UvA-DARE (Digital Alcohol and Brain Development in Adolescents and Young Adults: A Systematic Review of the Literature and Advisory Report of the Health Council of the Netherlands

Young people, whose brains are still developing, might entail a greater vulnerability to the eﬀects of alcohol consumption on brain function and development. A committee of experts of the Health Council of the Netherlands evaluated the state of scientiﬁc knowledge regarding the question whether alcohol negatively inﬂuences brain development in young people. A systematic literature search for prospective studies was performed in PubMed and PsychINFO, for longitudinal studies of adolescents or young adults ranging between 12 and 24 y of age at baseline, investigating the relation between alcohol use and outcome measures of brain structure and activity, cognitive functioning, educational achievement, or alcohol use disorder (AUD), with measures at baseline and follow-up of the outcome of interest. Data were extracted from original articles and study quality was assessed using the Newcastle-Ottawa Scale. A total of 77 studies were included, 31 of which were of suﬃcient quality in relation to the study objectives. There were indications that the gray matter of the brain develops abnormally in young people who drink alcohol. In addition, the more often young people drink or the younger they start, the higher the risk of developing AUD later in life. The evidence on white matter volume or quality, brain activity, cognitive function, and educational achievement is still limited or unclear. The committee found indications that alcohol consumption can have a negative eﬀect on brain development in adolescents and young adults and entails a risk of later AUD. The committee therefore considers it a wise choice for adolescents and young adults not to drink alcohol. Adv Nutr 2021;12:1379–1410.


