Neck Dissection and Survival Among Head and Neck Cancer Patients Undergoing Adjuvant Immunotherapy

ABSTRACT

Background

Some studies suggest that neck dissection (ND) should be avoided in candidates for immunotherapy because lymph nodes are primary sites for immunotherapy activation. Our study investigates ND utilization and associated differences in overall survival (OS) among patients with head and neck cancer (HNC) undergoing adjuvant immunotherapy.

Methods

The 2013–2018 National Cancer Database was retrospectively reviewed for patients with HNC undergoing surgery with curative intent, and adjuvant immunotherapy. Multivariable binary logistic and Cox regression models adjusted for patient demographics, clinicopathologic features, and treatment.

Results

Of 1335 patients satisfying inclusion criteria, 679 (50.9%) patients underwent ND: 94 (13.8%) had pN0, 109 (16.1%) had pN1, 411 (60.5%) had pN2, 60 (8.8%) had pN3, and 5 (0.7%) had pNx classification. On multivariable binary logistic regression, academic treatment facility, cT4, and cN1–3 classification were associated with higher odds of undergoing ND (p < 0.05); salivary, sinonasal, oropharyngeal, hypopharyngeal, and laryngeal primary sites were associated with decreased odds (p < 0.05). Compared with those undergoing neck observation, patients undergoing ND had worse OS (49.4% vs. 61.5%, p < 0.001) on Kaplan–Meier but not multivariable Cox (adjusted hazard ratio [aHR] 1.00, 95% confidence interval [CI] 0.82–1.24, p = 0.968) regression. Compared with adjuvant immunotherapy alone, the addition of radiotherapy (aHR 0.64, 95% CI 0.44–0.93) and chemoradiotherapy (aHR 0.56, 95% CI 0.37–0.86) were associated with higher OS (p < 0.025).

Conclusion

ND was utilized in approximately 51% of patients with HNC undergoing adjuvant immunotherapy. ND was not associated with worse OS, possibly related to the high rate of pN1–3 classification.

Level of Evidence

4

Some studies suggest that neck dissection (ND) should be avoided in candidates for immunotherapy because lymph nodes are primary sites for immunotherapy activation. Our study indicates that ND was utilized in approximately 51% of patients with head and neck cancer undergoing adjuvant immunotherapy. ND was not associated with worse overall survival, possibly related to the high rate of pN1–3 classification.

1. Introduction

Clinical evidence of lymph node (LN) metastasis warrants consideration of neck dissection (ND) in head and neck cancer (HNC) management because LN metastasis portends poor locoregional control, recurrence, distant metastasis, and worse survival [1, 2, 3, 4]. Increasing evidence supports consideration of ND for the clinically uninvolved neck because physical examination and imaging modalities lack sufficient sensitivity and accuracy in detecting occult nodal metastasis [5].

The recent advent of immunotherapies targeting cytotoxic T‐lymphocyte antigen 4, programmed cell death ligand‐1, tumor antigens, and cell cycle checkpoints has impacted HNC management, especially in the neoadjuvant setting, early in the disease course [6, 7]. Emerging literature suggests that ND should be avoided in candidates for immunotherapy because tumor‐draining LNs are primary sites for antigen‐presenting cell differentiation, T‐cell priming, and epitope spreading [3, 8, 9, 10, 11, 12, 13, 14]. Interestingly, excising uninvolved LNs is also reported to blunt responses to immunotherapy [15, 16]. Given that immunotherapy is reliant on host immune cells, antigen processing within LNs, and lymphatic vessels, our study of the National Cancer Database (NCDB) investigates ND utilization and associated differences in overall survival (OS) among patients with HNC undergoing adjuvant immunotherapy [17, 18, 19]. To our knowledge, our study is also the first to query a large, nationally representative database exclusively for patients with HNC undergoing immunotherapy.

2. Methods

2.1. Data Source

The NCDB, a national, hospital facility‐based, comprehensive clinical surveillance resource established in 1989, is jointly sponsored by the American Cancer Society and the American College of Surgeons (ACS) Commission on Cancer (CoC), contains data derived from > 1500 CoC‐accredited hospitals, and captures > 70% of all newly diagnosed cancers in the United States annually [20]. The ACS has executed a Business Associate Agreement that includes a data use agreement with each of its CoC‐accredited facilities. This study was deemed exempt from review by the Rutgers New Jersey Medical School Institutional Review Board because of the de‐identified nature of patient data. The ACS and CoC have not verified and are not responsible for the validity of the statistical analysis or conclusions derived herein.

