Log-rank P = .01.
Log-rank P = .002.
George JR, Yom SS, Wang SJ. Combined Modality Treatment Outcomes for Head and Neck CancerComparison of Postoperative Radiation Therapy at Academic vs Nonacademic Medical Centers. JAMA Otolaryngol Head Neck Surg. 2013;139(11):1118-1126. doi:10.1001/jamaoto.2013.4539
Patients with head and neck squamous cell carcinoma (HNSCC) who undergo surgical resection in an academic medical center (AC) often receive postoperative adjuvant external beam radiation therapy (RT) at non-ACs closer to home. Few data exist to compare outcomes of these populations.
To evaluate treatment metrics and outcomes in patients with HNSCC who underwent surgical resection at an AC and then received postoperative adjuvant external beam RT at an AC vs a non-AC.
Design, Setting, and Participants
Retrospective cohort study in 1 AC and several community RT non-ACs of patients with primary HNSCC treated with surgery at an AC followed by adjuvant therapy at an AC or a non-AC from January 1, 2002, to January 1, 2012.
We evaluated for between-groups differences in demographics, RT metrics, and survival outcomes. Subgroup analysis by tumor site was then performed.
Main Outcomes and Measures
Overall survival, disease-specific survival, and locoregional control rates.
A total of 286 patients underwent surgery at the University of California, San Francisco, followed by adjuvant therapy. A total of 214 patients were analyzed. Significant differences in demographic and oncologic variables emerged, including important differences in RT metrics. Patients treated at a non-AC received a lower total RT dose, lower fractional dose, more delays in RT initiation, more breaks in RT, and more early termination of RT. Adjuvant treatment at an AC was associated with improved survival on univariate but not multivariate analysis. Subgroup analysis by SCC tumor site normalized many of the differences between groups, yet still revealed persistent differences in RT metrics. On multivariate analysis, AC treatment was not an independent predictor of survival for any tumor site.
Conclusions and Relevance
Better oncologic outcomes were seen in the AC group on univariate analysis, but these improved outcomes were not found on multivariate analysis. Important differences in RT metrics were noted for non-AC treatment sites compared with AC sites. Subgroup analysis by tumor site demonstrated persistent differences in treatment metrics. Standardization of adjuvant HNSCC treatment according to national guidelines should be prioritized at non-AC treatment facilities.
Head and neck squamous cell carcinoma (HNSCC) comprises approximately 3% to 5% of all cancers in the United States. This year, an estimated 52 610 people (38 380 men and 14 230 women) will develop HNSCCs, and an estimated 11 500 deaths (8320 men and 3180 women) will occur.1 Management of advanced HNSCC is multidisciplinary and usually requires combined modality treatment. Because of the availability of specialty services, many patients with HNSCC receive some aspect of their care at an academic medical center (AC).
Several studies2,3 have addressed the effect of surgical volume of the treating institution on cancer patients’ outcomes. These studies4- 6 have demonstrated superior outcomes for teaching or large-volume institutions compared with low-volume or community hospitals. Similar findings are seen in analyses of several cancer types, including ovarian cancer, gastrointestinal tract cancers, breast cancer, sarcoma, nonseminoma, and thoracic tumors.7- 17 The difference has been attributed to the prevalence of experienced surgeons and improved coordination of adjuvant therapy in ACs, among other contributing factors.15- 17
Limited studies18,19 of HNSCC have found improved outcomes for some patients treated at high-volume hospitals or specialized cancer centers compared with lower-volume treatment facilities. To date, the effect of AC treatment has not been studied specifically for adjuvant therapy for HNSCC. At our institution, a large proportion of patients who underwent head and neck cancer surgery then undergo postoperative treatment at a non-AC, usually in their local area. Because postoperative radiation therapy (RT) usually requires daily sessions for several weeks, many patients find that completing adjuvant RT and/or chemotherapy at the AC is infeasible because this effort requires available transportation, housing, and social support. However, non-ACs may lack expertise in management of the complex tumor types in patients who undergo operation at an AC.
To evaluate treatment metrics and outcomes in patients with HNSCC who underwent surgical resection at an AC and then received postoperative adjuvant external beam RT at an AC vs a non-AC. Our hypothesis was that patients receiving adjuvant therapy at an AC would have improved RT coordination and delivery and better cancer-specific outcomes compared with patients receiving adjuvant therapy at a non-AC.
A retrospective cohort study was performed of all patients with primary HNSCCs who underwent primary surgical resection at the University of California, San Francisco (UCSF), followed by postoperative adjuvant external beam RT with or without chemotherapy. The Committee on Human Research at the UCSF granted approval for this study. The time range of surgical dates was from January 1, 2002, to January 1, 2012.
