Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2024 Mar 1.
Published in final edited form as: Psychol Addict Behav. 2021 Dec 23;37(2):285–293. doi: 10.1037/adb0000807

Longitudinal Relations Between Physical Activity and Alcohol Consumption Among Young Adults

Craig E Henderson 1, Laian Z Najjar 2, Chelsie M Young 3, J Leigh Leasure 2, Clayton Neighbors 2, Melissa L Gasser 4, Kristen P Lindgren 5
PMCID: PMC9218000  NIHMSID: NIHMS1758908  PMID: 34941330

Abstract

Objective:

Recent research has revealed positive associations between alcohol use and physical activity. However, findings from these studies have been inconsistent, and longitudinal designs have been underutilized. Therefore, the current study examined longitudinal associations between physical activity and alcohol use in a sample of young adults.

Method:

This study is a secondary analysis of 383 college students (57% female) who reported their drinking behaviors at three-month assessments over an approximately two-year period. Self-reported physical activity was examined for the first 9 months, and drinking was assessed over 21 months.

Results:

Analyses revealed that increases in the intensity of physical activity over the first 9 months predicted increases in drinking over the same time period; however, predictions over the subsequent year were non-significant. Conversely, increases in alcohol use over the first 9 months were associated with concurrent increases in duration of physical activity.

Conclusions:

Results extend previous cross-sectional research findings by indicating that positive associations between physical activity and alcohol use also are found longitudinally.

Keywords: drinking patterns, exercise, college students, health behavior

Despite decades of research, the relationship between alcohol and health remains poorly understood. While the relation between high levels of alcohol consumption and negative health outcomes is well-established, there is some evidence suggesting that moderate alcohol consumption may confer health benefits (Criqui, 1994; Fuchs et al., 1995; Klatsky, 1994), although more recent evidence suggests that even moderate amounts of alcohol are associated with some health risks (Knott et al., 2015). The relationship is further complicated when physical activity (PA) is considered. It makes intuitive sense that PA, with its multitude of health benefits, may counteract the deleterious effects of alcohol, and, consistent with that, the link between alcohol and mortality is weaker among physically active people (Perreault et al., 2016).

Generally speaking, health behaviors (such as eating a healthy diet, getting adequate sleep and not smoking) cluster together (Pronk et al., 2004). Alcohol drinking and PA are health behaviors that cluster in a somewhat surprising way—physically active people drink more alcohol. Indeed, multiple studies indicate that PA is positively associated with alcohol consumption across adulthood (Buscemi et al., 2011; Conroy et al., 2014; French et al., 2009; Lisha et al., 2011; Musselman & Rutledge, 2010). Mechanisms underlying this relationship are understudied, although both are rewarding, stress-relieving, and frequently occur socially. Accordingly, we have delineated potential “joint motives”, including stress relief, work hard-play hard, and social connection (Leasure et al., 2015).

The current literature on associations between PA and alcohol use is hampered by a lack of longitudinal studies. To date, the majority of studies have examined cross-sectional, between-person associations between PA and alcohol use and observed a consistent, positive association at the between-person level; that is, individuals who drink more tend to engage in more PA (Buscemi et al., 2011; Conroy et al., 2014; French et al., 2009; Leasure & Neighbors, 2014; Leichliter et al., 1998; Lisha et al., 2011; Piazza-Gardner & Barry, 2012). This trend has also been consistently observed among young adults (Dunn & Wang, 2003; Kokotailo et al., 1996; Leichliter et al., 1998; Nattiv & Puffer, 1991). For instance, a systematic review found that 7 out of 8 studies examining young adult college students and 6 out of 8 studies examining older, non-student adults reported a positive association between PA and alcohol use (Dodge & Clarke, 2018). Importantly though, this relationship appeared to change with higher rates of drinking, exhibiting a curvilinear relationship with heavy drinking (Lisha et al., 2013).

Thus far, only a handful of studies (Abrantes et al., 2017; Cho et al., 2018; Conroy et al., 2014; Graupensperger et al., 2018) have examined both between- and within-person PA and drinking associations over time, and the findings have been mixed. In a 21-day diary study, Conroy and colleagues (2014) reported an association between increased PA and increased same-day alcohol consumption in a community sample of individuals aged 19 to 89. They examined how the PA-alcohol use relation may unfold rapidly over days and also slowly over months. Their results indicated that time of the year (season) did not significantly impact the PA-alcohol relationship. However, day of the week did. In particular, the coupling of PA and alcohol use over three weeks appeared to be uniform across age, gender, and day of the week, with an overall increase in drinking on weekends.

Conversely, Abrantes et al. (2017) examined within-person effects on the PA-alcohol use relation over 90 days using the Timeline Followback (TLFB) for PA and drinking in a sample of young adult college students and revealed a negative within-person association between PA and drinking, indicating that on days when people were more physically active, they consumed less alcohol. Moreover, over the 90 days, increased PA during the week was associated with decreased drinking on weekends.

A third study by Graupensperger et al. (2018) evaluated the longitudinal PA-alcohol use relationship in a sample of college students. They collected data from participants at three time points every three months over an academic year. Their findings indicated a unidirectional association between PA and alcohol consumption. Specifically, they found that more drinking predicted more vigorous PA, but vigorous PA did not significantly predict future drinking.

