|Home | About | Journals | Submit | Contact Us | Français|
Tobacco smoking is associated with oropharynx cancer survival, but to what extent cancer progression or death increases with increasing tobacco exposure is unknown.
Patients with oropharynx cancer enrolled onto a phase III trial of radiotherapy from 1991 to 1997 (Radiation Therapy Oncology Group [RTOG] 9003) or of chemoradiotherapy from 2002 to 2005 (RTOG 0129) were evaluated for tumor human papillomavirus status by a surrogate, p16 immunohistochemistry, and for tobacco exposure by a standardized questionnaire. Associations between tobacco exposure and overall survival (OS) and progression-free survival (PFS) were estimated by Cox proportional hazards models.
Prevalence of p16-positive cancer was 39.5% among patients in RTOG 9003 and 68.0% in RTOG 0129. Median pack-years of tobacco smoking were lower among p16-positive than p16-negative patients in both trials (RTOG 9003: 29 v 45.9 pack-years; P = .02; RTOG 0129: 10 v 40 pack-years; P < .001). After adjustment for p16 and other factors, risk of progression (PFS) or death (OS) increased by 1% per pack-year (for both, hazard ratio [HR], 1.01; 95% CI, 1.00 to 1.01; P = .002) or 2% per year of smoking (for both, HR, 1.02; 95% CI, 1.01 to 1.03; P < .001) in both trials. In RTOG 9003, risk of death doubled (HR, 2.19; 95% CI, 1.46 to 3.28) among those who smoked during radiotherapy after accounting for pack-years and other factors, and risk of second primary tumors increased by 1.5% per pack-year (HR, 1.015; 95% CI, 1.005 to 1.026).
Risk of oropharyngeal cancer progression and death increases directly as a function of tobacco exposure at diagnosis and during therapy and is independent of tumor p16 status and treatment.
Tobacco smoking remains the principal risk factor for head and neck squamous cell carcinoma (HNSCC) worldwide. In addition to its etiologic role, smoking status at diagnosis (never, former, current) is associated with treatment response,1 risk of second primary cancers,2–4 and survival.5 Smoking during radiotherapy is also associated with treatment response and disease control,6 albeit inconsistently.7 However, the magnitude by which a patient's risk of cancer progression or death is affected by both cumulative measures of lifetime tobacco exposure at diagnosis and smoking during treatment is unknown.
Patients with oropharyngeal carcinoma (OPC) provide an opportunity to measure the impact of tobacco exposure before, during, or after treatment on outcomes, including progression-free survival (PFS), overall survival (OS), and risk of second primary tumors. Although tobacco smoking does not appear to be a strong cofactor for development of the human papillomavirus (HPV) –positive subset of OPC,8,9 there is increasing evidence that it may nevertheless alter the behavior and treatment response of this cancer. In preliminary studies, patients who were HPV-positive and were smokers at diagnosis were reported to have worse survival than those who were not,10,11 possibly because of an increased risk for both local recurrence and distant metastases.12
In a recent analysis,13 we found that tumor HPV status and tobacco smoking (≤ 10 or > 10 pack-years) were the two strongest, independent determinants of PFS and OS for patients with OPC treated by chemoradiotherapy. In this study, we sought to further investigate the impact of quantitative measures of tobacco smoking on survival outcomes in two trials of the Radiation Therapy Oncology Group (RTOG): RTOG 9003 and RTOG 0129.
Patients eligible for RTOG 9003 (Fig 1) and RTOG 0129 (Fig 2) had untreated, pathologically confirmed, stage III to IV (M0)14 squamous cell carcinoma of the oral cavity, oropharynx, hypopharynx, or larynx and were age ≥ 18 years. Patients with T1-2N1 or T1N2-3 were eligible for RTOG 9003 but were excluded from RTOG 0129. Karnofsky performance status was ≥ 60 for RTOG 9003 and more than 60 for RTOG 0129.15
In both trials, patients were stratified by tumor site, nodal stage, and performance status as previously reported.13,16 Patients in RTOG 9003 were assigned to one of four radiotherapy regimens: standard fractionation, hyperfractionation, accelerated hyperfractionation with split, and accelerated fractionation with concomitant boost.16 Patients in RTOG 0129 were assigned to cisplatin concurrent with either standard fractionation (cisplatin 100 mg/m2 days 1, 22, and 43) or accelerated fractionation with concomitant boost (cisplatin 100 mg/m2 on days 1 and 22).13
Lifetime cigarette exposure was prospectively determined at enrollment by use of a standardized questionnaire administered by clinical research staff. Surveys for both trials obtained data on ever use, current use, age at start, average number of cigarettes smoked per day, total years of smoking (RTOG 9003 only), or age stopped smoking (RTOG 0129 only). Data on current cigarette smoking (yes, no) were prospectively collected for patients in RTOG 9003 at the first follow-up visit after completion of radiotherapy.