We included all patients who presented with primary, untreated HNSCCs of the oral cavity, oropharynx, hypopharynx, and larynx. All stages were included. Patients included in this study did not receive any prior definitive surgery, RT, or chemotherapy before initial surgery at our institution. We excluded patients with a diagnosis of squamous carcinoma in situ, those with any history of head and neck irradiation, patients treated with preoperative or postoperative brachytherapy, and those who experienced local recurrence or distant metastasis before the initiation of adjuvant external beam RT. Nasopharyngeal, thyroid, salivary, sinonasal, and cutaneous carcinomas, including primary cutaneous squamous cell carcinoma, were excluded. We also excluded all patients with prolonged hospital stays (>4 weeks) after surgery due to wound breakdown and postoperative complications that led to delays in initiating RT.
All patients included in this study underwent combined modality therapy starting with primary tumor resection followed by adjuvant external beam RT or chemoradiation therapy. Oncologic resection was completed at the Division of Head and Neck Surgical Oncology in the Department of Otolaryngology–Head and Neck Surgery at the UCSF. External beam RT was performed either in the Department of Radiation Oncology at the UCSF (hereafter referred to as an AC) or at a community-based, nonacademic RT facility (hereafter referred to as a non-AC), with or without concurrent administration of chemotherapy. Administration of adjuvant chemotherapy was tabulated as a dichotomous (yes/no) variable only because chemotherapy treatment records initially collected were noted to be of universally poor reliability.
Data collection was performed using all electronic medical record systems in place at the UCSF from January 2002 to January 2012. Additional data were collected through communication with RT facilities and discussion with next of kin. More than 2500 patient records were screened for inclusion in this study using Current Procedural Terminology codes for head and neck cancer operations. Demographic data collected included information about sex, race, ethnicity, tobacco and alcohol use, insurance type, home zip code, and dates of treatment. Oncologic data were collected from the surgical pathology report and medical records. These included American Joint Committee on Cancer (AJCC) stage, TNM stage, HNSCC site and subsite, and detailed histopathologic data.
Adjuvant RT details were obtained for each included patient from the RT summary (RTS) provided by radiation oncology professionals at both the AC and non-AC treatment sites. These data included dates and type of treatment (intensity-modulated RT [IMRT] vs non-IMRT), total radiation dose, number of fractions, and dose per fraction. Information about treatment delays and their cause was also obtained. Total dose was recorded as the highest level of prescription dose delivered, universally cited as the “total dose” in the RTS. From the RTS, we also obtained information on RT breaks, early RT termination, and completion of the intended RT course. A break in RT was defined as a stoppage of RT lasting more than 4 days due to technical breakdown, patient illness, or treatment-associated toxic effects, as cited by the treating radiation oncologist. On the basis of the accepted literature, a delay in the initiation of RT was defined as an interval of greater than 56 days (8 weeks) from the time of postsurgical hospital discharge to the first day of adjuvant external beam RT.20
Outcome data collected included information about locoregional recurrence, distant metastasis, and death. Date of death was determined from the national Social Security Death Master File. Cause of death was obtained through medical records.
Missing data from the non-AC group were noted during the data collection process. Certain RT centers had closed by the time of this study, and data for patients treated at these centers could not be found. In other cases, RTSs from non-ACs could not be obtained despite multiple requests. Examination of the group of patients excluded because of missing data revealed a similar distribution of tumor sites and stages as those patients who were included for analysis in the non-AC group.
Descriptive analyses were performed on demographic, oncologic, histologic, and adjuvant treatment variables using Mann-Whitney, χ2, and Fisher exact tests for continuous, nominal, and dichotomous variables, respectively. Kaplan-Meier analyses were performed for overall survival (OS), disease-specific survival (DSS, survival for patients who died of HNSCCs only), and disease-free survival (DFS, time from treatment to locoregional recurrence or metastasis, whichever occurred sooner). These variables were all evaluated for statistical significance using Wilcoxon rank sum tests. Locoregional recurrence was evaluated using the Fisher exact test for a between-groups comparison. Univariate survival analyses were performed using Cox proportional hazards regression model to find associations of important variables with survival. Multivariate analyses were performed to determine the effect of AC treatment after adjustment for key variables identified from the univariate analysis.
Subgroup analysis by SCC tumor site was then performed to attempt to control for differences between AC and non-AC groups. We analyzed 3 subgroups: (1) oral cavity SCC, (2) oropharynx or unknown primary SCC, and (3) hypopharynx or larynx SCC. The aforementioned statistical measures were then repeated.