Introduction
In 2014 the Dutch government changed the legal drinking age from 16 to 18 y in order to protect children and adolescents from the risks of alcohol consumption, based on experts' advice to do so. The reason for this policy change was the emerging literature indicating that underage drinking may have detrimental effects on brain development (1)(2)(3)(4), besides the fact that acute effects of alcohol include a higher risk of accidents, violence, and other transgressive behavior (5)(6)(7)(8), and that chronic alcohol consumption increases the risk of many diseases and disorders (9)(10)(11)(12)(13)(14)(15)(16). That is why the Dutch advice for the general population is not to drink alcohol, or at least ≤1 glass/d. Especially for young people, alcohol is harmful. For example, they become intoxicated more quickly than adults (7,17). Furthermore, drinking at a young age is associated with drinking later in life (18,19). Also, it is widely assumed that alcohol negatively affects brain development, which continues into the late 30s (20).
In 2016, the Dutch State Secretary for Health, Welfare and Sport asked the Health Council of the Netherlands what, according to the latest scientific knowledge, is known about the effects of alcohol on the brain of young people between the ages of 12 and 24 y and whether such possible effects are reversible. One of the reasons for this request may have been that conflicting data concerning adverse effects were published since the policy change, including a large prospective study in the Netherlands (21) showing no adverse effects of adolescent binge drinking on a number of neuropsychological functions, which made it to the front page of a national newspaper (22). The State Secretary also asked for the consequences of alcohol consumption at a young age on the extent of use of alcohol in adulthood to be evaluated. A committee was formed from experts from different relevant areas of scientific expertise.
Over the past decade, a large number of scientific reviews have addressed the topic of alcohol consumption in relation to brain development in adolescents and young adults indicating detrimental effects of alcohol on brain development (1)(2)(3)(4)(23)(24)(25)(26)(27)(28)(29)(30)(31)(32)(33)(34)(35)(36)(37)(38)(39)(40)(41). We briefly highlight the findings from some of these reviews. Feldstein Ewing et al. (41) concluded in their systematic review (SR) of 21 observational studies, of which most were cross-sectional, that alcohol consumption during adolescence is associated with differences in both brain structure and function during development. Furthermore, based on 7 observational studies in adolescents, of which 1 was longitudinal and 6 were cross-sectional, Elofson et al. (4) concluded that alcohol consumption is associated with reduced white matter integrity, particularly in the superior longitudinal fasciculus. Based on a review of 38 observational, also mainly cross-sectional, studies in adolescents, Silveri et al. (34) concluded that differences in brain structure, white matter architecture, and brain function associated with alcohol consumption were mainly present in the frontal lobe (in 61% of the studies), followed by the temporal lobe (45% of the studies) and parietal lobe (32% of the studies). Alcohol consumption during adolescence or young adulthood may influence cognitive functions (2,24,30), such as attention, memory, decision-making, planning, and learning ability. In addition, alcohol consumption appears to be associated with automatically activated appetitive responses to alcohol cues, known as alcohol-related cognitive bias ("being hypersensitive for cues of alcohol in the environment"). Such cognitive biases are likely to contribute to the development of problem use (37). Adverse effects on cognitive function may, in turn, influence educational achievement, an important determinant of vocational success, income, health, social status, and quality of life (42). However, alcohol consumption is likely to also affect educational achievement directly, for example as a result of hangovers or sleep deprivation (43). A 2011 SR, however, reported mixed findings on the relation between alcohol consumption and educational consequences based on 3 longitudinal studies (18). In 2011, McCambridge et al. (18) performed an SR of prospective cohort studies into the adult consequences of late-adolescent alcohol use, with ≥3 y of follow-up. The authors concluded that there is consistent evidence that higher alcohol consumption in late adolescence continues into adulthood and is also associated with alcohol problems, including alcohol use disorder (AUD) or alcohol dependence (AD) (18). In 2014, Maimaris and McCambridge (44) performed an SR on the association between the age of first drink (AFD) and adult alcohol problems. Only cohort studies comprising general population samples were included, with a requirement of ≥3 y follow-up between the initial measurement of AFD in adolescence and the assessment of alcohol-related outcomes. Based on 5 studies (4 study samples), the authors concluded that there is some evidence for an association between AFD and AD, but this disappears with more rigorous control for confounding. The authors also mention that over-adjustment is a point of concern, because peer variables may lie on the causal pathway to adult outcomes as well as being implicated in earlier AFD.
The aforementioned reviews of human research (2,4,18,30,34,37,41,(44)(45)(46) point out several limitations regarding the interpretation of the available evidence, such as small sample sizes (18,41), the small number of longitudinal studies (4,18,30,34,41,44), overlap in study samples (18), and vulnerability to bias (18). Neurobiological differences may exist before the initiation of alcohol use. In addition, confounding or effect modification by gender, concurrent marijuana use, or comorbid psychiatric disorders may play a large role. Observed differences could also reflect antecedents of alcohol use, such as age of first use, family history of addiction, childhood maltreatment, or comorbid psychiatric conditions (34).
The committee decided to limit the scope of the review to 1) human studies, 2) with a prospective design, 3) on the following outcomes: brain structure and activity, cognitive function, educational achievement, and AUD. These 3 decisions will be further explained in what follows. Many reviews of experimental animal studies on the effects of alcohol are available (24,35,(47)(48)(49), yet the committee was not aware of any SR. In a recent review from Spear (47), human observational research and experimental animal research were presented together and compared. The review referred to 3 studies in which alcohol intake affected gray matter volume and white matter volume and quality in adolescent rats. Based on other experiments presented in the review, adolescent rats repeatedly exposed to ethanol vapors 1380 de Goede et al.  Population  Adolescents and young adults within the age range of 12-24 y at baseline  Intervention  Alcohol consumption  Comparator  Less or no alcohol consumption  Outcome  Measures of brain structure and activity, cognitive functioning, educational achievement, or alcohol  use disorder  Study design Human prospective studies with measures at baseline and follow-up of the outcomes of interest showed an aberrant electrophysiological pattern (decrease in P300 amplitude), consistent with a disruption in the development of the hippocampus, a brain area involved in memory. According to Spear, cognitive studies in rodents generally have revealed that repeated exposure to alcohol during adolescence has minimal effects on simple spatial learning tasks and on more challenging learning tasks like 5-choice serial reaction time tests. However, when the task demands require some degree of cognitive flexibility, deficits have often emerged. It was argued by the committee that, although animal studies have unique merit in delineating causal mechanisms in the effects of alcohol on the brain, they also have their limitations: studied dosages (sometimes unrealistically high) as well as studied outcomes in animal studies limit extrapolation to humans. The main drawback of cross-sectional studies is that they cannot disentangle causes and consequences. Neurobiological differences may have existed before the initiation of alcohol use and they could even be the cause of early drinking. Therefore, the committee focused on prospective studies with repeated measurements of the outcome in order to identify whether or not outcome differences were already present at baseline to get insight into reverse causation.
Because the committee hypothesized that changes in brain structure or activity could translate into changes of cognitive function and eventually educational performance, the committee decided to focus not only on measures of brain structure and brain activity, but also on the association between alcohol consumption and both cognitive function and educational achievement. For the question about the influence of alcohol consumption at a young age on the use of alcohol in adulthood, the committee focused on AUD, previously divided into 2 types of problematic drinking: alcohol abuse (AA) and AD. In AUD someone's activities, behavior, or relationships suffer from the use of alcohol and the person has difficulty stopping or cutting back alcohol use, or is addicted to alcohol.
Existing reviews either included both cross-sectional research and prospective studies, or were not systematic reviews, or were not sufficiently recent or specific. Therefore, the completeness of selection and judgment of the literature in existing reviews were uncertain. The committee therefore performed an SR of human prospective studies on alcohol consumption and both brain function and development and AUD in adolescents or young adults, including a quality assessment of the included studies.

