A and B, Unadjusted Kaplan-Meier survival curves are shown for patients with cT3N0M0 glottic laryngeal squamous cell carcinomas treated with surgery (alone or in combination with other treatments) or primary radiotherapy (RT) (log-rank χ21 = 0.76) and by treatment modality (log-rank χ25 = 53.53). C and D, Adjusted Kaplan-Meier survival curves are shown for patients with cT3N0M0 glottic laryngeal carcinomas treated with surgery (alone or in combination with other treatments) or primary RT (log-rank χ21 = 0.68) and by treatment modality (log-rank χ25 = 6.22) using inverse probability treatment weighting. CRT indicates chemoradiotherapy.
Kaplan-Meier survival curves of patients with cT3N0M0 glottic laryngeal squamous cell carcinomas treated with RT alone are separated by daily or altered fractionation schemes (log-rank χ21 = 1.97).
eFigure. Flow Diagram of Study Cohort
Customize your JAMA Network experience by selecting one or more topics from the list below.
Ko HC, Harari PM, Chen S, et al. Survival Outcomes for Patients With T3N0M0 Squamous Cell Carcinoma of the Glottic Larynx. JAMA Otolaryngol Head Neck Surg. 2017;143(11):1126–1133. doi:10.1001/jamaoto.2017.1756
What are the survival outcomes in patients with clinical T3N0M0 glottic laryngeal carcinoma who undergo definitive surgical vs primary radiotherapy-based approaches?
In this cohort study of 2622 patients with clinical T3N0M0 glottic laryngeal carcinoma treated with primary surgical or radiotherapy-based approaches, both groups exhibited similar 5-year adjusted survival rates (53% and 54%, respectively).
Results of randomized clinical trials substantiated by national outcomes data support the use of radiotherapy-based organ preservation approaches for patients with clinical T3N0M0 glottic larynx cancer without compromising overall survival.
Radiotherapy (RT)–based organ preservation approaches for patients with advanced laryngeal cancer have been established stepwise through prospective randomized clinical trials. However, broad adoption of these approaches has stimulated discussion about long-term results challenging their applicability in a heterogeneous patient population, most recently for patients with T3 disease.
To define outcomes in patients with clinical T3N0M0 glottic laryngeal cancer treated with definitive surgical and RT-based approaches.
Design, Setting, and Participants
This retrospective cohort study included patients treated from January 1, 2004, through December 31, 2013, with a median follow-up time of 58 months (range, 0-126.6 months) in the National Cancer Database. Of the 4003 patients with T3N0M0 disease, 2622 received definitive therapy defined by the study protocol. Data were obtained from the clinical oncology database sourced from hospital registry data that are collected from more than 1500 Commission on Cancer–accredited facilities. Data were analyzed from September 14, 2016, through April 24, 2017.
Radiotherapy, chemoradiotherapy, surgery, surgery and RT, or surgery and chemoradiotherapy.
Main Outcomes and Measures
Five-year overall survival (OS).
A total of 2622 patients (2251 men [85.9%] and 371 women [14.1%]; median age, 64 years [range, 19-90 years]) were included in the analytic cohort. In the overall patient cohort, the adjusted 5-year survival probability was 53%. No statistical differences were observed between the primary surgery (53%; 95% CI, 48%-57%) and primary RT (54%; 95% CI, 52%-57%) cohorts. In multivariate analysis, patient factors associated with decreased OS included age (hazard ratio [HR], 1.04; 95% CI, 1.03-1.04), insurance status (HR, 1.26; 95% CI, 1.06-1.50), and increasing comorbidity (HR, 1.20; 95% CI, 1.02-1.42).
Conclusions and Relevance
Current management of T3N0M0 glottic laryngeal cancer relies largely on RT-based organ preservation approaches. The present study substantiates randomized clinical trial data supporting the use of RT-based organ preservation approaches for patients with T3N0M0 glottic laryngeal cancer without compromising OS.
