Squamous cell carcinoma of the oropharynx

Abstract

BACKGROUND

The treatment of patients with squamous cell carcinoma (SCC) of the oropharynx remains controversial. No randomized trial has addressed adequately the question of whether surgery (S), radiation therapy (RT), or combined treatment is most effective.

METHODS

Treatment results from North American academic institutions that used S with or without adjuvant RT (S ± RT) or used RT alone or followed by neck dissection (RT ± ND) for patients with SCC of the tonsillar region or the base of tongue were compiled through a MEDLINE search (from 1970 to August, 2000) and from the references cited in each report. Studies were eligible for inclusion if they contained direct, actuarial (life-table), or Kaplan–Meier calculations for the following end points: local control, local-regional control, 5-year absolute survival, 5-year cause specific survival, or severe or fatal treatment complications. Weighted average results, which took into account series size, were calculated for each end point for the purposes of treatment comparison. Results and conclusions were based on data from 51 reported series, representing the treatment of approximately 6400 patients from the United States and Canada.

RESULTS

The results for patients with SCC of the base of tongue who underwent S ± RT versus RT ± ND, respectively, were as follows: local control, 79% versus 76% (P = 0.087); local-regional control, 60% versus 69% (P = 0.009); 5-year survival, 49% versus 52% (P = 0.2); 5-year cause specific survival, 62% versus 63% (P = 0.4); severe complications, 32% versus 3.8% (P < 0.001); and fatal complications, 3.5% versus 0.4% (P < 0.001). The results for patients with SCC in the tonsillar region who underwent S ± RT versus RT ± ND, respectively, were as follows: local control, 70% versus 68% (P = 0.2); local-regional control, 65% versus 69% (P = 0.1); 5-year survival, 47% versus 43% (P = 0.2); 5-year cause specific survival, 57% versus 59% (P = 0.3); severe complications, 23% versus 6% (P < 0.001); and fatal complications, 3.2% versus 0.8% (P < 0.001).

CONCLUSIONS

The information in this article provides a useful benchmark for evidence-based counseling of patients with SCC of the oropharynx. The rates of local control, local-regional control, 5-year survival, and 5-year cause specific survival were similar for patients who underwent S ± RT or RT ± ND, whereas the rates of severe or fatal complications were significantly greater for the S ± RT group. Furthermore, available data on the functional consequences of treatment suggest the superiority of RT ± ND. The authors conclude that RT ± ND is preferable for the majority of patients with SCC of the oropharynx. Cancer 2002;94:2967–80. © 2002 American Cancer Society.

DOI 10.1002/cncr.10567

The oropharynx consists of the tonsillar region (pillars and fossae), base of tongue, soft palate, and posterior and lateral oropharyngeal walls. Carcinoma arising from these sites usually is squamous in origin and is related strongly to smoking and, to a lesser degree, alcohol ingestion. Most commonly, carcinoma affects patients in the fifth through seventh decades in life: The incidence in men is two to five times greater than the incidence observed in women. In 2001, there will be approximately 30,100 new diagnoses of oral or pharyngeal carcinoma in the United States, resulting in 7800 deaths. 1

The management of patients with primary oropharyngeal squamous cell carcinoma (SCC) remains controversial. 2-12 It is often stated that radiation therapy (RT) and surgery (S) are equally effective for the treatment of patients with early-stage disease, whereas a combination of the two modalities is the standard of care for patients with advanced disease, with the presumption that two radical treatments are more effective than one. There are insufficient randomized data that adequately address the question of RT alone versus combined-modality treatment. The only randomized trial was conducted by the Radiation Therapy Oncology Group. 13 Seventy patients with oropharyngeal carcinoma received definitive RT (24 patients), preoperative RT and S (23 patients), or S and postoperative RT (23 patients). There were no significant differences in the end results.

Tumor control and survival are the usual measures of treatment efficacy. For patients with SCC of the oropharynx, local (primary site) control is particularly relevant. If ultimate local control is not achieved, then the patient likely will suffer a miserable death, with intractable pain, bleeding, infection, malodor, trismus, inability to handle secretions, suffocation, and wasting—a scenario not appreciably different than when President Ulysses S. Grant died of tonsillar carcinoma 115 years ago. 14 For this reason, if it can be proven clearly that S, RT, or combined-modality treatment is the superior local treatment, then the finding would be of considerable significance.

In addition to tumor control and survival, therapies for patients with malignant disease are judged by the complications, morbidity, and mortality that they induce. The choice between S and RT is not trivial, because these treatments produce different toxicities and functional outcomes.

In the absence of randomized trials, treatment decisions are based on retrospective data. The purpose of this article was to review the North American experience with RT alone or followed by neck dissection (RT ± ND) or S with or without adjuvant RT (S ± RT) for patients with SCC of the base of tongue or tonsil. There were insufficient published data on patients with soft palate and pharyngeal wall tumors to make meaningful comparisons.

MATERIALS AND METHODS

Treatment results from North American academic institutions that used S ± RT or RT ± ND for patients with SCC of the tonsillar region or the base of tongue were compiled from a MEDLINE search (1970 to August 2000) and from references cited in each study. Studies were eligible for inclusion if they contained direct, actuarial (life-table), or Kaplan–Meier calculations for the following end points: local control, local-regional (primary site and neck) control, 5-year absolute or cause specific survival, or severe or fatal treatment complications. Few reports contained all information. All available data for each selected end point are presented. Some institutions have published results more than once; the dates of patient treatment are included for each such report. In some instances, the dates of patient inclusion overlapped slightly, indicating that some patients were reported more than once. Entries in each table are chronological, according to year of publication.

