Identifying clinical predictors of response to repetitive transcranial magnetic stimulation for smoking cessation: Secondary analysis of a multicenter RCT

Tobacco smoking remains a leading cause of morbidity and mortality worldwide. While evidence-based treatments exist for nicotine dependence, many patients do not respond to or cannot tolerate them. Repetitive Transcranial Magnetic Stimulation (rTMS) is a novel, non-invasive neuromodulation treatment that stimulates parts of the brain involved in addiction [[1]Mehta D.D. Praecht A. Ward H.B. et al.A systematic review and meta-analysis of neuromodulation therapies for substance use disorders.Neuropsychopharmacology. Dec 12 2023; https://doi.org/10.1038/s41386-023-01776-0Crossref PubMed Scopus (0) Google Scholar]. Recently, the Brainsway H4 deep TMS coil was cleared by the FDA and Health Canada for the treatment for nicotine dependence, based on a pivotal multisite RCT with continuous quit rates of 19.4 % from rTMS compared to 8.7 % in the sham control group [[2]Zangen A. Moshe H. Martinez D. et al.Repetitive transcranial magnetic stimulation for smoking cessation: a pivotal multicenter double-blind randomized controlled trial.World Psychiatr. Oct 2021; 20: 397-404https://doi.org/10.1002/wps.20905Crossref PubMed Scopus (75) Google Scholar]. This trial included 262 smokers randomized to 3 weeks of daily active or sham rTMS to the lateral prefrontal and insular cortices, plus 3 once-weekly maintenance sessions, for a total of 18 sessions. Although this trial represents a major advance in the treatment of nicotine dependence, the overall response rates are modest, and more research is needed on how to further optimize treatment efficacy. Given that rTMS is a novel and resource-intensive mode of smoking cessation therapy, determining who benefits most from treatment may inform future device designs or clinical recommendations. The original trial reported an exploratory analysis showing that reductions in craving after the initial rTMS treatment session predicted abstinence in the active arm [[2]Zangen A. Moshe H. Martinez D. et al.Repetitive transcranial magnetic stimulation for smoking cessation: a pivotal multicenter double-blind randomized controlled trial.World Psychiatr. Oct 2021; 20: 397-404https://doi.org/10.1002/wps.20905Crossref PubMed Scopus (75) Google Scholar], and a follow-up study examined demographic correlates of response among treatment completers [[3]Gersner R. Barnea-Ygael N. Tendler A. Moderators of the response to deep TMS for smoking addiction.Front Psychiatr. 2022; 131079138https://doi.org/10.3389/fpsyt.2022.1079138Crossref PubMed Scopus (0) Google Scholar]. Our aim was to explore the possibility of constructing a predictive model based on clinical factors measured before treatment through a secondary analysis of the pivotal trial of rTMS for smoking cessation. We focused on symptom subtypes that we hypothesized to represent the neurobiological targets of rTMS for addiction. As in the primary analysis, our outcome was four-week continuous smoking abstinence at 18 weeks [[2]Zangen A. Moshe H. Martinez D. et al.Repetitive transcranial magnetic stimulation for smoking cessation: a pivotal multicenter double-blind randomized controlled trial.World Psychiatr. Oct 2021; 20: 397-404https://doi.org/10.1002/wps.20905Crossref PubMed Scopus (75) Google Scholar]. We included the 234 participants reported in the final intention-to-treat analysis of trial data, of whom 32 quit smoking successfully. We included variables associated with cessation in published studies: age, sex, previous quit attempts, and the Fagerström Test for Nicotine Dependence [FTND] [[4]Hymowitz N. Cummings K.M. Hyland A. Lynn W.R. Pechacek T.F. Hartwell T.D. Predictors of smoking cessation in a cohort of adult smokers followed for five years.Tobac Control. 