Improved survival using intensity-modulated radiation therapy in head and neck cancers: A SEER-Medicare analysis

Authors

Errata

This article is corrected by:

  1. Errata: Erratum: Improved survival using intensity-modulated radiation therapy in head and neck cancers: A SEER-Medicare analysis Volume 120, Issue 11, 1754, Article first published online: 22 February 2014

  • This work has been accepted for an oral presentation at the 2013 Annual Meeting of the American Society for Radiation Oncology (ASTRO), September 22-25, 2013, Atlanta, Georgia.

  • We acknowledge that this study used the linked SEER-Medicare database. The interpretation and reporting of these data are the sole responsibility of the authors. The authors acknowledge the efforts of the Applied Research Program, National Cancer Institute; the Office of Research, Development and Information, Centers for Medicare and Medicaid Services; Information Management Services (IMS), Inc.; and the SEER Program tumor registries in the creation of the SEER-Medicare database.

Abstract

BACKGROUND

Intensity-modulated radiation therapy (IMRT) is a technologically advanced, and more expensive, method of delivering radiation therapy with a goal of minimizing toxicity. It has been widely adopted for head and neck cancers; however, its comparative impact on cancer control and survival remains unknown. The goal of this analysis was to compare the cause-specific survival (CSS) for patients with head and neck cancers treated with IMRT versus non-IMRT from 1999 to 2007.

METHODS

CSS was determined using the Surveillance, Epidemiology, and End Results (SEER)-Medicare database and analyzed regarding treatment details, including the use of IMRT versus non-IMRT, using claims data. Hazard ratios (HRs) were estimated by the frailty model with a propensity score matching cohort and instrumental variable analysis.

RESULTS

A total of 3172 patients were identified. With a median follow-up of 40 months, patients treated with IMRT had a statistically significant improvement in CSS compared with those treated with non-IMRT (84.1% versus 66.0%; P < .001). When each anatomic subsite was analyzed separately, all respective subgroups of patients treated with IMRT had better CSS than those treated with non-IMRT. In multivariable survival analyses, patients treated with IMRT were associated with better CSS (HR = 0.72, 95% confidence interval  = 0.59 to 0.90 for propensity score matching; HR = 0.60, 95% confidence interval = 0.41 to 0.88 for instrumental variable analysis).

CONCLUSIONS

Patients with head and neck cancers who were treated with IMRT experienced significant improvements in CSS compared with patients treated with non-IMRT techniques. This suggests there may be benefits to IMRT in cancer outcomes, in addition to toxicity reduction, for this patient population. Cancer 2014;120:702–710. © 2013 American Cancer Society.

INTRODUCTION

Head and neck cancers will be diagnosed in approximately 54,000 patients in 2013 and will cause approximately 12,000 deaths.[1] Treatment algorithms have evolved with time; extensive, disfiguring surgeries had been standard in early days, but these have largely been replaced with radiation therapy (RT), either in a definitive or adjuvant role (after more limited surgery). RT itself, however, is not without toxicity; radiation is well known to result in long-term sequelae for survivors, including dysphagia, xerostomia, osteoradionecrosis, and dental caries, all of which can result in significant decrements in quality of life and can also require expensive medical interventions.[2-6]

In an attempt to minimize these chronic toxicities, intensity-modulated radiation therapy (IMRT) was developed with the goal of sparing normal tissues while delivering curative dose to the tumor targets. Given the close proximity of critical normal tissues to the targets in cancers of the head and neck, IMRT was widely and enthusiastically employed in practice. Several analyses of the Surveillance, Epidemiology, and End Results (SEER)-Medicare database have demonstrated significant increases in the use of IMRT for head and neck cancer from 2000 to 2005.[7, 8] Previous studies have focused on the utility of IMRT in improving quality of life and toxicity profiles compared to non-IMRT, which includes both 2-dimensional (2D) and 3-dimensional (3D) conventional radiation therapy. Overall, these studies have demonstrated improved dose-volume parameters for key critical structures, improved quality of life, and better xerostomia scores for patients treated with IMRT.[9-14]

