Department of Head and Neck Surgery, The University of Texas M. D. Anderson Cancer Center, Houston, Texas
Department of Epidemiology, The University of Texas M. D. Anderson Cancer Center, Houston, Texas
Corresponding author: Erich M. Sturgis, MD, MPH, Department of Head and Neck Surgery, Unit 1445, The University of Texas MD Anderson Cancer Center, 1515 Holcombe Boulevard, Houston, TX 77030; Fax: (713) 794-4662; email@example.com
We thank Stephanie Deming for editing the article.
A recent review of the Surveillance, Epidemiology, and End Results registry suggested that patients with index squamous cell carcinoma (SCC) of the oropharynx (SCCOP) are less likely to develop second primary malignancies (SPM) than patients with index SCC of nonoropharyngeal sites (oral cavity, larynx, hypopharynx). The objectives of this study were to determine the impact of index primary tumor site on SPM risk and to explore factors that potentially affect this risk within a large, prospectively accrued cohort of patients with index SCC of the head and neck (SCCHN).
A cohort of 2230 patients with incident SCCHN was reviewed for development of SPM. Kaplan-Meier analysis, log-rank testing, and Cox proportional hazards models were used to detect the impact of various factors, including index tumor site, on SPM risk.
The SPM rate was lower for patients with index SCCOP than for patients with index nonoropharyngeal cancer (P < .001). Among patients with SCCOP, former smokers had a 50% greater risk of SPM, and current smokers had a 100% greater risk of SPM than never-smokers (Ptrend = .008). Also among patients with SCCOP, those with the classic SCCHN phenotype had an SPM risk similar to that of patients with index nonoropharyngeal cancers; those with a typical human papillomavirus phenotype had a very low SPM risk. SPM most commonly occurred at nontobacco-related sites in patients with index SCCOP and at tobacco-related sites in patients with index nonoropharyngeal cancers.
Squamous cell carcinoma (SCC) of the head and neck (SCCHN) (oropharynx, oral cavity, larynx, and hypopharynx) is associated with high rates of subsequent second primary malignancy (SPM).[1, 2] Studies have demonstrated that the incidence of SPM in patients with index SCCHN is 2% to 3% per year, and the risk remains constant throughout the life of the patient.[1, 3-6] Although overall survival appears to be improving for some patients with SCCHN because of an evolution in the etiology of the disease, improvements in tobacco control and cessation, and therapeutic advances, SPM remains a major problem.[1, 7] SPM is a leading cause of long-term mortality among patients with SCCHN.[2, 3] The prognosis for patients who develop SPM is less favorable than the prognosis for patients with index SCCHN alone.[6, 8, 9]
A recent analysis of the Surveillance, Epidemiology, and End Results registry has revealed a dramatic shift in the rates of SPM development among patients with SCCHN. Whereas patients with SCC of the oropharynx (SCCOP) previously had high rates of SPM, they now have lower rates of SPM than patients with SCCHN at any other site. This change may reflect another impact of the current epidemic of SCCOP related to human papillomavirus (HPV).[10-12]
Smoking and alcohol use are the traditional risk factors for SCCHN. Smokers with SCCHN have a worse prognosis than nonsmokers, because smoking increases the risk of SPM development.[4, 6, 15] In smokers with index SCCHN, the majority of SPMs occur in the head and neck, lung, or esophagus.[4, 6, 16] Although the incidence of SCC of the oral cavity, hypopharynx, and larynx has decreased over the past 2 decades, the incidence of SCCOP has been rising approximately 5% per year.[10, 12] At the same time, the proportion of patients with SCCOP who have a limited history of smoking or alcohol use has been increasing, and recent studies have attributed many of these cases to HPV.[10, 11, 17] In contrast to patients with traditional, HPV-negative SCCOP, patients with HPV-associated SCCOP present at a younger age with limited or no history of smoking.[10, 17] Furthermore, patients with HPV-associated SCCOP have a lower incidence of SPM.[18, 19]
In light of the significant risk and increased mortality because of SPM in patients with SCCHN, it is crucial to refine our understanding of SPM risk among patients with index SCCHN to optimize screening and follow-up strategies and improve long-term outcomes. The objectives of this study were to determine the impact of index cancer site on the risk of SPM and to explore factors that potentially affect this risk within a large cohort of patients who had incident SCCHN treated at a tertiary cancer center in the United States.
