Utilizing Lung Cancer Risk Prediction Models to Promote Smoking Cessation: Two Randomized Controlled Trials

Abstract

Purpose:

The current project sought to examine whether delivery of lung cancer risk projections (calculated using the Liverpool Lung Project [LLP] risk model) predicted follow-up smoking status.

Design:

Two single-blinded randomized controlled trials.

Setting:

Stop Smoking Services in Liverpool (United Kingdom).

Participants:

Baseline current smokers (N = 297) and baseline recent former smokers (N = 216) were recruited.

Intervention:

Participants allocated to intervention groups were provided with personalized lung cancer risk projections, calculated using the LLP risk model.

Measures:

Baseline and follow-up questionnaires explored sociodemographics, smoking behavior, and lung cancer risk perceptions.

Analysis:

Bivariate analyses identified significant differences between randomization groups, and logistic regression models were developed to investigate the intervention effect on the outcome variables.

Results:

Lung cancer risk projections were not found to predict follow-up smoking status in the trial of baseline current smokers; however, they did predict follow-up smoking status in the trial of baseline recent former smokers (odds ratio: 1.91; 95% confidence interval: 1.03-3.55).

Conclusion:

The current study suggests that lung cancer risk projections may help maintain abstinence among individuals who have quit smoking, but the results did not provide evidence to suggest that lung cancer risk projections motivate current smokers to quit.

Keywords

smoking cessation cigarette smoking smoking cancer of lung lung cancer risk perception

Purpose

Lung cancer is the most commonly diagnosed cancer worldwide. In 2012, it was estimated that worldwide there were approximately 1.8 million new lung cancer cases and 1.6 million lung cancer mortalities.¹ In the United Kingdom alone, approximately 86% of lung cancer cases have been attributed to tobacco smoking.² Despite smoking cessation being identified as one of the most effective strategies in reducing lung cancer incidence,³ smoking cessation success remains at a mere 1% to 5% of smokers each year.⁴ Poor smoking cessation success rates demonstrate the need for new, innovative, and effective tobacco control ventures.

One of the most promising approaches to smoking cessation interventions for entire populations are tailored risk communications.⁵ A recent Cochrane review suggested that personalized risk information was associated with increased informed choice, increased knowledge, and more accurate risk perceptions.⁶ Furthermore, although relatively few studies have identified a causal association between generic risk communication and behavior change, the importance of tailoring risk communications to the individual characteristics of targets has previously been emphasized.⁷

Lung cancer risk prediction models are statistical models that estimate the probability of developing lung cancer within a given time period; such models consider and incorporate various risk factors. Lung cancer risk prediction models may offer a new opportunity for the delivery of tailored risk communications, as previous research has failed to explore the application of lung cancer risk prediction models in the context of smoking cessation. The Liverpool Lung Project (LLP) developed a lung cancer risk prediction model for predicting 5-year risk, based on 579 lung cancer cases and 1157 age- and sex-matched population-based controls.⁸ Various risk factors were incorporated into the model including age, sex, smoking history, occupational exposure to asbestos, prior diagnosis of pneumonia, prior diagnosis of malignant tumor (except lung cancer), and family history of lung cancer. The model demonstrates good discrimination between cases and controls, with a reported area under the curve of 0.71.⁸ Furthermore, the model has been validated within 3 independent populations (United Kingdom, Europe, and North America)⁹ and has been utilized to recruit high-risk individuals into the UK Lung Cancer Screening Trial.¹⁰

The data within this article were derived from 2 pragmatic randomized controlled trials (RCTs), which examined the feasibility and efficacy of providing personalized lung cancer risk projections to current and recent former smokers, with the intention of enhancing smoking cessation rates at 6-month follow-up. The current project sought to achieve this objective by delivering lung cancer risk projections, calculated using the LLP risk model among individuals recruited via a local Stop Smoking Services (SSS) in North West England. Stop Smoking Services typically aim to support smokers within local communities to quit by providing a range of pharmacotherapy products and behavioral therapies.¹¹ Secondarily, the effect of lung cancer risk projections on lung cancer risk perceptions was also explored in order to provide further insight regarding the relationships between risk communication, risk perception, and smoking behavior. If lung cancer risk projections using lung cancer risk prediction models do predict follow-up smoking status, this has important implications; delivery of lung cancer risk projections could reduce the burden of smoking-related diseases and in turn smoking-related mortalities.