To assess tumor status, physical examination and imaging studies (if indicated) were performed every 3 months for 18 to 24 months, every 4 to 6 months through year 3, every 6 months through year 5, and then annually. Both trials were approved by the institutional review boards of participating sites. Patients provided written informed consent.
Formalin-fixed, paraffin-embedded tumor specimens were evaluated for tumor p16 expression, an established surrogate for tumor HPV status in OPC, by immunohistochemistry using a mouse monoclonal antibody (MTM Laboratories, Heidelberg, Germany) visualized with the Ventana XT autostainer using the one-view secondary detection kit (Ventana Medical Systems, Tucson, AZ).17 p16 expression was scored as positive if strong and diffuse nuclear and cytoplasmic staining was present in ≥ 70% of the tumor cells.17
Our analysis was restricted to patients with OPC who had p16 determination because tumor p16 status is strongly associated with both smoking and survival. This post hoc subset analysis was not part of the original study protocol.
Our primary outcome of interest was the effect of pack-years of tobacco exposure on OS, defined as time from random assignment to death due to any cause. PFS was defined as time from random assignment to death or first documented relapse, categorized as locoregional (primary site or regional nodes) failure (LRF) or distant metastases (DMs). Death from index cancer without documented site of recurrence was considered LRF. Second primary tumors (SPTs) were evaluated separately. PFS, LRF, and DM are reported here instead of protocol-specified secondary end points (eg, locoregional control) to be consistent with prior analyses of RTOG 0129.13 Follow-up was calculated as days to the date of an event or last known date alive. OS and PFS rates were estimated by using the Kaplan-Meier method18and compared by log-rank test.19 Karnofsky performance status was converted to Zubrod performance status to facilitate comparisons. The cumulative incidence method20 and Gray's test21 were used to estimate and compare rates of LRF, DM, and SPT. Cox proportional hazards models22 were used to estimate hazard ratios (HRs); multivariable models were developed by minimizing Akaike information criteria23 by using the method of Shtatland.24 Cox regression was performed for patients with OPC with determined HPV status and smoking data. To investigate potential bias in estimates due to missing data, analyses were repeated for patients with OPC by using smoking values imputed with the Markov Chain Monte Carlo algorithm with a noninformative prior (SAS/STAT software, SAS Online Doc 9.2; SAS Institute, Cary, NC). Twenty data sets were created and the resulting analyses were combined per Rubin's formula.25
Patients were enrolled in RTOG 9003 from 1991 to 1997. Sixty percent (646 of 1,068) of the eligible patients had a diagnosis of OPC, and 29.4% (n = 190) of the 646 patients had tumor specimens available for p16 determination. No significant differences in baseline characteristics or outcomes were observed between patients with and without p16 determination (Appendix Table A1, online only). For RTOG 0129, 60% (433 of 721) of eligible patients enrolled from 2002 to 2005 had a diagnosis of OPC, and 73% (n = 316) of the 433 patients had p16 determination. The characteristics of the resulting study populations from RTOG 9003 and RTOG 0129 are listed in Table 1. Data on pack-years were missing for 15 and 56 patients for RTOG 9003 and 0129, respectively. Data on smoking during radiotherapy were missing for 15 patients for RTOG 9003.
In RTOG 9003, median age at start of smoking was 17 years (interquartile range [IQR], 15 to 20 years), and median cigarettes smoked per day was 20 cigarettes (IQR, 0 to 25 cigarettes). A median of 38 pack-years (IQR, 13 to 60 pack-years) and median 32 years of smoking (IQR, 21 to 44 years of smoking) were reported by patients with OPC.
To examine the independent effect of smoking on outcomes for RTOG 9003, we must first account for the effect of an important confounder, p16 status, as previously reported for RTOG 0129.13 p16 expression was found in 39.5% (95% CI, 32.5% to 46.4%) of patients with OPC. Characteristics of the p16-positive and p16-negative patients are listed in Table 1. p16-positive patients were more likely to be never smokers and had significantly lower cigarette smoking exposure, as measured by pack-years (median, 29 v 45.9 pack-years; P = .02) of smoking and cumulative years of smoking (median, 28 v 36 years of smoking; P = .002).
