|Home | About | Journals | Submit | Contact Us | Français|
Continued smoking after cancer diagnosis may adversely affect treatment effectiveness, subsequent cancer risk, and survival. The prevalence of continued smoking following cancer diagnosis is understudied.
In the multi-regional Cancer Care Outcomes Research and Surveillance cohort (lung cancer [N=2456], colorectal cancer [N=3063]), we examined smoking rates at diagnosis and 5 months following diagnosis and factors associated with continued smoking.
90.2% of lung and 54.8% of colorectal cancer patients reported ever smoking. At diagnosis, 38.7% of lung cancer and 13.7% of colorectal cancer patients were smoking; 14.2% of lung cancer and 9.0% of colorectal cancer patients were smoking 5 months post-diagnosis. Factors associated with continued smoking among non-metastatic lung cancer patients were: Medicare, other public/unspecified insurance, not having chemotherapy, not having surgery, prior cardiovascular disease, lower body mass index, lower emotional support, and higher ever daily smoking rates (all p<.05). Factors independently associated with continued smoking among non-metastatic colorectal cancer patients were male sex, high school education, being uninsured, not having surgery, and higher ever daily smoking rates (all p<.05).
Following diagnosis, a substantial minority of lung and colorectal cancer patients continue smoking. Lung cancer patients had higher rates of smoking at diagnosis and following diagnosis; colorectal cancer patients were less likely to quit smoking following diagnosis. Factors associated with continued smoking differed between the two groups. Future smoking cessation efforts should examine differences by cancer type, particularly when comparing cancers for which smoking is a well established risk factor versus cancers for which it is not.
Smoking accounts for at least 30% of all cancer deaths.1 Lung cancer is the leading cause of cancer death for both men and women, with 219,000 new cases per year;1 colorectal cancer is the third most common cancer in both men and women, with over 146,000 new cases per year.1 There is a clear association between tobacco use and lung cancer incidence; tobacco use is responsible for 87% of lung cancers. Studies suggest that smoking is associated with formation and recurrence of colorectal adenomatous polyps2 and with increased colorectal cancer incidence and mortality,3–6 although the smoking-attributable risk for colorectal cancer has not been definitively established.1
Recent estimates of smoking at lung cancer diagnosis are 20–30%,7–9 and smoking at colorectal cancer diagnosis is estimated to be 10–20%.2, 4, 10–12 Research suggests that quitting smoking upon lung cancer diagnosis can improve chances for treatment efficacy, reduces chances of secondary tumors, and may double chances of survival.13–18 To our knowledge, there have been no studies examining the effects of continued smoking on treatment efficacy in colorectal cancer patients. Although lung cancer survival rates have historically been quite low, this is likely to change if current and former smokers increasingly undergo computed tomography screening, which preliminary results of the National Lung Screening Trial19 suggest can decrease lung cancer mortality. As lung cancer patients are diagnosed at earlier, more curable stages, tobacco treatment will be even more relevant and critical. The 5-year relative survival rate for colorectal cancer is 66%;20 thus there is a significant majority of cancer survivors who will need tobacco treatment to prevent future smoking related illness and new cancers. At diagnosis, the vast majority of cancer patients report that they want to quit smoking;21–23 estimates of quitting range from 35–79%;22–25 about half of those who quit relapse within one year.13
Previous work examining smoking rates among cancer patients, and associated factors, have been limited by single site cross-sectional studies with small samples or studies conducted within the context of a clinical trial. Population-based estimates of smoking at lung cancer and colorectal cancer diagnosis are needed to understand the extent and nature of the problem. In addition, it is uncertain which factors are associated with continued smoking and thus should be targeted for identification and intervention.26 In this paper we examine factors associated with continued smoking among early stage patients, as stakes differ between these and advanced stage patients and therefore clinicians might manage smoking differently. This is the first large scale, population-based study to examine the smoking behaviors of lung and colorectal cancer patients at diagnosis and follow-up. Our objective was to assess rates of continued smoking following diagnosis and associated factors among patients with a cancer for which smoking is a strong risk factor versus cancers for which the smoking-attributable risk is unclear.
This study is part of a large multi-regional prospective study examining processes and outcomes of care for a population-based and health system-based cohort of cancer patients conducted by the Cancer Care Outcomes Research and Surveillance (CanCORS) Consortium.27 The cohort includes more than 10,000 patients diagnosed with lung or colorectal cancer during 2003–2005. Details on study design and procedures have been published previously.28, 29 The study was approved by the human subjects committees at all participating institutions.
