Predictors of Enrollment of Older Smokers in Six Smoking Cessation Trials in the Lung Cancer Screening Setting: The Smoking Cessation at Lung Examination (SCALE) Collaboration

Abstract

Significance

Increased rates of smoking cessation will be essential to maximize the population benefit of low-dose CT screening for lung cancer. The NCI’s Smoking Cessation at Lung Examination (SCALE) Collaboration includes eight randomized trials, each assessing evidence-based interventions among smokers undergoing lung cancer screening (LCS). We examined predictors of trial enrollment to improve future outreach efforts for cessation interventions offered to older smokers in this and other clinical settings.

Methods

We included the six SCALE trials that randomized individual participants. We assessed demographics, intervention modalities, LCS site and trial administration characteristics, and reasons for declining.

Results

Of 6285 trial- and LCS-eligible individuals, 3897 (62%) declined and 2388 (38%) enrolled. In multivariable logistic regression analyses, Blacks had higher enrollment rates (OR 1.5, 95% CI 1.2,1.8) compared to Whites. Compared to “NRT Only” trials, those approached for “NRT + prescription medication” trials had higher odds of enrollment (OR 6.1, 95% CI 4.7,7.9). Regarding enrollment methods, trials using “Phone + In Person” methods had higher odds of enrollment (OR 1.6, 95% CI 1.2,1.9) compared to trials using “Phone Only” methods. Some of the reasons for declining enrollment included “too busy” (36.6%), “not ready to quit” (8.2%), “not interested in research” (7.7%), and “not interested in the intervention offered” (6.2%).

Conclusion

Enrolling smokers in cessation interventions in the LCS setting is a major priority that requires multiple enrollment and intervention modalities. Barriers to enrollment provide insights that can be addressed and applied to future cessation interventions to improve implementation in LCS and other clinical settings with older smokers.

Implications

We explored enrollment rates and reasons for declining across six smoking cessation trials in the lung cancer screening setting. Offering multiple accrual methods and pharmacotherapy options predicted increased enrollment across trials. Enrollment rates were also greater among Blacks compared to Whites. The findings offer practical information for the implementation of cessation trials and interventions in the lung cancer screening context and other clinical settings, regarding intervention modalities that may be most appealing to older, long-term smokers.

Lung cancer mortality can be reduced by 20%–24% via low-dose computed tomographic lung cancer screening (LCS) and treatment of early stage disease.^1,2 In 2013, the United States Preventive Services Task Force (USPSTF) assigned a Grade B recommendation for annual LCS for eligible individuals aged 55–80 with at least a 30 pack-year smoking history, who are current smokers or who have quit in the past 15 y.³ Using these criteria, there are approximately four million current smokers who are eligible for LCS in the U.S.^4,5 Importantly, modeling studies have shown that the maximum health impact from LCS will only be achieved if current smokers undergoing lung screening receive assistance to stop smoking.^6,7 LCS may also provide a teachable moment by motivating smokers to consider reducing or quitting smoking.⁸

The NCI’s Smoking Cessation at Lung Examination (SCALE) Collaboration^9,10 includes eight ongoing randomized cessation trials occurring in the setting of LCS. The SCALE trials are assessing the impact of various methods of evidenced-based cessation counseling and medications on cessation outcomes, with results of the trials expected in 2021–2022. The purpose of the collaboration is to enable cross-project research by sharing study methods, measures, and data between the smoking cessation intervention trials. To facilitate this, the SCALE initiative established a Special Measures Collection¹¹ to generate pooled data that will contribute to the knowledge base on how to provide cessation programs most effectively to long-term smokers who are being screened for lung cancer. A primary advantage of pooled analyses is the ability to assess the role of trial level variables, including study design and intervention modalities, to better understand enrollment patterns.¹²

The existing literature on predictors of enrollment in cessation trials conducted in other settings suggests that race and intervention modality have impacted enrollment,^13–16 with lower rates among Blacks,^13,14,17 and higher rates through web-based recruitment methods^18,19 and the provision of free medications.^20–23 A Cochrane review reported that personal and proactive recruitment methods were associated with higher enrollment rates in smoking cessation studies.²⁴ The cessation trials previously conducted in the LCS setting^25–30 provided minimal information on predictors of trial enrollment, highlighting the need for further investigation on the barriers to enrollment in smoking cessation interventions in this setting.³¹