Included in the tables are the percentages of patients who had T4 tumors (advanced primary disease) or Stage IV tumors (advanced primary and/or neck disease). During the 30- year study period, the staging definitions have undergone several modifications. 15-17 In general, however, throughout the study period, the T4 designation usually implied the presence of one or more of the following findings: extension of primary tumor into soft tissues of the neck, trismus, tongue deviation or fixation (usually with dysarthria), bone destruction, or cranial nerve paralysis (the latter two are rare in patients with oropharyngeal tumors). Stage IV has generally meant T4 primary disease and/or the presence of lymph node metastases measuring > 3 cm, multiple lymph node metastases, or bilateral or contralateral lymph node involvement.

Weighted average results, which take into account series size, were calculated for ease of data comparison. Thus, the impact of each series was scaled in importance according to the number of patients treated. For each end point, a weighted average was calculated as follows: for each series, the number of patients treated was multiplied by the percentage of patients who satisfied the selected end point (for example, 5-year survival). The total number of patients who satisfied that end point from all series combined was then divided by the total number of all patients who were treated in the respective series. Series for which no data were available for a particular end point were excluded from both the numerator and the denominator. For comparison of two proportions, significance was calculated by a modified chi-square test. 18 Patients who developed local and/or regional recurrences were scored as failures even if, subsequently, they were salvaged successfully by S and/or RT. Patients who survived for 5 years were scored as survivors irrespective of their disease status at that time and irrespective of any prior failures and/or salvage procedures.

Serious complications of RT included conditions that required operation or hospitalization or that resulted in severe debility, including severe osteoradionecrosis, soft tissue necrosis, permanent gastrostomy, severe edema, serious neurologic injury, radiation-induced sarcoma, and complications of planned neck dissection or salvage surgery, when reported. Severe complications of S consisted of severe wound breakdown or fistula requiring surgical repair, osteomyelitis, permanent gastrostomy, permanent tracheostomy, chronic aspiration, carotid artery rupture, myocardial infarction, or other severe cardiopulmonary or central nervous system events. Results and conclusions are based on data from 51 reported series, representing treatment administered to approximately 6400 patients across the United States and Canada.

RESULTS

Base of Tongue

Local control

The local control rates in patients with SCC of the base of the tongue after undergoing S ± RT (390 patients; 13% had T4 lesions) or external-beam RT with or without an interstitial implant (806 patients; 13% had T4 lesions) are shown in Table 1. 3, 6, 19-31 The weighted-average local control rate was 79% in the S ± RT group compared with 76% in the RT alone group (P = 0.087) for all stages.

Table 1. Base of Tongue Carcinoma: Local Control According to T Stagea

Institution	No. of patients	T4 (%)	Boost technique	Local control (%)
				T1	T2	T3	T4	Overall
S with or without adjuvant RT
Washington University, St. Louis, MO (1983) 19	101	9	NA	ND	ND	ND	ND	74
Stanford University, Palo Alto, CA (1974–1982) (1985) 20	14	0	NA	ND	3/3	43	ND	60
M. D. Anderson Cancer Center, Houston, TX (1974–1984) (1990) 3	35	ND	NA	83		79		80
University of Pittsburgh, Pittsburgh, PA (1992) 6	13	0	NA	100	100	ND	ND	100
Memorial Sloan-Kettering Cancer Center, New York, NY (1973–86) 1992 21	31	39	NA	ND	84b	86b	84b	84b
Memorial Sloan-Kettering Cancer Center, New York, NY (1979–89) 1993 22	100	19	NA	ND	ND	ND	ND	82b
University of Pennsylvania, Philadelphia, PA (1997) 23	17	41	NA	77		80		77
Mayo Clinic, Rochester, MN (1971–1993) (1998) 24	79	0	NA	84	87	76	ND	82
Weighted average	390	13	—	—	—	—	—	79
RT with or without neck dissection
M. D. Anderson Cancer Center, Houston, TX (1954–1971) (1976) 25	174	17	EBRT	91	71	78	52	74
University of California, San Francisco, CA (1979) 26	48	10	EBRT	100	56	32	20	48
Stanford University, Palo Alto, CA (1974–1982) (1985) 20	14	14	^192Ir	ND	ND	ND	ND	86
M. D. Anderson Cancer Center, Houston, TX (1974–1984) (1990) 3	121	ND	EBRT	89		59	44	68
M. D. Anderson Cancer Center, Houston, TX (1984–1992) (1995) 27	54	2	EBRT	100	96	67	ND	85
Massachusetts General Hospital, Boston (b.i.d.) (1995) 28	90	0	EBRT	85	85	54	ND	ND
William Beaumont Hospital, Royal Oak, MI (1996) 29	20	25	^192Ir	100	89	100	80	88
Memorial Sloan-Kettering Cancer Center, New York, NY (1981–1995) (1998) 30	68	3	^192Ir	87	93	82	2/2	89
University of Florida, Gainesville, FL (2000) 31	217	19	EBRT	96	91	81	38	79
Weighted average	806	13	—	—	—	—	—	76

• NA: not applicable; ND: no data; ¹⁹²Ir: iridium 192 interstitial brachytherapy boost; EBRT: external beam radiation therapy; b.i.d.: twice a day; S: surgery; RT: radiation therapy.
• a Modified from Table 1 in Mendenhall et al., 2000. 31
• b First site of failure.

Local-regional control

The local-regional control rates in patients with SCC of the base of the tongue after undergoing S ± RT (370 patients) or RT ± ND (370 patients) are shown in Table 2. 17, 19, 20, 22-24, 27, 31-34 Although the RT series had a significantly greater proportion of patients with Stage IV disease (66% vs. 42%; P < 0.001), weighted-average, local-regional control rates for all stages were greater after RT ± ND (69%) than after S ± RT (60%; P = 0.009).

