Parent trial design

The present study is a secondary analysis of data collected from WTQL participants recruited between 7 June 2018 and 25 January 2023 for a 2×2×2×2 factorial experiment that evaluated the main and interactive effects of 4 intervention components: 1) Quitline counseling (1 call vs 4 calls); 2) NRT (2 weeks of nicotine patch vs 4 weeks of nicotine patch with nicotine lozenge); 3) SmokefreeTXT (enrollment information for the National Cancer Institute text messaging program vs no information); 4) Incentives for treatment engagement ($30 for each counseling call completed and staying in SmokefreeTXT for at least 6 weeks post-quit-date vs no incentives). All enrolled participants set a quit date within 2 weeks of enrollment and consented to follow-up assessments at 12 and 26 weeks following the quit date. Additional details are available via ClinicalTrials.gov (NCT03538938) and the primary outcome publication¹².

Participants

For participants to be eligible for recruitment outreach, they needed to satisfy the following criteria: aged ≥18 years, completed at least one counseling call in English with the WTQL in the past 3–6 months, reside in Wisconsin, not share a phone number with someone already enrolled in the trial, and have at least one indicator of socioeconomic disadvantage (Medicaid insurance, no health insurance, and/or no more than a high school education or GED).

Participants in the clinical trial had to meet additional inclusion criteria at recruitment: smoking at least for 6 of the past 7 days, smoking ≥5 cigarettes at least 1 of these smoking days, and being willing to set a quit date within 2 weeks. Female participants had to report that they were not pregnant, breastfeeding, or planning to become pregnant in the next 6 months. To receive NRT, participants had to meet standing WTQL criteria. Enrollment in SmokefreeTXT required controlled access to a mobile device and consent to join the program.

Procedures

The study proactively recruited adult participants 3–6 months after they enrolled in an index round of WTQL counseling (with or without NRT). Quit Line personnel trained in the study protocol made up to 5 attempts to reach prospective participants to assess their current smoking status using a single item: ‘Have you smoked at all, even a puff, in the last 7 days?’. Those reporting no smoking were congratulated and offered standard WTQL support. Those who reported any smoking in the past 7 days were given a brief introduction to the clinical trial and asked if they assented to eligibility screening. Those who passed the screening were asked to provide oral consent for participation. Those who declined trial participation were offered standard WTQL services. Consented trial participants completed a baseline phone survey, were randomized to 1 of 16 conditions, and completed a counseling call. This typically occurred in a single phone call. Participants were then mailed a welcome packet tailored to their randomized condition. Those eligible for medication were mailed NRT with use instructions.

Phone assessments occurred at 12 and 26 weeks post-quit and assessed use of tobacco products and smoking cessation treatments. Participants reporting no smoking in the past 7 days at the follow-up at 26 weeks, were asked to provide a saliva sample for cotinine testing. A sample collection kit was mailed to eligible participants, with a stamped return envelope. Researchers called participants to encourage sample return.

All recruitment, enrollment, and treatment activities were completed by WTQL personnel, while the post-quit assessments at 12 and 26 weeks were completed by research personnel at the University of Wisconsin. The Health Sciences Institutional Review Board (IRB) at the University of Wisconsin approved all study procedures.

Measures

Participant characteristics were assessed primarily with existing WTQL items gathered at registration. These items follow the Minimal Data Set (MDS) guidelines published by the North American Quitline Consortium¹³. Items analyzed for this study include: region of Wisconsin (based on 3-digit zip code and coded as Northeast [reference group], Southeast, Southern, Western, or missing); self-reported sex (female [reference group], male, or missing); age (18–44 years [reference group], 45–64 years, ≥65 years, or missing); race (Black/African American, White [reference group], Other, or missing); ethnicity (Hispanic or Latino, Not Hispanic/Latino [reference group] or missing); health insurance type (Medicaid, Medicare, Commercial or other non-Medicaid or -Medicare insurance [reference group], uninsured, or missing); education level (lower than grade 12/GED, grade 12/GED, some college or higher [reference group], or missing); types of tobacco product used (cigarettes only [reference group], cigar, pipe, or smokeless tobacco; or hookah, e-cigarettes, or other alternative nicotine device, or missing); time to first cigarette from the Fagerström test for nicotine dependence¹⁴ (within 30 min of waking, after 30 min of waking [reference group], or missing); and chronic health conditions (yes if reported asthma, chronic obstructive pulmonary disease, coronary artery disease, and/or diabetes, no if reported none of these conditions [reference group], or missing). All these variables were examined as predictors of inclusion in the trial to assess the representativeness of the sample retained at each step in the recruitment, enrollment, treatment, and outcome assessment process. None was considered to be a confounder.

