ORIGINAL ARTICLE. Five-Year Survival Rates and Time Trends of Laryngeal Cancer in the US Population

Transcription

1 ORIGINAL ARTICLE Five-Year Survival Rates and Time Trends of Laryngeal Cancer in the US Population Maura Cosetti, MD; Guo-Pei Yu, MD; Stimson P. Schantz, MD Objectives: To provide comprehensive temporal trend analysis of 5-year relative survival rates of laryngeal cancer using the Surveillance, Epidemiology, and End Results database; and to expand on prior reports by including inclusion of laryngeal tumor location, stage, age at diagnosis, treatment strategy, and histologic grade. Design: Retrospective cohort analysis using the Surveillance, Epidemiology, and End Results database of the National Cancer Institute. The Surveillance, Epidemiology, and End Results data were used to design 5 cohorts of patients with laryngeal cancer: , , , , and Fiveyear survival rates were analyzed according to tumor site, stage, and grade; age at diagnosis; and treatment strategy. The joinpoint regression model was used to assess survival trends and their statistical significance. Results: Among patients with supraglottic cancer, 5-year relative survival rates for distant disease worsened over time while rates for local and regional disease did not change (P =.01 and P.05, respectively). For localized glottic cancer, survival remained stable from to However, patients with regional and distant glottic cancer demonstrated a significant decrease in survival in the past 3 decades (P.001). This trend was independent of treatment strategy. Finally, the proportion of well-differentiated tumors in patients with regional laryngeal cancer decreased over time (P.001 for supraglottic and P=.007 for glottic). Conclusions: A decreasing 5-year survival trend was found among patients with glottic cancer who had regional disease and in all patients with distant disease. Histopathologic trends not previously reported in those with laryngeal cancer seem to parallel those seen in other tobacco-related cancers. These trends may reflect the effect of birth cohorts and implicate the relationship between carcinogenic exposure and host factors, rather than the influence of treatment. Arch Otolaryngol Head Neck Surg. 08;134(4):3-379 Author Affiliations: Departments of Otolaryngology (Drs Cosetti, Yu, and Schantz) and Biostatistics and Epidemiology Service (Dr Yu), The New York Eye and Ear Infirmary and New York Medical College, Valhalla (Drs Yu and Schantz), New York, New York. SEVERAL RECENT STUDIES HAVE suggested that the incidence of laryngeal cancer and other smoking-related cancers is declining in North America 1-3 and Western Europe. 4,5 One factor in this decline may be decreased exposure to carcinogens, specifically tobacco. It has been hypothesized that this declining incidence may reflect achievement in antismoking efforts. Diagnostic technology and treatment for laryngeal cancer have undergone significant changes during the past several decades. 1-6 However, the impact of these changes on survival rates of laryngeal cancer remains unclear. A recent study by Hoffman et al 1 reports declining survival among patients with laryngeal cancer in the past 2 decades. Stratification by tumor location is an important aspect of laryngeal cancer survival data because glottic cancer is known to impart a better prognosis than cancers originating in the supraglottis. 7,8 Notably, prior studies have not included other factors known to influence survival, such as histopathologic tumor grade, in their analyses. In this study, we analyzed changes in 5-year survival rate according to laryngeal subsites and tumor stage. We further considered the relationship between these survival trends and age at diagnosis, treatment strategy, and histologic tumor grade. METHODS We obtained data from the Surveillance, Epidemiology, and End Results (SEER) program of the National Cancer Institute 9 and applied information on cancer diagnosis only to the residents of 9 population-based registries (ie, 5 states [Connecticut, Hawaii, Iowa, New Mexico, and Utah] and 4 standard metropolitan areas [Atlanta, Georgia; Detroit, Michigan; San Francisco Oakland, California; and Seattle Puget Sound, Washington]). We used SEER computer software (Stat 5.2.2) to conduct survival rate analysis. We selected cases of laryngeal cancer that contained follow-up information. The Interna- (REPRINTED) ARCH OTOLARYNGOL HEAD NECK SURG/ VOL 134 (NO. 4), APR 08 3 Downloaded From: on /05/18 08 American Medical Association. All rights reserved.