Patients aged 21 and older diagnosed with lung or colorectal cancer were identified within weeks of their diagnosis and surveyed by telephone approximately 5 months after diagnosis. This analysis included only CanCORS participants who personally completed a full patient survey at baseline (N=5519) (versus having the baseline survey completed by a surrogate if they were too ill or deceased; those survey versions did not include detailed questions about smoking status). We further limited analyses to the 5388 patients for whom information on smoking status was available.
Surveys were conducted using computer assisted telephone interviews in English, Spanish and Chinese. The response rate30 was 51% and the cooperation rate was 60%; comparisons of responders and nonresponders have been described previously.31 Participants in CanCORS have been shown to be demographically similar to population-based samples with these cancers in Surveillance, Epidemiology, and End Results registries.32 Information about cancer site, histology, and stage at diagnosis was obtained from registry data and medical records.
Smoking around the time of diagnosis was determined based on three questions from the baseline survey. Participants were considered current smokers around the time of diagnosis if they answered “yes” to the baseline questionnaire question “Do you smoke cigarettes regularly now?” or if they answered “yes” to the question “Have you ever smoked cigarettes regularly?”, “no” to “Do you smoke cigarettes regularly now?” and answered the question “How old were you the last time you were smoking regularly?” with an age that was within 1 year of their age at diagnosis. Participants were considered former smokers around the time of diagnosis if they answered “yes” to the question “Have you ever smoked cigarettes regularly?”, “no” to “Do you smoke cigarettes regularly now?” and answered the question “How old were you the last time you were smoking regularly?” with an age that was > 1 year older than their age at diagnosis. Participants were considered current smokers at 5 months post-diagnosis if they answered “yes” to the question, “Do you smoke cigarettes regularly now?”
Sociodemographic variables used in the current study included sex, race, education, marital status, self-reported insurance, and age.
Cancer stage was dichotomized as stage IV vs. stages I–III. Cancer treatment included surgery, chemotherapy and radiation treatments based on patients’ reports. Cancer symptoms were assessed using the European Organization for Treatment Research of Cancer Quality of Life Questionnaire (EORTC QLQ)33 for lung cancer and colorectal cancer, using a 4-point Likert scale (“not at all” to “quite a bit”). The items are then converted to a 1–100 scale in the scoring procedure.
Patients reported on their history of cardiovascular disease, lung disease, and diabetes. We examined pain by using the Brief Pain Inventory34 worst pain item that asks patients to rate their worst pain in the past 4 weeks on a scale from 0–10 (0=no pain, 10=worst pain you can imagine). Finally, we calculated body mass index from participants’ self-reported height and weight.
Depression was measured using an 8-item version of the Center for Epidemiological Studies Depression Scale (CES-D) assessing the presence or absence of symptoms over the past week.35 Fatalism was assessed using a 4-item version of the Powe Fatalism Inventory36 in which patients rated their views about health and life on a 4-point Likert scale. Health status was reported using a single item from the Short Form Health Survey as excellent, very good, good, fair, or poor.37 Emotional support was measured using the 5-item Emotional/Informational Support scale from the Medical Outcomes Study Social Support Survey38 in which items are rated on a 5 point Likert scale (1=all of the time; 5=none of the time). Physician communication was measured using 5 items from the Hospital Consumer Assessment of Healthcare Providers and Systems39 on a 4-point Likert scale (1=always, 4=never).
Reported frequency of alcohol use in the year before diagnosis was categorized into heavy drinkers (≥4 days/week), social drinkers (1–3 days/week) and light/never drinkers (≤1 day/month). Finally, participants were asked their highest number of cigarettes smoked per day.
Item non-response was infrequent; we used multiple imputation to impute missing data for items other than the smoking items when conducting inferential statistics.40, 41 All data analyses were conducted using Predictive Analytic Software (formally Statistical Package for the Social Sciences; SPSS) version 18.0. Using chi-square tests and one-way ANOVAs we compared the sociodemographics, medical histories, smoking histories, and psychosocial and health behavior variables of lung and colorectal cancer patients. Two-sided p- values of <.05 were considered statistically significant.
