Total Laryngectomy Versus Larynx Preservation for T4a Larynx Cancer: Patterns of Care and Survival Outcomes

CME

International Journal of

Radiation Oncology biology

physics

www.redjournal.org

Clinical Investigation

Total Laryngectomy Versus Larynx Preservation for T4a Larynx Cancer: Patterns of Care and Survival Outcomes Surbhi Grover, MD, MPH,* Samuel Swisher-McClure, MD, MSHP,* Nandita Mitra, PhD,y Jiaqi Li, BS,y Roger B. Cohen, MD,z Peter H. Ahn, MD,* John N. Lukens, MD,* Ara A. Chalian, MD,x Gregory S. Weinstein, MD,x Bert W. O’Malley Jr, MD,x and Alexander Lin, MD* Departments of *Radiation Oncology, yBiostatistics and Epidemiology, zHematology Oncology, and x Otorhinolaryngology-Head and Neck Surgery, University of Pennsylvania, Philadelphia, Pennsylvania Received Oct 18, 2014, and in revised form Feb 18, 2015. Accepted for publication Mar 3, 2015.

Summary Although guidelines recommend total laryngectomy as initial management for T4a larynx cancer, we report that a majority of patients (64%) with T4a disease undergo larynx-preservation chemoradiation. Predictors of receiving total laryngectomy were lower burden of nodal disease and treatment at a high case-volume facility. Compared with patients initially treated with total laryngectomy, those undergoing organ preservation had shorter medial overall survival.

Purpose: To examine practice patterns and compare survival outcomes between total laryngectomy (TL) and larynx preservation chemoradiation (LP-CRT) in the setting of T4a larynx cancer, using a large national cancer registry. Methods and Materials: Using the National Cancer Database, we identified 969 patients from 2003 to 2006 with T4a squamous cell larynx cancer receiving definitive treatment with either initial TL plus adjuvant therapy or LP-CRT. Univariate and multivariable logistic regression were used to assess predictors of undergoing surgery. Survival outcomes were compared using Kaplan-Meier and propensity scoreeadjusted and inverse probability of treatmenteweighted Cox proportional hazards methods. Sensitivity analyses were performed to account for unmeasured confounders. Results: A total of 616 patients (64%) received LP-CRT, and 353 (36%) received TL. On multivariable logistic regression, patients with advanced nodal disease were less likely to receive TL (N2 vs N0, 26.6% vs 43.4%, odds ratio [OR] 0.52, 95% confidence interval [CI] 0.37-0.73; N3 vs N0, 19.1% vs 43.4%, OR 0.23, 95% CI 0.070.77), whereas patients treated in high case-volume facilities were more likely to receive TL (46.1% vs 31.5%, OR 1.78, 95% CI 1.27-2.48). Median survival for TL versus LP was 61 versus 39 months (P<.001). After controlling for potential confounders, LP-CRT had inferior overall survival compared with TL (hazard ratio 1.31, 95% CI 1.10-1.57), and with the inverse probability of treatmenteweighted

Reprint requests to: Alexander Lin, MD, Department of Radiation Oncology, Perelman School of Medicine, University of Pennsylvania, 3400 Civic Center Boulevard, TRC 2 W, Philadelphia, PA 19104. Tel: (215) 662-3198; E-mail: [email protected] NotedAn online CME test for this article can be taken at http:// astro.org/MOC. Int J Radiation Oncol Biol Phys, Vol. 92, No. 3, pp. 594e601, 2015 0360-3016/$ - see front matter Ó 2015 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.ijrobp.2015.03.004

S.G. and S.S.-M. contributed equally to this work. Conflict of interest: none. Supplementary material for this article can www.redjournal.org.

be

found

at

Volume 92  Number 3  2015

Patterns of care and survival for T4a larynx cancer

595

model (hazard ratio 1.25, 95% CI 1.05-1.49). This survival difference was shown to be robust on additional sensitivity analyses. Conclusions: Most patients with T4a larynx cancer receive LP-CRT, despite guidelines suggesting TL as the preferred initial approach. Patients receiving LP-CRT had more advanced nodal disease and worse overall survival. Previous studies of (non-T4a) locally advanced larynx cancer showing no difference in survival between LP-CRT and TL may not apply to T4a disease, and patients should be counseled accordingly. Ó 2015 Elsevier Inc. All rights reserved.