Although the goal of IMRT has been to spare normal tissues, the ability to provide more conformal dose distributions has also allowed dose escalation to tumor volumes and relative acceleration of dose delivery; it is widely accepted that high-dose areas receive greater than 2 Gy per fraction, which was the standard with non-IMRT conventional therapies.[13, 15, 16] Because of this, IMRT could theoretically improve or worsen cause-specific outcomes; namely, dose intensification could result in improved outcomes and cause-specific survival (CSS) if done well, or more conformality could risk marginal misses and worsened CSS if done poorly. Preliminary analyses of the SEER-Medicare database demonstrated no differences in survival between IMRT and non-IMRT for patients treated from 2002 through 2005.[17]

Because the use of IMRT has grown exponentially in the last decade, and the expertise of practitioners similarly has increased, the goal of this study was to analyze CSS for patients with head and neck cancer treated with IMRT as compared to those treated with non-IMRT to identify whether this widely accepted and comparatively expensive technique has benefit or risk with regard to oncologic outcomes.

MATERIALS AND METHODS

Data Source

The SEER-Medicare linked database was used to identify the cohort and outcomes of interest. The SEER program, which is a National Cancer Institute (NCI)-supported database, includes tumor registries in 16 geographic areas : Greater California, San Francisco-Oakland, Los Angeles, San Jose, Connecticut, Detroit, Seattle-Puget Sound, Atlanta, Rural Georgia, Iowa, Louisiana, New Mexico, Utah, Hawaii, Kentucky, and New Jersey, comprising approximately 25% of the United States population.[18] These registry data are linked to Medicare claims files by encrypted patient identifiers. The Medicare program provides payments for hospital, physician, and outpatient medical services for 97% of the US citizens who are ≥65 years of age.[19, 20] Diagnoses and procedures for each patient were identified using the Patient Entitlement and Diagnosis Summary File (PEDSF), as well as all available Medicare claims. All data were stripped of identifying information (de-identified) so that no protected health information could be traced to individual patients; the University of Texas MD Anderson Cancer Center institutional review board exempted this study.

Cohort Identification

We queried the database for patients with head and neck cancers treated between 1999 and 2007, as defined by SEER and International Classification of Diseases, 9th revision (ICD-9) codes, treated with radiation therapy, including both definitive and adjuvant RT. We included all histologies of the head and neck; the most common, as expected, was squamous cell carcinoma and its variants (91.2%). Patients were included only if they began definitive RT within 4 months or adjuvant RT within 6 months of diagnosis. This window was determined to allow for the possibility that some patients may have received neoadjuvant chemotherapy prior to local therapy. The SEER codes of interest were: cancers of the lip (01), tongue (02), floor of mouth (04), gum and other mouth (05), nasopharynx (06), tonsil (07), oropharynx (08), hypopharynx (09), other oral cavity and pharynx (10), nose, nasal cavity, and middle ear (37), and larynx (38). The ICD-9 diagnosis codes of interest were: cancers of the lip (140.x), tongue (141.x), floor of mouth (144.x), gum (143.x), oropharynx (146.x), mouth (145.x), nasopharynx (147.x), hypopharynx (148.x), ill-defined sites of the oral cavity, lip, and pharynx (149.x), nasal cavities, middle ear, and accessory sinuses (160.x), and laryngeal sites (161.x). For inclusion, the tumor needed to be pathologically confirmed, not diagnosed at death or autopsy, with a stage indicated, and no evidence of distant disease. For patients with laryngeal cancers (SEER code 38 and/or ICD-9 code 161.x), stage III and IV patients were included; early-stage patients (stages I and II) were excluded, because the latter are typically treated only with conventional non-IMRT techniques. Table 1 shows the algorithm for development of this cohort.