Methods

Sample

All participants were consented via community drop-in sessions delivered by a SSS in North West England between November 2013 and June 2014. The drop-in sessions enabled service users to attend at any point within a given time period, and the number of service users that attended the sessions ranged greatly (0-30). It should be noted that the number of occasions that participants had attended SSS drop-in sessions previously was not reported. Participants were aged 18 to 60 years, and participants were excluded from the project if they had previously been diagnosed with lung cancer. The first trial incorporated a sample of current smokers (ie, individuals who had smoked within the previous week; n = 302), however, at 6-month follow-up, 5 participants were either deceased or relocated to an untraceable address, resulting in a final sample of 297 participants (see Figure 1). The second trial incorporated a sample of recent former smokers (ie, those who had already stopped smoking and had not smoked at all in the 7 days prior to recruitment; n = 219), although at 6-month follow-up, 3 participants were either deceased or relocated to an untraceable address, resulting in a final sample of 216 participants (see Figure 1). At baseline, the median number of days abstinent in the second trial (of recent former smokers) was 39.0 days (interquartile range [IQR] = 21.0-75.0), with a minimum and maximum reported number being 7 and 600 days abstinence, respectively (although only 4 participants reported abstinence for over 1 year).

Figure 1.

Flow of participants through the 2 project trials.

Design and Procedure

Ethical approval was acquired for the project via Liverpool Central National Research Ethics Service Committee. Participants were made aware that they could withdraw from the study at any time, data were anonymized, strict confidentiality guidelines were adhered to, and participants were aware that the results derived from the data they provided may be published in a scientific journal.

The project entailed the implementation of 2 RCTs. The first RCT consisted of baseline current smokers and the second RCT consisted of baseline recent former smokers; the trials were undertaken in parallel. Although participants were designated to a trial based on smoking status, both trials followed the same design and procedure, which will now be described and illustrated in Figure 1.

Upon arriving at a community drop-in session delivered by a local SSS, service users were provided with a participant information sheet. Following service users’ consultations with a smoking cessation advisor, service users were introduced to a researcher by the smoking cessation advisor in order to avoid influencing service users’ decisions to participate. Service users were offered the opportunity to discuss trial participation in greater detail, privately but still within the drop-in session locality. Service users who were happy to participate were requested to sign a consent form and complete a baseline questionnaire. The researcher offered to complete the questionnaire with all participants and pens were provided for those who wished to complete the questionnaire without the support of the researcher.

Following completion of the baseline questionnaire, participants were stratified into one of the 2 trials, dictated by a participant’s baseline smoking status (ie, baseline current smokers or baseline recent former smokers); smoking status classification is detailed in the sample description. Within the respective trial, participants were subsequently randomized into one of the 2 groups—(1) the control group or (2) the intervention group. Randomization software was utilized to allocate participants on a 1:1 ratio (via the URL, http://www.randomization.com/). Participants were blinded to randomization group allocation; participants in both the intervention and control arms were informed that they would receive lung cancer risk information, but the nature of the information (ie, generic or personalized) was not disclosed until debriefing.

Participants allocated to the control groups for both trials were provided with simplistic, generic smoking risk communication in the form of a pamphlet. The generic pamphlet simply stated the association between smoking and lung cancer and highlighted that quitting smoking was the best thing to do to avoid many serious diseases, including lung cancer. Participants allocated to the intervention groups for both trials were provided with the same generic pamphlet as above, but additionally, they received the intervention (detailed later in this article).

Participants were informed that they would be contacted at 6 months to ascertain outcome variables. Six-month follow-up questionnaires were predominantly undertaken by telephone. Telephone calls were attempted 3 times, at different times of the day, before a paper questionnaire was dispatched to a participant’s address with a stamped addressed envelope and letter, requesting completion. Follow-up responses relied upon participants’ goodwill, as no financial (or other) incentives were used. At 6-month follow-up, participants were debriefed regarding randomization blinding and study aims. Participants allocated to the control arm were offered the intervention upon completing the follow-up questionnaire and being debriefed regarding blinding.

Measures

Questionnaires were completed at baseline and at 6-month follow-up. Baseline questionnaire measures included variables pertaining to sociodemographics, lung cancer risk factor exposure, smoking behavior, and lung cancer risk perceptions. Age, gender, ethnicity, marital status, highest educational attainment, and socioeconomic status were ascertained. The measures for ethnicity and highest educational attainment have been previously adopted.¹² The measure for marital status was based on the measure used as part of the LLP,⁸ whereas socioeconomic status was ascertained using the English Index of Multiple Deprivation ranks. The Index of Multiple Deprivation is a robust index, which uses 38 separate indicators of deprivation.¹³ The Index of Multiple Deprivation information was obtained using a website developed by Mimas at the University of Manchester,¹⁴ and the output was reported as ranks within 5 quintiles, as described elsewhere.¹⁵ Socioeconomic status and marital status variable levels were, however, subsequently transformed, due to some low cell frequencies. The transformed variable for socioeconomic status consisted of (1) most deprived (most deprived) and (2) least deprived (above average deprivation, average deprivation, below average deprivation, least deprived). The transformed variable for marital status consisted of (1) other (divorced, separated, widowed, and other); (2) single (single); and (3) married or living together (married, living together).

