No	National survey	917	Participants were categorized into younger [6, 7, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32] and older (35+) problem and non‐problem drinkers	100	The study surveyed 917 women in 1981, with a 1986 follow‐up of 143 problem drinkers (≥2 indicators of problem drinking) and 157 non‐problem drinkers (regular drinkers without issues). The sample included moderate‐to‐heavy drinkers (≥4 drinks/week), with high follow‐up rates (94% for problem drinkers and 93% for non‐problem drinkers)	5‐years	Face‐to‐face interview
Chaikelson et al. [7]	Concordia LDH	Canada	World War II army veterans	No	Participants were from Montreal and part of an adult functioning study	72	69 (4.28, 61–85)	0	The sample consisted of individuals with a skilled working‐class background, with most participants being married. 46 wives provided collateral reports on their husbands' drinking. Mean education level: 10 years (SD = 13.3)	33.1 ± 4.2 months (range = 26–40 months)	Face‐to‐face interview
Koenig et al. [26]	LDH	US	Vietnam War veterans	No	Family Twin Study of twins and their children with AD parents	1295 twins completed LDH, 420 had lifetime AD	In the full sample, 38.0 (SD = 2.6) in 1987; 40.3 (SD = 2.6) in 1990, 41.6 (SD = 2.5) in 1992; 45.8 (SD = 2.5) in 1995; and 50.9 (SD = 2.8) in 2000	0	Participants were selected for their alcohol use or parental history and enrolled in a longitudinal study using prospective and retrospective data. The sample included regular drinkers and lifetime ADs, recruited from a community setting for comparison of drinking patterns	The study compared retrospective LDH data from 2000 with prospective quantity‐frequency data collected in 1987, 1990, 1992 and 1995	Primarily conducted through in‐person interviews. Telephone interviews were also used

Abbreviations: AD, alcohol dependence; CAGE, Cut down, Annoyed, Guilty, Eye‐opener; score ≥2 suggests possible alcohol misuse; CLDH, cognitive lifetime drinking history; LDH, Lifetime drinking history; MAST, Michigan Alcoholism Screening Tool; MI, myocardial infarction; US, United States.

Test–retest reliability

Lifetime drinking history

Five studies assessed test–retest reliability of the LDH, including an LDH self‐administered questionnaire (LDH‐q).

The first study focused on evaluating the reliability of LDH over a 5‐year test–retest period in a sample of non‐treatment‐seeking, alcohol dependent patients [25]. Across two assessments, the study found a significant positive correlation (r = 0.68) for total drinking volume. The total number of drinking phases showed moderate reliability, with a correlation coefficient of r = 0.37, and correlations for specific measures ranged from r = 0.31 to r = 0.52. Initial drinking phase variables were moderately consistent, including QFI r = 0.40 and the proportion of alcohol consumed by type r = 0.59 for beer and r = 0.64 for spirits. Final drinking phase measures were also reliable, with correlations of r = 0.48 for drinking alone and r = 0.52 for drinking at home.

The second study included 83 treatment seeking individuals with alcohol‐related problems [6]. Participants were assessed on two occasions, approximately 4.8 months apart. Recall of drinking behaviour at age 40 was relatively consistent between these time points, with high test–retest correlations, as indicated by r = 0.71 for frequency, r = 0.57 for quantity and r = 0.65 for QFI. Recall of duration of drinking was reliable (r = 0.94), followed by total volume consumed (r = 0.80), daily average consumption (r = 0.68) and quantity per drinking session (r = 0.57). Beverage type‐specific measures ranged from r = 0.62 to r = 0.70.

The third study examined the reliability of retrospective recall for drinking during a specified year‐long period, resulting in a recall duration that ranged from 7.5 to 8 years [22]. Data captured included the number of days any alcohol was consumed, the total number of drinks consumed, the number of drinks per drinking day, the greatest number of standard drinks on any single day and the longest period of abstinence. Examining frequency‐related measures revealed correlations of r = 0.81 for days any alcohol was consumed, r = 0.67 for QFI, r = 0.66 for total drinks and r = 0.65 for the most significant number of drinks on a single day. Limited drinking days showed correlations of r = 0.53 (1–4 drinks) and r = 0.73 (1–6 drinks). Age of onset of drinking showed moderate reliability (r = 0.50) with 66.3% of younger problem drinkers and 51% to 54% of others reporting within 1 year of their initial response.

