Self-related neural response to tailored smoking-cessation messages. predicts quitting

Neural response to tailored messages and quitting Page 1 Supporting Online Material Self-related neural response to tailored smoking-cessation messages predicts quitting Hannah Faye Chua 1, S. Shaun Ho 2, Agnes J. Jasinska 3, Thad A. Polk 2,4, Robert C. Welsh 5, Israel Liberzon 2, Victor J. Strecher 1 1 Health Behavior and Health Education, 2 Psychiatry, 3 Neuroscience Program, 4 Psychology and 5 Radiology, University of Michigan, Ann Arbor, MI. SUPPLEMENTARY METHODS Participants Ninety-one right-handed smokers (51.6% males) participated. Participants were between 21-55 years old (mean age 37.5 ± 11.5 years). Eligibility criteria included having smoked a cigarette within the last 7 days, having smoked a minimum of 10 cigarettes on average per day and at least 100 cigarettes in a lifetime, and having an interest in quitting within the next 30 days. Participants were not enrolled in other smoking cessation programs and were not taking pharmacological treatment for smoking cessation during study enrollment 1. Participants smoked on average 16.65 ± 5.8 cigarettes a day for an average of 19.81 ± 11.24 years. Most of the participants had previously attempted to quit (84.6%). Participants were motivated to quit (M = 8.99 [SD = 1.23] on a 10-point scale) and were also confident about quitting (M = 8.13 [SD = 1.78] on a 10-point scale). Education level was distributed across different levels (20% completed high school, 36.3% completed some college, 23.1% completed a non-4-year college degree, 12.1% completed a 4-year college degree, and 7.7% completed a graduate degree). Across the participants, 73.6% reported being employed or self-employed and 23.1% reported attending school. Top reasons for quitting smoking reported by the participants were personal health reasons, financial costs of cigarettes, and desire to have more control of their lives. Participants had no prior history of head injury or major psychiatric (schizophrenia/bipolar) or neurological illness. Participants did not receive any diagnosis of mental health condition (including depression and other substance abuse) and did not use illicit drugs during the past year. All participants were native English speakers, had normal hearing, and had good visual acuity. Messages We created statements for each of the following 3 message types: Tailored, Untailored, and Neutral messages. The messages had comparable reading levels (approximately 6 th grade). The number of words, syllables, and characters across the statements was also controlled. We based the Tailored and Untailored messages on cognitive-behavioral methods of smoking cessation and relapse prevention 1. Tailored statements include self-relevant personalization and feedback about the smoker. It involves the degree to which there are

Neural response to tailored messages and quitting Page 2 recognizable features of the individual in the message (e.g., You have been smoking for 12 years. ), as well as references to the individual s state characteristics (e.g., levels of stress and coping) which were generated based on participants responses. We created Untailored statements based on information available from the Guide to Quitting Smoking section of the American Cancer Society website and other similar resources, which are generic smokingcessation messages. The topics for each Tailored and Untailored statement were constrained to make them comparable. Neutral statements served as the control condition and were facts used mostly from a different study (e.g., The weight of a carat is equal to one fifth of a gram ) 2. All participants received the same Untailored and Neutral statements, but Tailored statements varied depending on the participants survey responses. For the Tailored messages, a bank of statements was created based on the protocol used for a web-based tailored smoking-cessation program 1. By the term tailoring, we refer to a process consisting of: (a) an assessment of individual characteristics relevant to smoking cessation; (b) algorithms that use the assessment data to generate intervention messages relevant to the specific needs of the user; and (c) a protocol that delivers these messages to the smoker. For this study, we collected data from the participants via questionnaire, and the algorithm produced appropriate statements for each message type. Participants received 50 statements of each of the 3 message types. See samples in Supplementary Table 1. Supplementary Table 1. Sample statements from each message type presented as blocks Tailored messages A concern you have is being tempted to smoke when around other smokers. Something else that you feel will tempt you after you quit is because of a craving. You are worried that when angry or frustrated, you may light up. You have a lot of stress in your life right now, but you do feel you can manage it. When you feel stressed, you tend to seek out friends for support. You feel like your sister will help you stay on track once you quit. No one close to you is going to quit with you. You want to quit because you are tired of spending your money on cigarettes. You feel your health somewhat limits you, including even taking the stairs. You have tried to quit smoking before. Untailored messages Some people are tempted to smoke to control their weight or hunger. Smokers also light up when they need to concentrate. Certain moods or feelings, places, and things you do can make you want to smoke. We all face times when life can be stressful and we are not sure how to handle it all. There are many skills that can help smokers deal with stress. Someone you trust can provide support to you once you quit. Many people quit with another person so they can support each other. Many smokers quit because they are tired of spending money on cigarettes. Smokers are admitted to the hospital more often than nonsmokers. Most people need to try more than once to quit smoking for good. Neutral messages Oil was formed from the remains of animals and plants that lived millions of years ago. Sighted in the Pacific Ocean, the world's tallest sea wave was 112 feet. Wind is simple air in motion. It is caused by the uneven heating of the earth's surface by the sun.