Enrolling smokers in cessation interventions in the LCS setting is a major priority in order to maximize the population benefit of LCS. In the current paper, we used pooled data to assess rates and predictors of trial enrollment in six of the SCALE trials, which collectively have approached over 6000 older, long-term smokers who were eligible for LCS. We examined multiple predictors of enrollment, including participant demographic characteristics, intervention modality, LCS site characteristics, and trial characteristics. We also assessed the reasons for refusing enrollment in the cessation interventions to improve future outreach efforts in clinical settings that serve a high proportion of older adult smokers.

Methods

Overview

For these analyses, we included the six SCALE trials that randomized participants^9,10 and excluded the two trials that randomized screening sites³² or providers.³³ All individuals included in these analyses were eligible for both LCS and trial participation. Study procedures were approved by the institutional review boards of the six institutions and their affiliated LCS sites. We included data collected from the start of each trial through September 30, 2019, at which point five of the six trials were continuing to enroll participants. As a result, the final sample sizes of these trials will differ from what is presented here. The aims and methods of several of the trials are described in detail elsewhere.^9,32,34,35

Participants

The participant eligibility criteria for each trial are described in the Supplementary Table. All trials included participants who met the USPSTF eligibility criteria for LCS, with four trials utilizing the National Comprehensive Cancer Network’s (NCCN)³⁶ Group 1 criteria (55–80 y, 30+ pack years), and two trials utilizing NCCN’s broader Group 2 criteria that were recently recommended by the USPSTF³⁷ (50–80 y, 20+ pack years, plus one additional risk factor). Other differences between trials included the definition of current smoking (past 7 d vs past 30 d), readiness to quit (no requirement vs within next 12 wk), LCS completion (required vs not required), mental health exclusions (no exclusions vs some exclusions), and language (English only vs. English or Spanish).

Procedures

SCALE trial teams determined the available common data-set needed to conduct this comparison from the SCALE Special Measures Collection.¹¹ The project and analysis plans were developed during an annual SCALE meeting, with the primary goal of assessing the impact of trial interventions and methods on participation rates. The SCALE Steering Committee approved the planned analyses, the NCI team collated the pooled dataset, and the Georgetown Lombardi team conducted the analyses. The procedures and intervention modalities used in each trial are presented in Table 1 and the Supplementary Table.

Table 1.

Trial Administration Characteristics and LCS Site Characteristics

SCALE site and N approached for participation	MDACC (N = 619) (NCT03059940)	MGH (N = 443) (NCT03611881)	MUSC (N = 469) (NCT03927989)	MSK/NYU (N = 1103) (NCT03315910)	UMN (N = 982) (NCT02597491)	GU (N = 2669) (NCT03200236)
Methods to identify and reach potentially eligible participants	Radio, print, and social media ads seek smokers interested in LCS and smoking cessation. Interested callers screened for eligibility and approached for participation by the study team	Patients scheduled for LCS are prescreened for eligibility and approached for participation by the study team	Patients scheduled for LCS are approached for participation by the study team	Patients scheduled for LCS are prescreened by the LCS site staff. Eligible patients who provide permission to be referred to the study are approached for participation by the study team	Patients scheduled for LCS are identified via the EMR, invitation letter is mailed, phone follow-up by the study team. Also used provider referrals and approached patients in person at LDCT.	Patients scheduled for LCS are prescreened and approached for recruitment by the LCS sites
Methods used to enroll eligible participants	Phone, in-person	Phone	Phone, in-person	Phone	Phone, in-person	Phone
Counseling Method Used	Telephone, video-based and in-person counseling, quitline	Telephone and video-based counseling	Telephone counseling	Telephone and in-person counseling, quitline	Telephone and in-person counseling	Telephone counseling
Cessation Medications Offered	NRT plus Bupropion or Varenicline	NRT only	NRT only	NRT only	NRT plus Bupropion or Varenicline	NRT only
Study Burden: Post-intervention follow-up assessments	3 assessments: (6 wk, 3 mo, 6 mo)	2 assessments: (3 mo, 6 mo)	3 assessments: (1 mo, 3 mo, 6 mo,)	2 assessments: (3 mo, 6 mo)	6 assessments: (1 mo, 2 mo, 3 mo, 6 mo, 12 mo, 18 mo)	3 assessments: (3 mo, 6 mo, 12 mo)
Timing of Incentives	First payment at baseline	Delayed incentives	First payment at baseline	First payment at baseline	First payment at baseline	Delayed incentives
Ownership type of lung screening hospital sites	1 university owned	6 privately owned	5 university owned	6 privately owned, 2 university owned	1 privately owned, 1 publicly owned (VA), 1 university owned	8 privately owned
Hospital site location	1 urban site	6 urban sites	5 urban sites	6 urban, 2 urban cluster sites	3 urban sites	7 urban, 1 urban cluster site