The fourth study assessed test–retest for the LDH. A group of female participants surveyed for alcohol use in 1981 were approached 5 years later and asked to recall their drinking [23]. Agreements between the two measures were relatively consistent. For typical drinking patterns, Kendall's τ‐b and κ ranged from 0.26 to 0.49 for frequency and 0.34 to 0.47 for quantity per drinking day, with 60% to 88% of participants reporting consistent patterns. Correlations for assessing maximum frequency and quantity were lower in younger participants (r = 0.17–0.39) and higher in older participants (r = 0.35–0.58).

The final study examining test–retest reliability assessed 3255 participants in a Dutch lifestyle and health cohort study [24]. A self‐administered LDH‐q assessment was conducted twice, 1 year apart, with a 77.2% follow‐up response rate. The reliability of self‐reported drinking behaviours, assessed via intra‐class correlation coefficients (ICCs) and varied by sex and life stage. For total lifetime alcohol intake, ICCs were 0.75 for men and 0.70 for women, while for age of onset, they were 0.21 for men and 0.52 for women. The reliability of drinking frequency recall appeared to improve with age: during youth (12–18 years), ICCs were 0.48 for men and 0.42 for women; in young adulthood (19–27 years), 0.64 for men and 0.53 for women; in adulthood (28–45 years), 0.72 for men and 0.75 for women; in middle age (46–60 years), 0.78 for men and 0.79 for women; and for participants 61 and older, 0.80 for men and 0.77 for women.

The reliability of alcohol quantity per occasion varied. During youth, ICCs were 0.68 for men and 0.56 for women; in young adulthood, 0.75 for men and 0.62 for women; in adulthood, 0.72 for men and 0.57 for women; in middle age, 0.72 for men and 0.54 for women; and for participants 61 and older, 0.77 for men and 0.50 for women.

The QFI showed increasing reliability with age. Youth ICCs were 0.44 for men and 0.50 for women, rising in young adulthood to 0.74 for men and 0.66 for women; in adulthood, they increased to 0.74 for men and 0.83 for women; in middle age, they remained at 0.73 for men and 0.79 for women; and finally, they reached 0.88 for both sexes among those 61 and older.

Cognitive LDH

The cognitive LDH's test–retest reliability was evaluated in two studies. The index study examining cognitive LDH indicated high reliability for self‐reported lifetime alcohol consumption, with Spearman test–retest correlations for lifetime alcohol volume, from 0.74 to 0.85 [8]. Pearson's correlations also indicated a stable level of alcohol intake. Average lifetime alcohol consumption remained stable between visits, supporting the consistency of participants' responses.

For recall of the data on ‘times intoxicated’, recall appeared accurate between visits, as indicated by Spearman's correlation coefficients ranging from 0.76 to 0.92. The fixed and floating measurement tools showed similar reliability, with no meaningful differences in their test–retest correlations.

The second cognitive LDH study revealed high reliability for drinking behaviour phases, with ICCs of 0.856 for phases, 0.987 for ‘years of drinking’ recall and 0.930 for ‘average phase duration’ [27]. None of these measures showed significant differences across time points (non‐significant F‐tests). More specific data on self‐reported alcohol consumption also showed reliability across various measures of consumption, including frequency, quantity and type of alcohol consumed, with ICC values ranging from 0.660 to 0.991.

Concordia LDH

One study, using data collected in 1983 and published in 1994, assessed the test–retest reliability of the Concordia LDH, reporting correlations of 0.75 for quantity and 0.83 for duration over a median 28‐day interval (IQR = 21–37 days) [7]. Drinking habits before 1963 (a recall period of ≥27 years) exhibited a correlation of r = 0.68, while a slightly more reliable correlation of r = 0.78 was observed for drinking habits since 1963. The correlation for the most recent drinking month was r = 0.84. Similarly, for the current month's drinking among active drinkers, the correlation was r = 0.77. These findings are consistent with the anticipated result that more recent drinking behaviour measures are more reliable.

