Smoking has a remarkable ability to disguise itself. For some people, it feels like stress relief, comfort, routine, or even a “friend” during difficult moments. Yet behind the ritual sits one of the deadliest commercial addictions ever created: around half of long-term smokers will die from smoking unless they quit (Doll et al., 2004; Pirie et al., 2013).
Despite major reductions in smoking prevalence over recent decades, tobacco use is now increasingly concentrated among people experiencing socioeconomic disadvantage and marginalisation (Cornelius et al., 2023; OHID, 2024; ONS, 2021; Taylor et al., 2020). Attempts to stop smoking are frequently unsuccessful, particularly for those with mental health conditions (Taylor G., 2025), and relapse remains common even among people receiving evidence-based interventions (Rigotti et al., 2022). Given the substantial health harms and societal burden associated with smoking, there is a clear need to develop novel cessation approaches that improve sustained long-term abstinence.
Evidence-based smoking cessation treatment is clear on one thing: behavioural and psychological support improves quit rates (Hartmann-Boyce et al., 2021; Stead et al., 2016). Cochrane reviews consistently show that behavioural support (e.g., structured CBT, counselling, motivational interviewing, brief behavioural advice) increases the likelihood of long-term abstinence, particularly when paired with stop smoking medications. In English stop smoking services, the active components of behavioural interventions are well mapped out and standardised (NCSCT, 2019).
Against this backdrop, Wittekind and colleagues (2026) tested a psychological approach that is not offered in standard smoking cessation care: Approach Bias Modification, a computerised intervention designed to retrain the brain’s automatic cognitive responses to smoking cues. Evidence for Approach Bias Modification remains mixed and methodologically limited, highlighting the need for stronger trials (Cristea et al., 2016; Stephan Mühlig, 2017). In this new randomised controlled trial Wittekind et al., (2026) asked an important question:
Can retraining automatic “approach biases” towards cigarettes actually help people quit smoking compared to a cognitive-behavioural intervention?
Smoking is a master of disguise, but can Approach Bias Modification reveal the elf in the room?
Methods
Wittekind et al. conducted a randomised, controlled, double-blind superiority trial involving 351 adults with tobacco dependence recruited in Germany. All participants received a one-day cognitive behavioural smoking cessation intervention (“treatment as usual;” TAU) before being randomised to either Approach Bias Modification training, sham training, or TAU alone. Participants completed seven days of training, with the primary outcome being biochemically verified prolonged abstinence at six months using the Russell Standard criteria (West et al., 2005). The study used intention-to-treat analyses, included two control groups, and blinded participants and assessors to allocation where possible. However, fidelity of the behavioural intervention was not formally assessed.
Results
A total of 351 adults with tobacco dependence were included in the final intention-to-treat analysis:
- 119 received TAU plus Approach Bias Modification
- 115 received TAU plus sham training
- 117 received TAU alone.
Participants were 42 years old on average, smoked around 19 cigarettes per day, and had been smoking for approximately 24 years. Baseline characteristics were balanced across groups, suggesting randomisation was successful.
Primary analysis
- The primary outcome was prolonged smoking abstinence at six months, verified using self-report alongside biochemical confirmation using exhaled carbon monoxide.
- At follow-up:
- 19.3% of participants receiving Approach Bias Modification had quit smoking,
- compared with 17.4% receiving sham training and
- 16.2% receiving TAU alone.
- Statistical analysis found no statistically significant differences between groups, and the researchers did not conclude that Approach Bias Modification improved quit rates beyond standard behavioural treatment.
Absolute effects
Looking at the absolute effects helps place these findings in context. Compared with TAU alone, Approach Bias Modification was associated with an absolute increase in abstinence of 3.1 percentage points, approximately three additional quitters per 100 people treated. Compared with sham training, the difference was 1.9 percentage points. These are potentially clinically meaningful effects at population level but were accompanied by wide confidence intervals, meaning the true effect could range from benefit to little or no additional effect.
Secondary analysis
Secondary outcomes told a similarly nuanced story. Across all groups, participants reduced cigarette dependence, craving, cigarette consumption, and carbon monoxide levels over time. Average daily cigarette use approximately halved immediately after treatment, dropping from around 19 cigarettes per day at baseline to around 7 cigarettes per day post-intervention across groups, with some increase by six months but remaining below baseline. This suggests that the behavioural smoking cessation programme itself was effective.
Mechanistic results
The mechanistic findings were also notable. Although approach biases reduced over time, mediation analyses found no evidence that changes in cognitive bias explained smoking outcomes. Similarly, impulsivity and executive functioning did not appear to alter treatment response. In practical terms, this means the intervention changed some psychological measures, but those changes did not translate into measurable improvements in long-term smoking cessation.
Slightly more quitters, and slightly fewer puffs… but not enough evidence to declare a breakthrough.
Conclusions
Wittekind and colleagues found that adding Approach Bias Modification to standard smoking cessation treatment did not provide strong evidence for an improvement in long-term quit rates compared with either sham training or treatment as usual alone. Although smoking dependence, craving, and cigarette consumption reduced over time, these improvements occurred across all groups rather than specifically in the Approach Bias Modification condition.