Data were also collected pertaining to additional lung cancer risk factor exposure, as guided by the LLP risk model⁸; details regarding occupational exposure to asbestos, prior diagnosis of pneumonia, prior diagnosis of malignant tumor (except lung cancer) and family history of lung cancer were all established as part of the baseline questionnaire. Although these 4 variables were not of relevance to trial data analyses, they were required to calculate lung cancer risk projections.

Smoking behavior was also investigated. Smoking status was measured at baseline using 7-day point prevalence (PP). Seven-day PP is commonly used in smoking cessation trials and can be advantageous, as it captures the dynamic, real-life process of smoking cessation.¹⁶ Age started smoking, whether or not the participant lived with another smoker, and cigarettes per day (retrospectively where applicable) were recorded. Nicotine dependence was measured using the Fagerström test for nicotine dependence (FTND).¹⁷ The Fagerström test for nicotine dependence scores were calculated based on 6 items, and scores ranged from 0 to 10, low to high dependency. Baseline quit duration (in days) was also calculated among those in the second trial (ie, baseline recent former smokers).

Several lung cancer risk perceptions were additionally measured. Measures for perceived personal lung cancer risk, perceived lung cancer risk of the average smoker, and perceived relative risk of lung cancer were developed based on previously applied measures,¹⁸ whilst the measures for lung cancer worry and perceived lung cancer survival were adapted from a prior study.¹⁹

The 6-month follow-up questionnaire entailed several repeated measures, including smoking status (using 7-day PP), quit duration, and all aforementioned lung cancer risk perceptions, with the exception of perceived lung cancer survival, as this measure was not anticipated to be an outcome. Furthermore, an intention-to-treat approach was adopted at follow-up, which entailed classifying participants lost to follow-up as current smokers (with the exception of those who were deceased and those documented as having moved to an untraceable address).²⁰

Intervention

The researcher delivered the intervention to participants individually, immediately following completion of the baseline questionnaire, which took approximately 10 minutes.

The researcher brought a laptop computer to all drop-in sessions. A Microsoft Access database was saved on the laptop computer. The database incorporated formulae associated with the LLP risk model (see elsewhere)⁸ and included a user interface, enabling the researcher to input data collected from a participant’s questionnaire to ascertain a participant’s risk profile. The database was able to provide an individual risk profile, using the LLP risk model formulae, by inputting a participant’s age, gender, smoking duration, occupational exposure to asbestos, prior diagnosis of pneumonia, prior diagnosis of malignant tumor (except lung cancer) and family history of lung cancer. The database was able to incorporate all of these risk factors and estimate projected 5-year lung cancer risk at the age of 70 years.

Lung cancer risk at 70 years was estimated for all participants in the intervention groups, and estimates were provided for 2 hypothetical circumstances—(1) continued smoking from present until 70 years old and (2) smoking cessation from present until 70 years old. Providing 2 hypothetical estimates enabled the researcher to demonstrate the benefit of stopping smoking compared to continuing to smoke. For example, a 43-year-old male, with a smoking duration of 33 years, a previous diagnosis of pneumonia, no previous malignancies, no family history of lung cancer, and no previous asbestos exposure, demonstrated a 12% projected 5-year lung cancer risk at 70 years if he continued to smoke, whereas his projected 5-year lung cancer risk at 70 years was only 5% if he had stopped smoking at the age of 43 years.

Lung cancer risk projections were detailed verbally to participants, including a brief explanation as to which risk factors informed the projections, the meaning of projected 5-year lung cancer risk at 70 years, the projected difference between stopping and continuing to smoke, and a plain English summary of the LLP risk model. Participants were also offered the opportunity to ask questions about lung cancer risk projections. Furthermore, lung cancer risk projections were also presented in a pamphlet. The researcher calculated risk using the Microsoft Access database and discussed the resulting risk projections, with the aid of the pamphlet.

The pamphlet stated “If you were to continue smoking, your estimated risk of getting lung cancer between the ages of 70 and 74 years old will be […] % BUT if you quit smoking from now on, your estimated risk of getting lung cancer between the ages of 70 and 74 years old will be reduced to […] %.” Beside, each of the 2 estimations was an illustration that included 100 squares. The appropriate number of squares was colored red by the researcher to represent the percentage of risk for each of the estimations. The pamphlet also provided a background on the LLP risk model, highlighting that the projections were based on the information provided by the participant and that if this information differed from now until the age at which risk is projected, the results may also differ. Lastly, the pamphlet provided contact details if the participant wished to find out more about the LLP risk model or the information provided. Participants were provided with the pamphlet to keep.