Moderate reliability was observed in milestones related to drinking patterns (e.g. school, marriage and divorce). The correlation between age and the first drink was r = 0.66, while the correlation between age and the start of regular drinking was lower at r = 0.56. Test–retest correlations, calculated every 5 years between 1945 and 1985, ranged from 0.65 to 0.87, showing higher reliability with more recent recall.

Internal reliability

No studies of lifetime alcohol consumption measures reported on internal reliability.

Validity

Concurrent validity

The LDH‐q and Concordia LDH assessed concurrent validity [7, 24]. For the LDH‐q, mean alcohol intake in 5‐year intervals for the cohort (n = 3255) for the years 1950 to 1995 was calculated for comparison with per‐capita intake for these years in Holland [24]. The averages, in litres of alcohol consumed per capita, in the Dutch population and the cohort were similar, and they increased equally through the years. Age‐specific intakes, plotted against per capita intake, showed the same trends.

For the Concordia LDH, collateral history from the participant's spouse was compared to the individual's self‐report [7]. Retrospective reports of their spouse's drinking at the time of marriage strongly correlated with their participants' self‐report (r = 0.73). However, the correlation was slightly weaker than for spouse‐spouse agreement of current drinking. To rule out the possibility that the spouse was estimating past drinking based on current patterns, researchers examined the correlation between husbands' reports of their drinking at the two time points, which was low and not significant (r = 0.24), suggesting the spouse's report is valid. One further study of the LDH did not formally assess concurrent validity, but provided indirect support by demonstrating consistency with the Michigan Alcoholism Screening Tool (MAST), an established screening tool for AUD [6].

Construct validity

The LDH, LDH‐q and Concordia LDH each had one study evaluating construct validity. The study compared the LDH using retrospective data from 2000 with prospective QFI data from 1987, 1990, 1992 and 1995 [26]. Retrospective QFI scores were generally higher than prospective ones, particularly among alcohol‐dependent individuals. For the overall sample, significant differences were observed between 1987 and 1995 (Wilcoxon test, P < 0.001), with medians of 7.0 and 8.0 (1987) and 5.0 and 6.0 (1995), respectively. Among alcohol‐dependent drinkers, discrepancies were significant across all years, with the largest gap in 1987 (median QFI = 17.2 prospective vs. 30.0 retrospective).

Retrospective reports overestimated alcohol quantity, particularly among alcohol‐dependent individuals (mean QFI = 43.7 prospective vs. 72.9 retrospective, P < 0.001). In comparison, differences in frequency were more minor and less consistent (mean days/month: 10.5 prospective vs. 10.6 retrospective). A linear mixed‐effects model with random intercepts was used to assess test–retest reliability over time. This model confirmed significant differences in reported quantity (P < 0.001), but only minimal differences in frequency. Correlations between retrospective and prospective assessments demonstrated rank‐order stability, with coefficients ranging from 0.57 to 0.69 for alcohol‐dependent individuals and from 0.60 to 0.69 for the overall cohort (P < 0.01). Agreement improved for more recent data (1995 QFI correlation = 0.69 vs. 1987 = 0.60), reflecting shorter recall periods.

The LDH‐q showed correlations of 0.63 (past week) and 0.66 (past year) for male alcohol consumption, and 0.68 (past week) and 0.78 (past year) for females. [24]. The Concordia LDH was significantly correlated with the MAST (r = 0.60, P < 0.001) for lifetime alcohol intake, but not for current drinking (r = 0.05). This is consistent with the MAST's focus on long‐term alcohol‐related problems rather than recent consumption, supporting the Concordia LDH as a lifetime‐oriented measure. [7].

One LDH study indirectly supported construct validity, accurately captured expected lifetime alcohol use patterns, suggesting validity across life stages and drinking contexts [25]. Table 3 summarises the reliability and validity of included studies.