The landmark Department of Veterans Affairs (VA) Laryngeal Cancer Study1 pioneered the concept of nonsurgical laryngeal preservation with sequential chemoradiotherapy (CRT). The overall 2-year laryngeal preservation rate was 64%. However, 56% of patients with T4 disease required salvage laryngectomy compared with 29% of patients with lower T-stage tumors. The concept of laryngeal preservation was refined by the RTOG 91-11 trial, which demonstrated an improvement in laryngeal preservation with concurrent CRT.2 In contrast to the VA Laryngeal Cancer Study, the RTOG 91-11 trial excluded patients with large-volume T4a disease that penetrated through the thyroid cartilage or extended more than 1 cm into the tongue base. Reported 2-year rates of laryngeal preservation were 75% for sequential CRT and 88% for concurrent CRT. The overall survival (OS) rates were similar across all treatment arms at 55%. Given these data, radiotherapy (RT)-based laryngeal preservation approaches have been argued to be appropriate therapy for selected patients with low-volume T4a disease or less, leading to a gradual adoption of these techniques in the management of locally advanced laryngeal cancer.3 However, this change in practice has stimulated considerable discussion about long-term results and applicability to a heterogeneous patient population.
Several reports from retrospective database reviews suggest that the increased use of laryngeal preservation with CRT rather than surgical management in the context of locally advanced disease has contributed to decreasing survival among patients with laryngeal cancer.4-6 However, these database reviews also identified decreasing survival outcomes among patients with small-stage laryngeal cancers (T1N0 and T2N0) and stability or small gains in 5-year relative survival rates among patients with advanced laryngeal cancer.7 Leading experts in head and neck cancer have provided detailed review and interpretation of these data.7-9 Among several challenges to analyzing outcomes for different treatment strategies are the nonrandom selection of patients by clinicians for different treatment approaches and the significant heterogeneity of staging included within the category of advanced laryngeal cancer. For example, a national hospital-based cancer registry study included patients with stages III and IV advanced laryngeal cancer to evaluate factors associated with survival.4 Patients categorized as having stage III or IV cancer included highly diverse T and N stage groups such as T1N1, T3N0, and T4bN3. Retrospective comparisons of specific T and N stage groups might diminish the considerable disease burden diversity and treatment selection bias of clinicians.
Laryngeal carcinomas limited to the glottic larynx with vocal cord fixation, invasion of the inner cortex of the thyroid cartilage, and/or spread into the paraglottic fat space constitute T3 disease. Patients with T3 glottic laryngeal carcinomas constitute a significant proportion of patients enrolled in randomized clinical trials for treatment of locally advanced disease (79% in the RTOG 91-11 trial and 65% in the VA Laryngeal Cancer Study), with 50% to 75% presenting without nodal disease.1,2,10 Given the prevalence in this cohort of patients with advanced laryngeal cancers, evaluation of T3N0M0 cases with regard to management and survival outcomes is warranted. Because a randomized clinical trial evaluating outcomes between surgical and RT-based approaches in this patient cohort is unlikely, we used the National Cancer Database (NCDB) to perform an assessment of patterns of care and OS.
We performed a retrospective, observational cohort study using the NCDB. Data were obtained from the clinical oncology database sourced from hospital registry data that are collected from more than 1500 Commission on Cancer–accredited facilities. Patient data from January 1, 2004, through December 31, 2013, included 347 252 patients with head and neck cancer, of whom 4003 were defined as having clinically staged T3N0M0 (cT3N0M0) glottic laryngeal squamous cell carcinoma. This study was approved by the institutional review board of the University of Wisconsin School of Medicine and Public Health, Madison, which waived the need for informed consent for review of deidentified data.
Of the 4003 patients with T3N0 disease, 2622 received definitive therapy defined as surgery- or RT-based approaches. Treatment groups were designated as RT alone, CRT, surgery alone, surgery and RT, or surgery and CRT. The RT-alone group was subdivided into those with daily and accelerated fractionation schemes. The cohort was limited to those receiving external beam RT, for which the sum total dose to the primary tumor ranged from 5000 to 9000 cGy. Lower doses were not considered to be adequate definitive or adjuvant levels and likely represented palliative RT, and higher doses were considered to represent a clerical error. Patients receiving brachytherapy, radioisotopes, or intraoperative RT or with missing information with regard to definitive treatment types were excluded. Surgical techniques were cataloged as surgery performed on the primary tumor, including hemilaryngectomy (excluding supraglottic partial laryngectomy), total laryngectomy, and pharyngolaryngectomy. Primary tumor size was analyzed in millimeters; tumors with sizes listed as greater than 100 mm were also excluded on the basis that they were not clinically feasible for cT3N0M0 and likely represented a clerical error. Definition of the complete study cohort is shown in the eFigure in the Supplement.
The only outcome data analyzed included OS. Deaths that occurred after 5 years were considered to be unrelated to the disease process and therefore were censored. The NCDB records all first-course definitive treatment and eventual survival outcome without information regarding intervening disease progression, recurrence, or salvage therapies.