We then selected early stage (cancer stages I–III; those with unknown stage were not included in this analysis) participants who were smoking around the time of cancer diagnosis and assessed factors associated with continued smoking by conducting a series of univariate logistic regression equations using demographic, medical history, psychosocial, and health behavior variables. To qualify for each multivariable model, variables must have been associated with continued smoking in the univariate regressions (p <.25).42 All univariate and multivariate regression analyses controlled for study site.
Lung cancer patients were more likely than colorectal cancer patients to be female, white, older, less educated, and publicly insured. Lung cancer patients were also more likely to have a history of cardiovascular and lung disease, have a more advanced cancer, and less likely to have had surgery. Lung cancer patients had higher ratings of depression, poorer perceived health, lower ratings of perceived emotional support, and higher numbers of cigarettes smoked per day than colorectal cancer patients.
Overall, 90.2% of lung cancer and 54.8% of colorectal cancer patients had a history of ever smoking. Thirty percent of lung cancer and 14% of colorectal cancer patients reported to quit within 2 years prior to diagnosis (Figure 2). Rates of smoking around the time of diagnosis were 38.7% for lung cancer and 13.7% for colorectal cancer (p<.001) (Figure 3). By 5 months post-diagnosis, 14.2% of lung cancer and 9.0% of colorectal cancer (p<.001) patients remained smokers. Therefore, 63.0% of lung cancer patients versus 34.3% of colorectal cancer (p<.001) patients who had been smoking around the time of diagnosis had quit by 5 months following diagnosis. With respect to smoking rates by cancer stage, rates of smoking at the time of diagnosis were 37.4% for early stage lung cancer, and 41.5% for stage IV lung cancer (p=.07). For colorectal cancer patients, smoking rates at the time of diagnosis were 12.2% for early stage and 18.4% for advanced stage patients (p=.13). Smoking rates at 5 months post-diagnosis in lung cancer patients were 11.3% for early stage and 17.5% for advanced stage patients (p<.001). Smoking rates at 5 months post-diagnosis for colorectal cancer patients were 8.9% for early stage and 9.7% for advanced stage patients (p=.60).
Factors associated with continued smoking for lung cancer were being unmarried, having Medicare, other public, or unspecified insurance (vs. private), not having surgery, having more lung cancer symptoms, having a history of cardiovascular disease or lung disease, reporting worse pain in the past month, having a lower body mass index, reporting higher levels of depression, poorer perceived health, less emotional support, and having a higher reported highest number of cigarettes smoked per day (all p<.05). Significant factors associated with continued smoking for colorectal cancer patients were white race/ethnicity, older age, not having surgery, not having chemotherapy, reporting higher levels of depression, and higher rate of highest number of cigarettes per day (all p<.05).
Multivariable regression models are presented in Table 3. Factors associated with continued smoking among lung cancer patients were having Medicare, other public or unspecified insurance (vs. private) insurance, not having surgery, not having chemotherapy, a history of cardiovascular disease, a lower body mass index, lower levels of emotional support, and greater number of highest ever cigarettes smoked per day. Factors associated with continued smoking among colorectal cancer patients were male sex, having a high school education (vs. less than high school), not having health insurance (versus private), not having surgery, and greater number of highest ever cigarettes smoked per day.
Since the smoking rates and behaviors of cancer patients are not well understood, we used a population-based cohort to assess the prevalence of smoking at lung and colorectal cancer diagnosis, quitting behavior around the time of diagnosis, and factors associated with continued smoking post-diagnosis.
Almost all lung cancer patients had a history of ever smoking, and, although considerably lower, more than half of colorectal cancer patients reported ever smoking (higher than the general population rate of 42%).43 Rates of smoking at diagnosis were significantly higher for lung cancer versus colorectal cancer patients; approximately 1 in 3 lung cancer patients reported smoking at diagnosis, compared with 1 in 7 colorectal cancer patients. By around 5 months post-diagnosis, smoking rates had dropped to approximately 1 in 7 for lung cancer patients and 1 in 11 for colorectal cancer patients. While cancer rates did not differ significantly by stage in colorectal cancer patients or in lung cancer patients at the time of diagnosis, advanced stage lung cancer patients had higher rates of smoking following cancer diagnosis. This could reflect a sense from patients or providers that quitting smoking during advanced disease would not affect prognosis.
