Relapse in alcohol use disorder (AUD) is a significant deterrent on one’s road to recovery. Relapse substantially diminishes the impact of management strategies, especially in the treatment-seeking population. Understanding and addressing relapse predictors is of paramount importance in addressing this risk. In addition to the compelling evidence of the association of craving with relapse to drinking,[1–4] craving reduction can explain the treatment effects in AUD. Moreover, assessment of craving offers a potential opportunity to intervene through anti-craving medications and counseling for craving management.
A recent review on relapse in AUD highlighted craving and other factors like the severity of dependence, comorbidities of psychiatric and other substance use disorders, and negative emotions as potential associations with alcohol relapse. Interestingly, all the studies assessing craving in this review are from Western populations. We wondered whether craving assessment in the Indian population is feasible and has predictive utility in drinking outcomes. Even among treatment-seekers, alarmingly high relapse rates form a solid rationale for assessing craving as a potential factor for relapse prevention.[7–9]
In the Indian context, studies ranging in sample sizes (23–112) assessed the Penn Alcohol Craving Scale to capture craving.[10–12] Although all of these studies aimed to identify relapse, none followed up with the subjects for a longer time to assess its dynamic nature: whether craving changed over time and, most importantly, whether craving predicted relapse in subsequent visits. There are varying views among clinical practitioners treating alcohol use disorder regarding the clinical usefulness of craving assessments. This is because of mixed evidence on whether or not a strong association exists between craving and alcohol use. The initial hesitation to discuss craving among patients seeking care for alcohol use disorder in our setting adds further complexity to this dimension. There is no clarity on whether craving in the severe AUD population can be an objective treatment outcome. While consumption measures are considered outcome measures in all the treatment facilities, we see a lacuna in using craving assessments in this context. Also, in order to have a more comprehensive understanding of craving, a recent review suggested assessing temporal fluctuations, in addition to frequency and intensity.
To explore the feasibility of craving assessments before treatment initiation and subsequent follow-ups, we conducted a study among participants with severe AUD-seeking treatment at a tertiary treatment facility. Here, we show that high craving is associated with fewer abstinent days and lesser time to relapse during AUD treatment; most importantly, craving responds to treatment. Our findings demonstrate the utility of craving assessment in identifying relapse risk in an outpatient facility in the treatment-seeking AUD population. This will point us to identify an at-risk population for future relapse in AUD, and make better-targeted treatment approaches possible.
MATERIAL AND METHODS
Participants characteristics and craving data
This study was conducted at the outpatient de-addiction facility of a tertiary care center. This study was part of a PhD thesis on exploring epigenetic correlates of chronic alcohol consumption and relapse. Thus, the participants were restricted to males to avoid sex-dependent effects on the epigenome.
Consecutive participants seeking treatment for AUD at the outpatient department were invited to participate in the study. The institutional ethics committee approved the study, and written informed consent was obtained from all participants willing to participate. Any comorbid substance use other than nicotine was excluded.
Diagnosis of alcohol use disorder was based on the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5) criteria for AUD. All subjects had severe AUD as per the DSM-5 and endorsed craving criterion. Two medical professionals, including a psychiatrist, assessed all the participants. At this first contact, demographic details and alcohol consumption–related measures were acquired. The latter included a quantity-frequency assessment using the National Institute on Alcohol Abuse and Alcoholism - Quantity Frequency (NIAAA-QF) questionnaire, dependence severity using the Severity of Alcohol Dependence Questionnaire (SADQ), age at first alcohol use, and age of dependence. We calculated last month’s alcohol consumption to obtain a single measure of recent alcohol consumption. We achieved this from the typical drink/day and the number of drink days in a week (drinks/day × drink days in a week × 10) collected from the NIAAA-QF questionnaire. We calculated the dependence duration by subtracting the age of dependence from the age at the time of presentation. Pack-years (pack used per day × years of smoking) were collected for smoking history.
Penn Alcohol Craving Scale (PACS)
Craving was assessed according to the PACS. PACS has five items that capture the intensity of craving for alcohol from the frequency of thoughts on drinking, strength of craving, length of time spent in craving, difficulty in resisting alcohol when available, and subjective overall craving intensity in the past week. All of our participants spoke Tamil or Kannada language. A translated version of the questionnaire was used to capture craving. The same investigator administered craving assessments for all the participants.
