|Home | About | Journals | Submit | Contact Us | Français|
This study characterizes tobacco cessation patterns and the association of cessation with survival among lung cancer patients at Roswell Park Cancer Institute: an NCI Designated Comprehensive Cancer Center.
Lung cancer patients presenting at this institution were screened with a standardized tobacco assessment, and those who had used tobacco within the past 30 days were automatically referred to a telephone-based cessation service. Demographic, clinical information and self-reported tobacco use at last contact were obtained via electronic medical records and the RPCI tumor registry for all lung cancer patients referred to the service between October 2010 and October 2012. Descriptive statistics and Cox proportional hazards models were used to assess whether tobacco cessation and other factors were associated with lung cancer survival through May 2014.
Calls were attempted to 313 of 388 lung cancer patients referred to the cessation service. Eighty percent of patients (250/313) were successfully contacted and participated in at least one telephone-based cessation call; 40.8% (102/250) of persons contacted reported having quit at the last contact. After controlling for age, pack year history, sex, ECOG performance status, time between diagnosis and last contact, tumor histology, and clinical stage, a statistically significant increase in survival was associated with quitting compared to continued tobacco use at last contact (HR=1.79; 95% CI: 1.14-2.82) with a median 9 month improvement in overall survival.
Tobacco cessation among lung cancer patients after diagnosis may increase overall survival.
According to the 2014 Surgeon General's Report, smoking by cancer patients and survivors causes adverse outcomes including increased overall mortality, increased cause- specific mortality, and increased risk for second primary cancers.1 However, few studies have examined whether those who quit as a result of standardized tobacco use assessments with automated referral to cessation support are more likely to survive.2-5 The majority of published literature utilizes retrospective chart reviews, with non-standardized tobacco use determination, if recorded at all, to examine the association of tobacco use with cancer outcomes.3, 6-8 Far fewer studies have evaluated the potential benefit of smoking cessation for improving cancer outcomes, 1, 5-7 though data strongly suggest that smoking cessation may significantly improve survival for lung cancer.9
The need for standardized tobacco use assessments and cessation support has been recognized even though there is little prospective research on tobacco use and cancer survival.6, 7 Since October 2010, Roswell Park Cancer Institute (RPCI) has operated a Tobacco Assessment and Cessation Service (TACS), completing standardized tobacco use assessments for every patient in the thoracic center with automatic referrals to a dedicated tobacco cessation counseling service. Patients can elect to opt out of the service once contacted by a cessation specialist.10 This paper presents data on the following: 1) the lung cancer patients referred to the RPCI TACS; 2) quit rates of the referred patients; and 3) patient survival, comparing patients who stopped smoking with those who did not.
This study was approved by the Institutional Review Board of RPCI. The tobacco assessment and automated referral methods for the RPCI TACS have been previously described in detail.10 In brief, during the routine nursing assessment, all ambulatory patients are screened with a standardized tobacco assessment and if they report tobacco use in the past 30 days are automatically referred to a tobacco cessation counseling service. This service is staffed by cessation specialists who provide behavioral counseling with linkage to pharmacologic support if needed.11 Cessation specialists contact referred patients one to two weeks after referral, with call priority for patients referred as current users. RPCI TACS is a clinically designed and supported service that offers cessation support to patients, while relieving physicians of the burden of providing smoking cessation support while dealing with complex oncology issues. This service created a prospective cohort of thoracic clinic patients to serve as a base for this analysis exploring the association between cessation and survival.
The goal for the cessation counselors was to complete at least one successful contact attempt for as many patients as possible. Patients were considered not reached after five unsuccessful call attempts; if they were not reached, they were sent a letter informing them of the benefits of quitting and quitting tips, and inviting them to call the RPCI TACS for support. During successful phone calls, cessation specialists asked patients if they had been tobacco free for the prior 24 hours. The cessation specialist asked patients who reported cessation if they had been tobacco free for the prior 7 days.
Data were abstracted from the electronic medical record (e.g., age, marital status, referring clinic, pack-year history, ECOG performance status and smoking status at referral), cessation specialist notes (e.g., call dates and outcomes), and the RPCI tumor registry (e.g., race, ethnicity, clinical stage, vital status and survival time in months) between October 2010 and May 2014; survival duration was assessed in May 2014. Data were linked by patient medical record number, compiled into a main dataset and de-identified.