Abbreviations: MDACC: MD Anderson Cancer Center; MGH: Massachusetts General Hospital; MUSC: Medical University of South Carolina; MSK: Memorial Sloan Kettering; NYU: New York University School of Global Health; UMN: University of Minnesota; GU: Georgetown University

Measures

Participant Characteristics

(Supplementary Table).

Demographics

The demographic variables available for both groups (declined and enrolled) included age, gender, race, and ethnicity.

Enrolled/Declined status

Individuals were defined as enrolled after providing verbal and/or written informed consent and completing a baseline assessment. Decliners were eligible for the trial, were contacted for participation, but did not complete a baseline assessment. Active and passive (no response) refusers are included among the decliners.

Reasons for Declining Trial Enrollment (Figure 1)

Those who actively declined trial enrollment provided their reason(s) for declining. Investigators reached consensus on the seven most widely cited categories across the trials: (1) not interested/too busy, (2) not ready to quit, (3) not interested in research, (4) not interested in the particular cessation intervention offered by a trial, (5) other (e.g., health concerns, transportation barriers), (6) missing (e.g., not asked for a reason, refused to answer), and (7) passive decliners (eligible individuals who trial staff attempted to contact but were never reached).

Trial Intervention Characteristics (Table 1)

Regarding cessation medications, we categorized trials into two groups: (a) “NRT Only” (four trials that included nicotine replacement therapy (NRT) patches and/or lozenges) vs. (b) “NRT+” (two trials that included NRT plus prescription medication). Regarding counseling modality, we categorized trials into two groups: (a) “Phone Counseling” (three trials that conducted phone-based counseling, including state quitlines, phone-based counseling, and video-based counseling) vs. (b) “Multimodal Counseling,” (three trials that included phone plus in-person counseling).

Trial Administration Characteristics (Table 1)

Identifying and Reaching Potentially Eligible Participants.

Four trials utilized clinic lists of patients scheduled for LCS to identify and contact potentially eligible participants. One trial (MDACC) utilized radio, print, and social media advertisements and potentially eligible individuals then contacted the trial by phone or web to express interest in LCS and trial enrollment. Another trial (UMN) identified age-eligible individuals via the electronic medical record (EMR) and contacted those potentially eligible by letter, phone, in-person contact, or through provider referrals.

Study Enrollment Method.

This variable pertains to the method used for study enrollment, during which time the trial details were described (e.g., study burden and incentives), and eligible individuals provided informed consent and completed the baseline assessment. We stratified enrollment methods into two groups: (a) sites that used phone contact only, or “Phone Only” sites; and (b) sites that approached individuals by phone and/or in-person, “Phone+.”

Study Burden.

We used the number of follow-up assessments as a measure of study burden. The number of follow-up assessments was 2, 3, or 4+, and occurred over 6, 12, or 18 months post-randomization. We divided trials into two groups: (1) two trials that had two follow-up assessments vs. (2) four trials that had 3 or more follow-up assessments.

Timing of Financial Incentives.

Trials varied on the timing of incentive distribution: (1) four trials disbursed incentives starting at baseline vs. (2) two trials provided the first incentive after the first follow-up assessment.

Lung Cancer Screening Site Characteristics.

