The purpose was to determine if postradiotherapy (RT) neck dissection can be limited to the neck levels of residual adenopathy on post-RT computed tomography (CT).
The purpose was to determine if postradiotherapy (RT) neck dissection can be limited to the neck levels of residual adenopathy on post-RT computed tomography (CT).
In all, 274 patients with lymph node-positive head and neck squamous cell carcinoma were treated with definitive RT. All patients had a contrast-enhanced CT performed 4 weeks after completing RT to evaluate tumor response. Two hundred eleven heminecks were dissected, either planned pre-RT or because of residual adenopathy on post-RT CT. CT images were reviewed to determine the presence and location of residual adenopathy. Radiographic complete response (rCR) was defined as lymph node size ≤1.5 cm and normal radiographic morphology (no filling defects or calcifications). For each neck level the CT findings were correlated with neck dissection pathology.
Correlation of CT nodal response with neck dissection pathology revealed the following negative predictive values of rCR: level I, 100%; level II, 95%; level III, 98%; level IV, 96%; and level V, 96%. A subset analysis was performed on 61 neck levels with initially positive lymph nodes that completely responded to RT that were in a hemineck with residual lymphadenopathy elsewhere in the neck. Correlation of nodal response on CT to pathology indicated a negative predictive value of an rCR of 95% for this high-risk scenario. In 71 heminecks that underwent a selective neck dissection (defined as dissection of less than levels I-V) the 5-year neck control rate was 100%.
rCR on post-RT CT has a negative predictive value of ≥95% for each neck level. This suggests that limiting neck dissection based on post-RT CT is safe. Cancer 2008. © 2008 American Cancer Society.
Radiation therapy (RT) with concurrent chemotherapy has emerged as the treatment of choice for many patients with locally advanced head and neck cancer. Management of the post-RT neck in patients presenting with nodal metastases has evolved over the past several decades and remains controversial. Historically, planned neck dissection was performed on all patients with N2 and N3 nodal disease because these patients were known to have a high rate of regional failure. Several studies had demonstrated improved regional control and possibly cause-specific survival with a planned neck dissection.1, 2
In the 1990s several centers demonstrated low rates of neck failure in patients who had a clinical complete response after RT.3–6 These findings sparked interest in using the response to treatment to determine the need for a planned neck dissection after RT. At the University of Florida, computed tomography (CT) performed 4 weeks after completing RT was used to assess the nodal response in patients with cervical metastases. Analysis of the data showed that radiographic complete response (rCR) was associated with a pathologically negative neck dissection specimen 94% of the time.7, 8 Since the year 2000, patients who achieved an rCR in the neck have been observed instead of receiving a planned dissection, and the neck control rate in these patients has been 100%.7 It is our current policy at the University of Florida to determine the need for neck dissection after RT based on the nodal response on the post-RT CT.
Although there is still considerable debate about which patients can be spared a neck dissection after RT, very little attention has been given to determining what the extent of the neck dissection should be in patients with evidence of residual nodal disease. The traditional philosophy of most head and neck surgeons has been to remove all levels of lymph nodes (levels 1–5) with or without the nonlymphatic structures. This approach is associated with significant morbidity, including neck fibrosis, pain, and shoulder dysfunction secondary to spinal accessory nerve injury.9–11 Studies have shown that limiting the extent of neck dissection results in decreased morbidity.12–17 The hypothetical problem with a less extensive neck dissection is the risk of leaving behind regional disease. The purpose of this study was to determine if post-RT neck dissection can be limited to the level(s) of residual adenopathy, as determined by CT scan, without compromising disease control.
The records of 274 patients with lymph node-positive squamous cell carcinoma of the oropharynx, hypopharynx, larynx, or an unknown head and neck primary site treated with definitive RT at the University of Florida between 1990 and 2004 were retrospectively reviewed under Institutional Review Board approval. Patients with a prior neck dissection, a history of RT, no post-RT contrast-enhanced CT, or uncontrolled primary site at the time of the post-RT CT were excluded. Eight patients were excluded because they did not undergo a recommended neck dissection due to poor performance status (n = 2), unresectable disease in the neck (n = 3), and distant metastases discovered before the neck dissection was performed (n = 3). Patient and treatment characteristics are listed in Table 1. The median age was 58 years (range, 31–84 years).