Introduction

Methods and Materials

In 2014 the estimated incidence of larynx cancer in the United States will be 12,630, with 3610 deaths (1). The Department of Veterans Affairs (VA) Laryngeal Cancer Study established larynx preservation (LP) as a viable alternative to total laryngectomy (TL) for those with locally advanced disease (2). Radiation Therapy Oncology Group (RTOG) protocol 91-11 demonstrated that concurrent cisplatin/radiation (LP-CRT) achieved higher rates of LP than radiation therapy (RT) alone or sequential CRT, resulting in concurrent chemoradiation being adapted as the dominant form of LP (3, 4). The optimal treatment for advanced-stage larynx cancer remains unclear. The randomized trials infrequently included patients with T4a disease, comprising 26% of patients on the VA study (2) and 10% on RTOG 91-11 (in which patients with tumor penetrating through the thyroid cartilage or >1 cm into the base of tongue were ineligible) (3). In the VA study, >50% of those with T4a disease receiving LP ultimately needed salvage laryngectomy. Guideline recommendations for T4a larynx cancer specify TL as the initial treatment of choice (5, 6). However, some advocate that well-selected, low-volume T4a disease may be appropriate for LP, given good pretreatment larynx function and limited thyroid cartilage destruction (7). Concerns that LP is being offered too broadly and inappropriately have been fueled by evidence of declining survival, concurrent with the rise of LP and the declining use of TL (8). Furthermore, it was reported that patients with stage IV disease had improved overall survival with TL compared with LP (9). A population-based analysis in Canada revealed that TL resulted in better survival among T4a patients relative to LP (with 37% undergoing LP) (10). It is unclear how often T4a disease is treated with TL versus CRT in a non-study, practice-based setting in the United States. Our study aimed to: (1) determine what proportion of T4a patients receive the recommended standard treatment (TL); (2) report disease outcomes associated with TL versus CRT; and (3) identify factors that account for differences in treatment and outcomes. Our results will have implications on patient counseling for treatment, with the aim of changing practice patterns to comply with current treatment paradigms/guidelines.

Study design and data source This was a retrospective, observational, cohort study conducted with data from the National Cancer Data Base (NCDB). The NCDB is a large national cancer registry sponsored by the American College of Surgeons Commission on Cancer (COC) and the American Cancer Society. Approximately 70% of incident cancer cases in the United States are reported to the NCDB (11). The study was reviewed and determined to have exempt status by our institutional review board.

Cohort The study cohort is illustrated in Figure 1. We identified 6267 patients with pathologically confirmed nonmetastatic clinical stage T4a (American Joint Committee on Cancer staging manual [AJCC], 6th edition) invasive squamous cell carcinoma of the larynx. We excluded patients with prior malignancy, documented metastatic (M1) disease, or incomplete staging. We restricted our analysis to patients diagnosed between 2003 and 2006, because at the time of our study survival data were only available in NCDB for cases before 2007. Patients diagnosed before 2003 were staged with the AJCC 5th edition that did not distinguish T4a versus T4b. We defined treatment groups as either primary LP or primary surgery. Larynx preservation was defined as combination chemotherapy and radiation delivered either concurrently or sequentially because long-term survival outcomes are similar with either treatment approach (4). The NCDB records data regarding the first course of treatment, defined as all methods of treatment administered to the patient before either disease progression or recurrence (11). Because chemotherapy improves laryngectomyfree survival with LP (3, 4), and adjuvant RT is recommended postoperatively (2, 12-14), we excluded 1229 patients who did not receive care consistent within these standards, excluding those receiving RT alone, surgery alone, or surgery and chemotherapy without RT. We excluded 1049 patients who received only palliative therapy. We also excluded 205 patients who did not receive

596

International Journal of Radiation Oncology  Biology  Physics

Grover et al.

Fig. 1.

Definition of the study cohort.

treatment within 100 days of diagnosis, 52 patients with interrupted RT courses lasting >75 days, 148 patients receiving <50 Gy as part of LP-CRT, or <40 Gy postoperatively, and 138 patients with missing data related to the timing of RT completion or the total dose received. The final cohort consisted of 969 patients receiving either TL plus adjuvant RT/CRT (nZ353) or LP-CRT (nZ616).

Primary outcome Primary outcome was overall survival defined from the date of diagnosis to the date of death or last follow-up. Tumor registrars report patient follow-up to the NCDB annually, and COC accreditation standards require an annual 90% follow-up rate for all living analytic patients. Causespecific survival is not recorded in the NCDB.

