Skip to main content
NCBI home page
Addiction Biology logoLink to Addiction Biology
. 2024 Nov 18;29(11):e70001. doi: 10.1111/adb.70001

N‐acetylcysteine as a treatment for substance use cravings: A meta‐analysis

Emma L Winterlind 1,2,3,4, Samantha G Malone 1,2, Michael R Setzer 1, Mikela A Murphy 1,2,4,3, David Saunders 5, Joshua C Gray 1,✉
PMCID: PMC11572739  PMID: 39556483

Abstract

N‐acetylcysteine (NAC) may serve as a novel pharmacotherapy for substance use and substance craving in individuals with substance use disorders (SUDs), possibly through its potential to regulate glutamate. Though prior meta‐analyses generally support NAC's efficacy in reducing symptoms of craving, individual trials have found mixed results. The aims of this updated meta‐analysis were to (1) examine the efficacy of NAC in treating symptoms of craving in individuals with SUD and (2) explore subgroup differences, risk of bias and publication bias across trials. Database searches of PubMed, Cochrane Library and ClinicalTrials.gov were conducted in June and July of 2023 to identify relevant randomized control trials (RCTs). The meta‐analysis consisted of 9 trials which analysed data from a total of 623 participants. The most targeted substance in the clinical trials was alcohol (3/9; 33.3%), followed by tobacco (2/9; 22.2%) and multiple substances (2/9; 22.2%). Meta‐analysis, subgroup analyses and leave‐one‐out analyses were conducted to examine the treatment effect on craving symptoms and adverse events (AEs). Risk of bias assessments, Egger's tests and funnel plot tests were conducted to examine the risk of bias and publication bias. NAC did not significantly outperform placebo in reducing symptoms of craving in the meta‐analysis (SMD = 0.189, 95% CI = −0.015–0.393). Heterogeneity was very high in the meta‐analysis (99.26%), indicating that findings may have been influenced by clinical or methodological differences in the study protocols. Additionally, results indicate that there may be publication bias present. Overall, our findings are contrary to those of prior meta‐analyses, suggesting a limited impact of NAC on substance craving. However, the high heterogeneity and presence of publication bias identified warrants cautious interpretation of the meta‐analytic outcomes.

Keywords: N‐acetylcysteine, novel pharmacotherapy, substance craving

Prior meta‐analyses have found some support for N‐acetylcysteine as a pharmacological intervention for craving reduction in individuals with substance use disorder, despite individual trials showing mixed results. We conducted an updated meta‐analysis to examine its efficacy in reducing symptoms of craving in individuals with substance use disorder and explore subgroup differences, risk of bias, heterogeneity, and publication bias across trials. Contrary to prior meta‐analyses, we found that N‐acetylcysteine did not reduce craving symptoms, though we similarly identified high heterogeneity.

graphic file with name ADB-29-e70001-g003.jpg

1. INTRODUCTION

Substance use disorders (SUDs) are one of the most prevalent forms of psychopathology with a significant public health disease burden. In 2022 alone, approximately 48.7 million Americans met the criteria for any SUD. 1 Globally, SUDs contribute to multiple measures of public health disease burden, including years of life lost (YLLs), disability‐adjusted life years (DALYs) and years of life lived with disability (YLDs). One global study found that approximately 31.8 million DALYs, 16.7 million YPLs and 15 million YLDs were attributable to drug use in 2016. 2 Additionally, the national economic impact of alcohol and illicit drug misuse is estimated to be $249 and $193 billion, respectively. 3 Despite SUDs posing great public health and economic burdens, current treatments show limited efficacy with relapse rates being as high as 40–60%. 4 , 5 One particularly strong predictor of relapse is craving, defined as a “strong desire or urge to use” a substance. 6(p. 491 ) Indeed, a systematic review and meta‐analysis by Vafaie & Kober 7 found that a 1‐point increase in drug‐related cue and craving indicators more than doubled the odds of substance use or relapse at a later timepoint. Overall, reduction in drug cravings has been proposed as a potentially effective mechanism for reducing substance use. 8 , 9 , 10

Individuals with SUD, when presented with drug cues, generally show altered neurobiological craving and reactivity responses in areas of the brain associated with motivational processes, attentional bias, reward‐seeking and ‐processing and decision‐making. 11 Importantly, greater activation in these brain regions is associated with poorer treatment outcomes and relapse. 12 , 13 Some pharmacological treatments have been found to reduce both neural craving responses and self‐reported craving, thus demonstrating the potential of pharmacological interventions for craving reduction. 14

N‐acetylcysteine (NAC) is an over‐the‐counter (OTC) antioxidant, approved for the treatment of acetaminophen overdose, which may also serve as an effective pharmacological treatment for craving reduction, primarily due to its effect on glutamate functioning and signalling. The dysregulation of glutamate transmission is one of the most highly researched cellular neuroadaptations associated with chronic substance use. 15 , 16 Preclinical models indicate that NAC may reduce drug‐seeking behaviour by modulating the expression of the glutamate transporter 1 and cysteine‐glutamate antiporter, thereby regulating the release of synaptic glutamate in areas of the brain such as the nucleus accumbens core and prefrontal cortex, 15 , 17 , 18 , 19 , 20 leading to reduced reinstatement of drug‐seeking behaviour. These effects have been shown to be long‐lasting. 18 , 20

While several randomized control trials (RCTs) have examined the efficacy of NAC in SUD treatment, including in mitigating craving, many show mixed results. 21 Despite the mixed results found by individual RCTs, meta‐analyses support NAC's efficacy as a pharmacological treatment for craving reduction. 22 , 23 Given the rise in trials focused on the matter, we sought to conduct an updated meta‐analysis with the primary aims of (1) examining the efficacy of NAC in treating symptoms of craving in individuals with SUD and (2) exploring subgroup differences, risk of bias and publication bias across trials.

