Does familial risk for alcohol use disorder predict alcohol hangover?

All variables followed a normal distribution although tending towards skew and platykurtosis in some cases. However, where appropriate transforms could be identified (e.g. logarithm; inverse), analyses yielded identical results and so only non-transformed analyses are reported. During the online data collection, 10 participants omitted to input their age and nine participants did not provide an answer to the item “age of first intoxication”. These participants were included in the analysis in order to maximize analytic power.

This online survey has replicated previously observed findings of increased retrospective ratings of hangover symptom frequency among FHPs controlling for alcohol consumption. Interestingly, when family history and blood alcohol were examined simultaneously, neither variable moderated or mediated the other, suggesting that positive family history of AUD and drinking behaviour explain separate aspects of retrospectively rated hangover frequency. Our study addresses some of the concerns raised in the introduction around the need to control for hangover-prone alcohol consumption rather than typical consumption. It is possible that the retrospective hangover frequency estimates provided by FHPs reflect a bias in past month drinking estimates; however, these data support the previous studies showing that FHPs report greater hangover frequency than FHNs over the past year, over and above alcohol consumption. This occurred despite more relevant measures of alcohol consumption—estimated blood alcohol following the largest amount consumed over the past month and drinking frequency—being taken into account.

However, while these data suggest possible predictive value of hangover in the aetiology of AUD, an alternative interpretation is that FHP individuals may be subject to a cognitive bias in recall accuracy such that they tend to recall more frequent hangovers than they actually experience. To elucidate this, study 2 assessed concurrent hangover severity in FHP and FHN individuals interviewed the morning after having consumed alcohol. Rather than using the HSS which captures the frequency of experiencing a variety of hangover symptoms over a period of time, study 2 used the Acute Hangover Scale (Rohsenow et al. 2007), an instrument designed to assess hangover severity by capturing the experience of a variety of hangover symptoms in situ. It was predicted that hangover severity would not be increased in FHP individuals compared with FHN individuals in line with the more robust previous studies assessing concurrent hangover in relation to family history.

The distribution of males and females was approximately the same in the FHN and FHP groups, chi-square = 0.120, df = 1, p = 0.729, phi = 0.084. Of the variables listed in Table 2, only age differed across the FHN and FHP groups, t(47) = 2.430, p = 0.019, d = 0.709. Participants had consumed a mean of 87.3 g of alcohol, equivalent to around 1.2 g/kg.

The prediction that hangover severity would not be increased in FHP individuals compared with FHN individuals was supported by the data. Contrastingly, estimated blood alcohol the evening before (eBAC), a variable capturing the quantity and duration of alcohol consumption relative to body mass, was a significant predictor of hangover severity. This replicates some previous findings (Chapman 1970; Huntley et al. 2015; Kruisselbrink et al. 2017), although peak blood alcohol level has sometimes been found not to predict hangover severity (e.g. Ylikahri et al. 1974), probably due to a stronger influence of individual differences in studies with smaller samples (Penning et al. 2010).

Experimenter error led to the omission of one item from the F- and M-SMAST questionnaires, necessitating lowering the criterion for FHP categorization from scores of 5 and above to 4 and above. While ensuring no potential FHPs were lost, others would have been included that would not otherwise have been. Some FHNs might also have not been so categorized had the missing item been present. Still, there is variation in the criteria for identifying FHN and FHP individuals such that FHPs have occasionally been designated with F or M-SMAST scores as low as 2 (Newlin and Pretorious 1990). Furthermore, a well-validated 9-item version of the F and M-SMAST questionnaires omits the same item (Crews and Sher 1992). We conclude that the omission has not significantly impacted the analyses.

Study 1 found greater self-reported hangover frequency the previous year in FHP individuals after controlling for hangover-prone drinking styles. Study 2 found that concurrent hangover severity was not raised in FHP individuals. One possible explanation for this discrepancy is that FHPs experience more variable symptom profiles from one hangover to another. If FHPs have a variable hangover symptom profile over time (that is, a more variable intra-personal hangover symptom profile), then this should manifest as greater variability in hangover symptoms within a group of FHPs at any given point in time. This would be due to some FHPs fluctuating towards reporting more severe symptoms, and others less severe at the time point in question. However, Table 2 indicates similar SDs for the variable AHS across the FHP and FHN groups. As a further check on this possibility, we assessed a tally of hangover symptoms reported via the AHS in study 2. We counted the number of symptoms from the AHS that were reported at a moderate or more severe level and observed that FHN and FHP individuals showed very similar tallies of symptoms (respective means 3.7, 3.9; respective SDs 2.2, 1.9; t[47] = 0.26, p = 0.796). Therefore, this explanation seems unlikely.