Eligibility included patients seen in the RPCI thoracic center with histologically-confirmed lung cancer, diagnosed on or after October 1, 2010, and who reported tobacco use within the 30 days prior to initial evaluation (n=388). Established patients (i.e. not newly diagnosed with cancer), patients already treated outside of RPCI, and patients presenting for evaluation of biopsy results or for a second opinion were excluded (n=9). Patients in hospice or end-of-life care, as indicated by the EMR records or contact with a family member, were also excluded (n=27/388, 7.0%). Only patients referred to the cessation service between October 2010 and October 2012 and with at least one successful contact by the RPCI TACS were included in the survival analyses (n=250/388, 64.4%).
Some patients who had quit at initial referral were lost to follow-up (n=75/388, 19.3%), while others had one cessation contact before being lost to follow-up (n=137/388, 35.3%). Patients were included in this analysis if they participated in at least one telephone call with the RPCI TACS. To account for the loss to follow-up and variable nature of quitting, the main cessation indicator was self-reported smoking status at the last contact, which was obtained through May 2014. To be considered as having quit, a patient had to indicate that he or she was tobacco-free for at least 24 hours at time of last contact.
ECOG performance status was routinely documented in the EMR for all patients by a combination of nursing staff, advanced practice providers and physicians; and was dichotomized into those with a score of 0 (fully active, as if disease free) and those with a score of 1 or higher (restricted from strenuous activity or more severely limited, including deceased).12 To control for disease severity, ECOG status, along with tumor stage, were included in the final adjusted model. Clinical stage was included as a categorical variable with three categories; stages I and II combined served as the referent, with stage III and stage IV as comparison groups.
Descriptive statistics were used to examine the referral patterns, demographics, disease characteristics, self-reported quit rates at each contact, and self-reported status at the last contact. Frequencies and chi-square tests were used for categorical variables, ANOVA was used to analyze continuous variables and quit rates, and Kaplan-Meier with log rank test for survival time with quit rates. Survival analysis was conducted using a Cox proportional hazards model. Vital status (alive or deceased at the end of follow-up) and survival time in months were the main outcome variables for the survival analysis. The date of diagnosis served at the date of origin for survival time. Variables included in the Cox proportional hazards model with support in the literature included age at diagnosis, pack-years, days between diagnosis and last contact, sex, clinical stage, ECOG status, tumor histology (NSCLC versus other), and tobacco use status at time of referral.3, 13, 14
Between October 2010 and October 2012, 388 thoracic clinic referrals to RPCI TACS were lung cancer patients who met eligibility criteria for inclusion in these analyses with a diagnosis date between October 1, 2010 and October 31, 2012 (Figure 1). About 81% of those referred (313/388) had at least one cessation support call attempted, of which 250/313 (79.9%) patients participated; only 5 (1.3%) actively refused participation. Among 58 patients who did not participate, 27 (7.0%) were in end of life care or reported to be deceased; 26 (6.7%) could not be reached, and 5 (1.3%) were considered inappropriate referrals as never users or they had quit greater than 30 days before referral. 75/388 (19.3%) were not called, but were sent an information pamphlet about the benefits of cessation, quit tips and an invitation to call the service. There were, on average, 21.5 days between referral to the RPCI TACS and first contact.
Among the 250 patients who participated in at least one cessation support call, 149 (59.6%) patients had a follow-up call attempted, with 136 (91.3%) participating in at least one follow-up call (Figure 1). Due to staffing limitations, follow-up calls were not attempted to 101 persons (40.4%) who had one successful contact. No lung cancer patients actively refused participation at the first follow-up call. There were an average of 39.2 days between the initial cessation support call and the next follow-up call. Patients had the opportunity for a maximum of 8 cessation support calls, with a median of 2 calls completed (range: 1-8; mean=2.27).