2. METHODS

2.1. Searching strategy and study selection

One author conducted database and registry searches of PubMed, Cochrane Libraries and ClinicalTrials.gov (clinicaltrials.gov). Database searches were conducted on June 20th and July 17th, 2023. The following search terms were used: “addiction AND N‐acetylcysteine,” “abstinence AND N‐acetylcysteine,” “cessation AND N‐acetylcysteine,” “craving AND N‐acetylcysteine,” “dependence AND N‐acetylcysteine,” “substance use disorder AND N‐acetylcysteine,” “withdrawal AND N‐acetylcysteine” and “addictive disorder AND N‐acetylcysteine.” The search terms were drawn from those of Chang et al., 22 who, to our knowledge, conducted the most recent meta‐analysis on NAC's efficacy in treating symptoms related to SUDs. The “randomized control trial” and “clinical trial” options in PubMed were used to only retrieve randomized control trials (RCTs). The “search word variations” option in Cochrane Libraries was used to broaden the search scope. No language restrictions were applied. Two authors independently reviewed database and registry search results for the study selection process and a third author served as a third reviewer in the case of unresolvable discrepancies between the primary reviewers. Trial inclusion judgments were based on the reviewers' individual assessments of whether trials fit the inclusion and exclusion criteria. Initial agreement between the primary reviewers at the full‐text review stage was 87.1% (Cohen's Kappa [κ] = 0.746), indicating substantial agreement. 24 After discussion, 100% agreement was reached (κ = 1.0). All authors reviewed and agreed upon the final trials included in the meta‐analysis.

2.2. Inclusion and exclusion criteria

We used the following inclusion criteria: (a) randomized, placebo‐controlled trials with no crossover; (b) standardized assessment(s) of craving using validated measures; (c) only adjuvant treatments to NAC allowed were psychosocial or behavioural therapies (e.g., cognitive behavioural therapy, contingency management); and (d) include only participants who meet DSM‐IV, DSM‐5 or DSM‐5‐TR criteria for any SUD. An exception to criterion D was made for tobacco use trials; to be included in our meta‐analysis, trials examining the efficacy of NAC as a treatment for tobacco use had to include an eligibility criterion specifying that participants had to smoke ≥10 cigarettes per day. We excluded pre‐clinical trials, case reports, trial protocols, systematic reviews or meta‐analyses, non‐controlled trials, trials with adjuvant pharmacological or non‐psychosocial or ‐behavioural treatments to NAC, and trials not assessing substance use and craving. We only allowed for psychosocial or behavioural therapies as adjuvant treatments to NAC because these are the standard, evidence‐based treatments for SUD 25 and, in theory, would likely be the ones administered alongside NAC in clinical practice.

2.3. Data extraction

The primary outcome was craving. Baseline and endpoint craving data were independently extracted and then compared by two authors to ensure accuracy. If a trial reported scores at more than one time‐point after baseline we used the scores corresponding to the longest time period for which participants were blinded.

If potentially eligible trials did not include sufficient data or information for inclusion in our meta‐analysis, the authors were contacted three times to retrieve the data or information. Insufficient data or information for inclusion in the meta‐analysis include not reporting post‐intervention scores (means and standardized deviation [SD] or standard error [SE]) for the craving measure(s) used in their trial or not posting trial results to ClinicalTrials.gov (if the trial was only posted to ClinicalTrials.gov). If the authors did not respond after three contact attempts (k = 3) or were unable to provide the data requested (k = 2), the study was excluded (k = 5). One study 26 did not include follow‐up data in their publication but provided it upon request. Trial characteristics such as author names, inclusion criteria, demographics, substance treated, measure(s) used and intervention details were independently extracted by two authors and then compared to ensure accuracy. Trial characteristics are presented in Table 1. The total number of trials included in the present meta‐analysis (k = 9) was slightly lower than the generally recommended number (k = 10 35 , 36 ).

TABLE 1.

Trial characteristics.

Trial Inclusion criteria Baseline N (% male) Mean age (SD) Substance Intervention Measure(s) used (scores reported) Measure description
27 ≥15 cigarettes per day

I: 10 (40)

C: 12 (42)

I: 21.4 (2.07)

C: 20.25 (1.14)

 Tobacco

I: 3 days NAC 1800 mg 2x/day and 1 day 1800 mg 1x/day

C: Placebo

 QSU‐B (subscale and total score) 10‐item self‐report measure of craving with Likert‐type scale items and two subscales
28 Meet DSM‐IV AUD diagnostic criteria

I: 22 (90.9)

C: 24 (91.7)

I: 50.1 (11.3)

C: 56.5 (7.0)

 Alcohol

I: NAC 900 mg/day x 1 week; 1800 mg/day x 1 week, 2,700 mg/day x 1 week; 3,600 mg/day x 5 weeks

C: Placebo

 PACS & OCDS (total scores) 5‐item single‐factor self‐report measure of craving; 14‐item self‐report measure of craving with two subscales
26 Meet DSM‐IV diagnostic criteria for cannabis dependence

I: 45 (64.4)

C: 44 (75)

I: 18.8 (1.6)

C: 18.8 (1.6)

 Cannabis

I: NAC 2400 mg/day, contingency management, and cessation counselling x 8 weeks

C: Placebo, contingency management, and cessation counselling

 MCQ‐SF (subscale and total scores) 12‐item self‐report questionnaire with Likert‐scale items and four subscales
29 Meet DSM‐IV diagnostic criteria for SUD and PTSD or subthreshold PTSD

I: 13 (92.3)

C: 14 (100)

I: 48.2 (8.6)

C: 49.9 (8.1)

 Multiple

I: NAC 2400 mg/day and CBT x 8 weeks

C: Placebo and CBT x 8 weeks

 VAS (3 subscales) Visual Analogue Scale with three subscales (amount, intensity, frequency) of craving over past week
30 ≥1f cigarettes/day; ≥3 FTND

I: 24 (100)

C: 24 (100)

I: 37.04 (9.91)