Although all three studies utilized a longitudinal design, substantial methodological and participant-related differences exist between them. The length of time participants were studied varied from three weeks to nine months and measurements included daily diary, calendar-based prompts (e.g., TLFB), and self-report measures. Although both examined college students, specific discrepancies between Abrantes et al. (2017) and Graupensperger et al. (2018) may be attributed to differing study designs and measures. Further, the yet different findings by Conroy et al. (2014) may be better explained by participant characteristics, as they assessed participants across the lifespan. Studies have demonstrated patterns of alcohol use and physical activity unique to college students whereby drinking typically increases on weekends while physical activity levels are typically higher on weekdays (Finlay et al., 2012; Neighbors et al., 2011).

Nonetheless, college students represent a crucial population in which to explicate the link between PA and alcohol consumption. As emerging adults, their health behaviors are becoming established, with the potential to influence life-long health (reviewed in Leasure et al., 2015). As college is a time of transition, and in light of previous cross-sectional research documenting a positive association between physical activity and alcohol use, our primary aim was to examine the extent to which these cross-sectional findings extend longitudinally. Therefore, the present study examined the PA-alcohol consumption relation using a longitudinal research design over a 21-month period. The current research focused on elucidating the extent to which changes in PA co-occurred with changes in alcohol consumption during the course of 9 months, along with whether changes in PA over the same period of time were associated with subsequent changes in drinking over the next 12 months.

Method

Participants

Analyses for this study were conducted using existing data (Lindgren et al., 2016), which were collected between Fall 2013 and Summer 2015. The larger data set included 506 young adults (57% women) between 18 and 20 years of age (M=18.57, SD=.69) who were attending a large public university in the United States. To be eligible to participate in the study, participants were required to be: (1) full-time undergraduate students enrolled in their first or second year of study, (2) between 18 and 20 years of age, and (3) fluent in English.

The goals of the present research were to examine concurrent and prospective associations between changes in PA and changes in drinking. Participants who reported not drinking at all assessment points were excluded from the present study. This was done because there was no possibility of change in alcohol use for these participants. Although participants may not have engaged in exercise behavior, there were no participants who reported no PA as it was measured in the current study (see below). The resulting sample included 383 students, 57% of which identified as female. Eight percent of participants reported their ethnicity as Hispanic/Latinx. The race composition of participants was 54% White/Caucasian, 28% Asian/Asian American, 12% Multi-racial, 1% Black/African American, 1% American Indian/Alaska Native, and 2% Unknown or Declined to answer.

Procedures

Participants for the initial study were recruited utilizing information provided by the university registrar. Students who met the age (18 to 20 years) and student status (full-time, first- or second-year undergraduate) criteria received email invitations to participate in an online study. The primary study was described as a two-year study relating to alcohol use and cognition. If eligible, potential participants provided their informed consent and then completed the baseline assessment, which was then followed by seven follow-up surveys, each spaced three months apart. The assessments of the primary study comprised physical activity, alcohol use, and both explicit and implicit alcohol associations; however, the latter were not the focus of the present study. Please see Results for details on attrition. Each assessment was completed in an average of 50 minutes. Participants regularly received reminders to complete assessments in the form of emails, text messages (both a maximum of 11) and phone calls (a maximum of 2). In year one, participants were compensated $25 for each of the first three evaluations, $30 for the fourth evaluation, and an additional $5 if they completed all four assessments. In year two, participants were compensated $30 for each of the final four evaluations and an additional $10 if they completed all four assessments for the year. Those who completed fewer than four assessments were offered an additional $5 to complete the final evaluation. The Institutional Review Board where the research was conducted approved all procedures, and all participants provided consent.

Measures

Alcohol consumption.

Alcohol consumption was assessed with the AUDIT-C. The AUDIT-C comprises the first three items of the AUDIT (Babor et al., 1989). These items assess the frequency of drinking; typical quantity; and frequency of heavy drinking, each on a scale of 0 to 4. Scores reflect the sum of the three items with a possible range of 0 to 12. These items do not refer to a specific time frame (e.g., past month), and no specific time frame was given in the instructions for completing the AUDIT. This measure has been previously used without identifying a specific time frame in studies examining change in alcohol consumption (Boschuetz et al., 2020; Rubinsky et al., 2019).

Physical activity.

The International Physical Activity Questionnaire-Short (adapted from Craig et al., 2003) evaluated the regularity of various levels of PA. Participants were asked about the frequency over the past week (in both the number of days and typical minutes per day) with which they had engaged in walking (for at least ten consecutive minutes), moderate physical activities (such as bicycling at a normal pace), and vigorous physical activities (such as heavy lifting or aerobics). The total number of MET minutes was calculated by summing the MET minutes for each of these categories, including walking. The Compendium of Physical Activities (https://sites.google.com/site/compendiumofphysicalactivities/home) defines METs as the ratio of the working metabolic rate relative to the resting metabolic rate, and Craig et al. (2003) provide formulae for translating varying intensities of activities to metabolic equivalents. Heretofore, for simplicity and interpretability, we refer to MET minutes as intensity. Along with MET minutes, we also calculated a variable representing time spent exercising. Vigorous activity was calculated by multiplying days of the week when vigorous activity occurred by the total number of minutes per day. Likewise, moderate activity was calculated by multiplying days of the week when moderate activity occurred by the total number of minutes per day. Our duration of PA was calculated as the sum of these values. Including both PA intensity and duration may provide a more comprehensive understanding of the PA-alcohol use relationship. It may also help bridge any gaps in the literature as to discrepant findings, given that some studies have examined both duration and intensity (French et al., 2009), while others have examined only duration (Abrantes et al, 2017) and some only intensity (Conroy et al., 2014; Graupensperger et al., 2018).