Self-reported quitting status was assessed at each contact; status reported at each contact was independent of that at prior contact. At time of referral to the RPCI TACS, 20.0% of patients (n=50/250) self-reported having quit within the last 30 days (Table 1). At the initial phone contact, 71 of 250 patients (28.7%) reported having quit, including newly quit patients and those who maintained their status from the time of referral. Among 136 patients who completed at least one follow-up call, 56 (41.8%) self-reported having quit. In total, 102 patients (40.8%) reported at least 24 hours abstinence at last contact, including 65 (63.7%) who had quit before the initial contact.
In bivariate analyses (not shown), patients who had quit at last contact, as compared to current users, had more favorable performance status (47.1% ECOG 0 vs. 33.0% ECOG ≥1 had quit, p=0.024) and were more likely to be alive (48% alive vs. 33.6% deceased had quit, p=0.021). Patients who reported having quit at last contact had a higher mean survival time as compared with patients who reported continued tobacco use (25.2 versus 20.0 months, p=0.004; results not shown).
There was an average of 15.5 months (median: 15.0 months; range: 0-40 months) follow-up for survival time through May 2014. As of May 2014, 51.3% of the lung cancer population included in the final adjusted model were deceased (n=115/224, Table 2). Older age, advanced stage, and current tobacco use at last contact were associated with a higher risk of mortality with control for age at diagnosis, pack-years, days between diagnosis and last contact, sex, tobacco use at referral, ECOG status and tumor histology (Table 2). Current tobacco use at the last contact was associated with a significant increase in death compared to patients who self-reported being quit (HR=1.79, 95% CI: 1.14-2.82). This survival advantage for quitters, after adjusting for the demographic, disease and health characteristics previously mentioned, is depicted graphically in Figure 2. The median survival for patients who reported current tobacco use at last contact was 20 months versus 29 months for those who reported being quit at last contact (Figure 2).
Adjustment for the number of completed calls did not affect the results; the hazard ratio for current tobacco use at last contact remained the same (HR=1.78; 95% CI: 1.13-2.81; results not shown). Restricting the analysis to patients who participated in at least two calls yielded similar results (n=126; HR=2.19; 95% CI: 1.14-4.21; results not shown). Also, restricting to NSCLC patients who participated in one call, results remained similar (n=197; HR=2.01; 95% CI: 1.24-3.26; results not shown). There were only 27 patients with other lung cancer types; the hazard ratio for continuing to use tobacco was similar, but not statistically significant (n=27; HR=1.99; 95% CI: 0.30-12.99; results not shown).
Lung cancer patients who quit using tobacco had significantly better overall survival than patients who continued to use tobacco after diagnosis. This result persisted even with control for age at diagnosis, pack-year history, days between diagnosis and last contact, sex, clinical stage, performance status, tumor histology, and having quit at referral. This study provides evidence that lung cancer patients are interested in quitting tobacco. Though not a randomized trial, it provides limited, indirect evidence that quitting may improve survival.
The 2014 Surgeon General's Report concluded that the adverse effects of smoking on cancer patients are no longer in question and that the effects extend across virtually all cancer disease sites and for all treatment modalities.1 As a result, it is timely to explore methods for improving outcomes among cancer patients by targeting smoking cessation among those patients; one important component could be the provision of cessation support.6, 15 To our knowledge, this is one of the first studies to examine the association of tobacco cessation with survival among lung cancer patients who participated in a systematic tobacco use program combining automated referral with tobacco cessation services. Using a retrospective design, Warren et al. (2013) found that quitting smoking during the year prior to a cancer diagnosis was associated with improved survival among lung cancer patients compared to patients that continued to use tobacco.3 Among lung cancer patients treated between 1982 and 1998, current smokers had increased mortality compared to patients quitting within a year of their diagnosis (HR=1.38; 95% CI: 1.06-1.79) and compared to former smokers (HR=1.43; 95% CI: 1.13-1.81). Both the present study and previous studies provide evidence that continued smoking following a lung cancer diagnosis appears to reduce survival, suggesting the importance of providing cessation support.3, 11, 16
The majority of published studies reporting on the association of smoking with cancer outcomes depend upon retrospectively collected tobacco use status, using notes in medical records; these can contain very limited smoking information.3, 6, 15, 17 Prior to the more recent reliance on electronic records and implementation of “meaningful use”, tobacco use was rarely collected in a standardized, prospective manner to support clinical outcomes research among cancer patients.3 A strength of this study was the collection of tobacco use information at diagnosis and during follow-up contacts. An additional strength of the present study is the potential reduction in self-selection bias for participation in the tobacco cessation service through the use of an opt-out requirement. Only 1.3% of smokers contacted by phone actively refused participation in the initial call with a similar low rate of refusal (~5%) for follow-up calls. Our findings suggest that standardized tobacco use assessments and automatic referrals to cessation counselors assure universal access to cessation support services without reliance on actual clinician referrals.