2.2. Inclusion Criteria

A retrospective review of the NCDB was performed for adults with a histologically confirmed diagnosis of HNC between January 2013 and December 2018 and treated with curative intent, including surgery and adjuvant immunotherapy (Figure 1). HNC was identified using the International Classification of Diseases for Oncology, Third Edition (ICD‐O‐3) histology (“8070–8072” for conventional squamous cell carcinoma [SCC] and all remaining histology codes for Other), behavior (“3” for malignant tumors), and topography codes (“C00,” “C02,” “C03,” “C04,” “C05,” and “C06” for oral cavity; “C07” and “C08” for major salivary glands; “C11,” “C30.0,” and “C31” for sinonasal tract; “C01,” “C02.4,” “C05.1,” “C05.2,” “C09,” and “C10” for oropharynx; “C12” and “C13” for hypopharynx; and “C32” for larynx). Patients with a history of prior malignancy; treatment with palliative intent; clinically distant metastasis; treatment without surgery and adjuvant immunotherapy; unknown ND; adjuvant radiotherapy (aRT) volume outside the head and neck; non‐therapeutic aRT dose (< 44 or > 80 Gy); unknown aRT modality, volume, or dose; initiation of aRT > 90 days following surgery; unknown chemotherapy; treatment with hormone therapy; neoadjuvant therapy; unknown sequence of surgery, radiotherapy, and chemotherapy; or unknown vital status or survival were excluded.

FIGURE 1.

Inclusion criteria. CM, clinical metastasis; NCDB, National Cancer Database.

2.3. Variables

Patient data included age at diagnosis, sex, race, reporting facility type, Charlson‐Deyo comorbidity score (CDCS), history of prior malignancy, histology, primary site, grade, clinical tumor‐nodal‐metastasis, pathologic tumor‐nodal‐metastasis classification, human papillomavirus (HPV), pathologic extranodal extension, lymphovascular invasion (LVI), ND, regional lymph node yield (LNY), surgical margins, surgical length of stay (LOS), 30‐day readmission and 90‐day mortality following surgery, adjuvant therapy, vital status, and survival time. Patients with a CDCS of 0 had no recorded comorbid conditions. Grade was classified as low (i.e., well or moderately differentiated), high (i.e., poorly differentiated, undifferentiated, or anaplastic), or unknown. Surgery included local tumor destruction, local tumor excision, partial removal, total removal, radical removal, and unspecified surgery. ND was defined as the removal and examination of $\ge$ 18 LNs, a validated threshold in HNC (Figure 2) [21, 22, 23, 24]. aRT was defined as the delivery of therapeutic doses (44–80 Gy) of external beam radiation to the head and neck within 90 days of surgery. aCRT was defined as the delivery of any chemotherapy, regardless of the type or number of agents, within 14 days of aRT initiation. The NCDB directly defines immunotherapy (i.e., biologic response modifier) as biologic or chemical agents that alter the immune system or change the host response to tumor cells; starting in 2013, at the beginning of our study period, 6 medications previously classified as chemotherapy were reclassified as immunotherapy: alemtuzumab, bvacizumab, rituximab, trastuzumab, pertuzumab, and cetuxumab. The primary outcome of our study was 5‐year OS. Survival time was calculated as the time from diagnosis to either death or 5 years of follow‐up.

FIGURE 2.

Extent of regional lymph node examination.

2.4. Statistical Analysis

The chi‐square and Mann–Whitney U tests were utilized to compare categorical and continuous variables, respectively. A multivariable binary logistic regression model handling missing data with listwise elimination and adjusting for all significant variables in univariable regression was implemented to identify factors associated with undergoing ND. Kaplan–Meier analysis was performed with the log‐rank test to estimate 5‐year OS. Multivariable Cox proportional hazards regression models handling missing data with listwise elimination included all significant variables from univariable Cox regression. The proportionality of hazards was evaluated using time‐dependent covariables and was not violated. To account for nonlinear behavior of LNY, follow‐up loss, immortal time bias, and the possibly increased radiosensitivity of advanced tumors with poor prognosis, the following 3 sensitivity analyses were conducted: (1) analyzing LNY as a categorical variable, (2) censoring survival > 2 and > 3 years, and (3) sequential landmark survival analysis for patients surviving $\ge$ 6 and $\ge$ 12 months from diagnosis, as previously described [25]. The two‐sided threshold for statistical significance was set at p < 0.05. SPSS version 25 (IBM) and Microsoft Excel (Microsoft) were used for statistical analysis.

3. Results

3.1. Patient Demographics, Clinicopathologic Features, and Treatment

One thousand three hundred and thirty‐five patients satisfied the inclusion criteria. The median (interquartile range [IQR]) patient age at diagnosis was 63 (56–70) years. A high proportion of patients were male (N = 991, 74.2%), of White race (N = 1186, 89.2%), without comorbidities by the CDCS (N = 1073, 80.4%), and had conventional SCC (N = 1122, 84.0%) of the oropharynx (N = 583, 43.7%) classified as low‐grade (N = 516, 38.7%), cT2 (N = 446, 33.4%), and cN2 (N = 588, 44.0%) (Table 1). Most patients had negative surgical margins (N = 731, 62.5%) and underwent ND (N = 679, 50.9%). Immunotherapy + radiotherapy (N = 939, 70.3%) was the most common adjuvant therapy combination, followed by immunotherapy + chemoradiotherapy (N = 282, 21.1%), immunotherapy alone (N = 74, 5.5%), and immunotherapy + chemotherapy (N = 40, 3.0%). The median (IQR) time from surgery to adjuvant chemotherapy, aRT, and adjuvant immunotherapy was 47 (36–60), 47 (35–58), and 47 (34–63) days, respectively. The median (IQR) aRT dose was 60 (58–66) Gy. Following surgery, patients undergoing ND and neck observation had similar 30‐day readmission (N = 38 [5.8%] vs. N = 31 [4.9%], p = 0.467) and 90‐day mortality (N = 4 [0.6%] vs. N = 4 [0.6%], p = 0.959).