Covariates.

Several variables that have shown consistent associations with young adult drinking patterns were included as covariates. These variables consisted of biological sex, living arrangements (i.e., on campus, fraternity/sorority house, with parents, and other off-campus locations), and participation in official university sports teams. All information was collected through standard demographic and background information forms. With the exception of sports participation, all variables were collected at the baseline assessment. Sports participation was collected at 18- and 21-month follow-up assessments.

Transparency and Openness

The authors conducted an a priori power analysis to determine optimal sample size for testing the specific aims of the research grant that funded the parent study (Lindgren et al., 2016). A post hoc power analysis using an alpha of .05, sample size of 383, and an obtained effect size of approximately d = .50, revealed that power for the current study exceeded .90. We report study inclusion and exclusion criteria, along with all measures used in the study above, and we followed JARS guidelines in preparing the manuscript (Applebaum et al., 2018). All data, analysis code, and research materials are available by emailing the corresponding author. Data were analyzed using Mplus, version 8.1 (Muthén & Muthén, 1998–2021). The study’s design and its analysis were not pre-registered.

Data Analytic Approach

Latent growth curve (LGC) modeling was used to analyze individual changes in alcohol use and PA (measured as intensity [MET minutes] and duration). Although the calculation of MET minutes incorporates exercise time, we follow a precedent in the literature in which the two variables are analyzed and reported separately (Norton et al., 2010; Weinstock et al., 2016; Weinstock et al., 2020). As described below, missing data were accommodated using all available data in a method similar to full information maximum likelihood (FIML) estimation (specifically, Markov Chain-Monte Carlo estimation) under the assumption that the data were missing at random (MAR; Little & Rubin, 2002).

LGC modeling was conducted using Mplus Version 8.1 (Muthén & Muthén, 1998–2021). Given the complexity of the model we were estimating, we originally encountered problems with the model converging. Therefore, we adopted a strategy using Bayesian estimation, which is much more successful in producing model convergence (van de Schoot, et al., 2014). In some applications of Bayesian estimation, researchers can include prior knowledge regarding the distribution of effects that have been observed in previous literature that are combined with the data in hand to produce a posterior distribution (van de Schoot, et al., 2014). Given the lack of information from previous studies on our research question, we used non-informative priors, which entails using only the data from the current study.

Rather than the confidence intervals yielded in conventional statistical testing, Bayesian estimation produces credible intervals, which are set at the same locations in a probability distribution (e.g., the 2.5th and 97.5th percentiles; Ozechowski, 2014). A model fit statistic known as the potential scale reduction (PSR) indicates whether the model converges (or not), with a small PSR value (i.e., PSR < 1.05) indicating that convergence has occurred. Fit indices are also different with Bayesian estimation, as conventional fit indices (e.g., comparative fit index, root mean square error of approximation) are based on maximum likelihood estimation. Mplus uses the posterior predictive p (PPP) value, which is also more sensitive than chi-square in most situations. A PPP value greater than .05 indicates good model fit (Asparouhov & Muthén, 2017).

Evaluation of distributional assumptions for the alcohol use variables revealed that, even after excluding students who never drank, the proportion of non-drinkers at any given time point was relatively large (range from 27% to 37%). Therefore, we examined a series of models incorporating various transformations (e.g., log, square root), as well as using estimation methods designed to account for count data, such as number of drinks, to determine the optimal way to model change in alcohol use. Bayesian Information Criterion values indicated that a two-part growth model resulted in the best fitting model, and therefore, we examined study hypotheses using this modeling approach. Two-part growth models were specifically developed to address non-normality caused by a large number of participants reporting zeros. Two-part models simultaneously estimate the presence or absence of the behavior (drinking) and at the same time a continuous model (natural log-transformed) for those who reported some alcohol use at any time point. Both models are estimated in an LGC framework (Brown et al., 2005; Olsen & Schafer, 2001. The modeling approach was successful in bringing skewness and kurtosis within acceptable levels for applied data analysis (less than 2).

PA data were collected for the first four of eight assessments by study design. Further, as our research questions involved examining associations between short-term growth in PA and short- and long-term growth in alcohol use, change in alcohol use was modeled as two joint time periods (piecewise growth modeling), one representing change in alcohol use over the first year, the other representing change in alcohol use over the second year. Finally, we used a parallel process growth model, which allowed us to examine how change in PA predicted change in alcohol use. Parallel process models extend standard LGCs by simultaneously modeling growth in two behaviors and regressing the slope of one of the growth models on the slope of the other (Greenbaum & Dedrick, 2007; see Figure 1). These associations were examined in two parallel process models, one using MET minutes as the measure of PA, and the other using duration of PA. Biological sex, residence, and sports participation were included as covariates. Along with significance tests, we included standardized regression coefficients (betas) as a measure of effect size.

Figure 1.

Figure 1.

Open in a new tab

Parallel Process Growth Curve Model for Exercise Intensity Predictor and Alcohol Use Outcome

Results

Baseline Descriptive Statistics

Descriptive statistics for baseline study variables are presented in Table 1, and bivariate correlations between PA and alcohol use variables at each assessment point are presented in Table 2. The median frequency of drinking was two to four times per month, and the median number of drinks consumed on a typical drinking day was three to four. Participants reported spending an average of approximately 2 hours engaging in PA (not including walking) per week, during which they expended a little over 4,100 METs. Figure 2 displays trajectories for PA and alcohol use over the periods of time they were assessed. Concurrent associations between alcohol use and PA were, in general, small but statistically significant. All eight assessments were completed by approximately half of the participants. All 383 completed T1, 87% completed T2, 74% completed T3, 76% completed T4, 74% completed T5, 69% completed T6, 67% completed T7, and 65% completed T8. These numbers highly correspond to attrition rates reported by Lindgren et al. (2016).