Methods
The 10 committee members, covering the research fields of (alcohol) addiction, cognition, neurology, neuropsychology, neuroimaging, social sciences, epidemiology, and statistics, filled out declarations of interest, which were published (in Dutch) on the website of the Health Council (www. gezondheidsraad.nl). The committee performed an SR of peer-reviewed longitudinal studies of alcohol consumption by young people (adolescents and young adults) in relation to outcome measures of 1) brain structure and activity, 2) cognitive functioning including alcohol-related cognitive biases, 3) educational achievement, and 4) AUD. The SR was performed in accordance with the Meta-analysis and Systematic Reviews Of Observational Studies in Epidemiology (MOOSE) guidelines (see Supplemental Methods 1) (50).

Identification and quality appraisal of longitudinal studies
Published articles (in English) up to and including May 2018 were retrieved by the committee and librarian from PubMed and PsychINFO, and complemented by hand searches of reference lists and correspondence with researchers in the field. Included were longitudinal studies of alcohol consumption by adolescents and young adults within the age range of 12-24 y at baseline with repeated measurements of any of the 4 outcomes of interest ( Table 1). For the outcome educational achievement (mainly school dropout or highest attained degree), by definition, there are no differences yet at baseline. For reasons of consistency, within the topic of educational achievement, the committee also included studies with only 1 measurement of educational marks. For studies concerning AUD, we also included studies that lacked a baseline assessment of AUD for subjects aged 16 y or younger, because the committee regarded the risk of already existing AUD as low in this age group.
The committee excluded 1) studies on the acute effects of alcohol; 2) studies of specific subgroups, because findings could not be generalized to the general population (e.g., subjects with attention deficit hyperactivity disorder or speech and language impairment, patients in drug clinics, patients with bipolar disorder); 3) studies without a control group with no alcohol use; 4) studies with only combined use of alcohol and other substances (such as marijuana); and 5) studies in which the onset of alcohol consumption was assessed retrospectively, because of the risk of recall bias (51,52). In total, the committee included 77 studies Alcohol and brain development in young people 1381