Table 1. Algorithm for Cohort Identification From the SEER-Medicare Database
CriteriaNo.Percent
Entire head and neck cancer cohort172,708 
Head and neck cancer is the only cancer or first of multiple cancers87,38150.6%
Diagnosis 1999–200743,78825.4%
Age at diagnosis ≥66 and ≤8019,20511.1%
Part A and B and no HMO 1 year before and after diagnosis96575.6%
In Medicare for age only96175.6%
Histologically confirmed95585.5%
Stage not missing and nonmetastatic77284.5%
If larynx (SEER code 38), only stage III and IV59123.4%
Received either definitive or adjuvant radiation31881.8%
Excluded missing income31721.8%

Outcome Identification

The primary outcome analyzed in this study was CSS. Overall survival (OS) was determined from SEER records; CSS was determined by extracting cause of death data from death certificates as reported by SEER, with follow-up through May 2010.[21]

Definition of Explanatory Variables

Potential explanatory variables of interest in this analysis included age at diagnosis, sex, race, marital status, urban versus rural residence, SEER geographic region, neighborhood income level, neighborhood educational level, and stage at diagnosis. We also investigated the importance of medical comorbidities using the Charlson comorbidity index, a modified comorbidity score using inpatient and outpatient claims within a 12-month window prior to the date of diagnosis.[22-25]

Detailed radiation treatment-related characteristics were identified from Medicare claims data through the ICD-9 procedure codes or Current Procedural Terminology (CPT) codes for radiation treatment planning and delivery. The cohort was further categorized as having received IMRT if any IMRT delivery or planning CPT codes were present (77418, 77301, GO174-IMRT, GO178-IMRT). If patients did not have an IMRT planning or delivery code as part of their RT claims, then they were assumed to have received non-IMRT techniques, e.g., conventional 2D or 3D RT techniques. Details of radiation delivery, including dose and target volumes, are not available within the SEER-Medicare database and hence not included in the analysis. In addition to RT delivery, we identified treatment with surgery and chemotherapy using the SEER coding for receipt of these modalities as part of primary management for a patient's disease at the time of diagnosis.

Statistical Analysis

Statistical analyses were conducted using SAS statistical software, version 9.3 (SAS Institute, Cary, NC). The unadjusted association of each potential explanatory variable with the outcome of CSS was assessed using the log-rank test statistic for binary and categorical variables. Covariate-adjusted analyses were performed using: 1) the semiparametric log-normal frailty model, an extension of the Cox regression model, and 2) instrumental variable (IV) analysis to derive hazard ratios (HR) and 95% confidence intervals (CIs).

Because the utilization of IMRT likely involved a selection bias, propensity score matching was used to compare patients treated with IMRT with matched controls. Propensity scores were calculated by the use of a logistic model comparing the dependent variable of IMRT versus non-IMRT treatment, and the independent variables being age at diagnosis, sex, race/ethnicity, marriage, rural versus urban residence, income, education, year of diagnosis, Charlson comorbidity index, head and neck subsite, and physician experience. Physician experience was a variable determined for each treating physician based on the number of RT planning and simulation codes. Patients were matched 1:1 by nearest-neighbor methodology with caliper (0.2*standard deviation) and without replacement.[26] A frailty model accommodating random error for matched pairs was used to compare survival between those patients treated with IMRT and those treated with non-IMRT.

Although the propensity scoring model accounts for defined variables, there is a possibility that there is a bias affecting which patients are treated with IMRT based on unmeasured factors. To account for unmeasured factors, an IV analysis was performed using 2-step residual inclusion methods.[27, 28] The IV was defined as provider experience, based on quantity of IMRT simulation and planning charges in the 1 year prior to treatment, which were dichotomized at the median value. To use this model, an instrumental variable is chosen that is considered to be related to the intervention in question (in this case, the use of IMRT) but not directly related to survival outcomes. By analyzing the cohort by the IV, one is able to mitigate selection bias and unmeasured confounding inherent in large database studies using administrative data. For this analysis, residuals derived from the first-stage equation (logistic regression) were added as a covariate in the second-stage equation (survival analysis). Variables used in the first-stage equation are the same variables used in estimating propensity score along with the IV. HRs were estimated by the second-stage equation.

The propensity score model attempts to control for selection bias by assigning a weight based on the probability of a patient having received an intervention or treatment (in this case, IMRT) according to observable factors, thus balancing the cohort to mitigate this. The IV analysis, on the other hand, accounts for unmeasured potential sources of confounding by analysis based on a variable that is correlated to the treatment, but not the outcomes in question.