Table 2. Base of Tongue Carcinoma: Local-Regional Control According to the American Joint Committee on Cancer Staging Systema

Institution	No. of patients	Stage IV (%)	Local-regional control (%)
			Stage I	Stage II	Stage III	Stage IV	Overall
S with or without adjuvant RT
M. D. Anderson Cancer Center, Houston, TX (1964–1973) (1978) 32	34	76	ND	ND	ND	ND	56
Massachusetts General Hospital, Boston, MA (1962–1976) (1983) 33	25	40	67	80	29	20	40
Washington University, St. Louis, MO (1983) 19	101	45	100	75	67	40	57
Stanford University, Palo Alto, CA (1974–1982) (1985) 20	14	21	ND	100	33	33	43
Memorial Sloan-Kettering Cancer Center, New York, NY (1979–1989) (1993) 22	100	36	ND	ND	ND	ND	72b
University of Pennsylvania, Philadelphia, PA (1997) 23	17	59	ND	ND	ND	50	68
Mayo Clinic, Rochester, MN (1971–1993) (1998) 24	79	33	56	79	55	54	59
Weighted average	370	42	—	—	—	—	60
RT with or without neck dissection
Massachusetts General Hospital, Boston, MA (1962–1976) (1983) 33	29	52	ND	20	22	7	14
Memorial Medical Center, Long Beach, CA (1988) 34	70	57	ND	100	78	73	77
M. D. Anderson Cancer Center, Houston, TX (1984–1992) (1995) 27	54	63	ND	100	79	72	76
University of Florida, Gainesville, FL (2000) 31	217	71	100	100	83	65c	72
Weighted average	370	66	—	—	—	—	69

• ND: no data.
• a See American Joint Committee on Cancer, 1998. 17 Modified from Table 2 in Mendenhall et al., 2000.31
• b First site of failure.
• c Stage IVA, 64%; Stage IVB, 66%.

Absolute and cause specific survival

The absolute and cause specific survival rates in patients with SCC of the base of the tongue after undergoing S ± RT (500 patients) or RT ± ND (473 patients) are shown in Table 3. 19, 20, 22-24, 27, 30, 31, 34-37 Twice as many patients in the RT ± ND group had Stage IV disease (62% vs. 31%; P < 0.001). Despite this bias, the 5-year, weighted, cause specific survival rates were 62% in the S ± RT group compared with 63% in the RT ± ND group (P = 0.4). The 5- year absolute survival rates were 49% in the S ± RT group compared with 52% in the RT ± ND group (P = 0.2).

Table 3. Base of Tongue Carcinoma: Five-Year Survivala

Institution	No. of patients	T4 (%)	Stage IV (%)	Survival (%)
				Absolute	Cause specific
S with or without adjuvant RT
Mayo Clinic, Rochester, MN (1960–1967) (1972) 35	102	ND	7	44	ND
Washington University, St. Louis, MO (1983) 19	101	9	45	45	ND
Stanford University, Palo Alto, CA (1974–1982) (1985) 20	14	0	21	51	ND
Memorial Sloan-Kettering Cancer Center, New York, NY (1979–1989) (1993) 22	100	19	36	55	65
University of Pennsylvania, Philadelphia, PA (1997) 23	17	41	59	46 (3 yr)	ND
Mayo Clinic, Rochester, MN (1971–1993) (1998) 24	79	0	33	51	65
University of Pittsburgh, Pittsburgh, PA (1980–1987) (2000) 36	87	ND	ND	49	56
Weighted average	500	11	31	49	62
RT with or without neck dissection
Stanford University, Palo Alto, CA (1958–1980) (1983) 37	64	ND	50	35	ND
Memorial Medical Center, Long Beach, CA (1988) 34	70	17	57	35	60
M. D. Anderson Cancer Center, Houston, TX (1984–1992) (1995) 27	54	2	63	59	65
Memorial Sloan-Kettering Cancer Center, New York, NY (1981–1995) (1998) 30	68	3	51	87	ND
University of Florida, Gainesville, FL (2000) 31	217	19	71	50	64
Weighted average	473	14	62	52	63

• ND: no data.
• a Modified from Table 3 in Mendenhall et al., 2000. 31

Severe or fatal complications

The severe or fatal complication rates in patients with SCC of the base of the tongue after undergoing S ± RT (407 patients) or RT ± ND (842 patients) are shown in Table 4. 2, 3, 6, 19, 20, 22, 23, 25, 27, 29-32, 34, 38-40 Severe complications were noted in 32% of patients in the S ± RT group compared with 3.8% in the RT ± ND group (P < 0.001). Fatal complications occurred in 3.5% of patients in the S ± RT group compared with 0.4% of patients in the RT ± ND group (P < 0.001). Most of the excess mortality after S was reported in the early series.

Table 4. Base of Tongue: Severe Complicationsa

Institution	No. of patients	T4 (%)	Boost technique	Complications (%)
				Severe	Fatal
S with or without adjuvant RT
M. D. Anderson Cancer Center, Houston, TX (1964–1973) (1978) 32	34	41	NA	26	18
Indiana University, Indianapolis, IN (1983) 38	8	38	NA	38	12
Washington University, St. Louis, MO (1983) 19	101	9	NA	28	4
Stanford University, Palo Alto, CA (1985) 20	14	0	NA	64	0
M. D. Anderson Cancer Center, Houston, TX (1974–1984) (1990) 3	51	ND	NA	28	2
University of California, Los Angeles, CA (1990) 39	13	0	NA	23	0
University of Pittsburgh, Pittsburgh, PA (1992) 6	14	0	NA	0	0
Memorial Sloan-Kettering Cancer Center, New York, NY (1979–1989) (1993) 22	100	19	NA	ND	0
Mayo Clinic, Rochester, MN (1971–1986) (1993) 2	55	0	NA	49	4
University of Pennsylvania, Philadelphia, PA (1997) 23	17	41	NA	29	0
Weighted average	407	15	32	3.5
RT with or without neck dissection
Stanford University, Palo Alto, CA (1956–1973) (1976) 40	104	ND	EBRT	7	1
M. D. Anderson Cancer Center, Houston, TX (1954–1971) (1976) 25	174	17	EBRT	3	0
Stanford University, Palo Alto, CA (1974–1982) (1985) 20	14	14	^192Ir	7	0
Memorial Medical Center, Long Beach, CA (1988) 34	70	17	^192Ir	6	0
M. D. Anderson Cancer Center, Houston, TX (1974–1984) (1990) 3	121	ND	EBRT	2	0
M. D. Anderson Cancer Center, Houston, TX (1984–1992) (1995) 27	54	2	EBRT	0	0
William Beaumont Hospital, Royal Oak, MI (1996) 29	20	25	^192Ir	10	0
Memorial Sloan-Kettering Cancer Center, New York, NY (1981–95) (1998) 30	68	3	^192Ir	3	0
University of Florida, Gainesville, FL (2000) 31	217	19	EBRT	4	1
Weighted average	842	14	—	3.8	0.4