Study quality assessment and weighing of study quality
The risk of bias for each study was assessed with the Newcastle-Ottawa Scale (NOS) (128) (see Supplemental Methods 3 for the scoring method and Supplemental Table  1 for the scores per study), based on consensus between 2 independent judges (pairs of authors). The NOS rating system scores studies from 0 (highest risk of bias) to 9 (lowest risk of bias) on the nature of the study sample, exposure and outcomes assessments, baseline differences in the assessed outcome, attrition bias, and potential confounding. The committee judged gender, age, use of other drugs or smoking, externalizing behavior, and family history of AUD as important potential confounders. The committee judged studies where the outcome was assessed before the initiation of alcohol consumption to be of high value for the research questions. In that situation, the baseline measurements cannot (yet) be affected by alcohol consumption. In cases of large study samples or many statistical comparisons within a study, the possibility of chance findings is relatively high.  Therefore, the committee reported for each study whether results were based on a priori defined hypotheses (such as a priori defined brain "regions of interest"), or whether results were adjusted for multiple testing to limit chance findings. The committee also weighed whether results were based on independent data, i.e., different study populations.

Data extraction and data synthesis
Data were extracted using structured extraction forms which included information on the study sample, measurement of exposure and outcomes measures, statistical analysis (including covariates, stratification or matching factors, and correction for multiple testing), results, and limitations. All relevant exposure and outcome measures were extracted, based on the most extensive statistical models in terms of adjustment reported in the original studies. The committee judged studies with an NOS score of ≥7, with at least minimal adjustment for confounding, to be of sufficient quality and the remainder of the evidence of lower quality in relation to the study questions. In the description of the results, results were presented separately, if possible, for high school students and college/university students, i.e., providing a rough distinction between groups that differ in age, social circumstances, and drinking patterns. Conclusions of the committee were primarily based on the studies of sufficient quality, whereas the results of the studies with lower NOS scores were used as ancillary material. Conclusions were derived only if ≥3 studies were Alcohol and brain development in young people 1383 available with sufficient quality based on ≥3 different study populations. Regarding all outcomes, the large heterogeneity of studies did not allow quantitative conclusions.
Alcohol and brain development in young people 1387 62) and white matter integrity (56,59,62), and functional measures including task-related fMRI (60,63,67), taskrelated event-related potential (33,54,55,57), task-related connectivity (fMRI or magnetoencephalography) (69), and resting-state connectivity (66,68). In 1 study, the study sample was selected for having no lifetime experience with alcohol (56). In 7 studies (of which 6 were from the same research group), baseline alcohol consumption was limited (53,(59)(60)(61)(62)(63)69). NOS scores ranged between 4 and the maximum possible score of 9. In the majority of the studies (n = 12), the extent of attrition bias could not be evaluated because limited information was available about the participants who were excluded from the analyses (see Supplemental

Gray matter volume and cortical thickness.
There were 7 studies (in secondary school students and university students) that reported on gray matter: 3 of sufficient quality. All 3 studies (based on 3 study populations) of sufficient quality on the association between alcohol consumption and gray matter volumes showed reduced gray matter volumes or cortical thickness for higher levels of alcohol consumption (56,59,64), with the most consistent findings for the frontal lobe. In 1 of these 3 studies (59), higher alcohol consumption was related to both reduced gray and white matter volumes, but not with differences in cortical thickness. In 2 of these 3 studies, baseline alcohol consumption was low or absent and baseline differences of the outcome measures were absent (56,59). This strengthens the findings because reverse causation is unlikely. Three (61,62,65) out of the 4 (53, 61, 62, 65) lower-quality studies were consistent with a more rapid decline of gray matter volume, whereas 1 study of lower quality mainly suggested pre-existing gray matter volume differences, with groups (drinking initiators compared with nondrinkers) becoming more similar over time (53).