Table 1.

Descriptive Statistics for Exercise and Alcohol Use Variables

Variable BL 3mo 6mo 9mo 12mo 15mo 18mo 21mo
M (SD) M (SD) M (SD) M (SD) M (SD) M (SD) M (SD) M (SD)
AUDIT-C Score 3.58 (2.89) 3.33 (2.71) 3.13 (2.68) 3.23 (2.39) 2.51 (2.66) 2.28 (2.56) 2.22 (2.49) 2.26 (2.52)
MET minutes 4136.65 (3010.09) 3604.02 (3174.14) 3705.47 (3117.57) 4546.39 (4275.79) N/A N/A N/A N/A
Exercise Timeb 1.69 (1.27) 1.59 (1.39) 1.68 (1.39) 1.98 (1.79) N/A N/A N/A N/A
Open in a new tab

Note: BL = Basline, M = Mean, MET = Metabolic Equivalents, mo = months, SD = Standard Deviation.

a

Time in hours per week.

Table 2.

Bivariate Correlations Between Physical Activity and Alcohol Use at Study Assessment Points

Alc1 Alc2 Alc3 Alc4 Alc5 Alc6 Alc7 Alc8 Met1 Met2 Met3 Met4 Time1 Time2 Time3 Time4
Alc1 1.00
Alc2 0.90 1.00
Alc3 0.88 0.90 1.00
Alc4 0.76 0.82 0.84 1.00
Alc5 0.66 0.69 0.71 0.72 1.00
Alc6 0.60 0.64 0.66 0.65 0.85 1.00
Alc7 0.59 0.61 0.63 0.63 0.80 0.83 1.00
Alc8 0.48 0.51 0.51 0.51 0.62 0.71 0.77 1.00
Met1 0.12 0.14 0.14 0.15 0.15 0.14 0.11 0.08 1.00
Met2 0.09 0.10 0.12 0.12 0.08 0.08 0.05 0.04 0.61 1.00
Met3 0.12 0.13 0.16 0.16 0.16 0.16 0.11 0.13 0.52 0.54 1.00
Met4 0.16 0.16 0.15 0.15 0.16 0.14 0.10 0.06 0.33 0.36 0.43 1.00
Time1 0.16 0.18 0.17 0.17 0.14 0.16 0.12 0.10 0.82 0.51 0.42 0.26 1.00
Time2 0.06 0.10 0.11 0.10 0.07 0.07 0.03 <0.01 0.43 0.79 0.40 0.30 0.43 1.00
Time3 0.14 0.11 0.17 0.17 0.17 0.16 0.10 0.08 0.43 0.39 0.82 0.37 0.43 0.35 1.00
Time4 0.18 0.14 0.16 0.15 0.15 0.13 0.09 0.04 0.17 0.17 0.29 0.82 0.12 0.23 0.31 1.00
Open in a new tab

Figure 2.

Figure 2.

Open in a new tab

Change in Physical Activity and Alcohol Use Trajectories Over Time.

Note. Scales for the outcomes have been harmonized to facilitate comparison of the trajectories.

Changes in Alcohol Use and Changes in Physical Activity Over Time

Over both the short- (9 months) and long-term (21 months), a decreasing proportion of participants reported drinking on the AUDIT-C (the categorical part [presence/absence of drinking at any time point] of the two-part model) over time, with approximately 81% reporting any drinking on the AUDIT-C at the first assessment and decreasing to 60% at the last assessment. Results were significant over both short term (first 9 months of the study) (Mean Slope = −.60, standard error [SE] = .14, 95% credibility interval = −0.90 – −0.37) and long term (year 2) intervals (Mean Slope = −.14, SE = .06, CI = −0.26 – −0.01). Among those reporting drinking (the continuous part of the two-part model), consumption stayed relatively similar over the short (Mean Slope = .02, SE = .01, CI = −0.01 – 0.04) and long term (Mean Slope = .02, SE = .01, CI = −0.01 – 0.04). The model converged and the data fit the model reasonably well (PPP = .022).

PA measured by intensity remained relatively stable (Mean Slope = 0.07, SE = .13, CI = −0.17 – 0.33) with participants averaging 6.44 MET minutes (roughly equivalent to the value reported in Table 1 when untransformed) per week and significantly varying around that average (Variance = 0.61, SE = 0.47, CI = 0.05 – 1.77). PA measured by time significantly increased over the 12-month follow-up period (Mean Slope = 0.15, SE = .04, CI = 0.07 – 0.23) with significant variance around that average (Variance = 0.15, SE = 0.06, CI = 0.04 – 0.27). The activity data fit the linear models well (PPP = .087 and .389 for METs and time, respectively).

