Secular trends in the survival of patients with laryngeal carcinoma , 1995 – 2007

In Canada in the year 2013, 1050 new cases of laryngeal cancer and 390 deaths attributable to that cancer were reported1. The treatment of laryngeal cancer can be morbid, causing impaired vocalization, dysphagia, aspiration, and cosmetic deformity2–8. The subsite (glottis, supraglottis, or subglottis) and extent of cancer determine the treatment options available9. For early-stage laryngeal cancers, the treatment options include radiation and partial surgical excision of the larynx9. Advanced-stage cancers are treated with a combination of chemotherapy and radiation (“chemoradiation”) or laryngectomy (complete removal of the larynx)9. A Cochrane systematic review of treatment in early laryngeal cancer demonstrated no difference in survival for surgical excision compared with radiation10. Radiation offers improved voice outcome; partial surgical excision offers a shorter treatment duration and preserves radiation as a treatment option for recurrence or a second primary cancer11,12. In advanced laryngeal cancer, a landmark randomized controlled trial demonstrated no difference in overall survival for laryngectomy compared with chemoradiation, with higher laryngeal preservation rates in the chemoradiation group13. However, ABSTRACT


INTRODUCTION
In Canada in the year 2013, 1050 new cases of laryngeal cancer and 390 deaths attributable to that cancer were reported 1 .The treatment of laryngeal cancer can be morbid, causing impaired vocalization, dysphagia, aspiration, and cosmetic deformity [2][3][4][5][6][7][8] .The subsite (glottis, supraglottis, or subglottis) and extent of cancer determine the treatment options available 9 .For early-stage laryngeal cancers, the treatment options include radiation and partial surgical excision of the larynx 9 .Advanced-stage cancers are treated with a combination of chemotherapy and radiation ("chemoradiation") or laryngectomy (complete removal of the larynx) 9 .
A Cochrane systematic review of treatment in early laryngeal cancer demonstrated no difference in survival for surgical excision compared with radiation 10 .Radiation offers improved voice outcome; partial surgical excision offers a shorter treatment duration and preserves radiation as a treatment option for recurrence or a second primary cancer 11,12 .In advanced laryngeal cancer, a landmark randomized controlled trial demonstrated no difference in overall survival for laryngectomy compared with chemoradiation, with higher laryngeal preservation rates in the chemoradiation group 13 .However,

Background
Recent reports suggest a decline over time in the survival of patients newly diagnosed with laryngeal cancer in spite of developments in treatment practices.Our study set out to determine whether the survival of patients with laryngeal cancer in Ontario has changed over time.

Methods
This population-based cohort study of patients diagnosed with laryngeal cancer in the province of Ontario between 1995 and 2007 used data extracted from linked provincial administrative and registry databases.Its main outcomes were overall survival, laryngectomy-free survival, and survival ratio relative to an age-and sex-matched general population.

Results
The 4298 patients newly diagnosed with laryngeal cancer during the period of interest were predominantly men (n = 3615, 84.1%) with glottic cancer (n = 2787, 64.8%); mean age in the group was 66 years (interquartile range: 59-74 years).Patient demographics did not significantly change over time.Overall, 5-year survival was 57.4%; laryngectomy-free survival was 45.4%.Comparing patients from three eras (1995-1998, 1999-2003,  2004-2007) and adjusting for age, sex, and comorbidity status, we observed no differences in overall survival or laryngectomy-free survival over time.The 5-year relative survival ratio for patients with laryngeal cancer compared with an age-and sex-matched group from the general population was 81.1% for glottic cancer and 44.5% for supraglottic cancer.compared with radiation alone, chemoradiation was later confirmed in a meta-analysis to achieve higher overall survival for patients with advanced laryngeal cancer 14 .Thus, chemoradiation became an alternative to laryngectomy for advanced laryngeal cancer, achieving higher laryngectomy-free survival 14 .As a result of those clinical trials, the years since the mid-1990s have seen a trend toward endoscopic partial surgery for early-stage cancers and chemoradiation for advanced-stage cancers 15 .
Although clinical trials would support that shift in treatment, several population-based studies have shown a decline over time in the survival of patients newly diagnosed with laryngeal cancer [15][16][17] .The reasons for the discrepancy are unknown, but addressing them is important, because the discordance calls into question the current treatment paradigm.However, the generalizability of earlier population-based studies are limited because of incomplete participation in cancer registries, variable access to health care, lack of reports of laryngectomy-free survival, and no adjustment for patient comorbidities [15][16][17] .To overcome those limitations and to provide an accurate assessment of patient survival, we used linked health care administrative and registry databases to study the overall and laryngectomy-free survival of patients newly diagnosed with laryngeal cancer in Ontario over a 13-year period.We hypothesized that, in a universal health care system with near-complete reporting adjusted for patient comorbidities, the overall survival of laryngeal cancer patients would be seen to have remained unchanged over time and the laryngectomy-free survival would be seen to have improved.