Independent variables Patient and clinical characteristics included age, gender, clinical nodal stage (AJCC, 6th edition), laryngeal subsite (supraglottis, glottis, subglottis, or overlapping/not otherwise specified), comorbidities, treatment facility type (classified according to COC accreditation status), and treatment facility case volume (defined as average annual volume of larynx cancer patients of any stage receiving treatment at each facility). Patient comorbidities are reported as a truncated Charlson/Deyo score of 0, 1, or 2 because of the small number of cases with a score >2. Facilities were divided into quartiles of case volume, with only those within the highest quartile considered high volume, with the remainder considered low volume. Demographic characteristics included race and insurance status (insured, uninsured, or unknown).

Statistical analysis We used c2 tests and 2-sample t tests to compare the patient, clinical, and demographic characteristics of the treatment groups. Factors significantly associated with treatment on univariate analysis were included in a multivariable logistic regression model to further examine factors independently associated with receipt of TL. We used Kaplan-Meier methods to compare unadjusted survival probability estimates by treatment. Follow-up was calculated as the time from date of diagnosis until either date of death or last follow-up. We used propensity scoreeweighted Cox proportional hazards models to compare survival outcomes between patients receiving primary TL versus LP-CRT. We estimated the probability of receiving LP-CRT (the propensity score) for each patient according to relevant observed covariates using multivariable logistic regression. Covariates that were statistically significantly associated with treatment on univariate analysis or factors deemed to be clinically significant (patient age) were included in the logistic model. We evaluated the distribution of propensity scores for each treatment group and confirmed sufficient overlap in the distributions to ensure that the groups were comparable. We then grouped patients into quintiles according to their estimated propensity scores and used the CochraneManteleHaenszel test and linear regression to verify that measured covariates were balanced across all strata (Table 1). Our final Cox model incorporated inverse probability of treatment weights (IPTW) based on the propensity score to account for confounding due to measured covariates. For survival outcomes, this method has been shown to produce less-biased treatment effect estimates than stratification or covariate adjustment based

Volume 92  Number 3  2015 Table 1

Patterns of care and survival for T4a larynx cancer

597

Characteristics of the study population P

Characteristic All patients Age (y), mean (SD) Race White Black Hispanic Other/unknown Gender Male Female Year of diagnosis 2003 2004 2005 2006 Treatment facility type Community/other Community cancer Academic Facility case volume Low-volume center High-volume center Insurance status Not insured Insured Unknown Clinical nodal stage N0 N1 N2 N3 Unknown Charlson/Deyo score 0 1 >1 Primary site Glottis Supraglottis Subglottis Subsite overlapping/NOS Radiation dose received (Gy) <60 60-69.9 70

Larynx preservation therapy

Primary surgery plus adjuvant therapy

616 (63.6) 59.2 (10.4)

353 (36.4) 60.0 (9.4)

492 114 7 3

266 80 5 2

(79.9) (18.5) (1.1) (0.5)

Before adjustment

After propensity adjustment

.19 .44

.44

(75.4) (22.7) (1.4) (0.57) .40

487 (79.1) 129 (20.9) 99 160 160 197

(16.1) (26.0) (26.0) (32.0)

287 (81.3) 66 (18.7) 97 94 89 73

<.001

.92

<.001

.80

<.001

.56

(27.5) (26.6) (25.2) (20.7)

71 (11.5) 305 (49.5) 240 (39.0)

32 (9.1) 129 (36.5) 192 (54.4)

438 (71.0) 178 (28.9)

201 (56.9) 152 (43.1)

77 (12.5) 528 (85.7) 11 (1.8)

30 (8.5) 317 (89.8) 6 (1.7)

266 100 224 17 9

204 59 81 4 5

.16

(43.2) (16.2) (36.4) (2.8) (1.5)

454 (73.7) 134 (21.8) 28 (4.6) 20.45 53.9 3.42 22.24

(126) (332) (21) (137)

13.5 (83) 19.6 (121) 66.9 (412)

<.001

.96

.007

.96

<.001

.97

<.001

N/A

(57.8) (16.7) (23.0) (1.13) (1.42)

226 (64.0) 105 (29.8) 22 (6.2) 24.8 42.8 3.12 21.8

(114) (151) (32) (77)