MATERIALS AND METHODS
This research was approved by the Institutional Review Board of The University of Texas M. D. Anderson Cancer Center. All patients were participants in a prospective molecular epidemiologic protocol of incident SCCHN between 1995 and 2010 and completed an epidemiologic questionnaire. Patients were included if they had previously untreated, pathologically confirmed SCC of the oral cavity, oropharynx, hypopharynx, and/or larynx (N = 2524). Patients were excluded from the analysis if they were not treated at our center (N = 184), if they received only palliative care (N = 89), or if they had no follow-up after treatment (N = 21), leaving a final sample size of 2230 patients.
Throughout their treatment and post-treatment courses, patients had regularly scheduled clinical and radiographic examinations with treating physicians (typically quarterly the first year, 4 months the second year, 6 months the third year, and annually thereafter). On the basis of modified criteria of Warren and Gates, second lesions were considered SPMs if they were of different histopathologic type than the index tumor, if they occurred >5 years after treatment for the index tumor, and/or if they were clearly separated from the index tumor by normal epithelium, as indicated by findings on clinical and radiographic assessment. A synchronous tumor was defined as a malignancy diagnosed at the same time as or during the staging of the index SCCHN. Pulmonary lesions were considered to be SPMs if they had a nonsquamous histology or if they were isolated squamous lesions that appeared >5 years after the initial SCCHN and were believed to be SPMs by the thoracic medical oncologist and the thoracic surgeon. If there was a difference of opinion or question regarding the origin of a tumor (ie, recurrence vs SPM), then the second lesion was classified as a recurrence rather than an SPM. SPMs were then categorized into 1 of 3 classes: SCCHN; tobacco-related cancer, not SCCHN (cancers of the lung and bronchus, esophagus, or urinary bladder); or other. Nonmelanoma skin cancers were not recorded or considered to be SPMs.
Analysis was done using the Stata 12 statistical software package (StataCorp, College Station, Tex). The threshold for statistical significance was set at P < .05, and all tests were 2-sided. The Kaplan-Meier failure function was used to calculate the 3-year and 5-year rates of SPM within groups. Kaplan-Meier curves were generated to evaluate the proportion of patients with SPM, and log-rank tests were performed to detect statistically significant differences between groups. The time to event was calculated from the date of diagnosis of the index cancer to the date of diagnosis of SPM. Cox proportional hazards models were generated to calculate hazard ratios to detect differences in SPM risk according to demographic and clinical variables in patients with index SCCOP and those with index nonoropharyngeal cancer (oral cavity, hypopharynx, larynx) and to detect differences in SPM risk between patients with and without the classic SCCHN phenotype and patients with and without the typical HPV phenotype. The classic SCCHN phenotype was defined as SCCHN in a man aged >50 years who was a current smoker and ever-drinker. The typical HPV phenotype was defined as SCCOP in a non-Hispanic white man aged <60 years who was a never-smoker or former smoker and presented with a T1-T2/N1-N3 tonsil or base of tongue cancer. Multivariable models were created by minimizing Akaike information criterion.
The median follow-up for patients in the study who were still alive was 33.1 months for all patients, 36.1 months (range, 0.1-170.8 months) for patients with SCCOP, and 28.4 months (range, 0.1-166.8 months) for patients with nonoropharyngeal cancer. Five patients had synchronous SCCOP, 7 patients had synchronous nonoropharyngeal cancers (1 synchronous oral cavity cancer, 1 synchronous hypopharyngeal cancer, and 5 synchronous cancers in >1 nonoropharyngeal subsite), and 19 patients had synchronous tumors of both oropharyngeal and nonoropharyngeal sites.