Study Design and Setting
Residents of Ontario (2014 population estimate: 13,678,700 1 ) have universal access to hospital care and physician services.Those encounters are recorded in large population-based health care databases that are held at the Institute for Clinical Evaluative Sciences.We conducted a population-based retrospective cohort study of all patients with laryngeal carcinoma.The reporting of the study follows guidelines for observational studies (Table i) 18 , and the study itself was approved by the Sunnybrook Health Sciences Centre Research Ethics Board.

Data Sources
We used five linked databases: the Ontario Cancer Registry, which records data on all patients with nonskin cancers diagnosed in Ontario (mandatory reporting) 19,20 ; the Discharge Abstract Database held by the Canadian Institute for Health Information, which records all admissions to hospitals and includes information about diagnoses and procedures performed; the Ontario Health Insurance Plan database, which contains information on all fee-for-service physician claims for inpatient and outpatient services; the Registered Persons Database, which contains vital statistics about all permanent residents of Ontario; and the National Ambulatory Care Reporting System database, which collects data on ambulatory care visits, including day surgery, outpatient clinics, cancer clinics, and emergency department visits.We previously used those data sources to study secular trends in other conditions 21 .
For the present work, we used the Ontario Cancer Registry to identify laryngeal cancer patients and the Discharge Abstract Database, the National Ambulatory Care Reporting System, the Ontario Health Insurance Plan database, and the Registered Persons Database to define patient characteristics, baseline comorbidities, and patient outcomes.Diagnoses and procedures were determined using the codes in the International Classification of Diseases, 9th revision (pre-2002) and 10th revision (post-2002), and the Canadian Classification of Diagnostic, Therapeutic, and Surgical Procedures (pre-2002) and the Canadian Classification of Health Interventions (post-2002).Whenever possible, we applied database codes that had been validated by chart review [22][23][24][25][26][27] .The databases were linked using unique encoded identifiers (encrypted Ontario health card numbers that are unique to each resident eligible for health care services paid by the government) available starting July 1991, after the assignment of new health card numbers in Ontario.

Patients
We included patients 18 years of age and older with a new diagnosis of laryngeal carcinoma in the period 1995-2007.To allow for a complete 3-year look-back for baseline comorbidities, cohort accrual began on January 1, 1995.We restricted our cohort to patients who were residents of Ontario and who had a histologic diagnosis of squamous cell carcinoma.A prior validation study found a sensitivity of 89.8% and a positive predictive value of 96.8% for the diagnostic code for laryngeal cancer in the registry compared with a clinical database 28 .The date of the laryngeal cancer diagnosis ("index date") served as the start time for follow-up.
We further obtained information about the subsite and stage of the laryngeal cancer and about the location of treatment.Staging data were available only for the subpopulation diagnosed from 2004 to 2007.Registry staging ranged from i to iv and was derived from either the American Joint Committee on Cancer staging manual (6th or 7th edition) 29 .We classified patients staged i and ii as "early-stage."According to both versions of the staging manual, that grouping includes only patients with local disease and excludes patients with regional or distant    metastases.Patients staged iii and iv were grouped as "advanced-stage."That group included patients with advanced local disease and patients with regional or distant metastasis.

Outcomes
We assessed two primary outcomes in the 5 years following a new diagnosis for laryngeal cancer: overall survival and laryngectomy-free survival.We defined overall survival as the proportion of patients alive 5 years from the date of diagnosis and laryngectomy-free survival as the proportion of patients alive 5 years from the date of diagnosis with an intact larynx.Deaths (including out-of-hospital mortality) are well ascertained in the Registered Persons Database, which provides accurate mortality data for all Ontario residents 30 .We also assessed the survival ratio for the cohort relative to an age-and sex-matched general population.