Patient characteristics, including age, sex, and truncated Charlson/Deyo comorbidity scores (CDCS; range, 0 to ≥2, with higher scores indicating more comorbidities), were incorporated into the analysis. Demographics of the patient population included income, race, facility location as urban or nonurban, year of diagnosis, and insurance status. Treatment facility information was further categorized by facility volume and community vs academic.
Data were analyzed from September 14, 2016, through April 24, 2017. We used standard descriptive statistics to analyze the distribution of covariates throughout the cohort. All baseline demographics and patient characteristics were analyzed using Pearson χ2 tests, whereas continuous variables such as age, radiation dose, and tumor size were analyzed with the Kruskal-Wallis test. Univariate and multivariate analyses with Cox proportional hazards regression methods were used to determine factors correlated with OS. Appropriate effect size and 95% difference in effect between compared variables (95% CI) were used for reporting results. Factors found to be significant in univariate analysis were included and further selected by stepwise method in multivariate analysis. The Cox proportional hazards regression assumption was checked using the test based on Schoenfeld residuals,11 and the proportionality assumption was not violated in the final selected multivariate model.
Survival was evaluated using propensity score–adjusted Kaplan-Meier curves with inverse probability of treatment weighting (IPTW) and the log-rank test to compare OS among the different treatment groups. With use of IPTW, Kaplan-Meier estimators incorporated the inverse of the generalized multitreatment propensity score to account for confounding owing to measured covariates. Each individual received 1 propensity score corresponding to the probability of assignment to each treatment, conditional on baseline characteristics using multinomial logistic regression. Important covariates significantly associated with OS (P < .02) were included in the multinomial logistic model and underwent backward selection, and the final included covariates for adjustment were age, CDCS, insurance type, urban facility location, and facility type. We then calculated the weight of each patient as the inverse of the propensity score corresponding to the treatment received. The details and justification of using the generalized IPTW were provided by Imbens.12
We also grouped treatments into binary units (surgery vs nonsurgery) to compare patient survival using the IPTW technique. Moreover, to evaluate covariate balance, we grouped patients into quintiles according to their estimated binary treatment propensity scores and used the Cochrane-Mantel-Haenszel test and linear regression to verify that measured covariates were balanced across all strata. In addition, a subset survival analysis using a Kaplan-Meier estimator was performed for patients who received RT alone to compare patients with accelerated vs daily fractionation, by which propensity score adjustment was not used owing to good covariate balance between the 2 fractionation groups. Statistical analyses, including propensity scores and survival, were performed using R (version 3.2.2; R Foundation for Statistical Computing), SPSS (version 22; IBM), and SAS (version 9.4; SAS Institute Inc) software. All P values were 2-sided, and P ≤ .05 was considered to be statistically significant.
Baseline patient, disease, and treatment characteristics of the 2622 included patients (2251 men [85.9%] and 371 women [14.1%]) are detailed in Table 1. Median follow-up time for all patients was 58 months (range, 0-126.6 months). Median age was 64 years (range, 19-90 years) for all patients and similar for all therapeutic groups except those receiving RT alone, who had a median age of 72 years (range, 34-90 years). Most patients were white (2047 [78.1%]) and similar across all treatment groups. In the analytic cohort, 1965 patients (74.9%) were treated with RT-based approaches. Among the 657 patients undergoing surgery, 98 (14.9%) underwent partial laryngectomy. Factors associated with primary surgery or primary RT as defined by multivariate logistic regression are described in Table 2. Patients with a CDCS of 1 (odds ratio [OR], 1.90; 95% CI, 1.39-2.60) or 2 or greater (OR, 1.89; 95% CI, 1.20- 2.98), treatment at an academic center (OR, 2.08; 95% CI, 1.60-2.70), increasing tumor size (OR, 1.59; 95% CI, 1.40-1.80), and nonurban facility location (OR, 1.52; 95% CI, 1.10-2.08) were more likely to receive surgery. Those treated from 2010 to 2012 compared with 2004 to 2006 were more likely to receive RT-based treatment (OR, 0.67; 95% CI, 0.48-0.93).