Five trials enrolled participants from multiple hospital- and clinic-based LCS sites and one trial enrolled participants from a single hospital-based LCS site. We assessed hospital ownership (privately owned, publicly owned (e.g., VA), or university owned),³⁸ as well as location (urban: 50 000+ people, urban cluster: >2500 and <50 000 people).³⁹

Additional Trial Characteristics.

We excluded other eligibility criteria from the multivariable analyses (readiness to quit, past 7 d vs. past 30 d definition of current smoking at baseline, LCS completion, mental health status, and language [English only vs. English and Spanish]), as well as other trial characteristics (method of obtaining informed consent and 6 mo vs. ≥12 mo for the final follow-up assessment). We excluded these variables because: (a) a response category applied to only one trial, which resulted in a comparison of one trial vs. all of the others, or (b) there was collinearity with other variables already in the model (e.g., method of obtaining consent and method of enrollment).

Statistical Analyses

Prior to pooling the data, we explored the associations between the participant demographics and the enrolled/declined status variable within each trial (data not shown). Next, we pooled the data to formally evaluate the effects of both individual characteristics as well as the LCS site- and trial-level variables on enrollment. To determine the enrollment rate across the six trials, we calculated the proportion of those who enrolled divided by the total number of eligible individuals contacted for enrollment. We compared decliners to those who enrolled on demographics, intervention methods, and LCS site and trial-related characteristics (Table 2). Chi-square tests were used to evaluate the variables of interest. We excluded individuals reporting more than one race due to the small sample size (N = 84). Ethnicity was missing in almost one-third of those who declined participation and thus was not used in the multivariable analyses. We used logistic regression models to explore independent predictors of enrollment, including the demographic characteristics, intervention methods, LCS site characteristics, and trial administration variables (Table 3).

Table 2.

Comparison of Participants vs. Decliners across 6 SCALE Trials*

	Decliner 62.0% N = 3897	Participant 38.0% N = 2388	Total N = 6285	OR [95% CI]
Participant Characteristics	% (N) % (N) % (N)
Age  50–59 (ref)  60–69  70–80  Missinga	28.8 % (1034) 51.5% (1851) 19.7% (710) (302)	32.4% (722) 51.0% (1215) 16.6% (396) (5)	30.2% (1806) 51.3% (3066) 18.5% (1106) (307)	0.88 [0.78–0.99] 0.75 [0.64–0.87]
Gender  Male (ref)  Female  Other/Missing/Refuseda	53.8% (1970) 46.2% (1694) (233)	52.7% (1251) 47.3% (1122) (15)	53.4% (3221) 46.6% (2816) (248)	1.0 [0.94–1.2]
Race  White (ref)  Black.  Other/Multi-racea  Missing/Not reported  Refused/Unknowna	92.9% (3109) 7.1% (237) (42) (509)	90.1% (2039) 9.9% (224) (42) (83)	91.8% (5148) 8.2% (461) (84) (592)	1.4 [1.2–1.7]
Ethnicity  Non-Hispanic (ref)  Hispanic  Missing/Not reported Refused/Unknowna	93.5% (2426) 6.5% (170) (1301)	97.0% (2257) 3.0% (69) (62)	95.1% (4683) 4.9% (239) (1363)	0.44 [0.33–0.58]
Intervention Characteristics
Medication Type  NRT only (ref)  NRT+prescription	85.3% (3323) 14.7% (574)	57.0% (1361) 43.0% (1027)	74.5% (4684) 25.5% (1601)	4.4 [3.9–4.9]
Counseling Method  Phone Counseling only (ref)  Multimodal Counseling	64.8% (2525) 35.2% (1372)	44.2% (1056) 55.8% (1332)	57.0% (3581) 43.0% (2704)	2.3 [2.1–2.6]
LCS Site Characteristics
Hospital Type  Private (ref)  Public (VA)  University	73.5% (2863) 2.1% (80) 24.5% (954)	52.5% (1254) 11.0% (263) 36.5% (871)	65.5% (4117) 5.5% (343) 29.0% (1825)	7.5 [5.8–9.7] 2.1 [1.9–2.3]
Location  Urbanized areas (ref)  Urban clusters	78.3% (3053) 21.7% (844)	82.9% (1980) 17.1% (408)	80.1% (5033) 19.9% (1252)	0.75 [0.65–0.85]
Trial Administration Characteristics
Enrollment Method  Phone only (ref)  Phone+In Person	82.8% (3228) 17.2% (669)	73.5% (1756) 26.5% (632)	79.3% (4984) 20.7% (1301)	1.7 [1.5–1.9]
Number of Post-intervention Follow-up assessments  2 (ref)  3+	29.4% (1146) 70.6% (2751)	16.8% (400) 83.2% (1988)	24.6% (1546) 75.4% (4739)	2.0 [1.8–2.4]
Timing of Incentives  Starting with follow-up (ref)  Starting with baseline	55.0% (2142) 45.0% (1755)	40.6% (970) 59.4% (1418)	49.5% (3112) 50.5% (3173)	1.8 [1.6–2.0]