• NA: not applicable; ND: no data; EBRT: external beam radiation therapy; ¹⁹²Ir: iridium 192 interstitial brachytherapy boost; S: surgery; RT: radiation therapy.
• a Modified from Table 4 in Mendenhall et al., 2000. 31

Tonsillar Region

Local control

The local control rates for patients with SCC of the tonsillar region after undergoing S ± RT (406 patients; 12% had T4 lesions) or RT alone (1833 patients; 14% had T4 lesions) are shown in Table 5. 21, 28, 41-53 Local control rates were similar (70% after S ± RT vs. 68% after RT; P = 0.2).

Table 5. Tonsillar Squamous Cell Carcinoma: Local Control According to Tumor Statusa

Series	Treatment	No. of patients	T4 (%)	Local control (%)
				T1	T2	T3	T4	Overall
S with or without adjuvant RT
Memorial Sloan-Kettering Cancer Center, New York, NY (1973–1986) (1992) 21	S with or without RT	20	20	ND	4/4	100	50	89
Mayo Clinic, Rochester, MN (1970–1988) (1994) 41	S with or without RT	72	3	78	76	44	0/2	71
Roswell Park Cancer Institute, Buffalo, NY (1998) 42	S	56	ND	ND	ND	ND	ND	75
University of California Medical Center, Los Angeles, CA (1998) 43	S and RT	28	ND	80		61		64
Washington University Medical Center, St. Louis, MO (1998) 44	S and RT	230	14	80	71	65	58	68
Weighted average	406	12	—	—	—	—	70
RT with or without neck dissection
University of Michigan, Ann Arbor, MI (1955–1963) (1971) 45	RT	102	14	77		39	29	65
University of Michigan, Ann Arbor, MI (1960–1976) (1983) 46	RT	353	16	92	77	50	25	62
University of Maryland Hospital, Baltimore, MA (1984) 47	RT	185	14	94	80	51	19	58
Memorial Medical Center, Long Beach, CA (1985) 48	RT, ^192Ir	80	24	ND	ND	ND	ND	84
M. D. Anderson Cancer Center, Houston, TX (1968–1983) (1989) 49	RT	150	5	94	81	67	63	75
Mayo Clinic, Rochester, MN (1970–1988) (1994) 50	RT	21	43	ND	80	100	56	76
Massachusetts General Hospital, Boston, MA (b.i.d) (1995) 28	RT	102	0	91		80	ND	ND
University of Virginia, Charlottesville, VA (1995) 51	RT	53	0	83	78	ND	ND	81
Washington University Medical Center, St. Louis, MO (1998) 44	RT	154	23	76	63	59	33	56
University of California, Los Angeles, CA (1998) 43	RT	55	ND	65		31		44
Vancouver Cancer Center, Vancouver, BC (1999) 52	Ipsilateral RT	178	5	94	79	58	56	75
University of Florida, Gainesville, FL (2000) 53	RT	400	17	83	81	74	60	76
Weighted average	—	1833	14	—	—	—	—	68

• S: Surgery; RT: radiation therapy; ND: no data; ¹⁹²Ir: iridium 192 intestitial brachytherapy boost; b.i.d.: twice a day.
• a Modified from Table 2 in Mendenhall et al., 2000. 53

Local-regional control

The local-regional control rates for patients with SCC of the tonsillar region after undergoing S ± RT (281 patients) or RT ± ND (858 patients) are shown in Table 6. 41, 42, 48, 50, 52-57 A greater proportion of patients in the RT ± ND group had Stage IV disease (52% vs. 31%; P < 0.001). Despite this bias, local-regional control rates (65% after S ± RT; 69% after RT ± ND) were similar (P = 0.1).

Table 6. Tonsillar Squamous Cell Carcinoma: Local-Regional Control According to the American Joint Committee on Cancer Staging Systema

Series	Treatment	No. of patients	Stage IV (%)	Local-regional control (%)
				Stage I	Stage II	Stage III	Stage IV	Overall
S with or without adjuvant RT
University of California School of Medicine, San Francisco, CA (1986) 54	S with or without RT	40	ND	5/5	64	68		73
Memorial Sloan-Kettering Cancer Center, New York, NY (1969–1983) (1989) 55	S with or without RT	113	35	83		71	68	73
Mayo Clinic, Rochester, MN (1970–1988) (1994) 41	S with or without RT	72	25	73	69	53	56	63
Roswell Park Cancer Institute, Buffalo, NY (1998) 42	S	56	29	78		43		48
Weighted average	—	281	31	—	—	—	—	65
RT with or without neck dissection
University of Washington, Seattle, WA (1982) 56	RT	104	35	100	74	49	33	54
Memorial Medical Center, Long Beach, CA (1985) 48	R ^192Ir	80	49	3/3	100	85	56	75
Mayo Clinic, Rochester, MN (1970–1988) (1994) 50	RT	21	57	ND	ND	100	60	77
M. D. Anderson Cancer Center, Houston, TX (1983–1992) (1997) 57	RT	83	53	ND	76	65	87	77
Vancouver Cancer Center, Vancouver, BC (1999) 52	Ipsilateral RT	170	ND	92	75	55	59	ND
University of Florida, Gainesville, FL (2000) 53	RT	400	56	63	73	85	65, 52b	70
Weighted average	—	858	52	—	—	—	—	69

• S: surgery; RT: radiation therapy; ND: no data; ¹⁹²Ir: iridium 192 interstitial brachytherapy boost.
• a See American Joint Committee on Cancer, 1998. 17 Modified from Table 3 in Mendenhall et al., 2000.53
• b Stage IVA, 65%; Stage IVB, 52%.