A total of 286 patients underwent surgery at the UCSF followed by RT with or without chemotherapy from January 2002 to January 2012. A total of 214 patients were included in this study: 123 patients in the non-AC treatment group and 91 patients in the AC treatment group. Seventy-two patients were excluded because of the inability to obtain RT information, all from the non-AC group. Demographic characteristics of this cohort are listed in Table 1. The mean (SD) age at treatment was significantly higher for the non-AC group than for the AC group (62.0 [12.6] years vs 57.8 [13.3] years; P = .02). Patients in the non-AC group lived farther from the AC than those in the AC treatment group (mean [SD], 95.6 [184.0] miles vs 46.9 [64.5] miles; P < .001). The non-AC treatment group had a higher Hispanic proportion than the AC group (13.9% vs 4.4%, P = .03).
Significant differences in substance abuse history existed between groups (Table 2). Patients treated at non-ACs were more likely to be former or active tobacco users than those treated at an AC (80.3% vs 55.6%, P = .001) and were more likely to be heavy tobacco users with greater than a 30 pack-year history (39.3% vs 21.1%, P = .008). No difference in heavy alcohol use was found between groups (P = .76).
Clinicopathologic characteristics for this cohort are listed in Table 3. More than 70% of the patients treated in this study had stage IV disease, but no difference was found in distributions of the AJCC group stages between the AC and non-AC groups (P = .44). Significant differences between treatment groups were noted for HNSCC site (P = .02). Patients receiving treatment at a non-AC were more likely to have laryngeal SCC, whereas those treated at an AC were more likely to have oropharyngeal SCC (Table 3). Tumor status data were significantly different as well (P = .002). Non-AC patients were more likely to have T4 tumors, whereas AC patients were more likely to have T1 to T2 tumors (P = .001). Nodal status, however, was not different between groups (P = .17).
Histopathologic characteristics for this cohort revealed interesting differences between treatment groups. Patients receiving adjuvant external beam RT at a non-AC were more likely to have tumor necrosis (P = .02), perineural invasion (P = .02), and vascular invasion (P = .007) noted on histopathologic analysis. No differences were noted between groups in histologic tumor grade or nodal extracapsular extension, but patients receiving treatment at non-ACs were somewhat more likely to have positive margins, although this finding was not significant (11.5% vs 6.6%, P = .07).
Several notable differences were found in adjuvant treatment characteristics between AC and non-AC groups (Table 4). Although there was no difference in patients receiving chemoradiotherapy vs RT from each group (P = .48), patients treated at non-AC facilities were less likely to be treated with IMRT (53% vs 94%; P < .001). The postoperative adjuvant external beam RT for the non-AC group consisted of a significantly lower mean (SD) total RT dose than the AC group (57.1 [11.1] Gy vs 62.4 [53.8] Gy; P < .001) and a lower dose per fraction (1.9 [0.2] vs 2.1 [0.3] Gy; P < .001). A higher proportion of patients in the non-AC group had delays in starting treatment than those in the AC group (23.0% vs 10.0%; P < .001), more breaks in RT (16.4% vs 0%; P < .001), and more early termination of RT (9.8% vs 1.1%; P < .001). Finally, patients in the non-AC group were less likely to receive the intended full course of RT (74.6% vs 98.9%; P < .001).
Survival outcomes are highlighted in Table 5. Significant differences were noted in OS and DSS between the non-AC and AC groups (P = .01; Figure 1 and Figure 2). Three-year OS was 64.2% for non-AC patients vs 80.5% for AC patients, whereas 5-year OS was 53.7% for non-AC patients vs 74.3% for AC patients. Five-year DSS rates for the non-AC and AC groups were 54.9% and 80.2% (P = .002). The DFS rates for the non-AC and AC groups were 69.4% and 82.4%, respectively (P = .05). The locoregional control (LRC) rate was significantly lower for those receiving non-AC adjuvant treatment vs AC (78.5% vs 91.1%; P = .04).