TABLE 1.

Patient demographics, clinicopathologic features, and adjuvant therapy by neck dissection, n (%).

	Neck observation	Neck dissection	p	Total
No.	656 (49.1)	679 (50.9)	—	1335
Age at diagnosis, median years (IQR)	63 (56–71)	62 (55–70)	0.139	63 (56–70)
Sex
Male	478 (72.9)	513 (75.6)	0.262	991 (74.2)
Female	178 (27.1)	166 (24.4)		344 (25.8)
Race
White	588 (90)	598 (88.5)	0.586	1186 (89.2)
Black	44 (6.7)	50 (7.4)		94 (7.1)
Other	21 (3.2)	28 (4.1)		49 (3.7)
Reporting facility type
Academic/research	234 (36.4)	363 (54.8)	< 0.001	597 (45.7)
Non‐academic/research	409 (63.6)	299 (45.2)		708 (54.3)
CDCS
0	543 (82.8)	530 (78.1)	0.093	1073 (80.4)
1	74 (11.3)	96 (14.1)		170 (12.7)
$\ge$ 2	39 (5.9)	53 (7.8)		92 (6.9)
Histology
Conventional SCC	528 (80.5)	594 (87.5)	< 0.001	1122 (84.0)
Other	128 (19.5)	85 (12.5)		213 (16.0)
Primary site
Oral cavity	97 (14.8)	293 (43.2)	< 0.001	390 (29.2)
Major salivary glands	35 (5.3)	57 (8.4)		92 (6.9)
Sinonasal tract	69 (10.5)	10 (1.5)		79 (5.9)
Oropharynx	362 (55.2)	221 (32.5)		583 (43.7)
Hypopharynx	19 (2.9)	11 (1.6)		30 (2.2)
Larynx	74 (11.3)	87 (12.8)		161 (12.1)
Grade
Low	229 (34.9)	287 (42.3)	0.013	516 (38.7)
High	219 (33.4)	215 (31.7)		434 (32.5)
Unknown	208 (31.7)	177 (26.1)		385 (28.8)
cT classification
1	126 (19.2)	99 (14.6)	< 0.001	225 (16.9)
2	230 (35.1)	216 (31.8)		446 (33.4)
3	124 (18.9)	92 (13.5)		216 (16.2)
4	80 (12.2)	173 (25.5)		253 (19.0)
x	96 (14.6)	99 (14.6)		195 (14.6)
cN classification
0	223 (34)	188 (27.7)	0.007	411 (30.8)
1	137 (20.9)	132 (19.4)		269 (20.1)
2	273 (41.6)	315 (46.4)		588 (44.0)
3	12 (1.8)	15 (2.2)		411 (30.8)
x	11 (1.7)	29 (4.3)		269 (20.1)
pENE
No	—	309 (49.0)	< 0.001	451 (52.9)
Yes	—	321 (51.0)		401 (47.1)
LVI
No	279 (75.6)	338 (55.4)	< 0.001	617 (63.0)
Yes	90 (24.4)	272 (44.6)		362 (37.0)
Surgical margins
Negative	236 (45.9)	495 (75.6)	< 0.001	731 (62.5)
Positive	278 (54.1)	160 (24.4)		438 (37.5)
No. regional lymph nodes examined, median (IQR)	62 (55–70)	0 (0–5)	< 0.001	20 (0–38)
Surgical outcomes
LOS, median days (IQR)	0 (0–5)	36 (27–52)	< 0.001	2 (0–7)
30‐day readmission	31 (4.9)	38 (5.8)	0.467	69 (5.3)
90‐day mortality	4 (0.6)	4 (0.6)	0.959	8 (0.6)
Adjuvant therapy
Immunotherapy	37 (5.6)	37 (5.4)	0.049	74 (5.5)
Immunotherapy + chemotherapy	14 (2.1)	26 (3.8)		40 (3.0)
Immunotherapy + RT	481 (73.3)	458 (67.5)		939 (70.3)
Immunotherapy + CRT	124 (18.9)	158 (23.3)		282 (21.1)
Adjuvant RT
Duration, median days (IQR)	49 (43–52)	45 (42–49)	< 0.001	46 (43–51)
Dose, median Gy (IQR)	66 (56–70)	60 (60–66)	< 0.001	60 (58–66)
Adjuvant therapy timing, median days (IQR)
Diagnosis to surgery	16 (0–34)	35 (22–50)	< 0.001	28 (10–44)
Surgery to chemotherapy	42 (33–56)	50 (40–62)	< 0.001	47 (36–60)
Surgery to radiotherapy	41 (33–55)	50 (41–62)	< 0.001	47 (35–58)
Surgery to immunotherapy	40 (28–59)	50 (40–65)	< 0.001	47 (34–63)

Note: Bold values are statistically significant (p < 0.05).