RESULTS

Study Cohort Characteristics and Survival Outcomes

We identified 3172 patients who met our criteria for inclusion. The median follow-up time was 40 months. The mean age at diagnosis was 72.2 years (range, 66-80 years). Table 2 summarizes the demographic and treatment characteristics of the entire cohort. The most common site of cancer was oral cavity and oropharynx (58.3%), followed by larynx (all stages III and IV) (18.0%) and hypopharynx (10.7%). Of the entire cohort, 1056 patients (33.3%) were treated with IMRT and 2116 (66.7%) were treated with non-IMRT. Of the entire cohort, 1018 patients (32.1%) had surgery and adjuvant RT, and 2154 patients (67.9%) had definitive RT. Patients who received IMRT had statistically significant differences in several covariables compared to those that received non-IMRT (Table 3).

Table 2. Demographics and Characteristics of the Entire Cohort as Well as the Propensity Score–Matched Cohort
CharacteristicEntire Cohort (N = 3172)Propensity Score–Matched Cohort (N = 1894)Pa
N(%)N(%)
  1. a

    P values represent the result of the chi-square test of the variable (patient characteristics) between the entire cohort and the propensity-matched cohort.

  2. b

    Indicates statistically significant.

  3. Abbreviations: IMRT, intensity modulated radiotherapy; RT, radiation therapy.

Age at diagnosis66–691111(35.0%)699(36.9%).397
70–741193(37.6%)695(36.7%) 
75–80868(27.4%)500(26.4%) 
SexMale2119(66.8%)1302(68.7%).154
Female1053(33.2%)592(31.3%) 
EthnicityWhite non-Hispanic2574(81.1%)1556(82.2%).771
Black non-Hispanic260(8.2%)145(7.7%) 
Hispanic158(5.0%)95(5.0%) 
Other180(5.7%)98(5.2%) 
Marital statusMarried1791(56.5%)1122(59.2%).121
Unmarried1228(38.7%)694(36.6%) 
Unknown153(4.8%)78(4.1%) 
Rural/urbanUrban2874(90.6%)1726(91.1%).532
Rural298(9.4%)168(8.9%) 
Income quartile1st quartile793(25.0%)474(25.0%)1.000
2nd quartile793(25.0%)473(25.0%) 
3rd quartile794(25.0%)474(25.0%) 
4th quartile792(25.0%)473(25.0%) 
Education quartile1st quartile755(23.8%)451(23.8%)1.000
2nd quartile755(23.8%)452(23.9%) 
3rd quartile756(23.8%)449(23.7%) 
4th quartile748(23.6%)450(23.8%) 
Unknown158(5.0%)92(4.9%) 
Charlson comorbidity index01759(55.5%)1024(54.1%).001b
1739(23.3%)424(22.4%) 
2+460(14.5%)259(13.7%) 
Unknown214(6.7%)187(9.9%) 
Tumor siteOral cavity and oropharynx1848(58.3%)1198(63.3%)<.001b
Nasopharynx196(6.2%)108(5.7%) 
Hypopharynx339(10.7%)189(10.0%) 
Nose, nasal cavity, middle ear217(6.8%)149(7.9%) 
Larynx572(18.0%)250(13.2%) 
Type of radiationIMRT1056(33.3%)947(50.0%)<.001b
Non-IMRT2116(66.7%)947(50.0%) 
RegimenAdjuvant RT1018(32.1%)609(32.2%)0.964
Definitive RT2154(67.9%)1285(67.8%) 
Overall survivalAlive1389(43.8%)967(51.1%)<.001b
Dead1783(56.2%)927(48.9%) 
Cause-specific survivalAlive2284(72.0%)1471(77.7%)<.001b
Dead888(28.0%)423(22.3%) 
Table 3. Factors Associated With Use of IMRT Versus Non-IMRT for the Entire Cohort and Propensity Score–Matched Cohort
 Entire Cohort (N = 3172)Propensity Score–Matched Cohort (N = 1894)
TotalIMRT (N)(%)P ValueTotalIMRT (N)(%)P Value
  1. a

    Covariates used to estimate propensity score.

  2. b

    Indicates statistically significant.