2 tional Classification of Diseases for Oncology site codes used for laryngeal cancer are C3 (glottic), C321 (supraglottic), and C322 to C329 (all other laryngeal cancers). To better estimate the survival of primary laryngeal cancer, we excluded patients whose laryngeal cancer represented a second primary malignant neoplasm. To compare changes in survival rates, we designed 5 patient cohort periods: , , , , and Each of these 5 patient cohorts consists of data from a 2-year registry and contains data for an observation period of 5 years. The design of the cohorts ensured that the individual 5-year observation period for each cohort did not overlap. We assumed that the variation of survival rates of the 5 patient cohorts represented the trend of laryngeal cancer survival from 1977 to 03. Five-year survival data from these patient cohorts were further analyzed according to 2 subsites of the larynx (glottic and supraglottic) and tumor stage at diagnosis (localized, regional, distant, and unknown). In the SEER database, localized disease is defined as a malignancy limited to the organ of origin. 9(p6) Regional disease is defined as 1) extension beyond the limits of the organ of origin directly into surrounding organs or tissues; 2) into regional lymph nodes by way of the lymphatic system; or 3) by a combination of extension and regional lymph nodes. 9(pp5-7) Finally, the SEER program characterizes distant disease as tumor that spread to parts of the body remote from the primary site, either by direct extension or by discontinuous metastasis. 9(p8) Cause of death, as indicated on the death certificate, was grouped into 3 broad categories: laryngeal cancer, all other cancer, and noncancer. Three categories were used to calculate causespecific survival rates for laryngeal cancer (ie, death due to primary laryngeal cancer, death due to metastatic disease or second primary cancer, and non cancer-related death). In addition, treatment modality, including surgery or radiation therapy, was assessed. Within the SEER database, treatment with radiation encompasses all forms of therapy, including external beam, radioactive implants, radioisotopes, or a combination. 9 Information on chemotherapeutic agents is not included in the database. Further analysis was done using patient age at diagnosis. Within each cohort, patients were grouped into 3 broad age categories: to 49 years, to 64 years, and 65 years and older. The Mantel-Haenszel method was used to calculate overall rate ratios (RRs) for 5-year survival rates, period-specific RRs, and 95% confidence intervals (CIs). The survival rates of the period were defined as a reference category and those of subsequent periods were used in comparison. In the analysis, we described age-adjusted incidence rates for glottic and supraglottic cancers to explain the variation of survival over time. To test statistical significance of incidence trends and obtain estimated annual percentage change, we used joinpoint regression to analyze the age-adjusted rates based on 2-year combination. The joinpoint regression model is a statistical method that describes changing trends over successive segments of time and the amount of increase or decrease within each segment. 11,12 The statistics derived from these models are the estimated annual percentage change and 95% CIs of the estimated annual percentage changes. RESULTS INCIDENCE Age-adjusted incidence rates for glottic and supraglottic cancer are shown in Figure 1. Overall, incidence trends at both sites varied with age and stage. A significant decrease in the incidence of local supraglottic cancer over time was found among all age groups (Figure 1D-F). In the oldest age group (those 65 years), the incidence of regional supraglottic disease increased (Figure 1F). For glottic cancer, the incidence of local and regional disease decreased significantly in the youngest age group (those aged -49 years) (Figure 1A). In the - to 64- year-old subjects and in those 65 years and older, this decreased incidence was seen only in localized glottic disease (Figure 1B and C). Overall, incidence trends for regional glottic cancer were nonsignificant in both older age groups (Figure 1B and C). SURVIVAL Figure 2 depicts 5-year relative survival rates for supraglottic cancer. In , the 5-year survival was 63.0% for localized disease, 46.1% for regional disease, and 6.4% for distant disease. While survival rates for local and regional stage supraglottic cancer have not significantly changed over time, a significant decrease was found in 5-year survival of distant stage supraglottic cancer (Figure 2C). Survival trends of glottic cancer also differ by stage (Figure 3). The most current 5-year survival statistics by stage are as follows: from , 5-year survival was.6% for localized disease,.6% for regional disease, and 27.4% for distant disease. For localized disease (Figure 3A), 5-year relative survival rates were nearly identical for and (89.6% and.6%, respectively). Thus, no significant survival trend was found. On the other hand, a nearly % decrease in survival over time was observed for regional and distant glottic disease (RRs, 0. and 0.74, respectively; P.001 for both). Specifically, the 5-year relative survival rate of regional glottic cancer decreased from 78.6% in to.6% in (Figure 3B); and for distant stage disease, it decreased from 52.7% to 27.4% (Figure 3C). There was no significant change in survival for unstaged disease (Figure 3D). Additional analysis of the decreased survival trend of regional glottic cancer is demonstrated in Table 1. Patients with regional glottic cancer are grouped by age at diagnosis, and survival rates are categorized by cause, thereby distinguishing cancer- and non cancer-related deaths in each age group. In the - to 64-year-old subjects, the 5-year glottic cancer and all cancer specific survival rates significantly decreased between and While not significant, the exact opposite trend was seen in the other 2 age groups: glottic and all cancer related survival rates increased over time in the - to 49-year-old and the 65 years and older age groups. In non cancer-specific survival rates, all groups showed a slight, but insignificant, decrease over time. The RRs (95% CIs) by age are as follows (ratios were calculated by dividing the survival rate of by the survival rate of ). For laryngeal cancer, in the - to 49-year-old group, the RR was 1.1 (95% CI, ); in the - to 64-year-old group, the RR was 0.8 (95% CI, ) (P=.002); and in the 65 years and older group, the RR was 1.1 (95% CI, ). For all cancer, in the - to 49-year-old group, the RR was 1.1 (95% CI, ); in the - to 64-year-old group, the RR was 0.8 (REPRINTED) ARCH OTOLARYNGOL HEAD NECK SURG/ VOL 134 (NO. 4), APR Downloaded From: on /05/18 08 American Medical Association. All rights reserved.