Neural response to tailored messages and quitting Page 3 The true color of ocean water is neither blue nor white. We can measure the age of a fish by using a magnifying glass to count the rings on its scales. Social bees live in complex societies of 10,000-50,000 members. The African elephant has larger ears than the Indian elephant. The killer whale is the largest member of the dolphin family. Unlike other turtles, sea turtles cannot retract their legs and heads into their shells. fmri Tasks Participants completed the Messages Task and the Self-Appraisal Task during fmri. Messages Task. We presented three types of messages: Tailored, Untailored, and Neutral. Each statement was visually presented on the screen and simultaneously delivered in audio 3. There were 5 runs, 2 blocks of each message type per run, 5 messages per block, for the total of 150 messages presented. Each block lasted 24 seconds. We presented the Message blocks in pseudorandom order within and across participants. Fixations lasting between 4-10 seconds (averaging 7s) occurred after each message block. Each run also started and ended with a 10- second fixation. Self-Appraisal Task. The paradigm was adopted from Schmitz and Johnson 4 and Schmitz, Rowley, Kawahara & Johnson 5. There were two task conditions: self-referential and word valence judgment conditions. Adjectives were presented and participants had to make a selfreferential or word valence judgment, with yes or no responses. Example adjectives used were shy, happy, and analytical. In the self-referential judgment condition, participants had to decide whether an adjective described them or not. In the valence judgment condition, participants had to decide whether the adjective was of positive valence or not. Adjectives were presented for 3 seconds plus a 1 second inter-stimulus interval. Each block contained 6 adjectives of 4 seconds each for a total of 24s. Each run included 5 blocks of each condition in alternating cycles. There were two runs. The order of presentation of adjectives was counterbalanced across participants. Each run started and ended with a 10s fixation. The task lasted 8.67 minutes. Procedure Participants completed a phone-based eligibility screening. We invited those who were eligible to participate in the study. Participants completed 3 sessions, plus a follow-up call. Baseline Session (Session 1). Participants completed the MRI safety screening form and gave written consent. They also completed a standard baseline assessment about their smoking history and psychosocial, health, and demographic characteristics relevant to smoking cessation. We used the responses to create the tailored messages. fmri Session (Session 2). Participants completed the Messages Task and Self-Appraisal Task during fmri. After the scan, we gave participants a surprise memory task to assess their attention to the Messages Task in the scanner. Sixty statements were presented to the participants. Thirty were shown earlier in the scanner and thirty were new. There were an equal number of statements from each message type in both the old and new statements. In the task, we told the participants that they would be presented a series of statements, some of which they were shown earlier during the scan. Their task was to indicate whether the statement was old or new, by

Neural response to tailored messages and quitting Page 4 pressing specific response keys. After, we gave each participant one nicotine patch to be worn on the morning of their intervention session. Intervention Session (Session 3). Their third visit, scheduled within one week from their fmri scan, was the day the participants received the intervention and were also instructed to quit smoking. Participants were administered a complete web-based tailored smoking-cessation program developed by the University of Michigan Center for Health Communications Research 1. Participants spent 20-50 minutes going through the intervention materials. Self-relevant personalization and feedback about the smoker (similar to the messages presented in the scanner), as well as additional tailored motivational and instructional messages were provided in the intervention. We did not collect any additional behavioral ratings to compare message perception during scanner session and intervention session. We provided participants with a 10-week supply