Five-year absolute and cause specific survival

The 5-year absolute and cause specific survival rates for patients with SCC of the tonsillar region after undergoing S ± RT (321 patients; 44% had Stage IV disease) or RT ± ND (2276 patients; 49% had Stage IV disease) are shown in Table 7. 21, 40, 42, 45-47, 49, 50, 52, 53, 57-64 The absolute survival rates were 47% in the S ± RT group compared with 43% in the RT ± ND group (P = 0.2). The cause specific survival rates were 57% in the S ± RT group compared with 59% in the RT ± ND group (P = 0.3)

Table 7. Tonsillar Squamous Cell Carcinoma: Survivala

Series	Treatment	No. of patients	Stage IV (%)	Five-year survival (%)
				Absolute	Cause specific
S with or without adjuvant RT
University of Virginia Medical Center, Charlottesville, VA (1981) 58	S with or without RT	37	51	ND	54
Upstate Medical Center, Syracuse, NY (1982) 59	S and RT	47	45	ND	57
University of Cincinnati Medical Center, OH (1985) 60	S and RT	82	39	ND	56
East Orange VA Medical Center, NJ (1986) 61	S with or without RT	63	48	38	ND
Memorial Sloan-Kettering Cancer Center, New York, NY (1973–1986) (1992) 21	S and RT	20	65	54	ND
Mayo Clinic, Rochester, MN (1970–1988) (1994) 50	S and RT	16	56	74	ND
Roswell Park Cancer Institute, Buffalo, NY (1998) 42	S	56	29	ND	61
Weighted average	—	321	44	47	57
RT with or without neck dissection
University of Michigan, Ann Arbon, MI (1955–1963) (1971) 45	RT	102	ND	34	40
Stanford University, Palo Alto, CA (1956–1973) (1976) 40	RT	140	ND	33	ND
Yale University, New Haven, CT (1977) 62	RT	65	ND	37	ND
University of Virginia Medical Center, Charlottesville, VA (1981) 58	RT	45	51	ND	27
University of Michigan, Ann Arbor, MI (1960–1976) (1983) 46	RT	353	31	53	ND
University of Maryland Medical Center, Baltimore, MD (1984) 47	RT	185	46	30	42
Princess Margaret Hospital, Toronto, Ontario (1985) 63	RT	372	54	38	54
East Orange VA Medical Center, NJ (1986) 61	RT	81	60	20	ND
M. D. Anderson Cancer Center, Houston, TX (1968–1983) (1989) 49	RT	150	ND	ND	70
Stanford University, Palo Alto, CA (1975–1990) (1993) 64	RT	37	51	63	ND
Mayo Clinic, Rochester, MN (1970–1988) (1994) 50	RT	21	57	31	77
M. D. Anderson Cancer Center, Houston, TX (1983–1992) (1997) 57	RT	83	53	60	71
Vancouver Cancer Center, Vancouver, BC (1999) 52	RT	242	ND	48	62
University of Florida, Gainesville, FL (2000) 53	RT	400	56	49	70
Weighted average	—	2276	49	43	59

• S: surgery; RT: radiation therapy; ND: no data.
• a Modified from Table 4 in Medenhall et al., 2000. 53

Severe or fatal complication rates

The severe or fatal complication rates for patients with SCC of the tonsillar region after undergoing S ± RT (616 patients) or RT ± ND (2308 patients) are listed in Table 8. 41, 44, 46, 48, 50, 54-58, 61, 63-67 Severe complications (23% in the S ± RT group vs. 6% in the RT ± ND group; P < 0.001) and fatal complications (3.2% vs. 0.8%, respectively; P < 0.001) were both significantly more frequent in the S ± RT group.

Table 8. Tonsillar Squamous Cell Carcinoma: Severe Complicationsa

Institution	Treatment	No. of patients	Complications (%)
			Severe	Fatal
S with or without adjuvant RT
University of Cincinnati, Cincinnati, OH (1971) 65	Preop RT and S	32	34	0
University of Okalahoma, Oklahoma City, OK (1973) 66	Preop RT and S	30	ND	10
University of Virginia Medical Center, Charlottsville, VA (1981) 58	S with or without RT	52	23	ND
University of California Medical Center, San Francisco, CA (1986) 54	Preop RT and S	34	18	6
East Orange VA Medical Center, East Orange, NJ (1986) 61	Preop RT and S	49	31	8
Memorial Sloan-Kettering Cancer Center, New York, NY (1969–1983) (1989) 55	S with or without RT	117	18	0
Mayo Clinic, Rochester, MN (1970–1988) (1994) 41	S with or without RT	72	21	0
Washington University Medical Center, St. Louis, MO (1998) 44	Preop RT and S	144	24	5
S and Postop RT	86	24	2
Weighted average	616	23	3.2
RT with or without neck dissection
Massachusetts General Hospital, Boston, MA (1972) 67	RT	262	11	ND
University of Virginia Medical Center, Charlottsville, VA (1981) 58	RT	52	11	ND
University of Washington, Seattle, WA (1982) 56	RT	104	4	2
University of Michigan, Ann Arbor, MI (1963–1976) (1983) 46	RT	353	2	0.2
Memorial Medical Center, Long Beach, CA (1985) 48	RT, ^192Ir	80	3	0
Princess Margaret Hospital, Toronto, Ontario (1985) 63	RT	372	3	0
East Orange VA Medical Center, East Orange, NJ (1986) 61	RT	81	22	9
University of California School of Medicine, San Francisco, CA (1986) 54	RT	131	5	ND
Stanford University, Palo Alto, CA (1975–1990) (1993) 64	RT	37	5	0
Mayo Clinic, Rochester, MN (1994) 50	RT	21	5	0
M. D. Anderson Cancer Center, Houston, TX (1983–1992) (1997) 57	RT	83	2	0
Washington University Medical Center, St. Louis, MO (1998) 44	RT	154	10	1
Vancouver Cancer Center, Vancouver, BC (1999) 52	Ipsilateral RT	178	3	0
University of Florida, Gainesville, FL (2000) 53	RT	400	8	1
Weighted average	—	2308	6	0.8