Associations Between Physical Activity and Alcohol Use Over Time

Increases in PA intensity over the first year were associated with increases in alcohol use over the same time period for both categorical and continuous parts of the model (Categorical: b = 0.79, SE = 0.58, CI = 0.15 – 2.49, ß = .59; Continuous: b = 0.13, SE = 0.10, CI = 0.01 – .41, ß = .38). The model converged and produced adequate model fit (PPP = .037). Effect sizes were in the moderate range. None of the covariates were associated with changes in drinking. However, sex and living arrangement were associated with baseline levels of drinking, with males drinking more than females (b = −0.24, SE = 0.06, CI = −0.36 – −0.12, ß = .18) and those living in Greek housing drinking more than those living on campus (b = 0.86, SE = 0.10, CI = 0.68 – 1.05, ß = .55). In turn, those who reported living on campus drank more than those living with their parents (b = −0.25, SE = 0.11, CI = −0.46 – −0.03, ß = .11). Change in PA intensity over the first year did not continue to predict drinking over the second year of the study (Categorical: b = 0.41, SE = 0.48, CI = −0.41 – 1.53, ß = .39; Continuous: b = 0.01, SE = 0.04, CI = −0.06 – 0.10, ß = .04). Likewise, change in exercise duration over the first year of the study did not continue to predict drinking over the second year (Categorical: b = 0.74, SE =1.28, CI = −1.43 – 3.64, ß = .08; Continuous: b = −0.02, SE = 0.10, CI = −0.21 – 0.19, ß = .04).

Change in Drinking Predicting Change in Physical Activity

Because the statistically significant effects regarding the association between change in PA and change in drinking occurred concurrently (i.e., they did not extend through the 2nd follow-up year), we also examined the extent to which change in drinking over the first 9 months predicted change in PA over the same time period. The model converged and produced good model fit (PPP = .070). An increase in the proportion of individuals reporting any drinking (the categorical part of the model) was associated with increases in PA duration (b = 0.82, SE = 0.30, CI = 0.27 – 1.46, ß = 1.04). None of the covariates examined in the previous models were significantly associated with PA time. Further, the association between increases in drinking and PA intensity was not statistically significant.

Discussion

The purpose of this study was to examine whether alcohol consumption is longitudinally associated with PA among young adults. In this secondary data analysis, college students completed measures of alcohol consumption and PA at multiple assessments every three months over an approximately two-year time span. Although the sample should not be considered a heavy drinking sample (40% of the sample reporting no alcohol consumption at the terminal assessment point), their self-reported alcohol use is consistent with a recent report published by the National Institute of Alcoholism and Alcohol Abuse (2020), which states that 53% of college students reported drinking in the past month. Two-part latent growth curve models revealed that increases in the intensity of PA over 9 months were associated with increases in alcohol use concurrently through the same 9-month period. Further, as the proportion of students reporting any drinking increased, so did the average amount of time they spent exercising. These results are consistent with previous cross-sectional research in that PA and alcohol use are positively correlated, although in the current study, the correlation was quite small. However, they extend previous research by examining these associations longitudinally and by examining the relationship over a more extensive timeframe during an important developmental period, namely the early college years. This is a time during which both alcohol use and PA show a marked change (Calestine et al., 2017; Sher & Rutledge, 2006). The literature also indicates that lifestyle patterns that emerge from this period of development frequently last well into adulthood (Sparling & Snow, 2002; Williams et al., 2002).

The current findings build on extant research on temporal relations between alcohol consumption and PA (Graupensperger et al., 2018) by using two-part longitudinal models that suggest a bidirectional relationship between drinking and PA depending on the PA metric in question. Increases in PA intensity were associated with increases in alcohol use, and, reciprocally, increases in alcohol use were associated with increases in PA time. Using a cross-lagged design, Graupensperger et al. (2018) found that alcohol use was a leading indicator of vigorous PA, in that individuals who drank more engaged in more vigorous PA at later time points. This suggests that, for at least some college students, individuals may become aware of their drinking quantity and frequency and make a conscious decision to counteract negative health effects associated with drinking by engaging in more PA. Furthermore, in light of the current findings indicating that increases in PA intensity were associated with increases in drinking, some caution may be warranted, perhaps in the form of awareness programs. For instance, Weinstock and colleagues have introduced exercise regimens to heavy drinking college students with the goal of decreasing alcohol intake and preventing the development of alcohol use disorders (AUD) among these young emerging adults (Weinstock et al., 2016; Weinstock et al., 2008). But, if increased PA intensity, or as a recent study suggests, physical fitness (Shuval et al., 2021), leads to more drinking, as per the current study findings, then these exercise programs may be counterproductive. Together, these findings warrant more longitudinal research that may further illuminate the directionality of the relation between PA and alcohol consumption.

The findings also suggest more research is needed examining temporal relations between PA and alcohol use to form stronger conclusions regarding which behavior precedes the other and for whom, as recent research suggests that drinking motives, feelings states, and compensatory behaviors moderate PA-alcohol use relations (Abrantes et al., 2017). Additionally, more research focused on the intensity of PA is in order, as the patterns of relations—including the direction of the predictive relation—differed depending on whether PA intensity or PA duration were examined. Such studies should also examine in more depth the context in which drinking and alcohol use occur together. For example, social networking models could be used to study with whom individuals drink and exercise when these activities occur (e.g.., exercise that involves drinking afterward); and motives for engaging in PA and alcohol use jointly (e.g., weight motives; guilt and shame; Dodge & Clarke, 2018).

The findings from the current study also suggest the need for more research on short-term, dynamic (e.g., daily) longitudinal associations between PA and alcohol use. Findings from the limited literature available are inconclusive, with one study suggesting a positive link (Conroy et al., 2014), one a negative link (Abrantes et al., 2017), and two indicating a more complex relation, depending on contextual variables and temporal sequencing (Cho et al., 2018). This line of research is in its infancy but could prove effective in the future development of interventions designed to decrease students’ problematic drinking and facilitate their health.