The participants were treated as usual during the study period. Treatment included a detoxification regimen (with benzodiazepines) or anti-craving medications (baclofen, naltrexone, acamprosate) as appropriate. The participants all sought treatment in the outpatient clinic and were given a date for a follow-up visit.
During follow-up visits, treatment outcome was assessed by the participants’ abstinence/relapse/lapse status. We defined abstinence as complete alcohol-free days during the follow-up period. If the participant had relapsed or lapsed to alcohol, alcohol consumption in the interim period was recorded by timeline followback (TLFB) method. From this, we calculated the percentage of abstinent days by dividing abstinent days by follow-up days (i.e., abstinent days/follow-up days × 100). The participants were reassessed for PACS craving and prescribed anti-craving as appropriate.
Only participants for whom baseline total PACS scores were available were included in the final sample (N = 264). Those lost to follow-up in any subsequent visits were censored at the last visit. Those who remained abstinent through the study duration were also censored as the outcome (drinking) did not occur. Only in an event of relapse, the time to relapse is noted as the time to drink.
The schema of the study is provided in Figure 1.
Correlations among demographics and alcohol-related measures, craving at initial presentation, and follow-up measures were assessed using Spearman or Pearson correlation, as appropriate. Strength of correlations were based on the guidelines provided by Cohen et al.
We ran a Cox regression survival analysis with time to drink as the dependent variable to examine whether baseline craving predicted time to drink. We dichotomized craving as high and low based on the recent study identifying ≥15 total PACS scores as an indicator of clinically significant alcohol craving (PACS total score ≥15 was considered to have high craving). Days elapsed from an outpatient visit to the day of resumption of drinking were used as an indicator for days to drink. We hypothesized that high-craving scorers would have fewer days to drink. To control for other confounders that could potentially affect the time to drink, we performed a multivariate Cox regression analysis with age at presentation, current alcohol consumption over the past month, treatment provided, and dependence severity from the SADQ. We dichotomized all the continuous variables based on median splits and treated them as binary variables (median age = 36 years, median SADQ = 27.5, median past month alcohol consumption = 336 grams, median pack-years = 20).
We assessed the reduction in craving scores captured over time to understand whether craving responded to treatment. For this, PACS scores at initial presentation and follow-up visits (1 and 2) were analyzed using repeated measures ANOVA. We also conducted a two-way repeated measures ANOVA for abstainers or those who had relapsed as available from the follow-up treatment outcome. For this, we held craving scores as a dependent variable and time and abstinence status (abstinent/relapsed) as independent variables. No appreciable difference in repeated craving measures was observed when abstainers and those who relapsed at each follow-up were considered.
All analyses were conducted in RStudio. In all the statistical tests, a P value ≤ 0.05 was considered significant.
With a larger aim to assess craving construct in outpatient facilities amongst the treatment-seeking population in severe AUD, we recruited 264 participants with a mean age of 36 years at initial presentation. By assessing the craving at initial presentation and subsequent follow-ups, we aimed to explore the predictive utility of the craving construct in predicting the percentage of days abstinent and the time to relapse.
All the participants reported some craving at initial presentation (range of total PACS scores in our data at initial presentation was 6–28). Detailed baseline information of the participants is available in Table 1. All the participants endorsed the craving criterion in the DSM-5 for AUD diagnosis, and all belonged to the severe AUD spectrum of the DSM-5. The majority belonged to the Hindu religion (98%), were married (68%), and had been dependent on alcohol for an average duration of nine years. The vast majority (95%) consumed alcohol more than half a week, with a typical 134 g per drinking day. Around 68% had a comorbid nicotine use disorder with average pack-years of 100. There were no significant associations between craving at initial presentation and alcohol-related measures, as summarized in Table 2.
Craving measures and follow-up information
After assessing 264 participants at initial presentation, the first follow-up data were available for about half of the baseline participants (N = 139) with a median follow-up of seven days (IQR: 7, 14). About 22% maintained complete abstinence until the first follow-up. The median craving scores at this first follow-up visit were 10 (IQR: 8, 13). The median PDA at follow-up 1 (FU1) was 71% (60%, 92.5%).
On the second follow-up, information was available for 84 participants. At the end of 14 median days of follow-up from the day of initial presentation, about 18% remained abstinent, and the median craving recorded was 3 (0, 5). About 36% (N = 30) reported no craving, as recorded by 0 in total PACS scores. The median PDA at follow-up 2 (FU2) was 67% (50%, 73%).