Relation Between Demographic and Clinical Variables and Risk of Second Primary Malignancy
Of all 2230 patients, 236 (10.6%) developed an SPM (3-year SPM rate, 7.6%) (Table 1). Older age at diagnosis of the index SCCHN was associated with a greater risk of SPM. The greater the level of smoking exposure, as measured by pack-years or smoking status, the higher the SPM rate (Table 1). Patients with SCCOP had a lower 3-year SPM rate (5.2%) than patients with nonoropharyngeal cancer (8.3%) and had a lower incidence of SPM over the course of follow-up (Fig. 1). Moderate-to-severe comorbidity also was associated with a higher rate of SPM.
Table 1. Factors Associated With the Risk of Second Primary Malignancy
Total No. of Patients (%)
No. of Patients With SPM (%)
3-Year SPM Rate [95% CI], %
5-Year SPM Rate [95% CI], %
Abbreviations: CI, confidence interval; SPM, second primary malignancy.
For current and former smokers only; pack-years data were missing for 42 patients, including 2 (4.6%) who developed SPM.
Nonoropharynx sites were oral cavity, larynx, and hypopharynx.
Grade was unavailable for 274 patients, including 28 (10.2%) who developed SPM.
Multivariable Analysis of Factors Influencing Second Primary Malignancy Risk by Index Primary Tumor Site
Among patients with index SCCOP, age <60 years, being a former smoker or never-smoker, fewer pack-years of exposure among smokers, lower comorbidity score, negative lymph node (N0) category, and TNM stage I or II were associated with a lower risk of SPM in univariate analysis; however, after multivariable adjustment, only smoking status remained a significant predictor of SPM risk (Table 2). In contrast, among patients with index nonoropharyngeal cancer, older age, male sex, being a current or former smoker, more pack-years of exposure among smokers, and being an ever-drinker were associated with a higher risk of SPM in univariate analysis; however, after multivariable adjustment, only age ≥45 years remained a significant predictor of SPM risk. Among patients with index SCCOP, the risk of SPM increased 1% per year (adjusted hazard ratio, 1.01; 95% confidence interval, 0.98-1.03); and, among patients with index nonoropharyngeal cancer, the risk of SPM increased 2% per year (adjusted hazard ratio, 1.02; 95% confidence interval, 1.00-1.03).
Table 2. Factors Associated With the Risk of Second Primary Malignancy by Index Tumor Site
Grade: Well/moderately well differentiated vs poorly differentiated)e
TNM stage: III/IV vs I/II
Treatment: Single-modality vs multimodality
Results of detailed subgroup analyses of the impact of age, smoking, alcohol use, and HPV phenotype on SPM risk are presented in Table 3. Among patients with index SCCOP, those aged ≥60 years, current smokers, and smokers with ≥20 pack-years of exposure had 3-year SPM rates (Table 3) similar to the rate of patients with index nonoropharyngeal cancer (8.3%) (Table 1). Similarly, patients with index SCCOP who had a classic SCCHN phenotype had a 3-year SPM rate (Table 3) similar to that of patients with index nonoropharyngeal cancer. Finally, although HPV status was not available for most patients in this large retrospective cohort, patients with index SCCOP who had a typical HPV phenotype had a very low SPM rate (Table 3, Fig. 2). Among patients with index nonoropharyngeal cancer, SPM rates were strongly associated with increased age, and there was a trend toward greater risk with smoking exposure (Table 3).
Table 3. Selected Factors Associated With Risk of Second Primary Malignancy by Index Tumor Site
Abbreviations: CI, confidence interval; HR, hazard ratio; HPV, human papillomavirus; NA, not applicable; Ref, referent category; SCCHN, squamous cell carcinoma of the head and neck; SPM, second primary malignancy.
The analysis included the 19 patients who had synchronous index tumors at both oropharyngeal and nonoropharyngeal sites.
The analysis was adjusted for age, race, smoking status, alcohol use, comorbidity index, tumor subsite, T classification, and N category.
The analysis was adjusted for age, race, smoking status, alcohol use, comorbidity index, tumor site, T classification, and N category.
Criteria were age ≥50 years, male, current smoker, and ever-drinker.
Criteria were age <60 years, male, non-Hispanic white, never/former smoker, tonsil/base of tongue site, and T1-T2/N1-N3 disease.
