Using ecological momentary assessments and Fitbit data to examine daily associations between physical activity, affect, and alcohol cravings in patients with alcohol use disorder

Introduction

Alcohol Use Disorder (AUD) is highly prevalent in the US population (lifetime prevalence: 29.1%, twelve-month prevalence: 13.9%) (1) and associated with significant societal and economic burden (2,3). Despite existing behavioral and pharmacological treatments for AUD, relapse rates remain high (4–6). Therefore, identifying adjunctive intervention approaches during early recovery may improve alcohol outcomes. One such intervention approach, physical activity (PA), has received increasing attention over the last decade as a relapse prevention strategy (7).

To date, few rigorously controlled exercise intervention studies among patients newly entering AUD treatment have been conducted. And, among these few existing studies, the findings for the effect of exercise on drinking outcomes have been equivocal. For example, Giesen and colleagues (8) noted modest positive effects on drinking outcomes, whereas in a more recent review, Hallgren and colleagues (9) concluded that exercise interventions did not significantly reduce alcohol consumption or related problems in adults with AUD. It is important to note that existing studies possess a number of limitations such as focusing on structured aerobic exercise, enrolling primarily male samples, and omitting objective measurement of physical activity (7).

Despite the uncertainty regarding the benefit of exercise on drinking outcomes, continued research in this area is warranted due to the robust effects of physical activity on important contributors to AUD treatment. Negative affect, positive affect, and alcohol craving are associated with alcohol relapse risk in early recovery (10). Physical activity has demonstrated acute and long-term increases in positive affect, decreases in negative affect, and reduced cravings (11–13). For example, in an experimental study of heavy drinkers, 15 minutes of exercise reduced cravings for alcohol significantly more than sitting (13). Further, in patients in early recovery from AUD, single 30-minute bouts (14) and even 10-minute bouts (15) of moderate-intensity exercise also reduced cravings for alcohol, decreased negative affect and increased positive affect.

These findings are consistent with broader literature on the effects of bouts of physical activity on affective changes. For example, experimental studies have found that acute physical exercise reduces negative affect (16–18) and may even have greater impact on increasing positive affect (19). The largest increases in positive affect have been observed when pre-exercise positive affect was low or following low-intensity exercise (20). Interestingly, brief bouts of activity (e.g., 7-minute bouts) were as effective in improving positive affect as longer bouts (e.g., 60 minutes).

However, to date, no study has examined naturalistic physical activity in outpatients receiving AUD treatment. Previous research suggests that patients in AUD treatment are interested in increasing their physical activity, specifically walking, as a strategy to aid sobriety (21). It is not known whether such physical activity at home would result in improved affect and reduced cravings. The use of ecological momentary assessment (EMA) coupled with objective measurement of physical activity provides a stringent methodological approach to detect the effects of physical activity on affect and alcohol craving.

While there are no studies that have employed EMA and objectively measured physical activity among patients with AUD, a number of studies have been conducted in community samples. A pair of EMA studies that included objective assessment of physical activity found no relationship between physical activity, measured by steps taken, and subsequent changes in positive or negative affect among samples of older adults and individuals with depression (23,24). Likewise, a review of 14 EMA studies of naturalistic exercise and affect found inconsistent evidence that negative affect improved following exercise, but consistent evidence that positive affect improved (22). These effects are similar in pattern but reduced in strength when compared to experimental evidence.

The purpose of this study is to examine the effect of naturalistic engagement in physical activity in the early weeks of recovery in a treatment-seeking sample of patients with AUD. We employed a three-week EMA protocol that assessed positive affect, negative affect, and alcohol cravings, along with objective measurement of steps taken and bouts of physical activity via Fitbit device. We examined associations between physical activity and the three outcomes (i.e., negative affect, positive affect, alcohol cravings) at three different levels: daily, next day, and hourly. We hypothesized that the same relationships would emerge in all three analyses: physical activity would be associated with lower negative affect, higher positive affect, and lower craving for alcohol.

Method

Results

Discussion

The current study examined the effect of naturalistic physical activity (PA), measured as number of steps and bouts of activity recorded on Fitbit devices, on alcohol cravings and affect in a sample of adults recovering from alcohol use disorder (AUD). Results showed that taking more steps in a day than one’s average daily step count was associated with higher positive affect and lower alcohol cravings in the same day, although the effect sizes were very small. The strength of the relationship between step count and positive affect was variable among participants, with most having a positive slope and a few participants having a neutral slope. Recording a bout of PA (at least 10 minutes of consecutive activity) was associated with slightly higher positive affect the next day for some participants, while for others there was no relationship. Contrary to hypotheses, none of the PA variables were predictive of affect or alcohol cravings at the hourly level, though a significant random effect revealed that some participants experienced increases in negative affect in the hour after a bout of physical activity, and others experienced decreases. This variability between subjects may have lead to the null relationship observed in the sample as a whole, though our sample size was also small for between-person comparisons.