Limitations

Some noteworthy limitations of the current study should be considered when interpreting its findings. First, the dataset used for this secondary analysis was limited to nine months of PA; whereas alcohol data spans 21 months. Future longitudinal studies should extend the observation period for both behaviors to more accurately explicate the PA-alcohol use association. It is also important to note that PA and drinking were not assessed with identical instructions regarding the time frame. Specifically, the measure of PA asked about the previous week, whereas items assessing drinking did not provide a specific time frame, which may impose limitations on the coupling of the two behaviors. Therefore, future studies providing a more specific timeframe in measures assessing both behaviors may yield a more comprehensive understanding of their longitudinal association. A further limitation regarding the measurement of key variables, the AUDIT-C measures an agglomeration of drinking behaviors, including frequency, quantity, and typical drinking. The current study is an initial foray into examining physical activity-drinking relations longitudinally. Future studies examining relations between physical activity and more specific drinking behaviors (heavy episodic drinking, drinking intensity, drinking quantity, etc.), not to mention aspects of physical activity (e.g., relations at different activity thresholds), are needed to advance this line of research.

Both PA and alcohol use were self-reported through surveys using retrospective recall. Therefore, some errors about accurate recall may have influenced the data; however, previous research supporting the validity of alcohol consumption and PA self-report measures among young, healthy adults partially addresses this limitation (Babor et al., 1989; Craig et al., 2003; Kokotailo, et al., 1996). As the technology exists to collect activity data in real time, future studies will certainly want to take advantage of more objective methods for collecting PA data. Our sample is also limited in generalizability because it focused on following early college-age students over time, and the demographics are not representative of the broader population. Future studies examining PA-alcohol relations across the lifespan are needed to further illuminate the complex relations between PA and alcohol use.

Conclusion

To summarize, there is a need to examine the PA-alcohol consumption association using a variety of research designs, such as more intensive assessment periods embedded in longer timeframes and ecological momentary assessment. Additionally, potential motives, moderators, and other variables that may act as confounding factors should continue to be explored. The field is coming to a greater understanding of the dynamics of the PA-alcohol relation; however, much more work needs to be done to capture how PA and drinking are interrelated within and between individuals. Such research, in turn, has the potential to facilitate the positive benefits PA may pose as an intervention for alcohol use while also identifying those for whom PA may be associated with increased alcohol use.

Public Health Statement:

This longitudinal study indicated that changes in physical activity were positively associated with alcohol use over the same 9-month period but did not predict changes over the subsequent year. The association between concurrent changes suggests that the linkage of alcohol use with exercise may become a lasting connection given that the college years are a period in which lifelong health habits form.

Acknowledgment

Preparation of this article was supported by National Institute on Alcohol Abuse and Alcoholism Grants R21AA026380 (PI: Neighbors) and R01AA021763 (PI: Lindgren).