White matter volume and integrity.
Regarding white matter volume, 2 (56, 59) studies of sufficient quality in adolescents showed inconsistent results [a reduced increase of white matter volume over time (56) compared with no difference in relation to alcohol consumption (59)]. At baseline, outcome measures were similar between the alcohol groups in both studies (56,59). A third study in adolescents, that had lower quality, suggested a lower increase of white matter volume in initiators of binge drinking than in nondrinkers. Baseline differences of the outcome were not reported (62). For white matter integrity, 1 study of sufficient quality showed a lower increase in fractional anisotropy (FA; a lower FA reflects disturbed integrity of white matter) in adolescent alcohol initiators than in noninitiators. At baseline, the outcomes did not differ between the groups (56). In a study of lower quality in university students, no difference was found for FA between sustained binge drinkers and a reference group of nonbinge drinkers. Baseline differences of the outcome were not reported (68).

Brain activity.
The 4 studies of sufficient quality on brain activity outcome measures all focused on binge drinking (54,57,58,63). One of those 4 studies focused on cognitive bias as the outcome (58). In that study, there was no difference in the performance measure of cognitive bias between persistent binge-drinking students and nondrinking students (see also the section on "Cognitive functioning"). The brain activity measures linked to the behavioral measure, however, did differ between the groups (58). In the other 3 studies of sufficient quality (54,57,63), 1 in adolescents and 2 in students, the behavioral measures regarding the cognitive test did not differ between binge drinkers and non-binge drinkers. In 2 out of the 3 studies, differences in brain activity (electroencephalogram, fMRI), linked to the cognitive functions studied, were observed (57,63). In 1 of the 2 studies in which binge drinkers showed different brain activity compared with non-binge drinkers, the participants did not drink yet or only drank limited amounts of alcohol at baseline (63). The diverse nature of the brain activity studies, however, limits drawing conclusions. In half of the studies of lower quality, based on 5 study populations, differences in brain activity were observed (60,68,69), whereas in the other half no differences were found according to alcohol consumption (55,66,67 67,69,72,73,[76][77][78]. Seven studies differed at baseline for the outcome measure of interest or baseline differences of the outcome were not reported (53,58,60,69,76,77,79). In 8 studies no adjustment for relevant confounders was made (55,67,(69)(70)(71)(73)(74)(75). Eight studies took adjustment for multiple testing into account (21,67,73,(75)(76)(77)(78)(79). The committee judged 8 studies to be of sufficient quality based on NOS score (21,54,63,72,(76)(77)(78)(79).
In total, 12 studies on high school students were found (21,53,60,63,67,69,73,74,(76)(77)(78)(79), 6 of which were of sufficient quality (21,63,(76)(77)(78)(79). Five of them were based on 1 American cohort (63,(76)(77)(78)(79). Participants from this cohort were alcohol naïve or had a very low level of alcohol consumption at baseline. In 1 of these American studies, no difference was found in cognitive functions between those who initiated binge drinking and nondrinkers (78). In the other 4, differences were found on several cognitive functions between alcohol consumers and nondrinkers, where alcohol consumers showed relatively poor outcomes compared with controls or where more drinks or starting at a younger age was associated with relatively poor cognitive outcomes (63,76,77,79). One of the American studies found an association between higher alcohol consumption and improvements in working memory (76). In the sixth study of high quality (a Dutch cohort, with 77%-95% alcohol-naïve participants at baseline and no initial differences of the outcome), no associations were found between alcohol consumption (including binge drinking) and cognitive functioning (21). In the remaining 6 studies of lower quality (53,60,67,69,73,74), based on 4 cohorts, no associations were found between alcohol consumption and cognitive functioning or cognitive biases [only 1 study (74) was available on this outcome].
All 7 studies that were conducted among college/university students focused on sustained binge drinking (54,55,58,(70)(71)(72)75). Two of these studies were of high quality; they used data from the same Spanish cohort (54,72). No differences between sustained binge drinkers and non-binge drinkers with regard to visual attention were found (54), whereas an association between sustained binge drinking and relatively poor working memory was observed (72). The outcome measures of interest did not differ at baseline (54,72). Four (55,70,71,75) out of the 5 (55,58,70,71,75) studies of lower quality were from the same Spanish cohort. One of these studies found an association between higher alcohol consumption and relatively poor cognitive functioning (71). The other 3 Spanish studies did not find any differences (55,70,75). The last study, based on Belgian students, focused on cognitive biases and found no differences between binge drinkers and non-binge drinkers (58).