C: 33.13 (9.55)

 Tobacco

I: NAC 2400 mg/daily x 2 weeks

C: Placebo

 QSU & VAS (total scores)

32‐item self‐report measure of craving with Likert‐type scale items and four subscales; Visual Analogue Scale measuring craving
31 Meet DSM‐5 diagnostic criteria for PTSD and AUD and/or other SUD

I: 49 (89.8)

C: 41 (95.1)

I: 52.99 (11.51)

C: 49.13 (13.23)

 Multiple

I: NAC 2400 mg/day and CBT x 8 weeks

C: Placebo and CBT x 8 weeks

 OCDS (total score) 14‐item self‐report measure of craving with two subscales
32 Meet DSM‐5 diagnostic criteria for methamphetamine dependence

I: 76 (59)

C: 77 (60)

I: 37.5 (8.4)

C: 37.9 (7.9)

 Methamphetamine

I: NAC 2400 mg/day x 12 weeks

C: Placebo x 12 weeks

 CEQ (total score) 10‐item self‐report measure of craving using Visual Analogue Scale assessing intensity and frequency of cravings
33 Meet DSM‐5 criteria for AUD and PTSD or subthreshold PTSD

I: 93 (38.7)

C: 89 (37.1)

I: 40.9 (13.2)

C: 39.7 (12.4)

 Alcohol

I: NAC 2400 mg/day and CBT x 12 weeks

C: Placebo and CBT x 12 weeks

 OCDS (2 subscales) 14‐item self‐report measure of craving with two subscales
34 Meet DSM‐IV diagnostic criteria for AUD

I: 21 (62)

C: 21 (62)

I: 48.5 (9.9)

C: 48.6 (11.7)

 Alcohol

I: NAC 2400 mg/day x 4 weeks

C: Placebo x 4 weeks

 PACS (total score) 5‐item single‐factor self‐report measure of craving
Open in a new tab

Note: For more information on craving measures, see Miele et al. 42 Aggregate scores were calculated for trials which reported only subscale scores or total scores for more than one measure. 29 , 30 , 33 If total scores and subscale scores were reported, 26 , 27 effect size was calculated using total score.

Abbreviations: AUD, alcohol use disorder; C, control; CEQ, Craving Experiences Questionnaire; DSM‐IV, Diagnostic and Statistical Manual of Mental Disorders, 4th edition; DSM‐5, Diagnostic and Statistical Manual of Mental Disorders, 5th edition; FTND, Fägerstrom test for nicotine dependence; I, Intervention; MCQ‐SF, Marijuana Craving Questionnaire Short Form; OCDS, Obsessive Compulsive Drinking Scale; PACS, Penn Alcohol Craving Scale; PTSD, post‐traumatic stress disorder; QSU(‐B), Questionnaire of Smoking Urges (Brief); SUD, substance use disorder; VAS, visual analogue scale.

2.4. Risk of bias assessments

Two authors independently assessed the methodological quality of each study based on the revised Cochrane Risk of Bias (RoB) tool for randomized control trials 37 before coming together to reach consensuses. Another author was assigned as a third reviewer in the case of unresolvable discrepancies between the primary reviewers; however, all discrepancies were able to be resolved through discussion between the two primary reviewers.

2.5. Statistical analysis

The meta‐analysis was conducted in accordance with PRISMA guidelines. 38 The meta‐analysis was not pre‐registered and no study protocol was prepared for publication. The meta‐analysis and subgroup analyses were all performed using the metafor package 39 in R version 4.3.3.

Cohen's d, a type of standardized mean difference (SMD), was independently calculated for each study by two authors and then compared to ensure accuracy. Several trials reported total scores for more than one measure or only subscale scores, but not total scores, for a measure. 28 , 29 , 30 , 33 Rather than selecting only one measure or subscale to include in our meta‐analysis, or treating each measure as an independent study, we calculated an aggregated score using two methods: by aggregating effect sizes in the MAd package in R 40 and by manually totalling scores and dividing by the number of measures or subscales. Both methods resulted in the same aggregated effect size for all the relevant trials. Logarithmic risk ratios (logRRs) were calculated to compare reports of adverse events (AEs) between placebo and intervention groups. Confidence intervals (CIs) were automatically calculated for SMDs and RRs by the metafor package when trials were fitted to a model. The package also calculates several heterogeneity statistics; we focus on the I 2 statistic when reporting the results. The random effects (RE) model was chosen to best account for between‐study variability and heterogeneity.

In the case of one study, 32 Cohen's d for the craving measure was found to be null and, thus, the standard error (SE) was also found to be null. To account for this, we transformed the SE to a small decimal (0.001) while maintaining the null effect size. To account for any bias this may introduce, we conducted a subgroup analysis which excluded this trial but maintained all other trials included in the meta‐analysis (SA1). Moreover, we were interested to see whether the substance type treated may impact the effect of treatment. To examine this, we conducted a subgroup analysis (SA2) similar to the one conducted by Chang et al. 22 which focused on trials examining alcohol to assess whether NAC had a greater treatment effect for alcohol, the substance type with the most recent and greatest number of trials, 28 , 33 , 34 than for other substances. Finally, we conducted a subgroup analysis (SA3) examining group differences in logRRs for AEs. We report on all meta‐analytic and subgroup findings.

First, the meta‐analysis including all nine trials was conducted, and alongside it, the first subgroup analysis excluding McKetin et al. 32 (SA1; Figure 2). The second subgroup analysis examining only trials treating alcohol use was then conducted (SA2; Figure 3). Finally, the subgroup analysis of risk ratios was conducted to compare the risk of AEs between placebo and intervention groups (SA3; Figure 4). Leave‐one‐out‐analyses (LOOAs) were conducted to examine the influence of each study on the overall outcomes of the meta‐analysis and the subgroup excluding McKetin et al. 32 Additionally, publication bias was assessed through the use of Begg's funnel plots and Egger's tests. 41

FIGURE 2.