33.1 (117) 52.7 (186) 14.2 (50)

Abbreviation: NOS Z not otherwise specified. Values are number (percentage) unless otherwise noted.

on the propensity score (15). Notably, in this case, covariate adjustment based on the propensity score yielded nearly identical results. We undertook a sensitivity analysis to assess the potential effects of an unknown or unmeasured confounder on hazard ratio (HR) estimates of the association between treatment and survival (16). We recalculated HRs and 95%

confidence intervals (CIs) for the association of treatment with overall survival under hypothesized conditions, varying both the prevalence of the unmeasured confounder as well as the relative hazard of death associated with the confounder. Additional subset analysis was performed limited to those receiving full-dose RT (70 Gy for LPCRT [nZ412] and 60 Gy after TL [nZ216]).

598

Grover et al.

We also performed a landmark analysis as an additional sensitivity analysis to assess for potential immortal time bias that might be present in the surgery cohort. Because we limited the study cohort to patients receiving surgery and at least adjuvant RT, there is a risk of immortal time bias if patients suffering acute perioperative death do not receive adjuvant therapy and are therefore excluded. The potential influence of immortal time bias and the use of landmark analyses in this clinical setting have been previously well described (17). Consistent with prior studies, we selected a landmark time frame of 3 months, expecting this to adequately account for any potential immortal time bias between surgery and adjuvant RT (18, 19). This resulted in 19 additional patients excluded. We repeated all survival analyses in this cohort, and our results were similar and remained statistically significant. Survival analyses were performed using Stata Version 12 (StataCorp, College Station, TX). Propensity score analyses and sensitivity analyses were performed using R version 2.15.1 (R Foundation for Statistical Computing, Vienna, Austria). Statistical significance was set at a Z 0.05, with all tests 2-tailed.

Results Patient characteristics Patient characteristics are shown in Table 1. Factors predictive of TL from multivariable logistic regression are shown in Table 2. Sixty-four percent of patients received LP-CRT versus TL (36%). Patients with advanced nodal disease (N2 vs N0: 26.6% vs 43.4%; OR 0.52, 95% CI 0.37-0.73, P<.001) and supraglottic (vs glottis) location (31.3% vs 47.5%; OR 0.51, 95% CI 0.36-0.72, P<.001) were less likely to undergo TL. Patients treated at high case-volume facilities were more likely to undergo TL (46.1% vs 31.5%; OR 1.78, 95% CI 1.27-2.48, P<.001). There was a significant decrease in TL during the years of the study (2003 vs 2006: 49.5% vs 27.0%; OR 0.36, 95% CI 0.24-0.54, P<.001). Neither insurance status nor treatment at an academic facility were significant predictors of TL.

Survival analysis Unadjusted Kaplan-Meier estimates of overall survival are shown in Figure 2. Median overall survival was 61 months among patients receiving TL and 39 months for patients receiving LP-CRT (P<.001). In a Cox proportional hazards model adjusted for age, year of diagnosis, treatment facility characteristics, nodal stage, larynx subsite, and comorbidity index, LP-CRT was associated with a statistically significant increase in the hazard of death (Table 3; HR 1.31, 95% CI 1.10-1.57, PZ.003). A second Cox proportional hazards model that included IPTW found similar results (HR 1.25, 95% CI

International Journal of Radiation Oncology  Biology  Physics Table 2 Predictors of total laryngectomy (TL) use in patients with T4a larynx cancer Proportion of patients receiving Unadjusted OR Characteristic TL (%) (95% CI) Age Year of diagnosis 2003 49.5 2004 37.0 2005 35.7 2006 27.0 Clinical N stage N0 43.4 N1 37.1 N2 26.6 N3 19.1 Unknown 35.7 Charlson/Deyo score 0 33.2 1 43.9 >1 44.0 Treatment facility Community/ 31.1 other Community 29.7 cancer Academic 44.4 Facility case volume Low-volume 31.5 center 46.1 Highvolume center Laryngeal subsite Glottis 47.5 Supraglottis 31.3 Subglottis 34.4 Overlapping/ 36.0 NOS Insurance status Not insured 28.0 Insured 37.5 Unknown 35.3

Adjusted OR (95% CI)

1.01 (1.00-1.02) 1.00 (0.98-1.01) 0.60 (0.41-0.88)* 0.60 (0.40-0.89)* 0.57 (0.39-0.83)* 0.50 (0.33-0.75)* 0.38 (0.26-0.56)y 0.36 (0.24-0.54)y