  3. Abbreviations: IMRT, intensity modulated radiotherapy; RT, radiation therapy.

Age at Diagnosisa66–69 y1111401(36.1%).048b699357(51.1%).491
70–74 y1193381(31.9%) 695335(48.2%) 
75–80 y868274(31.6%) 500255(51.0%) 
SexaMale2119723(34.1%).1601302640(49.2%).275
Female1053333(31.6%) 592307(51.9%) 
EthnicityaWhite non-Hispanic2574871(33.8%).1081556784(50.4%).679
Black non-Hispanic26071(27.3%) 14566(45.5%) 
Hispanic15848(30.4%) 9546(48.4%) 
Other18066(36.7%) 9851(52.0%) 
Marital statusaMarried1791632(35.3%).007b1122561(50.0%).634
Unmarried1228386(31.4%) 694351(50.6%) 
Unknown15338(24.8%) 7835(44.9%) 
Rural/urbanaUrban2874974(33.9%).026b1726866(50.2%).628
Rural29882(27.5%) 16881(48.2%) 
Income quartilea1st quartile793228(28.8%).004b474223(47.0%).489
2nd quartile793268(33.8%) 473237(50.1%) 
3rd quartile794264(33.2%) 474244(51.5%) 
4th quartile792296(37.4%) 473243(51.4%) 
Education quartilea1st quartile755292(38.7%)<.001b451233(51.7%).607
2nd quartile755267(35.4%) 452230(50.9%) 
3rd quartile756240(31.7%) 449229(51.0%) 
4th quartile748205(27.4%) 450212(47.1%) 
Unknown15852(32.9%) 9243(46.7%) 
Charlson comorbidity Indexa01759588(33.4%).5151024526(51.4%)<.001b
1739258(34.9%) 424226(53.3%) 
2+460144(31.3%) 259129(49.8%) 
Unknown21466(30.8%) 18766(35.3%) 
Tumor siteaOral cavity and oropharynx1848679(36.7%)<.001b1198608(50.8%)<.001b
Nasopharynx19699(50.5%) 10865(60.2%) 
Hypopharynx339106(31.3%) 189105(55.6%) 
Nose, nasal cavity, middle ear21776(35.0%) 14973(49.0%) 
Larynx57296(16.8%) 25096(38.4%) 
RegimenAdjuvant RT1018323(31.7%).199609297(48.8%).461
 Definitive RT2154733(34.0%) 1285650(50.6%) 
Overall survivalAlive1389611(44.0%)<.001b967541(55.9%)<.001b
 Dead1783445(25.0%) 927406(43.8%) 
Cause-specific survivalAlive2284888(38.9%)<.001b1471794(54.0%)<.001b
Dead888168(18.9%) 423153(36.2%) 

Of the 3172 patients, a total of 1783 (56.2%) died. A total of 888 patients (28.0%) reportedly died of head and neck cancer–related causes. The median OS was 40 months (range, 3-133 months). On univariate analysis of CSS, statistically significant improvements in CSS were related to head and neck subsite (P = .0024), use of IMRT (P < .0001), lower Charlson comorbidity index (P < .0001), younger age (P < .0001), marital status (married > single; P < .0001), race/ethnicity (white non-Hispanic > Hispanic > black non-Hispanic; P = .0001), and income (higher > lower; P = .0363). Patients treated with IMRT had a statistically significant improvement in CSS compared to those treated with non-IMRT (38.9% versus 18.9%; P < .0001) (Fig. 1). The impact of factors on CSS is shown in Table 4.

Figure 1.

Impact of intensity-modulated radiation therapy (IMRT) on cause-specific survival. Kaplan-Meier curve depicting the cause-specific survival with time for patients treated with IMRT (dashed) compared to those treated with non-IMRT (solid).

Table 4. Impact of Factors on Cause-Specific Survival Using Propensity Scoring and Instrumental Variable Methods
Factor Propensity Scoring Frailty Model (n = 1894)Instrumental Variable 2-Stage Residual Inclusion Method (n = 3114)
HR95% CIHR95% CI
  1. Abbreviations: CI, confidence interval; HR, hazard ratio; IMRT, intensity modulated radiotherapy.