Statistical Analysis
We categorized the study period into three eras: 1995-1998, 1999-2003, and 2004-2007.We chose those eras because staging data were available only for patients diagnosed from 2004 to 2007, which allowed for division of the cohort into three approximately equal periods.We applied chi-square tests to compare differences in baseline characteristics for patients in the three eras.Kaplan-Meier curves were generated, and associated log-rank tests were performed by era for each of the two primary outcomes, overall and grouped by cancer subsite and stage.We used a Cox proportional hazards model to investigate the association between era of diagnosis and each outcome (adjusting for the influence of potential confounders) and Schoenfeld residuals to verify the proportional hazards assumption 31 .
We further produced 5-year relative survival ratio (rsr) estimates.The 5-year rsr is a measure of disease severity and prognosis; it is used in comparisons addressing the impact of cancer on life expectancy.We calculated 5-year rsrs by dividing the percentage of patients alive 5 years from the date of diagnosis by the expected 5-year survival for a matched group of individuals in the general population.The latter group was matched on age, sex, diagnosis date, and rural (population < 10,000) or urban residence.
All analyses were performed using the SAS software application (version 9.3: SAS Institute, Cary, NC, U.S.A.).We interpreted 2-tailed p values less than 0.05 as statistically significant.

Cohort Description and Demographics
The registry search identified 4341 patients newly diagnosed with laryngeal cancer (squamous-cell histology) from January 1, 1995, to December 31, 2007.We excluded 34 patients because of an invalid provincial health care number or missing demographic information.Another 8 patients were excluded because they were not Ontario residents, and 1 patient was excluded because the reported death date fell on or before the date of diagnosis.The final cohort included 4298 patients (Figure 1).Table ii presents the baseline characteristics of the patients.The cohort consisted primarily of patients with a diagnosis of glottic cancer (64.8%, n = 2787).Just a little more than a quarter of the patients had a diagnosis of supraglottic cancer (28.2%, n = 1210).The remainder of the patients had a diagnosis of subglottic cancer (1.8%, n = 76) or a nonspecific diagnosis (5.2%, n = 225).The patients were predominantly men (84.1%), and mean age in the cohort was 66 years (interquartile range: 59-74 years).Patients were more likely to come from lower-income neighbourhoods (46.7% low income vs. 33.4% high income).Most of the patients (70.6%) had not been hospitalized in the 3 years before their laryngeal cancer diagnosis; however, 42.3% had a history of chronic lung disease.We observed no major differences in patient demographics or comorbidities in the three eras.
Table ii presents the stage distribution of the patients (when available) overall and by subsite.Among patients with a diagnosis of glottic cancer, 58.6% presented with early-stage disease, and 18.9% presented with advanced-stage disease.In contrast, patients with a diagnosis of supraglottic cancer more commonly presented with advanced-stage disease (57.2%).Staging information was missing or miscoded for 313 of the 1368 patients belonging to the staging data era (22.9%).

Five-Year Survival Trends
The 5-year overall survival for all patients diagnosed with laryngeal cancer was 57.4% (Figure 2), yielding a rsr of 65.5%.The 5-year laryngectomy-free survival rate was 45.4% (Figure 2).Comparing patients diagnosed with laryngeal cancer in the three eras, we observed no statistically significant change in overall survival nor in laryngectomy-free survival for glottic or supraglottic cancer (Figure 3).
Among patients with glottic cancer, the 5-year overall survival rate, the laryngectomy-free survival rate, and the rsr were 67.1%, 55.5%, and 81.1% respectively.In early-stage glottic disease, 5-year overall survival was 76.8%; it was 43.4% in advanced-stage disease (Figure 4).
Among patients with supraglottic disease, the 5-year overall survival rate, the laryngectomy-free survival rate, and the rsr were 39.5%, 28.0%, and 44.5% respectively.In early-stage disease, the 5-year overall survival was 49.3%; it was 34.5% in advanced-stage disease (Figure 4).
Trends in laryngectomy-free survival were similar whether considering glottic compared with supraglottic cancers or early-stage compared with advanced-stage disease (Figure 4).

DISCUSSION
Among patients newly diagnosed with laryngeal cancer over a 13-year period, our study showed no difference in overall survival or laryngectomy-free survival after adjustment for age, comorbidity score, year of diagnosis, and individual comorbidities.Additionally, 5-year laryngectomy-free survival remained poor, at 55.5% for glottic cancer and 28.0% for supraglottic cancer.We had hypothesized that laryngectomy-free survival would have improved for laryngeal cancer patients from 1995 to 2007, but we found that not to be the case.

