image-removebg-preview

Physiology-based relapse prevention tools are creating excitement, but can they work?

  • Home
  • Research
  • Physiology-based relapse prevention tools are creating excitement, but can they work?

Researchers are increasingly looking to identify biomarkers of relapse risk that could drive the next generation of digital interventions to support relapse prevention. This study used wearable biosensors to monitor people in early recovery, examining associations between craving, physiology, and substance use that could inform and drive the future of relapse prevention tools.

Stay on the frontiers of
recovery science
with the free, monthly
Recovery Bulletin

l

WHAT PROBLEM DOES THIS STUDY ADDRESS?

There has been tremendous growth in the digital health tools, with numerous smartphone apps now available that can support substance use disorder recovery. The best of these apps may be particularly useful because they can support individuals’ day-to-day recovery efforts between therapy sessions and mutual-help meetings. At the same time, these apps have a significant limitation – to be effective they require end-users have the wherewithal to use them in moments of high risk.

This is a problem because many people in early recovery struggle to identify and label emotional states and lack awareness of situational factors that increase risk for substance use lapses. To build on these existing recovery-supportive smartphone apps, biosensors embedded in smartwatches and fitness monitors might be harnessed to monitor the physiological signatures of emotional arousal, like heart rate and skin conductance (a measurement of subtle changes perspiration associated with stress), that may indicate a heightened risk of relapse, and when needed, trigger in-the-moment coaching using these relapse prevention apps.

A major challenge in developing this technology, however, is that changes in heart rate and skin conductance are influenced by many factors in addition to relapse-related emotional arousal. Further, although many studies have shown an association between craving and substance use lapses, the craving/substance use relationship is also known to be influenced by numerous individual and situational factors. In reality, although craving may precipitate a substance use lapse, lapses don’t occur every time someone experiences craving (e.g., people may use cognitive and behavioral coping skills to cope with the craving and avoid relapse). Taken together, even if physiological markers of emotional arousal related to states like craving can be identified, this might have limited utility for preventing substance use lapses.

As a necessary first step for teasing out these issues, the researchers in this study explored associations between self-reported craving, substance use lapses, and physiological arousal measured by heart rate and skin conductance in a sample of people in early recovery from alcohol use disorder.


HOW WAS THIS STUDY CONDUCTED?

This was an observational study including 10 individuals in outpatient treatment for moderate-severe alcohol use disorder in the Netherlands. Participants wore a heart rate and skin conductance monitor (similar to a Fitbit) for 100 days, during which they were also surveyed 8 times a day via a smartphone app which asked them to report on their emotional states and any substance use behaviors.

The researchers had two primary research questions: 1) Does craving lead to substance use, and 2) is craving associated with changes in physiological arousal reflected by heart rate and skin conductance? As a secondary aim, the researchers also explored if contextual factors including movement, stress, social situation, self-belief in one’s ability to cope, and nicotine craving influenced the relationships between craving and substance use, as well as craving and physiological arousal.

Participants were surveyed every 3 hours from 7am to 4am the following day (i.e., 8 surveys per 24hrs), and were given 1 hour to complete surveys after notification. To incentivize survey completion, participants were compensated with €1 for each survey completed. The biomonitor was worn on the wrist throughout the day and removed for charging overnight. Lapses were defined as any substance use since the last survey, and craving was reported on a scale of 0-10. Movement was also assessed using internal sensors in the biosensor device.

Additional self-report measures in the daily surveys included: 1) Positive and negative affect, 2) energy levels, 3) stress, 4) social situations [i.e., ‘no social activity/work’, ‘friend/family’, ‘terrace/restaurant’, ‘party’, or ‘other’], 5) whether hobbies or religious activities were available, 6) whether drinking was permitted, 7) belief in effectiveness of coping skills, and 8) nicotine craving.

Given this study had a very small sample, to explore associations between craving and alcohol use lapses, and craving and physiological arousal, the researchers examined predictor variables (e.g., craving and affect) according to whether they were above or below average, rather than specific scale score. Relatedly, the researchers also used a statistical method that allowed them to create measures of the strength of association between variables with a very small sample.


WHAT DID THIS STUDY FIND?

Compliance with daily surveys was low.

The researchers surveyed participants throughout the day and night, so they assumed some surveys would be missed each day while participants were sleeping. However, even after taking this into consideration, as well as the protracted period of monitoring (100 days), survey response rates were fairly low. After the researchers excluded 3 participants from all their analyses who were deemed to be non-responders, survey response rates ranged from 13% to 82%, with an average of 66%.

The relationship between craving and alcohol use varied greatly.