Our findings show substantial quit activity prior to, during, and immediately following a cancer diagnosis. Lung cancer patients were more likely than colorectal cancer patients to quit around the time of diagnosis and following their diagnosis. Due to the widespread public knowledge that smoking causes lung cancer, it is plausible that lung cancer patients associate their diagnosis with smoking and quit accordingly, while colorectal cancer patients may not associate their diagnosis with smoking and thus may be less likely to change their smoking behavior as a results of their diagnosis. Another possibility is that oncology providers are more likely to talk to lung cancer patients about quitting. The few studies that have been conducted on physician communication show that there is a great need to educate cancer patients about the risks of continued smoking after cancer diagnosis as well as offer assistance for cancer patients to quit.44–47 Our results suggests that future smoking cessation efforts should examine differences by cancer type, particularly when comparing cancers for which smoking is a well established risk factor versus a cancer for which it is not.
Many lung and colorectal cancer patients quit around the time of diagnosis, resulting in a significant group of cancer patients who are relatively new former smokers vulnerable to relapse. The 2009 American Society for Clinical Oncology’s Quality Oncology Practice Initiative recommends that cigarette smoking status be documented by the second medical visit and tobacco treatment counseling be recommended to patients receiving cancer treatment.48 Our findings emphasize the importance of following these recommendations and starting smoking cessation and relapse prevention conversations when a cancer patient presents with a suspected malignancy, regardless of whether or not it is an established smoking-related cancer. This is a critical time to address tobacco treatment, as research has shown that the closer to the time of diagnosis that smoking cessation treatment is delivered, the higher the likelihood for continued abstinence post-treatment.7, 22, 24, 49, 50
Lung and colorectal cancer patients had different baseline profiles as well as many different factors associated with continued smoking following diagnosis. Smoking cessation treatment programs should have disease-specific targeted components to address the variations in risk subgroups and modifiable factors. Smoking cessation interventions targeted for lung cancer patients should be designed to enhance support.
There were two factors that were significantly associated with continued smoking in both groups: not having surgical treatment and highest ever number of cigarettes smoked per day. Because of the negative effects of smoking on wound healing, many surgeons insist on abstinence from cigarettes before surgery. This type of aggressive intervention around smoking during other cancer treatments could reduce treatment complications and boost efficacy. 7, 13–17, 51, 52 Highest number of cigarettes smoked per day could be a proxy for addiction and underscores the need for pharmacological support to aid cancer patients to quit smoking as recommended by the 2008 U.S. Public Health Service Treating Tobacco Use and Dependence Clinical Practice Guideline.53 However, little progress has been made to adapt these guideline recommendations to the needs of cancer patients or implement these guidelines in the cancer care setting.9
Our findings are strengthened by the large, geographically diverse sample of cancer patients. Nevertheless, our findings must be viewed in consideration of several limitations. First, smoking status was assessed by self-report and is thus a likely underestimate of the smoking prevalence in this population. Recent analyses conducted by Dr. Park suggest that lung cancer patients may underreport their smoking rate; without biochemical validation of self report, there is the possibility of deception.9 Second, the patient survey did not specifically ask if patients were smoking at the time of diagnosis, so we determined smoking status based on estimates of timing of quitting. Third, patients whose surrogates completed the baseline survey were not included because smoking questions were not asked of surrogates. These patients overall had more advanced stage, and possibly had higher smoking rates. This resulted in lung cancer patients who are not necessarily geographically-divers of the population of lung cancer patients who smoke in that there was a high proportion of earlier stage patients. Finally, the small number of colorectal cancer patients who were smokers reduced our statistical power to detect factors that might be associated with continued smoking after colorectal cancer diagnosis.
We thank Ms. Hannah Pajolek, B.A. for her assistance in preparing this manuscript.
Funding: The work of the CanCORS Consortium was supported by grants from the NCI to the Statistical Coordinating Center (U01 CA093344) and the NCI supported Primary Data Collection and Research Centers (U01 CA093332, U01 CA093324, U01 CA093348, U01 CA093329, U01 CA093339, U01 CA093326, CRS 02-164) and the ACS (Park MSRG 005–05-CPPB).
Financial Disclosures: Dr. Rigotti has consulted for Pfizer, Free & Clear and served as site PI for research grants from Pfizer and Nabi Biopharmaceuticals. No other authors have any disclosures.