*Data complete as of September 30 2019 (N = 6285)

^aNot included in the measures of association analysis

Table 3.

Multivariable Logistic Regression Models Predicting Enrollment Collapsed Across SCALE Trials (N = 5602)

	Model 1	Model 2	Model 3	Model 4
Participant Characteristics	OR [95% CI]	OR [95% CI]	OR [95% CI]	OR [95% CI]
Age  50–59 (ref)  60–70  70–80	0.86 [0.76–0.97] 0.75 [0.64–0.88]	0.99 [0.87–1.1] 0.91 [0.77–1.1]	0.87 [0.77–0.99] 0.77 [0.65–0.90]	0.99 [0.87–1.1] 0.92 [0.77–1.1]
Race  White (ref)  Black	1.4 [1.2–1.7]	1.4 [1.1–1.7]	1.4 [1.2–1.7]	1.5 [1.2–1.8]
Intervention Characteristics
Medication Type  NRT only (ref)  NRT+prescription		6.0 [5.2–6.9]		6.1 [4.7–7.9]
Counseling Methoda  Phone Counseling only (ref)  Multimodal Counseling			2.4 [2.2–2.7]
LCS Site Characteristics
Hospital Type  Private (ref)  Public (VA)  University				1.2 [.77–1.8] 0.66 [0.49–0.90]
Location  Urbanized areas (ref)  Urban clusters				1.1[0.95–1.3]
Trial Administration Characteristics
Enrollment Method  Phone only (ref)  Phone+In Person				1.6 [1.2–1.9]
Number of Post-Intervention  Follow-up Assessments  2 (ref)  3+				1.2 [0.99–1.5]
Timing of Incentives  Start with follow-up (ref)  Start with baseline				1.0 [0.79–1.3]

^aDue to collinearity between the counseling method and medication type variables, Models 2 and 3 were conducted with these variables separately. Due to the stronger association with enrollment for medication type, this variable was used in the final model.

For the significant univariate predictors (p < 0.05), collinearity diagnostics were performed by calculating the variance inflation factor (VIF) and using a VIF value >10 as an indicator of multicollinearity.⁴⁰ To explore independent predictors of enrolling in the SCALE trials, we conducted a series of four multivariable logistic regression models, with demographic characteristics (Model 1), demographic plus trial intervention variables (Models 2 & 3), and previous predictors plus LCS site characteristics and trial administration characteristics (Model 4). Due to collinearity between the counseling method and medication type variables, Models 2 and 3 were conducted with these variables separately. Due to the stronger association with enrollment for medication type, this variable was used in the final model.

SAS version 9.4 was used to pool the data and SPSS version 26 was used for the statistical analyses.

Results

Trial Enrollment

Of the 6285 (average n = 1047.5 per trial; range = 443 to 2669) trial eligible individuals who were approached for enrollment, 2388 (38%; range = 18%–70%) consented and completed the baseline assessment. Univariate analyses indicated that trial participants were significantly younger, more likely to be Black, and less likely to be Hispanic compared to the 3897 (62%) who actively declined or could not be reached (Table 2). Regarding the intervention characteristics, “NRT+” trials and “Multimodal Counseling” trials had significantly higher enrollment rates compared to the “NRT Only” and the “Phone Counseling” trials, respectively (Table 2). Regarding LCS site characteristics and trial characteristics, enrollment was significantly higher in public and university hospitals compared to private hospitals, and was significantly lower among urban cluster hospitals compared to urban hospitals. Compared to their respective counterparts, enrollment was significantly higher among trials that approached individuals by phone and in-person, had 3 or more follow-up assessments, and provided incentives immediately following the baseline assessment (Table 2).