Abbreviations: CDCS, Charlson‐Deyo comorbidity score; CRT, chemoradiotherapy; cTN, clinical tumor‐nodal; IQR, interquartile range; LOS, length of stay; LVI, lymphovascular invasion; pENE, pathologic extranodal extension; RT, radiotherapy.

Categorizing age at diagnosis as 18–49 years (N = 137, 10.3%), 50–59 years (N = 362, 27.1%), 60–69 years (N = 460, 34.5%), 70–79 years (N = 285, 21.3%), and $\ge$ 80 years (N = 91, 6.8%), utilization of ND and immunotherapy + chemoradiotherapy was highest among age 18–49 years (Figure 3).

FIGURE 3.

Utilization of neck dissection and adjuvant therapy by age. CRT, chemoradiotherapy; ND, neck dissection; RT, radiotherapy.

On multivariable binary logistic regression, academic treatment facility, cT4, and cN1–3 classification were associated with higher odds of undergoing ND (p < 0.05); salivary, sinonasal, oropharyngeal, hypopharyngeal, and laryngeal primary sites were associated with decreased odds (p < 0.05) (Table 2).

TABLE 2.

Kaplan–Meier analysis of 5‐year OS (%) by neck dissection with log‐rank comparisons within strata.

	Neck observation	Neck dissection	p	Total
Overall	61.5	49.4	< 0.001	55.2
Histology
Conventional SCC	61.8	48.9	< 0.001	54.9
Other	60.0	52.4	0.320	56.8
Primary site
Oral cavity	34.2	39.3	0.695	38.1
Major salivary glands	51.7	49.5	0.908	50.3
Sinonasal tract	52.6	26.3	0.170	48.3
Oropharynx	78.2	76.3	0.811	77.5
Hypopharynx	31.3	—	0.892	25.5
Larynx	42.9	30.6	0.122	36.1
Grade
Low	56.8	46.2	0.027	50.8
High	68.0	53.1	0.003	60.3
Unknown	59.9	50.2	0.092	55.3
cT classification
1	76.9	63.5	0.037	70.9
2	66.2	60.9	0.417	63.6
3	49.0	28.5	0.003	39.5
4	32.3	39.1	0.260	37.0
x	74.1	53.3	0.005	63.3
cN classification
0	52.7	48.5	0.565	50.8
1	73.2	53.2	0.003	63.0
2	63.1	49.0	< 0.001	55.4
3	52.9	45.5	0.791	48.7
x	76.5	44.7	0.125	53.1
pENE
No	65.0	55.3	0.066	58.2
Yes	50.4	43.9	0.180	45.1
LVI
No	64.3	53.8	0.017	58.4
Yes	45.5	41.5	0.392	42.5
Surgical margins
Negative	63.4	51.1	0.002	55.0
Positive	61.3	41.9	0.001	53.9
Adjuvant therapy
Immunotherapy	45.5	27.9	0.054	36.2
Immunotherapy + chemotherapy	33.3	44.2	0.565	40.3
Immunotherapy + RT	62.8	49.6	< 0.001	56.2
Immunotherapy + CRT	65.0	55.0	0.050	59.2

Note: Bold values are statistically significant (p < 0.05).

Abbreviations: CRT, chemoradiotherapy; cTN, clinical tumor‐nodal; LVI, lymphovascular invasion; OS, overall survival; pENE, pathologic extranodal extension; RT, radiotherapy; SCC, squamous cell carcinoma.

3.2. 5‐Year OS

On Kaplan–Meier analysis, patients undergoing ND had worse 5‐year OS than those undergoing neck observation (49.4% vs. 61.5%, p < 0.001) (Figure 4, Table 3). The 5‐year OS for patients undergoing adjuvant immunotherapy alone, adjuvant immunotherapy + chemotherapy, adjuvant immunotherapy + radiotherapy, and adjuvant immunotherapy + chemoradiotherapy was 36.2%, 40.3%, 56.2%, and 59.2%, respectively (p < 0.001).

FIGURE 4.

Kaplan–Meier analysis of 5‐year overall survival for 656 patients undergoing neck observation and 679 patients undergoing neck dissection.

TABLE 3.

Univariable and multivariable binary logistic regression predicting preoperative factors associated with undergoing ND.