References

  1. Abrantes AM, Scalco MD, O’Donnell S, Minami H, Read JP (2017). Drinking and exercise behaviors among college students: between- and within-person associations. Journal of Behavioral Medicine, 40, 964–977. doi: 10.1007/s10865-017-9863-x [DOI] [PubMed] [Google Scholar]
  2. Applebaum M, Cooper H, Kline RB, Mayo-Wilson E, Nezu AM, & Rao SM (2018). Journal article reporting standards for quantitative research in psychology: The APA publications and communications board task force report. American Psychologist, 73, 3–25. doi: 10.1037/amp0000191 [DOI] [PubMed] [Google Scholar]
  3. Asparouhov T & Muthén BO (2010). Bayesian analysis using Mplus: Technical implementation. Retrieved from http://www.statmodel.com/dowload/BayesAdvantages18.pdf. Accessed June 22, 2017
  4. Babor TF, De La Fuente MF, Saunders JB, Grant M (1989). AUDIT - The alcohol use disorders identification test: Guidelines for use in primary health care. Geneva, Switzerland: World Health Organization. [Google Scholar]
  5. Boschuetz N, Cheng S, Mei L, & Loy VM (2020). Changes in alcohol use patterns in the United States during COVID-19 pandemic. Wmj, 119(3), 171–176 [PubMed] [Google Scholar]
  6. Brown EC, Catalano RF, Fleming CB, Haggerty KP, Abbott RD (2005). Adolescent substance use outcomes in the raising healthy children project: A two-part latent growth curve analysis. Journal of Consulting and Clinical Psychology, 73, 699–710. doi: 10.1037/11855-007 [DOI] [PubMed] [Google Scholar]
  7. Buscemi J, Martens MP, Murphy JG, Yurasek AM, & Smith AE (2011). Moderators of the relationship between physical activity and alcohol consumption in college students. Journal of American College Health, 59, 503–509. doi: 10.1080/07448481.2010.518326 [DOI] [PubMed] [Google Scholar]
  8. Calestine J, Bopp M, Bopp CM, & Papalia Z (2017). College student work habits are related to physical activity and fitness. International Journal of Exercise Science, 10, 1009–1017. [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Cho D, Armeli S, Weinstock J, & Tennen H (2018). Daily-and person-level associations between physical activity and alcohol use among college students. Emerging Adulthood, 4, 219–222. doi: 10.1177/2167696818809760 [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Conroy DE, Ram N, Pincus AL, Coffman DL, Lorek AE, Rebar AL, & Roche MJ (2014). Daily physical activity and alcohol use across the adult lifespan. Health Psychology, 34, 653–660. doi: 10.1037/hea0000157 [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Craig CL, Marshall AL, Sjöström M, Bauman AE, Booth ML, Ainsworth BE, Pratt M, Ekelund U, Yngve A, Sallis JF, & Oja P (2003). International physical activity questionnaire: 12 country reliability and validity. Medicine and Science in Sports and Exercise, 35, 1381–1395. doi: 10.1249/01.MSS.0000078924.61453.FB [DOI] [PubMed] [Google Scholar]
  12. Criqui MH (1994). Alcohol and the heart: implications of present epidemiologic knowledge. Contemporary Drug Problems, 21, 125–142. doi: 10.1177/009145099402100113 [DOI] [Google Scholar]
  13. Dodge T & Clarke P (2018). Testing weight motives and guilt/shame as mediators of the relationship between alcohol use and physical activity. Addictive Behaviors, 77, 131–136. doi: 10.1016/j.addbeh.2017.09.018 [DOI] [PubMed] [Google Scholar]
  14. Dunn MS & Wang MQ (2003). Effects of physical activity on substance use among college students. American Journal of Health Studies, 18, 126–132. [Google Scholar]
  15. Finlay AK, Ram N, Maggs JL, & Caldwell LL (2012). Leisure activities, the social seekend, and alcohol use: Evidence from a daily study of first-year college students. Journal of Studies on Alcohol and Drugs, 73(2), 250–259. 10.15288/jsad.2012.73.250 [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. French MT, Popovici I, & Maclean JC (2009). Do alcohol consumers exercise more? Findings from a national survey. American Journal of Health Promotion, 24, 2–10. doi: 10.4278/ajhp.0801104 [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Fuchs CS, Stampfer MJ, Colditz GA, Giovannucci MD, Manson JE, Kawachi I, Hunter DJ., Hankinson SE, Hennekens CH, Rosner B, Speizer FE, & Willett WC (1995). Alcohol consumption and mortality among women. New England Journal of Medicine, 332, 1245–1250. doi: 10.1056/NEJM199505113321901 [DOI] [PubMed] [Google Scholar]
  18. Graupensperger S, Wilson O, Bopp M, & Evans MB (2018). Longitudinal association between alcohol use and physical activity in US college students: evidence for directionality. Journal of American College Health, 68, 155–162. doi: 10.1080/07448481.2018.1536058 [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Greenbaum PE & Dedrick RF (2007). Changes in use of alcohol, marijuana, and services by adolescents with serious emotional disturbance: A parallel process growth mixture model. Journal of Emotional and Behavioral Disorders, 15, 21–32. doi: 10.1177/10634266070150010301 [DOI] [Google Scholar]
  20. Klatsky AL (1994). Epidemiology of coronary heart disease: influence of alcohol. Alcoholism: Clinical and Experimental Research, 18, 88–96. doi: 10.1111/j.1530-0277.1994.tb00886.x [DOI] [PubMed] [Google Scholar]
  21. Knott CS, Coombs N, Stamatakis E, & Biddulph JP (2015). All cause mortality and the case for age specific alcohol consumption guidelines: pooled analyses of up to 10 population based cohorts. BMJ, 350, h384. doi: 10.1136/bmj.h384 [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Kokotailo PK, Henry BC, Koscik RE, Fleming MF, & Landry GL (1996). Substance use and other health risk behaviors in collegiate athletes. Clinical Journal of Sport Medicine, 6, 183–189. [DOI] [PubMed] [Google Scholar]
  23. Leasure JL & Neighbors CN (2014). Impulsivity moderates the association between physical activity and alcohol consumption. Alcohol, 48, 361–366. doi: 10.1016/j.alcohol.2013.12.003 [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Leichliter JS, Meilman PW, Presley CA, & Cashin JR (1998). Alcohol use and related consequences among students with varying levels of involvement in college. Journal of American College Health, 46, 257–263. doi: 10.1080/07448489809596001 [DOI] [PubMed] [Google Scholar]
