Longitudinal study of smoking patterns in relation to the development of smoking-related secondary primary tumors in patients with upper aerodigestive tract malignancies
The authors set out to assess the correlation between smoking-related second primary tumor (SPT) development and cigarette smoking habits after diagnosis and definitive treatment in patients with early-stage head and neck squamous cell carcinoma who were enrolled in a placebo-controlled randomized chemoprevention trial of 13-cis-retinoic acid.
Longitudinal data collected for 10 years after the index diagnosis are presented for 1190 patients. Cox proportional hazards regression models were used to examine the effects of changes in smoking behavior on smoking-related SPT development.
One-third of all patients who quit smoking within 12 months before randomization experienced recurrence, compared with 6.9% and 10.4% of all never-smokers and former smokers, respectively. Approximately 16% of all current smokers stopped smoking, and nearly 22% of current smokers developed SPTs, compared with 14.5%, 13.2%, and 8.8% of all recent smokers, former smokers, and never-smokers, respectively. The probability of developing a smoking-related SPT was highest among patients who were current smokers at randomization. These patients, regardless of whether they ceased smoking during follow-up, were nearly three times more likely than patients who had never smoked to develop a smoking-related SPT. In contrast, former smokers and recent quitters who continued to abstain from smoking during follow-up were approximately 1.5 times more likely to develop an SPT compared with patients who had never smoked.
Patients who continue to smoke after the successful treatment of their index head and neck malignancies have a substantially higher risk of developing smoking-related SPTs. Cancer 2004. © 2004 American Cancer Society.
Smoking is the main primary risk factor for the development of upper aerodigestive tract malignancies.1 The American Cancer Society estimates that 27,700 incident malignancies of the oral cavity and pharynx and 9500 laryngeal malignancies will be diagnosed in the United States in 2003.2 More than 80% of all head and neck malignancies diagnosed each year are associated with tobacco use.1, 3, 4
A history of smoking at the diagnosis of the index cancer is a universally recognized risk factor for second malignancy among patients with head and neck squamous cell carcinoma (HNSCC), with intense tobacco use having an especially notable cumulative effect on risk in this setting.5–9 The effects of the diagnosis and/or treatment of cancer on smoking cessation and abstinence, particularly at discrete time points postdiagnosis/posttreatment, have been examined by others.10–12 Nonetheless, most studies evaluating the effects of tobacco smoking and/or alcohol consumption on second primary tumor (SPT) risk have been cross-sectional investigations,13, 14 and to our knowledge, there are few studies in the literature regarding continuous patterns of smoking after diagnosis of the index HNSCC.
The current summary, based on longitudinal data collected for 10 years after patients' index diagnoses of HNSCC, deals with the cigarette smoking patterns of 1190 patients with early-stage HNSCC who had been enrolled in an unblinded, randomized, placebo-controlled multicenter trial of 13-cis-retinoic acid (13cRA) for the chemoprevention of SPTs. Data collected through January 21, 2003, were used to assess cigarette smoking patterns subsequent to index HNSCC detection and study enrollment. The current article describes the overall net changes in smoking status that were observed over this interval.
MATERIALS AND METHODS
Eligibility Criteria and Clinical Study Design
Participant accrual for the current study began in November 1991 and concluded in June 1999. Patients were registered through the Radiation Therapy Oncology Group (Philadelphia, PA), The University of Texas M. D. Anderson Cancer Center (Houston, TX), the Clinical Community Oncology Program (Houston, TX), the Southwest Oncology Group (San Antonio, TX), and Cancer and Leukemia Group B (Chicago, IL). The following were the primary patient inclusion criteria: 1) diagnosis of HNSCC of the upper aerodigestive tract (oral cavity, larynx, or pharynx); 2) age ≥ 18 years; 3) personal history of an American Joint Committee on Cancer staging system (1988) Stage I–II primary malignant lesion; 4) freedom from this primary malignancy for ≥ 16 weeks at the time of recruitment; 5) receipt of final treatment for this malignancy within the last 3 years; and 6) adequate liver performance, adequate renal functioning, and the presence of normal hematologic parameters. Among the excluded patients were those who had concurrent malignancies other than localized nonmelanomatous skin cancer, those who had multiple primary head and neck tumors, and those who had a history of malignancy at sites other than the head and neck. All inclusion and exclusion criteria have been described in detail by Hudmon et al.15 and Khuri et al.16
All patients signed a statement of informed consent. Patients were assigned with equal probability to receive treatment with either 13cRA or placebo. The treatment allocation process took disease stage (I or II), primary tumor site (oral cavity, pharynx, or larynx), and smoking status (never, former, recent, or current) into account as stratifying factors. Participation required patients to receive either 13cRA or placebo for a total of 3 years. Patients were evaluated at 3, 6, 9, 12, 16, 20, 24, 28, and 36 months after randomization, with additional evaluations performed at 6-month intervals thereafter for 4 years.