|Characteristic||No. (%) of patients|
|Base of tongue||92|
|Posterior pharyngeal wall||11|
|Unknown head and neck primary||11 (4)|
|Multiple synchronous primaries||6 (2)|
|Radiation therapy technique|
|Hyperfractionation (twice daily)||190 (69)|
|Conventional (once daily)||39 (14)|
|Concomitant boost||45 (16)|
RT was usually given with conventional opposed lateral fields using CT planning with 6-MV photons or cobalt-60. The median dose was 74.4 Gy (range, 56.4–79.6 Gy) predominantly administered at 1.2 Gy/fraction (69%) twice daily to the primary site and upper neck. A subset of patients (14%) was treated with once-daily fractionation administered at 1.8 to 2.0 Gy per fraction. An off-cord reduction was made at 40 to 45 Gy and 8 to 10 MeV electrons were used to supplement the dose to the tissues overlying the spinal cord. For oropharyngeal, hypopharyngeal, and laryngeal primary tumors, a second mucosal reduction was made at 60 Gy. Forty-five (16%) patients were treated with intensity-modulated radiotherapy (IMRT), in which the RT was delivered using a concomitant boost plan (primary field treated at 1.8 Gy/fraction daily plus a boost of 1.5 Gy in the afternoons of the last 12 days of treatment). The low neck was treated with an anterior field prescribed to Dmax, typically with 50 Gy at 2 Gy/fraction with a mid-neck boost to the involved neck of 10 to 20 Gy at 2 Gy/fraction. The median total treatment time (from RT start to RT end) was 47 days (range, 27–65 days).
Chemotherapy was administered in 116 patients (42%). One hundred one (87%) patients received concurrent chemotherapy and 15 (13%) patients received neoadjuvant chemotherapy. The latter was discontinued in 1997. Concurrent chemotherapy consisted of cisplatin in 59 patients, administered as 6 mg/m2 daily, 30 mg/m2 weekly, or 80–100 mg/m2 bolus every 3 weeks. In 13 patients cisplatin was administered intraarterially for 4 cycles as outlined by an institutional protocol. Concurrent carboplatin and paclitaxel were administered in 26 patients, usually given in doses of 100 and 45 mg/m2, respectively, for 5 to 7 weekly cycles. Concurrent fluorouracil (800–1000 mg/m2 for 4 days) and cisplatin (20 mg/m2 for 4 days) times 2 cycles was administered in 13 patients. Induction chemotherapy usually consisted of fluorouracil 600 to 1000 mg/m2 for 4 days and cisplatin 80 to 100 mg/m2 bolus times 3 cycles.
Posttreatment nodal response was evaluated by contrast-enhanced CT at a median of 29 days after completing RT. All 274 patients had CT images available on soft copy (ie, digital film, which can be manipulated on a workstation) for blinded re-review by a neuroradiologist. Neck levels I through V18 on the initially involved hemineck were assessed for lymph node size and the presence of internal focal defects (lucency, enhancement, or calcifications). An rCR was defined as the absence of any lymph nodes greater than 1.5 cm in greatest axial dimension and lymph nodes with any focal lucency, focal enhancement, or focal calcifications. No patient had a PET scan as part of the post-RT evaluation.
One hundred ninety patients underwent post-RT neck dissection at a median of 47 days after completing RT. The neck dissection was performed more than 60 days after the post-RT CT in 6 patients. These 6 patients were excluded from any further analysis because of the extended delay between imaging and neck dissection, which could confound study results because of a change in burden of disease from the time of the CT scan. Of the 184 patients whose neck dissections were performed within 60 days after the date of the CT, 27 patients underwent a bilateral neck dissection and 157 underwent a unilateral neck dissection, resulting in 211 dissected heminecks that were included in the analysis. All of the 27 patients who underwent a bilateral neck dissection had bilateral nodal disease before treatment. Seventy-six patients did not undergo a post-RT neck dissection, corresponding to 110 heminecks. Most of the neck dissections were planned pre-RT until the year 2000, after which neck dissections were recommended only if there was evidence of residual nodal disease on the 4-week post-RT CT scan. The extent of neck dissection was recorded: 71 heminecks underwent a selective neck dissection (defined as the removal of any combination of neck levels other than I-V) and 140 heminecks underwent a complete neck dissection (defined as the removal of levels I-V). All neck dissection specimens were separated during surgery into levels I through V and sent to pathology as separate specimens. Standard pathologic processing of the specimens involved sectioning them at 4-μm slices before microscopic examination.
Patients were seen in the clinic for follow-up every 4 to 6 weeks for the first year after treatment, gradually extending to yearly appointments after 5 years. Patients who were unavailable for clinical follow-up were contacted by telephone. Minimum follow-up for all surviving patients was 2 years. Observed median follow-up was 4.2 years (range, 0.2–14.1 years), and among the 158 living patients the median follow-up was 5.2 years (range, 2.0–14.1 years). Five patients were lost to follow-up and censored at the time of last follow-up.