Median adjusted survival in the entire cohort was 59.8 months (95% CI, 53.5-65.2 months). The 5-year OS among patients receiving primary RT-based or surgical management was 54% (95% CI, 52%-57%) and 53% (95% CI, 48%-57%), respectively (Figure 1A). The 5-year OS among patients receiving definitive treatment modalities was 52% (95% CI, 46%-58%) for surgery alone, 54% (95% CI, 45%-63%) for surgery with RT, 50% (95% CI, 35%-63%) for surgery with CRT, 35% (95% CI, 28%-42%) for daily RT, 42% (95% CI, 31%-53%) for accelerated RT, and 59% (95% CI, 55%-62%) for CRT (Figure 1B). Multivariate analysis using a Cox proportional hazards regression model was performed and demonstrated that increasing age (hazard ratio [HR], 1.04; 95% CI, 1.03-1.04), CDCS of 1 (HR, 1.20; 95% CI, 1.02-1.42) or 2 or greater (HR, 2.24; 95% CI, 1.81-2.76), and nonprivate or unknown insurance (HR, 1.26; 95% CI, 1.06-1.50) were associated with worse OS (Table 3). We then performed IPTW Kaplan-Meier analyses using covariates significantly associated with OS for adjustments that included age, CDCS, insurance type, urban location, and facility type. Adjusted 5-year OS among patients receiving primary radiotherapy was 54% (95% CI, 52%-57%); for patients receiving surgical management, 51% (95% CI, 46%-56%) (Figure 1C). The adjusted 5-year OS among patients receiving surgery alone was 51% (95% CI, 44%-58%); surgery with RT, 53% (95% CI, 42%-62%); surgery with CRT, 38% (95% CI, 24%-53%); daily RT, 50% (95% CI, 39%-60%); accelerated RT, 53% (95% CI, 37%-66%); and CRT, 57% (95% CI, 54%-60%) (Figure 1D). The IPTW Kaplan-Meier analysis of daily RT compared with accelerated RT revealed median survival of 31.5 months (95% CI, 25.3-41.2 months) and 51.9 months (95% CI, 30.2-75.6 months), respectively (Figure 2).
We evaluated the OS among 2622 patients with cT3N0M0 laryngeal squamous cell carcinoma treated with curative intent using surgical and RT-based approaches. Overall survival at 5 years was approximately 53% across all treatment types. Although we found no differences in OS on Kaplan-Meier analyses, daily fractionated RT was shown to be associated on multivariate analysis with inferior OS, unlike accelerated RT. This finding is consistent with randomized clinical trial data comparing RT schedules consisting of 5 fractions per week vs accelerated 6 fractions per week in laryngeal cancer and supports consideration for treating patients who select RT-based laryngeal preservation approaches but are not eligible for concurrent CRT with accelerated RT fractionation regimens.13
Randomized clinical trial data are available to support the use of RT-based laryngeal preservation approaches for patients with advanced laryngeal squamous cell carcinoma.1,2 To ensure the highest probability of success, patients should have low-volume T4a disease or less as defined by the RTOG 91-11 trial and receive concurrent CRT. After the widespread adoption of RT-based laryngeal preservation approaches, however, retrospective population-based studies5,6,14,15 noted a decrease in OS among patients with advanced laryngeal cancer, specifically patients with T4a disease. These data prompted debate regarding the universal applicability and effectiveness of RT-based approaches for patients with advanced laryngeal cancer.
A recent Surveillance, Epidemiology, and End Results (SEER) registry data analysis10 compared outcomes in 451 patients with T3 glottic laryngeal carcinomas treated with surgical vs nonsurgical modalities. After controlling for covariates significantly associated with survival, nonsurgical therapy was associated with inferior outcomes compared with surgery alone (adjusted HR, 0.68; 95% CI, 0.49-0.94) and compared with surgery with adjuvant therapy (adjusted HR, 0.75; 95% CI, 0.57-0.98). These data lead to the conclusion that patients with T3 glottic laryngeal carcinoma achieve a significant and clinically meaningful improvement in OS with surgery. This interpretation, however, is tempered by notable limitations of the data available in the SEER database. The nonsurgical arm combined an undefined percentage of patients receiving chemotherapy, RT, or CRT. In particular, RT was dichotomized as none or external beam RT, which raises the possibility of inclusion of patients receiving noncurative, palliative RT doses—a reasonable assumption given the median age of 75 years in that cohort. Furthermore, despite limited clinical trial data, chemotherapy is not considered to be a curative therapy in locally advanced laryngeal cancer but was included in the nonsurgical cohort.16 Several smaller studies14,15 have contributed data supporting a theme of inferior OS outcomes among patients with T3 glottic laryngeal cancer treated with primary RT-based approaches. Herein, we report data supporting similar OS among patients undergoing definitive surgical or RT-based approaches. Furthermore, in comparing this analysis with the preceding NCDB series of locally advanced laryngeal cancers by Hoffman et al,6 the survival outcomes among patients with T3N0M0 glottic laryngeal carcinomas were similar, although not directly comparable. This difference may be attributable to our series reporting 5-year OS, whereas the previous series reported 5-year relative survival. In addition, unlike previous reports, we used IPTW to adjust for confounders between treatment groups in survival analyses.