Tumor recurrences and SPT diagnoses were histologically confirmed by the centralized M. D. Anderson Endpoint Review Committee using the modified criteria established by Warren and Gates.17 An SPT was defined as a new malignancy whose histologic type differed from that of the index lesion, a malignancy that was identical to the index lesion in terms of histologic type but that arose > 3 years after the treatment of the index lesion, or a malignancy that was separated from the index lesion by > 2 cm of clinically normal epithelium. Patients who developed SPTs were required to discontinue use of the study medication and were withdrawn from the trial.
Patients provided data on sociodemographic factors, clinical characteristics, tobacco use, and alcohol consumption in a structured questionnaire administered at study entry and at each follow-up visit. Data on tobacco use included cigarette, pipe, and cigar smoking habits; chewing tobacco and snuff use habits; the average number of cigarettes smoked per day; and the ages at which the patient began and ceased smoking, if applicable. Patients were categorized as never-smokers (patients who smoked fewer than 100 cigarettes or 5 packs in their lifetime), former smokers (patients who quit smoking > 12 months before randomization), recent quitters (patients who quit smoking < 12 months before randomization), and current smokers (patients who continued to smoke at the time of randomization).
The current analysis focused only on smoking-related SPTs, which were defined as those found in the larynx, oral cavity, pharynx, lung, esophagus, other upper digestive tract site, or bladder.
Missing cigarette counts
Before the analysis was performed, a conservative method was used to replace missing cigarette counts from patients' postrandomization study visits. For each patient with missing counts, the smoking cessation date and the cigarette counts that were available were reviewed. If this review yielded no indication of a history of smoking, then all missing counts were programmatically replaced with zeroes; otherwise, missing counts were not replaced with any definite value.
Smoking patterns in relation to index diagnosis and randomization
Using all available cigarette counts from patients' follow-up visits, patterns of change in smoking status were examined according to smoking status at randomization. For each patient, follow-up duration (in months) was calculated from the date of the index diagnosis. Next, the average number of cigarettes smoked per day was plotted against time since the index diagnosis for each patient, with the results stratified according to smoking status at randomization and also according to treatment type. Follow-up data were reviewed on a patient-by-patient basis to assess patterns of change in smoking habits; each patient was classified as either having experienced a change in smoking behavior or not having experienced a change in smoking behavior. For never-smokers, former smokers, and recent quitters (as classified at randomization), any nonzero cigarette count at a follow-up visit was considered indicative of a change in smoking behavior. In contrast, for those categorized as current smokers at randomization, smoking behavior was considered unchanged unless the patient ceased smoking and remained abstinent such that all cigarette counts subsequent to a given follow-up visit were zero.
Time to SPT was defined as the number of years between randomization and SPT diagnosis. All SPT-free patients were censored, and time to SPT was calculated as the time (in years) from randomization to most recent follow-up. Kaplan–Meier estimates of smoking-related SPT–free survival were computed according to smoking status at randomization using a Cox proportional hazards regression model18 adjusted for time from index diagnosis to randomization and also for age at diagnosis. Additional Cox models, including one in which cigarette counts at follow-up were treated as time-dependent covariates, were used to evaluate the effects of changes in smoking behavior on SPT risk. Only patients with at least two follow-up visits subsequent to randomization were included in the Cox model involving time-dependent covariates.