of nicotine patch (6-weeks of 21 mg, 2- weeks of 14-mg, and 2-weeks of 7-mg patches) as recommended by the manufacturers to people who smoke at least 10 cigarettes a day. Participants were debriefed, compensated, and received printed smoking-cessation materials, including a copy of their tailored smoking-cessation program as well as a booklet developed by the National Cancer Institute (i.e., Clearing the Air: Quit Smoking Today). Four-month Follow-up. Four months after the intervention session, trained research staff conducted the follow-up interviews with participants. The interview focused on questions regarding smoking status and smoking-cessation activities 1. The primary outcome measure was the standard 7-day abstinence (7-day point prevalence abstinence). This metric has been demonstrated to have an extremely high correlation with 24-hour point prevalence, and 30-day prolonged abstinence. These three measures provide the same information when used as outcome measures in smoking-cessation studies 6. We also compensated the participants for their participation in the follow-up interview. fmri Acquisition Scanning occurred on a 3T GE Signa Excite 2 scanner (Milwaukee, Wisconsin), beginning with a structural acquisition (T1-overlay) (repetition time [TR] = 250 ms, echo time [TE] = 7 ms, flip angle [FA] = 75 o, field of view [FOV] = 220 mm, 43 oblique axial slices, 256 x 256, slice thickness 3.0 mm). A T2*-weighted, spiral-in acquisition sequence 7 was used during the actual task (gradient echo, TR = 2000 ms, TE = 30 ms, FA = 90 o, FOV = 220 mm, 64 x 64, slice thickness 3.0 mm). A high-resolution T1 scan was also obtained (three-dimensional spoiled-gradient echo [3D-SPGR] with inversion recovery prep, time of inversion = 400 ms, TR = 9.0 ms, TE = 1.8 ms, FA = 15 o, FOV = 260 mm, 128 slices, 256 x 256, 1.2 mm slice). fmri Analyses First, data were sinc-interpolated in time, slice-by-slice. A six-parameter, least-squares minimization, motion correction algorithm, using the MCFLIRT program (FSL Analysis Group, FMRIB, Oxford, UK) was applied to realign all functional data. Subsequent processing was done using SPM2 (Wellcome Institute of Cognitive Neurology, London, UK). The T1-overlay was co-registered with a functional scan. Then the 3D-SPGR, high-resolution image was coregistered to the T1-overlay. The high-resolution image was anatomically normalized to the

Neural response to tailored messages and quitting Page 5 MNI152 template. The resulting transformation parameters were then applied to the coregistered functional volumes. A 5-mm smoothing was applied to the functional volumes. The fmri time-series were analyzed using a modified general linear model. Messages Task included blocks of Tailored, Untailored, and Neutral messages. The regressors for the Self- Appraisal Task were self-referential judgment and valence judgment blocks. Regressors were convolved with a canonical hemodynamic response function and a time derivative was calculated. For all of the models, movement parameters were included as covariates. Parameter estimates for the different trials were obtained at each voxel within the brain. Contrast images were smoothed an additional 5 mm before entering into one-sample t-tests during the second-level random effects analysis. The main contrast results for Self-Appraisal Task (Self > Valence) and Messages Task (Tailored > Neutral, Untailored > Neutral) were thresholded at whole-brain family-wise correction of.05 with voxel extent of 50, and the specific contrast results for Tailored > Untailored were thresholded at whole-brain correction using false discovery rate of.05 with voxel extent of 50. Supplemental Results Behavioral differences between Quitters and Non-Quitters The results included the data from 87 participants who were available during the followup smoking-cessation outcome assessment. We could not reach 4 of the 91 subjects during their follow-up interview period. Out of 87 participants available during follow-up assessment, two reported their quit status, but declined to complete the entire follow-up interview. Quitters and Non-Quitters groups were comparable in age, education, years of smoking, number of cigarettes smoked per day, motivation to quit or confidence in quitting, and reasons for quitting. There was no difference in the reported nicotine patch use between Quitters and