Multivariable Logistic Regression Models

We conducted four sequential multivariable models to explore the independent predictors of trial enrollment (Table 3). In Model 1, both younger age and Black race were associated with higher enrollment. In Model 2, which included both age and race, “NRT+” trials had substantially higher enrollment rates (OR = 6.0, 95% CI 5.2, 6.9) compared to “NRT Only” trials. In Model 3, due to the collinearity between medication type and counseling method (see Table 2), we entered counseling method alone, which indicated that “Multimodal Counseling” trials had significantly better enrollment rates than “Phone Counseling Only” trials (OR = 2.4, 95% CI 2.2, 2.7). In Model 4, given that the medication method variable had a much higher odds ratio than counseling method, we included medication method, along with all other variables that had a significant univariate association with the outcome. There was a higher odds of enrollment among Blacks, among individuals accrued to trials that offered “NRT+,” and to trials conducting enrollment by “Phone+In Person” methods. There was a lower odds of enrollment among individuals accrued to trials conducted at university hospitals (vs. private hospitals).

Sensitivity Analysis

The two trials (MDACC and UMN) that offered NRT and prescription medication were also two of the three trials that used “Phone+In Person” methods of enrolling individuals. Because we were unable to evaluate the unique contribution of these two variables on enrollment, a sensitivity analysis was conducted by entering MDACC and UMN independently with the “NRT Only” trials. This analysis indicated that each “NRT+” trial independently had higher enrollment compared to “NRT Only” trials.

Reasons for Declining Trial Enrollment

Across the six trials, 25.4% of the responses given provide potentially addressable reasons that can be explored in future trials in order to increase enrollment (Figure 1): “not ready to quit” (8.2%), “not interested in research” (7.7%), “not interested in the intervention offered” (6.2%), and “other reasons” (3.3%; e.g., health concerns, transportation barriers). The remaining 74.6% of potential participants did not provide reasons that can be easily addressed, including passive decline/never reached (32.8%), missing (5.3%), and “too busy/not interested” (36.6%), which needs further exploration to understand whether the interventions appear burdensome or possibly ineffective. Reasons for declining enrollment did not differ by gender (p = 0.29) or race (p = 0.33). However, individuals in their 50s were more likely to passively decline (43.1%) compared to their older counterparts (35.5% among those 60–69 years old and 31.7% among those 70–80 years old; p < 0.001). Older smokers (70–80 years old) were more likely to report not being ready to quit as a reason for declining (11.7%) versus 50–59 and 60–69 year olds, (6.7% and 9.6%, respectively; p < 0.001). When reasons for declining were collapsed into active and passive reasons, individuals in their 50s were more likely to passively decline (78.1%) compared to their older counterparts (73.9% among those 60–69 years old and 73.4% among those 70–80 years old; p < 0.05). This finding is consistent with the uncollapsed reasons for declining.

Figure 1.

Distribution of Reasons for Declining Trial Participation

Discussion

This cross-project pooled analysis provides a comprehensive evaluation of enrollment rates and the predictors of trial enrollment in six of the SCALE intervention trials. Demographic characteristics, trial and site characteristics, as well as the tobacco treatment intervention methods offered by the trials are common across many randomized cessation trials. As such, the results presented here can inform the implementation of cessation interventions conducted in both research and clinical settings,³¹ particularly relevant in contexts with a high prevalence of older smokers and where smoking cessation support is outside the usual scope of care.

Several important findings emerged from these analyses. Although there was some variation across the six trials, on average, over one-third of eligible smokers enrolled, which is similar to some cessation trials,^13,41 although somewhat lower,²⁴ and somewhat higher,⁴²compared to other trials, depending on the denominator used.⁴² Despite considerable resources and effort across all trials, the large proportion of decliners illustrates the importance of understanding and improving enrollment methods in cessation interventions for older smokers who are seeking LCS.