All reported P values are two sided. Analyses were performed using SAS (SAS Institute, Cary, NC)19 and S-PLUS software (Mathsoft Inc., Seattle, WA).20
For the 1190 patients enrolled in the current study, the median time from index diagnosis to randomization was 1.0 years (range, 0.4–4.5 years). The median follow-up duration (from the time of randomization) was 4.9 years (range, 0–9.9 years), with 75% of all patients having been followed for ≥ 3.1 years. These data correspond to a median follow-up duration of 6.1 years from the time of the index diagnosis (range, 0.9–11.6 years). Median follow-up was slightly longer for never-smokers and former smokers (5.0 years for each group) compared with recent quitters (4.6 years) and current smokers (4.5 years). The median follow-up durations in the 13cRA and placebo arms were 4.9 and 5.0 years, respectively, and the median follow-up duration for the 173 patients diagnosed with smoking-related SPTs was 4.5 years (range, 0.1–9.9 years).
SPTs were diagnosed in 248 of 1190 patients (20.8%). Although the study population was predominantly male (79.0%), gender was not associated with SPT development (P = 0.84). Patients diagnosed with SPTs were, however, significantly older (median age, 65 years) compared with SPT-free patients (median age, 61 years; P < 0.0001).
Seventy percent of the SPTs observed (173 of 248) were smoking-related lesions. Nearly 22% of all current smokers, compared with 8.8%, 13.2%, and 14.5% of all never-smokers, former smokers, and recent smokers, respectively, were diagnosed with smoking-related SPTs (P = 0.002) (Table 1). In addition, Kaplan–Meier estimates of SPT-free survival were lowest among patients who were current smokers at the time of randomization. Smoking-related SPT site distributions also differed significantly across the smoking status groups defined at randomization (P < 0.0001) (Table 2). Among former smokers, recent quitters, and current smokers, > 50% of all smoking-related SPTs occurred in the lung, whereas only 14.3% of all smoking-related SPTs found in never-smokers were lung lesions. In contrast, nearly 80% of all smoking-related SPTs diagnosed in never-smokers (11 of 14) were located in the oral cavity, compared with only 17.1%, 22.6%, and 15.4% of all smoking-related SPTs detected in former smokers, recent quitters, and current smokers, respectively.
Table 1. Kaplan–Meier Estimates of Smoking-Related SPT–Free Survival by Smoking Status at Randomization
|Never||160|| 14 (8.8)||0.98||(0.95, 1.00)||0.97||(0.94, 1.00)||0.97||(0.94, 1.00)||0.95||(0.91, 0.98)||0.93||(0.88, 0.97)||0.88||(0.82, 0.95)|
|Former||575|| 76 (13.2)||0.99||(0.98, 1.00)||0.97||(0.95, 0.98)||0.94||(0.92, 0.96)||0.91||(0.88, 0.93)||0.88||(0.85, 0.91)||0.86||(0.82, 0.89)|
|Recent||214|| 31 (14.5)||0.97||(0.95, 0.99)||0.94||(0.91, 0.97)||0.90||(0.85, 0.94)||0.86||(0.81, 0.92)||0.84||(0.78, 0.90)||0.83||(0.77, 0.89)|
|Current||241|| 52 (21.6)||0.96||(0.94, 0.99)||0.91||(0.87, 0.95)||0.87||(0.82, 0.91)||0.83||(0.78, 0.88)||0.77||(0.71, 0.84)||0.70||(0.63, 0.78)|
|All patients||1190||173 (14.5)||0.98||(0.97, 0.99)||0.95||(0.94, 0.96)||0.92||(0.91, 0.94)||0.89||(0.87, 0.91)||0.86||(0.83, 0.88)||0.82||(0.80, 0.85)|
Table 2. Site Distribution of Smoking-Related SPTs by Smoking Status at Randomizationa
|Larynx||0 (0.0)||12 (15.8)||3 (9.7)||5 (9.6)||20 (11.6)|
|Oral cavity||11 (78.6)||13 (17.1)||7 (22.6)||8 (15.4)||39 (22.5)|
|Pharynx||0 (0.0)||7 (9.2)||4 (12.9)||2 (3.8)||13 (7.5)|