Second Primary Malignancy Locations of Second Primary Malignancies by Index Primary Tumor Site
Among patients with index nonoropharyngeal cancer, the most common sites of SPMs were sites traditionally associated with tobacco-related cancers, including the lung, esophagus, and bladder, and the lung was the most common site (Table 4). In contrast, patients with index SCCOP tended to develop SPMs in other sites. No cases of anogenital SPM developed in this cohort except for 1 cervical SPM in a woman who had an index nonoropharyngeal cancer.
Table 4. Second Primary Malignancy Sites in Patients With Index Squamous Cell Carcinoma of the Oropharynx or Nonoropharyngeal Cancera
Index Tumor Site: No. of Patients With SPM (%)
Abbreviations: SCCHN, squamous cell carcinoma of the head and neck; SPM, second primary malignancy.
The analysis did not include the 19 patients who had synchronous index tumors at oropharyngeal and nonoropharyngeal sites.
Tobacco-related, not SCCHN
Lung and bronchus
In this study, patients with SCCOP had a lower risk of SPM than patients with nonoropharyngeal cancer. Overall, smoking and older age were associated with the risk of SPM. Among patients with SCCOP, being a current or former smoker and having greater exposure to smoking were associated with the risk of SPM in multivariable analysis. Among patients with nonoropharyngeal cancer, age was a strong predictor of SPM risk in multivariable analysis, but smoking and alcohol use also seemed to modify SPM risk. The most common locations of SPMs in patients with index SCCOP were nontobacco-related sites, whereas the most common locations of SPMs in patients with index nonoropharyngeal cancer were tobacco-related sites.
Dose-response relations between smoking and the risk of SPM also were suggested. The associations between smoking and SPM risk are consistent, and previous studies demonstrated a direct association between increased smoking exposure and increased risk of SPM. However, the increase in risk may be minimal after patients have exceeded a certain threshold. In our study, while a smoking history of greater than 40 pack-years raised the overall risk of SPM, this risk was only slightly higher than the risk associated with a history of 20 to 39 pack-years in patients with non-oropharyngeal cancer and was lower than the risk associated with a history of 20-39 pack-years in patients with SCCOP. Others have reported similar patterns above certain levels of smoking exposure, and these patterns were similar to those observed in other smoking-related cancers, such as lung, esophageal, and bladder cancer.[22, 23] Tomek and McGuirt concluded that, once a critical level of cellular damage because of smoking has occurred, subsequent exposure may not have any further impact on the risk of developing SPM.
Although some studies have indicated that tobacco cessation after diagnosis of an index SCCHN reduces the risk of SPM,[24, 25] other studies have not reported a consistent benefit.[15, 26] Although we did not have data on smoking exposures after diagnosis, we did observe that, both in patients with index SCCOP and in patients with index nonoropharyngeal cancer, former smokers had lower SPM rates than current smokers. Although HPV status was not available for most patients in this large cohort, it is likely that a high proportion of the patients with SCCOP in the cohort had HPV as the primary etiologic agent, given that 38% of these SCCOP patients were never-smokers. We might expect HPV positivity to modify the impact of smoking history and smoking cessation on SPM risk. In light of our observation that both former smokers and light smokers with index SCCOP had very low SPM rates, HPV positivity may have affected the SPM risk in this subset of patients. Regardless, we actively promote tobacco cessation among patients with index SCCHN, not only because of the potential benefit in terms of reducing SPM risk but also to improve therapeutic outcomes and reduce toxic effects and complications of surgery and radiation therapy.
Previously, it was believed that, in patients with index SCCHN, most SPMs arose in the head and neck region followed by the lung.[4, 6, 27] However, recent studies have demonstrated that the lung is the most common site of SPM in patients with index SCCHN, and our data are in agreement with those findings.[16, 28] The tendency of patients with SCCHN to develop SPM in sites traditionally associated with tobacco-related cancers is consistent with the concept of field cancerization first suggested by Slaughter et al However, within the index SCCOP group, we observed that a high percentage of SPMs occurred in sites not traditionally associated with tobacco, among which the prostate was the most common site. Although some have suggested that patients with SCCOP are at increased risk for SPM at HPV-related sites,[16, 30] we did not observe an HPV-associated field effect; only a single SPM was observed at an HPV-related site (cervix) in a patient with index nonoropharyngeal cancer.