• S: surgery; RT: radiation therapy; Preop: preoperative; Postop: postoperative; ND: no data; ¹⁹²Ir: iridium-192 interstitial brachytherapy boost.
• a Modified from Table 5 in Mendenhall et al., 2000. 53

Additional Analyses

Considering the rather wide range of reported outcomes, especially in the early years, we suspect some heterogeneity of treatment technique and/or patient selection among the various institutions. To minimize the impact of including data that may be considered unrepresentative of the current state of the art, each end point was analyzed four additional ways for each treatment type: 1) analysis limited to the three largest patient series, 2) analysis limited to the three series reporting the most favorable results, 3) analysis limited to the three most recent series, and 4) analysis limited to eight U.S. institutions widely recognized for expertise in the management of patients with carcinoma of the head and neck that have published their data on multiple occasions and that represent all geographic regions of the United States (Memorial Sloan-Kettering Cancer Center, New York, NY; Stanford University, Palo Alto, CA; M. D. Anderson Cancer Center, Houston, TX; University of Florida, Gainesville, FL; University of California, San Francisco, CA; Mayo Clinic, Rochester, MN; Washington University, St. Louis, MO; and Massachusetts General Hospital, Boston, MA). Reanalysis by these methods resulted in no change in any conclusion and provided reassurance that the conclusions were not distorted or diluted by the inclusion of heterogeneous, unrepresentive patient series. Whenever significant differences were noted between S and RT, as noted below, the differences were always in favor of the RT ± ND group compared with the S ± RT group.

Table 9 shows local control, local-regional control, survival, and complication data for the three largest RT series and the three largest S series for each end point for which data were available on both base of tongue carcinoma and tonsillar carcinoma. Of 12 data comparisons between S ± RT and RT ± ND, results significantly favored RT ± ND in five instances and S ± RT in none (P < 0.05).

Table 9. Local Control, Local-Regional Control, Survival, and Complication Rates for the Three Largest Radiation Therapy and Surgical Series

Variable	S with or without RT		RT with or without ND		P value
	No. of patients	%	No. of patients	%
Local control
Base of tongue	280	79	512	75	0.14
Tonsil	358	70	938	67	0.36
Local-regional control
Base of tongue	280	63	337	74	0.002
Tonsil	241	64	587	68	0.289
Five-year survival
Base of tongue
Absolute	303	48	355	54	0.11
Cause specific	266	58	341	63	0.17
Tonsil
Absolute	99	47	1125	47	0.86
Cause specific	185	58	1014	62	0.26
Complications
Base of tongue
Severe	207	34	512	3.2	< 0.0001
Fatal	256	2.4	512	0.4	0.02
Tonsil
Severe	419	22	1125	4.4	< 0.0001
Fatal	419	2	1125	0.4	0.0036

• S. surgery; RT. radiation therapy; ND. neck dissection.

Table 10 shows local control, local-regional control, survival, and complication data for the three most recently reported RT series and the three most recently reported S series for each end point for which data were available for both base of tongue carcinoma and tonsillar carcinoma. Of 12 data comparisons, results significantly favored RT ± ND in five instances and S ± RT in none.

Table 10. Local Control, Local-Regional Control, Survival, and Complication Rates for the Three most Recent Radiation Therapy and Surgical Series

Variable	S with or without RT		RT with or without ND		P value
	No. of patients	%	No. of patients	%
Local control
Base of tongue	196	82	305	82	0.92
Tonsil	314	69	633	73	0.32
Local-regional control
Base of tongue	196	66	337	75	0.046
Tonsil	241	64	498	72	0.028
Five-year survival
Base of tongue
Absolute	266	52	339	59	0.080
Cause specific	266	62	341	63	0.74
Tonsil
Absolute	99	47	725	50	0.65
Cause specific	185	58	725	67	0.022
Complications
Base of tongue
Severe	86	37	305	4.2	< 0.0001
Fatal	172	1.3	305	0.7	0.62
Tonsil
Severe	419	22	732	7.2	< 0.0001
Fatal	419	2	732	0.7	0.06

• S. surgery; RT. radiation therapy; ND. neck dissection.

Table 11 shows local control, local-regional control, survival, and complication data for the three RT series that reported the most favorable results for each endpoint for which data were available compared with the same data for the most favorable S series for both base of tongue carcinoma and tonsillar carcinoma. Of 12 data comparisons, results significantly favored RT ± ND in six instances and S ± RT in none.

Table 11. Local Control, Local-Regional Control, Survival, and Complication Rates for the Three Radiation Therapy and Surgical Series Reporting the Most Favorable Results

Variable	S with or without RT		RT with or without ND		P value
	No. of patients	%	No. of patients	%
Local control
Base of tongue	223	83	102	88	0.005
Tonsil	148	75	154	83	0.11
Local-regional control
Base of tongue	218	65	337	75	0.014
Tonsil	225	70	184	76	0.006
Five-year survival
Base of tongue
Absolute	193	53	331	59	0.115
Cause specific	266	62	341	63	0.74
Tonsil
Absolute	99	47	473	55	0.175
Cause specific	185	58	641	70	0.0016
Complications
Base of tongue
Severe	61	19	349	2.2	< 0.0001
Fatal	145	0	191	0.0	1.0
Tonsil
Severe	223	19	1065	2.6	< 0.0001
Fatal	223	0	771	0.0	1.0

• S: surgery; RT: radiation therapy; ND: neck dissection.