0.77 0.47 0.31 0.72

(0.53-1.11) (0.34-0.64)y (0.10-0.93)* (0.24-2.19)

0.85 0.52 0.23 0.60

(0.57-1.26) (0.37-0.73)y (0.07-0.77)* (0.18-1.96)

1.57 (1.16-2.13)* 1.82 (1.31-2.52)y 1.58 (0.88-2.82) 1.51 (0.80-2.84) -

-

0.94 (0.59-1.49) 0.86 (0.53-1.41) 1.78 (1.12-2.81)* 1.55 (0.92-2.61) -

-

1.86 (1.42-2.45)y 1.78 (1.27-2.48)y

0.50 (0.37-0.69)y 0.51 (0.36-0.72)y 0.58 (0.27-1.25) 0.66 (0.29-1.49) 0.62 (0.43-0.91)* 0.66 (0.44-0.98)*

1.54 (0.99-2.40) 1.51 (0.93-2.45) 1.40 (0.48-4.12) 1.34 (0.42-4.31)

Abbreviations: CI Z confidence interval; OR Z odds ratio. Other abbreviation as in Table 1. * P<.05. y P<.001.

1.05-1.49, PZ.01). Subset analysis of those receiving fulldose RT for LP-CRT (70 Gy, nZ412) or after TL (60 Gy, nZ236) also showed worse survival with LPCRT (HR 1.30, 95% CI 1.05-1.62, PZ.02). Landmark analysis using a propensity scoreeadjusted Cox proportional hazards model and IPTW continued to find a statistically significant association between LP and the risk of death (Table E1 [available online at www.redjournal.org]; HR 1.27, 95% CI 1.06-1.52, PZ.01 and HR 1.21, 95% CI 1.01-1.44, PZ.04).

Volume 92  Number 3  2015

Patterns of care and survival for T4a larynx cancer

1.00 0.25 0.50 0.75 0.00

Survival Probability

Kaplan-Meier survival estimates

0 Number at risk Larynx Preservation 616 Total Laryngectomy 353

2

4

6

8

10

139 118

30 32

0 0

Time (Years) 349 245

236 180

Larynx Preservation Therapy Primary Surgery + Adjuvant Therapy

Fig. 2. Overall survival for larynx preservation and primary surgery plus adjuvant therapy.

Sensitivity analysis is shown in Table 4. Baseline HR for this analysis was from the IPTW model. We hypothesized that functional status is a potential unmeasured confounder because functional status is not recorded in the NCDB, and arguably poor functional status could be associated with inferior survival. We varied the prevalence of poor functional status in the LP-CRT group and the strength of confounding to understand the extent of imbalance between treatment groups on this unmeasured confounder that would be necessary to affect the statistical significance of our results. For example, if the unmeasured variable was a moderate confounder (HR 1.5), a 4-fold difference in the prevalence of the unmeasured confounder (ie 80% vs 20%) would render our results not statistically significant. If the unmeasured variable was a strong confounder (HR 2.0), a 3-fold difference in the prevalence of the unmeasured confounder (60% vs 20%) would erase the statistical significance of our results. The results of our sensitivity analysis indicated that the observed results are robust and that any unmeasured confounder would have to be associated with markedly differential prevalence and survival to affect the significance of our results.

Discussion Although the efficacy and safety of organ preservation in the treatment of advanced larynx cancer has been well established (2-4), previous studies suggested worse survival with organ preservation compared with TL for stage IV larynx cancer (9, 20). Laryngeal cancer comprises a wide spectrum of tumor and nodal stage, and patient selection is important, as reflected in the guidelines supporting initial TL for patients with T4a disease (5, 6). As recently as 1998 the majority of patients with stage IV larynx cancer were treated with TL (9), but recent studies have shown increasing usage rates of RT in place of surgery, with no apparent impact on survival (21). However, the patterns of

599

care and outcomes specifically for treating T4a cancer in the modern era in the United States have not been well described. Our study is the first to do so, revealing that most patients with T4a disease (64%) are being treated with initial LP-CRT, with significantly worse survival than those treated with initial TL. On multivariable analysis, patients with advanced nodal disease and supraglottis subsite were less likely to receive TL, whereas those treated in facilities with high case volume were more likely to receive TL. Patients receiving TL experienced significantly greater median survival than those Table 3 analysis