The recent epidemic of HPV-associated SCCOP has altered the incidence and sites of SPMs for patients diagnosed with SCCOP, as suggested by Morris et al. HPV-associated SCCOP often occurs in middle-aged, white men who have limited or no history of smoking. The different risk-factor profile of this population, reflecting a unique oncologic process, may result in different clinical presentations, fewer comorbidities, decreased risk of SPMs, and improved outcomes.[1, 2, 17-19] However, within the SCCOP group, we observed that patients with the traditional SCCHN risk factors, including smoking history and classic SCCHN phenotype, experienced a high rate of SPM. This is consistent with the traditional behavior and poor outcome of SCCOP before the HPV epidemic.
The current study has several limitations. Most important, we were not able to report the results segregated according to HPV status. However, in a recent study by Peck et al analyzing 182 patients with index SCCOP from this same cohort for whom HPV serology data were available, we observed very low 5-year SPM rates for HPV-seropositive patients (5.6%) and more typical 5-year SPM rates for HPV-seronegative patients (14.6%); furthermore, among the 34 HPV-seropositive patients who were never-smokers, the 5-year SPM rate was 0%. We also observed that the SPM rate decreased over time for patients with SCCOP, whereas the rate did not change significantly among patients with nonoropharyngeal cancer (data not shown). This is consistent with the concept of increasing prevalence of HPV in SCCOP tumors leading to a lower risk for SPM. In addition, our findings are similar to those of 2 smaller reports in which the tumor HPV status in patients with SCCOP was documented.[18, 19] In all 3 studies (ours and the preceding 2 smaller studies), 63% to 75% of SPMs in patients with HPV-negative disease were identified in traditional smoking-associated locations (head and neck, lung and bronchus, esophagus, and bladder). When taken together, only 46% (32 of 70) of the SPMs among patients with HPV-associated disease occurred in smoking-associated locations. Although previous studies suggested that SCCOP is also associated with SPMs at sites traditionally associated with HPV, such as the cervix, penis, and anus,[16, 30] we did not observe any SPMs at anogenital sites among patients with index SCCOP, and only 2 anogenital cancers (both cervix) of 60 SPMs were reported among patients with HPV-associated oropharyngeal cancer in the studies by Ang et al and Huang et al.[18, 19]
Another limitation of our study was that patients' smoking status was obtained through a prospective epidemiologic questionnaire. Recall bias may have occurred, especially in heavier smokers, who may have underestimated packs smoked per day. A further limitation was that data on continued smoking and alcohol use or tobacco-cessation efforts after diagnosis of the index SCCHN were not collected. Finally, although the length of follow-up was reasonable, this analysis of SPM incidence is a retrospective review of the outcomes of a prospectively accrued patient cohort, and loss to follow-up and lack of standard SPM screening may have biased the outcomes. However, the median follow-up for patients with index SCCOP was actually greater than that for patients with index nonoropharyngeal cancer (36 months vs 28 months).
In conclusion, in patients with SCCHN, the index primary tumor site affects the risk of SPM and the sites of SPM. Patients with index SCCOP who have limited or no smoking exposure and/or a typical HPV phenotype are at particularly low SPM risk, whereas smokers with any index SCCHN are at elevated SPM risk, especially current smokers and those with greater smoking exposure. Accurate understanding of the unique risk factors and carcinogenesis of SCCHN at particular subsites is essential to establishing more effective secondary disease-prevention strategies and improving long-term patient outcomes.
This research was supported in part by the National Institutes of Health through The University of Texas M. D. Anderson Cancer Center Support Grant, CA016672 and through R01 ES 11740 and R01 CA 131274 to Q.W.. Kristina Dahlstrom was supported by award no. R25CA057730 from the National Cancer Institute. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Cancer Institute or the National Institutes of Health.