Of the 10 study participants, 4 did not report any alcohol use lapses. As such, these individuals were excluded from this aspect of the researchers’ analyses. The remaining participants reported between 6 and 28 alcohol use lapses with an average craving rating of 1.45, with scores ranging from 0.40 to 3.87 (total possible range of 0-10).

The association between craving and alcohol use lapses in the 3 hours before assessment ranged from 0.00 (i.e., no association) to 0.24 (i.e., a weak association), suggesting that above average craving did not often precede a lapse. However, the association between craving and alcohol use lapse during an assessment ranged from 0.19 (i.e., a weak association) to 0.90 (i.e., a very strong association), indicating for some participants, craving rarely occurred in the context of a lapse, while for others it was common.

The relationship between craving and heart rate was generally weak.

Associations between craving in the 3 hours prior to assessment and heart rate ranged from 0.00 (i.e., no association) to 0.24 (i.e., a weak association), indicating that on the whole, craving did not have a strong relationship with heart rate. Closer inspection of these associations, however, indicted that for two participants craving usually cooccurred with increases in heart rate (88% and 94% of the time), but at the same time, heart rate increases also commonly occurred in the absence of craving. Findings were similar for the association between craving and heart rate during assessments, with correlations ranging from 0.02 (i.e., virtually no association) to 0.29 (i.e., a small association).

The relationship between craving and skin conductance varied greatly across participants.

Associations between craving in the 3 hours before assessment and skin conductance ranged from 0.00 (i.e., no association) to 0.85 (i.e., a very strong association), reflecting a great deal of variability among study participants. Notably however, closer inspection of these results indicated that associations were weak for most participants, with all except one participant having a correlation less than 0.30. Findings were very similar for the associations between craving and skin conductance during assessments, with correlations ranging from 0.02 (i.e., virtually no association) to 0.84 (i.e., a very strong association). Again, only one participant demonstrated a high correlation with all other having a correlation of 0.32 or less.

Individual characteristics were not found to markedly influence craving/physiological associations.

Movement, positive and negative affect, energy levels, stress, social situations, whether hobbies or religious activities were available, whether drinking was permitted, belief in effectiveness of coping skills, and nicotine craving were not found to markedly impact associations between craving and heart rate, and craving and skin conductance.


WHAT ARE THE IMPLICATIONS OF THE STUDY FINDINGS?

Biosensor-derived relapse prevention tools hold much promise because they could one day potentially alert individuals to relapse risk before they themselves are consciously aware of risk and thereby buffer key substance use disorder relapse vulnerabilities and support recovery. Before this happens, however, research needs to be done to tease out associations between emotional states and substance use lapses, and emotional states and physiological arousal.

The researchers’ findings are perhaps most notable for suggesting that past 3-hour craving is not a strong predictor of subsequent alcohol use lapses, and only two of the individuals in this study generally experienced above average craving during an alcohol use lapse.

Findings are also notable for the large degree of variability among participants in the relationship between craving and physiological measures. Although the association between craving and heart rate and craving and skin conductance was generally fairly weak, a small portion of the sample demonstrated a strong craving/skin conductance relationship. Part of the challenge in developing biosensor-driven relapse prevention tools is going to be teasing out the causes and conditions of such variability among different people. Though the researchers explored individual and situational factors that may influence these associations, nothing notable was found.

While on the whole these findings don’t bode well for the idea of biosensor-derived relapse prevention tools, several major limitations of this study should be considered. 1) This was a pilot study with a very small sample size. Larger studies are needed to see if the researchers’ findings replicate. 2) Heart rate is a somewhat crude measure of autonomic nervous system/emotional arousal and is unlikely to be markedly affected by subtle changes in craving

on the order observed in this sample. 3) The researchers did not appear to remove movement artefacts from the raw physiological recordings meaning that this could have muddied the waters as to the true picture. Given the accuracy of wrist-worn biosensor devices like the one used in this study are highly impacted by movement, it would be wise for future studies to incorporate approaches that ensure better reliability of physiological data.

It should be also noted that a number of studies using more sophisticated techniques to manage movement artefacts have shown fairly good accuracy in identifying emotional arousal from heart rhythm recordings. For instance, other researchers have previously been able to achieve stress detection accuracy of 88% using laboratorybased studies, and 72% using ambulatory heart rhythm monitoring like the authors of this study. These findings suggest that more sophisticated approaches may ultimately bear fruit and may pave the way for future biosensor-derived relapse prevention tools.