In considering our results, several limitations must be noted. First is that, although this study was prospective, it was not a clinical trial. There is no control group; this study utilized a cohort generated from a standard-of-care program established to reach all patients. The goal of this study was less to document that the tobacco cessation service improved patient survival than to utilize a natural experiment: a prospective cohort of lung cancer patients who participated in this clinical service, to explore the relationship between cessation and survival.
There is a potential for survival bias, given that patients who survived longer have more opportunities to quit and to relapse. We explored this by examining the length of time (in days) between diagnosis and the last contact; no significant associations with having quit were noted. In each model studied, even controlling for disease severity and time since diagnosis, having quit at last contact was a significant predictor of subsequent survival, again suggesting that quitting after diagnosis may improve survival for lung cancer patients. Guarantee time bias resulting in spurious significant findings is limited in this analysis; the study is restricted to patients who participated in at least one telephone call, therefore removing patients who were not called or did not participate in the service and could not have their tobacco use status assessed. Inclusion of a time-dependent variable for having quit at last contact suggested that there is no interaction between time and the quit status variable, that the Cox proportional hazards assumption was upheld, and is therefore an appropriate model (HRinteraction=0.978; 95% CI: 0.949-1.008).
Having tobacco use status based upon self-report and the imperfect accuracy of patient self-report is a study limitation.18 Adding biochemical confirmation of tobacco use status among patients participating in the cessation service is a goal for future studies.
Another limitation of this study is the limited sample size and loss to follow-up. At the beginning of the service priority was placed on making at least one telephone call to each new patient referred from the thoracic clinic as a current user, which reduced the number of patients with follow-up calls. Patients were contacted for a follow-up call only after attempts to contact new referrals were completed in a week. The goal of this approach was to provide at least some cessation support to as many patients as possible. Calling patients for the first and follow-up contacts was done randomly, as supported by the lack of statistically significant differences in patient demographic and disease characteristics. Current users at referral were significantly more likely than former users to be called for a first contact. The service was designed with current users as first priority; as a result, 83.3% of current users, but only 71.9% former users were called for the first contact (p=0.017).
Similarly, the limited follow-up data for patients beyond the first call also limited the tobacco use status information available after the first call. Patients with only one contact by the cessation service may have quit afterwards, but this information is not available. This limitation represents one of the challenges associated with longitudinal study design. We explored multiple ways of representing having quit while exploring this issue, and quitters consistently survived significantly longer than current users. Adding the number of successful counseling support calls in the model did not change the results. Similarly, restricting the final model to patients who participated in at least two calls led to similar results (n=126; HR=2.19; 95% CI: 1.14-4.21).
Additional research, including data derived from larger cohorts and from cessation trials, is needed to justify the provision of tobacco cessation services as part of comprehensive care for cancer patients. As we accrue data for additional patients and extend follow-up, we plan to explore other outcomes including disease-free survival and quality of life, and to extend these analyses to other cancer sites.
The present results are unique in demonstrating that lung cancer patients who reported quitting smoking at or following their diagnosis had significantly better survival than that of patients who continued to use tobacco after diagnosis. This represents a possibly important message for both clinicians and patients regarding the importance of smoking cessation.
Conflicts of Interest and Source of Funding: Dr. Cummings has a grant from Pfizer examining an inpatient hospital tobacco cessation service. In addition, Drs. Cummings and Mahoney both serve as expert witnesses for the plaintiffs in tobacco litigation cases. This work was supported in part by the Roswell Park Alliance Foundation, NCI R25CA113951, and the American Cancer Society (GWW, MRSG 11-031-01-CCE).