TABLE 3.

Summary of the reliability and validity of the included studies.

Reference	Measure	Reliability		Validity
		Test–retest	Test–retest value	Concurrent criterion‐related validity	Concurrent criterion‐related validity value	Content and/or construct validity	Construct validity value
Jacob et al. [25]	LDH	Yes: reliability coefficient. Specific test not reported	The drinking volume showed a strong correlation (r = 0.68, P < 0.01) across all assessments. Drinking phases had moderate correlations: total (r = 0.37), drinking (r = 0.31) and abstinent (r = 0.52). Initial phase variables were consistent, notably QFI (r = 0.40), beer (r = 0.59) and spirits (r = 0.64). Final phase correlations linked QFI with drinking alone (r = 0.48) and at home (r = 0.52). At age 40, test–retest correlations were highest for frequency (r = 0.71), quantity (r = 0.57) and QFI (r = 0.65)	No	NA	Not formally assessed, although single dimensionality of lifetime alcohol patterns implied	The LDH demonstrated construct validity by accurately capturing lifetime alcohol use patterns, aligning with established drinking behaviours across life stages and contexts
Russel et al. [8]	CLDH	Yes: Pearson's r and Spearman's r	Strong test–retest reliability (Spearman: 0.74–0.85, Pearson: similar). Consumption remained stable, with high reliability for self‐reported intoxication (Spearman: 0.76–0.92). Fixed and floating versions showed comparable reliability with no significant differences	No	NA	No	NA
Skinner et al. [6]	LDH	Yes: reliability coefficient. Specific test not reported. Data transformed to reduce skewness	Duration of drinking (high, r = 0.94), total volume (strong, r = 0.80), daily average (moderate, r = 0.68), quantity per session (lower, r = 0.57) and beverage type (r = 0.62–0.70)	Not formally or statistically assessed	Indirect support provided for validity by providing parallel data with the LDH and an established screening tool for alcohol dependence (MAST) at 2 time points	No	NA
Melo et al. [27]	Modified CLDH	Responses from 2 interviews were compared using ICC	Drinking phases (ICC = 0.856), years drinking (ICC = 0.987) and phase duration (ICC = 0.930), all with non‐significant F tests. ICC for frequency, quantity and alcohol type ranged 0.660–0.991	No	NA	No	NA
Sobell et al. [22]	LDH	Reliability assessed using correlation coefficients	Frequency (r = 0.81), total drinks (r = 0.66), QFI (r = 0.67), max drinks/day (r = 0.65), limited drinking (1–4 drinks: r = 0.53, 1–6 drinks: r = 0.73)	Yes	Prediction accuracy ranged from 72%–91% across three domains of alcohol problems in the Recent Life Changes Questionnaire	No	NA
Friesema et al. [24]	LDH	Assessed using κ for binary variables or single‐score ICC for a 1‐way model	The ICC for total lifetime intake was 0.75 for men and 0.70 for women, with age of onset ICC at 0.21 for men and 0.52 for women Frequency of drinking and quantity per occasion increased with age. In youth (12–18 years), frequency was 0.48 for men and 0.42 for women, while quantity was 0.68 for men and 0.56 for women. ICCs rose in young adulthood (19–27 years) to 0.64 and 0.53 for frequency, 0.75 and 0.62 for quantity. Among adults (28–45 years), frequency reached 0.72 and 0.75 and quantity 0.72 and 0.57. In middle age (46–60 years), frequency was 0.78 and 0.79, quantity 0.72 and 0.54. Among the elderly (61 + years), frequency was 0.80 and 0.77 and quantity 0.77 and 0.50 For QFI, ICCs increased from 0.44 and 0.50 in youth to 0.74 and 0.66 in young adulthood, 0.74 and 0.83 in adulthood, 0.73 and 0.79 in middle age, peaking at 0.88 for both in the elderly	Yes	Face validity‐mapping of data derived from the LDH—visually matched/tracked per capita alcohol consumption	Yes: compared the LDH with previously self‐reported QF baseline questionnaire data at within the last 12‐months self‐report and the last week recall (Pearson's r)	Alcohol last week 0.63 for males and 0.68 for females. Alcohol last 12‐months 0.66 for males and 0.78 for females
Harris et al. [23]	Modified LDH	Yes: Pearson's and Kendall's τ‐b and κ	Correlation for age of onset was around 0.5, with 66.3% of younger problem drinkers and 51%–54% of others reporting onset within 1 year of their initial report For drinking patterns, frequency reliability (Kendall's τ‐b and κ) ranged 0.26–0.49 and quantity per drinking day ranged 0.34–0.47. Consistency was high, with 60–88% reporting the same frequency and quantity across time points Maximum frequency and quantity correlations were lower in younger participants (τ‐b 0.17–0.39) and higher in older participants (0.35–0.58)	No	NA	No	NA
Chaikelson et al. [7]	Concordia LDH	Yes: correlation coefficient	Correlations were 0.76 for lifetime drinking, 0.68 for pre‐1963 drinking and 0.78 for drinking since 1963. The last reported month had the highest correlation (r = 0.84), followed by the current month for active drinkers (r = 0.77). Age at first drink had r = 0.66 and age at regular drinking onset had r = 0.56. Correlations calculated every 5 years (1945–1985) ranged 0.65–0.87, with higher reliability for more recent years	Yes‐ collateral history from spouse. Current drinking and drinking at time of marriage	0.87 for current drinking and 0.72 at time of marriage (median length of marriage 42‐years)	Yes: MAST	MAST scores correlated 0.60 for lifetime alcohol consumed and 0.05 for current drinking
Koenig et al. [26]	LDH	No	NA	No	NA	Yes: mean differences and correlations were examined. ANOVA assessed age of first drinking over time, PROC mixed tested retrospective vs. prospective QF differences, and the Wilcoxon signed‐rank test analysed median differences. Retrospective recall was compared to prospective data at 13, 10, 8 and 5 years	The age at first regular drinking ranged 0.42–0.58 (P < 0.01), with a mean of 19 years across surveys and 20 years for LDH. QFI scores showed the greatest mean differences in lifetime AD drinkers, with median variations in 1987 and 1995. AD men tended to overestimate past intake, especially for quantity, while frequency differences were smaller. Correlation coefficients between prospective and retrospective reports ranged 0.60–0.69 overall, 0.59–0.69 for non‐dependent individuals and 0.57–0.69 for those with AD