All statistical analyses were completed using SAS and JMP software (SAS Institute, Cary, NC). Neck control rates were assessed using the Kaplan-Meier product limit method. The log-rank test statistic allowed detection of statistically significant differences; a P-value of <.05 was interpreted as statistically significant. A Fisher exact test was used to assess the ability of selected prognostic factors to predict for radiographic response.
The 5-year outcome rates for all 266 patients were as follows: neck control, 94%; primary-site control, 90%; cause-specific survival, 80%; and overall survival, 65%.
In all, 184 patients underwent a post-RT neck dissection, 27 of which were bilateral, corresponding to the 211 involved heminecks dissected. The neck control rate at 5 years was 100% for the 32 dissected heminecks that met the criteria for an rCR, and 96% for the 178 dissected heminecks that did not meet the criteria for an rCR. Seventy-six patients did not undergo a post-RT neck dissection, corresponding to the 110 involved heminecks not dissected. The neck control rate at 5 years was 100% for the 68 heminecks that met the criteria for an rCR, and 88% for the 42 heminecks that did not meet the criteria for an rCR when the CT images were reviewed for this analysis. Reasons for not undergoing dissection in the 42 undissected heminecks that did not meet the criteria for an rCR were patient refusal in 3 heminecks, comorbidities in 5 heminecks, and interpretation of the post-RT CT as a complete response at the time the patient was evaluated in 34 heminecks.
For all 211 dissected heminecks, correlation of complete response by CT with the neck dissection specimen by neck level demonstrated negative predictive values of 100% for level I, 95% for level II, 98% for level III, 96% for level IV, and 96% for level V (Table 2).
|Neck level||No. of neck levels that met following criteria||NPV of rCR|
|Path (−), CT (−)||Path (−), CT (+)||Path (+), CT (−)||Path (+), CT (+)|
A subset analysis was performed to determine the negative predictive value of an rCR for the specific high-risk scenario in which an initially positive lymph node completely responds to RT, but there is evidence of residual lymphadenopathy elsewhere in the hemineck. Sixty-one neck levels met the following criteria: 1) initially positive before treatment as determined by pretreatment CT; 2) determined to have completely responded to RT on re-review of post-RT CT; 3) neck level included in dissection; and 4) re-review of post-RT CT showed residual lymphadenopathy in a different neck level within the same hemineck. Three of the 61 neck levels had positive neck dissection pathology, resulting in a negative predictive value of 95% for an rCR in this high-risk scenario (Table 3).
|No. of neck levels that met criteria*||No. of neck levels with positive neck dissection pathology||NPV of rCR|
Seventy-one heminecks underwent a selective neck dissection (dissection of less than levels I-V) with a 5-year neck control rate of 100%. This was not significantly different (P = .06) from the 5-year neck control rate of 95% for the 140 heminecks that underwent a complete dissection of levels I through V (Fig. 1). Table 4 shows the distribution of nodal and tumor stage in the selective neck dissection group compared with the complete neck dissection group. Because this was not a prospective comparison, there were some differences in the distribution of nodal stages between the 2 groups. Specifically, there was a higher proportion of N2C necks in the selective neck dissection group compared with the complete dissection group (32% vs 23%), but a higher proportion of N3 necks in the complete dissection group compared with the selective neck dissection group (21% vs 10%). Neither of these differences was statistically significant. T stage was distributed evenly between the groups.
|Distribution||Selective dissection N = 71||Dissection of levels I–V N = 140|
|No. (%)||No. (%)|
|N1||7 (10)||13 (9)|
|N2a||8 (11)||15 (11)|
|N2b||26 (37)||50 (36)|
|N2c||23 (32)||32 (23)|
|N3||7 (10)||30 (21)|
|T0||4 (6)||6 (4)|
|T1||7 (10)||16 (11)|
|T2||35 (49)||67 (48)|
|T3||12 (17)||26 (19)|
|T4||13 (18)||25 (18)|
The current approach to the post-RT neck is to perform a neck dissection of levels I through V in patients who have evidence of residual lymphadenopathy after RT, which is usually determined by CT scan. Although there has been much controversy over how to determine which patients may be spared a neck dissection after RT, little attention has been given to the concept of decreasing the extent of neck dissection in patients who have residual lymphadenopathy. Our data suggest that there is a 5% risk of leaving behind disease in the neck if the neck dissection is limited to the levels of residual adenopathy as determined by the post-RT CT. Correlation of neck dissection pathology with the post-RT CT reading shows that an rCR has a negative predictive value of at least 95% for each neck level. This is further supported by the finding that the 5-year neck control rate in patients with an rCR who did not undergo a neck dissection and were followed was 100%. The high negative predictive value of CT is true even for the high-risk scenario in which an initially positive lymph node completely responds to RT, but there is evidence of residual lymphadenopathy elsewhere in the hemineck. For example, in a patient who initially presents with lymphadenopathy in right levels II and III, and then on the 4-week post-RT CT shows an rCR in level III but residual lymphadenopathy in level II, there is a 5% risk of leaving behind residual disease in the neck if only level II is dissected.