Non-Quitters. A multivariate analysis of variance was undertaken to determine differences in reported behaviors between Quitters and Non-Quitters four months after the intervention. Pillai s Trace was used as a global test for any group differences on all variables. There were a significant difference between Quitters and Non-Quitters (Pillai s trace =.368, p <.001) on the reported behaviors. We examined in depth which variables were significantly different between Quitters and Non- Quitters. Quitters were more likely to change their stress response since the program, rather than turn to smoking, Ms.80 vs..58, F(1,83) = 5.24, p =.03; Quitters were more likely to avoid situations that trigger smoking since the program,, Ms.69 vs..43,f(1,83) = 6.30, p =.01; Quitters felt a lot better about their health since the program, Ms 4.24 vs. 3.48,F(1,83) = 19.17, p <.001; Quitters have made fewer attempts to quit during the past 4-month period, Ms 2.11 vs. 5.40, F(1,83) = 11.28, p =.001; Quitters were overall more satisfied with the study program, Ms 8.89 vs. 8.23, F(1,83) = 4.32, p <.05; Quitters were overall more motivated to continue to stay smoke-free, Ms 9.42 vs. 8.43, F(1,83) = 8.88, p <.005, and were overall more confident that they would not start smoking again, Ms 8.73 vs. 7.78, F(1,83) = 4.10, p <.05; Quitters reported experiencing less severe withdrawal symptoms during the first 3 days of quitting, Ms 5.38 vs. 6.88, F(1,83) = 5.57, p =.02. Self Appraisal Task Behavioral Results. The Self-Appraisal Task consisted of the self-referential condition and the word valence condition. Using repeated measures ANOVA, we examined task

Neural response to tailored messages and quitting Page 6 performance in both conditions with quitting status as a between-subjects variable. Overall, participants rated that 57.8% of the adjectives described them and 51.7% of the adjectives were positive in valence. This difference in judging frequency was significant, F(1,83) = 32.66, p <.001. (Out of 87 participants, there were 85 useful and complete behavioral data used in this analysis. The behavioral data set was missing for one participant, and another participant repeatedly gave the same response.) Non-Quitters were slightly more likely than Quitters to respond yes that an adjective described them or was of positive valence, Ms 56.3% vs. 53.3%, respectively, F(1, 83) = 4.02, p =.05. There was no interaction effect of quitting status and performance in self-referential and word valence judgments in the Self-Appraisal Task. fmri Results. Consistent with previous findings 4,5, self-referential evaluation, in contrast to valence judgment, of adjectives activated the medial prefrontal cortex (MPFC) and precuneus/posterior cingulate regions across the 87 subjects. See Supplementary Figure 1 and Supplementary Table 2. Activations in these regions were not correlated with quitting. The results were presented at a statistical threshold of p <.05 with whole brain correction using family-wise error and extent threshold of 50 voxels. These results were saved as a mask to identify brain regions associated with self-related processing. x = 3 Supplementary Figure 1. Activations in the MPFC and precuneus/posterior cingulate during self-referential judgment, relative to valence judgment, in the Self-Appraisal Task, across 87 subjects. The coordinates are in MNI space. Supplementary Table 2. Peak activations of self-referential evaluation, relative to valance judgments, in the Self- Appraisal Task. Region Gyri / Subregions Precuneus / Posterior Cingulate Parietal lobe Middle cingulate gyrus Cuneus MPFC Medial frontal gyrus Superior frontal gyrus Anterior cingulate gyrus Medial orbital gyrus Inferior frontal gyrus L, insula Caudate Putamen BA x, y, z a T Z b k d 31, 7, 24, 23, 30, 5, 29, 19, 6 10, 9, 32, 11, 8, 25, 47, 24 3, 51, 33 11.27 9.70 c 1166 3, 60, 12 12.20 11.01 c 2159