  1. Compliance with daily surveys was low, decreasing confidence in the quality of the survey data.
  2. The Empatica E4 biomonitor uses fairly basic biosensors affixed on a part of the body that’s generally not considered optimal for measuring heart rate or skin conductance (i.e., the wrist).
  3. This study included a very small sample of 10 individuals, of which several were omitted from analyses for poor survey compliance. This work needs to be replicated in much larger samples.
  4. A strength of the Empatica E4 is that it uses the same heart rate measurement technology as commercially available smartwatches and fitness trackers, meaning these findings are more likely to generalize to commonly used devices. At the same time, this study was ultimately aiming to identify an association between craving and heart rate and skin conductance. Ideally this would have initially been done using gold standard measurement practices to first determine if a relationship exists, before seeing if the relationship remains identifiable under sub-optimal recording conditions.
  5. The Empatica E4 biomonitor includes a skin conductance sensor. In reality, this kind of sensor is rarely found in commercially available smartwatches and fitness trackers. As such, even if skin conductance is eventually found to be a biomarker of craving, it will probably have limited utility.
  6. This study focused on heart rate, which is a crude cardiac indicant of emotion-driven autonomic arousal. Heart rate variability—the variance in inter-heartbeat intervals—may have been a better marker the generally subtle changes in craving observed in this study.

BOTTOM LINE

Biosensor-based relapse prevention is innovative and timely, however, these researchers’ preliminary findings speak to the challenges associated with detecting emotional arousal from physiological markers like heart rate and skin conductance. More work is needed before it can be determined if such interventions have a future.


  • For individuals and families seeking recovery: More work is needed before the hope of biosensor-based relapse prevention tools is realized. This doesn’t negate the utility of existing relapse prevention mobile phone apps like ACHESS, which can support individuals seeking addiction recovery. Though these apps are not considered  first-line treatments, they can be a useful supplement to established addiction treatments such as cognitive behavioral therapy, FDA-approved medications, and community-based mutual-help organizations. 
  • For treatment professionals and treatment systems: More work is needed before the hope of biosensor-based relapse prevention tools is realized. This doesn’t negate the utility of existing relapse prevention mobile phone apps like ACHESS, which can support individuals seeking addiction recovery. Though these apps are not thought of as first-line treatments, they can be a useful supplement to established addiction treatments such as cognitive behavioral therapy, FDA-approved medications, and community-based mutual-help organizations. Such apps also have clinician facing dashboards, which can support treatment adherence and patient monitoring. 
  • For scientists: The idea of a biosensor-based relapse prevention tool continues to hold promise, however, these researchers’ preliminary findings speak to the challenges associated with detecting emotional arousal from physiological markers like heart rate and skin conductance. At the same time, other researchers have developed models for real-time emotion detection from psychophysiological indices like heart rate variability. More work is needed, however, to fine tune these models. Additional studies are also needed that tease out the nuanced relationship between craving and substance use.
  • For policy makers: More work is needed before the hope of biosensor-based relapse prevention tools is realized. This doesn’t negate the utility of existing relapse prevention mobile phone apps like ACHESS, which can support individuals seeking addiction recovery. These apps can be a useful supplement to established addiction treatments such as cognitive behavioral therapy, FDA-approved medications, and community-based mutual-help organizations. Increasing access to these apps through legislation supporting reimbursement for technology-based addiction interventions like this has great potential to improve public health.  

CITATIONS

van Lier, H. G., Noordzij, M. L., Pieterse, M. E., Postel, M. G., Vollenbroek-Hutten, M. M. R., de Haan, H. A., & Schraagen, J. M. C. (2022). An ideographic study into physiology, alcohol craving and lapses during one hundred days of daily life monitoring. Addictive Behaviors Reports, 16, 100443. doi: 10.1016/j.abrep.2022.100443


Stay on the Frontiers of
recovery science
with the free, monthly
Recovery Bulletin

l

WHAT PROBLEM DOES THIS STUDY ADDRESS?

There has been tremendous growth in the digital health tools, with numerous smartphone apps now available that can support substance use disorder recovery. The best of these apps may be particularly useful because they can support individuals’ day-to-day recovery efforts between therapy sessions and mutual-help meetings. At the same time, these apps have a significant limitation – to be effective they require end-users have the wherewithal to use them in moments of high risk.

This is a problem because many people in early recovery struggle to identify and label emotional states and lack awareness of situational factors that increase risk for substance use lapses. To build on these existing recovery-supportive smartphone apps, biosensors embedded in smartwatches and fitness monitors might be harnessed to monitor the physiological signatures of emotional arousal, like heart rate and skin conductance (a measurement of subtle changes perspiration associated with stress), that may indicate a heightened risk of relapse, and when needed, trigger in-the-moment coaching using these relapse prevention apps.