Analysis

Both trial data sets were analyzed using the same analytical approach. Bivariate tests were undertaken to explore differences in baseline participant characteristics across randomization groups. Relationships were ascertained using χ² test or Fisher exact test as appropriate for categorical variables. For continuous variables, Mann-Whitney U tests were utilized to identify significant differences between variable levels. Logistic regression analyses were subsequently conducted to explore the effect of the intervention on follow-up smoking status, perceived personal lung cancer risk, perceived average smoker lung cancer risk, perceived relative risk of lung cancer, and lung cancer worry. Purposeful selection of covariates and potential confounders were included as per previous guidance (ie, the logistic regression models adjusted for variables significant at the level of 25%).²¹ All analyses were performed using IBM-SPSS statistical software (version 22.0).²²

It should be noted that since there is a paucity of research regarding the provision of personalized lung cancer risk projections, the anticipated effect of the intervention was uncertain. A power calculation was undertaken to determine the required sample size for the trial considering baseline current smokers. The calculation indicated that a sample size consisting of 785 baseline current smokers (randomized on a 1:1 basis) was required to detect a 10% difference in smoking cessation, considering 80% power for a 5% 2-sided type 1 error, as guided by the literature.²³

Results

Trial 1: Baseline Current Smokers

The first trial explored the intervention effect among baseline current smokers. The median age for the sample overall was 42.0 years (IQR = 31.0-51.0) and most baseline current smokers were female (n = 177, 59.6%), white (n = 271, 92.2%), and single (n = 157, 53.2%). Participants marginally tended to have higher qualifications (ie, achieving qualifications beyond General Certificate of Secondary Education level; n = 154, 53.1%), although the vast majority were classified within the most deprived quintile with regard to socioeconomic status (n = 255, 86.1%). Table 1 displays the distribution of baseline participant characteristics across randomization groups in respect of the first trial.

Table 1.

Baseline Participant Characteristics by Randomization Group Among Baseline Current Smokers in Trial 1.

Baseline Variable	Control (n = 146, 49.2%)	Intervention (n = 151, 50.8%)	P Value
Age, median (IQR)	42.5 (33.0-51.0)	41.0 (30.0-50.0)	.154^a
Gender			.343
Female	83 (56.8)	94 (62.3)
Male	63 (43.2)	57 (37.7)
Ethnicity^b			.802
White	134 (91.8)	137 (92.6)
Other	12 (8.2)	11 (7.4)
Marital status^b			.823
Other	21 (14.4)	23 (15.4)
Single	76 (52.1)	81 (54.4)
Married or living together	49 (33.6)	45 (30.2)
Highest educational attainment^b			.803
Basic or no qualifications	66 (46.2)	70 (47.6)
Higher qualifications	77 (53.8)	77 (52.4)
Socioeconomic status^b			.003^a
Most deprived	117 (80.1)	138 (92.0)
Least deprived	29 (19.9)	12 (8.0)
Age started smoking, median (IQR)	15.0 (13.0-17.0)	15.0 (13.0-17.0)	.600
Living with another smoker^b			.423
No	51 (34.9)	58 (39.5)
Yes	95 (65.1)	89 (60.5)
FTND, median (IQR)	6.0 (4.0-7.0)	6.0 (4.0-7.0)	.854
Cigarettes per day, median (IQR)	20.0 (11.3-25.0)	20.0 (14.5-20.0)	.668
Perceived personal lung cancer risk^b			.514
Low	23 (16.4)	28 (19.2)
Moderate	53 (37.9)	61 (41.8)
High	64 (45.7)	57 (39.0)
Perceived average smoker lung cancer risk^b			.688
Very low to moderate	33 (23.2)	39 (26.9)
Somewhat high	54 (38.0)	49 (33.8)
Very high	55 (38.7)	57 (39.3)
Perceived relative risk of lung cancer^b			.024^a
About the same to a little higher risk	26 (18.8)	31 (21.4)
Twice as high risk	17 (12.3)	29 (20.0)
Five times higher risk	37 (26.8)	48 (33.1)
Ten times higher risk	58 (42.0)	37 (25.5)
Lung cancer worry^b			.986
Rarely or never	43 (30.1)	44 (29.7)
Sometimes	53 (37.1)	54 (36.5)
Often or all the time	47 (32.9)	50 (33.8)
Perceived lung cancer survival^b			.869
Less than a quarter	51 (36.4)	54 (37.8)
About a quarter	44 (31.4)	39 (27.3)
About half	32 (22.9)	37 (25.9)
About three-quarters to nearly all	13 (9.3)	13 (9.1)

Abbreviations: FTND, Fagerström Test of Nicotine Dependence; IQR, interquartile range.

^a P < .25.

^bFigures do not equate to 297 due to some missing data.