Neural response to tailored messages and quitting Page 7 L Inferior Parietal Lobule 39, 19, 40 45, 72, 42 8.05 6.93 228 L Angular Gyrus R Inferior Parietal Lobule 39, 40 57, 63, 36 6.37 5.75 53 R Angular Gyrus L, left; R, right. a Stereotactic coordinates of the peak voxel from the Montreal Neurological Institute atlas, left/right (x), anterior/posterior (y), and superior/inferior (z), respectively. b SPM-generated Z score, significant at FWE p <.05. c Estimated Z score, significant at FWE p <.05. d Spatial extent of the cluster in voxels (5 voxel minimum). Messages Task Tailored messages and quitting. There was no difference in the type and length of Tailored statements received by Quitters and Non-Quitters. This rules out the possibility that any differences found between Quitters and Non-Quitters was due to incomparable stimuli received by the participants. Behavioral Results. Overall participants seemed to have been paying attention in the scanner during the Messages Task, as supported by results in the surprise memory task outside the scanner. The average proportion of correctly identifying an old statement was 0.73, which was greater than chance level of.50, t(87) = 13.32, p <. 001. Also, interestingly, participants seemed to accurately remember the Tailored statements more than the Untailored statements, contrast F(1,85) = 43.81, p <.001, and the Neutral statements, contrast F(1,85) = 44.27, p <.001. Similarly, the Untailored statements were remembered better than the Neutral statements, contrast F(1,85) = 7.01, p <. 01. The mean recognition accuracies for the Tailored, Untailored, and Neutral statements were 0.84, 0.72, and 0.63, respectively. This indicates preliminary evidence for levels of processing effect 8 with Tailored messages. While Tailored messages were remembered better than Untailored and Neutral messages, there was no relationship between better memory performance for the Tailored messages and smoking cessation (r = 0.05, ns). There was also no correlation between memory performance and dmpfc activation during Tailored messages (r =.09, ns). These results suggest that neither attention to messages nor memory performance predicts quitting, although self-related response in dmpfc during tailored message processing does. In the surprise recognition memory task, we presented only 10 old statements from the 50 statements for each message type presented in the scanner, and showed an equal number of new statements; the researchers have no information if memory performance for the statements would differ if participants were tested with all old statements with an equal number of lures. The researchers also have no information if memory performance between the two groups would have changed as a function of time. fmri Results: Tailored or Untailored messages relative to Neutral messages. We identified the neural activation patterns associated with the Tailored or Untailored smokingcessation messages relative to the Neutral messages in the whole sample. As expected, the Tailored and Untailored smoking-cessation messages produced a robust pattern of activation in the MPFC and precuneus/posterior cingulate regions (Supplementary Figures 2a & 2b).

Neural response to tailored messages and quitting Page 8 Supplementary Figures 2c & 2d show that activations in the MPFC and precuneus/posterior cingulate regions overlapped with regions associated with self-appraisal processes, using the mask generated from the Self-Appraisal Task. These results extend previous work by Chua et al. 4, who showed differential activation patterns between High- and Low-Tailored messages in contrast to Untailored smoking-cessation messages, but did not contrast both Tailored and Untailored messages with Neutral messages, and the overlap between these activations with independently identified self-related processing regions. Supplementary Tables 3 and 4 list the brain regions activated more by the Tailored or Untailored messages, relative to the Neutral messages, and that survived the family-wise error correction of <.05 and cluster extent equal or greater than 50 voxels. A. x = 6 B. x = 6 C. x = 6 D. x = 6 Supplementary Figure 2. Activations in the MPFC and precuneus/posterior cingulate regions during processing of (A) Tailored messages and (B) Untailored messages, relative to Neutral messages, across 87 subjects. Common areas of activation in the MPFC and precuneus between (C) Tailored messages and self-related processing (Self > Valence), and between (D) Untailored messages and self-related processing. The coordinates are in MNI space. Supplementary Table 3. Peak activations to Tailored messages, relative to Neutral messages, in the Messages Task. Region Gyri / Subregions MPFC Superior frontal gyrus Medial frontal gyrus Anterior cingulate gyrus Middle frontal gyrus Supplementary motor area Medial orbital gyrus L Inferior Frontal Gyrus L inferior orbital gyrus L middle temporal gyrus L temporal pole L inferior temporal gyrus L superior temporal gyrus L insula BA x, y, z a T Z b k d 10, 9, 32, 11, 6, 8, 24 0, 54, 30 10.50 9.04 c 2219 47, 21, 38, 20, 45 51, 0, 36 9.84 8.47 c 599