Radiotherapy type       
 Non-IMRT      
 IMRT0.720.590.900.600.410.88
Age       
 66–70 y      
 70–74 y1.020.801.301.120.951.32
 75–80 y1.301.011.671.301.091.54
Subsites       
 Hypopharynx      
 Larynx1.020.681.530.790.631.00
 Nasopharynx0.750.431.310.720.511.02
 Nose, nasal cavity, middle ear0.820.511.330.780.571.07
 Oral cavity0.950.671.350.820.671.01
 Oropharynx0.560.370.870.610.470.79
Ethnicity       
 White      
 Black non-Hispanic1.671.212.321.461.181.82
 Hispanic1.170.751.811.030.761.39
 Other1.000.631.601.130.470.79
Marital status       
 Married      
 Unmarried1.321.071.631.301.131.49
 Unknown1.470.932.340.940.681.31
Comorbidity       
 0      
 11.180.811.541.181.001.39
 2+1.581.182.121.451.201.74
 Unknown1.370.991.881.451.131.82
Year       
 1999–2002      
 2003–20070.510.400.660.650.530.79

On univariate analysis of OS, statistically significant improvements in OS were related to tumor subsite (P = .0013), use of IMRT (P = .0469), lower Charlson comorbidity index (P < .0001), younger age (P < .0001), marital status (married > single; P < .0001), race/ethnicity (white non-Hispanic > Hispanic > black non-Hispanic; P < .0001), income (higher > lower; P = .0009), and education (more > less; P = .0346).

IMRT Use and Survival Outcomes

Given that the use of IMRT was nonuniform among the patient population, a propensity score model was created to understand the impact of IMRT versus non-IMRT on CSS. A propensity score–matched cohort was created, composed of 1894 patients (947 pairs). Table 2 shows the demographic characteristics of this matched cohort, demonstrating that the input variables are balanced based on the use of IMRT; the balance of characteristics is shown in the full cohort and propensity score–matched cohort in Table 3. With this matched-pair analysis, patients treated with IMRT had significantly improved CSS compared with patients treated with non-IMRT (HR = 0.72, CI = 0.59 to 0.90) as shown in Table 4. The IV analysis also shows improvement in head and neck cancer–specific mortality among those treated with IMRT with a HR of 0.60 (95% CI = 0.41 to 0.88) (Table 4). This analysis also confirms previous studies that have shown improvements in CSS with other factors, including marital status and disease subsite. Overall survival among patients receiving IMRT versus those who did not in this analysis was not significant in the covariate-adjusted model of the matched cohort and was of borderline significance (P = .0469) using the log-rank test of the entire cohort.

DISCUSSION

In this analysis, IMRT use was associated with a significant improvement in CSS compared to treatment with non-IMRT techniques. This is a unique, and controversial, finding. IMRT was first adopted as a method of providing definitive RT while sparing normal tissues, with a goal of improving side effects, namely xerostomia. A series of phase 2 and 3 trials have demonstrated that it is effective in the purported goal of reducing toxicity.[12, 13, 15, 16] In addition, both these prospective trials and multiple single-institution reports indicate locoregional control and survival levels on par (or at least not inferior) to those expected from historical reports of conventional, non-IMRT radiation.[15, 29-31] However, our current study offers the first data to suggest that IMRT may improve cancer-control outcomes, in addition to the improved toxicity profiles reported by other investigators.

IMRT itself allows improved conformality of radiation treatment, which is especially important in the head and neck, where critical normal tissues are in close proximity to the targets. Historically, conventional 2D and 3D RT uses large fields and a series of field reductions to provide sequentially higher doses to the primary tumor. IMRT has the advantage of being able to target the tumor itself with greater conformality and simultaneously reducing dose to normal tissues. As a result, dose intensification, both total dose and dose per fraction, is feasible with IMRT where it was not with non-IMRT techniques. Although tight margins and the opportunity for geographic miss are possible drawbacks to IMRT, the increased conformality also provides the potential to optimize normal tissue sparing and provide curative dose to adjacent tumor targets, which could conversely make it more effective. For these reasons, it is has always been conceivable that IMRT would provide a benefit in cancer-control outcomes. This is the first report to document such a finding in a large population-based database.