Educational achievement
The committee identified 30 longitudinal studies (  (102), or drinking (yes/no) at baseline (81,84,94,99,107). None of the studies was based on an alcohol-naïve population at baseline. Outcomes included school marks (81,83,90,97,105,108) and level of educational attainment or dropout (19, 80, 82, 84-89, 91-96, 98-108). The NOS scores ranged between 4 and 9. In 16 studies the extent of the attrition bias could not be evaluated, because limited information was available about the participants who were excluded from the analyses (Supplemental Table 3

Level of education.
The committee identified 10 studies of sufficient quality on the association between alcohol use and level of educational attainment in high school students (19,82,85,87,89,98,100,(106)(107)(108). In 5 of these higher alcohol consumption was associated with a higher risk of achieving a lower level of education (85,87,(106)(107)(108). In 1 of the studies an association in the opposite direction was observed: i.e., higher alcohol used was associated with a lower likelihood of dropout (89). In the other 4 studies of sufficient quality no differences were observed between drinkers and nondrinkers (19,82,85,100). Associations were found in studies that focused on increasing levels of alcohol use as well as in studies that focused specifically on binge drinking. Of the remaining 10 studies of lower quality on educational attainment (80, 84, 86, 92-94, 99, 101, 102, 105), 6 found an association with a lower level of education in drinkers (86, 92-94, 99, 101), 3 did not find a significant association (84,102,105), and in 1 study those who used more alcohol were more likely to attain a tertiary education (80).

School marks.
Three studies of sufficient quality on alcohol use and school marks achieved were identified in high school students (81,90,108). One of them found an association between alcohol use at young age and lower school marks (108). The other 2 studies did not find evidence for an association (81,90). Of the remaining 3 studies of lower quality (83, 97, 105), 2 found an association with lower school marks (83,97). The other found no significant association (105). There were no studies of sufficient quality on the association between alcohol use and educational attainment in college/university students. Of the remaining 6 studies of lower quality (88,91,95,96,103,104), 4 found an association of heavier alcohol use with a Alcohol and brain development in young people 1393     Study quality/risk of bias was assessed with the Newcastle-Ottawa Scale (0-9); for clarification see Supplemental Methods 3 and Supplemental Table 3. 3 Cohort included in Green et al. (80). 4 No sufficient adjustment for relevant confounders, therefore not qualified by the committee as a study of sufficient quality.
Alcohol and brain development in young people 1399 higher likelihood of relatively worse school performance (88,91,95,103) and 2 found no significant association (96,104).

Amount or frequency of drinking.
The committee identified 2 studies (based on 3 study populations of adolescents and university students) of sufficient quality regarding amount or frequency of alcohol consumption as exposures of interest (19,127). In both studies, a higher frequency of alcohol consumption was associated with a higher risk of later AUD, whereas a higher amount of alcohol consumption was associated with a higher risk of later AUD in 1 of the 2 studies (127). In the second study, the association of a higher alcohol quantity with a higher risk of later AUD was no longer statistically significant after Bonferroni corrections for multiple testing (19). From the remaining 12 studies of lower quality (93, 105, 106, 109-112, 115, 119-121, 123), based on 10 study populations, 10 out of 12 (93, 105, 106, 109-112, 115, 120, 123) observed an association between a higher level of alcohol consumption and a higher risk of AUD. The other 2 (119, 121) did not find an association between level of alcohol consumption and risk of later AUD.