Participant demographics were reviewed at enrollment, and additional characteristics were assessed in clinical trial participants. These included annual household income from all sources and the national percentile of the Area Deprivation Index (ADI, Version 3.0)^15,16 for the participant’s address at enrollment. The ADI ranks census blocks based on Census and American Community Survey measures of education, employment, housing, and poverty¹⁵. Higher scores on the ADI indicate greater neighborhood socioeconomic disadvantage.

Outcome variables included binary indicators of reaching/completing each phase in the requitting process: outreach (e.g. 1 if reached, 0 if not reached), screening, eligibility, enrollment, treatment engagement (i.e. completing at least 1 counseling call, receiving NRT, and using NRT), and reporting abstaining from smoking for at least 7 days at the post-quit-date follow-up at 12 weeks and 26 weeks. At the follow-up at 26 weeks, self-reported abstinence was confirmed by salivary cotinine (cutoff 10 ng/mL: Forensic Fluids Laboratories, Kalamazoo, MI). Additional outcomes included specific reasons for exclusion at initial outreach and eligibility screening, and binary indicators of completion of the follow-up at 12 weeks and 26 weeks, and of submitting a saliva sample among those reporting abstinence at the follow-up at 26 weeks. For all outcomes, missing cases were treated as not having the outcome (e.g. not reached, not screened, not abstinent). Some outcomes occurred at very low frequencies and were not analyzed, including exclusion due to pregnancy, breastfeeding, or planning to become pregnant (n=18) and not completing at least one counseling call (n=81).

Statistical analysis

Descriptive statistics were used to characterize the full sample of people eligible for trial outreach, for those enrolled in the trial, and for those not enrolled in the trial. Descriptive analyses also captured retention and loss at each step in trial flow to inform the development of coding schemes for predictors. Variables with many missing cases were coded in a separate missing category rather than excluded because differential rates of missingness across items suggest that these data are unlikely to be missing at random and instead may reflect participant attitudes that are non-ignorable. We elected not to use multiple imputation to address missing data because the missingness is potentially informative. In primary analyses, the ADI percentile was treated as a continuous variable. Sensitivity analyses used a dummy indicator of ADI (≥ median of 85) to assess whether this altered the pattern of results.

Logistic regression analyses were conducted for each outcome with sufficient variability for analyses of representativeness. We conducted both bivariate models with a single predictor and multiple logistic regression models with Benjamini-Hochberg corrections¹⁷ to control the false discovery rate at alpha=0.05. Corrected multiple logistic regression analyses were considered primary and are summarized in Supplementary file Figure S1. We did not have a priori hypotheses regarding interactions between patient characteristics and intervention components, but conducted exploratory analyses using hierarchical multiple logistic regression to examine such effects. Analyses were conducted in IBM SPSS Statistics for Windows, version 28.0.0.0 (IBM Corp., Armonk, NY).

Descriptive results

Flow through study procedures, from initial outreach to confirming abstinence 6 months after a target quit date, is shown in Supplementary file Figure S1. A total of 14125 people met the prerequisites for proactive outreach from the quitline. Characteristics of this sample are presented in Table 1. As shown in Table 1, participants were drawn from all regions of Wisconsin, with the greatest representation in the most densely populated region of the state, the Southeast (50.4%). The majority of participants (62.2%) were aged 45–64 years, with fewer <45 years (23.3%) or >65 years (14.2%). More than half of the participants were female (57.8%), while 41.8% were male. Four percent of participants identified as Hispanic or Latino. In terms of race, 26.3% identified as Black or African American, 64.4% as White, 6.5% as another race or multiple races. The majority of participants were covered by Medicaid insurance (56.1%), while 19.4% were covered by Medicare, 12.5% were covered by commercial or other insurance, and 9.9% reported having no insurance. In collinearity checks, correlations among predictor variables were modest (all r<0.37).