Multivariable and propensity-adjusted survival

Characteristic

Adjusted HR (95% CI)

P

Multivariable Cox proportional hazards model Treatment with larynx 1.36 (1.14-1.63) <.001 preservation therapy Age (continuous) 1.03 (1.02-1.03) <.001 Year of diagnosis .23 2003 (ref) 2004 1.14 (0.90-1.44) 2005 0.98 (0.77-1.25) 2006 1.21 (0.94-1.54) Treatment facility type .98 Community/other (ref) Community cancer 0.97 (0.73-1.29) Academic 0.98 (0.73-1.33) Facility case volume .99 Low volume (ref) High volume 1.00 Clinical N stage .004 N0 (ref) N1 1.02 (0.81-1.30) N2 1.40 (1.15-1.69) N3 0.90 (0.49-1.64) Unknown 0.61 (0.37-1.37) Charlson/Deyo score .05 0 (ref) 1 1.06 (0.87-1.28) >1 1.53 (1.09-2.15) Primary site .13 Glottis (ref) Supraglottis 1.24 (1.01-1.54) Subglottis 1.24 (0.76-2.01) Overlapping/NOS 1.27 (0.99-1.62) Propensity scoreeadjusted Cox proportional hazards model Treatment with larynx 1.31 (1.10-1.57) .003 preservation therapy IPTW Cox proportional hazards model Treatment with larynx 1.25 (1.05-1.49) .01 preservation therapy Subset analysis of patients receiving full-dose radiation IPTW Cox proportional hazards model Treatment with larynx 1.30 (1.05-1.62) .02 preservation therapy Abbreviations: HR Z hazard ratio; IPTW Z inverse probability of treatmenteweighted. Other abbreviations as in Tables 1 and 2.

600

International Journal of Radiation Oncology  Biology  Physics

Grover et al.

Table 4 Sensitivity analysis of the effects of an unmeasured confounder (UC) on the HR of death after larynx preservation therapy Prevalence of UC in larynx preservation therapy group 0.4 0.4* 0.4* 0.6 0.6* 0.6* 0.8 0.8* 0.8*

Prevalence of UC in surgery group

UC HR

0.2 0.2* 0.2* 0.2 0.2* 0.2* 0.2 0.2* 0.2*

1.1 1.5* 2* 1.1 1.5* 2* 1.1 1.5* 2*

Treatment HR adjusted for UC (95% CI) 1.26 1.18 1.10 1.24 1.09 0.97 1.22 1.01 0.86

(1.06-1.50) (0.99-1.40)* (0.93-1.31)* (1.04-1.47) (0.92-1.29)* (0.81-1.15)* (1.23-1.44) (0.85-1.20)* (0.72-1.02)*

Abbreviations as in Tables 2 and 3. * Scenarios in which the effects of a UC would render the association of treatment with survival no longer statistically significant.

receiving LP-CRT. These associations remained consistent in multiple additional analyses that adjusted for baseline differences between the treatment groups, as well as in sensitivity analysis and landmark analysis. Additional subset analysis limited to patients receiving the full prescribed RT dose in both the LP-CRT (70 Gy) and TL (60 Gy) cohorts again revealed superior overall survival with TL. Although this study does not definitively answer why this is so (locoregional failure vs complications of LP-CRT compromising survival), we believe that the survival advantage associated with TL is most likely due to high rates of locoregional failure that cannot be adequately salvaged with laryngectomy, given the suboptimal long-term cancer-specific outcomes seen in patients from RTOG 91-11 (which largely excluded T4a disease) (4). Our study findings reflect a real-world analysis that complements what we can learn from data generated in a randomized and well-controlled clinical trial. In situations where prospective randomized evidence is lacking, observational studies can help to answer important questions regarding comparative effectiveness. Given the relative paucity of randomized evidence to guide treatment decisions for patients with T4a larynx cancer, the data from our study provide important information regarding both patterns of care and the comparative effectiveness of the 2 established treatment approaches for locally advanced larynx cancer (2-4). Our study brings to light several other important issues. Unexpectedly, we found that most patients with T4a disease were treated with organ preservation. Patients treated at high case-volume facilities were more likely to receive primary TL. It is possible that surgeons at these medical centers feel more comfortable recommending and performing major oncologic surgery, compared with surgeons in lower case-volume facilities. Patient access to highvolume surgeons may be limited, and patients may have strong preferences to pursue treatment options offered