We had information for two subsequent follow-ups with abstinence and relapse status; however, no craving scores were available for these visits. Information about follow-up visits is available in Table 3.
Censoring and days to drink
The data were censored at the previous visit if any participant was lost to follow-ups and if the event of drinking did not occur, that is, those who remained abstinent. On the whole, the range of follow-up was 0–355 days with mean days to drink being 11.34 days.
High craving is associated with a shorter time to resume drinking in AUD
We performed a Cox regression survival analysis to explore the influence of craving on time to drink [Figure 2]. We found that when considered alone, high craving (PACS total scores ≥ 15) at the time of treatment-seeking significantly increased the risk for fewer days to drink (HR = 1.62, 95% CI = 1.05 to 2.05, P = 0.030) during treatment. The median days to drink were approximately 10 days less with high craving at initial presentation (8 days) than those with lower craving at initial presentation (18 days).
High craving is marginally associated with a shorter time to drink in AUD when controlling for covariates
To investigate the effects of other potential confounders for time to resumption to drinking while on treatment for AUD, we performed a multivariate Cox regression analysis for craving at treatment initiation on relapse event, with age, current alcohol consumption, medication at initial presentation, and dependence severity as covariates [Figure 3]. We observed that the main effect of craving in the final model was marginally significant (HR = 1.54, 95% CI = 0.99 to 2.41, P = 0.057). Among the covariates, we observed that those treated with drugs other than benzodiazepines (anti-craving drugs) reported a reduced risk of relapse (HR = 0.54, 95% CI = 0.33 to 0.86, P = 0.01). All other covariates were non-significant.
Associations among craving and outcome measures
To explore how craving and abstinence during follow-up are associated, we prepared a correlation matrix to depict the inter-correlations among craving scores (at initial presentation and follow-ups) and the percentage of abstinent days during follow-ups [Table 4]. We observed that craving at initial presentation was negatively correlated with immediate follow-up (follow-up 1) abstinent days (r = −0.22, CI: −0.38, −0.04). There were small cross-sectional correlations between craving and abstinent days only for the follow-up visits; that is, craving at follow-up 1 significantly negatively correlated with abstinent days at follow-up 1 (r = −0.23, CI: −0.39, −0.06); similarly for follow-up 2, significantly negatively correlated with craving scores at follow-up 2 (r = −0.28, CI: −0.49, −0.05). This suggests that craving at initial presentation is only associated with the proximal abstinent days, whereas craving is associated with the cross-sectional abstinent status during follow-up.
Craving comes down with treatment initiation
Finally, we ran a repeated measures ANOVA to understand whether craving responded to treatment. We observed that craving was statistically significantly different at three time points: initial presentation and follow-up 1 and 2 (F (2, 166) = 347. 81, P < 0.0001, generalized eta squared = 0.72). Post hoc analyses with a Bonferroni adjustment revealed that all the pairwise differences were statistically significant (P < 0.05) [Figure 4]. No appreciable difference in repeated craving measures was observed when relapsers and abstainers at each follow-up were analyzed separately; P < 0.001 for both abstainers and relapsers.
We did not find any significant differences when exploring the reduction in craving scores over time for abstainers and those who had relapsed at FU1 and FU2. This suggests that, once treatment is initiated, craving decreases with time irrespective of drinking.
Craving, one of the criteria for diagnosing alcohol use disorder, is associated with relapse, as evident in the Western population. However, its utility in severe AUD in the Indian context is yet to be disentangled. We examined whether craving assessments were feasible and useful in a treatment-seeking population while being treated at an outpatient facility in a tertiary care hospital. We found that during treatment for severe AUD, high craving at treatment initiation was associated with fewer abstinence days, and high-craving scorers relapsed sooner than those who had less craving. Most importantly, craving responded to treatment during follow-ups. No longitudinal assessments on craving have been attempted so far in the Indian population for its utility in predicting relapse in treatment-seekers. In this scenario, we place our findings as a validation of longitudinal craving assessments in the Indian population and its predictive utility for proximal relapse risk in severe AUD.