The authors concluded that:
this randomised controlled trial in a large sample of adults does not provide evidence that Approach Bias Modification, when used as an add-on to smoking cessation treatment, improves long-term abstinence rates.
Same destination, different routes: all groups improved, but no clear winner emerged.
Strengths and limitations
This was a well-conducted randomised controlled trial with several important methodological strengths. The researchers used biochemical verification of smoking abstinence, intention-to-treat analyses, double-blinding for the training conditions, and included both a sham-training and treatment-as-usual control group. The intervention was also theory-driven and a plausible mechanistic target: automatic approach biases towards smoking cues.
However, I am not convinced the trial was adequately powered to detect clinically realistic smoking cessation effects. The study appears powered for relatively large absolute differences between groups, but most effective smoking cessation interventions produce modest improvements in quit rates, often in the region of 10-15 percentage points (Stead et al., 2016). With 115–119 participants per arm, the trial would likely have had limited statistical power to detect these smaller, but clinically meaningful differences. The observed abstinence rates numerically favoured Approach Bias Modification + TAU (19.3%) over sham training + TAU (17.4%) and TAU alone (16.2%), but confidence intervals were wide and overlapping. An imprecise finding here should therefore not automatically be interpreted as evidence of “no effect.”
There are also interesting conceptual issues. The intervention was compared against an intensive cognitive behavioural smoking cessation programme that included well established motivational and behavioural techniques. This raises the possibility of a ceiling effect: when participants already receive high-quality behavioural support, it may be difficult for an adjunctive computerised intervention to demonstrate additional benefit. In that sense, the findings may say more about comparative effectiveness than outright inefficacy.
Attrition is another important consideration. Dropout rates were higher in the treatment-as-usual-only arm, potentially introducing attrition bias. The authors classified all missing participants as relapsed smokers, which is standard in cessation research, but this assumption may disproportionately disadvantage groups with poorer retention, like the sham group (92/115, 80%) and Approach Bias Modification group (99/119, 83%). Furthermore, most training sessions occurred at home, reducing control over adherence and potentially diluting intervention fidelity.
Finally, the broader clinical question may not simply be “does bias modification outperform CBT?”, but whether it offers an additional treatment option for people who engage less well with traditional behavioural approaches. Smoking cessation is rarely one-size-fits-all, and patient choice may matter as much as slight differences in efficacy estimates.
The elves checked for bias… but who checked whether the trial could detect realistic quit rates?
Implications for practice
So, should this trial change practice? Probably not immediately, but nor do I think it closes the door on Approach Bias Modification for smoking cessation. The headline finding from this study is easy to oversimplify:
Approach Bias Modification did not significantly improve quit rates.
But smoking cessation research is rarely that straightforward. The intervention achieved numerically higher abstinence rates than both comparator groups, with quit rates approaching 19.3% at six months. In smoking cessation, those are not trivial outcomes. Many established behavioural and pharmacological interventions produce modest absolute improvements in quit rates, and the reality is that helping even a small additional proportion of people stop smoking can translate into substantial population health gains.
Importantly, this trial tested Approach Bias Modification as an add-on to an already intensive cognitive behavioural smoking cessation intervention. Participants were not receiving minimal care; they were receiving structured behavioural support delivered by trained clinicians. In that context, expecting a large additional treatment effect from a brief computerised intervention may simply be unrealistic. The more meaningful question may be whether Approach Bias Modification offers another acceptable option within a broader menu of cessation support, particularly for people who struggle to engage with traditional approaches.
I also do not think this evidence should sit in isolation. The logical next step is synthesis rather than dismissal. This study should be incorporated into an updated systematic review and meta-analysis alongside previous Approach Bias Modification trials. At present, the evidence base remains fragmented, underpowered, and methodologically heterogeneous. Larger pragmatic trials are still needed, particularly studies embedded within real-world healthcare systems and studies comparing different delivery models, intensities, and patient groups.
There are also wider policy implications. NICE is currently exploring digital technologies to support smoking cessation in secondary care through its Early Value Assessment programme. Approach Bias Modification is potentially well aligned with this agenda. Because these interventions are computerised, scalable, and potentially low cost, they fit closely with the NHS “analogue to digital” ambitions outlined in the UK 10-Year Health Plan. If effective, these approaches could theoretically be integrated into NHS stop smoking pathways, offered remotely, and delivered at scale with minimal workforce burden.
But this is where implementation science collides with reality. One of the greatest barriers in UK healthcare is not necessarily generating promising evidence, it is translating that evidence into commissioned NHS services. Academic groups are rarely equipped to rapidly scale digital interventions, navigate procurement systems, or secure market access. Industry partnerships are often essential. Yet even when interventions show promise, offer potential cost savings, and align with NHS priorities, achieving adoption within routine care can feel painfully slow.
Perhaps that brings us back to the opening story. Smoking addiction thrives on automatic habits, repeated thousands of times over years. Maybe changing those habits will also require persistence: not one “magic bullet” intervention, but multiple complementary tools working together. Approach Bias Modification may not be the breakthrough some hoped for, but this trial suggests it may still deserve a seat at the table.
