Craving cannabis: a meta-analysis of self-report and psychophysiological cue—reactivity studies

Melissa M. Norberg*, David J. Kavanagh, Jake Olivier, Stephanie Lyras

*Corresponding author for this work

Research output: Contribution to journalReview articlepeer-review

32 Citations (Scopus)


Aims: To estimate the magnitude of cannabis cue–reactivity responses in people who regularly use cannabis. Methods: EMBASE, PUBMED, PsycINFO, PsycEXTRA and Project CORK were searched for within-subject comparison studies published between January 2000 and December 2014. Eligible studies compared regular cannabis users' subjective craving or psychophysiological responding when exposed to cannabis cues to their responses to neutral cues and/or to their responses during a baseline assessment. Eligible studies presented original data and were written in English. Fourteen studies met inclusion criteria, but data were not available for two. The 12 included studies contained 393 participants (12–97 participants per study; mean age range: 16.4–32.6 years). Standardized effects sizes were computed for heart rate, skin conductance, skin temperature, blood pressure, event-related potentials (Pz P300) and self-reported craving (Visual Analogue Scales and Marijuana Craving Questionnaire) and synthesized using a multivariate meta-regression model. Results: On average, cannabis users experienced moderate cue–reactivity, as determined by comparisons with baseline conditions [d = 0.59, 95% confidence interval (CI) = 0.43, 0.74, P < 0.001] and with neutral cue comparisons (d = 0.47, 95% CI = 0.32, 0.62, P < 0.001). This model, which excluded studies at a greater than low risk of bias and included PzP300 studies as a moderator (due to having larger effect sizes), had a moderately low level of heterogeneity of effect size (I2 = 45.6%, Q = 56.97, P = 0.003) and showed no evidence of publication bias. Cue–reactivity was most noticeable (large effect size) when assessed by Pz P300 waves (d = 1.71, 95% CI = 0.77, 2.64, P = 0.0003; I2 = 86%, Q = 36.30, P < 0.001) and least noticeable (trivial effect size) when examining heart rate (d = 0.14, 95% CI = −0.34, 0.62, P = 0.58; I2 = 80%, Q = 20.0, P < 0.001). These subgroup models demonstrated high heterogeneity of effect size. Conclusions: Regular cannabis users experience moderate to extremely intense cue–reactivity, such that their attentional biases towards cannabis cues are much stronger than their perceptions of craving for the drug.

Original languageEnglish
Pages (from-to)1923-1934
Number of pages12
Issue number11
Publication statusPublished - Nov 2016


  • conditioning
  • craving
  • cues
  • drug effects
  • marijuana
  • association learning


Dive into the research topics of 'Craving cannabis: a meta-analysis of self-report and psychophysiological cue—reactivity studies'. Together they form a unique fingerprint.

Cite this