The multivariable analyses revealed that both intervention modality and trial characteristics predicted enrollment. The offer of NRT+prescription cessation medication was the strongest predictor of trial enrollment. This finding corroborates prior research on medication giveaway studies for smoking cessation, which show that giving free medication increases enrollment.^20,21 The higher enrollment in the “NRT+” trials may have resulted because NRT is more commonly utilized than prescription medication, and participants may be seeking different strategies than those they have previously utilized. In addition, the financial value of the free prescription medications is significant, as they are typically more costly than NRT. The two trials that offered NRT + prescription medication were also two of the three trials that used the phone + in person methods of enrolling individuals, and thus we were unable to distinguish the independent impact of these variables on enrollment. Thus, the data suggest that tobacco treatment programs integrated into the LCS setting should consider using multiple modes of enrolling participants²⁴ and providing a choice of multiple medication and counseling modalities to increase participation among older smokers. This finding complements the evidence that multi-modality cessation interventions are also most efficacious in achieving abstinence among those eligible for LCS.⁴³ However, LCS programs would need to consider methods to address the potential practical limitations of medication management and cost, including the need for in-person or telehealth visits in this setting. While enrollment rates from a clinical trial may vary from smoking cessation programs incorporated as standard of care, the findings from these analyses shed light on potential ways to increase enrollment in both clinical and research settings.

The analyses also indicated that the smokers who enrolled did not differ on age and gender compared to the smokers who declined trial participation. The unexpected but encouraging finding that Blacks were more likely to enroll relative to Whites suggests that approaching smokers about cessation at the time of LCS registration may promote engagement of Black smokers in cessation treatment. This is of particular importance as prior studies have found that compared to Whites, Blacks who smoke are less likely to be assessed for tobacco use, to receive advice to quit by their providers,^44,45 or to enroll in cessation trials.^13,17,46 However, this finding is consistent with a prior study exploring tobacco treatment engagement and cessation among National Lung Screening Trial participants, in which Blacks were more likely to enroll in tobacco treatment and to express interest in quitting, compared to Whites.⁴⁷ This illustrates the importance of examining individual characteristics among LCS participants, to understand how this subset of smokers may differ from more well-known populations of smokers. Although this is a promising finding, the overall percentage of Black participants in the SCALE trials was low, indicating that it will be important for future trials to include sites in areas of the country that serve a greater proportion of Black communities.

The specific reasons for trial refusal provided by 25% of the decliners suggest the need to ask older, long-term smokers about preferences for types of cessation interventions or other methods of delivery that they would find acceptable. Further, addressing concerns about research participation and about not being ready to quit will be necessary. Including all smokers, regardless of their readiness to quit, is important given that the majority of individuals who enrolled in prior cessation trials conducted in the LCS setting were not ready to quit within the next 30 d.^25,27,29,30 Trials that have enrolled smokers not ready to quit have reported increased readiness to quit as well as abstinence, indicating that their inclusion is worthwhile.^48,49 One limitation in our analysis is that the classification of reasons for refusal were created post-hoc and fit to reasons that varied by trial. While categories of reasons were agreed upon by the trial investigators, they lacked a more formal assessment. Similarly, data were missing on reasons for refusal for many participants. Further research is needed to determine whether these barriers remain prevalent outside the scope of a clinical trial.

Several variables had significant univariate associations with enrollment, although they were not significant in the final multivariable model (e.g., age, the number of follow-up assessments, and the timing of when incentives were provided). It may be useful to consider these variables in the design of future clinical trials to further assess their association with enrollment. Additionally, although individuals in their 60s were the largest age group overall, those in their 50s were more likely to participate than to decline, suggesting that older individuals may require additional resources for accrual (e.g., targeted communication about the benefits of quitting at older ages). Finally, for trials with the ability to provide financial incentives, doing so at enrollment may be more effective than waiting until the follow-up assessments.⁵⁰