Outreach

Phone contact was achieved for 6019 (42.6%) of those called and varied across participants. After Benjamini-Hochberg correction in multiple logistic regression analyses, rates of reaching people were higher for: middle-aged (45–64 years: AOR=1.69; 95% CI: 1.55–1.83) or older (≥65 years: AOR=2.33; 95% CI: 2.07–2.63) people than in those <45 years; men (AOR=1.16; 95% CI: 1.08–1.25 vs women); Black people (AOR=1.18; 95% CI: 1.08–1.30 vs White people); those with Medicare (AOR=1.19; 95% CI: 1.05–1.34 vs commercial insurance); those who used electronic or other alternative nicotine delivery systems (AOR=1.46; 95% CI: 1.25–1.70 vs those who did not); and those who reported having at least one of the four chronic health conditions assessed (AOR=1.12; 95% CI: 1.05–1.21 vs no chronic condition). Rates of reaching people via outreach were lower for: people who did not report their ethnicity at their recent quitline registration (AOR=0.69; 95% CI: 0.52–0.92 vs non-Hispanic or Latino people); those without health insurance (AOR=0.79; 95% CI: 0.68–0.90 vs commercial insurance); and those who smoked within 30 min of waking (AOR=0.90; 95% CI: 0.83–0.98 vs after >30 min). Full results from bivariate and multiple logistic regression analyses are shown in Supplementary file Table S1.

A total of 1635 (11.6% of the full sample) individuals reached by phone were excluded due to reporting not smoking at the re-treatment outreach call. Longer latency to smoke after waking at initial quitline registration predicted greater odds of reporting abstinence at outreach (AOR=1.52; 95% CI: 1.32–1.74). No other predictors were significantly related to abstinence at outreach after correction (Supplementary file Table S2).

Screening

Of those reporting smoking at outreach, 2375 (16.8% of the full sample) completed trial screening. Screening completion rates were significantly higher among participants who identified as Black (AOR=2.32; 95% CI: 1.95–2.76 vs White) and those who had one of the assessed chronic health conditions (AOR=1.34; 95% CI: 1.18–1.53 vs no chronic condition). Screening completion rates were lower among: participants who were >65 years (AOR=0.65; 95% CI: 0.53–0.80 vs those <45 years); those who did not complete high school (AOR=0.61; 95% CI: 0.50–0.74) and those who completed high school or a GED but did not attend college (AOR=0.64; 95% CI: 0.54–0.75) versus those who had some college; and those not reporting their ethnicity at registration (AOR=0.47; 95% CI: 0.27–0.82 vs non-Hispanic or Latino participants). Full results are given in Supplementary file Table S3.

Eligibility

A total of 1744 (12.3% of the full sample) met trial eligibility criteria (Supplementary file Table S4). Eligibility likelihood was elevated in those who smoked within 30 min of waking (AOR=1.88; 95% CI: 1.49–2.36 vs those who waited >30 min). People who identified as Black (AOR=0.63; 95% CI: 0.50–0.79 vs those who identify as White) and people who did not report their ethnicity (AOR=0.25; 95% CI: 0.11–0.56 vs non-Hispanic/Latino participants) passed screening at lower rates. Black individuals were excluded for not smoking enough in the past 7 days at twice the rate of White individuals (AOR=2.03; 95% CI: 1.57–2.62). Those who did not report their ethnicity were nearly 4 times more likely to be ineligible due to insufficient smoking (AOR=3.92; 95% CI: 1.72–8.97). See Supplementary file Table S5 for full results. Unwillingness to set a quit date did not differ significantly across populations after correction (Supplementary file Table S6).

Enrollment

A total of 1316 (9.3% of the full sample) participants consented and enrolled in the parent trial. Reasons for not enrolling are summarized in Supplementary file Figure S1, and the characteristics of those enrolled versus not enrolled are reported in Table 1, along with the p-values for chi-squared tests. In multiple regression analyses with correction (Supplementary file Table S7), people with a high school education or GED (AOR=0.60; 95% CI: 0.45–0.79 vs those with at least some college) and those who did not report their ethnicity (AOR=0.12; 95% CI: 0.03–0.45 vs non-Hispanic or Latino participants) were less likely to enroll than their counterparts.