Abbreviations: AD, alcohol dependence; ANOVA, analysis of variance; CLDH, cognitive lifetime drinking history; ICC, interclass coefficient; LDH, lifetime drinking history; MAST, Michigan Alcoholism Screening Tool; max, maximum; MI, myocardial infarction; QF, quantity frequency; QFI, quantity frequency index.

Summary of reliability and validity findings

Across the included studies, test–retest reliability for lifetime alcohol exposure instruments was generally moderate to strong. Reported correlation coefficients ranged from 0.60 to over 0.90, with most falling within accepted thresholds for moderate‐to‐strong psychometric evaluations. Validity evidence was more variable. Concurrent validity was typically assessed using self‐report screening tools [e.g. MAST, Alcohol Use Disorders Identification Test (AUDIT)], retrospective QF measures or collateral history (spouse). One study compared reported alcohol use to national per capita consumption. No studies incorporated prospective drinking diaries or biological measurement of alcohol exposure to validate self‐report. Construct validity was evaluated in a subset of studies. It was generally supported through associations with theoretically expected patterns, including higher lifetime estimates among individuals with alcohol dependence, stronger correlations with instruments such as the MAST, and consistency across life stages. Although the number of validity studies was limited, available data suggest that these instruments perform reasonably well in capturing cumulative alcohol exposure, particularly in structured settings.