3 A 0.8 D 0.5 Localized disease Regional disease 0.7 Age-Adjusted Rate (1/0 000) B 9 E Age-Adjusted Rate (1/0 000) C 9 F Age-Adjusted Rate (1/0 000) Year of Diagnosis Year of Diagnosis Figure 1. Age-adjusted incidence rates for glottic and supraglottic cancers from 1973 to 03 by age and tumor stage. A, Glottic cancer in patients aged to 49 years; B, glottic cancer in patients aged to 64 years; C, glottic cancer in patients 65 years and older; D, supraglottic cancer in patients aged to 49 years; E, supraglottic cancer in patients aged to 64 years; and F, supraglottic cancer in patients 65 years and older. In A, the estimated biannual percentage change (EBAPC) (95% confidence interval [CI]) was 7.3 ( 8.7 to 5.8) for localized disease and 3.7 ( 5.5 to 1.9) for regional disease (P.001 for both). In B, the EBAPC (95% CI) was 2.4 ( 0.1 to 5.0) (P=.06) for localized disease from 1973 to 1985, 9.2 (.6 to 7.7) (P.001) for localized disease from 1985 to 03, and 1.5 ( 3.0 to 0) (P=.05) for regional disease. In C, the EBAPC (95% CI) was 2.1 ( ) (P=.004) for localized disease from 1973 to 1995, 12.2 ( 17.8 to 6.3) (P=.001) for localized disease from 1995 to 03, and 0.7 ( 1.0 to 2.3) (P=.) for regional disease. In D, the EBAPC (95% CI) was 8.6 ( 12.0 to 5.2) (P.001) for localized disease and 1.6 ( 0.3 to 3.7) (P=.) for regional disease. In E, the EBAPC (95% CI) was 4.0 (1.1 to 9.3) (P=.11) for localized disease from 1973 to 1987, 11.1 ( 15.0 to 6.9) (P.001) for localized disease from 1987 to 03, 9.6 (0.1-.1) (P=.05) for regional disease from 1973 to 1983, and 1.6 ( 4.2 to 1.0) (P=.) for regional disease from 1983 to 03. In F, the EBAPC (95% CI) was 7.8 ( ) (P=.04) for localized disease from 1973 to 1983, 4.1 ( 6.3 to 1.8) (P=.002) for localized disease from 1983 to 03, 7.9 ( ) (P.001) for regional disease from 1973 to 1987, and 1.0 ( 3.3 to 1.4) (P=.39) for regional disease from 1987 to 03. (95% CI, ) (P.001); and in the 65 years and older group, the RR was 0.9 (95% CI, ). For non cancerspecific data, in the - to 49-year-old group, the RR was 0.9 (95% CI, ); in the - to 64-year-old group, the RR was 0.9 (95% CI, ); and in the 65 years and older group, the RR was 1.0 ( ). Examining trends in survival of regional stage glottic cancer by age and birth cohort revealed an overall de- (REPRINTED) ARCH OTOLARYNGOL HEAD NECK SURG/ VOL 134 (NO. 4), APR Downloaded From: on /05/18 08 American Medical Association. All rights reserved.

4 A 0 C (n = 59) [.5%] (n = 37) [19.1%] (n = 37) [21.9%] (n = 35) [6.4%] (n = 87) Relative (n = 198) [68.2%] (n = 225) [67.1%] (n = 2) [63.7%] (n = 211) [63.0%] (n = 363) B 0 D 0 Relative (n = 264) [46.5%] (n = 332) [41.0%] (n = 366) [46.3%] (n = 424) [46.1%] (n = 43) (n = 13) [41.2%] (n = 28) [46.9%] (n = 25) [44.2%] (n = 19) [46.9%] (n = 31) Figure 2. Time trends of 5-year relative survival rates by tumor stage for supraglottic cancer. A, Local stage; B, regional stage; C, distant stage; and D, unstaged. The overall Mantel-Haenszel method-calculated relative ratio of 5-year survival (other cohorts vs the cohort) was 0.95 in A, 0.96 in B, 0.52 inc,and 1.06 in D. The numbers in parentheses after each period indicate the number of patients at the start of the study, and the percentages in brackets indicate the 5-year relative survival rates. In C, for , the rate ratio (RR) was 0.63 (95% confidence interval [CI], ) (P.05); for , the RR was 0.72 (95% CI, ) (P.05); and for , the RR was 0.21 (95% CI, ) (P=.01). crease in survival in patients of late cohorts (Figure 4). The effect of birth cohort was particularly evident in the survival curve of the group aged to 59 years, with a significant decrease beginning in the 1924 to 1929 cohort (Figure 4B). Restated, patients with regional glottic cancer in this age group had a lower 5-year survival if they were born after Analysis according to treatment strategy is shown in Figure 5. In the - to 64-year-old age group, a decreasing survival trend was observed across all treatments, including surgery only, radiation only, and combined surgery and radiation. Cancer-specific 5-year survival of those with regional glottic cancer did not vary with treatment modality. A significant trend in survival was found in non cancer-specific mortality in patients treated with surgery only (Figure 5D). Further analysis on the proportion of patients treated with each modality is depicted in Table 2. Over time, there is a significant trend toward increased use of radiation therapy and decreased treatment with surgery alone. Last, regional stage tumors in this age group were classified according to histologic tumor grade (Table 3 and Table 4). The proportion of well-differentiated tumors in patients with regional glottic disease decreased significantly over time. In , 28.2% of regional disease was histologically well differentiated, significantly higher than the 15.8% of well-differentiated disease seen in No significant time-dependent changes in tumor grade were found in localized disease. A similar trend was found in regional supraglottic disease. COMMENT Release of newly expanded data from the SEER database program has prompted a reexamination of survival rates for a multitude of cancer types. Laryngeal cancer was unique among this group of 24 cancers as the only type to experience a decrease in 5-year survival rates. 1 In a recent analysis of the SEER database, Carvalho et al 13 report survival trends for various sites of head and neck cancer. In their analysis of laryngeal cancer, they report a decrease in 5-year survival for local disease, no change for patients with regional disease, and an improvement in 5-year survival in distant or late-stage laryngeal cancer. These results seem to conflict with the findings of (REPRINTED) ARCH OTOLARYNGOL HEAD NECK SURG/ VOL 134 (NO. 4), APR Downloaded From: on /05/18 08 American Medical Association. All rights reserved.