The median follow-up among surviving patients in RTOG 9003 was 9.3 years (range, 0.3 to 13.2 years) at the data cut point (August 11, 2005). The 5-year OS for patients with OPC in RTOG 9003 was 31.0% (95% CI, 24.3% to 37.7%; Fig 3A). There were 49 deaths among p16-positive and 104 deaths among p16-negative patients. In Kaplan-Meier analysis, patients with p16-positive tumors had better OS (Fig 3B) and PFS than patients with p16-negative tumors (log-rank test P < .001 for both). The 5-year OS rates were 49.0% (95% CI, 37.5% to 60.6%) and 19.6% (95% CI, 12.2% to 26.9%), and PFS rates were 43.6% (95% CI, 32.2% to 55.0%) and 19.0% (95% CI, 11.8% to 26.2%), respectively. LRF was lower for p16-positive patients (28.9% v 54.9% at 5 years; P < .001) but DMs (11.1% v 13.0% at 5 years; P = .71) and SPTs (13.8% v 11.4% at 5 years; P = .40) were not.
In the study population, 30 patients experienced SPTs, of which 14 were among p16-positive patients (n = 75) and 16 were among p16-negative patients (n = 115), respectively, at the data cut point. The only factor found to be significantly associated with risk of SPT was smoking exposure at diagnosis. The hazard of SPT increased 1.5% per pack-year (HR, 1.015; 95% CI, 1.005 to 1.026) or 1.5% per year of smoking (HR, 1.015; 95% CI, 0.994 to 1.037).
Smoking exposure at diagnosis was also strongly associated with OS in RTOG 9003. The hazard of death was more than two-fold higher among individuals with more than 10 versus ≤ 10 pack-years of tobacco smoking at diagnosis (HR, 2.10; 95% CI, 1.35 to 3.25; log-rank P < .001; Fig 3C). This was equivalent to an absolute difference in 5-year OS of 30% (95% CI, 6.9% to 53.1%) between the two smoking exposure groups.
Smoking exposures remained important predictors of survival, even after accounting for the strong effects of tumor p16 status and other important prognostic factors (Zubrod performance status, T stage, and N stage; Tables 2 and and3).3). When evaluated as a continuous variable, the hazard of death increased by approximately 1.0% per pack-year and by approximately 2% per year of smoking. The increased hazard of progression per pack-year and per year of smoking was quite similar to that for death (Table 3).
Smoking exposure was also an independent predictor of LRF in RTOG 9003. LRF was reported for 75 patients and was more common among individuals with more than 10 versus ≤ 10 pack-years of tobacco smoking at diagnosis (64 of 135 events v 11 of 40 events; 5-year LRF: 48.3% v 25.6%; Gray's P = .01). Even after adjustment for p16 status, performance status, and T stage, LRF was more common among individuals with more than 10 versus ≤ 10 pack-years of tobacco smoking (HR, 2.14; 95% CI, 1.09 to 4.18; P = .03). Age and treatment assignment were neither important predictors nor confounders in these multivariable analyses.
In RTOG 9003, current smoking status (yes, no) was available for the period of radiotherapy (accessed a median of 32 days [range, 0 to 105 days] after the end of radiotherapy). Smoking during radiotherapy significantly increased the hazard of death (Fig 3D), even after adjustment for pack-years and other factors (HR, 2.19; 95% CI, 1.46 to 3.28) or after adjustment for years of smoking and other factors (HR, 1.87; 95% CI, 1.23 to 2.84). Smoking during radiation similarly increased the hazard of progression (Table 3). No differences in rates of severe mucositis (grade ≥ 3) or radiotherapy treatment breaks (≥ 5 days) were observed in individuals who did or did not smoke during radiotherapy (data not shown).
To enhance our understanding of the effect of smoking on survival outcomes for patients with OPC, we examined the effect of several measures of tobacco exposure on survival outcomes in RTOG 0129. Medians of 20 pack-years (IQR, 0 to 40.5 pack-years) and 25 years of smoking (IQR, 0 to 35 years of smoking) were reported by patients with OPC who had p16 determination in that trial.
The proportion of patients with OPC in RTOG 0129 with p16-positive tumors was 68.0% (95% CI, 62.9% to 73.2%). p16-positive patients were more likely than p16-negative patients to be never smokers and had significantly lower cigarette smoking exposure, as measured by pack-years and cumulative years of smoking (Table 1).
The median follow-up among surviving patients in RTOG 0129 was 4.9 years (range, 1.6 to 6.4 years), and 5-year OS for patients with OPC was 66.8% (95% CI, 61.4% to 72.1%; Fig 4A). OS was significantly worse (HR, 2.81; 95% CI, 1.72 to 4.58; log-rank P < .001) for patients with OPC with more than 10 versus ≤ 10 pack-years of tobacco smoking (Fig 4B). This remained the case even after adjustment for other factors (Table 3).