1997; 6 (Suppl 2): S57-S62https://doi.org/10.1136/tc.6.suppl_2.s57Crossref PubMed Scopus (505) Google Scholar,[5]West R. Evins A.E. Benowitz N.L. et al.Factors associated with the efficacy of smoking cessation treatments and predictors of smoking abstinence in EAGLES.Addiction. Aug 2018; 113: 1507-1516https://doi.org/10.1111/add.14208Crossref PubMed Scopus (53) Google Scholar]. In addition, we included symptom scores characterizing the behavioural phenotypes associated with the neurocircuitry of addiction. Non-invasive neuromodulation techniques such as rTMS can selectively target aberrant neurocircuits involved in reward, cognition, and emotions that are implicated in substance use disorders [[6]Koob G.F. Volkow N.D. Neurobiology of addiction: a neurocircuitry analysis.Lancet Psychiatr. Aug 2016; 3: 760-773https://doi.org/10.1016/S2215-0366(16)00104-8Abstract Full Text Full Text PDF PubMed Scopus (1856) Google Scholar], which are characterized by maladaptive changes in incentive salience (driven by areas in the basal ganglia), reward deficit and stress (extended amygdala), and executive dysfunction (prefrontal cortex). Modulating these brain regions may effectively treat the addicted brain. To reflect these domains, we summed z-scores to produce 3 summary measures: negative emotionality (the emotionality subscale of the tobacco craving questionnaire [TCQ] + Minnesota Nicotine Withdrawal Scale); executive functioning (TCQ compulsivity score); and incentive salience (TCQ expectancy score + visual analogue scale measuring craving). Any model including all variables would have had too many parameters for the number of events and would have been at high risk of over-fitting and non-replicability. We therefore performed two analyses. First, we ran one-at-a-time logistic regression models, each including all main effects and one interaction term. These results are shown in Table 1, with 99 % confidence intervals to emphasize the issue of multiplicity. No interaction p-values would survive any common correction for multiplicity.Table 1Results of one-at-a-time interaction testing with logistic regression.Interaction with treatment arm (active = 1, sham = 0)ORSEzpL99 %U99 %AgeaZ-scores.0.390.43−2.180.0290.131.19Sex (female = 1)1.680.880.590.550.1716.28Lifetime quit attemptsaZ-scores.0.980.42−0.040.970.332.89Executive functioningaZ-scores.,bExecutive functioning = Tobacco craving questionnaire (TCQ).1.160.420.360.720.393.46Incentive salience aZ-scores.,cIncentive salience = TCQ expectancy score and visual analogue scale measuring craving.2.410.432.060.040.807.24Negative emotionalityaZ-scores.,dNegative emotionality = the emotionality subscale of the TCQ and Minnesota Nicotine Withdrawal Scale.0.930.4−0.170.870.342.61FTNDaZ-scores.0.720.4−0.820.410.262.01OR: Odds ratio, SE: Standard error, z: z-score, p: p-value, L99 %: lower bound 99 % confidence interval, U99 %: upper bound 99 % confidence interval.a Z-scores.b Executive functioning = Tobacco craving questionnaire (TCQ).c Incentive salience = TCQ expectancy score and visual analogue scale measuring craving.d Negative emotionality = the emotionality subscale of the TCQ and Minnesota Nicotine Withdrawal Scale. Open table in a new tab OR: Odds ratio, SE: Standard error, z: z-score, p: p-value, L99 %: lower bound 99 % confidence interval, U99 %: upper bound 99 % confidence interval. Second, we used the glmnet R package [[7]Friedman J. Hastie T. Tibshirani R. Regularization paths for generalized linear models via coordinate descent.J Stat Software. 2010; 33: 1-22Crossref PubMed Google Scholar] to fit a lasso model. Lasso performs variable selection through regularization, “shrinking” coefficients such that weaker predictors are excluded from the model, and is well-suited to the development of replicable models. We used a logistic model with all variables above and their interactions with treatment arm, and used default standardization, with cross-validation to select the penalization parameter (lambda). Four participants (1.7 %) did not supply values for number of previous quit attempts; as this was a very small amount of missing data, we simply used mean substitution to retain these cases. As is general practice, we selected lambda based on the +1SE rule [[8]Krstajic D. Buturovic L.J. Leahy D.E. Thomas S. Cross-validation pitfalls when selecting and assessing regression and classification models.J Cheminf. Mar 29 2014; 6: 10https://doi.org/10.1186/1758-2946-6-10Crossref PubMed Scopus (573) Google Scholar]. The final lasso model included only treatment