5 A C Relative (n = 634) [89.6%] (n = 0) [.3%] (n = 676) [86.7%] (n = 916) [.6%] (n = 378) (n = 21) [52.7%] (n = 23) [48.2%] (n = 11) [39.5%] (n = 16) [27.4%] (n = 22) B 0 D 0 Relative (n = 244) [78.6%] (n = 291) [63.9%] (n = 252) [64.7%] (n = 334) [.6%] (n = 834) (n = ) [69.2%] (n = 39) [76.5%] (n = 87) [61.6%] (n = 38) [73.2%] (n = 53) Figure 3. Time trends of 5-year relative survival rates by tumor stage for glottic cancer. A, Local stage; B, regional stage; C, distant stage; and D, unstaged. The Mantel-Haenszel method-calculated relative ratio of 5-year survival (other cohorts vs the cohort) was 1.00 in A, 0. in B, 0.74 in C, and 1.02 in D. The numbers in parentheses after each year indicate the number of patients at the start of the study, and the percentages in brackets indicate the 5-year relative survival rates. In B, for , the rate ratio (RR) was 0.81 (95% confidence interval [CI], ); for , the RR was 0.82 (95% CI, ); and in , the RR was 0.77 (95% CI, ) (P.001 for all periods). this study. One explanation for this discrepancy lies in the years of data included in each analysis. Carvalho et al use the SEER data from 1974 to 1997 in their calculation of overall survival trends; however, in their analysis of stage-specific survival, they include only the data from 1983 to In doing so, they exclude a period of relatively improved survival for regional and distant disease reported from 1974 to Exclusion of these data would contribute to the variance in results between studies. In addition, Carvalho et al grouped all subsites of laryngeal cancer, rather than a separate analysis of glottic vs supraglottic cancer, thus adding to the differences between studies. A recent report by Hoffman et al 1 also examined trends in laryngeal cancer survival in the past 2 decades. Using the National Cancer Data Base and SEER data, their study reports a decrease in survival of patients with glottic and supraglottic laryngeal cancer that varied by TNM stage. 1 In general, the results of the present study support those of Hoffman et al and report a more pronounced decline in survival in patients with advanced stage disease. The present report uses the SEER database program to analyze 5-year survival in patients with laryngeal cancer by age, laryngeal subsite, tumor stage, histologic grade, and treatment strategy. Overall, the data depict a significant decline in survival in patients with regional stage glottic cancer; from 1977 through 03, a nearly % decrease in 5-year survival rates of advanced glottic cancer was observed. In addition, a significant decrease was found in survival rates of distant disease of glottic and supraglottic cancer during these 2 decades. There are multiple factors that may explain these results, including (1) the phenomenon of stage drift; (2) changing trends in treatment; (3) spurious data from the early cohorts; (4) the nature of relative survival rates and the impact of comorbidities; (5) a change in causative factors or carcinogenic exposure, such as tobacco; and (6) a change in the disease itself. To begin, the confounding influence of stage drift can be created by advances in diagnostic technology. Specifically, a lesion classified as early-stage disease in the pre computed tomographic scan era would be diagnosed as advanced when more sophisticated imaging is used. Therefore, a trend toward increased incidence of advanced disease may reflect the introduction and application of improved diagnostic technology and not a change in the (REPRINTED) ARCH OTOLARYNGOL HEAD NECK SURG/ VOL 134 (NO. 4), APR Downloaded From: on /05/18 08 American Medical Association. All rights reserved.