The adjusted hazard of death or progression associated with several common measures of cumulative tobacco exposure for patients enrolled onto RTOG 0129 is provided in Table 3. As was observed for RTOG 9003, the hazard of death (OS) and progression (PFS) increased by approximately 1.0% per pack-year and by approximately 2% per year of smoking. Similarly, in RTOG 0129, LRF was more common among p16-negative than p16-positive patients (30 of 215 v 38 of 101 events; 5-year LRF, 38.6% v 14.3%; Gray's P < .001) and for those with more than 10 versus ≤ 10 pack-years of tobacco smoking at diagnosis (43 of 150 events v 13 of 110 events; 5-year LRF, 29.3% v 11.9%; Gray's P = .001). Even after adjustment for p16 status, performance status, and T stage, LRF was more common among individuals with more than 10 versus ≤ 10 pack-years of tobacco smoking (HR, 1.98; 95% CI, 1.03 to 3.80; P = .04).
In our prior analysis of RTOG 0129,13 we demonstrated that tumor HPV status and tobacco exposure (≤ 10 or >10 pack-years) were the strongest determinants of survival for patients with OPC. Here we demonstrate that risk of cancer progression or death and SPTs increased as a direct function of quantitative measures of tobacco exposure at diagnosis and that the effect strength was independent of treatment by radiotherapy or chemoradiotherapy. Thus, significant changes in both the HPV-attributable proportion and tobacco exposure may, taken together, contribute to improvements in absolute survival over calendar time for patients with OPC. Furthermore, smoking during radiotherapy may further compromise treatment outcome.
The increased prevalence of p16-positive patients and the decline in tobacco exposure we observed when comparing the study population for RTOG 9003 with that for RTOG 0129 are consistent with increases in incidence for HPV-positive OPC26,27 and declines in smoking prevalence at the population level in the United States.28 Although the per pack-year increase in risk of progression or death was the same regardless of p16 status, declines in tobacco exposure were more marked for the p16-positive group, likely increasing their relative survival benefit over calendar time.
Smoking is known to increase all-cause29 and cancer-specific mortality,30 and therefore our findings are not unexpected. For patients with early stage HNSCC, risk of death has been associated with smoking status at diagnosis2,4and increased with increasing categories of exposure to tobacco as measured in pack-years or years of smoking.31 Because tumor HPV status is strongly associated with both smoking status and survival, it is important to examine the effect of smoking on survival after accounting for the effect of tumor HPV status. Our data indicated that risk of cancer progression or death increased directly as a function of pack-years or total number of years of smoking, even after accounting for HPV status. Because deaths unrelated to cancer and from unknown cause are included as events in analyses of OS and PFS, competing causes of mortality reasonably expected to be more pronounced among heavy versus light or nonsmokers may account for associations between smoking and OS and PFS. However, the increased hazard of LRF we observed in association with smoking suggests a possible direct effect on treatment response and/or disease control. Interpreting our data from a molecular perspective, the probability that an OPC will acquire genetic hits imparting resistance to DNA-damage-induced cell death increases directly with smoking exposure. Further study is clearly warranted before incorporating measures of smoking exposure into treatment decision making.