|Lung||2 (14.3)||31 (40.8)||15 (48.4)||29 (55.8)||77 (44.5)|
|Esophagus||0 (0.0)||3 (3.9)||0 (0.0)||4 (7.7)||7 (4.0)|
|Other upper digestive tract site||1 (7.1)||1 (1.3)||0 (0.0)||0 (0.0)||2 (1.2)|
|Bladder||0 (0.0)||9 (11.8)||2 (6.5)||1 (1.9)||12 (6.9)|
|Other||0 (0.0)||0 (0.0)||0 (0.0)||3 (5.8)||3 (1.8)|
|Total||14 (100.0)||76 (100.0)||31 (100.0)||52 (100.0)||173 (100.0)|
The median number of follow-up visits per patient for which cigarette counts were available was 13. Current smokers at the time of randomization had a median of 11 follow-up visits for which cigarette counts were available, whereas never-smokers, former smokers, and recent quitters had available cigarette counts for a median of 14, 13, and 12 follow-up visits, respectively. Finally, cigarette counts were available for a median of 12 follow-up visits per patient in the 13cRA arm, compared with 13 visits per patient in the placebo arm.
One-third of all patients who quit smoking < 12 months before randomization subsequently presented with changes in smoking behavior (namely, resumption of smoking), compared with 6.9%, 10.4%, and 16.2% of all never-smokers, former smokers, and current smokers, respectively. The majority of never-smokers (90.9%) and former smokers (68.3%) in whom changes in smoking behavior were observed had only a single nonzero cigarette count during follow-up. In contrast, nearly half of all patients defined as recent quitters at the time of randomization (48.6%) had more than 3 nonzero cigarette counts during follow-up. More than 69% of all current smokers in whom changes in smoking habits were observed quit smoking and remained abstinent for ≥ 12 months.
Table 3 presents data on the relative risk of smoking-related SPT development according to smoking behavior during follow-up. Specifically, patients defined as current smokers at randomization were nearly 3 times more likely than never-smokers to develop a smoking-related SPT, regardless of whether they subsequently quit (relative risk [RR], 2.94; 95% confidence interval [CI], 1.30–6.66) or continued smoking (RR, 2.75; 95% CI, 1.49–5.07). In addition, former smokers and recent quitters who remained abstinent during follow-up were approximately 1.5 times more likely to develop an SPT (95% CI, 0.84–2.60) compared with never-smokers, although this difference was not statistically significant. Finally, never-smokers, former smokers, and recent quitters who began or resumed smoking during follow-up were 1.2 times more likely to develop a smoking-related SPT (95% CI, 0.59–2.46) compared with those who did not begin or resume smoking during follow-up.
Table 3. Smoking-Related SPT Risk by Smoking Behavior During Follow-Up
|Never smoked||149||14 (9.4)||1.00|| |
|Remained abstinentc||647||90 (13.9)||1.48||(0.84, 2.60)|
|Started smokingd||145||16 (11.0)||1.20||(0.59, 2.46)|
|Stopped smokinge||39||10 (25.6)||2.94||(1.30, 6.66)|
|Continued smokingf||195||40 (20.5)||2.75||(1.49, 5.07)|
Cox modeling revealed no significant predictive effect associated with time-dependent cigarette counts at follow-up (Table 4); however, smoking status at randomization was found to be predictive of smoking-related SPT risk. Specifically, current smokers were 3 times more likely than never-smokers to develop a smoking-related SPT (RR, 3.28; 95% CI, 1.69–6.38). The corresponding RR estimates for former smokers and recent quitters were 1.41 and 1.77, respectively, although neither estimate was statistically significant.