6 Table 1. Time Trend of Cause-Specific 5-Year Observed Survival Rates by Age for Patients With Regional Stage Glottic Cancer a Cause-Specific 5-y Observed Survival Age Group, y No. of Patients Alive at Start Laryngeal Cancer Specific All Cancer Specific Noncancer Specific (7.7) 71.3 (8.6) 95.1 (4.8) (8.0) 78.7 (8.5).0 (6.7) (6.1) 78.6 (7.8) 95.5 (4.4) (7.8) 81.1 (8.5) 85.8 (7.7) (2.9) 83.2 (3.6).4 (2.9) (3.7).4 (4.2) 81.5 (3.6) (4.0) 69.5 (4.8) 84.7 (3.9) (3.9) 63.3 (4.2) 81.9 (3.8) (4.9) 68.3 (5.0) 77.7 (4.7) (4.5) 61.4 (4.7) 78.0 (4.5) (4.6) 63.4 (4.9) 74.9 (4.5) (3.7) 62.1 (4.1) 75.1 (3.9) a Data are given as rate (SE), measured in percentages, for each age group at each year. aggressiveness or behavior of the disease itself. The overall trends in our data support a decreasing incidence in local disease and an increase in regional disease over the period studied. On the other hand, stage drift would likely lead to a positive survival trend in categorical groupings. For example, as radiologic sophistication restaged early disease as more advanced, the advanced cancer group would be diluted with patients who demonstrate an earlier form of the disease. Thus, stage drift would likely manifest in an overall increase in survival of advanced stage cancer. However, this was not demonstrated in our results. In fact, our data support the opposite (ie, a decrease in survival of regional stage glottic cancer). Thus, the complex phenomenon of stage drift cannot solely account for the results of this study. Not only the diagnostic modalities but also the treatment of laryngeal cancer has changed in the past years. More important, nonsurgical treatment, specifically radiation and chemotherapy, was introduced and widely applied to laryngeal carcinoma. 1 As depicted in Figure 5, however, the decrease in survival of regional glottic cancer is likely not entirely attributable to a change in treatment strategy. The magnitude of decrease in cancerspecific survival rates was approximately % for all treatment modalities, including surgery, radiation, and combined therapy, from 1977 to Throughout the period examined, there was a significant trend toward more frequent use of radiation therapy alone and decreased use of surgery alone. While reflecting an evolving treatment paradigm in laryngeal cancer, patients in each treatment modality demonstrated decreased survival over time. Evidence of the insignificance of treatment modality is also found in the stable survival trends among patients with regional glottic cancer of the age groups to 49 years and 65 years and older. If treatment plays a role in promoting survival, it should impact all ages rather than only patients aged to 64 years, unless different treatments are used at different ages. These data are not an appropriate evaluation of therapeutic effectiveness of a given treatment modality. However, they do demonstrate a consistent decrease in relative survival independent of changes or trends in treatment strategy. Analysis of the data must mention the possibility of spurious results. In Figure 4, patients in the earliest cohorts demonstrate strikingly good survival. It is possible that these data are spurious and inappropriately influence the analysis toward significance. Another potentially confounding factor is the influence of comorbid disease. Survival rates in this study refer to relative survival rates. As the treatment of comorbidities improves with time, patients in later cohorts may survive their comorbid disease and succumb to laryngeal carcinoma. This is somewhat unlikely, however, because this confounding variable should influence all patients within a given cohort equally. Finally, the decrease in survival of patients with regional glottic carcinoma may be explained by a change in causative factors, including carcinogenic exposure. The relationship between carcinogens, specifically tobacco and alcohol, and laryngeal cancer has long been recognized Shifting trends in tobacco use may be implicated in observed changes in survival statistics and disease characteristics. Although the absolute number of tobacco users has decreased dramatically since 1965, recent reports 17,18 suggest that this may disproportionately represent the occasional or limited tobacco user. From 1993 to 00, the percentage of smokers who report everyday tobacco use has increased. 17,18 Thus, while overall tobacco use has clearly decreased, there seems to be a trend toward increased intensity among remaining smokers. Prior research 19 examining carcinogenic exposure on a cellular level has linked intensity of carcinogenic exposure, amount of genetic damage, and carcinogenesis. This positive relationship supports the idea that increasing intensity of tobacco exposure may be related to the increased incidence and decrease survival of late-stage disease. (REPRINTED) ARCH OTOLARYNGOL HEAD NECK SURG/ VOL 134 (NO. 4), APR Downloaded From: on /05/18 08 American Medical Association. All rights reserved.

7 A D (n = 25) [92.0%] (n = 26) [79.7%] (n = 19) [72.3%] (n = 34) [63.7%] (n = 94) (n = 44) [83.0%] (n = 46) [68.5%] (n = 49) [.3%] (n = 56) [63.9%] (n = 23) B 0 E (n = 44) [82.7%] (n = 48) [55.0%] (n = 34) [62.4%] (n = 57) [62.9%] (n = 3) (n = ) [.7%] (n = 36) [63.5%] (n = 33) [73.0%] (n = 46) [69.4%] (n = 99) C (n = 44) [78.8%] (n = 62) [77.1%] (n = ) [73.4%] (n = 51) [63.5%] (n = 134) Figure 4. Cancer-specific survival rates by age and birth cohort for patients with regional stage glottic cancer. A, Patients aged to 54 years; B, patients aged 55 to 59 years; C, patients aged to 64 years; D, patients aged 65 to 69 years; and E, patients aged to 74 years. The overall Mantel-Haenszel method-calculated rate ratio (RR) of cancer-specific survival was 0.78 (95% confidence interval [CI], ) (P=.003) in A, 0.73 (95% CI, ) (P.001) in B, 1.66 (95% CI, ) (P=.) in C, 0.81 (95% CI, ) (P=.005) in D, and 1.35 (95% CI, ) (P=.03) in E. The numbers in parentheses after each year indicate the number of patients at the start of the study, and the percentages in brackets indicate the 5-year relative survival rates. In A, for 1929 to 1934, the RR was 0.87 (95% CI, ) (P=.39); for 1935 to 19, the RR was 0.79 (95% CI, ) (P=.18); and for 1941 to 1946, the RR was 0.69 (95% CI, ) (P=.03). In B, for 1924 to 1929, the RR was 0.67 (95% CI, ) (P=.009); for 19 to 1935, the RR was 0.76 (95% CI, ) (P=.08); and for 1936 to 1941, the RR was 0.75 (95% CI, ) (P=.04). In D, for 1914 to 1919, the RR was 0.83 (95% CI, ) (P=.18); for 19 to 1925, the RR was 0.85 (95% CI, ) (P=.24); and for 1926 to 1931, the RR was 0.77 (95% CI, ) (P=.04). In addition, histologic changes in disease characteristics seem to parallel trends in tobacco use and disease survival. To our knowledge, trends in histopathologic differentiation among laryngeal cancer tumors have not been previously reported. However, this phenomenon has been examined in other tobacco-related cancers, specifically lung cancer. The leading histologic type of lung cancer has undergone a well-documented shift from a squa- (REPRINTED) ARCH OTOLARYNGOL HEAD NECK SURG/ VOL 134 (NO. 4), APR Downloaded From: on /05/18 08 American Medical Association. All rights reserved.