A number of bivariate tests were undertaken to examine the relationships between randomization groups and the aforementioned baseline participant characteristics (see Table 1). Age (P = .154), socioeconomic status (P = .003), and perceived relative risk of lung cancer (P = .024) were adjusted throughout the logistic regression analyses, as these variables were significant at the level of 25%.²¹ There were no significant associations between randomization group and gender, ethnicity, marital status, highest educational attainment, age started smoking, living with another smoker, FTND, cigarettes per day, perceived personal lung cancer risk, perceived average smoker lung cancer risk, lung cancer worry, and perceived lung cancer survival.

Logistic regression analyses were conducted to explore the intervention effect on 6-month follow-up outcome variables. The intervention failed to predict any of the 6-month follow-up outcome variables, including 7-day PP (smoking status; P = .658), perceived personal lung cancer risk (P = .785), perceived average smoker lung cancer risk (P = .950), perceived relative risk of lung cancer (P = .580), and lung cancer worry (P = .455).

Trial 2: Baseline Recent Former Smokers

The second trial explored the intervention effect among baseline recent former smokers. The median age was 44.0 years (IQR = 37.0-52.8). The majority of baseline recent former smokers were female (n = 118, 54.6%) and white (n = 197, 91.6%), while single was the most frequently reported marital status (n = 92, 43.0%). Participants marginally tended to have lower qualifications (ie, achieving General Certificate of Secondary Education level or below; n = 114, 53.0%), and again, the vast majority were classified within the most deprived quintile in relation to socioeconomic status (n = 182, 84.3%).

Several bivariate analyses were undertaken to examine the relationships between randomization group and baseline participant characteristics (see Table 2). Age (P = .122), gender (P = .243), ethnicity (P = .241), marital status (P = .178), highest educational attainment (P = .001), and quit duration (P = 0.156) were adjusted throughout the logistic regression analyses, as these variables were significant at the level of 25%.²¹ There were no significant associations between randomization group and socioeconomic status, age started smoking, living with another smoker, FTND, cigarettes per day, perceived personal lung cancer risk, perceived average smoker lung cancer risk, perceived relative risk of lung cancer, lung cancer worry, and perceived lung cancer survival.

Table 2.

Baseline Participant Characteristics by Randomization Group Among Baseline Recent Former Smokers in Trial 2.

Baseline Variable	Control (n = 103, 47.7%)	Intervention (n = 113, 52.3%)	P Value
Age, median (IQR)	43.0 (36.0-52.0)	46.0 (38.5-53.0)	.122^a
Gender			.243^a
Female	52 (50.5)	66 (58.4)
Male	51 (49.5)	47 (41.6)
Ethnicity^b			.241^a
White	92 (89.3)	105 (93.8)
Other	11 (10.7)	7 (6.3)
Marital status^b			.178^a
Other	16 (15.5)	22 (19.8)
Single	51 (49.5)	41 (36.9)
Married or living together	36 (35.0)	48 (43.2)
Highest educational attainment^b			.001^a
Basic or no qualifications	43 (41.7)	71 (63.4)
Higher qualifications	60 (58.3)	41 (36.6)
Socioeconomic status^b			.937
Most deprived	87 (84.5)	95 (84.1)
Least deprived	16 (15.5)	18 (15.9)
Age started smoking, median (IQR)	15.0 (14.0-17.0)	15.0 (14.0-17.0)	.812
Living with another smoker^b			.567
No	80 (77.7)	84 (74.3)
Yes	23 (22.3)	29 (25.7)
FTND, median (IQR)	6.0 (5.0-7.0)	6.0 (4.0-8.0)	.943
Cigarettes per day, median (IQR)	20.0 (20.0-30.0)	20.0 (15.0-30.0)	.461
Quit duration, median (IQR), days	35.0 (21.0-70.0)	46.0 (26.0-76.0)	.156^a
Perceived personal lung cancer risk^b			.987
Low	33 (33.0)	35 (32.1)
Moderate	43 (43.0)	47 (43.1)
High	24 (24.0)	27 (24.8)
Perceived average smoker lung cancer risk^b			.277
Very low to moderate	19 (18.6)	22 (20.2)
Somewhat high	51 (50.0)	43 (39.4)
Very high	32 (31.4)	44 (40.4)
Perceived relative risk of lung cancer^b			.447
About the same to a little higher risk	14 (13.9)	21 (19.3)
Twice as high risk	18 (17.8)	24 (22.0)
Five times higher risk	29 (28.7)	23 (21.1)
Ten times higher risk	40 (39.6)	41 (37.6)
Lung cancer worry^b			.937
Rarely or never	36 (35.3)	37 (33.0)
Sometimes	41 (40.2)	46 (41.1)
Often or all the time	25 (24.5)	29 (25.9)
Perceived lung cancer survival^b			.959
Less than a quarter	40 (40.4)	47 (43.1)
About a quarter	30 (30.3)	32 (29.4)
About half	21 (21.2)	23 (21.1)
About three-quarters to nearly all	8 (8.1)	7 (6.4)

Abbreviations: FTND, Fagerström Test of Nicotine Dependence; IQR, interquartile range.