Age of onset of alcohol consumption.
Regarding the age of onset of alcohol consumption, the committee identified 9 studies (113, 114, 116-118, 122, 124-126), of which 5 were included in a previous SR on the relation between AFD or early drinking and later-life AUD (44). Four additional studies were found by the committee that were not included (117) in the SR (44) or are more recent (116,118,122).
In all 4 studies of sufficient quality (all in adolescents) (113,116,124,126), drinking at a younger age (113,116,124) or getting drunk after the first drinking episode (126) was associated with a higher risk of later AUD. In 1 of the 4 studies, the association was more pronounced for regular use initiation than for any drinking initiation (113). In the second of the 4 studies, the age of any drinking initiation was not associated with later AUD, whereas the age of first getting drunk was the strongest predictor of later AUD (126). In the third study, the size of the RR of later AUD for drinking before the age of 14 y was similar to the size of the RR associated with getting drunk before the age of 14 y (116). The fourth study did not have information on the level of alcohol consumption in relation to drinking age (124). From the remaining 4 studies (based on 4 study populations) that were not of sufficient quality (117,118,122,125), 1 study (118) found an association between a younger starting age of alcohol consumption and an increased risk of later AUD. In the other 3 studies no such associations were observed (117,122,125).

Discussion
The committee aimed to provide a systematic overview of human prospective studies on alcohol consumption and both brain function and development and AUD in adolescents and young adults. In the research results assessed, the committee found indications that alcohol consumption can have a negative influence on brain development in adolescents and young adults and entails a risk of later AUD.

Brain structure and function
Regarding brain structure and function, the committee concluded that there are indications based on rather consistent findings across studies that the volume of gray matter in the brain shows an accelerated decline in young people who drink. The consequences of an accelerated decline of gray matter, however, are not yet clear. No judgment could be made on the relation between alcohol consumption and white matter volume or integrity because not enough studies of sufficient quality were available. For alcohol consumption and brain activity, the study designs were too diverse to make a judgment. Of note, in general, a difference in brain activity not paralleled by behavioral effects is difficult to interpret. A decrease in brain activity could be explained as impaired brain functioning, whereas the opposite (an increased brain function) could be explained as a compensatory mechanism of an affected brain to perform normally. Regarding the neuroimaging and neurophysiology outcome measures, the total number of studies was low and the study quality and study designs in terms of population, exposure, and outcome measures varied widely, limiting the possibility of drawing overall conclusions on the effect of alcohol on the brain. Furthermore, the available studies generally were not comprised of large groups of subjects and focused on groups that showed extreme differences in terms of their alcohol consumption. It may be that the majority of adolescents are part of the middle group, which was often not studied. In    addition, studying only extreme groups makes it impossible to study the shape of associations (i.e., dose-response or threshold effects).

Cognitive functioning
Regarding cognitive functioning, the committee concluded that the relation between alcohol consumption and cognitive functioning in young people is still unclear, because the available studies of sufficient quality were largely based on data from only 2 study populations. Other caveats were that a variety of cognitive tests have been used, making the results difficult to compare, and that in some studies a large number of cognitive outcomes (test results) have been reported, which increases the possibility of chance findings.

Educational achievement
Regarding educational achievement, in approximately half of the available studies, adolescents who drank alcohol performed worse at school than young people who did not drink: they achieved a lower level of education or left school without a diploma. In the analyzed studies, it is difficult, however, to establish the extent to which the risk of early school dropout at the start of the study already differed among the participants. As a result, it is possible that poorer educational achievement is not the result of alcohol consumption but the cause. It can also not be ruled out that the associations found were caused by a so-called "third factor" (129) associated with both alcohol consumption and educational achievement. For example, personality characteristics such as risk-seeking behavior may be related both to higher alcohol consumption and to skipping classes or problem behavior. Two of the included studies that found an association between higher alcohol use and lower school performance (1 of sufficient quality) doubted the causality of their findings and discussed alternative explanations for their findings (87,101). Altogether, the committee therefore concluded that the connection between alcohol consumption and educational achievement is still unclear.