8 A 0 D (n = 41) [84.8%] (n = 38) [66.8%] (n = 19) [.3%] (n = 26) [57.8%] (n = 64) (n = 41) [95.0%] (n = 38) [73.2%] (n = 19) [67.5%] (n = 26) [73.4%] (n = 64) B 0 E (n = 37) [82.1%] (n = 45) [73.3%] (n = ) [69.5%] (n = 51) [66.6%] (n = 145) (n = 37) [85.6%] (n = 45) [87.9%] (n = ) [89.5%] (n = 51) [85.4%] (n = 145) C 0 F (n = ) [82.8%] (n = 47) [68.7%] (n = ) [71.0%] (n = 54) [67.9%] (n = 96) (n = ) [88.6%] (n = 47) [86.7%] (n = ) [87.4%] (n = 54) [87.3%] (n = 96) Figure 5. Survival rates for patients with regional stage glottic cancer aged to 64 years. A, Cancer-specific rates for those who underwent surgery only; B, cancer-specific rates for those who underwent radiation only; C, cancer-specific rates for those who underwent surgery and radiation; D, non cancer-specific rates for those who underwent surgery only; E, non cancer-specific rates for those who underwent radiation only; and F, non cancer-specific rates for those who underwent surgery and radiation. The overall Mantel-Haenszel method-calculated rate ratio of survival was 0.76 (P.001) in A, 0.85 (P=.04) in B, 0.84 (P=.03) in C, 0.76 (P.001) in D, 1.02 (P=.83) in E, and 0.98 (P=.) in F. In A, for , the rate ratio (RR) was 0.79 (95% confidence interval [CI], ) (P=.11); for 1989 to 19, the RR was 0.83 (95% CI, ) (P=.34); for 1995 to 1996, the RR was 0.68 (95% CI, ) (P=.03). In B, for , the RR was 0.89 (95% CI, ) (P=.); for , the RR was 0.85 (95% CI, ) (P=.36); and for , the RR was 0.81 (95% CI, ) (P=.17). In C, for , the RR was 0.83 (95% CI, ) (P=.27); for , the RR was 0.86 (95% CI, ) (P=.36); and for , the RR was 0.82 (95% CI, ) (P=.22). In D, for , the RR was 0.77 (95% CI, ) (P=.02); for , the RR was 0.71 (95% CI, ) (P=.01); and for , the RR was 0.77 (95% CI, ) (P=.03). mous cell carcinoma to adenocarcinoma, and this evolution has paralleled the change in cigarette composition and tobacco use. 3, The introduction and widespread use of filtered cigarettes in the 19s is believed to have changed smoking behaviors and may account for the histologic trends observed in lung cancer. 3,,21 Thun et al (p15) suggest that the increase in lung adenocarcinoma since the 19s is more consistent with changes in smoking behavior and cigarette design than with diagnostic advances. Our histologic analysis of regional (REPRINTED) ARCH OTOLARYNGOL HEAD NECK SURG/ VOL 134 (NO. 4), APR Downloaded From: on /05/18 08 American Medical Association. All rights reserved.