Table 12 shows local control, local-regional control, survival, and complication data for eight geographically distributed cancer centers in the United States that are recognized widely for their expertise in the management of patients with carcinoma of the head and neck. Not all centers reported results for each end point. Some centers reported data more than once. All available reports were used to compile final results for both base of tongue carcinoma and tonsillar carcinoma. Of 11 data comparisons, results significantly favored RT ± ND in four instances and S ± RT in none.

Table 12. Local Control, Local-Regional Control, Survival, and Complication Rates for Eight U.S. Hospitals

Variable	S with or without RT		RT with or without ND		P value
	No. of patients	%	No. of patients	%
Local control
Base of tongue	360	79	696	75	0.19
Tonsil	322	70	725	72	0.57
Local-regional control
Base of tongue	353	60	300	67	0.056
Tonsil	225	70	504	71	0.651
Five-year survival
Base of tongue
Absolute	396	49	403	55	0.085
Cause specific	179	65	271	64	0.89
Tonsil
Absolute	36	63	681	47	0.051
Cause specific	n.d.	n.d.	654	70	n.d.
Complications
Base of tongue
Severe	255	34	752	3.5	< 0.0001
Fatal	355	3.8	752	0.4	< 0.0001
Tonsil
Severe	453	22	1088	8.0	< 0.0001
Fatal	453	2.4	695	0.8	0.034

• S: surgery; RT: radiation therapy; ND: neck dissection n.d. no data.

DISCUSSION

This report presents a retrospective, nonrandomized, multi- institutional, pooled analysis of therapeutic results from over 6000 patients from U.S. and Canadian academic institutions who were treated with RT ± ND or S ± RT for SCC of the base of the tongue or the tonsillar region. There are problems in comparing data from nonrandomized patient series, including the influence of patient selection as well as differences in staging systems, treatment delivery, and end results reporting. Nevertheless, because there are no adequate randomized data, we believe that this information represents a useful benchmark for evidence-based counseling of patients with SCC of the oropharynx, and we also believe that this is the first serious attempt to compile all evidence, both favorable and unfavorable, relative to the efficacy, complications, and morbidity of S or RT.

The reality for patients with oropharyngeal carcinoma is that treatment selection often depends on where the patient is referred: Patients who are referred to institutions that favor RT usually receive that treatment, whereas patients who are referred to institutions that favor S undergo an operation. The usual basis for such decision making is institutional preference, prior training, recent experience (positive or negative) with a similar patient, or a recent journal article. Although some statisticians may challenge the methods and conclusions of the current article and their utility as a means of treatment selection, probably few would argue that the aforementioned institutional bias justifications for recommending one treatment over another are more valid.

The number of publications in the field of oncology is increasing rapidly. Most clinicians cannot remain abreast of all relevant information, which is scattered widely through the medical literature and often is fragmented, confusing, or biased. Summary articles traditionally are informal reviews that reflect one or another school of thought rather than an entire data base, as presented herein. Although randomized trials or meta-analyses are the gold standards for guiding treatment, literature reviews like the current report have considerable potential to improve clinical practice when no randomized trials exist. We believe it is unlikely that the questions addressed in this article will ever be studied in a randomized trial.

The most striking findings in this review were 1) the marked similarities in rates of local control, local-regional control, absolute survival, and cause specific survival between the two treatment modalities and 2) the marked disparity in rates of severe or fatal complications, both of which occurred four to nine times more frequently after S ± RT than after RT ± ND. Tables 4 and 8 summarize the rates of readily measurable complications, such as osteoradionecrosis, soft tissue necrosis, carotid artery rupture, and pharyngocutaneous fistula. The tables do not address the much more difficult to quantify frequencies of other types of functional or cosmetic losses.

The individual and collective morbidities of S and RT are difficult to categorize and transcend any single definition. Physicians generally define complications as unexpected adverse sequelae. To the patient, however, some expected sequelae, although not statistically scored as complications, may be equally debilitating. Successful treatment by S or by RT may produce very different end results as perceived by the patient. For example, a patient with T4 base of tongue carcinoma who has undergone total glossectomy, total laryngectomy, bilateral ND, and postoperative RT who heals uneventfully and experiences no postirradiation complication is scored statistically as complication free, despite suffering considerable physiologic, psychologic, and socioeconomic injury. The same patient treated with RT ± ND remains more intact anatomically and, assuming treatment is successful and free of complications, retains better function. Conversely, a patient with a T1 or superficial T2 tonsillar region tumor with minimal or absent lymph node metastasis may have a better functional outcome after S plus ipsilateral ND than with high-dose RT, thus avoiding the expected sequelae of xerostomia.

Quality-of-life outcomes are critical measures of a patient's well being that should be included in comprehensive assessments of treatment outcome. Functional losses may result from anatomic deficits caused by resection of tissues or from xerostomia, taste changes, fibrosis, pain, or trismus, all of which affect ability to eat and communicate. In modern surgical series for base of tongue carcinoma, 15–20% of patients have required total laryngectomy; 2, 3, 22 15–20% of patients have required mandibulectomy; 3, 22, 23 and 10– 25% of patients have required permanent gastrostomy, tracheostomy, or both. 2, 7, 23 Approximately 15–20% of patients with tonsillar carcinoma who undergo S require segmental mandibulectomy. 41

There is a paucity of literature on functional outcome in patients who are treated for oropharyngeal malignancies. Harrison et al. 68 performed detailed functional assessments on 30 patients who underwent external plus interstitial RT ± ND at Memorial Sloan-Kettering Cancer Center for base of tongue carcinoma using the performance scale of List et al. 69 Patients analyzed their ability to eat in public, understandability of speech, and normalcy of diet (scoring range, 0–100, with 100 representing no impairment). Although the assessment was subjective, it provided important information about patient perception of outcome. Mean scores were 83%, 93%, and 75%, respectively. The latter score was related mainly to xerostomia, an expected sequelae of RT. Seven of eight patients with T3 tumors rated understandability of speech at 100%, a result difficult to achieve with S, which would require at least partial glossectomy. In a separate article, Harrison et al. 70 noted that the majority of working patients maintained their prediagnosis employment status and earning potential after RT. Horwitz et al., 29 using the same performance status scale, noted mean scores of 89% and 94% for understandability of speech and normalcy of diet, respectively, in an analysis of 20 patients who were treated with external plus interstitial RT at William Beaumont Hospital (Royal Oak, MI).

