Multicenter Validation of Recursive Partitioning Analysis Classification for Patients With Squamous Cell Head and Neck Carcinoma Treated With Surgery and Postoperative Radiotherapy

Int. J. Radiation Oncology Biol. Phys., Vol. 68, No. 1, pp. 119 –125, 2007 Copyright © 2007 Elsevier Inc.

Printed in the USA. All rights reserved 0360-3016/07/$–see front matter

C

SaaMfr

pnTl

doi:10.1016/j.ijrobp.2006.12.002

LINICAL INVESTIGATION Head and Neck

MULTICENTER VALIDATION OF RECURSIVE PARTITIONING ANALYSIS CLASSIFICATION FOR PATIENTS WITH SQUAMOUS CELL

HEAD AND NECK CARCINOMA TREATED WITH SURGERY AND POSTOPERATIVE RADIOTHERAPY

ANJA JONKMAN, M.D.,* JOHANNES H. A. M. KAANDERS, M.D., PH.D.,†

CHRIS H. J. TERHAARD, M.D., PH.D.,‡ FRANK J. P. HOEBERS, M.D.,§

PIET L. A. VAN DEN ENDE, M.D.,¶ ODA B. WIJERS, M.D., PH.D.,� LIA C. G. VERHOEF, M.D., PH.D.,#

MARTIN A. DE JONG, M.D.,** C. RENÉ LEEMANS, M.D., PH.D.,††

AND JOHANNES A. LANGENDIJK, M.D., PH.D.*

*Department of Radiation Oncology, University Medical Center Groningen, University of Groningen, Groningen, The Netherlands;†Department of Radiation Oncology, Radboud University Nijmegen Medical Center, Nijmegen, The Netherlands; ‡Department of

Radiation Oncology, University Medical Center Utrecht, Utrecht, The Netherlands; §Department of Radiation Oncology,Antoni van Leeuwenhoek Hospital/Netherlands Cancer Institute, Amsterdam, The Netherlands; ¶MAASTRO Clinic, Maastricht,

The Netherlands; �Radiotherapy Institute Friesland, Leeuwarden, The Netherlands; #Arnhem Radiotherapy Institute, Arnhem,The Netherlands; **Department of Radiation Oncology, Leiden University Medical Center, Leiden, The Netherlands; ††Department

of Otolaryngology/Head and Neck Surgery, VU University Medical Center, Amsterdam, The Netherlands

Purpose: To validate the recursive partitioning analysis (RPA) classification system for squamous cell head and neck cancer as recently reported by the VU University Medical Center. Methods and Materials: In eight Dutch head and neck cancer centers, data necessary to classify patients according to the RPA system were retrospectively collected from the charts of a group of 780 patients treated between 1989 and 2003. The patients in this validation group were classified according to the RPA classification system. For each endpoint, the 5-year values and hazard ratios were calculated and compared with the results of the VU University Medical Center. The RPA classification system was considered valid if the hazard ratio of the validation population was within the 95% confidence interval of the VU University Medical Center study population. Results: The locoregional control rate was 82%, 75%, and 63% at 5 years for those with class I, II, and III, respectively (p < 0.0001). The hazard ratio for the locoregional control rate relative to class I was 1.44 (95% confidence interval, 0.97–2.16) for class II and 2.37 (95% confidence interval, 1.57–3.57) for class III. Similar results were found for the distant metastasis, overall survival, and disease-free survival rates. Conclusion: The RPA classification system for head and neck squamous cell carcinoma in the postoperative setting, which was originally designed at one center, proved to be valid in a multicenter setting among patients included in a national multicenter study. This validated RPA classification scheme can be used to assess standard treatment strategies for head and neck squamous cell carcinoma in the postoperative setting, as well as in the design of future prospective studies. © 2007 Elsevier Inc.

Head and neck cancer, Postoperative radiotherapy, Surgery, Recursive partitioning analysis, Prognostic factors.

ao(

msot

TL

A

INTRODUCTION

quamous cell carcinoma of the head and neck (HNSCC) is relatively rare group of malignant tumors, accounting for lmost 10% of all new cancer patients in The Netherlands. any of these patients are treated with primary surgery

ollowed by adjuvant treatment, including postoperative adiotherapy (RT), with or without concomitant chemother-

Reprint requests to: Johannes A. Langendijk, M.D., Ph.D., De­artment of Radiation Oncology, University Medical Center Gro­ingen, P.O. Box 30001, Groningen 9300 RB The Netherlands. el: (�31) 50-361-1190; Fax: (�31) 50-361-1692; E-mail: j.a.