9 Table 2. Treatment of Patients With Regional Stage Glottic Cancer, Aged to 64 Years, Over Time a Type of Treatment (n = 113) (n = 136) (n = 3) (n = 142) (n = 331) b Surgery alone 41 (36.3) 38 (27.9) 19 (18.4) 26 (18.3) 64 (19.3) Radiation alone 37 (32.7) 45 (33.1) (29.1) 51 (35.9) 145 (43.8) Surgery plus radiation (26.5) 47 (34.6) (48.5) 54 (38.0) 96 (29.0) No surgery or no radiation 3 (2.7) 0 2 (1.9) 6 (4.2) 3 (0.9) Surgery or radiation refused 2 (1.8) 0 1 (1.0) 2 (1.4) 1 (0.3) Unknown surgery or radiation 0 6 (4.4) 1 (1.0) 3 (2.1) 22 (6.6) a Data are given as number (percentage) of patients. Column percentages may not total 0 because of rounding = (P.001) for testing the trend of treatment modality. b The increase in number of patients with glottic cancer is because of an increase in the population included in the Surveillance, Epidemiology, and End Results (SEER) database. The SEER regional catchment areas increased from 9 in 1973 to 1991 to 13 in 1992 to 1999 and to 17 in 00 to 03. Table 3. Histologic Grade of Glottic Cancer by Year of Diagnosis for Patients Aged to 64 Years a Year of Diagnosis Stage of Cancer Regional Well differentiated 33 (28.2) 29 (.6) 18 (17.3) 27 (18.6) 53 (15.8) Moderately differentiated 46 (39.3) 55 (39.0) 51 (49.0) 76 (52.4) 192 (57.3) Poorly differentiated 15 (12.8) 21 (14.9) 18 (17.3) 22 (15.2) 57 (17.0) Undifferentiated 0 3 (2.1) 1 (1.0) 2 (1.4) 1 (0.3) Unknown 23 (19.7) 33 (23.4) 16 (15.4) 18 (12.4) 32 (9.6) Total b 117 (0.0) 141 (0.0) 4 (0.0) 145 (0.0) 335 (0.0) Localized Well differentiated 81 (27.6) 94 (29.8) (26.7) 78 (25.9) 125 (26.2) Moderately differentiated 91 (31.0) 9 (34.6) 93 (35.5) 135 (44.9) 213 (44.6) Poorly differentiated 19 (6.5) (6.3) 25 (9.5) 27 (9.0) 29 (6.1) Undifferentiated (0.7) 2 (0.4) Unknown 3 (35.0) 92 (29.2) 74 (28.2) 59 (19.6) 9 (22.8) Total b 294 (0.0) 315 (0.0) 262 (0.0) 1 (0.0) 478 (0.0) a Data are given as number (percentage) of patients in each period for each stage of cancer. For regional data: 2 1 = 7.25 (P =.007) for testing the trend of proportion of well-differentiated tumor. For localized data: 2 1 = 0.81 (P =.37) for testing the trend of proportion of well-differentiated tumor. b The sum of column percentages may not total 0 because of rounding. Table 4. Histologic Grade of Supraglottic Cancer by Year of Diagnosis for Patients Aged to 64 Years a Year of Diagnosis Stage of Cancer Localized Well differentiated 22 (18.3) (16.9) 15 (15.0) (11.1) (12.5) Moderately differentiated 43 (35.8) 54 (45.8) (.0) 49 (54.4) 83 (51.9) Poorly differentiated 22 (18.3) (16.9) 26 (26.0) 17 (18.9) 25 (15.6) Undifferentiated 2 (1.7) 2 (1.7) 4 (4.0) 2 (2.2) 1 (0.6) Unknown 31 (25.8) 22 (18.6) 15 (15.0) 12 (13.3) 31 (19.4) Total b 1 (0.0) 118 (0.0) 0 (0.0) (0.0) 1 (0.0) Regional Well differentiated 16 (11.3) 23 (13.6) (5.6) 15 (8.6) 28 (5.6) Moderately differentiated 65 (45.8) 58 (34.3) 89 (.3) 81 (46.3) 263 (52.6) Poorly differentiated 26 (18.3) 43 (25.4) 59 (33.3) 62 (35.4) 137 (27.4) Undifferentiated 4 (2.8) 6 (3.6) 1 (0.6) 4 (2.3) 9 (1.8) Unknown 31 (21.8) 39 (23.1) 18 (.2) 13 (7.4) 63 (12.6) Total b 142 (0.0) 169 (0.0) 177 (0.0) 175 (0.0) 0 (0.0) a Data are given as number (percentage) of patients. For localized stage: 2 1 = 2.82 (P =.09) for testing the trend of proportion of well-differentiated tumor. For regional stage, 2 1 = 9.42 (P.001) for testing the trend of proportion of well-differentiated tumor. b The sum of column percentages may not total 0 because of rounding. (REPRINTED) ARCH OTOLARYNGOL HEAD NECK SURG/ VOL 134 (NO. 4), APR Downloaded From: on /05/18 08 American Medical Association. All rights reserved.