	OR (95% CI)	p	aOR a (95% CI)	p
Age at diagnosis (years)	0.99 (0.98–1.00)	0.099
Sex
Male	Ref
Female	0.87 (0.68–1.11)	0.262
Race
White	Ref
Black	1.12 (0.73–1.70)	0.605
Other	1.31 (0.74–2.33)	0.358
Reporting facility type
Academic	2.12 (1.70–2.65)	< 0.001	2.13 (1.66–2.73)	< 0.001
Non‐academic	Ref		Ref
CDCS
0	Ref
1	1.33 (0.96–1.84)	0.087
$\ge$ 2	1.39 (0.91–2.14)	0.132
Histology
Conventional SCC	Ref		Ref
Other	0.59 (0.44–0.80)	0.001	0.82 (0.55–1.23)	0.344
Primary site
Oral cavity	Ref		Ref
Major salivary glands	0.54 (0.33–0.87)	0.012	0.54 (0.31–0.94)	0.029
Sinonasal tract	0.05 (0.02–0.10)	< 0.001	0.05 (0.02–0.11)	< 0.001
Oropharynx	0.20 (0.15–0.27)	< 0.001	0.18 (0.12–0.25)	< 0.001
Hypopharynx	0.19 (0.09–0.42)	< 0.001	0.15 (0.06–0.35)	< 0.001
Larynx	0.39 (0.26–0.57)	< 0.001	0.34 (0.22–0.52)	< 0.001
Grade
Low	Ref		Ref
High	0.78 (0.61–1.01)	0.062	1.16 (0.86–1.57)	0.335
Unknown	0.68 (0.52–0.89)	0.004	1.07 (0.78–1.48)	0.673
cT classification
1	Ref		Ref
2	1.20 (0.87–1.65)	0.278	1.09 (0.76–1.55)	0.638
3	0.94 (0.65–1.38)	0.766	0.90 (0.58–1.41)	0.658
4	2.75 (1.89–4.00)	< 0.001	2.15 (1.35–3.41)	0.001
x	1.31 (0.89–1.93)	0.166	1.22 (0.80–1.87)	0.356
cN classification
0	Ref		Ref
1	1.14 (0.84–1.55)	0.395	1.78 (1.22–2.58)	0.003
2	1.37 (1.06–1.76)	0.015	2.13 (1.54–2.96)	< 0.001
3	1.48 (0.68–3.25)	0.324	2.59 (1.02–6.62)	0.046
x	3.13 (1.52–6.43)	0.002	4.93 (2.21–11.03)	< 0.001

Note: Bold values are statistically significant (p < 0.05).

Abbreviations: aOR, adjusted odds ratio; CDCS, Charlson‐Deyo comorbidity score; CI, confidence interval; cTN, clinical tumor‐nodal; OR, odds ratio; Ref, reference; SCC, squamous cell carcinoma.

^a N = 1305; total number of neck dissection events: 764.

On multivariable Cox regression, age at diagnosis, cT3–4, and cN2–3 classification were associated with worse OS (p < 0.05); salivary, oropharyngeal primary sites, immunotherapy + radiotherapy (adjusted hazard ratio [aHR] 0.64, 95% confidence interval [CI] 0.44–0.93), and immunotherapy + chemoradiotherapy (aHR 0.56, 95% CI 0.37–0.86) were associated with higher OS (p < 0.025) (Table 4). ND (aHR 1.00, 95% CI 0.82–1.24, p = 0.968) was not associated with OS.

TABLE 4.

Univariable and multivariable Cox proportional hazards regression models.

	HR (95% CI)	p	aHR a (95% CI)	p
Age at diagnosis (years)	1.03 (1.02–1.04)	< 0.001	1.02 (1.01–1.03)	< 0.001
Sex
Male	Ref
Female	1.14 (0.94–1.40)	0.187
Race
White	Ref		Ref
Black	1.48 (1.07–2.03)	0.016	1.03 (0.74–1.43)	0.867
Other	1.92 (1.28–2.87)	0.002	1.51 (1.00–2.28)	0.051
Reporting facility type
Academic	0.96 (0.80–1.15)	0.682
Non‐academic	Ref
CDCS
0	Ref		Ref
1	1.09 (0.84–1.40)	0.516	0.97 (0.75–1.25)	0.801
$\ge$ 2	1.43 (1.04–1.95)	0.026	1.25 (0.91–1.73)	0.171
Histology
Conventional SCC	Ref
Other	0.93 (0.73–1.19)	0.558
Primary site
Oral cavity	Ref		Ref
Major salivary glands	0.66 (0.47–0.94)	0.020	0.61 (0.42–0.88)	0.008
Sinonasal tract	0.73 (0.50–1.06)	0.096	0.70 (0.46–1.09)	0.113
Oropharynx	0.26 (0.20–0.33)	< 0.001	0.29 (0.22–0.39)	< 0.001
Hypopharynx	1.43 (0.89–2.29)	0.135	1.40 (0.86–2.29)	0.180
Larynx	1.05 (0.82–1.35)	0.702	0.91 (0.70–1.19)	0.510
Grade
Low	Ref		Ref
High	0.74 (0.60–0.92)	0.006	0.97 (0.78–1.22)	0.801
Unknown	0.85 (0.68–1.07)	0.164	0.85 (0.66–1.09)	0.203
cT classification
1	Ref		Ref
2	1.42 (1.03–1.98)	0.035	1.14 (0.82–1.60)	0.427
3	3.01 (2.15–4.21)	< 0.001	1.98 (1.39–2.82)	< 0.001
4	3.38 (2.44–4.69)	< 0.001	1.74 (1.23–2.48)	0.002
x	1.39 (0.95–2.04)	0.092	1.39 (0.93–2.08)	0.105
cN classification
0	Ref		Ref
1	0.65 (0.49–0.85)	0.002	1.03 (0.78–1.37)	0.830
2	0.87 (0.71–1.07)	0.178	1.59 (1.26–2.00)	< 0.001
3	1.05 (0.57–1.93)	0.875	1.99 (1.06–3.72)	0.032
x	1.04 (0.63–1.71)	0.889	1.28 (0.75–2.18)	0.370
Neck dissection
No	Ref		Ref
Yes	1.45 (1.21–1.74)	< 0.001	1.00 (0.82–1.23)	0.968
Surgical margins
Negative	Ref
Positive	1.02 (0.84–1.24)	0.858
Adjuvant therapy
Immunotherapy	Ref		Ref
Immunotherapy + chemotherapy	0.93 (0.54–1.59)	0.792	1.00 (0.57–1.77)	0.986
Immunotherapy + RT	0.48 (0.34–0.67)	< 0.001	0.64 (0.44–0.93)	0.018
Immunotherapy + CRT	0.44 (0.30–0.65)	< 0.001	0.56 (0.37–0.86)	0.007

Note: Bold values are statistically significant (p < 0.05).