Our findings are consistent with McHugh et al. and Stohs et al. who reported that craving predicts relapse in subsequent follow-ups. In line with this, our study strengthens the current claim that high craving is associated with relapse in AUD and is highly relevant in treatment settings for AUD. In addition to this, our results also suggest that those who have high cravings can relapse quickly. A quicker relapse can bring down the patients’ confidence over AUD treatment and can be viewed as a failure of efforts to stay abstinent. This can be a critical factor in potentially affecting their return to treatment. With immense efforts to retain patients in AUD treatment, assessing craving at treatment initiation can help clinicians identify an at-risk group for early resumption of drinking. If identified to have a high craving at treatment initiation, this group at risk of relapse can benefit from being given a shorter time for follow-up than traditional weekly turn-ups. Checking their relapse status over the phone during the interim period until the next visit is also a viable option. Strengthening the craving reduction by aiding them with psychological interventions might be immensely beneficial.
Our results suggest that craving was negatively associated with abstinent days only proximally, specifically in the week after treatment began. We did not find evidence for associations of craving at initial presentation and abstinent days in subsequent follow-ups. This is in contrast to a study on treatment-seekers in the Indian population. Although said study concluded that craving was associated with relapse after a month, it was limited by sample size and that no anti-craving medication or other treatment had been provided. On the other hand, our study included anti-craving medications for all who completed successful detoxification. The lack of association between craving at initial presentation and abstinent status in subsequent follow-ups could be because of the treatment initiation. This assumption is strengthened by our finding on reduction of craving in subsequent follow-ups and that high craving was no longer predictive of time to relapse when covariates, including medication, are added to the model. We observed that using medications other than benzodiazepines was an indicator of reduced relapse risk. This could be interpreted as a prescription of anti-craving at initial presentation instead of benzodiazepines, which only indicates a better status regarding withdrawal severity at presentation. Thus, an association with reduced relapse risk in anti-craving treatment compared with benzodiazepines is explainable.
We did not find craving to be associated with drinking or dependence-related measures at initial presentation. As Flannery et al. suggested, this indicates a good discriminant validity of the PACS. However, we also observed that craving became more correlated with cross-sectional abstinent days during treatment. This result suggests that the more the number of days a subject remained abstinent, the lesser the craving they reported during follow-up visits. This finding can be leveraged to engage treatment-seekers who feel craving is an overpowering criterion for their continuation of drinking. They can be explained about the possibilities of craving lowering with medication and by encouraging to maintain more abstinent days.
Limitations and strengths
Our study has a few limitations. This study was restricted only to male subjects. The skewed high prevalence of AUD among males and the study being part of the epigenome study restricted only to males limited us in exploring the gender differences in craving and its dynamicity. It was particularly challenging to overcome the initial hesitancy of the participants to discuss craving. However, the authors note that developing a good rapport with the treatment-seeking subjects is a helpful way to elicit craving, particularly in an outpatient setting. Authors underscore that participants are responsive to discussing craving once rapport is established and show greater engagement. A test-retest was not possible as the inherent nature of the study to assess craving in the outpatient setting. Future studies should capture craving construct in in-patient settings to confirm our findings.
Despite these limitations, our results build on the existing evidence of craving and its association with alcohol relapse. We provide new insight into the relationship between craving and time to resume alcohol among treatment-seekers aiming for abstinence. These results should be taken into account when considering outpatient care. While previous research on craving in the Indian context has focused on relapse, we provide evidence from repeated measures assessment of craving construct by capturing its dynamicity and responsiveness to treatment.
In summary, we conclude that craving assessment is highly feasible and valuable in an outpatient facility among the Indian population seeking treatment for AUD. The potential of the craving construct in identifying the immediate risk of relapse is further strengthened in this study. Also, the usefulness of anti-craving medications in mitigating craving and, thus, its associated risk of relapse will be helpful for clinicians treating those with AUD.
Financial support and sponsorship
The parent epigenetic study was funded partly by the TVS foundation, India, and by a grant from the Department of Science and Technology (DST), Government of India. None of the affiliated institutions (ICMR) or DST, or TVS foundation played a further role in the study design, collection, analysis, or interpretation of data, in writing and in the decision to submit the paper for publication.
Conflicts of interest
There are no conflicts of interest.
The authors acknowledge that all the patients consented to the study. The first author, SS, remembers and appreciates the conversation with the late Dr. Nadia Chaudhri, Professor, Concordia University at the GRC Alcohol and Nervous System 2020 held at Galveston, Texas. Her enthusiasm for the findings and the translational utility she saw in this work were an inspiration to write this research paper.