Neural response to tailored messages and quitting Page 9 L fusiform gyrus R Inferior Frontal Gyrus 47, 38 54, 24, 3 8.05 6.93 284 R inferior orbital gyrus R insula R temporal pole R superior temporal gyrus Precuneus / Posterior Cingulate 7, 31, 23, 24, 30 6, 60, 39 12.52 10.78 c 1820 Middle cingulate gyrus Cuneus L Inferior Parietal Lobule 40, 39, 7 51, 63, 45 12.17 10.01 c 668 L angular gyrus L superior temporal gyrus L middle temporal gyrus L Middle Frontal Gyrus 8, 6, 9 39, 15, 48 9.48 7.80 143 L Middle Temporal Gyrus 21, 22 54, 30, 6 8.68 7.33 172 R Inferior Parietal Lobule 40, 39 57, 51, 36 8.13 6.99 321 R angular gyrus R supramarginal gyrus R superior temporal gyrus R Middle Temporal Gyrus 21, 38 51, 9, 36 7.30 6.42 69 R inferior temporal gyrus R Cerebellum 27, 81, 39 7.09 6.27 101 R Middle Frontal Gyrus 9, 6, 8 42, 18, 45 7.05 6.25 51 R Middle Temporal Gyrus 21, 22 57, 30, 6 6.90 6.13 65 L Cerebellum 24, 81, 36 6.35 5.73 103 L, left; R, right. a Stereotactic coordinates of the peak voxel from the Montreal Neurological Institute atlas, left/right (x), anterior/posterior (y), and superior/inferior (z), respectively. b SPM-generated Z score, significant at FWE p <.05. c Estimated Z score, significant at FWE p <.05. d Spatial extent of the cluster in voxels (50 voxel minimum). Supplementary Table 4. Peak activations to Untailored messages, relative to Neutral messages, in the Messages Task. Region Gyri / Subregions MPFC Medial frontal gyrus Superior frontal gyrus Middle frontal gyrus Anterior cingulate gyrus Medial orbital gyrus Superior orbital gyrus Precuneus / Posterior Cingulate Middle cingulate gyrus Cuneus L Inferior Parietal Lobule L angular gyrus L supramarginal gyrus L superior temporal gyrus R Superior Frontal Gyrus R frontal medial gyrus R supplementary motor area R Inferior Parietal Lobule R angular gyrus BA x, y, z a T Z b k d 10, 9, 32, 8, 11, 24 3, 63, 6 8.46 7.19 1574 7, 31, 23, 24, 6 0, 21, 36 10.76 9.14 c 1761 40, 39 48, 63, 42 9.95 8.71 c 423 8, 10, 9, 6 27, 54, 33 7.46 6.53 277 40, 39 57, 54, 45 7.27 6.40 201

Neural response to tailored messages and quitting Page 10 R supramarginal gyrus R superior temporal gyrus L Inferior Temporal Gyrus 20, 21 54, 6, 39 7.27 6.40 105 L middle temporal gyrus L Middle Temporal Gyrus 21 66, 27, 15 6.85 6.10 83 L inferior temporal gyrus L, left; R, right. a Stereotactic coordinates of the peak voxel from the Montreal Neurological Institute atlas, left/right (x), anterior/posterior (y), and superior/inferior (z), respectively. b SPM-generated Z score, significant at FWE p <.05. c Estimated Z score, significant at FWE p <.05. d Spatial extent of the cluster in voxels (50 voxel minimum). fmri Results: Tailored messages > Untailored messages. Supplementary Figure 3 and Supplementary Table 5 list the brain regions preferentially engaged by Tailored messages. These regions showed greater activation to Tailored messages relative to Untailored messages. A. x = 6 B. z = 3 Supplementary Figure 3. Activations in (A) the dmpfc, pre-supplementary motor area, precuneus and (B) the inferior frontal gyri during processing of Tailored messages relative to Untailored messages, across 87 subjects. The coordinates are in MNI space. Supplementary Table 5. Peak activations to Tailored messages, relative to Untailored messages, in the Messages Task. Region BA x, y, z a Z b k c Gyri / Subregions dmpfc 9, 8, 10, 6 3, 48, 36 5.52 204 Medial frontal gyrus Superior Frontal Gyrus R Inferior Frontal Gyrus 47, 45, 46 51, 24, 3 6.19 379 R inferior orbital gyrus R middle frontal gyrus R insula L Inferior Frontal Gyrus 47, 45, 44 48, 27, 9 5.46 325 L inferior orbital gyrus L middle frontal gyrus L temporal lobe L insula L Precuneus 31, 7 9, 51, 36 3.83 60 L middle cingulate gyrus L cuneus Pre-Supplementary Motor Area 9, 8, 10, 6 3, 24, 57 6.20 140 Superior Frontal Gyrus Supplementary Motor Area L Superior Temporal Gyrus L middle temporal gyrus L angular gyrus 39, 22 54, 54, 15 4.75 172