[email protected]

119

py, in the case of adverse prognostic factors, such as close r positive surgical margins (1– 6) and extranodal spread ENS) (3, 5, 7–9).

To make the right decision with regard to adjuvant treat-ent after primary surgery, information of the outcome of

pecific prognostic groups is essential. Moreover, the choice f adjuvant treatment (e.g., total dose, overall treatment ime, and addition of chemotherapy) after primary surgery

herapeutic Radiology and Oncology (ESTRO) Meeting 2006, eipzig, Germany, October 8 –12, 2006. Conflict of interest: none. Received Sept 15, 2006, and in revised form Dec 1, 2006.

ccepted for publication Dec 1, 2006.

120 I. J. Radiation Oncology ● Biology ● Physics Volume 68, Number 1, 2007

wp

r(ssgma

pfiosdtc

P

espRstme

tDWtt12w(ebMFpip9tdAtc1

P

ot

itthp(tEdfe

R

NISRno

S

Mp(cptpdp

R

udtcptvwl(ws(impGtttTaito

ill depend on the risk of locoregional failure and the robability of distant metastases (9, 10).

Recently, the VU University Medical Center (VUMC) eported on the results of recursive partitioning analysis RPA) among patients with HNSCC treated with curative urgery and postoperative RT (11). This RPA classification ystem allowed for a clear distinction of three prognostic roups with regard to locoregional control (LRC), distant etastasis-free interval (DMFI), disease-free survival (DFS),

nd overall survival (OS). One of the main criticisms of this report was that only

atients from one center were included and that the classi­cation system had not been validated. Therefore, the aim f this study was to validate the VUMC-RPA classification ystem (VUMC-RPA) in a multicenter setting (RPA vali­ation). More specifically, the hypothesis to be tested was hat this classification system would be significantly asso­iated with LRC, DMFI, DFS, and OS.

METHODS AND MATERIALS

atient characteristics To be included in this study, patients had to meet the same

ligibility criteria as used in the VUMC (i.e., SCC of the mucosal urfaces of the oral cavity [excluding the lip], oropharynx, hypo­harynx, or larynx treated with curative surgery and postoperative T). Excluded were those with macroscopic residual tumor after

urgery, distant metastases before initiating postoperative RT, pa­ients treated with chemotherapy, patients treated with intensity-odulated RT, and patients treated with excision biopsy only for

arly-stage tonsillar carcinoma. In The Netherlands, most patients with head and neck cancer are

reated at 10 specialized centers, all of which are members of the utch Head and Neck Oncology Cooperative Group (Nederlandse erkgroep Hoofd-Hals Tumoren). For the purposes of this study,

he data were collected retrospectively from the records of patients reated at eight of these centers. The aim was to collect data from 00 patients from each center with a minimal follow-up period of 4 months. Assessment of the data started in July 2005; therefore, e started to select consecutive patients treated up to July 2003

minimal follow-up period of 2 years) aiming at 100 patients for ach center. Eventually, a total of 780 consecutive patients treated etween January 1990 and July 2003 were included (University edical Center Groningen, 204 patients; Radiotherapy Institute

riesland, 90 patients; University Medical Center Utrecht, 96 atients; MAASTRO Clinic, 97 patients; Leiden University Med­cal Center, 80 patients; University Medical Center Nijmegen, 98 atients; Antoni van Leeuwenhoek Hospital/Dutch Cancer Center, 0 patients; and Arnhem Radiotherapy Institute, 25 patients). In he University Medical Center Groningen, data from an existing atabase that already included 204 patients were used. In the rnhem Radiotherapy Institute, a limited number of patients were

reated with postoperative RT after 1990. The pretreatment patient haracteristics are listed in Table 1. The TNM classification of 997 was used for tumor staging.