A major challenge in developing this technology, however, is that changes in heart rate and skin conductance are influenced by many factors in addition to relapse-related emotional arousal. Further, although many studies have shown an association between craving and substance use lapses, the craving/substance use relationship is also known to be influenced by numerous individual and situational factors. In reality, although craving may precipitate a substance use lapse, lapses don’t occur every time someone experiences craving (e.g., people may use cognitive and behavioral coping skills to cope with the craving and avoid relapse). Taken together, even if physiological markers of emotional arousal related to states like craving can be identified, this might have limited utility for preventing substance use lapses.

As a necessary first step for teasing out these issues, the researchers in this study explored associations between self-reported craving, substance use lapses, and physiological arousal measured by heart rate and skin conductance in a sample of people in early recovery from alcohol use disorder.


HOW WAS THIS STUDY CONDUCTED?

This was an observational study including 10 individuals in outpatient treatment for moderate-severe alcohol use disorder in the Netherlands. Participants wore a heart rate and skin conductance monitor (similar to a Fitbit) for 100 days, during which they were also surveyed 8 times a day via a smartphone app which asked them to report on their emotional states and any substance use behaviors.

The researchers had two primary research questions: 1) Does craving lead to substance use, and 2) is craving associated with changes in physiological arousal reflected by heart rate and skin conductance? As a secondary aim, the researchers also explored if contextual factors including movement, stress, social situation, self-belief in one’s ability to cope, and nicotine craving influenced the relationships between craving and substance use, as well as craving and physiological arousal.

Participants were surveyed every 3 hours from 7am to 4am the following day (i.e., 8 surveys per 24hrs), and were given 1 hour to complete surveys after notification. To incentivize survey completion, participants were compensated with €1 for each survey completed. The biomonitor was worn on the wrist throughout the day and removed for charging overnight. Lapses were defined as any substance use since the last survey, and craving was reported on a scale of 0-10. Movement was also assessed using internal sensors in the biosensor device.

Additional self-report measures in the daily surveys included: 1) Positive and negative affect, 2) energy levels, 3) stress, 4) social situations [i.e., ‘no social activity/work’, ‘friend/family’, ‘terrace/restaurant’, ‘party’, or ‘other’], 5) whether hobbies or religious activities were available, 6) whether drinking was permitted, 7) belief in effectiveness of coping skills, and 8) nicotine craving.

Given this study had a very small sample, to explore associations between craving and alcohol use lapses, and craving and physiological arousal, the researchers examined predictor variables (e.g., craving and affect) according to whether they were above or below average, rather than specific scale score. Relatedly, the researchers also used a statistical method that allowed them to create measures of the strength of association between variables with a very small sample.


WHAT DID THIS STUDY FIND?

Compliance with daily surveys was low.

The researchers surveyed participants throughout the day and night, so they assumed some surveys would be missed each day while participants were sleeping. However, even after taking this into consideration, as well as the protracted period of monitoring (100 days), survey response rates were fairly low. After the researchers excluded 3 participants from all their analyses who were deemed to be non-responders, survey response rates ranged from 13% to 82%, with an average of 66%.

The relationship between craving and alcohol use varied greatly.

Of the 10 study participants, 4 did not report any alcohol use lapses. As such, these individuals were excluded from this aspect of the researchers’ analyses. The remaining participants reported between 6 and 28 alcohol use lapses with an average craving rating of 1.45, with scores ranging from 0.40 to 3.87 (total possible range of 0-10).

The association between craving and alcohol use lapses in the 3 hours before assessment ranged from 0.00 (i.e., no association) to 0.24 (i.e., a weak association), suggesting that above average craving did not often precede a lapse. However, the association between craving and alcohol use lapse during an assessment ranged from 0.19 (i.e., a weak association) to 0.90 (i.e., a very strong association), indicating for some participants, craving rarely occurred in the context of a lapse, while for others it was common.

The relationship between craving and heart rate was generally weak.

Associations between craving in the 3 hours prior to assessment and heart rate ranged from 0.00 (i.e., no association) to 0.24 (i.e., a weak association), indicating that on the whole, craving did not have a strong relationship with heart rate. Closer inspection of these associations, however, indicted that for two participants craving usually cooccurred with increases in heart rate (88% and 94% of the time), but at the same time, heart rate increases also commonly occurred in the absence of craving. Findings were similar for the association between craving and heart rate during assessments, with correlations ranging from 0.02 (i.e., virtually no association) to 0.29 (i.e., a small association).

The relationship between craving and skin conductance varied greatly across participants.

Associations between craving in the 3 hours before assessment and skin conductance ranged from 0.00 (i.e., no association) to 0.85 (i.e., a very strong association), reflecting a great deal of variability among study participants. Notably however, closer inspection of these results indicated that associations were weak for most participants, with all except one participant having a correlation less than 0.30. Findings were very similar for the associations between craving and skin conductance during assessments, with correlations ranging from 0.02 (i.e., virtually no association) to 0.84 (i.e., a very strong association). Again, only one participant demonstrated a high correlation with all other having a correlation of 0.32 or less.