^a P < .25.

^bFigures do not equate to 216 due to some missing data.

Logistic regression analyses were subsequently conducted to explore the intervention effect on 6-month follow-up outcome variables (see Table 3). The intervention was found to significantly predict 7-day PP (smoking status) at 6 months (odds ratio: 1.91; 95% confidence interval: 1.03-3.55); however, the intervention failed to predict follow-up perceived personal lung cancer risk (P = .711), perceived average smoker lung cancer risk (P = .567), perceived relative risk of lung cancer (P = .874), and lung cancer worry (P = .869). Thus, the findings suggest that lung cancer risk projections may promote abstinence among individuals who have recently quit smoking, but the results suggest they do not motivate current smokers to quit.

Table 3.

Outcome Variables at 6-Month Follow-Up Among Recent Former Smokers.

Outcome Variables	Control, n (%)	Intervention, n (%)	Odds Ratio (95% CI)^a	P Value
Primary outcome variable
Seven-day point prevalence				.040^b
Current	66 (64.1)	56 (49.6)	1.00 (base^d)
Former	37 (35.9)	57 (50.4)	1.91 (1.03-3.55)
Secondary outcome variables
Perceived personal lung cancer risk^c				.711
Low	17 (25.8)	23 (29.5)	1.00 (base)
Moderate	27 (40.9)	34 (43.6)	0.76 (0.30-1.92)
High	22 (33.3)	21 (26.9)	0.66 (0.24-1.79)
Perceived average smoker lung cancer risk^c				.567
Very low to moderate	11 (16.2)	10 (12.8)	1.00 (base)
Somewhat high	20 (29.4)	25 (32.1)	1.69 (0.54-5.26)
Very high	37 (54.4)	43 (55.1)	1.77 (0.61-5.17)
Perceived relative risk of lung cancer^c				.874
About the same to a little higher risk	6 (9.4)	8 (10.5)	1.00 (base)
Twice as high risk	10 (15.6)	8 (10.5)	0.82 (0.18-3.67)
Five times higher risk	15 (23.4)	17 (22.4)	1.24 (0.31-4.95)
Ten times higher risk	33 (51.6)	43 (56.6)	1.26 (0.37-4.32)
Lung cancer worry^c				.274
Rarely or never	29 (42.6)	33 (42.3)	1.00 (base)
Sometimes	27 (39.7)	26 (33.3)	1.01 (0.46-2.21)
Often or all the time	12 (17.6)	19 (24.4)	2.18 (0.79-6.00)

Abbreviation: CI, confidence interval.

^aMultivariate logistic regression: adjusted for baseline age, gender, ethnicity, marital status, highest educational attainment, and quit duration.

^b P < .05.

^cFigures do not equate to 216 due to some missing data.

^dBase refers to the baseline variable level to which the other variable levels are compared.

Discussion

Previous research has suggested that tailored risk communications provide one of the most promising approaches to smoking cessation interventions for entire populations,⁵ suggesting that the application of lung cancer risk prediction models in the context of smoking cessation could prove highly beneficial in promoting smoking cessation success rates. The current study identified that lung cancer risk projections were associated with follow-up smoking status among baseline recent former smokers but not among baseline current smokers. To our knowledge, this is the first study to evaluate the utility of a lung cancer risk prediction model in the context of smoking cessation. Our results demonstrate that the output produced from lung cancer risk prediction models (such as the LLP risk model⁸) can be adapted to deliver tailored lung cancer risk projections to the general public.

Behavior change theory may provide some explanation as to why a significant effect was identified in the trial of baseline recent former smokers but not in the trial of baseline current smokers, as well as considering why provision of lung cancer risk projections failed to predict follow-up lung cancer risk perceptions in either of the trials. One of the most dominant models of behavior change that has been applied extensively to smoking behavior^24
-26 is the transtheoretical model of change (TTM).²⁷ The TTM proposes and systematically incorporates several concepts considered influential to behavior change, including the stages of change and the processes of change.^27,28 Furthermore, the TTM stipulates that specific processes of change, such as “reinforcement management”, may be more applicable among individuals progressing from the active stage of change to the maintenance stage of change (ie, baseline recent former smokers), whereas processes of change, such as “self-liberation,” may be more applicable to current smokers progressing from the preparation stage of change to the action stage of change (ie, baseline current smokers). This might suggest that lung cancer risk projections may provide recent former smokers with a reinforcing message pertaining to behavior change, although this message may be less applicable to current smokers preparing to quit smoking. Further research is required to fully understand this mechanism.