  25. Lindgren KP, Neighbors C, Teachman BA, Baldwin SA, Norris J, Kaysen D, Wiers RW, & Gasser ML (2016). Implicit alcohol associations, especially drinking identity, predict drinking over time. Health Psychology, 35, 908–918. doi: 10.1037/hea0000396 [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Lisha NE, Martens M, & Leventhal AM (2011). Age and gender as moderators of the relationship between physical activity and alcohol use. Addictive Behaviors, 36, 933–936. doi: 10.1016/j.addbeh.2011.04.003 [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Lisha NE, Sussman S, Fapa F, & Leventhal AM (2013). Physical activity and alcohol use disorders. American Journal of Drug & Alcohol Abuse, 39(2), 115–120. doi: 10.3109/00952990.2012.713060 [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Little RJA & Rubin DB (2002). Statistical analysis with missing data. 2nd ed. New York: Wiley. [Google Scholar]
  29. Musselman JR & Rutledge PC (2010). The incongruous alcohol-activity association: physical activity and alcohol consumption in college students. Psychology of Sport and Exercise, 11, 609–618. doi: 10.1016/j.psychsport.2010.07.005 [DOI] [Google Scholar]
  30. Muthén LK & Muthén BO (1998–2021). Mplus user’s guide. 7th ed. Los Angeles, CA: Muthén & Muthén. [Google Scholar]
  31. National Institute on Alcohol Abuse and Alcoholism (2020). Fall semester: A time for parents to discuss the risks of college drinking. https://pubs.niaaa.nih.gov/publications/CollegeFactSheet/CollegeFactSheet.pdf.
  32. Nattiv A & Puffer JC (1991). Lifestyles and health risks of collegiate athletes. The Journal of Family Practice, 33, 585–590. [PubMed] [Google Scholar]
  33. Neighbors C, Atkins DC, Lewis MA, Lee CM, Kaysen D, Mittmann A, Fossos N, & Rodriguez LM (2011). Event specific drinking among college students. Psychology of Addictive Behaviors, 25(4), 702–707. 10.1037/a0024051 [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Norton K, Norton L, & Sadgrove D (2010). Position statement on physical activity and exercise intensity terminology. Journal of Science and Medicine in Sport, 13, 496–502. doi: 10.1016/j.jsams.2009.09.008 [DOI] [PubMed] [Google Scholar]
  35. Olsen MK & Schafer JL (2001). A two-part random-effects model for semicontinuous longitudinal data. Journal of the American Statistical Association, 96, 730–745. Doi 10.1198/016214501753168389 [DOI] [Google Scholar]
  36. Ozechowski TJ (2014). Empirical Bayes MCMC Estimation for modeling treatment processes, mechanisms of change, and clinical outcomes in small samples. Journal of Consulting and Clinical Psychology, 82, 854–867. doi: 10.1037/a0026802 [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Perreault K, Bauman A, Johnson N, Britton A, Rangul V, & Stamatakis E (2016). Does physical activity moderate the association between alcohol drinking and all-cause, cancer and cardiovascular diseases mortality? A pooled analysis of eight British population cohorts. British Journal of Sports Medicine, 51, 651–657. doi: 10.1136/bjsports-2016-096194 [DOI] [PubMed] [Google Scholar]
  38. Piazza-Gardner AK & Barry AE (2012). Examining physical activity levels and alcohol consumption: are people who drink more active? American Journal of Health Promotion, 26, e95–e104. doi: 10.4278/ajhp.100929-LIT-328 [DOI] [PubMed] [Google Scholar]
  39. Pronk NP, Anderson LH, Crain AL, Martinson BC, O’Connor PJ, Sherwood NE, & Whitebird RR (2004). Meeting recommendations for multiple healthy lifestyle factors. Prevalence, clustering, and predictors among adolescent, adult, and senior health plan members. American Journal of Preventive Medicine, 27, 25–33. doi: 10.1016/j.amepre.2004.04.022 [DOI] [PubMed] [Google Scholar]
  40. Rubinsky AD, Chavez LJ, Berger D, Lapham GT, Hawkins EJ, Williams EC, & Bradley KA (2019). Utility of routine alcohol screening for monitoring changes in alcohol consumption. Drug and alcohol dependence, 201, 155–160. doi: 10.1016/j.drugalcdep.2019.03.010 [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. Sher KJ & Rutledge PC (2007). Heavy drinking across the transition to college: Predicting first-semester heavy drinking from precollege variables. Addictive Behaviors, 32, 819–835. doi: 10.1016/j.addbeh.2006.06.024 [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. Shuval K, Leonard D, Chartier K, Barlow CE, Fennis BM, Katz DL, Abel K, Farrell SW, Pavlovic A, & DeFina LF (2021). Fit and tipsy? The interrelationship between cardiorespiratory fitness and alcohol consumption and dependence. Sports and Exercise. Manuscript published ahead of print. [DOI] [PubMed] [Google Scholar]
  43. Sparling PB & Snow TK (2002). Physical activity patterns in recent college alumni. Research Quarterly for Exercise and Sport, 73, 200–205. doi: 10.1080/02701367.2002.10609009 [DOI] [PubMed] [Google Scholar]
  44. van de Schoot R, Kaplan D, Denissen J, Asendorpf JB, Neyer FJ, & van Aken MAG (2014). A gentle introduction to bayesian analysis: Applications to developmental research. Child Development, 85, 842–860. doi: 10.1111/cdev.12169 [DOI] [PMC free article] [PubMed] [Google Scholar]
  45. Weinstock J, Barry D, & Petry NM (2008). Exercise-related activities are associated with positive outcome in contingency management treatment for substance use disorders. Addictive Behaviors, 33, 1072–1075. doi: 10.1016/j.addbeh.2008.03.011 [DOI] [PMC free article] [PubMed] [Google Scholar]
  46. Weinstock J, Petry NM, Pescatello LS, & Henderson CE (2016). Motivational interventions for exercise are not related to reductions in college student drinking. Psychology of Addictive Behaviors, 30, 791–801. doi: 10.1080/07448481003686034 [DOI] [PMC free article] [PubMed] [Google Scholar]
  47. Weinstock J, Petry NM, Pescatello LS, Henderson CE, & Nelson CR (2020). Randomized Clinical Trial of Exercise for Non-Treatment Seeking Adults with Alcohol Use Disorder. Psychology of Addictive Behaviors. Psychology of Addictive Behaviors, 34, 65–75. doi: 10.1037/adb0000506 [DOI] [PMC free article] [PubMed] [Google Scholar]
  48. Williams PG, Holmbeck GN, & Greenley RN (2002). Adolescent health psychology. Journal of Consulting and Clinical Psychology, 70, 828–842. doi: 10.1037/0022-006X.70.3.828 [DOI] [PubMed] [Google Scholar]

RESOURCES