FIGURE 2

Open in a new tab

Meta‐analysis forest plot. The plot depicts a meta‐analysis and subgroup analysis examining the overall effect of NAC on the primary outcome, as well as a test for subgroup differences. Note that values above the line of no effect indicate favorability towards the intervention (NAC) group.

FIGURE 3.

FIGURE 3

Open in a new tab

Alcohol trials subgroup forest plot. The plot depicts a subgroup analysis comparing trials examining alcohol to all other trials. Note that values above the line of no effect indicate favorability towards the intervention (NAC) group.

FIGURE 4.

FIGURE 4

Open in a new tab

Log risk ratio forest plot. The plot depicts a subgroup analysis of risk of adverse events. Note that values below the line of no effect indicate lower risk of AEs for the intervention (NAC) group.

3. RESULTS

3.1. Trial characteristics

Figure 1 displays the literature screening and selection process. The initial search yielded 955 records, 636 of which were excluded for being duplicates, leaving a total of 319 records for title and abstract screening. Another potentially eligible study 34 was identified by one of the authors after the initial database searches were completed; the study was added to the review process, thus leaving a total of 320 records for title and abstract screening. After titles and abstracts were screened for relevance, a total of 31 records remained for full‐text review. Two trials were excluded for including a pharmacological adjunct to NAC and two for including another ineligible adjuvant treatment, four for being the wrong article type, nine for insufficient assessment(s) of craving (e.g., use of unstandardized assessments, no assessments or assessment only at baseline) and five for having insufficient data following the contact attempts. The two trials excluded for including other non‐pharmacological ineligible adjuvant treatments were associated trials with the same dataset. A total of nine trials were included in the final meta‐analysis. The trials were published between 2013 and 2023 and had sample sizes ranging from 22 to 153, with a total of 699 participants at baseline. However, due to attrition within the trials, the total number of participants analysed was 623.

FIGURE 1.

FIGURE 1

Open in a new tab

PRISMA flow diagram depicting study selection process.

The measures of craving used were all self‐report. The Craving Experiences Questionnaire (CEQ), Obsessive Compulsive Drinking Scale (OCDS) and Penn Alcohol Craving Scale (PACS) assessed generalized craving (i.e., general senses of craving, usually across the past week). The Visual Analogue Scale (VAS), employed by Back et al. 29 and Schulte et al., 30 can be used to measure both generalized or acute (i.e., current or immediate sense of craving) craving. Back et al. 29 used the VAS in a generalized fashion, assessing intensity, frequency and amount of craving over the past week, whereas Schulte et al. 30 did not specifically document in what fashion they used the scale. The Questionnaire of Smoking Urges (QSU) measures acute craving and the Marijuana Craving Questionnaire (MCQ) measures both acute and generalized craving. All measures except the VAS and CEQ are explicitly designed to measure craving for a specific substance. None of the trials used cue‐induction prior to craving assessment and all measures were administered to participants in‐person during follow‐ups. For more information on these and other craving measures, see Miele et al. 42 Table 1 describes trial characteristics in further detail.

3.2. Craving symptoms

Baseline data were reported in all trials. Follow‐up data were reported in eight trials; one study did not report follow‐up data but provided these data upon request. 26 We first conducted the meta‐analysis and SA1. In the meta‐analysis, we found that NAC did not significantly outperform placebo in reducing symptoms of craving (SMD = 0.189, 95% CI = −0.015–0.393). Very high heterogeneity was observed across trials (I 2 = 99.26%). The results of SA1 further confirmed the findings of the meta‐analysis (SMD = 0.217, 95% CI = −0.012–0.446, I 2 = 97.33%). A test for group differences (meta‐analysis versus subgroup) identified no significant differences between the groups (p = 0.87), indicating little to no impact of including the null effect size study on the overall outcome of the meta‐analysis. Figure 2 depicts a forest plot with outcomes of the RE models for the meta‐analysis, subgroup analysis and the test for subgroup differences.

SA2, the subgroup analysis examining only trials which were targeting alcohol use, also yielded insignificant results (SMD = 0.074, 95% CI = −0.277–0.425, I 2 = 97.11%). The test for subgroup differences (alcohol trials versus all other trials) identified no significant differences between the groups (p = 0.43). Figure 3 depicts a forest plot with outcomes of the RE models for SA2 and tests for subgroup differences.

LOOAs were conducted to examine the influence of each study on the outcomes of the meta‐analysis and SA1. The LOOAs identified Morley et al. 34 as a significantly influential study in both the meta‐analysis (SMD = 0.235, 95% CI = 0.025–0.444, p = 0.03) and SA1 (SMD = 0.273, 95% CI = 0.040–0.506, p = 0.02), indicating that if this study were to be excluded from the meta‐analysis or subgroup analysis, it would change the overall results to indicate that NAC did significantly outperform placebo. The LOOA on SA1 also identified Back 33 (SMD = 0.257, 95% CI = 0.002–0.511, p = .049) and Schulte et al. 30 (SMD = 0.257, 95% CI = 0.006–0.508, p = 0.045) as significantly influential trials.

3.3. Adverse events

Eight trials reported on AEs experienced by participants, while one did not. 26 There were 16 serious adverse effects reported in the experimental groups across trials, only one of which was conservatively estimated to have a possible relation to the intervention (syncopal episode 29 ). Gastrointestinal upset (e.g., stomach cramps and discomfort, diarrhea, nausea) and headache were the most commonly reported non‐severe AEs.