The study has some limitations. Firstly, it was not possible to recruit the optimal number of smokers in each of the trials and achievement of abstinence was substantially lower than anticipated for the power calculation regarding the trial of baseline current smokers (actual 21%, compared to an expected 26%); we therefore had insufficient statistical power to conclude superiority of the intervention. A larger trial or extension to the current project would certainly be beneficial to explore whether the results are replicable. Secondly, the current project relied on self-reports with regard to smoking status. Future studies that explore the impact of tailored smoking risk communication should always endeavor to biochemically verify self-reported smoking status.²⁰ Despite this, the value of self-reported smoking status should not be underestimated; 1 review surmised that sensitivity means and specificity means of self-reported smoking status were both high when compared with biochemical indices.²⁹

It should also be noted that 7-day PP was used to measure smoking status at baseline and at 6-month follow-up, a measure that has been argued to be highly advantageous, enabling the dynamic, real-life process of smoking cessation to be captured.¹⁶ Some researchers, however, recommend the use of prolonged abstinence (ie, self-reported continuous abstinence) in addition to PP to further enhance reliability.³⁰ Future research might benefit from inclusion of both measures to improve confidence in the result that abstinence was maintained throughout the follow-up period. Lastly, a clustered RCT design may have also been beneficial. Although attempts were made to avoid contamination of randomization blinding by delivering the intervention to participants in private, some participants may have inadvertently disclosed their randomization group in contact with service users following the SSS drop-in session; it should be noted, however, that this was not apparent.

In terms of behavior change theory, the current study supports the notion that stage-based interventions may be particularly beneficial in promoting long-term smoking cessation; a recent Cochrane review summarized that the effectiveness of stage-based interventions for smoking cessation remains unclear.²⁶ If a future, larger trial is able to replicate the current study findings, a cost–benefit analysis would be beneficial to consider delivery of lung cancer risk projections within SSS on a wider scale. Stop Smoking Services quit rates have remained fairly constant within recent years in England³¹; therefore, new, innovative, and effective interventions would certainly be welcomed nationally and, undoubtedly, internationally. Furthermore, lung cancer risk projections (using the LLP risk model) can be obtained using a simple database, thus, enabling nonclinicians to communicate lung cancer risk with minimal training. It may also be feasible to deliver lung cancer risk projections in alternate settings to local SSS, such as general practitioner surgeries and hospital settings; however, further research is required to explore the effect of providing lung cancer risk projections among other smoking populations.

The current study showed that provision of lung cancer risk projections predicted 6-month follow-up smoking status among baseline recent former smokers but not among baseline current smokers. The delivery of lung cancer risk projections using risk models such as the LLP risk model may improve long-term smoking cessation rates, which could subsequently reduce the burden of smoking-related diseases and mortalities; further research is required.

SO WHAT?

What is already known on this topic?

Smoking cessation is one of the most effective strategies in reducing lung cancer incidence, and tailored risk communications have been identified as one of the most promising approaches to smoking cessation. Lung cancer risk prediction models may offer a new opportunity for the delivery of tailored risk communications.

What does this article add?

To our knowledge, this is the first study to explore the application of a lung cancer risk prediction model in the context of smoking cessation. The findings suggest that lung cancer risk projections may promote abstinence among individuals who have recently quit smoking, but they do not motivate smokers to quit.

What are the implications for health promotion practice or research?

Risk communication delivery using lung cancer risk prediction models may greatly improve long-term smoking cessation rates, which may in turn reduce the burden of smoking-related diseases and mortalities.

Footnotes

Acknowledgments

The authors would like to acknowledge the staff and clients from Roy Castle FagEnds (Roy Castle Lung Cancer Foundation) for their support with the study. Furthermore, the authors would like to thank Liverpool Clinical Commissioning Group and Liverpool Primary Care Trust for funding the study.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This study was funded by Liverpool Clinical Commissioning Group and Liverpool Primary Care Trust.

References

Ferlay

Soerjomataram

Dikshit

. Cancer incidence and mortality worldwide: sources, methods and major patterns in GLOBOCAN 2012. Int J Cancer. 2015;136(5):E359–E386.

Parkin

. Tobacco-attributable cancer burden in the UK in 2010. Br J Cancer. 2011;105(suppl 2):S6–S13.

Thun

DeLancey

Center

Jemal

Ward

. The global burden of cancer: priorities for prevention. Carcinogenesis. 2010;31(1):100–110.

Song

Raftery

Aveyard

Hyde

Barton

Woolacott

. Cost-effectiveness of pharmacological interventions for smoking cessation: a literature review and a decision analytic analysis. Med Decis Making. 2002;22(suppl 5):S26–S37.

Velicer

Prochaska

Redding

. Tailored communications for smoking cessation: past successes and future directions. Drug Alcohol Rev. 2006;25(1):49–57.