The results should be interpreted in light of the study’s methodological limitations. First, we collapsed trials using the state quitline with those conducting their own phone-based counseling and/or video-based counseling for the phone counseling modality variable. We acknowledge the varied experience across these modalities, but this was necessary to achieve the goal of comparing enrollment rates between trials offering only remote methods of counseling vs. remote and in-person counseling. Future studies should explore how various types of remote counseling relate to willingness to enroll in a cessation intervention. Although both the “NRT+” and “Multimodal Counseling” interventions were significantly associated with greater enrollment, due to the collinearity of these variables, we were unable to determine which additional component was most appealing to participants. Because the trials assessed eligible individuals prior to obtaining informed consent, we were unable to collect additional data (e.g., nicotine dependence, SES) on those who declined, and there were missing demographic data, particularly on ethnicity. This prevented its inclusion in the multivariable analyses, and thus the enrollment of Hispanics in cessation trials in this setting requires further study. Further limitations are that the eligibility criteria were not identical across trials (Supplementary Table) and the reasons for declining were not fully harmonized, resulting in some missing data for a few of the reasons. We also note we did not account for the potential correlation of enrollment outcomes within each trial. In other words, we have assumed that knowing the enrollment outcome of an individual from a trial will not help us predict the enrollment outcome of a different individual from the same trial. We addressed the potential limitations of pooled analyses by using common definitions and coding for the variables of interest, estimating the study-specific effects using logistic regression, dividing the studies into groups based on characteristics (e.g., trial administration characteristics, LCS site variables) that could be associated with heterogeneity, utilizing VIF to identify multicollinearity, and conducting sensitivity analyses.¹² It should be noted that the two trials with both in-person enrollment and inclusion of prescription medication (MDAAC and UMN) differed from the other trials in an additional way, as these two trials reached out to potentially eligible participants, rather than using lists of patients already registered for LCS. It is possible that the way individuals were approached impacted enrollment, but due to the overlap, we cannot tease apart these components. Future research will be needed to examine these aspects of trial design independently.

The strengths of this pooled analysis include that as a part of the SCALE Collaboration, over 6000 smokers were systematically offered participation in six cessation trials in the LCS setting, using a harmonized data set to compare those who enrolled vs. declined trial participation. These results may be generalizable primarily to cessation intervention trials conducted with older adults. However, the results may also provide practical information that can inform the implementation of cessation interventions in LCS facilities and other clinical settings, regarding the counseling and medication intervention modalities that may be most appealing to older, long-term smokers. These analyses suggest that interest in using certain types of medication may differ across smokers and thus offering patients additional choices of cessation interventions may be beneficial to enrollment, as demonstrated by the two trials with the highest enrollment (MDA, UMN).

In summary, our findings synthesized rates and predictors of enrollment for six of the SCALE smoking cessation intervention trials for LCS participants. We found the highest rates of enrollment when NRT + prescription medications were offered. Additionally, using multiple methods of accrual was associated with higher enrollment. In order to maximize the reach of tobacco treatment interventions, it is critical that LCS programs eliminate barriers for integration of tobacco treatment services, that multimodal approaches are utilized to engage smokers, and that a wide range of smoking cessation treatments are offered to encourage enrollment.

Supplementary Material

A Contributorship Form detailing each author’s specific involvement with this content, as well as any supplementary data, are available online at https://academic.oup.com/ntr.

Funding

This research was funded by the National Cancer Institute: R01 CA207228 (Taylor), P30 CA051008 (Weiner). R01 CA207078 (Cinciripini), P30 CA016672 (Pisters), R01 CA196873 (Joseph), R01 CA207442 (Ostroff/Shelley), P30CA008748 (Thompson), R01CA218123 (Park/Rigotti/Haas), R01 CA207229 (Toll), K07CA214839 (Rojewski), R01CA207158 (Foley). The research was also supported in part by the generous philanthropic contributions to The University of Texas MD Anderson Moon Shots Program. The content of this article reflects the views of the authors. The funders had no role in the design of the study; the collection, analysis, and interpretation of the data; the writing of the manuscript; and the decision to submit the manuscript for publication.

Declaration of Interests

Dr. Minnix reports personal fees from UptoDate, Dr. Rigotti reports personal fees from UpToDate and personal fees from Achieve Life Sciences, Dr. Toll reports personal fees from Pfizer and personal fees from Expert Testimony. All of these consultant positions are outside the submitted work. There are no additional conflicts of interest to report.