athologic findings All pathology reports were reviewed. In the case of insufficient

r unclear information, the local pathologists were asked to review

nformation: type of surgery of the primary site; size of the primary umor; surgical margins (free, �5 mm; close, 1–5 mm; and posi­ive, �1 mm); angioinvasion (yes/no); squamous cell carcinoma as istologic type (yes/no); grade of differentiation (well/moderate/ oor); perineural growth (yes/no); dysplasia in surgical margins yes/no); type of neck dissection; number of lymph nodes inves­igated; number of tumor-positive nodes; number of nodes with NS; pathologic T stage; and pathologic N stage. In line with the efinitions of Ang et al. (10), patients without adverse prognostic actors, who had no indication for postoperative RT, were consid­red low-risk patients and were not included in this analysis.

PA classification RPA class I (intermediate risk) includes patients without Stage

3 nodes, free surgical margins (�5 mm), and no ENS. RPA class I (high risk) includes patients with one positive node with ENS or tage T1, T2, or T4 tumors with close or positive surgical margins. PA class III (very high risk) includes patients with Stage N3 eck, two or more positive nodes with ENS, or Stage T3 with close r positive surgical margins.

urgery All patients underwent curative surgery of the primary tumor. inor surgery of the primary tumor, including local excision, was

erformed in 177 patients (23%). In the remaining 603 patients 77%), surgery consisted of more extensive procedures, such as ommando resection, partial or total glossectomy, partial or total haryngectomy, marginal or segmental mandibulectomy, and par­ial or total laryngectomy. In some patients, a combination of these rocedures was performed. In 481 patients (62%), unilateral neck issection, and in 189 patients (24%), bilateral neck dissection was erformed.

adiotherapy All patients were treated with conventional radiation techniques

sing direct simulation or a planning CT procedure. Among the ifferent institutions, some heterogeneity was present regarding otal dose, dose per fraction, and radiation technique. In most ases, the initial clinical target volume consisted of the original rimary tumor site, with margins varying from 1.5 to 2.0 cm and he lymph node areas on both sides of the neck. The initial target olume was generally irradiated by two opposing lateral beams ith or without one AP beam or two opposing AP beams for the

ower jugular and supraclavicular lymph nodes. In 86 patients 11%), unilateral irradiation was applied (e.g., in the case of ell-lateralized tumors and/or in the case of unilateral neck dis­

ection with no pathologic lymph nodes [pN0]). In 31 patients 4%), only local RT was applied. The median total dose to the nitial target volume was 46 Gy (range, 20.0 –70.4 Gy), and the edian dose per fraction was 2.0 Gy (range, 1.8 –2.3 Gy). In 750

atients (96%), the total dose to the initial target volume was �40 y. In the case of a total dose of 50 Gy to the initial target volume,

he spinal cord was shielded after 40 Gy, with subsequent radiation o the anterior part of the neck with photons and electron beams for he posterior triangle. This technique was used in 319 cases (41%). he clinical target volume of the boost included the high-risk reas, including the original primary tumor site and pathologically nvolved neck node levels. The total dose to the target volume of he boost varied from 30 to 72 Gy (median dose, 64 Gy) depending n the presence of risk factors, such as surgical margin status, the

121 Validation of RPA classification in HNSCC ● A. JONKMAN et al.

G

A

T

N

P

G

S

P

E

A

D

odotwfme

S

r

Dtdrad

cLmc

Table 1. Pretreatment characteristics and u

Variable Patients (n) 5-y LR

ender Male 552 (71) Female 228 (29)

ge (y) 0–59 397 (51) �60 383 (49)

stage T1 36 (5) T2 182 (23) T3 157 (20) T4 405 (52)

stage N0 273 (35) N1 140 (18) N2a 17 (2) N2b 236 (30) N2c 98 (13) N3 16 (2)

rimary site Larynx 217 (28) Hypopharynx 66 (8) Oropharynx 149 (19) Oral cavity 348 (45)

rade of differentation Well differentiated 76 (10) Moderately differentiated 429 (55) Poorly differentiated 205 (26) Missing 70 (9)

urgical margins primary site Wide surgical margins 263 (34) Close surgical margins (1–5 mm) 213 (27) Positive surgical margins (�1 mm) 304 (39)

erineural invasion No 597 (77) Yes 182 (23) Missing 1

xtracapsular nodal extension No 513 (66) Yes 267 (34)

ngioinvasive growth No 655 (84) Yes 124 (16) Missing 1

ysplasia in surgical margin No 664 (85) Yes 115 (15)