Abbreviations: aHR, adjusted hazard ratio; CDCS, Charlson‐Deyo comorbidity score; CI, confidence interval; CRT, chemoradiotherapy; cTN, clinical tumor‐nodal; HR, hazard ratio; OS, overall survival; Ref, reference; RT, radiotherapy; SCC, squamous cell carcinoma.

^a N = 1329; total number of uncensored death events: 482.

3.3. pTN Classification

Among 656 patients undergoing neck observation, 166 (25.3%) had pT1, 151 (23.0%) had pT2, 86 (13.1%) had pT3, 83 (12.7%) had pT4, and 170 (25.9%) had pTx classification.

Among 679 patients undergoing ND, 119 (17.5%) had pT1, 199 (29.3%) had pT2, 124 (18.3%) had pT3, 223 (32.8%) had pT4, and 14 (2.1%) had pTx classification. Among these patients, 94 (13.8%) had pN0, 109 (16.1%) had pN1, 411 (60.5%) had pN2, 60 (8.8%) had pN3, and 5 (0.7%) had pNx classification.

3.4. Secondary Outcomes

Compared with neck observation, ND was associated with longer LOS (4.00 marginal days, 95% CI 3.26–4.75 days, p < 0.001) but similar 90‐day mortality (adjusted odds ratio 1.29, 95% CI 0.75–2.23, p = 0.355) on multivariable binary logistic regression adjusting for patient demographics and clinicopathologic features.

3.5. HPV and Oropharyngeal Cancer

Four hundred and seventy‐three patients with oropharyngeal cancer had known HPV status: 71 (15.0%) were HPV negative and 402 (85.0%) were HPV positive.

Among HPV‐negative patients, 41 underwent neck observation and 30 underwent ND; 5‐year OS was 67.7% and 51.0%, respectively, on Kaplan–Meier analysis (p = 0.181). ND was not associated with OS on multivariable Cox regression (aHR 2.25, 95% CI 0.68–7.49, p = 0.186).

Among HPV‐positive patients, 258 underwent neck observation and 144 underwent ND; 5‐year OS was 80.9% and 87.9%, respectively, on Kaplan–Meier analysis (p = 0.092). ND was not associated with OS on multivariable Cox regression (aHR 0.53, 95% CI 0.27–1.01, p = 0.054).

3.6. LNY

Instead of binarizing LNY at 18 LNs to define ND, LNY was categorized as 0 (N = 423, 32.0%), 1–5 (N = 98, 7.4%), 6–10 (N = 36, 2.7%), 11–15 (N = 63, 4.8%), 16–20 (N = 85, 6.4%), 21–25 (N = 91, 6.9%), 26–30 (N = 94, 7.1%), 31–35 (N = 86, 6.5%), 36–40 (N = 64, 4.8%), and $\ge$ 40 (N = 282, 21.3%) LNs to identify a possible inflection point; 5‐year OS was 61.6%, 71.2%, 64.3%, 44.6%, 53.4%, 52.4%, 48.6%, 55.6%, 53.4%, and 46.3%, respectively, on Kaplan–Meier analysis (p = 0.001). Compared with LNY of 0, LNY of 1–5 (aHR 0.68, 95% CI 0.42–1.10, p = 0.119), 6–10 (aHR 0.90, 95% CI 0.46–1.74, p = 0.750), 11–15 (aHR 1.29, 95% CI 0.83–1.99, p = 0.253), 16–20 (aHR 0.99, 95% CI 0.65–1.49, p = 0.957), 21–25 (aHR 1.04, 95% CI 0.69–1.55, p = 0.864), 26–30 (aHR 1.06, 95% CI 0.72–1.56, p = 0.780), 31–35 (aHR 0.92, 95% CI 0.61–1.40, p = 0.710), 36–40 (aHR 0.89, 95% CI 0.56–1.40, p = 0.615), and $\ge$ 40 (aHR 0.93, 95% CI 0.70–1.24, p = 0.622) LNs were all not associated with OS on multivariable Cox regression.

3.7. 2‐Year and 3‐Year OS

On Kaplan–Meier analysis, patients undergoing ND had worse 2‐year (70.5% vs. 79.5%) and 3‐year OS (61.9% vs. 72.6%) than those undergoing neck observation (p < 0.001). However, on multivariable Cox regression, ND was not associated with worse 2‐year (aHR 1.00, 95% CI 0.82–1.24, p = 0.968) or 3‐year OS (aHR 1.00, 95% CI 0.82–1.24, p = 0.968) than neck observation.