Taken together with prior research showing no increase in concurrent hangover incidence in FHPs, it appears that FHP individuals are not more susceptible to hangover severity or incidence per se, but that they evidence a retrospective bias in reporting increased hangover frequency over the previous year. One strength of the present research is that the questionnaires used to assess hangover frequency (study 1) and hangover severity (study 2) share a number of items including the three most commonly reported hangover symptoms: thirst, tiredness and headache (Rohsenow et al. 2007). This overlap of hangover assessment provides some confidence that a similar hangover construct was assessed across the two studies and consequently strengthens the case that the differing frames of reference (retrospective vs. immediate) underlie the findings. Nevertheless, future research could test the retrospective bias hypothesis more directly by asking drinkers to monitor drinking and hangover symptoms concurrently over a period of time, then retrospectively to report their experiences for the same period at a later time. If FHPs have a retrospective bias, they should show larger discrepancies between the self-monitoring and retrospective measures compared to FHNs.

That FHPs may have become sensitized to past hangover experiences is analogous to a variety of cognitive distortions observed among alcohol-dependent individuals. For example, Johnsen et al. (1994) found individuals with AUD showed slowed responses on alcohol-related words within a Stroop task. Our finding is novel because people at risk of AUD are thought to be insensitive to future adverse consequences relative to the short-term gains of intoxication (Cantrell et al. 2008); in other words, it is thought that for some drinkers the perceived positive benefits of consuming alcohol outweigh the influence of negative events. This is supported by data estimating that between 17% (Patrick and Maggs 2011) and 53% (Mallett et al. 2008) of college students rate hangover as not an adverse consequence of drinking.

Yet our data (study 1) suggest that people at familial risk for AUD are well aware of, and indeed may over-estimate, hangover frequency relative to drinking behaviour, compared with FHNs. It is not clear why this might be. One possible explanation of such a bias may be a veneration of hangover among individuals with a propensity for AUD as was suggested in a study of young Norwegian adults who talked about gathering the day after a party as a ritualistic and enjoyable way of coping with the discomfort of a hangover (Fjær 2012). However, guilt about drinking forms one of the 10 questions on the Alcohol Use Disorders Identification Test (AUDIT; Babor et al. (2001) and such guilt may encompass negative aspects of hangover, for example, letting down family members due to incapacitation the morning after drinking. Previously, it has been found that talking about hangovers in expressing intentions to avoid negative consequences of alcohol was marginally predictive of reduced alcohol consumption in college students (Mallett et al. 2015). Exploiting cognitive bias towards overestimating hangover frequency, perhaps foregrounding unambiguously negative aspects of hangover, may help practitioners support AUD-prone individuals in better managing their alcohol consumption. However, this recommendation is offered only tentatively pending further research assessing the extent to which negative aspects of hangover underlie such bias.

One limitation of both studies is that, while the analyses identified specific variables predicting hangover frequency and severity, the R ² values associated with these models indicate that only a modest amount of variance in hangover, between 13 and 16%, was explained. It is known that factors beyond blood alcohol and family history of alcohol use disorder influence hangover outcomes, for example, guilt about drinking (Harburg et al. 1993). Further research should assess psychosocial influences on hangover alongside risk factors for AUD.

In addition, study 2 could be argued to be under-powered to detect differences in hangover severity across the FHP and FHN groups. However, the 17 FHPs in study 2 are comparable with the 16 and 20 FHPs in the two previous studies that purport to show a relationship between concurrent hangover severity and family history status (McCaul et al. 1991; Span and Earlywine 1999). Furthermore, the effect size of the FHP vs. FHN difference in hangover severity (AHS) score was d = 0.03, which is a very small effect size. Power analyses indicate that in excess of N = 34,000, participants would be required to detect an effect of this size, for alpha = 0.05 and power = 0.80. On this basis, a lack of statistical power seems an unlikely explanation for the null effects of study 2.

A further limitation is that the increased hangover frequency over the previous 12 months reported by FHP individuals in study 1 could be an artefact of inadequate control of alcohol consumption. However, given this is the fourth survey finding an increased frequency in retrospectively reported hangover among FHP individuals, and that study 1 controlled for hangover-prone drinking (quantity and frequency), this too seems unlikely. It is relevant that none of the retrospective estimates of alcohol consumption used in study 1 mediated the relationship between family history status and hangover frequency, but rather family history and alcohol consumption levels explained separate aspects of hangover frequency.

In conclusion, people at risk of developing AUD due to a positive family history do not appear to experience concurrent hangovers as any more severe than people with no family history yet FHPs tend retrospectively to overestimate hangover frequency. Further research should assess the extent to which negative aspects of hangover underlie this apparent cognitive bias. If negative aspects of hangover are important then practitioners may be able to exploit this to help AUD-prone individuals manage their alcohol consumption more effectively.