Abbreviations: LRC � locoregional control; CI � confidence iData in parentheses are percentages.

f lymph node metastases with ENS. In 736 patients (94%), a total ose of �55 Gy was administered to the primary site. In the case f ENS, 246 (92%) of 267 patients received a total dose of �55 Gy o the involved lymph node area. In some patients, the total dose as lower because postoperative RT had been previously stopped

or different reasons, including patient refusal, excessive acute orbidity, comorbid diseases, tumor progression, and/or the pres­

nce of distant metastases.

tatistical analysis The primary endpoint, which was LRC, was defined as no tumor

ecurrence above the clavicles within the irradiated area. The e

te analysis regarding locoregional control

(%) Log–rank p Uncorrected hazard ratio 95% CI

0.030 1.00 0.72 0.53–0.97

0.540 1.00 0.92 0.69–1.21

0.477 1.00 0.75 0.40–1.41 0.63 0.33–1.21 0.66 0.36–1.20

�0.001 1.00 1.73 1.08–2.77 0.44 0.06–3.2 3.51 2.41–5.11 2.03 1.22–3.37 6.37 2.97–13.6

�0.001 1.00 1.49 0.80–2.77 1.46 0.92–2.34 2.19 1.50–3.20

0.301 1.00 1.41 0.82–2.42 1.56 0.89–2.78 1.27 0.62–2.60

�0.001 1.00 1.05 0.71–1.56 1.67 1.19–2.34

0.005 1.00 1.56 1.14–2.12 NA

�0.001 1.00 2.31 1.74–3.06

0.519 1.00 1.13 0.78–1.66 NA

0.751 1.00 0.94 0.63–1.40

l.

MFI was defined as no development of distant metastases (with he event being distant metastasis, and censored occurring for eath from any cause). The DFS was defined as no disease recur­ence and no death from any cause. Overall survival was defined s no death from any cause. All events were calculated from the ate of surgery. For the purposes of this study, the patients were retrospectively

lassified according to the VUMC-RPA classification system (11). RC, DMFI, DFS, and OS were estimated using the Kaplan-Meier ethod on univariate analysis. Differences between curves were

ompared using the log–rank test for statistical significance. For

nivaria

C rate

76 69

73 75

63 70 76 75

86 77 93 56 73 55

83 76 76 66

79 74 71 77

79 77 67

76 67 NA

80 59

74 70 NA

73 75

nterva

ach endpoint, the 5-year values and hazard ratios (HRs) were

122 I. J. Radiation Oncology ● Biology ● Physics Volume 68, Number 1, 2007

LHDHDHOH

grees o

ccdtCiwfTNmdEoy(

5

U

n(pa

M

m(httpbugtnr

R

i(

iwi6Dfgi3Ot(

F

Table 2. Outcome acc

Variable I (intermediate ri

ocoregional control at 5 y (%) 82% azard ratio (95% confidence interval) 1.00 istant metastases free interval at 5 y (%) 92% azard ratio (95% confidence interval) 1.00 isease-free survival at 5 y (%) 59% azard ratio (95% confidence interval) 1.00 verall survival at 5 y (%) 60% azard ratio (95% confidence interval) 1.00

Abbreviations: RPA � recursive partioning analysis; DF � de

alculated and compared with the results of the VUMC-RPA. The lassification system was considered valid if the HR of the vali­ation study population was within the 95% confidence interval of he VUMC-RPA study population. Multivariate analysis using the ox model was also performed to allow for identification of

ndependent prognostic factors for LRC. The following factors ere evaluated as prognostic indicators of LRC: gender (male vs.

emale), age (split at the median level of �60 years vs. �60 years), stage (ordinal: T1, T2, T3, and T4), N stage (ordinal: N0, N1, 2a, N2b, N2c, and N3), grade of differentiation (ordinal: well, oderately, and poorly differentiated), surgical margin status (or­

inal: free [�5 mm], close [1–5 mm], and positive [�1 mm]), NS (not present, present in one node, and present in more than ne node), tumor site (nominal: oral cavity, oropharynx, hypophar­nx, and larynx), perineural invasion (yes vs. no), angioinvasion yes vs. no), and dysplasia in the surgical margin (yes vs. no).