Though the effect sizes were small, bouts of moderate-intensity PA were associated with slightly higher positive affect in the same day and the next day, but not at the hourly level. These findings are in contrast to those observed in the experimental literature, where increases in positive affect are typically observed following acute bouts of PA (17, 20). While it is important to interpret our results with caution due to the small sample (i.e., 71 bouts across 14 people), they do point toward important key differences between experimental versus naturalistic examinations of the acute effects of PA on affect. Indeed, there is an emerging literature where EMA has been paired with objective measurement of PA during naturalistic, extended periods (e.g., over 10 days, (23,24) that has resulted in findings that do not consistently align with experimental results. It is possible that demand characteristics inherent to experimental designs influence participants’ reporting of affect during and after PA, whereas this is less likely to affect ratings of affect in an EMA study. In addition, given our naturalistic study design, it is possible that stressful daily events (e.g., hurrying to the bus stop) were classified as a bout of PA, which would likely lead to a different relationship between PA and affect from what is typically observed in experimental studies. Further, we did not find strong evidence that more general physical activity (as measured by steps) was associated with improved mood. It is possible that steps—the only indicator of PA in this study—do not accurately mark bouts of PA, or fail to capture an essential aspect of PA that contributes to PA’s effect on affect in laboratory-based studies (e.g., intention, perceptions). These differences between naturalistic and experimental studies could explain the discrepancy in results. To examine these possibilities, future studies may need to carefully examine physical activity to determine what activity a participant is engaging in, and compare the same participants’ reports during EMA and experimental protocols.

Interestingly, at the hourly level, we observed significant individual variability in the relationship between bouts of PA and negative affect. Some subjects have no relationship between PA and negative affect, and some had a positive relationship. This result could mean that physical activity leads to acute increases in negative affect for some people and no effect for others, but due to the small sample size (only 14 bouts of PA occurred within one hour of an EMA assessment, and these bouts came from only 9 participants), further research with more participants is needed to confirm this result. If this finding is replicated in larger studies, identifying moderators of this relationship may help tailor PA recommendations for those who can most affectively benefit from bouts of PA.

Regarding alcohol cravings, we did not observe strong evidence that alcohol cravings were associated with physical activity, contrary to our hypotheses and some previous literature (13–15). This result is consistent, however, with a recent review of the effects of exercise on alcohol outcomes (9) and new evidence that not all individuals with AUD experience decreased craving following bouts of exercise (43). It is also possible that participants in this naturalistic study were not exposed to alcohol cues, hampering our ability to observe variation in cravings. Future EMA studies should consider including reports of cue exposures that may influence alcohol cravings (e.g., presence of alcohol in their environment or experience of stressful events).

Of note, although our participants took an average of over 8,000 steps per day, very few participants engaged in regular, 10-minute bouts of moderate-intensity PA during the first few weeks following the intensive partial hospitalization program. This was true despite expressed interest in increasing PA as a relapse prevention strategy in this sample, installation of the Fitbit app on participants’ phones, and the provision of a Fitbit activity tracker. This suggests that patients with AUD may need additional support (e.g., PA counseling) to help them increase PA during early recovery that extends beyond using a Fitbit alone. Fitbits have the potential to facilitate self-monitoring and goal setting, two theoretically-relevant strategies associated with increased adherence and maintenance of PA (32). Yet, the efficacy of Fitbits for increasing PA remain to be determined. Preliminary evidence from many patient populations suggest that people wear the device, find it appealing, and show moderate effect sizes in increased steps/day (33,34), but only when coupled with a behavioral intervention such as telephone-delivered “health coaching” (35–38). When Fitbits alone have been given to participants, minimal change in PA is observed (39–41).

There are several limitations that merit discussion. First, unlike emerging studies pairing EMA with objectively measured PA, the current study used a Fitbit device to measure step count rather than accelerometer. While this is indeed a limitation, there is a growing body of evidence suggesting that Fitbit devices offer a valid alternative for assessing PA (42). Second, the sample was small and very homogeneous (24 out of 25 participants identified as White), and only included patients who completed a partial hospital treatment program. As such, the relationships observed here may not generalize well to the broader population of patients with AUD (e.g., it is possible that physical activity may impact alcohol cravings more for people who have not already received treatment). Third, although alcohol use was measured, we were unable to use it as an outcome in this study due to extremely low levels of endorsement, raising questions as to whether the observed relationships would be different if predicting actual use, rather than cravings. Relatedly, we did not measure exposure to alcohol cues, limiting our ability to investigate whether cravings would have been reduced in response to alcohol cues, as has been found in experimental studies. Fourth, the EMA completion rate in this study was below many other studies at 62% (Jones et al., 2019), and there is evidence that participants completed more assessments on more active days. This could be a sign of reactivity, since participants knew they were enrolled in a study monitoring physical activity. We did not employ any methods to reduce reactivity in this protocol. Lastly, despite being in the early days of sobriety, mean levels of negative affect and craving were low, potentially impacting the ability of PA to influence these constructs. Similarly, although the sample did not engage in frequent bouts of physical activity (i.e., consecutive minutes of moderate-intensity stepping), base step count rates were still relatively high (over 8,000/day), and it is possible that increases in PA may have different effects in less physically active samples.

Conclusion

The current study used EMA to examine the daily, next day, and hourly effects of naturalistic physical activity (PA), as measured by Fitbit devices, on negative and positive affect and alcohol cravings in a sample of patients in the first three weeks of recovery from alcohol use disorder (AUD). We found that positive affect was slightly higher in the same day and the day after PA for many individuals, but not all participants had this relationship. Similarly, at the hourly and daily levels, we found that negative affect was positively associated with bouts of PA for some people and negatively associated for others. PA was associated with slightly lower alcohol cravings at the daily level, but not hourly or next day. Future research is necessary to understand for whom naturalistic bouts of PA are associated with improvements in mood, and the extent to which individuals with AUD specifically benefit from PA in their natural environment.