closer to home. Patient preferences regarding these 2 divergent treatment approaches are also likely to have contributed to these findings. For example, it is unknown whether patients receiving LP-CRT were initially counseled to undergo TL but ultimately refused, or whether counseling by their treating physicians presented organ preservation as a viable, equivalent alternative despite the relative lack of evidence supporting its use specifically for T4a disease. If the latter is true, our results should serve as a cautionary tale in medical decision making and the potential risk of broadly applying results from prospective clinical trials to a population not included in these trials. Another question to consider is whether inferior outcomes with LP-CRT were observed in our study owing to differences in follow-up care and salvage treatment delivered. Up to half of patients with T4 larynx cancer undergoing LP-CRT will ultimately require salvage surgery, and failure to follow up and intervene early can ultimately lead to worse-thanexpected survival rates (as supported by the surgical analysis of salvage laryngectomy from the RTOG 91-11 trial) (22). The rapid separation of the survival curves between TL and LP patients suggests that many patients managed with initial LP may have had persistent disease at the end of therapy and succumbed rapidly to progressive disease. Therefore, one may question whether patients undergoing organ preservation had stringent and appropriate follow-up for detection of recurrence, and whether failure to intervene with timely salvage surgery could have ultimately affected their outcomes. Alternatively, we acknowledge that the rapid separation of the survival curves may alternatively be secondary to a difference in prognosis between the groups based on factors not obtained from the NCDB, despite the sensitivity analysis. However, we also performed a landmark analysis as an additional sensitivity analysis in an effort to assess for potential immortal time bias that might be present in the surgery cohort and found that the survival difference favoring those receiving TL persisted. The limitations of our study are largely due to the nature of the NCDB as a resource for studying cancer care and patient outcomes. First, our survival analysis was limited to comparisons of patient overall survival because cancerspecific survival and local recurrence data are not available in the NCDB. Second, we acknowledge other potential selection biases: (1) detailed patient-specific tumor characteristics, such as tumor volume, for which there is heterogeneity among patients and which has been shown to impact significantly on overall survival (23), are not available in the data source; and (2) there were likely patients who were medically inoperable, thereby obligating them to LP. This limits us from examining potential selection factors that could identify patients with T4a larynx cancer who may be more suitable candidates for LP-CRT. In addition, other aspects of patient care, such as treatment quality assurance and the adequacy of follow-up examinations (to intervene with surgical salvage in a timely fashion) are not available. Because of our sensitivity analyses, we believe the survival difference between the

Volume 92  Number 3  2015

2 treatments to be robust and that a large difference in an unmeasured confounder would have to be associated with markedly differential prevalence and survival to affect our results. However, we acknowledge that such a large difference may still exist because it is related to the therapy administered, and thus remains a limitation of our study. In summary, we find that in routine practice settings, outside the controlled environment of clinical trials, organ preservation chemoradiation is the most commonly used treatment approach for T4a larynx cancer and is associated with worse survival when compared with primary TL (the accepted standard of care). We are unable to determine from the data whether this reflects an inappropriate shift away from the “standard of care” or whether this is an appropriate use of LP-CRT due to factors and reasons not captured in the NCDB. Our finding of worse survival with LP-CRT reflect the findings by other comparative effectiveness studies (9, 20) but contrast with others showing no difference in survival with definitive RT (21). However, none of these previous studies were limited to patients with T4a disease, and this therefore highlights the risks of broadly applying results from studies that included but were not limited to our specific patient population. Given the complexity of care associated with treatment, we advocate that patients with advanced laryngeal cancer be seen, staged, and evaluated by a multidisciplinary team in a high-volume center, with standard and active follow-up care similar to that used on the clinical trials that form the basis of our current practice. We hope that our findings will help inform clinical practice and guide patiente physician discussions to ultimately allow for better informed decisions and patient outcomes.