Individual characteristics were not found to markedly influence craving/physiological associations.

Movement, positive and negative affect, energy levels, stress, social situations, whether hobbies or religious activities were available, whether drinking was permitted, belief in effectiveness of coping skills, and nicotine craving were not found to markedly impact associations between craving and heart rate, and craving and skin conductance.


WHAT ARE THE IMPLICATIONS OF THE STUDY FINDINGS?

Biosensor-derived relapse prevention tools hold much promise because they could one day potentially alert individuals to relapse risk before they themselves are consciously aware of risk and thereby buffer key substance use disorder relapse vulnerabilities and support recovery. Before this happens, however, research needs to be done to tease out associations between emotional states and substance use lapses, and emotional states and physiological arousal.

The researchers’ findings are perhaps most notable for suggesting that past 3-hour craving is not a strong predictor of subsequent alcohol use lapses, and only two of the individuals in this study generally experienced above average craving during an alcohol use lapse.

Findings are also notable for the large degree of variability among participants in the relationship between craving and physiological measures. Although the association between craving and heart rate and craving and skin conductance was generally fairly weak, a small portion of the sample demonstrated a strong craving/skin conductance relationship. Part of the challenge in developing biosensor-driven relapse prevention tools is going to be teasing out the causes and conditions of such variability among different people. Though the researchers explored individual and situational factors that may influence these associations, nothing notable was found.

While on the whole these findings don’t bode well for the idea of biosensor-derived relapse prevention tools, several major limitations of this study should be considered. 1) This was a pilot study with a very small sample size. Larger studies are needed to see if the researchers’ findings replicate. 2) Heart rate is a somewhat crude measure of autonomic nervous system/emotional arousal and is unlikely to be markedly affected by subtle changes in craving

on the order observed in this sample. 3) The researchers did not appear to remove movement artefacts from the raw physiological recordings meaning that this could have muddied the waters as to the true picture. Given the accuracy of wrist-worn biosensor devices like the one used in this study are highly impacted by movement, it would be wise for future studies to incorporate approaches that ensure better reliability of physiological data.

It should be also noted that a number of studies using more sophisticated techniques to manage movement artefacts have shown fairly good accuracy in identifying emotional arousal from heart rhythm recordings. For instance, other researchers have previously been able to achieve stress detection accuracy of 88% using laboratorybased studies, and 72% using ambulatory heart rhythm monitoring like the authors of this study. These findings suggest that more sophisticated approaches may ultimately bear fruit and may pave the way for future biosensor-derived relapse prevention tools.


  1. Compliance with daily surveys was low, decreasing confidence in the quality of the survey data.
  2. The Empatica E4 biomonitor uses fairly basic biosensors affixed on a part of the body that’s generally not considered optimal for measuring heart rate or skin conductance (i.e., the wrist).
  3. This study included a very small sample of 10 individuals, of which several were omitted from analyses for poor survey compliance. This work needs to be replicated in much larger samples.
  4. A strength of the Empatica E4 is that it uses the same heart rate measurement technology as commercially available smartwatches and fitness trackers, meaning these findings are more likely to generalize to commonly used devices. At the same time, this study was ultimately aiming to identify an association between craving and heart rate and skin conductance. Ideally this would have initially been done using gold standard measurement practices to first determine if a relationship exists, before seeing if the relationship remains identifiable under sub-optimal recording conditions.
  5. The Empatica E4 biomonitor includes a skin conductance sensor. In reality, this kind of sensor is rarely found in commercially available smartwatches and fitness trackers. As such, even if skin conductance is eventually found to be a biomarker of craving, it will probably have limited utility.
  6. This study focused on heart rate, which is a crude cardiac indicant of emotion-driven autonomic arousal. Heart rate variability—the variance in inter-heartbeat intervals—may have been a better marker the generally subtle changes in craving observed in this study.

BOTTOM LINE

Biosensor-based relapse prevention is innovative and timely, however, these researchers’ preliminary findings speak to the challenges associated with detecting emotional arousal from physiological markers like heart rate and skin conductance. More work is needed before it can be determined if such interventions have a future.