At the University of Florida, Moore et al. 71 used the same performance scale to analyze functional outcome in 49 patients with base of tongue carcinoma who were alive and free of disease 2–23 years after external beam RT ± ND. Functional results were similar to those reported by Harrison et al. 68 and Horwitz et al.29 after external plus interstitial RT. Function after both types of RT was superior to reported functional results after S and postoperative RT at Memorial Sloan-Kettering Cancer Center; 72 this was true for both patients with early-stage disease (T1–T2) and patients with advanced-stage disease (T3–T4). In all three RT series, there were nonsignificant trends toward declining function with increasing T stage; for patients who underwent S and postoperative RT at Memorial Sloan-Kettering Cancer Center, functional status deteriorated markedly when comparing patients who had T1–T2 disease with patients who had T3–T4 disease (P = 0.0014). This is consistent with the fact that larger tumors require more extensive operations. Because RT does not involve tissue resection, the observed functional decline with increasing T stage suggests more extensive tumor-related tissue destruction and scar formation as the reasons. Functional results after RT at the University of Florida did not deteriorate with follow-up beyond 5 years. Planned postradiotherapy ND had no impact on function scores.

Nisi et al. 73 performed the same functional analysis on 21 patients (S alone, 13 patients; S and postoperative RT, 8 patients) who underwent operation at the Mayo Clinic. Eighty-one percent of patients had early-stage disease (T1–T2). For patients who underwent S alone, performance scores were similar to those reported after RT. Patients who underwent S and received postoperative RT had worse function. In separate analyses, the same authors 2, 24 showed that S alone seldom was applicable to this patient group, because local-regional failure rates, even for patients with early-stage disease (clinical Stage I–II), were 40% when postoperative RT was not administered, rendering the functional results of S alone largely irrelevant. We are not aware of detailed quality-of-life assessments after treatment for tonsillar carcinoma.

Functional impairment caused by resection of large oropharyngeal lesions represents a significant challenge, despite the sophistication of modern surgical reconstruction. Regional lymphatic metastasis from these tumors is often bilateral, at multiple levels, or massive. The morbidities from extensive neck dissection(s) and postoperative RT in such patients invariably require further compound rehabilitative and reconstructive challenges. For patients with base of tongue carcinoma (which has an 85% incidence of lymph node metastases and a 35% incidence of bilateral nodal involvement at diagnosis), treatment limited to the ipsilateral neck represents incomplete therapy. 24 The ease and success of bilateral neck irradiation is another factor that favors RT for these patients.

For patients with tonsillar carcinoma, the majority of lymph node metastasis in early-stage disease is to the ipsilateral neck. T1 and superficial T2 lesions are suitable for S and ipsilateral ND with little or no functional impairment. If neck disease is minimal or absent, then RT may be avoided. S for patients with more advanced lesions requires complex reconstruction, 10 does not eliminate the need for postoperative RT, does not increase the cure rate compared with RT ± ND, is associated with a greater incidence of complications, and, we believe, results in less satisfactory function than RT ± ND.

In summary, this review does not support the claim that S ± RT optimizes the chances for the successful treatment of patients with oropharyngeal carcinoma. For no stage or site of disease did S ± RT produce clearly superior rates of tumor control or survival compared with RT ± ND. The incidences of severe and/or fatal complications were several-fold greater in patients who underwent operation. Furthermore, although post- treatment function has been studied incompletely, available data suggest the superiority of RT ± ND. These findings lead us to conclude that RT ± ND is preferable to S ± RT for most patients with SCC of the oropharynx.

The authors hasten to mention that the results and conclusions of the current article, which addresses only patients with carcinoma of the oropharynx, are not generalizable to patients with carcinoma at other head and neck sites. Different conclusions may be reached in analyses of other anatomic regions because rates of local control, long-term morbidity, and function after S or RT differ for each anatomic region based on different tumor sensitivities, anatomic constraints, and functional anatomy. For example, it is the authors' opinion that most patients with tumors of the oral cavity (e.g., tumors of the oral tongue, floor of mouth, buccal mucosa, retromolar trigone, alveolar ridge, and hard palate) are managed better by S ± RT than RT ± ND. Although local control rates for patients with T1–T2 tumors of the floor of the mouth or the oral tongue who were treated with primary RT were similar to the rates achieved by patients who underwent S, we believe that long-term complication rates (mostly bone and/or soft tissue necrosis) were greater after high-dose RT (which often involves an interstitial implant) than after S ± RT. For patients with tumors at other oral cavity sites (e.g., tumors of the alveolar ridge, retromolar trigone, buccal mucosa, and hard palate), the local control rates after RT ± ND were inferior to those after S ± RT.

Just as surgical options and ingenious reconstructive techniques have evolved over the years, so have radiotherapeutic options. Increasingly, patients with early-stage tonsil carcinoma are treated with ipsilateral RT techniques that limit morbidity, especially xerostomia, by using mixed-beam, wedge pair, three-dimensional conformal or intensity-modulated RT techniques. Hyperfractionation 74, 75 and accelerated fractionation have been proven superior to conventional fractionation in randomized trials. Neoadjuvant chemotherapy before RT is an effective organ- sparing strategy for patients with SCC of the larynx 76 and hypopharynx 77, 78 and showed promise as an organ-sparing tool in a retrospective analysis of patients with T4 oropharyngeal SCC at the University of Florida. 79 Concurrent chemotherapy and RT regimens have consistently produced higher rates of both local-regional control and survival compared with the rates produced by RT alone in randomized trials 80-85 and hold promise for continued improvement in both tumor control and functional outcome.