Univariate analyses revealed several variables to be significantly associated with survival (Table 6). Adjuvant treatment at an AC was significantly associated with improved survival (hazard ratio [HR], 0.53; 95% CI, 0.32-0.88; P = .02). Insurance status with Medicare (HR, 1.86; 95% CI, 1.09-3.2; P = .02) and Medicaid (Medi-Cal) (HR, 1.95; 95% CI, 1.05-3.6; P = .04) were associated with significantly poorer survival. Patients with a heavy alcohol history (>30 drinks per week) had poorer survival (HR, 1.78; 95% CI, 1.11-2.9; P = .02), as did patients with a history of heavy tobacco use (>30 pack-years) (HR, 2.0; 95% CI, 1.11-3.6; P = .02). Oropharyngeal tumors had a borderline association with improved survival (HR, 0.45; 95% CI, 0.19-1.05; P = .06), whereas T4 tumors and N2c nodal disease were associated with poorer survival compared with T1 and N0 disease, respectively (P = .07 and <.001, respectively). Use of IMRT conferred a survival benefit over use of non-IMRT (HR, 0.49; 95% CI, 0.29-0.81; P = .006), and receiving a full course of RT was associated with improved survival over receiving less than a full course of RT, although this association did not reach statistical significance (HR, 0.53; 95% CI, 0.25-1.10; P = .09). Finally, patients who had early RT termination had significantly worse survival (HR, 2.5; 95% CI, 1.20-5.3; P = .02).
Multivariate analyses were performed to determine the adjusted effect of AC treatment and to reveal whether demographic or clinicopathologic differences accounted for differences in survival noted between the groups. A selected multivariate survival analysis is given in Table 7. In this model, AC treatment was revealed to have a protective but not significant effect when the model was adjusted for sex, tumor stage, AJCC stage, and smoking history (model 1: HR, 0.75; 95% CI, 0.44-1.29; P = .30). The only variables consistently associated with poorer survival on multivariate analysis were heavy smoking history (model 1: HR, 1.8; 95% CI, 1.11-2.9; P = .02) and T stage (model not shown, P = .005-.02).
Significant demographic and clinicopathologic differences were noted between the AC and non-AC groups in the above analyses. Notably, there was a greater frequency of nonsmokers with oropharynx cancer in the AC group (43.4% vs 21.1% in the non-AC group) and a greater frequency of smokers with larynx cancer in the non-AC group (30.3% vs 17.0% in the AC group). Therefore, a subgroup analysis by SCC tumor site was performed. This eliminated many but not all of the significant demographic and clinicopathologic differences between the AC and non-AC groups (data not shown). However, greater T stage was still noted for patients with oral cavity SCC (P = .04), and ethnic differences were observed among non-AC patients with SCC of the hypopharynx or larynx (higher proportion of Hispanic patients in the non-AC group, P = .03).
The RT metrics remained markedly different between the AC and non-AC groups within the subgroup analysis. Non-AC patients with SCC of the oral cavity, oropharynx or unknown primary, and hypopharynx or larynx had lower total radiation doses (P < .001, P = .01, and P = .04, respectively) and lower fractional doses (P < .001, P < .001, and P = .004, respectively). However, multivariate analyses of these subgroups revealed that AC treatment was not independently associated with survival (data not shown).
This single-institution study is a 10-year analysis of differences in postoperative adjuvant treatment administered at an AC vs a non-AC. We evaluated differences in postoperative therapy received by patients treated at an AC and patients treated at a non-AC. We noted significant differences in RT characteristics between these 2 groups. The adjuvant external beam RT received by the non-AC group consisted of a significantly lower total RT dose, lower fractional dose, and less frequent use of IMRT compared with the non-AC group. Patients in the non-AC group also experienced significantly more delays in treatment, more treatment breaks, and more early RT termination.
Reasons for these findings are not immediately clear. Demographic and clinicopathologic differences were noted between the groups. Non-AC patients were older and had a greater tobacco and alcohol abuse burden. Patients receiving non-AC treatment had more laryngeal cancers, whereas those treated at the AC had more oropharyngeal cancers. These findings reveal broad demographic and clinicopathologic differences between patients receiving combined modality treatment at the AC and those receiving adjuvant treatment at non-ACs.
To adjust for the broad differences between the AC and non-AC groups, subgroup analysis by SCC tumor site was performed. The subgroup comparison eliminated most but not all of the demographic and clinicopathologic differences between the groups. However, significant differences in RT metrics persisted across all tumor sites.
Patients in the AC group had improved OS, DFS, and LRC rates compared with the non-AC group on univariate analysis. However, on multivariate analysis, treatment at an AC was not an independent predictor of survival, suggesting that factors other than RT center might have contributed to the observed differences in oncologic outcome. It is possible that differences in RT metrics, which remained significant on subgroup analysis, could have accounted for poorer oncologic outcomes observed in the non-AC group.
Prior literature has demonstrated improved head and neck cancer survival at ACs. Chen and colleagues21,22 found that treatment at high-volume hospitals resulted in improved survival for laryngeal cancer, whereas Akman and colleagues23 demonstrated improved LRC in laryngeal cancer patients treated at specialized cancer centers. In 2009, Cheung and colleagues24 demonstrated significantly improved survival at high-volume centers for low-, medium-, and high-grade HNSCCs. Finally, Kubicek et al25 demonstrated a trend for improved overall survival in patients who received definitive chemoradiotherapy at an AC vs a non-AC.