DISCUSSION

This is the first scoping review to comprehensively assess the reliability and validity of instruments used to capture lifetime alcohol consumption. We identified nine studies that met inclusion criteria for the assessment of reliability and validity that assessed three instruments. Overall, across the nine studies, test–retest reliability, construct validity and concurrent validity were found to be at least moderate, indicating that individual recall of historical alcohol exposure using valid instruments is generally reliable. Despite conjecture regarding recall bias and stigma as limitations of lifetime alcohol exposure instruments, the available data suggest recall bias does not consistently undermine reliability, particularly when structured tools are used. Nevertheless, given the limited number of validity studies and the lack of contemporary cohorts, further research is needed to better characterise the extent and impact of these biases. Strengthening and updating these instruments is also essential given their potential utility in translational research, particularly for investigating the long‐term effects of alcohol exposure on clinical complications such as cirrhosis or dementia. The capacity to better control for cumulative exposure also provides a means to better assess the influence of potential effect‐modifiers of alcohol exposure such as genetic, metabolic or environmental factors [10, 11].

Recall bias is a similarly important, and arguably more pervasive, limitation in lifetime alcohol exposure measurement. Recall of alcohol use over decades predisposes to bias, whether through memory gaps, blurred timelines or reliance on imprecise averages. In the instruments reviewed, all attempt to overcome this in some way by scaffolding recall, whether through age‐based phases, life event cues or visual prompts, however, the extent to which these strategies meaningfully reduce error remains unclear. Like stigma, recall bias introduces non‐random error that is difficult to quantify and is likely influenced by demographic and clinical factors. Future work should investigate how anchoring techniques affect recall accuracy and whether measurement performance varies systematically across different populations.

The instruments appear to be valid and reliable in a variety of patient populations, including both treatment‐seeking and alcohol dependent, and are effective independent of sex or other demographic factors. This generalisability increases their utility in a range of research settings, from population studies and epidemiological surveys to individual clinical assessments. Additionally, observed discrepancies (e.g. a 1‐year variation in the recall of the age of onset of drinking), seem unlikely to be clinically significant [23].

There are some limitations to these data. All included studies were conducted in the 1980s and 1990s. Although these studies are seminal, re‐assessing their reliability and validity in more contemporary populations is important, given the changes in societal attitudes toward alcohol and alcohol consumption. A lack of comprehensive validity assessments, with only three studies formally evaluating validity, is an important area that requires future work. None of the included studies reported confidence intervals or standard errors for their psychometric estimates, limiting our ability to assess the precision and statistical uncertainty of the reported reliability and validity values. Sample size is also a concern with many studies, including a relatively small number of participants. Most instruments also fail to adequately capture binge or HED directly, relying instead on proxies like peak intake. This is a critical gap, because HED drives many alcohol‐related harms. Future tools should consider incorporating structured prompts or calendar techniques to assess episodic patterns more effectively.

People seeking treatment for alcohol dependence were relatively under‐represented. Treatment seeking individuals are usually more accurate about their alcohol consumption than the general population or participants with AUD not seeking treatment [28]. This may be because of their motivation to change behaviour and to reflect on past drinking and associated harms, as well as often being seen in a structured, clinical environment. This may explain the findings of the Vietnam Twin Era study, where participants with a diagnosis of alcohol dependence who were not seeking treatment provided larger estimates of their drinking with increasing passage of time [26]. Similar findings have been reported elsewhere [29], although others have shown stability [30]. These limitations underscore the need for updated, comprehensive and diverse studies to enhance the reliability, validity and applicability of tools measuring lifetime alcohol exposure.

A comprehensive, broad and pragmatic definition of validity, like those used in prior studies, was used for this review as it provided a foundation for evaluating self‐reported measures of lifetime alcohol consumption. Unlike criterion or content‐based approaches, construct validity emphasises theoretical coherence by integrating diverse evidence of alcohol exposure and risk, such as screening tools for hazardous drinking [31, 32]. Using a multi‐dimensional approach provides a more nuanced assessment of cumulative lifetime alcohol exposure. For instance, when performed by the included studies, additional validation against MAST, spouse reports and per‐capita consumption allowed lifetime alcohol exposure to be situated within a broader theoretical framework, making it less reliant on history alone, which enhanced the accuracy and utility of the instrument.