Only a few reports have attempted to use neck dissection pathology to determine if decreasing the extent of post-RT neck dissection is feasible. Doweck et al.19 analyzed 76 patients with lymph node-positive oropharyngeal carcinoma treated with hyperfractionated RT. All patients with N2 and N3 disease had planned neck dissections; patients with N1 disease underwent neck dissection for residual adenopathy only. By correlating the neck dissection pathology to the post-RT CT reading by neck level, they demonstrated that the false-negative rate of the post-RT CT was negligible for levels I and V, and concluded that post-RT neck dissections could be safely limited to levels II through IV in the absence of radiologic evidence of residual disease in levels I and V. Boyd et al.20 reviewed the neck dissection pathology of 25 patients with lymph node-positive squamous cell carcinoma of the head and neck who were treated with definitive RT and planned neck dissection. Only 1 of 28 neck dissection specimens revealed tumor outside of levels II through IV. Based on the predictable pattern of residual disease, they concluded that the post-RT neck dissection could be safely limited to levels II-IV in most patients.
We also demonstrated that there was no decrease in neck control when a less extensive neck dissection was performed in carefully selected patients. Seventy-one heminecks underwent a selective neck dissection with a 5-year neck control rate of 100%. Neck control was not different compared with the 140 heminecks that underwent a complete dissection of levels I through V. This result compares favorably with the results reported in the literature, although data addressing this issue is quite sparse. Robbins et al.21 reported a neck control rate of 97% for 92 heminecks in patients with stage III or IV head and neck cancer treated with RT and concurrent intraarterial cisplatin followed by a selective or superselective (removal of levels II and III only) neck dissection for residual adenopathy (which was initially performed on all patients with N2 or N3 disease). Similarly, Stenson et al.22 demonstrated a neck control rate of 98% for 70 heminecks in patients with stage III or IV head and neck cancer who underwent a planned selective neck dissection after 1 of 4 concomitant chemoradiation protocols.
The major advantage of decreasing the extent of the post-RT neck dissection is reduced morbidity. Neck fibrosis, shoulder pain, and shoulder dysfunction are well-known side effects associated with neck dissections which remove nodal levels I through V, especially if the spinal accessory nerve is damaged or sacrificed.23, 24 There are no studies comparing the morbidity of dissecting only 1 or 2 neck levels versus a selective neck dissection. However, several studies have shown that decreasing the extent of neck dissection by performing a selective neck dissection instead of a complete dissection of levels I through V is associated with decreased morbidity and improved quality of life. Chepeha et al.17 assessed shoulder function after neck dissection and found that patients receiving a modified radical neck dissection sparing the spinal accessory nerve had significantly worse shoulder function than patients receiving a selective neck dissection. Inoue et al.14 assessed both functional outcome and pain in patients who underwent a neck dissection and were able to show that patients had decreased pain and increased shoulder and neck function when levels IV and V were not included in the dissection. By analyzing shoulder function scores and electromyograms, Sobol et al.10 demonstrated that patients who underwent a supraomohyoid neck dissection had minimal loss of shoulder function compared with patients who underwent a modified radical neck dissection. Several reports13, 15, 16 have shown that patients who had level V dissected had significantly more shoulder dysfunction, pain, and electrophysiologic abnormalities compared with patients who did not have level V dissected. Using the University of Washington quality of life scale, Kuntz and Weymuller12 concluded that the more extensive the neck dissection the greater the negative impact on quality of life, especially with regard to shoulder function.
In summary, an rCR on the post-RT CT has a negative predictive value of at least 95% for each neck level, even in the high-risk scenario of residual adenopathy in an adjacent neck level. These data suggest that removing only the levels of residual adenopathy, which may decrease morbidity, will result in a 5% risk of recurrence in the neck.