From analogue to digital: can Approach Bias Modification find its way into the NHS toolbox?
Statement of interests
Dr Taylor was not involved in this study, does not know the study authors personally, and was not involved in peer review or editorial decisions relating to publication of this paper. However, Dr Taylor has research expertise in smoking cessation and has functioned as Principal Investigator on trials of smoking cessation interventions, including both digital and face-to-face cognitive behavioural treatments.
Dr Taylor acknowledges research funding from Cancer Research UK (CRUK), the Causality in Healthcare AI Hub (funded by EPSRC and UKRI), and the NIHR Bristol Biomedical Research Centre (NIHR203315), University Hospitals Bristol and Weston NHS Foundation Trust, and the University of Bristol.
Dr Taylor previously worked at a health economics research agency whose clients included pharmaceutical companies and has received consultancy fees from publicly funded public health organisations.
Dr Taylor is a Trustee of the Society for the Study of Addiction and is a member of the Ethical Medicines Industry Group, the University-Industry Contracting Partnership, and the University Industry Innovation Network.
The views expressed in this blog are those of the author and do not necessarily reflect those of the funders, affiliated organisations, or memberships listed above.
ChatGPT was used to assist with proofreading and generating captions.
Editor
Edited by Éimear Foley. ChatGPT assisted with language refinement and formatting during the editorial phase.
Links
Primary paper
Charlotte Wittekind, Keisuke Takano, Franziska Motka, Markus Winkler, Gabriela Werner, Thomas Ehring, Tobias Rüther. 2026. Approach Bias Modification as an Add-On to Smoking Cessation Treatment: A Randomized Controlled Trial. American Journal of Psychiatry 183, 240–250. https://doi.org/10.1176/appi.ajp.20250189
Other references
Cornelius, M. E., Loretan, C. G., Jamal, A., Davis Lynn, B. C., Mayer, M., Alcantara, I. C., & Neff, L. (2023). Tobacco Product Use Among Adults – United States, 2021. MMWR Morb Mortal Wkly Rep, 72(18), 475-483. https://doi.org/10.15585/mmwr.mm7218a1
Cristea, I. A., Kok, R. N., & Cuijpers, P. (2016). The Effectiveness of Cognitive Bias Modification Interventions for Substance Addictions: A Meta-Analysis. PLoS ONE, 11(9), e0162226. https://doi.org/10.1371/journal.pone.0162226
Doll, R., Peto, R., Boreham, J., & Sutherland, I. (2004). Mortality in relation to smoking: 50 years’ observations on male British doctors. BMJ, 328(7455). https://doi.org/10.1136/bmj.38142.554479.AE
Hartmann-Boyce, J., Livingstone-Banks, J., Ordóñez-Mena, J. M., Fanshawe, T. R., Lindson, N., Freeman, S. C., Sutton, A. J., Theodoulou, A., & Aveyard, P. (2021). Behavioural interventions for smoking cessation: an overview and network meta‐analysis.
NCSCT. (2019). Standard Treatment Programme.
OHID. (2024). Official Statistics: Smoking profile for England: statistical commentary.
ONS. (2021). Deprivation and the impact on smoking prevalence, England and Wales: 2017 to 2021.
Pirie, K., Peto, R., Reeves, G. K., Green, J., & Beral, V. (2013). The 21st century hazards of smoking and benefits of stopping: a prospective study of one million women in the UK. The Lancet, 381(9861), 133-141. https://doi.org/10.1016/S0140-6736(12)61720-6
Rigotti, N. A., Kruse, G. R., Livingstone-Banks, J., & Hartmann-Boyce, J. (2022). Treatment of Tobacco Smoking: A Review. JAMA, 327(6), 566-577. https://doi.org/10.1001/jama.2022.0395
Stead, L., Koilpillai, P., Fanshawe, T., Lancaster, T., Stead, L., Koilpillai, P., Fanshawe, T., & Lancaster, T. (2016). Combined pharmacotherapy and behavioural interventions for smoking cessation. Cochrane Database, 3(3).
Stephan Mühlig, J. P., Johannes Lindenmeyer, Mike Rinck, Reto Cina, and Reinout W. Wiers. (2017). Applying the ‘Cognitive Bias Modification’ concept to smoking cessation – A Systematic Review. Journal of Addicition Research and Practice, 62(6). https://doi.org/https://doi.org/10.1024/0939-5911/a000454
Taylor, G., Itani, T., Thomas, K., Rai, D., Jones, T., Windmeijer, F., Martin, R., Munafò, M., Davies, N., & Taylor, A. (2020). Prescribing prevalence, effectiveness, and safety of smoking cessation medicines in patients with mental disorders. NTR, 22(1).
Taylor, G. (2025) Smoking and depression: a burning issue for the elf care community. The Mental Elf, 10 Apr 2025.
West, R., Hajek, P., Stead, L., & Stapleton, J. (2005). Outcome criteria in smoking cessation trials: proposal for a common standard. Addiction, 100(3), 299-303. https://doi.org/10.1111/j.1360-0443.2004.00995.x