SA3 was conducted to examine group differences in RRs for AEs. Though eight trials reported AEs, one study did not report occurrences of AEs between the intervention and control group. 29 Another trial reported a greater number of adverse events than the number of participants for both groups 32 and another reported zero AEs for both groups, 30 thereby making the calculation of risk ratios for these trials impossible. Of note, McKetin et al. 32 appeared to have employed the most thorough assessment of AEs out of all eight trials, coding AEs by Systems Organ Class according to the Medical Dictionary for Regulatory Activities (MedDRA 43 ). Only one other study mentioned a standardized assessment of AEs, 29 whereas all other trials either did not report how they collected data on AEs or simply mentioned asking participants if they experienced any side effects. All in all, four trials were excluded from this subgroup analysis, 26 , 29 , 30 , 32 leaving a total of five trials to be included. The results indicate no significant differences in risk for AEs by group (NAC versus placebo; logRR = 0.90, 95% CI = 0.68–1.19). Figure 4 depicts a forest plot with outcomes of the RE model for the AE subgroup analysis.

3.4. Risk of bias and publication bias

The methodological quality (i.e., risk of bias), as assessed by the Cochrane Risk of Bias tool, 37 of the included trials are described in Table 2 and Figure 5. Regarding overall bias, seven trials were rated as high risk and two as some concerns. The third domain of the Cochrane RoB tool, missing outcome data, was a consistent source of high bias across trials. Additional sources of bias identified by the reviewers include underpowered trials and lack of pre‐specified plans of analysis.

TABLE 2.

Risk of bias judgments.

Trial Domain 1 Domain 2 Domain 3 Domain 4 Domain 5 Overall judgment
27 Low Low Low Low Some concerns Some concerns
28 Some concerns Low High Low Some concerns High
26 Some concerns Low High Some concerns Some concerns High
29 Low High High Low Low High
30 Some concerns High High Low Some concerns High
31 Low Low High Low Low High
32 High Some concerns Low Some concerns Low High
33 Low Low High Low Low High
34 Low Low Low Low Some concerns Some concerns
Open in a new tab

Note: Domain 1: Bias arising from the randomization process; Domain 2: Bias due to deviations from intended intervention; Domain 3: Bias due to missing outcome data; Domain 4: Bias in measurement of the outcome; Domain 5: Bias in selection of the reported result.

FIGURE 5.

FIGURE 5

Open in a new tab

Risk of bias bar plot. The figure depicts overall and domain‐specific risk of bias judgments each included study.

Two Begg's funnel plots were created to assess publication bias of the included trials, one for the meta‐analysis and one for SA1. Since the plots looked virtually identical, only the plot which includes all trials is presented (Figure 6). Both funnels exhibited asymmetry, potentially indicating publication bias. Additionally, Egger's tests 41 were conducted on each funnel plot to further assess publication bias. If significant, results of Egger's tests indicate that publication bias favoring statistically significant or positive trials may be more likely to be present. Both tests indicated statistically significant asymmetry (ps < .05), further indicating publication bias. These findings are contrary to those of a prior meta‐analysis on the subject. 23

FIGURE 6.

FIGURE 6

Open in a new tab

Begg's funnel plot. The plot depicts the outcome of a funnel plot including all trials in the meta‐analysis.

4. DISCUSSION

Our meta‐analysis found that NAC was not superior to placebo in reducing cravings in individuals with SUD. These findings contrast with those of Chang et al. 22 and Duailibi et al., 23 who both found that NAC was superior to placebo in treating substance use cravings and other symptoms associated with SUD (e.g., withdrawal and depression). However, our meta‐analysis included several more recent and large trials, 31 , 32 , 33 , 34 most of which had smaller treatment effect sizes than those included in the prior meta‐analyses.

Smaller, earlier trials are more prone to inflate treatment effects (i.e., ‘small‐study effects’), potentially as a result of lesser methodological quality. 44 This may increase the likelihood of publication bias, the tendency for trials with statistically significant or positive results to be published over trials with null findings. 45 Indeed, the two trials with the largest effect sizes included in this meta‐analysis 27 , 29 also had the smallest sample sizes (Ns = 27 and 22, respectively). We identified significant asymmetry in the funnel plots, a tool considered to be effective in detecting small‐study effects, 46 though it may be considered unreliable when used in meta‐analyses with less than 10 trials. 45 While it appears likely that publication bias primarily influenced the asymmetry, it is worth noting that publication bias is only one of many potential reasons for funnel plot asymmetry (e.g., low methodological quality, selection biases, true heterogeneity, data irregularities, artifact 41 ).

This meta‐analysis is not without its limitations. The most notable limitation was the relatively low number of trials included (k = 9), which is slightly below the recommended number of trials for a meta‐analysis (k = 10 35 , 36 ). This was in part due to the number of studies for which we were unable to acquire sufficient data to include in the meta‐analysis (k = 5). Crossover trials were also excluded after considering the documented risks and limitations of including such trial designs in meta‐analyses. Such risks and limitations include period and carryover effects, insufficient within‐individual treatment comparisons, and higher dropout and heightened impact of missing data on trial outcomes. 47 However, only two small (Ns = 12 and 23) trials were excluded for the sole reason of being of a crossover study design. The lower number of trials included in the meta‐analysis also impacted our ability to explore potential sources of heterogeneity with statistical confidence.

Focusing solely on craving reduction may also have limited our conclusions. We chose to assess craving reduction because of evidence suggesting that it is a notable mechanism of behavior change leading to reductions in substance use, 9 a finding which has been translated across various substance types, including alcohol, cannabis, and cocaine. 8 , 9 , 10 Moreover, our meta‐analysis specifically sought to expand upon the prior meta‐analyses conducted by Chang et al. 22 and Duailibi et al., 23 both of which shared a primary outcome of craving reduction. Still, we recognize that craving reduction is only one of several important areas of addiction management and, as such, focusing solely on craving reduction as the primary outcome of the meta‐analysis limits our ability to understand how NAC may impact other important areas of SUD management and treatment, such as substance use reduction and functional impairment improvement. Moreover, examining other treatment areas such as substance use reduction may have allowed for greater analytical exploration and insight into substance‐specific treatment effects of NAC. To expand upon the findings of the present meta‐analysis, we suggest that future meta‐analyses explore the efficacy of NAC in treating various areas of SUD management, including quantity and frequency of use, abstinence, and relapse prevention.