RESULTS

The median follow-up was 46 months. The LRC rate after years was 73% for the entire study population.

nivariate analysis On univariate analysis, the following variables were sig­

ificantly associated with LRC: gender (p � 0.030), N stage p � 0.001), primary site (p � 0.001), surgical margins at rimary site (p � 0.001), perineural invasion (p � 0.005), nd ENS of lymph node metastases (p � 0.001; Table 1).

ultivariate analysis On multivariate analysis, ENS (p � 0.004), surgical argin status (p � 0.023), N stage (p � 0.001), and gender

p � 0.047) were significantly associated with LRC. Some eterogeneity was present regarding target volume defini­ion and total radiation dose among the different institu­ions. In this population, a relatively small proportion of atients (15%) were not treated with standard bilateral RT ut with unilateral RT, including the primary site and the nilateral neck or the primary site only. These latter two roups included a selection of patients with well-lateralized umors and, in most cases, a pathologically node-negative eck (pN0). Moreover, none of these patients developed

to RPA classification

RPA class

II (high risk) III (very high risk) Log-rank test p (DF � 2)

75% 63% �0.0001 1.55 (1.03-2.33) 2.51 (1.64-3.82)

88% 73% �0.0001 1.44 (0.78-2.66) 3.99 (2.21-7.23)

47% 34% �0.0001 1.39 (1.10-1.77) 1.97 (1.52-2.55)

50% 36% �0.0001 1.34 (1.05-1.70) 1.93 (1.49-2.51)

f freedom; LRC � locoregional control.

PA classification system Of the 780 patients, 182 (23%) were classified as being at

ntermediate risk (RPA class I), 384 (49%) at high risk RPA class II), and 214 (28%) at very high risk (class III).

After applying the RPA classification system to the val­dation database, the RPA class was significantly associated ith all endpoints (Table 2). The LRC rate was 82% for the

ntermediate-risk group, 75% for the high-risk group, and 3% for the very-high-risk group (p � 0.0001; Fig. 1). The MFI rate was 92% for the intermediate-risk group, 88%

or the high-risk group, and 73% for the very-high-risk roup (p � 0.0001; Fig. 2). The DFS rate was 59% for the ntermediate-risk group, 47% for the high-risk group, and 4% for the very-high-risk group (p � 0.0001; Fig. 3). The S rate was 60% for the intermediate-risk group, 50% for

he high-risk group, and 36% for the very-high-risk group p � 0.0001; Fig. 4).

As shown in Table 2, almost all HRs were within the 95%

ig. 1. Locoregional control according to recursive partitioning

ording

sk)

nalysis (RPA) classification.

123 Validation of RPA classification in HNSCC ● A. JONKMAN et al.

Ft

c(H

A

pdat

Fa

F(

d(t

vfgcrcO

gtR

irpTaaqacbpri

ig. 2. Distant metastasis-free interval according to recursive par­itioning analysis (RPA) classification.

onfidence interval of the VUMC-RPA, except for class III LRC), class III (DFS), and class II (OS). In all cases, the Rs were lower than those found in the VUMC study.

dditional analyses In contrast to what was found in the VUMC-RPA study

opulation, in the RPA validation study population, no ifference regarding LRC was found between those with T3 nd T4 tumors with positive or close surgical margins. In his subset, the 5-year LRC rate for those with Stage T3

ig. 3. Disease-free survival according to recursive partitioning

ig. 4. Overall survival according to recursive partitioning analysis RPA) classification.

isease was 71% vs. 67% for those with Stage T4 disease p � 0.51). This finding persisted after correcting for po­ential confounders.