Comparison with Other Studies
We found no change in laryngectomy-free survival during the study period, which conflicts with results reported in clinical trials 13,34 .The reasons for that finding are speculative, because our data sources were unable to provide accurate treatment data.
Compared with cancer patients from the general population, patients enrolled in clinical trials might experience improved survival for several reasons.Patients enrolled in clinical trials are highly selected, are followed more closely, and are treated at high-volume academic centres.Enrolled patients tend to be younger, white, educated, insured, and healthconscious; they also tend to have fewer comorbidities than the general population of patients with cancer.It is possible that, after the U.S. Veterans Affairs Study published in 1991 13 (a landmark study that demonstrated equivalent survival and improved laryngeal preservation for chemoradiation compared with surgery), treatment practices in Ontario changed in favour of chemoradiation before the period covered by our study.Another possible explanation is that the selection of patients for laryngectomy might   not have changed in Ontario since the mid-1990s.Treatment practices might have changed in favour of chemoradiation and away from laryngectomy as primary treatment; however, salvage laryngectomy after failed chemoradiation might account for stable laryngectomy-free survival overall.Finally, the results of the Veterans Affairs study might have been used to treat patients who would not have met the inclusion criteria for the study, possibly resulting in worse overall survival because of the toxicity of the treatment.Our results demonstrating no change in overall survival are in keeping with population-based studies from Australia, the Netherlands, Spain, and the United States (using data from the Surveillance, Epidemiology and End Results database) in patients with a new diagnosis of laryngeal carcinoma (Table viii) [35][36][37][38][39] .Using the U.S. National Cancer Database, both Hoffman et al. 17 and Chen et al. 15 found a decline in the survival of patients newly diagnosed with laryngeal cancer during the periods 1985-1996 and 1985-2007 respectively.Taken together, the research on population survival trends in patients with laryngeal cancer has not shown an improvement in either overall survival or laryngectomy-free survival.

Strengths and Limitations
This study has several strengths.To our knowledge, this is the first population-based study to report the trend of laryngectomy-free survival, which is an important outcome in the evaluation of treatment efficacy in the patient population of interest.Procedural and diagnostic codes were well-documented, with little chance of misclassification.The validated diagnostic code for laryngeal cancer in our data sources has excellent operational characteristics 28 .Our reported 5-year rsr (65.5%) and 5-year overall survival (57.4%) are true and consistent with published Canadian Cancer Statistics data for newly diagnosed laryngeal cancer during 2006-2008, which reported a 5-year rsr of 63% and a 5-year overall survival of 55% 1 .The distributions of patient age, sex, subsite, and stage were consistent with details published in other population-based studies 15,17 .
Our data sources and study also have limitations.We lacked staging information before 2004.Additionally, from 2004 onward, staging data for 23% of the patients were missing-a percentage similar to that observed for missing data other studies 15 .We lacked access to detailed treatment information.Disease-specific survival was not calculated, because cause of death has a low sensitivity in cancer registries and population databases 28 .We chose instead to use the rsr, which is a more ap-  propriate measure for population survival analyses because it determines whether the cancer has shortened life compared with the projected lifespan of age-, sex-, and residence-matched individuals.Secular changes in database codes and coding practices could have influenced some of our results; however, most codes were stable during the period of interest.

CONCLUSIONS
To date, clinical trials have demonstrated improved laryngeal preservation and a meta-analysis has shown improved survival in laryngeal cancer patients treated with chemoradiation.However, in routine practice, the 5-year overall and laryngectomy-free survivals for patients with laryngeal cancer in Ontario has not changed since 1995.Our results serve as a call to action to determine the factors that are preventing the gains observed in clinical trials from being realized in routine care.

ACKNOWLDGMENTS
This study was supported by the Institute for Clinical Evaluative Sciences (ices) Western.An annual grant from the Ontario Ministry of Health and Long-Term Care (mohltc) funds ices.Core funding for ices Western is provided by the Academic Medical Organization of Southwestern Ontario (amoso), the Schulich School of Medicine and Dentistry (ssmd), Western University, and the Lawson Health Research Institute (lhri).This project was supported by a London Regional Cancer Program (lrcp) Catalyst Grant.The results, opinions, and conclusions are those of the authors and are independent from the funding sources.No endorsement by ices, amoso, ssmd, lhri, lrcp, or mohltc is intended or should be inferred.The funding agency had no role in the design and conduct of the study; collection, management, analysis, or interpretation of the data; preparation, review, or approval of the manuscript; nor in the decision to submit the manuscript for publication.The first author (SDM) had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