The high levels of heterogeneity across trials utilized in the meta‐analysis possibly limit our conclusions. The nine trials included in this analysis often did not share methodologies or measurements of craving, making it difficult to draw accurate conclusions from the aggregate results regarding the utility of NAC as a treatment for SUD cravings. Particularly, the variability in craving measures used and the use of aggregated craving scores may have impacted findings. Additionally, because trials vary in the SUD treated, it is difficult to determine whether NAC serves as a more effective treatment for certain substances over others. However, the findings from the subgroup analysis examining only alcohol‐focused trials support the findings of Chang et al., 22 who similarly conducted a subgroup analysis of alcohol‐focused trials and found a nonsignificant treatment effect of NAC on craving, and Duailibi et al., 23 who found that substance type did not moderate treatment effects.

Regarding the high heterogeneity, it is possible that the I 2 estimate for the present meta‐analysis was inflated and results were more homogenous than predicted. Evidence suggests that I 2 demonstrates considerable bias when there are few trials included in analyses. When there is minimal true heterogeneity, this measure often increases it artificially, but when there is large true heterogeneity, this measure often decreases it artificially. 48 Because of this, the I 2 heterogeneity statistic should be interpreted cautiously and the real extent to which heterogeneity impacted our results is unclear.

Another important consideration which may have impacted results is the quality of the NAC product itself. It is possible that there were variations in quality (e.g., purity, potency, preparation) and regulatory status of the NAC products used in the included trials of this meta‐analysis. The quality of NAC products was not systematically assessed in the present study or in prior meta‐analyses. 22 , 23 Though NAC has undergone a rigorous regulatory process in order to become an FDA‐approved medication for the treatment of acetaminophen overdose, it is also a widely available OTC dietary supplement for which the FDA exercises limited enforcement discretion. 49 Since OTC supplements are unregulated by the FDA, without guarantee of purity or medication‐substance content, the effectiveness of a given supplement may be limited. As such, it is recommended that NAC products are sourced from suppliers which meet United States Pharmacopeia (USP) standards. 50

This meta‐analysis corroborated the findings of prior large RCTs with null results, despite prior meta‐analytic evidence indicating that NAC is a promising treatment option for substance craving. Large RCTs will enhance generalizability, reduce potential bias and amplify study power to detect real effects. Similarly, considering homogeneity in trial design (e.g. trial protocols, outcome measures, and the SUD treated) will enhance the ability of RCTs to evaluate the efficacy of NAC as a SUD treatment more accurately. Additionally, researchers should publish negative findings when applicable to help prevent publication bias. These remediations will improve the ability of future meta‐analyses to clarify NAC treatment effects and identify moderators of treatment effects.

AUTHOR CONTRIBUTIONS

Emma L. Winterlind: Database searches; study review and selection; data extraction; data analysis; risk of bias assessments; table and figure creation; writing; revisions. Samantha G. Malone: Study review and selection; risk of bias assessments; writing; revisions. Michael R. Setzer: Data extraction; table and figure creation; revisions. Mikela A. Murphy: Initial database searches; revisions. David Saunders: Revisions. Joshua C. Gray: Conceptualization; writing; revisions; supervision.

CONFLICT OF INTEREST STATEMENT

The authors have no commercial or financial conflicts of interests to disclose.

DISCLAIMER

The opinions and assertions herein are those of the authors and do not necessarily reflect the official views of the Henry M. Jackson Foundation for the Advancement of Military Medicine, Inc. Moreover, the opinions and assertions herein do not necessarily reflect the official views of the Department of Defense, Uniformed Services University, the National Institute on Alcohol Abuse and Alcoholism, the US Government, and do not imply endorsement by the Federal Government.

Winterlind EL, Malone SG, Setzer MR, Murphy MA, Saunders D, Gray JC. N‐acetylcysteine as a treatment for substance use cravings: A meta‐analysis. Addiction Biology. 2025;29(11):e70001. doi: 10.1111/adb.70001

Funding information This work was supported by funding from the National Institute on Alcohol Abuse and Alcoholism (R01AA030041) and the Department of Defense (HU0001‐22‐2‐0066).

DATA AVAILABILITY STATEMENT

The R code used to conduct analyses and create forest and funnel plots can be accessed using this link: https://github.com/ewinterli/NAC-meta-analysis.