Table 4. Cox Proportional Hazards Model Assessing Time-Dependent Predictive Effect of Cigarette Counts during Follow-Up on Time to Smoking-Related SPT Development
|Cigarette count||−0.0026||1.00||(0.98, 1.02)||0.820|
|Smoking status at randomization|| || || || |
| Never|| ||1.00|| || |
| Former||0.3436||1.41||(0.79, 2.50)||0.240|
| Recent||0.5723||1.77||(0.93, 3.39)||0.084|
| Current||1.1891||3.28||(1.69, 6.38)||<0.001|
|Treatment type|| || || || |
| Placebo|| ||1.00|| || |
| 13cRA||−0.1835||0.83||(0.61, 1.13)||0.240|
To our knowledge, there have been relatively few prospective analyses documenting the association between continued smoking and SPT risk after the diagnosis and treatment of an index upper aerodigestive tract malignancy; most studies have examined smoking history before the index diagnosis, and such studies have demonstrated that prediagnosis cigarette smoking is a significant risk factor for subsequent SPT development.5, 6 Of the investigators who previously examined postdiagnosis smoking habits and their relation to SPT risk, Castigliano21 and Schottenfeld et al.22 were unable to demonstrate a decrease in risk among those who had quit smoking after their index diagnoses, whereas Moore8 and Silverman et al.23 found that quitters did experience a reduction in risk. In addition, Day et al.5 reported that former smokers had a significantly lower SPT risk compared with current smokers, although they did not find the cessation of smoking after the index diagnosis to be associated with a reduction in risk.
Compared with previous reports, the current study involves considerably more extensive follow-up (75% of all patients observed for ≥ 3.1 years and 25% observed for ∼6.6 years after randomization; median follow-up, 6.1 years since the index diagnosis) and allows more accurate evaluation of the long-term benefits of smoking cessation after the diagnosis of an index upper aerodigestive tract malignancy. (Patients in the study conducted by Day et al.5 had a median follow-up duration of 27 months.) Overall, 8.8% of never-smokers developed smoking-related SPTs, compared with 21.6% of current smokers. The corresponding figures for former smokers and recent quitters were similar to one another (13.2% and 14.5%, respectively). Also noteworthy was the finding that nearly 7% of all never-smokers (11 of 160) developed oral malignancies; of these 11 patients, 1 reported the consumption of chewing tobacco, 1 reported engaging in pipe smoking, and the remaining 9 did not report any pre-enrollment pipe or cigar smoking or consumption of snuff or chewing tobacco.
The probability of developing a smoking-related SPT was highest among patients classified as current smokers at randomization. These patients were nearly three times as likely as patients who had never smoked to develop a smoking-related SPT, regardless of whether they ceased smoking during follow-up. In contrast, former smokers and recent quitters who remained abstinent during follow-up were only approximately 1.5 times more likely to develop an SPT compared with never-smokers. Our findings indicate that cessation of smoking subsequent to the index diagnosis may be beneficial with regard to SPT risk reduction and that continued smoking is associated with a threefold increase in risk. Furthermore, data obtained in the current study suggest that recent quitters, who were likely to have quit as a result of the symptoms of their disease (or its treatment), have the highest risk of recidivism; more than 30% of all recent quitters reported having smoked at some point during follow-up.
One possible shortcoming of the current study is that results may have been affected by a recall bias with regard to reported cigarette consumption during follow-up; however, to minimize this type of bias, we opted to use a conservative analysis method. Another possibility is that patients whose smoking habits changed did not accurately report cigarette counts at follow-up, causing risk estimates to be biased downward. The current investigation also was limited by the scarcity of data regarding the use of pipes, cigars, chewing tobacco, and snuff; ultimately, such data were not incorporated into our analyses.
Using the largest and most unique HNSCC data set available to date, the current study assessed the net effects of various risk factors on time to smoking-related SPT development. The results of this prospective study indicate that a significant association exists between time to SPT development and continued cigarette smoking after the treatment of HNSCC. Thus, the clinical and prognostic impact of effective smoking cessation interventions for patients presenting with index malignancies is obvious.