  • For individuals and families seeking recovery: More work is needed before the hope of biosensor-based relapse prevention tools is realized. This doesn’t negate the utility of existing relapse prevention mobile phone apps like ACHESS, which can support individuals seeking addiction recovery. Though these apps are not considered  first-line treatments, they can be a useful supplement to established addiction treatments such as cognitive behavioral therapy, FDA-approved medications, and community-based mutual-help organizations. 
  • For treatment professionals and treatment systems: More work is needed before the hope of biosensor-based relapse prevention tools is realized. This doesn’t negate the utility of existing relapse prevention mobile phone apps like ACHESS, which can support individuals seeking addiction recovery. Though these apps are not thought of as first-line treatments, they can be a useful supplement to established addiction treatments such as cognitive behavioral therapy, FDA-approved medications, and community-based mutual-help organizations. Such apps also have clinician facing dashboards, which can support treatment adherence and patient monitoring. 
  • For scientists: The idea of a biosensor-based relapse prevention tool continues to hold promise, however, these researchers’ preliminary findings speak to the challenges associated with detecting emotional arousal from physiological markers like heart rate and skin conductance. At the same time, other researchers have developed models for real-time emotion detection from psychophysiological indices like heart rate variability. More work is needed, however, to fine tune these models. Additional studies are also needed that tease out the nuanced relationship between craving and substance use.
  • For policy makers: More work is needed before the hope of biosensor-based relapse prevention tools is realized. This doesn’t negate the utility of existing relapse prevention mobile phone apps like ACHESS, which can support individuals seeking addiction recovery. These apps can be a useful supplement to established addiction treatments such as cognitive behavioral therapy, FDA-approved medications, and community-based mutual-help organizations. Increasing access to these apps through legislation supporting reimbursement for technology-based addiction interventions like this has great potential to improve public health.  

CITATIONS

van Lier, H. G., Noordzij, M. L., Pieterse, M. E., Postel, M. G., Vollenbroek-Hutten, M. M. R., de Haan, H. A., & Schraagen, J. M. C. (2022). An ideographic study into physiology, alcohol craving and lapses during one hundred days of daily life monitoring. Addictive Behaviors Reports, 16, 100443. doi: 10.1016/j.abrep.2022.100443


Share this article

l

WHAT PROBLEM DOES THIS STUDY ADDRESS?

There has been tremendous growth in the digital health tools, with numerous smartphone apps now available that can support substance use disorder recovery. The best of these apps may be particularly useful because they can support individuals’ day-to-day recovery efforts between therapy sessions and mutual-help meetings. At the same time, these apps have a significant limitation – to be effective they require end-users have the wherewithal to use them in moments of high risk.

This is a problem because many people in early recovery struggle to identify and label emotional states and lack awareness of situational factors that increase risk for substance use lapses. To build on these existing recovery-supportive smartphone apps, biosensors embedded in smartwatches and fitness monitors might be harnessed to monitor the physiological signatures of emotional arousal, like heart rate and skin conductance (a measurement of subtle changes perspiration associated with stress), that may indicate a heightened risk of relapse, and when needed, trigger in-the-moment coaching using these relapse prevention apps.

A major challenge in developing this technology, however, is that changes in heart rate and skin conductance are influenced by many factors in addition to relapse-related emotional arousal. Further, although many studies have shown an association between craving and substance use lapses, the craving/substance use relationship is also known to be influenced by numerous individual and situational factors. In reality, although craving may precipitate a substance use lapse, lapses don’t occur every time someone experiences craving (e.g., people may use cognitive and behavioral coping skills to cope with the craving and avoid relapse). Taken together, even if physiological markers of emotional arousal related to states like craving can be identified, this might have limited utility for preventing substance use lapses.

As a necessary first step for teasing out these issues, the researchers in this study explored associations between self-reported craving, substance use lapses, and physiological arousal measured by heart rate and skin conductance in a sample of people in early recovery from alcohol use disorder.


HOW WAS THIS STUDY CONDUCTED?

This was an observational study including 10 individuals in outpatient treatment for moderate-severe alcohol use disorder in the Netherlands. Participants wore a heart rate and skin conductance monitor (similar to a Fitbit) for 100 days, during which they were also surveyed 8 times a day via a smartphone app which asked them to report on their emotional states and any substance use behaviors.

The researchers had two primary research questions: 1) Does craving lead to substance use, and 2) is craving associated with changes in physiological arousal reflected by heart rate and skin conductance? As a secondary aim, the researchers also explored if contextual factors including movement, stress, social situation, self-belief in one’s ability to cope, and nicotine craving influenced the relationships between craving and substance use, as well as craving and physiological arousal.

Participants were surveyed every 3 hours from 7am to 4am the following day (i.e., 8 surveys per 24hrs), and were given 1 hour to complete surveys after notification. To incentivize survey completion, participants were compensated with €1 for each survey completed. The biomonitor was worn on the wrist throughout the day and removed for charging overnight. Lapses were defined as any substance use since the last survey, and craving was reported on a scale of 0-10. Movement was also assessed using internal sensors in the biosensor device.