3.8. 6‐Month and 12‐Month Landmark Survival Analysis

41 (3.2%) and 208 (15.6%) patients had survival times < 6 and < 12 months, respectively. After censoring survival times < 6 months, ND was associated with worse 5‐year OS on Kaplan–Meier (50.8% vs. 63.8%) but not multivariable Cox regression (aHR 1.03, 95% CI 0.83–1.27, p = 0.816). After censoring survival times < 12 months, ND was also associated with worse 5‐year OS on Kaplan–Meier (60.7% vs. 72.2%, p < 0.001) but not multivariable Cox regression (aHR 1.10, 95% CI 0.85–1.44, p = 0.471).

4. Discussion

Clinical evidence of LN metastasis warrants consideration of ND when pursuing surgical management of HNC. Several randomized controlled trials (RCTs) have precipitated hope that immunotherapy may reduce LN metastasis, but response rates remain poor, especially in recurrent or distantly metastatic disease [3, 26, 27]. Increasing evidence suggests that LNs are important sites for immunotherapy activation, raising questions regarding the potential unintended adverse effects of ND [19, 28]. Our study investigating ND among patients with HNC undergoing adjuvant immunotherapy suggests that ND is frequently utilized and is not associated with worse OS.

A high proportion of patients (51%) in our cohort underwent ND. Academic treatment facility, cT4, and cN1–3 classification were associated with undergoing ND. Increased utilization of ND at academic facilities may be driven by access to inpatient hospital resources, tumor boards, and multidisciplinary specialists. Salivary, sinonasal, oropharyngeal, hypopharyngeal, and laryngeal primary sites were associated with undergoing neck observation. Importantly, the increased prevalence of multimorbidity, malnutrition, and poor performance status in elderly patients may contraindicate surgical interventions such as ND, but our study did not detect associations between older age, higher CDCS, and undergoing neck observation. Many mucosal head and neck primary sites have a high propensity for LN metastasis and warrant aggressive neck management, such as elective ND in the setting of cN0 disease to prophylactically remove clinically negative LNs that have a significant probability of harboring occult metastasis [29].

ND was not associated with worse OS after adjusting for patient demographics, clinicopathologic features, and treatment; sensitivity analysis categorizing LNY in increments of 5 LNs did not associate more extensive nodal dissections with OS. Anecdotal experiences of surgeons associating ND with worse survival may be driven by differences in baseline clinicopathologic features, as evidenced by patients in our cohort undergoing ND also having a higher cTN classification. These findings conflict with recent studies suggesting the possible benefit of LN preservation in improving immunotherapy response. LNs are primary sites for (1) migration of lymphocytes, neutrophils, and antigen‐presenting cells such as dendritic cells, and (2) the filtration of antigenic and infectious substances from peripheral tissues, and therefore have a unique role in facilitating the acquisition of adaptive anti‐tumor immunity [14, 16, 19, 30, 31, 32, 33, 34, 35, 36, 37, 38]. Patent lymphatic vessels, however, permit drainage and colonization of tumor cells in LNs, forming LN metastasis [14]. As tumor colonization progresses, immuno‐stimulation in LNs tends to shift to immunosuppression, contributing to tumor immune escape, a dynamic editing effect characterized by defective antigen presentation, T‐cell regulation, depleted T‐cell potency, and uncontrolled cell proliferation due to tumor‐specific immune tolerance [19, 39, 40, 41, 42, 43]. The existence of an immunosuppressed cellular niche in metastatic LNs is supported by studies demonstrating worse survival among patients found to have LN metastasis despite neoadjuvant immunotherapy [44]. Metastatic LNs likely exert immunosuppressive effects and have less value for preservation because unresected LN metastasis accelerates disease progression despite aggressive adjuvant therapy.

Compared with physical examination and imaging, pathologic evaluation with ND provides the most accurate diagnosis of LN metastasis [45]. Sparing patients with a low probability of LN metastasis from undergoing ND could theoretically preserve quality of life and optimize immunotherapy response. For example, ND is associated with increased medical costs, operative time, and aesthetic and functional morbidities such as scarring, shoulder weakness, lymphedema, and impaired sensation [46]. Another consideration for neck management is sentinel LN biopsy, which preserves the majority of the neck lymphatics while offering accurate pathologic staging. Of note, research on cancer phylogenies suggests that LN and distant metastasis may have independent clonal origins, and lymphatic vessels may not be the direct pathway for distant metastasis [47, 48]. Considering the important role of LNs in immunotherapy response, arguments favoring their preservation have become compelling [14]. Techniques capable of accurately diagnosing LN metastasis without extensive nodal dissections, such as those involving deep learning artificial intelligence, liquid biopsy of blood for tumor biomarkers, and nanoparticle lymphatic imaging, are currently under development to better identify candidates for neck observation [14, 49, 50, 51, 52].