AUD
Regarding AUD, the committee concluded that there are indications that starting drinking at a young age is associated with a higher risk of developing AUD later in life. The more often young people drink, or the younger they start, the higher this risk. In all available studies of sufficient quality, adverse associations were observed. It was, unfortunately, not possible to quantitatively summarize these findings because studies differed substantially in the figures reported, precluding the aggregation of findings. For example, the measures of age of onset were different across the studies and the reference groups. The committee in addition notes that starting age of drinking as a measure of exposure to alcohol has its limitations (44,130,131). The question is, whether the age of the first experience with drinking alcohol is as important a risk factor as the starting age of regular alcohol consumption or the age of getting drunk for the first time (113,131). Several authors doubted the usefulness of the concept of drinking onset (44,130,131). In addition, some studies supported the idea that the first experience with alcohol, as such, is not as strong a risk factor for later problems as are experiences of amounts of more than just a few sips (113,131). Also a "third factor" (129) could play a role here.

Limitations
The committee wants to address some general limitations of this review. First, it is not possible to perfectly assess alcohol consumption in observational research, because the information on alcohol consumption is based on self-report by research participants, which is influenced, among other factors, by memory. In addition, alcohol consumption can vary over time. Based on extensive research it is known that the accuracy of self-reporting on alcohol is enhanced when 1) people are alcohol free when interviewed; 2) written assurances of confidentiality are provided to the participants; 3) people are interviewed in a setting that encourages honest reporting; and 4) participants are asked clearly worded objective questions (e.g., "Did you get drunk last night?") (132). Depending on one's personality, alcohol consumption may be underestimated (by providing socially acceptable answers) or overestimated on purpose (showing off) (18,130). Although the committee weighed the quality of the alcohol exposure assessment using the NOS, it is not possible to judge to what extent self-reporting has affected the results. Second, although the committee evaluated other substance use and externalizing behavior as confounding factors (NOS), it is still difficult to disentangle the role of alcohol from other often clustered risk factors for the outcomes studied here: brain development or AUD. Third, the variability in measurements of alcohol consumption impaired drawing conclusions for the degree of alcohol consumption, e.g., binge drinking. In addition, comparison groups varied widely between studies, e.g., binge drinking was often compared with non-binge drinking and not with nondrinking. Fourth, the committee could also not answer the question whether the consequences of alcohol consumption are reversible because hardly any studies were available. Finally, publication or reporting bias may have played a role (18,44,(133)(134)(135)(136).
To improve the evidence base of the impact of alcohol consumption on the developing brain, additional populationbased studies are urgently needed. These studies would need to include measurements of the outcomes of interest before the initiation of alcohol consumption in order to disentangle causes and consequences. Future studies should cover a wide range of measures of brain structure and function. That would, for example, enable studying functional consequences of aberrations of brain structure or activity in relation to cognitive function, or the impact of cognitive impairments after alcohol consumption in relation to educational achievements or AUD in later life. To this respect, a few ongoing initiatives should be mentioned (137,138). Furthermore, data on the reversibility of findings are of importance. Finally, data of Alcohol and brain development in young people 1405 non-Western countries are needed to enable extrapolations of the findings.
Because of the several potential limitations, which are largely inherent to this field of research, the committee has been cautious with drawing firm conclusions, first, by requesting a minimum of 3 comparable studies of sufficient quality based on nonoverlapping study populations and, second, by incorporating additional points of concern in the weighing for specific endpoints.

Conclusion
This SR suggested that alcohol consumption can have a negative influence on brain development of young people and entails a risk of later AUD. Moreover, based on previous research it is known that alcohol consumption leads to risky behavior, due to acute effects of alcohol, and health risks in the longer term. The committee therefore considers it a wise choice for adolescents and young adults not to drink alcohol.