Edwards

Naik

Ahmed

. Personalised risk communication for informed decision making about taking screening tests. Cochrane Database Syst Rev. 2013;(2):CD001865.

Gerrard

Gibbons

Reis-Bergan

. The effect of risk communication on risk perceptions: the significance of individual differences. J Nati Cancer Inst Monogr. 1999;(25):94–100.

Cassidy

Myles

van Tongeren

. The LLP risk model: an individual risk prediction model for lung cancer. Br J Cancer. 2008;98(2):270–276.

Raji

Duffy

Agbaje

. Predictive accuracy of the Liverpool Lung Project risk model for stratifying patients for computed tomography screening for lung cancer: a case-control and cohort validation study. Ann Intern Med. 2012;157(4):242–250.

10.

Field

Duffy

Baldwin

. The UK Lung Cancer Screening Trial: a pilot randomised controlled trial of low-dose computed tomography screening for the early detection of lung cancer. Health Technol Assess. 2016;20(40):1–146.

11.

National Institute for Health and Care Excellence (NICE). Public health guidance 10. Stop Smoking Services. London, United Kingdom: National Institute for Health and Care Excellence. 2008. https://www.nice.org.uk/guidance/ph10. Accessed October 7, 2016.

12.

Sherratt

Marcus

Robinson

Newson

Field

. Electronic cigarette use and risk perception in a Stop Smoking Service in England. Addict Res Theory. 2015;23(4):336–342.

13.

Deas

Robson

Wong

Bradford

. Measuring neighbourhood deprivation: a critique of the Index of Multiple Deprivation. Environ Plann C Gov Policy. 2003;21(6):883–903.

14.

GeoConvert. Mimas. http://geoconvert.mimas.ac.uk/. 2014. Accessed October 7, 2016.

15.

McRonald

Yadegarfar

Baldwin

. The UK Lung Screen (UKLS): Demographic profile of first 88,897 approaches provides recommendations for population screening. Cancer Prev Res (Phila). 2014;7(3):362–371.

16.

Velicer

Prochaska

. A comparison of four self-report smoking cessation outcome measures. Addict Behav. 2004;29(1):51–60.

17.

Heatherton

Kozlowski

Frecker

Fagerstrom

. The Fagerström Test for Nicotine Dependence: a revision of the Fagerström Tolerance Questionnaire. Br J Addict. 1991;86(9):1119–1127.

18.

Weinstein

Marcus

Moser

. Smokers’ unrealistic optimism about their risk. Tob Control. 2005;14(1):55–59.

19.

Rutten

LJF

Blake

Hesse

Augustson

Evans

. Illness representations of lung cancer, lung cancer worry, and perceptions of risk by smoking status. J Cancer Educ. 2011;26(4):747–753.

20.

West

Hajek

Stead

Stapleton

. Outcome criteria in smoking cessation trials: proposal for a common standard. Addiction. 2005;100(3):299–303.

21.

Hosmer

Lemeshow

Sturdivant

. Applied Logistic Regression. 3rd ed. Hoboken, NJ: Wiley; 2013.

22.

IBM Corp . IBM SPSS Statistics for Windows, Version 21.0. Armonk, NY: IBM Corp; 2012.

23.

Parkes

Greenhalgh

Griffin

Dent

. Effect on smoking quit rate of telling patients their lung age: the Step2quit randomised controlled trial. BMJ. 2008;336(7644):598–600.

24.

Atak

. A transtheoretical review on smoking cessation. Int Q Community Health Educ. 2007;28(2):165–174.

25.

Robinson

Vail

. An integrative review of adolescent smoking cessation using the transtheoretical model of change. J Pediatr Health Care. 2012;26(5):336–345.

26.

Cahill

Lancaster

Green

. Stage-based interventions for smoking cessation. Cochrane Database Syst Rev. 2010;(11):CD004492.

27.

Prochaska

DiClemente

. Stages and processes of self-change of smoking: toward an integrative model of change. J Consult Clin Psychol. 1983;51(3):390–395.

28.

Prochaska

DiClemente

. Toward a comprehensive model of change. In: Miller

Heather

, eds. Treating Addictive Behaviors. New York, NY: Plenum; 1986:3–27.

29.

Patrick

Cheadle

Thompson

Diehr

Koepsell

Kinne

. The validity of self-reported smoking: a review and meta-analysis. Am J Public Health. 1994;84(7):1086–1093.

30.

Hughes

Carpenter

Naud

. Do point prevalence and prolonged abstinence measures produce similar results in smoking cessation studies? A systematic review. Nicotine Tob Res. 2010;12(7):756–762.

31.

National Centre for Smoking Cessation and Training (NCSCT). Local Stop Smoking Services. Service and delivery guidance 2014. http://www.ncsct.co.uk/usr/pub/LSSS_service_delivery_guidance.pdf. Accessed October 7, 2016.