Among the 4 adjuvant therapy combinations analyzed in our study, the addition of radiotherapy was associated with the highest OS. Radiotherapy is the most utilized adjuvant therapy in HNC because of its established and predictable safety profile, widespread availability, and potentially immunostimulatory effects by increased antigen presentation occasionally progressing to an abscopal effect [36]. Several RCTs investigating immunoradiotherapy combinations, however, have failed to demonstrate a therapeutic benefit compared with either modality alone, possibly reflecting suboptimal study design, choice of outcomes, and administration of radiotherapy according to standard schedules and target volumes [34, 35, 36, 53]; an ongoing RCT is expected to provide more conclusive evidence [54]. One important consideration in HNC is LN‐sparing aRT, which is associated with improved anti‐tumor responses both locally within the irradiated tumor and tumor‐draining LNs, and systemically within distant LNs and secondary tumors [18, 34, 35, 53, 54, 55, 56, 57, 58]. Studies documenting locoregional recurrence despite LN‐sparing aRT confirm the function of tumor‐draining LNs as supportive niches for cancer dissemination; for these patients, a viable strategy may involve delaying LN irradiation and ND for several weeks until immunotherapy can establish adaptive anti‐tumor immunity, and then initiating comprehensive adjuvant therapy to eradicate persistent malignancy and prevent clinicopathologic upstaging [35]. Another viable strategy is initiating immunotherapy before surgery, when the lymphatics are intact, such as in Keynote 689. Although adjuvant immunotherapy + chemotherapy has been associated with remission in several cancers, the addition of chemotherapy without radiotherapy did not improve OS in our cohort [59].

Older age, cT3–4, and cN2–3 classification were associated with worse OS, aligning with existent literature [60]. Major salivary gland and oropharyngeal primary sites were associated with higher OS, possibly related to changing demographics increasing the incidence of HPV‐positive oropharyngeal cancer, which tends to occur in younger, healthier patients and portends a more favorable response to treatment, as our data reflect [61, 62, 63, 64, 65, 66].

Limitations inherent in the retrospective study of the NCDB include the possibility of inaccurate histologic diagnosis and variable miscoding. The NCDB does not report specific medical comorbidities, tobacco use, tumor mutations, history of neck irradiation, imaging studies, depth of invasion, extent of LVI, metastatic volume, intraoperative tumor spillage, close margins, PNI, multidisciplinary tumor board recommendations, quality of life, locoregional recurrence, and disease‐free survival. The NCDB also does not report (1) whether surgical margins were obtained from the specimen or tumor bed, (2) whether radiotherapy only targeted tumor‐draining LNs or spared uninvolved LNs, (3) the number and type of immunotherapy agents administered, (4) whether immunotherapy and radiotherapy were administered concurrently or sequentially, and (5) treatment delays, interruptions, and extent of completion. Regarding treatment sequence, the NCDB does not differentiate between neoadjuvant systemic therapies (i.e., patients undergoing neoadjuvant immunotherapy cannot be distinctly identified from those undergoing neoadjuvant chemotherapy), limiting our analysis to the adjuvant setting, despite increasing evidence suggesting that immunotherapy has the most significant survival benefit if administered early in the disease course before surgery, radiotherapy, or pathologic evaluation of nodal metastasis can be attempted [6]. Nevertheless, ND is possibly less associated with neoadjuvant immunotherapy response than adjuvant immunotherapy response, especially if sufficient time elapses between neoadjuvant immunotherapy administration and surgery, supporting our exclusion of patients undergoing neoadjuvant systemic therapy. National Comprehensive Cancer Network recommendations for HNC do not include adjuvant immunotherapy, representing a selection bias. Many patients undergoing immunotherapy have distant metastasis, but these patients are typically not candidates for surgery with curative intent and were therefore excluded, representing another selection bias. In fact, < 10% of approximately 19,000 potential patients on the initial database query satisfied our strict inclusion criteria, because a high proportion of patients had unknown ND, treatment, or treatment sequence; our cohort, although large and drawn from a national network of CoC‐accredited facilities, may not be representative of all patients with HNC undergoing immunotherapy (Figure 1). A minority of patients underwent surgery without ND and without aRT (N = 51, 7.7%); our study is therefore unable to draw conclusions regarding immunotherapy in patients with intact lymphatics because ND and radiotherapy may ablate these structures.

5. Conclusion

Approximately 51% of patients with HNC undergoing surgery + adjuvant immunotherapy also underwent ND. Academic treatment facility, cT4, and cN1–3 classification were associated with higher odds of undergoing ND; salivary, sinonasal, oropharyngeal, hypopharyngeal, and laryngeal primary sites were associated with decreased odds. ND was not associated with worse OS, possibly related to the high rate of pathologic nodal metastasis and previous reports of tumor progression shifting LNs from a beneficial immune‐stimulatory to a harmful immunosuppressive phenotype. Although speculative, the benefit of neck observation in improving immunotherapy response likely occurs only if LNs are free of disease, but accurately identifying LN metastasis without ND remains challenging despite advances in diagnostic technologies. ND should be offered whenever indicated because the benefits of removing possibly metastatic LNs likely outweigh the theoretical risk of reducing the efficacy of adjuvant immunotherapy. A well‐designed RCT may elucidate optimal neck management among patients with HNC undergoing immunotherapy and which patients can safely avoid ND without compromising survival.

Conflicts of Interest

The authors declare no conflicts of interest.