Despite the nonstatistically significant difference in OS, we found clinically meaningful differences in outcomes between patients receiving surgery with adjuvant CRT and the other cohorts that warrant discussion. To prevent staging biases between the surgical groups that had pathologic data and the RT cohorts that did not, all patients underwent clinical staging for analysis. As a result, the cohort receiving surgery with adjuvant CRT was enriched with patients with unfavorable pathologic factors, including positive surgical margins17 and regional nodal disease,18 that likely contributed to the difference in survival rates. For example, 23 patients in that cohort (32.4%) had positive surgical margins and 16 (40.0%) had nodal metastases, whereas patients receiving surgery and adjuvant RT included only 15 (9.1%) with positive surgical margins and 40 (33.0%) with nodal metastases.
Single-institution studies have reported that increased tumor volume, a source of variability within T3N0 laryngeal cancers, has an adverse effect on tumor control and survival outcomes in patients treated with primary RT.19 In this study, tumor size defined by the largest single dimension was not significantly associated with OS in the multivariate analysis, suggesting that tumor size is not a strong surrogate for tumor volume or that, after controlling for other factors such as treatment approach and patient comorbidities, the influence of tumor volume is diminished.
Several significant limitations to this study are inherent to the nature of retrospective data available in the NCDB. One major limitation is the lack of outcomes data. Despite recording the OS, NCDB does not include disease-free and cancer-free survival and does not include data regarding local, regional, or distant disease control. Furthermore, the percentage of patients undergoing RT-based laryngeal preservation approaches that required salvage laryngectomy is not reported. As such, our analysis does not report on laryngeal preservation but rather on OS among patients undergoing surgical or RT-based approaches. Also, long-term toxic effects of treatment and quality of life are not reported, which are important factors to consider when counseling patients about treatment options. In addition, treatment selection bias has a powerful effect on comparative clinical results. Although certain factors are considered to be objective and measurable, such as patient age, other factors are less objective with high interobserver variability, such as patient fitness and motivation for treatment and objective assessment of tumor resectability. Finally, smoking status is not recorded in the NCDB, which has been shown to be associated with survival among patients with laryngeal cancers.20 These issues underscore the importance of caution in reaching conclusions regarding comparisons of outcomes for nonrandomized patient cohorts who are selected to undergo distinct treatment approaches.
Counseling patients with advanced laryngeal cancer regarding treatment approaches is complex and has broad implications regarding quality of life and survival outcome. Randomized data support the use of RT-based laryngeal preservation approaches in appropriately selected patients. However, population-based studies have questioned the applicability of this approach in certain subgroups of patients, including those with T3N0 glottic laryngeal carcinomas.10 We queried the NCDB and identified 2622 patients treated definitively with surgery or RT-based laryngeal preservation approaches and observed equivalent OS among all compared groups, corroborating randomized clinical trial data.
Corresponding Author: Matthew E. Witek, MD, Department of Human Oncology, Carbone Comprehensive Cancer Center, University of Wisconsin School of Medicine and Public Health, 600 Highland Ave, K4/B100-0600, Madison, WI 53792 (email@example.com).
Accepted for Publication: July 15, 2017.
Published Online: October 12, 2017. doi:10.1001/jamaoto.2017.1756
Author Contributions: Drs Ko and Witek had full access to all of 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: Ko, Harari, Wieland, Kimple, Witek.
Acquisition, analysis, or interpretation of data: Ko, Chen, Yu, Baschnagel, Kimple, Witek.
Drafting of the manuscript: Ko, Chen, Baschnagel, Witek.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Ko, Chen, Yu.
Administrative, technical, or material support: Harari, Baschnagel.
Study supervision: Harari, Wieland, Baschnagel, Kimple, Witek.
Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest and none were reported.
Funding/Support: This study was supported in part by Comprehensive Cancer Center grant P30 CA014520-UW from the National Cancer Institute, National Institutes of Health (NIH) and by grant P50DE026787 from the Wisconsin Head and Neck Specialized Programs of Research Excellence.
Role of the Funder/Sponsor: The sponsors had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Disclaimer: The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.