DISCUSSION

This multicenter retrospective study was performed to alidate the single-center VUMC-RPA classification system or HNSCC treated with surgery and postoperative RT. In eneral, the results revealed that assignment of this RPA lassification system to a multicenter population of patients esulted in statistically significant differences for all clini­ally relevant endpoints, including LRC, DMFI, DFS, and S among the three classes. It should be emphasized that even though some hetero­

eneity was present regarding target volume definition and otal radiation dose among the different institutions, the PA classification system remained valid. Recently, the results of two prospective randomized stud­

es were presented in which patients with HNSCC were andomly assigned to receive postoperative RT alone vs. ostoperative RT plus concomitant chemotherapy (12, 13). hese trials have provided strong evidence for better LRC nd OS when chemotherapy was added to postoperative RT mong patients with unfavorable tumor characteristics. The uestion arises as to whether the RPA classification system lso holds true for those patients receiving postoperative oncomitant chemoradiotherapy. This question remains to e answered and is the subject of a separate analysis among atients who received postoperative chemoradiotherapy. A ecently performed comparative analysis also identified pos­tive surgical margins and/or ENS as important prognostic

124 I. J. Radiation Oncology ● Biology ● Physics Volume 68, Number 1, 2007

tpspbpiat

tcbsatReFstbRtiR

bsAodwottf

Temctwtbpidpomutwwhr(fcdfttp

caoa

he most benefit from the addition of chemotherapy to ostoperative RT (14). The translation of the results of that tudy to the RPA classification system could imply that atients classified as being at high or very high risk might enefit most from concomitant chemoradiotherapy in the ostoperative setting. A question that needs to be addressed s whether the classification according to the RPA system is better tool for the selection of patients for this combined reatment.

The HRs found in the present study were somewhat lower han the HRs found in the VUMC-RPA study population. In ontrast to the validation study population, the initial data­ase included patients from only one center. It should be tressed that the possibility of interobserver variability mong the pathologists from the different centers, as well as he differences regarding the indications for postoperative T among the different institutions, might have had some ffect on the distinctive power of the classification system. or these reasons, and from a statistical viewpoint, it is not urprising that the HRs in the initial database were greater han in the validation database. Another explanation could e that the LRC rate in the intermediate-risk group of the PA validation population, in particular, was lower than

hat in the VUMC-RPA population (82% vs. 92%), result­ng in smaller differences among the three classes in the PA validation population. One unexpected finding by Langendijk et al. (11) was

etter LRC for T4 tumors compared with T3 tumors in the ubset of patients with close or positive surgical margins. dditional analysis of the relationship between T stage and ther variables in the VUMC study revealed a significant ifference regarding the location of the primary tumor site, ith more laryngeal tumors in the T4 category and more ropharyngeal tumors in the T3 category. However, addi­ional analysis of the prognostic significance of T stage in his subset, stratified by tumor site, still revealed better LRC

REFEREN

152: 456–463.

1

he investigators concluded that this finding could not be xplained by differences in the primary tumor sites but was ost likely a result of the larger size of the T3 tumors

ompared with the T4 tumor size. However, this observa­ion was not confirmed by the results of the present study in hich no difference was noted between the T3 and T4

umors. A possible explanation for this discrepancy could e that the selection of patients for either surgery plus ostoperative RT or a nonoperative approach with concom­tant chemoradiotherapy or RT alone differs among the ifferent institutions. In the period of inclusion, the ap­roach at the VUMC was more surgery oriented than at the ther institutions, which could have resulted in relatively ore patients with advanced and unfavorable T3 disease

ndergoing primary surgery and postoperative RT than at he other institutions. It is extremely difficult to assess hether the worse prognosis of those with T3 compared ith those with T4 as mentioned by Langendijk et al. (11) as also been found by others. This is because in many eports T3 and T4 have been considered as one category 15). In some studies, no difference between T3 and T4 was ound (16), but in other studies, those with T3 did signifi­antly better than those with T4 (17). Thus, it is difficult to raw definitive conclusions as to whether a significant dif­erence exists between T3 and T4 tumors in the postopera­ive setting. Despite these differences, the RPA classifica­ion system still allowed for a clear distinction of three rognostic groups.

CONCLUSION

The RPA classification is a valid method to predict out­omes among patients with HNSCC treated with surgery nd postoperative RT. It may provide a tool for the selection f high-risk patients most likely to benefit from intensified djuvant treatments such as concomitant chemoradiother­

CES

1. Bastit L, Blot E, Debourdeau P, et al. Influence of the delay of adjuvant postoperative radiation therapy on relapse and sur­vival in oropharyngeal and hypopharyngeal cancers. Int J Radiat Oncol Biol Phys 2001;49:139 –146.