This study highlights that although survival outcomes are difficult to compare across heterogeneous populations, RT quality indicators can be measured and differences observed among RT centers. Academic medical centers may offer an advantage in timeliness and coordination of care and access to more HNSCC-specific support services. This may prevent early termination of the RT course, which was associated with survival in this study. Better adherence to national RT guidelines at the AC in this study may have led to a mean postoperative dose (62 Gy, all SCC sites) significantly higher than the mean postoperative RT dose of the non-AC group (57 Gy, all SCC sites). The mean AC dose, but not the mean non-AC dose, is in keeping with the 2012 National Comprehensive Cancer Network recommended postoperative RT dose of 60 to 66 Gy for an HNSCC (www.nccn.org/index.asp). Our analysis suggests the importance of coordinated multidisciplinary care to achieve successful completion of recommended treatment protocols for head and neck cancer.
It is a common scenario that transfer of care occurs from a surgical AC to a non-AC for postoperative care. To our knowledge, no other authors have exclusively studied outcome differences between ACs and non-ACs for patients with HNSCC treated with combined modality treatment consisting of primary surgery followed by postoperative adjuvant external beam RT. Lassig and colleagues26 compared RT at ACs vs non-ACs for both primary and adjuvant treatment of HNSCCs and found a significant survival advantage for the AC group. The improved survival could not be explained by differences in treatment characteristics because that study did not find differences in dose, fraction, rate of treatment delays, treatment breaks, or incomplete treatment course between the AC and non-AC groups.
Our study has several limitations. Because of the stringent inclusion criteria, our sample size was limited, decreasing our study’s overall statistical power. In addition, the 2 groups had significant demographic and clinicopathologic differences, which made comparisons of oncologic outcomes between the 2 groups difficult, as indicated by the results of our multivariate analysis. Another important limitation of this study was lack of availability of some outside treatment records. However, we believe the group of patients excluded because of lack of treatment records was likely similar to those included in the non-AC group. Another limitation stems from the comparison of one AC to multiple non-ACs. Evaluation of a single AC treatment institution limits the generalizability of our findings to other ACs. Multiple non-ACs were studied and were not individually evaluated for quality and survival outcomes. It is possible that some non-ACs had better outcomes than others; therefore, our findings cannot be applied to each treatment site but rather to non-AC treatment facilities generally. Our study did not address patient comorbidities, such as systemic illness, which could affect tolerance for full-course adjuvant chemoradiotherapy and therefore OS. All patients in this study underwent surgery, indicating that they had a satisfactory baseline level of physical health. All patients were judged to be in adequate condition to receive fractionated long-course external beam RT. Similarly, we did not collect data on socioeconomic status, which could have been another factor that affected outcomes.
In conclusion, adjuvant treatment at an AC was not an independent predictor of survival on multivariate analysis. However, significant differences were noted in the metrics of adjuvant RT provided at ACs compared with non-ACs. Our findings suggest that adherence to national guidelines for postoperative adjuvant external beam RT in head and neck cancer should be emphasized. The ACs should assess the quality of care received by their patients in all phases of treatment.
Submitted for Publication: October 23, 2012; final revision received June 22, 2013; accepted July 21, 2013.
Corresponding Author: Steven J. Wang, MD, Department of Otolaryngology–Head and Neck Surgery, University of California, San Francisco, 2233 Post St, UCSF PO Box 1225, San Francisco, CA 94115 (email@example.com).
Published Online: September 19, 2013. doi:10.1001/jamaoto.2013.4539.
Author Contributions: Dr George 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.
Study concept and design: All authors.
Acquisition of data: All authors.
Analysis and interpretation of data: All authors.
Drafting of the manuscript: George.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: George, Wang.
Administrative, technical, or material support: All authors.
Study supervision: All authors.
Conflict of Interest Disclosures: None reported.
Previous Presentation: This study was presented at the American Head and Neck Society 2013 Annual Meeting; April 11, 2013; Orlando, Florida.
Additional Contributions: Diane Thulin, BS, and Erin Shugard, MS, of the Departments of Otolaryngology–Head and Neck Surgery and Radiation Oncology at UCSF, respectively, provided assistance in record preparation. Emily R. Nelson, MD, School of Medicine at UCSF, provided data collection assistance.
Corrections: This article was corrected online November 22, 2013, for omission of a word in the title and a typographical error in the abstract and on November 25, 2013, for a second typographical error in the abstract.