REFERENCES

  • 1. Substance Abuse and Mental Health Services Administration . 2022 National Survey on Drug Use and Health (NSDUH) Releases. 2022. [PubMed]
  • 2. Degenhardt L, Charlson F, Ferrari A, et al. The global burden of disease attributable to alcohol and drug use in 195 countries and territories, 1990–2016: a systematic analysis for the global burden of disease study 2016. Lancet Psychiat. 2018;5(12):987‐1012. doi: 10.1016/S2215-0366(18)30337-7 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3. U.S Department of Health and Human Services . Addiction and substance misuse reports and publications. 2023. https://www.hhs.gov/surgeongeneral/reports-and-publications/addiction-and-substance-misuse/index.html
  • 4. Brecht ML, Herbeck D. Time to relapse following treatment for methamphetamine use: a long‐term perspective on patterns and predictors. Drug Alcohol Depend. 2014;139:18‐25. doi: 10.1016/j.drugalcdep.2014.02.702 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5. McLellan AT, Lewis DC, O'Brien CP, Kleber HD. Drug dependence, a chronic medical illness implications for treatment, insurance, and outcomes evaluation. Jama. 2000;284(13):1689‐1695. doi: 10.1001/jama.284.13.1689 [DOI] [PubMed] [Google Scholar]
  • 6. American Psychiatric Association . Diagnostic and Statistical Manual of Mental Disorders. 5th ed. Substance Abuse and Mental Health Services Administration; 2013. doi: 10.1176/appi.books.9780890425596 [DOI] [Google Scholar]
  • 7. Vafaie N, Kober H. Association of drug cues and craving with drug use and relapse: a systematic review and meta‐analysis. JAMA Psychiat. 2022;79(7):641‐650. doi: 10.1001/jamapsychiatry.2022.1240 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8. Sayette MA. The role of craving in substance use disorders: theoretical and methodological issues. Ann Rev Clin Psychol. 2016;12(1):407‐433. doi: 10.1146/annurev-clinpsy-021815-093351 [DOI] [PubMed] [Google Scholar]
  • 9. Stone BM, Gray KM, Tomko RL, McRae‐Clark AL, Sherman BJ. Differential mechanisms of behavior change in cannabis use disorder treatments: functional improvements and clinical implications. Int J Mental Health Addict. 2024;1‐30. doi: 10.1007/s11469-023-01231-7 [DOI] [Google Scholar]
  • 10. Witkiewitz K, Pfund RA, Tucker JA. Mechanisms of behavior change in substance use disorder with and without formal treatment. Annu Rev Clin Psychol. 2022;18(1):497‐525. doi: 10.1146/annurev-clinpsy-072720-014802 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11. Devoto F, Zapparoli L, Spinelli G, Scotti G, Paulesu E. How the harm of drugs and their availability affect brain reactions to drug cues: a meta‐analysis of 64 neuroimaging activation studies. Transl Psychiat. 2020;10(1):429. doi: 10.1038/s41398-020-01115-7 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12. MacNiven KH, Jensen ELS, Borg N, Padula CB, Humphreys K, Knutson B. Association of neural responses to drug cues with subsequent relapse to stimulant use. JAMA Netw Open. 2018;1(8):e186466. doi: 10.1001/jamanetworkopen.2018.6466 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13. Regier PS, Jagannathan K, Franklin TR, et al. Sustained brain response to repeated drug cues is associated with poor drug‐use outcomes. Addict Biol. 2021;26(5):e13028. doi: 10.1111/adb.13028 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14. Cabrera EA, Wiers CE, Lindgren E, Miller G, Volkow ND, Wang GJ. Neuroimaging the effectiveness of substance use disorder treatments. J Neuroimmune Pharmacol. 2016;11(3):408‐433. doi: 10.1007/s11481-016-9680-y [DOI] [PubMed] [Google Scholar]
  • 15. Kalivas PW. The glutamate homeostasis hypothesis of addiction. Nat Rev Neurosci. 2009;10(8):561‐572. doi: 10.1038/nrn2515 [DOI] [PubMed] [Google Scholar]
  • 16. Volkow ND, Michaelides M, Baler R. The neuroscience of drug reward and addiction. Physiol Rev. 2019;99(4):2115‐2140. doi: 10.1152/physrev.00014.2018 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17. Baker DA, McFarland K, Lake RW, et al. Neuroadaptations in cystine‐glutamate exchange underlie cocaine relapse. Nat Neurosci. 2003;6(7):743‐749. doi: 10.1038/nn1069 [DOI] [PubMed] [Google Scholar]
  • 18. Moro F, Giannotti G, Caffino L, et al. Lasting reduction of nicotine‐seeking behavior by chronic N‐acetylcysteine during experimental cue‐exposure therapy. Addict Biol. 2020;25(4):e12771. doi: 10.1111/adb.12771 [DOI] [PubMed] [Google Scholar]
  • 19. Roberts‐Wolfe D, Kalivas P. Glutamate transporter GLT‐1 as a therapeutic target for substance use disorders. CNS Neurol Disord Drug Targets. 2015;14(6):745‐756. doi: 10.2174/1871527314666150529144655 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20. Zhou W, Kalivas PW. N‐acetylcysteine reduces extinction responding and induces enduring reductions in cue‐ and heroin‐induced drug‐seeking. Biol Psychiatry. 2008;63(3):338‐340. doi: 10.1016/j.biopsych.2007.06.008 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21. Greenberg NR, Farhadi F, Kazer B, Potenza M, Gustavo AA. Mechanistic effects and use of N‐acetylcysteine in substance use disorders. Curr Behav Neurosci Rep. 2022;9(4):124‐143. doi: 10.1007/s40473-022-00250-3 [DOI] [Google Scholar]
  • 22. Chang CT, Hsieh PJ, Lee HC, Lo CH, Tam KW, Loh EW. Effectiveness of N‐acetylcysteine in treating clinical symptoms of substance abuse and dependence: a meta‐analysis of randomized controlled trials. Clin Psychopharmacol Neurosci. 2021;19(2):282. doi: 10.9758/cpn.2021.19.2.282 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23. Duailibi MS, Cordeiro Q, Brietzke E, et al. N‐acetylcysteine in the treatment of craving in substance use disorders: systematic review and meta‐analysis. Am J Addict. 2017;26(7):660‐666. doi: 10.1111/ajad.12620 [DOI] [PubMed] [Google Scholar]
  • 24. McHugh ML. Interrater reliability: the kappa statistic. Biochem Med. 2012;22(3):276‐282. doi: 10.11613/bm.2012.031 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25. National Institute on Drug Abuse . Drugs, brains, and behavior: the science of addiction. In: Treatment and Recovery. National Institute on Drug Abuse; 2011. https://nida.nih.gov/publications/drugs-brains-behavior-science-addiction/treatment-recovery [Google Scholar]