Additional self-report measures in the daily surveys included: 1) Positive and negative affect, 2) energy levels, 3) stress, 4) social situations [i.e., ‘no social activity/work’, ‘friend/family’, ‘terrace/restaurant’, ‘party’, or ‘other’], 5) whether hobbies or religious activities were available, 6) whether drinking was permitted, 7) belief in effectiveness of coping skills, and 8) nicotine craving.

Given this study had a very small sample, to explore associations between craving and alcohol use lapses, and craving and physiological arousal, the researchers examined predictor variables (e.g., craving and affect) according to whether they were above or below average, rather than specific scale score. Relatedly, the researchers also used a statistical method that allowed them to create measures of the strength of association between variables with a very small sample.


WHAT DID THIS STUDY FIND?

Compliance with daily surveys was low.

The researchers surveyed participants throughout the day and night, so they assumed some surveys would be missed each day while participants were sleeping. However, even after taking this into consideration, as well as the protracted period of monitoring (100 days), survey response rates were fairly low. After the researchers excluded 3 participants from all their analyses who were deemed to be non-responders, survey response rates ranged from 13% to 82%, with an average of 66%.

The relationship between craving and alcohol use varied greatly.

Of the 10 study participants, 4 did not report any alcohol use lapses. As such, these individuals were excluded from this aspect of the researchers’ analyses. The remaining participants reported between 6 and 28 alcohol use lapses with an average craving rating of 1.45, with scores ranging from 0.40 to 3.87 (total possible range of 0-10).

The association between craving and alcohol use lapses in the 3 hours before assessment ranged from 0.00 (i.e., no association) to 0.24 (i.e., a weak association), suggesting that above average craving did not often precede a lapse. However, the association between craving and alcohol use lapse during an assessment ranged from 0.19 (i.e., a weak association) to 0.90 (i.e., a very strong association), indicating for some participants, craving rarely occurred in the context of a lapse, while for others it was common.

The relationship between craving and heart rate was generally weak.

Associations between craving in the 3 hours prior to assessment and heart rate ranged from 0.00 (i.e., no association) to 0.24 (i.e., a weak association), indicating that on the whole, craving did not have a strong relationship with heart rate. Closer inspection of these associations, however, indicted that for two participants craving usually cooccurred with increases in heart rate (88% and 94% of the time), but at the same time, heart rate increases also commonly occurred in the absence of craving. Findings were similar for the association between craving and heart rate during assessments, with correlations ranging from 0.02 (i.e., virtually no association) to 0.29 (i.e., a small association).

The relationship between craving and skin conductance varied greatly across participants.

Associations between craving in the 3 hours before assessment and skin conductance ranged from 0.00 (i.e., no association) to 0.85 (i.e., a very strong association), reflecting a great deal of variability among study participants. Notably however, closer inspection of these results indicated that associations were weak for most participants, with all except one participant having a correlation less than 0.30. Findings were very similar for the associations between craving and skin conductance during assessments, with correlations ranging from 0.02 (i.e., virtually no association) to 0.84 (i.e., a very strong association). Again, only one participant demonstrated a high correlation with all other having a correlation of 0.32 or less.

Individual characteristics were not found to markedly influence craving/physiological associations.

Movement, positive and negative affect, energy levels, stress, social situations, whether hobbies or religious activities were available, whether drinking was permitted, belief in effectiveness of coping skills, and nicotine craving were not found to markedly impact associations between craving and heart rate, and craving and skin conductance.


WHAT ARE THE IMPLICATIONS OF THE STUDY FINDINGS?

Biosensor-derived relapse prevention tools hold much promise because they could one day potentially alert individuals to relapse risk before they themselves are consciously aware of risk and thereby buffer key substance use disorder relapse vulnerabilities and support recovery. Before this happens, however, research needs to be done to tease out associations between emotional states and substance use lapses, and emotional states and physiological arousal.

The researchers’ findings are perhaps most notable for suggesting that past 3-hour craving is not a strong predictor of subsequent alcohol use lapses, and only two of the individuals in this study generally experienced above average craving during an alcohol use lapse.

Findings are also notable for the large degree of variability among participants in the relationship between craving and physiological measures. Although the association between craving and heart rate and craving and skin conductance was generally fairly weak, a small portion of the sample demonstrated a strong craving/skin conductance relationship. Part of the challenge in developing biosensor-driven relapse prevention tools is going to be teasing out the causes and conditions of such variability among different people. Though the researchers explored individual and situational factors that may influence these associations, nothing notable was found.