To more thoroughly evaluate this finding, while taking into account possible selection bias as to who received treatment with IMRT, a propensity score model analysis was used. IMRT is not universally employed or available in all radiation oncology practices, and it may also be subject to selection bias in its utilization. Furthermore, it is an advanced technique in which greater experience is likely to result in better outcomes. To take into account the nonuniform application of IMRT, the model took into account diagnosis, sex, race/ethnicity, marriage, rural versus urban residence, income, education, year of diagnosis, Charlson comorbidity index, head and neck subsite, and physician experience. Even with these variables modeled, patients treated with IMRT still had a CSS benefit. This suggests that, controlling for both factors that influence who receives IMRT for head and neck cancer as well as these clinically relevant variables, IMRT is still superior to non-IMRT techniques with respect to CSS.

Although the propensity scoring matching accounts for the known variables, there is still a possibility that there is bias in the 2 groups (IMRT versus non-IMRT) as a result of unmeasured factors. To account for this, an IV model was used; this also demonstrated that there is a statistically significant benefit to IMRT with increased CSS. Similar methods have been used to account for confounders in observational data in the study of prostate cancer and renal cancer.[32, 33] Coupled with the propensity matching analysis, these data suggest that the observed improvement in CSS was robust for patients treated with IMRT compared to non-IMRT for head and neck cancer.

A recent study looking at the cost-effectiveness of IMRT in the treatment of oropharynx cancer, one of the subsites included in our current analysis, used a Markov model to estimate the incremental cost per quality-adjusted life-year (QALY) gained by IMRT.[34] In their model, IMRT use was assumed to lower the probability of the development of xerostomia. However, patients with either treatment strategy were estimated to have the same rates of disease control. Even with this assumption of equivalent cancer control outcomes, the authors concluded that IMRT was cost-effective, with a need to treat fewer than 2 patients with IMRT to avoid a case of severe chronic xerostomia with an incremental cost of $4532. Our analyses suggest that, in addition to decreasing the long-term toxicity, the outcomes of patients may not be equivalent. Hence, the cost-effectiveness of IMRT would be even more substantial. This merits further study.

The analysis and conclusions described here should be considered hypothesis-generating. As with all observational studies using population-based claims data, this study has limitations. First, claims data in the Medicare database are not collected for research purposes; as a result, there is a degree of uncertainty regarding their completeness and accuracy. Furthermore, the data does not reflect patient and physician preferences. Second, the Medicare database used here only captures patients older than 65 years of age and does not capture complete data on younger patients or those patients with managed care coverage. In addition, the SEER-Medicare database does not include information regarding radiation dose or target volume; both of these factors may have significant impact on the outcomes of RT, especially with IMRT, and the lack of this data does limit the analysis. Finally, there is no data on human papillomavirus status, which may also significantly affect tumor control and survival in patients with head and neck cancers and may not be evenly distributed between the cohort treated with IMRT and that treated with non-IMRT. Despite these drawbacks, this data source provides a comprehensive population-based resource for studying patterns of care and outcomes throughout the United States.

IMRT is a comparatively expensive technology that was employed on the basis of theoretical benefits to patients with head and neck cancer in the reduction of acute and long-term toxicity. Multiple retrospective and now prospective studies have shown this to be a reality.[13, 15, 16] However, IMRT is able to better target and dose-escalate specific regions of concern for tumor control; hence, improved cancer-control outcomes, while never the goal, were always a possibility. This is the first study suggesting that IMRT provides benefits in CSS in patients with head and neck cancers. As the population of patients with head and neck cancer grows younger and more prone to both toxicity and recurrence, a technology that could improve both may not only be comparatively more effective but may also be more cost-effective by reducing needs for costly salvage therapy for recurrences and for side effect management expenses, even if the upfront RT costs are higher due to use of a more resource-intensive RT technique. Although it is unlikely that future trials will be designed comparing IMRT and conventional radiation with cancer control as an outcome, given the results of prior trials designed for toxicity endpoints, further study is necessary to understand the long-term impact of IMRT on cancer-specific endpoints, economic evaluations of its utility, and best methods for minimizing cancer care and survivorship costs.

FUNDING SOURCES

No specific funding was disclosed.

CONFLICT OF INTEREST DISCLOSURE

The authors made no disclosures.

Ancillary