Treatment engagement

Among 1316 trial enrollees, 1235 (93.8%) completed at least one counseling call with a Quit Coach. Because this rate was so high, we did not analyze differences in completion of counseling calls across participant groups. We did not analyze the number of calls completed, as this was a manipulated experiment, and roughly half of the participants had access to only one proactive counseling call. Rates of NRT use were more variable across participants. A total of 1060 (80.5%) enrollees were eligible for and agreed to use NRT. At the follow-up phone survey at 12 weeks, 966 (73.4%) of enrollees reported they had used NRT since enrolling. Rates of any use of NRT at the follow-up at 26 weeks, were lower with 555 (42.2%) of enrollees reporting use since the last assessment. Rates of receiving NRT and using NRT at 12 weeks did not vary significantly across enrollee populations, but any use of NRT at 26 weeks was reported more frequently by participants aged ≥65 years (AOR=2.42; 95% CI: 1.51–3.86 vs those <45 years) and by Black participants (AOR=1.56; 95% CI: 1.14–2.14 vs White participants). See Supplementary file Tables S8-S10 for full results. Treating ADI as binary (≥ the median ADI of 85 vs <85) did not alter the pattern of results compared with analyses that treated ADI as continuous.

Exploratory analyses of NRT engagement did not detect differences in receiving or reporting using NRT as a function of patient characteristics and randomly assigned NRT condition. We detected no significant interactions between randomly assigned NRT condition (4 weeks of nicotine patch and lozenge vs 2 weeks of nicotine patch) and patient characteristics following correction.

Abstinence

In analyses in which missing=smoking, 25.5% of all enrollees reported no smoking in the past 7 days at the follow-up at 12 weeks, and 12.3% of enrollees provided salivary biochemical confirmation of abstinence (<10 ng/mL cotinine) at the follow-up at 26 weeks. Enrolled participants who identified as Hispanic or Latino were 2.89 times more likely (95% CI: 1.49–5.63) to report abstinence at 12 weeks (Supplementary file Table S10) and 5.73 (95% CI: 2.68–12.21) times more likely to have biochemically verified abstinence at 26 weeks (Supplementary file: Table S11 and Figure S1) than were non-Hispanic/Latino participants. Men were less likely (AOR=0.49; 95% CI: 0.33–0.73) than women to have confirmed abstinence at the follow-up at 26 weeks. No other differences survived correction (Supplementary file Tables S11 and S12). Treating ADI as binary (≥ the median ADI of 85 vs <85) did not alter the pattern of results compared with analyses that treated ADI as continuous. Exploratory moderation analyses with correction did not detect any significant interactions between patient characteristics and the main effects of the 4 intervention components. Hispanic ethnicity and male sex remained significant with interactions when treatment conditions were included in models. No predictors of abstinence were significantly related to follow-up completion or saliva sample returns in Benjamini-Hochberg-corrected multiple logistic regression analyses.

Limitations

Certain factors limit generalization from these results. First, the nature of the population, treatments, and outreach methods may limit generalization beyond Wisconsin quitline participants. For instance, the initial, low-intensity quitline treatment that prospective trial participants received (one coaching call and 2 weeks of a single NRT medication, plus web-based and mailed materials) likely affected the types of individuals available for re-treatment; these findings may not generalize to quitline programs that offer more coaching and NRT. Also, some treatment elements varied over the study period: e.g. the standard text messaging service offered by the quitline was not available for much of the study, and an enhanced treatment program for American Indian populations was added during the study. Also, analyses were limited to variables collected routinely by the quitline, some variables (e.g. race, ethnicity, insurance type) were often missing, and residual confounding may have affected results in adjusted models. Further, race and ethnicity were assessed using broad categories, with limited representation of non-Black minoritized populations. Finally, self-report data are subject to biases, and rates of returning saliva samples for biochemical verification of abstinence at follow-up at 26 weeks were low, which may have contributed to misclassification of abstinence status.