Neural response to tailored messages and quitting Page 11 L parietal lobule L supramarginal gyrus R Superior Temporal Gyrus 22, 21 57, 30, 0 4.24 88 R middle temporal gyrus L Cerebellum 15, 78, 27 4.08 62 L Dorsolateral Prefrontal Cortex 9, 8, 6 45, 12, 45 4.91 184 L Middle Frontal Gyrus L precentral gyrus L medial frontal gyrus L inferior frontal gyrus R Dorsolateral Prefrontal Cortex 9, 8 42, 12, 45 4.29 90 R middle frontal gyrus R medial frontal gyrus R inferior frontal operculum L Middle Temporal Gyrus L superior temporal gyrus 21, 22 69, 42, 0 3.85 58 L, left; R, right; dmpfc, dorsomedial prefrontal cortex a Stereotactic coordinates of the peak voxel from the Montreal Neurological Institute atlas, left/right (x), anterior/posterior (y), and superior/inferior (z), respectively. b Z score, significant at FDR p <.05. c Spatial extent of the cluster in voxels (50 voxel minimum). Conjunctive Regions of Interest (ROI) mask. The overlap of self-related processing [Self > Valence, FWE.05, k = 50] and tailored message processing [Tailored > Untailored, FDR.05, k =50] yielded 3 common regions: dorsomedial prefrontal cortex (128 voxels), precuneus (60 voxels), and angular gyrus (30 voxels). These regions were both preferentially engaged by self-related processing and by Tailored messages. These regions were saved as a mask and treated as regions-of-interest in examining whether neural response in these regions during Tailored messages (relative to Neutral messages) predicted quitting. Results are presented in the main manuscript. x = -6 z = 33 Supplementary Figure 4. Conjunctive ROI mask from self-related processing and tailored message processing.

Neural response to tailored messages and quitting Page 12 Although it is true that the Tailored messages used second person pronouns ( you ) more than the Untailored messages did, the conjunctive mask required that the activation be present for both the Tailored > Untailored and the Self > Valence contrasts. Exactly the same words were used for the self-reference and valence judgment conditions (and no pronouns at all). Thus, the association between neural response to tailored messages and quitting cannot be due to a change in pronouns in the Tailored messages. Other outcome-predictive regions. We also examined if there are activations in regions outside the conjunctive mask that correlated with quitting. Using the regions preferentially engaged by Tailored messages over linguistically comparable Untailored messages (Supplement Table 5; FDR.05, k = 50) as regions-of-interest, and that did not overlap with self-related processing regions as defined by the Self-Appraisal task results, we found that greater activation in bilateral inferior frontal gyri during Tailored messages positively correlated with quitting (left BA 47; peak xyz coordinates: 51, 24, 3; ß =.54, S.E. =.19, Wald χ 2 = 8.29, p =.004; O.R. = 1.72; right BA 47; peak xyz coordinates: 48, 27, 9; ß =.69, S.E. =.23, Wald χ 2 = 8.9, p =.003; O.R. = 1.98). Of course, the role of inferior frontal gyrus in predicting quitting suggests that there could be other mechanisms involved, e.g., inhibitory control processing 9 or autobiographical memory 10 and related cognitive control processes 11 engaged by the tailored messages. Activity in other parts of the self-related processing network (ventral and rostral mpfc areas) also predicted quitting, but these areas are not included because they were not significantly activated by the Tailored > Untailored contrast. fmri Results and Gender: We did not find any gender differences in any of the fmri results in the Messages Task or the Self-Appraisal Task. References: 1. Strecher, V.J., et al. Am J Prev Med 34, 373-381 (2008). 2. Chua, H.F., Chen, W. & Park, D.C. Gerontology 52, 306-313 (2006). 3. Chua, H.F., Liberzon, I., Welsh, R.C. & Strecher, V.J. Biol Psychiatry 65, 165-168 (2009). 4. Schmitz, T.W. & Johnson, S.C. Neuroimage 30, 1050-1058 (2006). 5. Schmitz, T.W., Rowley, H.A., Kawahara, T.N. & Johnson, S.C. Neuropsychologia 44, 762-773 (2006). 6. Velicer, W.F. & Prochaska, J.O. Addict Behav 29, 51-60 (2004). 7. Noll, D.C., Genovese, C.R., Vazquez, A.L., O'Brien, J.L. & Eddy, W.F. Magn Reson Med 40, 633-639 (1998). 8. Craik, F. L. Memory 10, 305-318 (2002). 9. Swick, D., Ashley, V., & Turken, A.U. BMC Neurosci 2008, 9:12 (2008). 10. Svoboda, E., McKinnon, M.C. & Levine, B. Neuropsychologia 44, 2189-2208 (2006). 11. Badre, D. & Wagner, A.D. Neuropscyhologia 45, 2883-901 (2007).