arm. We therefore did not find evidence that our set of variables can predict outcomes. Although results must be considered non-significant, we note that there is some suggestion in Table 1 that treatment effects were larger among younger people and those with higher levels of abnormal incentive salience for tobacco smoking, as characterized by a higher score on a composite of the VAS and TCQ expectancy subscale. As noted, one other report has looked at predictors of response. That analysis considered demographic factors (except sex), included the treatment completers only, and was broadly exploratory, with no control of multiplicity or overfitting [[3]Gersner R. Barnea-Ygael N. Tendler A. Moderators of the response to deep TMS for smoking addiction.Front Psychiatr. 2022; 131079138https://doi.org/10.3389/fpsyt.2022.1079138Crossref PubMed Scopus (0) Google Scholar]. Our analysis included the originally-reported sample, focuses on clinical factors (e.g., FTND scores) and symptom clusters theoretically targeted by rTMS, and attempted to identify a replicable model. We did not find evidence of treatment effect heterogeneity or of predictive utility for our variables. However, future studies might usefully consider the roles of incentive salience and age. Limitations of the current analysis include the small sample size and low number of primary outcome events. However, the trial is considered large for an rTMS RCT, and is certainly the largest to date for smoking cessation [[1]Mehta D.D. Praecht A. Ward H.B. et al.A systematic review and meta-analysis of neuromodulation therapies for substance use disorders.Neuropsychopharmacology. Dec 12 2023; https://doi.org/10.1038/s41386-023-01776-0Crossref PubMed Scopus (0) Google Scholar]. We also lacked specific, validated assessments for executive function, incentive salience, and negative emotionality, and therefore adapted measures based on face validity. The mechanisms underlying the effectiveness of rTMS for addiction are not fully understood. Exploring treatment effect heterogeneity may identify brain regions of particular importance and enable targeting to people most likely to benefit. Future trials could usefully include hypothesis-driven, neurobiological assessments to elucidate mechanisms of action. Progress in identifying predictors of response will advance if the field can develop a consensus on 1) areas of functioning most likely to predict response; and 2) the best measures of these domains. As large trials of rTMS are difficult to conduct, it is important for even smaller studies to report such findings, so these can be collated and summarized. Pre-specification of these analyses in future trial protocols would greatly reduce concerns about multiplicity and post hoc analysis. Victor M. Tang: Writing – review & editing, Writing – original draft, Resources, Project administration, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Laurie Zawertailo: Writing – review & editing, Supervision, Conceptualization. Peter Selby: Writing – review & editing, Supervision, Resources. Abraham Zangen: Writing – review & editing, Data curation. Dhvani Mehta: Writing – review & editing, Methodology, Data curation. Tony P. George: Writing – review & editing, Supervision. Bernard Le Foll: Writing – review & editing, Supervision, Conceptualization. Kristina M. Gicas: Writing – review & editing, Methodology, Conceptualization. Matthew E. Sloan: Writing – review & editing, Methodology. Scott Veldhuizen: Writing – review & editing, Writing – original draft, Validation, Project administration, Methodology, Investigation, Formal analysis, Conceptualization. The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: Dr. B. Le Foll has been provided a coil for a Transcranial magnetic stimulation (TMS) study from Brainsway. Dr. A. Zangen is an inventor of deep TMS coils and has financial interest in BrainsWay which produces and markets these coils. This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors. We appreciate Brainsway for the data shared through the Investigator Initiated Studies program and feedback on the manuscript provided by Dr. Colleen Hanlon.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.