10 laryngeal cancer also documents a shift toward less differentiation, which seems to parallel decreasing survival trends and changing patterns in tobacco use. These data address the complex interaction between causative agents, including known carcinogens, such as tobacco and alcohol, and host factors that affect disease manifestation and patient survival. The present study used fixed patient cohorts to describe temporal trends of survival in laryngeal cancer. Each of these cohorts had at least a 5-year observation period for survival, and these windows did not overlap each other in time. These methodological considerations are important in the use of cancer registry patients for whom follow-up periods are varied. In general, cancer survival is relatively high during the first several years. Because the last patient cohorts had observation periods of less than 5 years, their survival rate may be higher than those of earlier cohorts, with a greater window of observation. Our analysis excluded patients with second or later cancer primary tumors. Such exclusion is necessary for analysis of cause-specific survival. Inherent in any database analysis, however, is the possibility of incorrect diagnosis or inappropriate attribution of mortality to non cancer-related death when, in fact, a second primary or metastatic disease may be culpable. In conclusion, decreasing 5-year relative survival trends were demonstrated among patients with glottic cancer and regional disease and in patients with both glottic and supraglottic cancer as well as distant disease. Histopathologic trends not previously reported in laryngeal cancer seem to parallel those seen in other tobacco-related cancers. These trends may reflect the effect of birth cohorts with a unique exposure to carcinogens and changing trends in tobacco use over time, rather than an influence of treatment modality. Further research into the biological, genetic, and environmental interactions between carcinogenic exposure, specifically tobacco, and laryngeal cancer is warranted. Submitted for Publication: January 26, 07; final revision received August 21, 07; accepted September 4, 07. Correspondence: Stimson P. Schantz, MD, Department of Otolaryngology, The New York Eye and Ear Infirmary, 3 E 14th St, New York, NY 003 Author Contributions: Drs Cosetti, Yu, and Schantz had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Cosetti, Yu, and Schantz. Acquisition of data: Yu. Analysis and interpretation of data: Cosetti, Yu, and Schantz. Drafting of the manuscript: Cosetti and Schantz. Critical revision of the manuscript for important intellectual content: Cosetti, Yu, and Schantz. Statistical analysis: Yu. Administrative, technical, and material support: Schantz. Study supervision: Schantz. Financial Disclosure: None reported. Previous Presentation: This study was presented as a poster at the Annual Meeting of The Triologic Society; May -22, 06; Chicago, Illinois. REFERENCES 1. Hoffman HT, Porter K, Karnell LH, et al. Laryngeal cancer in the United States: changes in demographics, patterns of care and survival. Laryngoscope. 06; 116(9, pt 2)(suppl 111): McKean-Cowdin R, Feigelson HS, Ross RK, Pike MC, Henderson BE. Declining cancer rates in the 19s. J Clin Oncol. 00;18(11): Wingo PA, Ries LA, Giovino GA, et al. Annual report to the nation on the status of cancer, , with a special section on lung cancer and tobacco smoking. J Natl Cancer Inst. 1999;91(8): Capocaccia R, Micheli A, Berrino F, et al. Time trends of lung and larynx cancers in Italy. Int J Cancer. 1994;57(2): Raitiola HS, Pukander JS. Changing trends in the incidence of laryngeal cancer. Acta Oncol. 1997;36(1): Shah JP, Karnell LH, Hoffman HT, et al. Patterns of care for cancer of the larynx in the United States. Arch Otolaryngol Head Neck Surg. 1997;123(5): Silvestri F, Bussani R, Stanta G, Cosatti C, Ferlito A. Supraglottic versus glottic laryngeal cancer: epidemiological and pathological aspects. ORL J Otorhinolaryngol Relat Spec. 1992;54(1): Stell PM. Prognosis in laryngeal carcinoma: tumour factors. Clin Otolaryngol Allied Sci. 19;15(1): National Cancer Institute, US National Institutes of Health, Surveillance Research Program, Cancer Statistics Branch, Surveillance, Epidemiology and End Results (SEER) Program. SEER*Stat Database: Incidence SEER 17 Regions Limited-Use, Nov 05 Submission ( varying): Linked to County Attributes: Total U.S., Accessed April 06.. Hankey BF, Ries LA, Edwards BK. The Surveillance, Epidemiology, and End Results Program: a national resource. Cancer Epidemiol Biomarkers Prev. 1999; 8(12): Kim HJ, Fay MP, Feuer EJ, Midthune DN. Permutation tests for joinpoint regression with applications to cancer rates. Stat Med. 00;19(3): Hankey BF, Feuer EJ, Clegg LX, et al. Cancer surveillance series: interpreting trends in prostate cancer part I: evidence of the effects of screening in recent prostate cancer incidence, mortality, and survival rates. J Natl Cancer Inst. 1999; 91(12): Carvalho AL, Nishimoto IN, Califano JA, Kowalski LP. Trends in incidence and prognosis for head and neck cancer in the United States: a site-specific analysis of the SEER database. Int J Cancer. 05;114(5): Parkin DM, Pisani P, Lopez AD, Masuyer E. At least one in seven cases of cancer is caused by smoking: global estimates for Int J Cancer. 1994;59(4): Cattaruzza MS, Maisonneuve P, Boyle P. Epidemiology of laryngeal cancer. Eur J Cancer B Oral Oncol. 1996;32B(5): IARC Monographs on the Evaluation of Carcinogenic Risk to Humans: Tobacco Smoking. Vol 38. Lyon, France: International Agency for Research on Cancer, World Health Organization; Centers for Disease Control and Prevention (CDC). Cigarette smoking among adults: United States, MMWR Morb Mortal Wkly Rep. 1994;43(): Centers for Disease Control and Prevention (CDC). Cigarette smoking among adults: United States, 00. MMWR Morb Mortal Wkly Rep. 02;51(29): Schantz SP, Yu GP. The epidemiology of head and neck cancer. In: Fu YS, Wenig B, Abemayor E, Wenig BS, eds. Head and Neck Pathology: With Clinical Correlations. New York, NY: Churchill Livingstone Inc; 01: Thun MJ, Lally CA, Flannery JT, Calle EE, Flanders WD, Heath CW Jr. Cigarette smoking and change in the histopathology of lung cancer. J Natl Cancer Inst. 1997;89(21): Alberg AJ, Samet JM. Epidemiology of lung cancer. Chest. 03;123(1)(suppl):21S- 49S. (REPRINTED) ARCH OTOLARYNGOL HEAD NECK SURG/ VOL 134 (NO. 4), APR Downloaded From: on /05/18 08 American Medical Association. All rights reserved.