Future work should address: (1) how much detail respondents can reliably recall; (2) how current drinking influences memory of past use; and (3) what methods or data sources can resolve inconsistencies and improve reliability [9]. These questions are central to understanding inherent limitations in gathering patient history, advancing the reliability and validity of lifetime alcohol exposure instruments and how innovative approaches may overcome the limitations of patient recall and the instruments we use to capture it.

One promising avenue for addressing these challenges lies in the integration of alcohol biomarkers, such as PEth, with lifetime alcohol exposure measures [33]. Alcohol biomarkers offer some objectivity to complement self‐reported recall‐based estimates of lifetime drinking. By linking retrospective recall data to biological markers of alcohol consumption, a calibration point can be established for extrapolating past drinking patterns. Similarly, combining different recall techniques, such as Timeline Followback assessments, AUDIT and MAST, with alcohol biomarkers may help to ‘triangulate’ the history, identify and resolve inconsistencies, particularly when under‐reporting occurs because of stigma [34, 35, 36, 37].

Methods that enhance recall, such as calendars and the plotting of milestones on a timeline (e.g. leaving school, marriage and deaths), could also enhance the accuracy of respondents' accounts of their drinking behaviour [7, 8, 27]. These were part of the cognitive LDH and Concordia LDH. By using anchor points for significant life events to help respondents recall their more distant alcohol consumption behaviour, the burden of recall on respondents may be eased, leading to more accurate data [1, 3]. However, no study has ever compared lifetime measures of alcohol exposure with and without visual milestone aids that could assist in overcoming socio‐cultural barriers for the assessment. Future work should examine the administration of both long‐term and short‐term alcohol exposure instruments in parallel to help contextualise their relative performance and clarify their respective roles in clinical and research settings.

Alongside visual aids, cognitive interviewing also offers a complementary strategy to improve measurement. Exploring how respondents interpret and answer questions can help identify ambiguous terms, cultural assumptions or misunderstandings in the wording of items. Incorporating this approach could strengthen item clarity and response accuracy, especially when recalling behaviour across decades.

In conclusion, accurately measuring lifetime alcohol exposure is critical to understanding alcohol's role in disease. Although a valid assessment of lifetime alcohol exposure is essential to understanding the role of alcohol consumption for specific disease processes, this scoping review suggests that existing instruments for lifetime exposure are valid and reliable and can overcome concerns regarding biases. Our study identifies areas for future work, particularly using a more integrative multi‐dimensional approach, using ancillary assessment tools (e.g. AUDIT) to assess both exposure and harm and the use of independent biomarkers to supplement patient history. This has the potential to enhance the reliability and validity of lifetime alcohol exposure tools, reinforcing their relevance in understanding clinical, epidemiological, histological and translational outcomes.

AUTHOR CONTRIBUTIONS

Andrew J. Palmer: Conceptualization (equal); formal analysis (lead); investigation (lead); methodology (lead); project administration (lead); writing—original draft (lead). Jason P. Connor: Formal analysis (equal); investigation (supporting); methodology (supporting); project administration (equal); supervision (supporting); writing—original draft (supporting); writing—review and editing (lead). Gerald Holtmann: Supervision (supporting); writing—review and editing (equal). John B. Saunders: Supervision (supporting); writing—review and editing (equal). Katie Rice: Data curation (supporting). Jeremy Yeo: Data curation (equal). Andrea Huang: Data curation (supporting). Paul J. Clark: Conceptualization (equal); formal analysis (equal); project administration (equal); supervision (lead); writing—original draft (supporting); writing—review and editing (equal).

DECLARATION OF INTERESTS

None.

Palmer AJ, Connor JP, Holtmann G, Saunders JB, Rice K, Yeo J, et al. Measuring lifetime alcohol exposure: A scoping review and implications for translational research and alcohol‐related harm. Addiction. 2026;121(3):472–487. 10.1111/add.70208

DATA AVAILABILITY STATEMENT

Data available on request from the authors.