References 1. Siegel R, Ma J, Zou Z, et al. Cancer statistics, 2014. CA Cancer J Clin 2014;64:9-29. 2. Induction chemotherapy plus radiation compared with surgery plus radiation in patients with advanced laryngeal cancer. The Department of Veterans Affairs Laryngeal Cancer Study Group. N Engl J Med 1991;324:1685-1690. 3. Forastiere AA, Goepfert H, Maor M, et al. Concurrent chemotherapy and radiotherapy for organ preservation in advanced laryngeal cancer. N Engl J Med 2003;349:2091-2098. 4. Forastiere AA, Zhang Q, Weber RS, et al. Long-term results of RTOG 91-11: A comparison of three nonsurgical treatment strategies to preserve the larynx in patients with locally advanced larynx cancer. J Clin Oncol 2013;31:845-852. 5. Forastiere AA, Ang KK, Brizel D, et al. Head and neck cancers. J Natl Compr Canc Netw 2008;6:646-695.

Patterns of care and survival for T4a larynx cancer

601

6. Pfister DG, Spencer S, Brizel DM, et al. Head and neck cancers, Version 2.2014. Clinical practice guidelines in oncology. J Natl Compr Canc Netw 2014;12:1454-1487. 7. Corry J, Peters L, Kleid S, et al. Larynx preservation for patients with locally advanced laryngeal cancer. J Clin Oncol 2013;31:840-844. 8. Hoffman HT, Porter K, Karnell LH, et al. Laryngeal cancer in the United States: Changes in demographics, patterns of care, and survival. Laryngoscope 2006;116:1-13. 9. Chen AY, Halpern M. Factors predictive of survival in advanced laryngeal cancer. Arch Otolaryngol Head Neck Surg 2007;133:12701276. 10. Dziegielewski PT, O’Connell DA, Klein M, et al. Primary total laryngectomy versus organ preservation for T3/T4a laryngeal cancer: A population-based analysis of survival. J Otolaryngol Head Neck Surg 2012;41(Suppl 1):S56-S64. 11. American College of Surgeons Commission on Cancer. National Cancer Data Base. Available at: www.facs.org/cancer/ncdb/index. html. Accessed January 28, 2014. 12. Bernier J, Cooper JS, Pajak TF, et al. Defining risk levels in locally advanced head and neck cancers: A comparative analysis of concurrent postoperative radiation plus chemotherapy trials of the EORTC (#22931) and RTOG (# 9501). Head Neck 2005;27:843-850. 13. Bernier J, Domenge C, Ozsahin M, et al. Postoperative irradiation with or without concomitant chemotherapy for locally advanced head and neck cancer. N Engl J Med 2004;350:1945-1952. 14. Cooper JS, Pajak TF, Forastiere AA, et al. Postoperative concurrent radiotherapy and chemotherapy for high-risk squamous-cell carcinoma of the head and neck. N Engl J Med 2004;350:1937-1944. 15. Austin PC. The performance of different propensity score methods for estimating marginal hazard ratios. Stat Med 2013;32:2837-2849. 16. Mitra N, Heitjan DF. Sensitivity of the hazard ratio to nonignorable treatment assignment in an observational study. Stat Med 2007;26: 1398-1414. 17. Park HS, Gross CP, Makarov DV, et al. Immortal time bias: A frequently unrecognized threat to validity in the evaluation of postoperative radiotherapy. Int J Radiat Oncol Biol Phys 2012;83:13651373. 18. Yu JB, Decker RH, Knisely JP. The role of postoperative radiation therapy (PORT) in the treatment of extrahepatic bile duct cancer: A Surveillance, Epidemiology, and End Results (SEER) populationbased investigation. J Gastrointest Cancer 2008;39:11-21. 19. Yu JB, Wilson LD, Dasgupta T, et al. Postmastectomy radiation therapy for lymph node-negative, locally advanced breast cancer after modified radical mastectomy: Analysis of the NCI Surveillance, Epidemiology, and End Results database. Cancer 2008;113: 38-47. 20. Megwalu UC, Sikora AG. Survival outcomes in advanced laryngeal cancer. JAMA Otolaryngol Head Neck Surg 2014;140:855-860. 21. Zhang H, Travis LB, Chen R, et al. Impact of radiotherapy on laryngeal cancer survival: A population-based study of 13,808 US patients. Cancer 2012;118:1276-1287. 22. Weber RS, Berkey BA, Forastiere A, et al. Outcome of salvage total laryngectomy following organ preservation therapy: The Radiation Therapy Oncology Group trial 91-11. Arch Otolaryngol Head Neck Surg 2003;129:44-49. 23. Hsin LJ, Fang TJ, Tsang NM, et al. Tumor volumetry as a prognostic factor in the management of T4a laryngeal cancer. Laryngoscope 2014;124:1134-1140.