While on the whole these findings don’t bode well for the idea of biosensor-derived relapse prevention tools, several major limitations of this study should be considered. 1) This was a pilot study with a very small sample size. Larger studies are needed to see if the researchers’ findings replicate. 2) Heart rate is a somewhat crude measure of autonomic nervous system/emotional arousal and is unlikely to be markedly affected by subtle changes in craving

on the order observed in this sample. 3) The researchers did not appear to remove movement artefacts from the raw physiological recordings meaning that this could have muddied the waters as to the true picture. Given the accuracy of wrist-worn biosensor devices like the one used in this study are highly impacted by movement, it would be wise for future studies to incorporate approaches that ensure better reliability of physiological data.

It should be also noted that a number of studies using more sophisticated techniques to manage movement artefacts have shown fairly good accuracy in identifying emotional arousal from heart rhythm recordings. For instance, other researchers have previously been able to achieve stress detection accuracy of 88% using laboratorybased studies, and 72% using ambulatory heart rhythm monitoring like the authors of this study. These findings suggest that more sophisticated approaches may ultimately bear fruit and may pave the way for future biosensor-derived relapse prevention tools.


  1. Compliance with daily surveys was low, decreasing confidence in the quality of the survey data.
  2. The Empatica E4 biomonitor uses fairly basic biosensors affixed on a part of the body that’s generally not considered optimal for measuring heart rate or skin conductance (i.e., the wrist).
  3. This study included a very small sample of 10 individuals, of which several were omitted from analyses for poor survey compliance. This work needs to be replicated in much larger samples.
  4. A strength of the Empatica E4 is that it uses the same heart rate measurement technology as commercially available smartwatches and fitness trackers, meaning these findings are more likely to generalize to commonly used devices. At the same time, this study was ultimately aiming to identify an association between craving and heart rate and skin conductance. Ideally this would have initially been done using gold standard measurement practices to first determine if a relationship exists, before seeing if the relationship remains identifiable under sub-optimal recording conditions.
  5. The Empatica E4 biomonitor includes a skin conductance sensor. In reality, this kind of sensor is rarely found in commercially available smartwatches and fitness trackers. As such, even if skin conductance is eventually found to be a biomarker of craving, it will probably have limited utility.
  6. This study focused on heart rate, which is a crude cardiac indicant of emotion-driven autonomic arousal. Heart rate variability—the variance in inter-heartbeat intervals—may have been a better marker the generally subtle changes in craving observed in this study.

BOTTOM LINE

Biosensor-based relapse prevention is innovative and timely, however, these researchers’ preliminary findings speak to the challenges associated with detecting emotional arousal from physiological markers like heart rate and skin conductance. More work is needed before it can be determined if such interventions have a future.


  • For individuals and families seeking recovery: More work is needed before the hope of biosensor-based relapse prevention tools is realized. This doesn’t negate the utility of existing relapse prevention mobile phone apps like ACHESS, which can support individuals seeking addiction recovery. Though these apps are not considered  first-line treatments, they can be a useful supplement to established addiction treatments such as cognitive behavioral therapy, FDA-approved medications, and community-based mutual-help organizations. 
  • For treatment professionals and treatment systems: More work is needed before the hope of biosensor-based relapse prevention tools is realized. This doesn’t negate the utility of existing relapse prevention mobile phone apps like ACHESS, which can support individuals seeking addiction recovery. Though these apps are not thought of as first-line treatments, they can be a useful supplement to established addiction treatments such as cognitive behavioral therapy, FDA-approved medications, and community-based mutual-help organizations. Such apps also have clinician facing dashboards, which can support treatment adherence and patient monitoring. 
  • For scientists: The idea of a biosensor-based relapse prevention tool continues to hold promise, however, these researchers’ preliminary findings speak to the challenges associated with detecting emotional arousal from physiological markers like heart rate and skin conductance. At the same time, other researchers have developed models for real-time emotion detection from psychophysiological indices like heart rate variability. More work is needed, however, to fine tune these models. Additional studies are also needed that tease out the nuanced relationship between craving and substance use.
  • For policy makers: More work is needed before the hope of biosensor-based relapse prevention tools is realized. This doesn’t negate the utility of existing relapse prevention mobile phone apps like ACHESS, which can support individuals seeking addiction recovery. These apps can be a useful supplement to established addiction treatments such as cognitive behavioral therapy, FDA-approved medications, and community-based mutual-help organizations. Increasing access to these apps through legislation supporting reimbursement for technology-based addiction interventions like this has great potential to improve public health.  

CITATIONS

van Lier, H. G., Noordzij, M. L., Pieterse, M. E., Postel, M. G., Vollenbroek-Hutten, M. M. R., de Haan, H. A., & Schraagen, J. M. C. (2022). An ideographic study into physiology, alcohol craving and lapses during one hundred days of daily life monitoring. Addictive Behaviors Reports, 16, 100443. doi: 10.1016/j.abrep.2022.100443


Share this article