2. Chen TY, Emrich LJ, Driscoll DL. The clinical significance of pathological findings in surgically resected margins of the primary tumor in head and neck carcinoma. Int J Radiat Oncol Biol Phys 1987;13:833– 837.

3. Amdur RJ, Parsons JT, Mendenhall WM, et al. Postoperative irradiation for squamous cell carcinoma of the head and neck: An analysis of treatment results and complications. Int J Radiat Oncol Biol Phys 1989;16:25–36.

4. Mantravadi RV, Haas RE, Liebner EJ, Skolnik EM, Apple­baum EL. Postoperative radiotherapy for persistent tumor at the surgical margin in head and neck cancer. Laryngoscope 1983;93:1337–1340.

5. O’Brien CJ, Smith JW, Soong SJ, Urist MM, Maddox WA. Neck dissection with and without radiotherapy: Prognostic factors, patterns of recurrence and survival. Am J Surg 1986;

6. Parsons JT, Mendenhall WM, Stringer SP, Cassisi NJ, Million RR. An analysis of factors influencing the outcome of postoperative irradiation for squamous cell carcinoma of the oral cavity. Int J Radiat Oncol Biol Phys 1997;39:137– 148.

7. Carter RL, Bliss JM, Soo KC, O’Brien CJ. Radical neck dissec­tions for squamous carcinomas: Pathological findings and their clinical implications with particular reference to transcapsular spread. Int J Radiat Oncol Biol Phys 1987;13:825– 832.

8. Bartelink H, Breur K, Hart G, et al. The value of postoperative radiotherapy as an adjuvant to radical neck dissection. Cancer 1983;52:1008 –1013.

9. Peters LJ, Goepfert H, Ang KK, et al. Evaluation of the dose for postoperative therapy of head and neck cancer: First report of a prospective randomized trial. Int J Radiat Oncol Biol Phys 1992;26:3–11.

0. Ang KK, Trotti A, Brown BW, et al. Randomized trial ad­dressing risk features and time factors of surgery plus radio­therapy in advanced head-and-neck cancer. Int J Radiat Oncol

125 Validation of RPA classification in HNSCC ● A. JONKMAN et al.

1

1

1

1

1

1

1

1. Langendijk JA, Slotman BJ, van der Waal I, et al. Risk-group definition by recursive partitioning analysis of pa­tients with squamous cell head and neck carcinoma treated with surgery and postoperative radiotherapy. Cancer 2005; 104:1408 –1417.

2. Bernier J, Domenge C, Ozsahin M, et al. Postoperative irradia­tion with or without concomitant chemotherapy for locally ad­vanced head and neck cancer: European Organization for Re­search and Treatment of Cancer trial 22931. N Engl J Med 2004;350:1945–1952.

3. Cooper JS, Pajak TF, Forastiere AA, et al. Postoperative concur­rent radiotherapy and chemotherapy for high-risk squamous-cell carcinoma of the head and neck: Radiotherapy Oncology Group 9501/Intergroup. N Engl J Med 2004;350:1937–1944.

4. Bernier J, Cooper JS, Pajak TF, et al. Defining risk levels in

ysis of concurrent postoperative radiation plus chemotherapy trials of the EORTC (#22931) and RTOG (# 9501). Head Neck 2005;27:843– 850.

5. Muriel VP, Tejada MR, de Dios Luna del Castillo J. Time-dose-response relationships in postoperatively irradiated pa­tients with head and neck squamous cell carcinomas. Ra­diother Oncol 2001;60:137–145.

6. Nguyen-Tan PF, Le QT, Quivey JM, et al. Treatment results and prognostic factors of advanced T3-T4 laryngeal carci­noma: The University of California, San Francisco (UCSF) and Stanford University Hospital (SUH) experience. Int J Radiat Oncol Biol Phys 2001;50:1172–1180.

7. Hinerman RW, Mendenhall WM, Morris CG, et al. Post­operative irradiation for squamous cell carcinoma of the oral cavity: 35-year experience. Head Neck 2004;26:984 –