CONFLICT OF INTEREST DISCLOSURES
We have read and understood Current Oncology's policy on disclosing conflicts of interest, and we declare that we have none.
Describe all statistical methods, including those used to control for confounding.Methods (b) Describe any methods used to examine subgroups and interactions.Methods (c) Explain how missing data were addressed.Methods (d) If applicable, explain how loss to follow-up was addressed.Not applicable (e) Describe any sensitivity analyses.Methods Results Participants 13 (a) Report numbers of participants at each stage of study-for example, numbers potentially eligible, examined for eligibility, confirmed eligible, included in the study, completing follow-up, Give reasons for nonparticipation at each stage.Methods (c) Consider use of a flow diagram.
Give the characteristics of study participants (for example, demographic, clinical, social) and information on exposures and potential confounders.

figure 1
figure1 Selection of patients for the laryngeal cancer cohort.

e92
Current OnCOlOgy-VOlume 22, number 2, April 2015 Copyright © 2015 Multimed Inc.Following publication in Current Oncology, the full text of each article is available immediately and archived in PubMed Central (PMC).

Table ii (
b) Indicate the number of participants with missing data for each variable of interest.Results(c) Summarize follow-up time (for example, average and total time).ResultsOutcome data15 Report numbers of outcome events or summary measures over time.

Current OnCOlOgy-VOlume 22, number 2, April 2015
Copyright © 2015 Multimed Inc.Following publication in Current Oncology, the full text of each article is available immediately and archived in PubMed Central (PMC).

table ii
Baseline characteristics of patients newly diagnosed with laryngeal cancer inOntario, 1995-2007 Current OnCOlOgy-VOlume 22, number 2, April 2015Copyright © 2015 Multimed Inc.Following publication in Current Oncology, the full text of each article is available immediately and archived in PubMed Central (PMC).
Charlson comorbidity index and history of all individual comorbidities were examined using administrative codes during the 3 years before the cancer diagnosis date (not including the diagnosis date).c Available starting late 2003; provided only for years of diagnosis after 2004.
a Chosen such that all patients with staging data were included in one group.b

table iii
Hazard ratios for 5-year mortality after a diagnosis of laryngeal cancer For cancer diagnosis site and 18 patient characteristics [age, socioeconomic status, score on the Charlson comorbidity index, sex, year of diagnosis, alcoholism, cancer (excluding laryngeal), carotid endarterectomy, chronic kidney disease, chronic liver disease, chronic lung disease, coronary artery disease, dementia, diabetes, heart failure, myocardial infarction, pneumonia, stroke].Pts = patients; py = person-years; hr = hazard ratio; ci = confidence interval. a

e94 Current OnCOlOgy-VOlume 22, number 2, April 2015
Copyright © 2015 Multimed Inc.Following publication in Current Oncology, the full text of each article is available immediately and archived in PubMed Central (PMC).

table iv
Hazard ratios for death after a diagnosis of laryngeal cancer = hazard ratio; ci = confidence interval; cci = Charlson comorbidity index. hr

table v
Hazard ratios for 5-year mortality or laryngectomy after a diagnosis of laryngeal cancer a For cancer diagnosis site and 18 patient characteristics [age, socioeconomic status, score on the Charlson comorbidity index, sex, year of diagnosis, alcoholism, cancer (excluding laryngeal), carotid endarterectomy, chronic kidney disease, chronic liver disease, chronic lung disease, coronary artery disease, dementia, diabetes, heart failure, myocardial infarction, pneumonia, stroke].Pts = patients; py = person-years; hr = hazard ratio; ci = confidence interval.

e95 Current OnCOlOgy-VOlume 22, number 2, April 2015
Copyright © 2015 Multimed Inc.Following publication in Current Oncology, the full text of each article is available immediately and archived in PubMed Central (PMC).

table vi
32,33d ratio for death or laryngectomy after a diagnosis of laryngeal cancer For cancer diagnosis site and 18 patient characteristics [age, socioeconomic status, score on the Charlson comorbidity index, sex, year of diagnosis, alcoholism, cancer (excluding laryngeal), carotid endarterectomy, chronic kidney disease, chronic liver disease, chronic lung disease, coronary artery disease, dementia, diabetes, heart failure, myocardial infarction, pneumonia, stroke].bViolates the proportional hazards assumption.The correspond-ing hazard ratios can be interpreted as an average effect over the 5-year follow-up period32,33.hr = hazard ratio; ci = confidence interval; cci = Charlson comorbidity index. a

e96 Current OnCOlOgy-VOlume 22, number 2, April 2015
Copyright © 2015 Multimed Inc.Following publication in Current Oncology, the full text of each article is available immediately and archived in PubMed Central (PMC).table vii Death or laryngectomy among laryngeal cancer patients, by era a Patients were followed to a maximum of 5 years.