1. Martins JS, Fogelman N, Wemm S, Hwang S, Sinha R. Alcohol craving and withdrawal at treatment entry prospectively predict alcohol use outcomes during outpatient treatment. Drug Alcohol Depend 2022;231:109253.
2. McHugh RK, Fitzmaurice GM, Griffin ML, Anton RF, Weiss RD. Association between a brief alcohol craving measure and drinking in the following week. Addiction 2016;111:1004–10.
3. Schneekloth TD, Biernacka JM, Hall-Flavin DK, Karpyak VM, Frye MA, Loukianova LL, et al. Alcohol craving as a predictor of relapse
. Am J Addict 2012;21 (Suppl 1):S20–6.
4. Stohs ME, Schneekloth TD, Geske JR, Biernacka JM, Karpyak VM. Alcohol craving predicts relapse
after residential addiction treatment. Alcohol Alcohol 2019;54:167–72.
5. Subbaraman MS, Lendle S, van der Laan M, Kaskutas LA, Ahern J. Cravings as a mediator and moderator of drinking outcomes in the COMBINE Study. Addiction 2013;108:1737–44.
6. Cavicchioli M, Vassena G, Movalli M, Maffei C. Is craving a risk factor for substance use among treatment-seeking individuals with alcohol and other drugs use disorders?A meta-analytic review. Drug Alcohol Depend 2020;212:108002.
7. Soundararajan S, Narayanan G, Agrawal A, Murthy P. Personality profile and short-term treatment outcome in patients with alcohol dependence:A study from South India. Indian J Psychol Med 2017;39:169–75.
8. Nadkarni A, Velleman R, Bhatia U, Fernandes G, D'souza E, Murthy P. Home-detoxification and relapse
prevention for alcohol dependence in low resource settings:An exploratory study from Goa, India. Alcohol 2020;82:103–12.
9. Sharma A, Das K, Sharma S, Ghosh A. Feasibility and preliminary effectiveness of internet-mediated “relapse
prevention therapy”for patients with alcohol use disorder:A pilot study. Drug Alcohol Rev 2022;41:641–5.
10. Holla B, Karthik S, Biswal J, Viswanath B, Jayarajan D, Bharath RD, et al. Brain functional magnetic resonance imaging cue-reactivity can predict baclofen response in alcohol use disorders. Clin Psychopharmacol Neurosc 2018;16:290–301.
11. Rampure R, Inbaraj LR, Elizabeth CG, Norman G. Factors contributing to alcohol relapse
in a rural population:Lessons from a camp-based de-addiction model from rural Karnataka. Indian J Community Med 2019;44:307–12.
12. Kharb R, Shekhawat LS, Beniwal RP, Bhatia T, Deshpande SN. Relationship between craving and early relapse
in alcohol dependence:A short-term follow-up study. Indian J Psychol Med 2018;40:315–21.
13. American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders: Diagnostic and Statistical Manual of Mental Disorders. Fifth. Arlington, VA: American Psychiatric Association. Available from: https://doi.org/10.1176/appi.books.9780890425787
. [Last accessed 2022 Aug 06].
14. Greenfield TK. Ways of measuring drinking patterns and the difference they make:Experience with graduated frequencies. J Subst Abuse 2000;12:33–49.
15. Stockwell T, Murphy D, Hodgson R. The severity of alcohol dependence questionnaire:Its use, reliability and validity. Br J Addict 1983;78:145–55.
16. Flannery BA, Volpicelli JR, Pettinati HM. Psychometric properties of the Penn Alcohol Craving Scale. Alcohol Clin Exp Res 1999;23:1289–95.
17. Sobell LC, Sobell MB. Timeline follow-back Litten RZ, Allen JP. Measuring Alcohol Consumption:Psychosocial and Biochemical Methods. Totowa, NJ: Humana Press; 1992:41–72.
18. Prinja S, Gupta N, Verma R. Censoring in Clinical Trials:Review of Survival Analysis Techniques. Indian J Community Med [Internet 2010;35:217–21. Available from: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2940174/
. [Last accessed on 2022 Nov 17].
19. Cohen J. Statistical Power Analysis for the Behavioral Sciences [Internet]. 2nd ed., Routledge; 1988. Available from: https://doi.org/10.4324/9780203771587
. [Last accessed 2022 Aug 06].