Browman et al6 originally reported that smoking during radiotherapy reduced response rates and 2-year survival for patients with head and neck cancer. However, in a subsequent report, smoking during radiotherapy was not an independent predictor of survival after accounting for prior tobacco use.7 Chen et al32 recently reported reduced 5-year rates of locoregional control, disease-free survival, and OS among patients with head and neck cancer who continued to smoke after diagnosis (“active smokers”) who were matched to smokers who had quit. However, differences in baseline tobacco exposure could not be excluded as the explanation for these findings because median pack-years among the active smokers was twice that of the comparison group (40 v 20 pack-years). Although we accounted for prior exposure, we acknowledge that the excess mortality we observed in association with smoking during radiation therapy may not be independent of continued smoking beyond radiotherapy.31
There are several possible explanations for why smoking during radiotherapy might reduce effectiveness of therapy. Smoking during radiation therapy has been reported to increase the severity of mucositis, thus increasing the frequency of treatment breaks for smokers, which are known to decrease disease control (although not observed in this study).32 Tissue hypoxia, commonly observed in head and neck cancers, is known to be associated with reduced survival,33 and hypoxic modification strategies have shown some benefit with regard to locoregional control.34 Supporting evidence that smoking exacerbates tissue hypoxia includes smoking-induced tissue hypoxia in healthy human smokers35 and reduced radiation control of cancers by carbon monoxide inhalation in animal models.36 In addition, use of antioxidant vitamin supplementation during radiotherapy increased risk of disease recurrence only among those who smoked during radiotherapy5 and not among those who smoked in the year prior to or subsequent to radiotherapy. Alternate biologic mechanisms in addition to hypoxia induction include nicotine interactions with both the mitogen-activated protein kinase and Akt pathways, which may inhibit apoptosis in response to therapy37,38 and reduction by nicotine of the cytotoxic effects of cisplatin and radiation in head and neck cancer cell lines in vitro.39
Meta-analyses of randomized controlled trials for patients with locoregionally advanced HNSCC have estimated that cisplatin-based concurrent chemoradiotherapy confers an approximately 8% absolute improvement in 5-year survival when compared with radiotherapy alone.40 When compared with HPV-negative patients, patients with HPV-positive tumors have increased response rates to cisplatin-based induction chemotherapy41,42 and to radiotherapy. Whether p16-positive and p16-negative patients have a differential response to the addition of cisplatin to radiotherapy is unknown. Chaturvedi et al27 recently reported that, from 1984 to 2004 in the United States, OS significantly increased for individuals with HPV-positive but not HPV-negative OPC. How the adoption of organ preservation chemoradiotherapy after 1999 may have contributed to this increase for the patient with OPC remains unknown.43 Given the nonoverlapping time periods of enrollment and differences in eligibility criteria for RTOG 9003 and RTOG 0129, we are unable to inform this question. At this time, a randomized controlled trial for the HPV-positive patient comparing radiation versus concomitant cisplatin radiation appears unlikely to be performed.
Our data underscore the importance of measuring tobacco exposure in the context of clinical trials. Indeed, it has previously been recommended that all cooperative groups assess tobacco exposure via a standardized and centralized questionnaire.44 In RTOG 1016, a phase III trial for HPV-positive patients with OPC that will compare accelerated radiotherapy in combination with either cisplatin or cetuximab, mandatory assessment of tobacco exposure will be performed by use of validated instruments.45,46 Our data on smoking during radiotherapy also strongly support the implementation of smoking cessation programs and studies to evaluate the effect of smoking cessation on disease control.
|Characteristic||All Patients With OPC|
|Without p16 Determination (n = 456)||With p16 Determination (n = 190)|
|Zubrod performance status|
|Tonsillar fossa, tonsil||171||37.5||89||46.8|
|Base of tongue||170||37.3||57||30.0|
|Age started smoking, years||390||85.5||158||83.2|
|Cigarette use, years||437||95.8||178||93.7|
|Cigarettes smoked per day||434||95.2||182||95.8|
NOTE. Karnofsky performance status was collected and converted to Zubrod performance status. Anemia is defined as hemoglobin ≤ 13.5 g/dL for men and ≤ 12.5 g/dL for women. A pack-year is defined as the equivalent of smoking one pack of cigarettes per day for 1 year.
Abbreviations: AFX-C, accelerated fractionation with concomitant boost radiotherapy; AFX-S, accelerated fractionation with split radiotherapy; AJCC, American Joint Committee on Cancer; HFX, hyperfractionation radiotherapy; NOS, not otherwise specified; OPC, oropharyngeal carcinoma; Q1-Q3, quartile 1 to quartile 3; SFX, standard fractionation.
See accompanying editorial on page 2030
Supported by Grants No. RTOG U10 CA21661 and CCOP U10 CA37422 from the National Cancer Institute and No. NIDCR DE16631 from the Pennsylvania Commonwealth Universal Research Enhancement Program.
Presented in part at the 46th Annual Meeting of the American Society of Clinical Oncology, Chicago, IL, June 4-8, 2010.
The contents of this article are the sole responsibility of the authors and do not necessarily represent the official views of the NCI or the Commonwealth of Pennsylvania.
Authors' disclosures of potential conflicts of interest and author contributions are found at the end of this article.
The author(s) indicated no potential conflicts of interest.
Conception and design: Maura L. Gillison, Andy Trotti, Sharon Spencer, Christine H. Chung, K. Kian Ang
Provision of study materials or patients: Andy Trotti, Sharon Spencer, K. Kian Ang
Collection and assembly of data: Maura L. Gillison, Richard Jordan, Weihong Xiao, William H. Westra, Jonathan Harris, K. Kian Ang
Data analysis and interpretation: Maura L. Gillison, Qiang Zhang, Weihong Xiao, Jonathan Harris, K. Kian Ang
Manuscript writing: All authors
Final approval of manuscript: All authors