  • 26. Roten AT, Baker NL, Gray KM. Marijuana craving trajectories in an adolescent marijuana cessation pharmacotherapy trial. Addict Behav. 2013;38(3):1788‐1791. doi: 10.1016/j.addbeh.2012.11.003 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27. Schmaal L, Berk L, Hulstijn KP, Cousijn J, Wiers RW, Van Den Brink W. Efficacy of N‐acetylcysteine in the treatment of nicotine dependence: a double‐blind placebo‐controlled pilot study. Eur Addict Res. 2011;17(4):211‐216. doi: 10.1159/000327682 [DOI] [PubMed] [Google Scholar]
  • 28. Yoon G. N‐acetylcysteine in alcohol dependence. ClinicalTrials. 2013. https://classic.clinicaltrials.gov/ct2/show/NCT00568087 [Google Scholar]
  • 29. Back SE, McCauley JL, Korte KJ, et al. A double‐blind, randomized, controlled pilot trial of N‐acetylcysteine in veterans with posttraumatic stress disorder and substance use disorders. J Clin Psychiat. 2016;77(11):12406. doi: 10.4088/JCP.15m10239 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30. Schulte MHJ, Goudriaan AE, Kaag AM, et al. The effect of N‐acetylcysteine on brain glutamate and gamma‐aminobutyric acid concentrations and on smoking cessation: a randomized, double‐blind, placebo‐controlled trial. J Psychopharmacol. 2017;31(10):1377‐1379. doi: 10.1177/0269881117730660 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31. Back S. NAC for treating comorbid PTSD and SUD. ClinicalTrials. 2021. https://classic.clinicaltrials.gov/ct2/show/NCT02911285 [Google Scholar]
  • 32. McKetin R, Dean OM, Turner A, et al. N‐acetylcysteine (NAC) for methamphetamine dependence: a randomised controlled trial. EClin Med. 2021;38:101005. doi: 10.1016/j.eclinm.2021.101005 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 33. Back S. Clinical trial for alcohol use disorder and post traumatic stress disorder (PTSD). ClinicalTrials. 2023. https://classic.clinicaltrials.gov/ct2/show/NCT02966873 [Google Scholar]
  • 34. Morley KC, Peruch S, Adams C, et al. N‐acetylcysteine in the treatment of alcohol use disorder: a randomized, double‐blind, placebo‐controlled trial. Alcohol Alcohol. 2023;58(5):553‐560. doi: 10.1093/alcalc/agad044 [DOI] [PubMed] [Google Scholar]
  • 35. Borenstein M, Higgins JPT. Meta‐analysis and subgroups. Prev Sci. 2013;14(2):134‐143. doi: 10.1007/s11121-013-0377-7 [DOI] [PubMed] [Google Scholar]
  • 36. Richardson M, Garner P, Donegan S. Interpretation of subgroup analyses in systematic reviews: a tutorial. Clin Epidemiol Glob Health. 2019;7(2):192‐198. doi: 10.1016/j.cegh.2018.05.005 [DOI] [Google Scholar]
  • 37. Higgins JPT, Altman DG, Gøtzsche PC, et al. The Cochrane Collaboration's tool for assessing risk of bias in randomised trials. BMJ (Online). 2011;343(7829):d5928. doi: 10.1136/bmj.d5928 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38. Page MJ, McKenzie JE, Bossuyt PM, et al. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. The BMJ. 2021;372:n71. doi: 10.1136/bmj.n71 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 39. Viechtbauer W. Conducting meta‐analysis in R with the metafor package. J Stat Softw. 2010;36(3):1‐48. doi: 10.18637/jss.v036.i03 [DOI] [Google Scholar]
  • 40. Del Re, A. C. , & Hoyt, W. T. MAd: meta‐analysis with mean differences, 2014.
  • 41. Egger M, Smith GD, Schneider M, Minder C. Bias in meta‐analysis detected by a simple, graphical test. Br Med J. 1997;315(7109):629‐634. doi: 10.1136/bmj.315.7109.629 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 42. Miele C, Cabé J, Cabé N, et al. Measuring craving: a systematic review and mapping of assessment instruments. What about sexual craving? Addiction. 2023;118(12):2277‐2314. doi: 10.1111/add.16287 [DOI] [PubMed] [Google Scholar]
  • 43. Jaasu NM, Kamaraj R, Seetharaman R. MedDRA (medical dictionary for regulatory activities). Res J Pharm Technol. 2018;11(10):4751‐4754. doi: 10.5958/0974-360X.2018.00866.1 [DOI] [Google Scholar]
  • 44. Dechartres A, Trinquart L, Boutron I, Ravaud P. Influence of trial sample size on treatment effect estimates: meta‐epidemiological study. BMJ (Online). 2013;346(7908):f2306. doi: 10.1136/bmj.f2304 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 45. Dalton JE, Bolen SD, Mascha EJ. Publication bias: the elephant in the review. Anesth Analg. 2016;123(4):812‐813. doi: 10.1213/ANE.0000000000001596 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 46. Egger M, Jüni P, Bartlett C, Holenstein F, Sterne J. How important are comprehensive literature searches and the assessment of trial quality in systematic reviews? Empirical study. Health Technol Assess. 2003;7(1):1‐76. doi: 10.3310/hta7010 [DOI] [PubMed] [Google Scholar]
  • 47. Li T, Yu T, Hawkins BS, Dickersin K. Design, analysis, and reporting of crossover trials for inclusion in a meta‐analysis. PLoS ONE. 2015;10(8):e0133023. doi: 10.1371/journal.pone.0133023 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 48. Von Hippel PT. The heterogeneity statistic I2 can be biased in small meta‐analyses. BMC Med Res Methodol. 2015;15(1):1‐8. doi: 10.1186/s12874-015-0024-z [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 49. U.S. Department of Health and Human Services . Policy regarding N‐acetyl‐L‐cysteine: guidance for industry. Food and Drug Administration; 2022. https://www.fda.gov/media/157784/download?attachment [Google Scholar]
  • 50. Tomko RL, Jones JL, Gilmore AK, Brady KT, Back SE, Gray KM. N‐acetylcysteine: a potential treatment for substance use disorders. Curr Psychiat. 2018;17(6):30‐55. [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Data Availability Statement

The R code used to conduct analyses and create forest and funnel plots can be accessed using this link: https://github.com/ewinterli/NAC-meta-analysis.

Articles from Addiction Biology are provided here courtesy of Wiley

RESOURCES