A, Modified radical neck dissections (MRNDs) performed per month; the mean number of MRNDs per month increased over the study period. B, Supraomohyoid neck dissections (SONDs) performed per month; the mean monthly number of SONDs remained relatively constant. In both panels, dot represents the number of cases, and a straight line represents a linear trend.
A, Mean operation times of the MRND subgroups (1, earliest, least experienced, through 5, latest, most experienced), consisting of 10 consecutive cases each. The mean operation time for MRND subgroup 5 was 212.4 minutes, which was 29% shorter than that for MRND subgroup 1 (298.1 minutes) (P = .001). B, Mean hospital stays showed no statistically significant differences between the subgroups of the MRND group. C, Mean duration of drainage showed no statistically significant differences between subgroups of the MRND group. D, Mean amount of drainage showed no statistically significant differences between the subgroups of the MRND group.
A, Mean operation times of the SOND subgroups (1, earliest, least experienced, through 5, latest, most experienced). The mean operation time for SOND subgroup 4 was 106.1 minutes, which was 53% shorter than that for SOND subgroup 1 (226.5 minutes) (P = .001). B, Mean hospital stays showed no statistically significant differences between the subgroups of the SOND group. C, Mean duration of drainage showed no statistically significant differences between the subgroups of the SOND group. D, Mean amount of drainage showed no statistically significant differences between the subgroups of the SOND group.
Kim WS, Ban MJ, Chang JW, Byeon HK, Kim H, Han JH, Koh YW, Choi EC. Learning Curve for Robot-Assisted Neck Dissection in Head and Neck CancerA 3-Year Prospective Case Study and Analysis. JAMA Otolaryngol Head Neck Surg. 2014;140(12):1191-1197. doi:10.1001/jamaoto.2014.2830
Robot-assisted neck dissection (RAND) for the management of regional metastases is a recently developed technique in the field of head and neck cancer that uses a robotic surgical system. This is the first report that estimates the learning curve for RAND.
To evaluate a learning curve for RAND according to the types of neck dissection and report clinical outcomes.
Design, Setting, and Participants
Prospective case study of 90 patients undergoing neck dissection between May 2010 and April 2013 at a university tertiary referral center.
Fifty modified radical neck dissections (MRNDs) and 40 supraomohyoid neck dissections (SONDs) were performed by a single surgeon using a robotic surgical system to treat head and neck cancer. The MRND and the SOND groups were subdivided into 5 and 4 consecutive subgroups of 10 patients each, respectively.
Main Outcomes and Measures
Perioperative parameters were compared, including mean operation time, duration and amount of drainage, length of hospital stay, and postoperative complications.
Significant decreases in mean operation time were observed as experience performing RAND increased in both the MRND and the SOND groups. The mean operation time for the MRND group decreased by 29% over the course of our study (initial subgroup, 298.1 minutes; last subgroup, 212.4 minutes). The mean operation time for the SOND group decreased by 53% over the course of our study (initial subgroup, 226.5 minutes; last subgroup, 106.1 minutes). There were no significant differences between subgroups for the other perioperative parameters.
Conclusions and Relevance
We explored the RAND learning curve in a single institution over a 3-year period. This study can be used as a timeline reference for institutions where the RAND procedure will be adopted as an alternative procedure.
The cutting-edge application of transoral robotic surgery (TORS) to head and neck cancer surgery has now been adopted worldwide for the surgical treatment of selected cases of oropharyngeal cancers.1,2 We devised procedures for robot-assisted neck dissection (RAND) of regional nodal metastases in head and neck cancers that are slowly gaining in popularity among head and neck surgeons regardless of patient’s ethnicities.3- 7 However, the timeline references for RAND for head and neck cancer regarding operation time, technical complexity of the procedure, and possible perioperative morbidities have not yet been established. Even experienced head and neck surgeons can expect to face a long learning curve if they do not have prior experience with endoscopic surgery before performing RAND. In this study, we evaluated the learning curve for RAND by head and neck surgeons who were experienced with endoscopic procedures.
A total of 90 patients who underwent RAND for head and neck cancer in the Department of Otorhinolaryngology, Yonsei University College of Medicine, in Seoul, South Korea, from May 2010 to April 2013 were included in this prospective study. The patients all provided written informed consent for study participation, and the institutional review board of Yonsei University College of Medicine approved this study. The study population included patients who (1) underwent modified radical neck dissection (MRND) or selective neck dissection I to III or supraomohyoid neck dissection (SOND) for treatment of head and neck cancer and (2) had no history of neck surgery or radiotherapy. The exclusion criteria were (1) history of neck surgery or radiotherapy, (2) unresectable lymph node metastases, and/or (3) distant metastases at the initial presentation.
Patients were divided into 2 groups by the type of neck dissection (ND) they received. A total of 50 patients underwent MRND, and a total of 40 patients underwent SOND. The MRND group and the SOND group were subdivided into 5 and 4 consecutive subgroups, respectively, with 10 patients in each group. Patient demographic data were analyzed, including patient age, sex, primary tumor site, and cancer stage. The operation time for ND (the time interval between the skin incision and the extraction of the ND specimen), length of hospital stay, duration and amount of drainage, and postoperative complications were recorded and compared between subgroups.
The senior surgeon (Y.W.K) performed all the RAND procedures in this study. Prior to the beginning of the study period, he had performed more than 300 endoscopic head and neck surgical procedures in an academic setting, including thyroidectomies, parathyroidectomies, and submandibular sialoadenectomies.
Our operative procedures for RAND are inspired by those used for the previously reported robotic face-lift thyroidectomy,8 which used the face-lift incision for the access to the thyroid gland. Our techniques are described in detail elsewhere4,9 and are briefly summarized herein.
For all procedures, the patient was placed in the supine position with the neck extended. Either a retroauricular (RA) or a modified face-lift (MFL) incision was used. The incision was made around the RA sulcus and extended along the hairline. For the MFL incision, a preauricular fold incision was added to the RA incision. After the subplatysmal skin flap was elevated, the marginal mandibular branch of the facial nerve was identified. The perifacial lymph nodes were then dissected, after the distal facial vessels were controlled. The lymphofatty tissue found inferior to the parotid gland tail was dissected. Then a dissection along the inferior border of mandible was performed, during which the spinal accessory nerve (SAN) was identified and preserved. The dissection of tissues of levels II, upper III, and VA were performed under direct vision.
After a self-retaining retractor was applied to the skin flap, a dual-channel, 30° endoscope was inserted in an upward direction through the RA of the MFL incision. Harmonic curved shears and Maryland forceps were then inserted on either side of the endoscope. All fibrofatty tissue between the anterior bellies of the digastric muscle was dissected with Harmonic curved shears (Ethicon Inc). The proximal facial artery was sealed with the Harmonic curved shears or with vascular clips. For lymphofatty tissues of levels IIA and III, the medial side of the carotid sheath was dissected using robotic instruments. Then the dissection of level IV and VB tissues was performed. The thoracic duct or lymphatic duct was controlled with vascular clips. A closed suction drain was placed before the wound was closed.
The patient was placed in the supine position with the neck extended. The skin incision and the flap elevation were performed in the same fashion as described for robot-assisted MRND. The ND focusing on lymphofatty tissues of levels II and upper III was undertaken with direct vision first through an RA or MFL incision. The robotic surgical system for SOND used the same settings as for MRND. After the robotic surgical system was docked, level I and any remnant level III tumors were dissected using the robotic system. To secure the working space, the sternocleidomastoid muscle was retracted laterally by the surgical assistant. After the removal of ND specimens, a closed-suction drain was placed.
The t test and Pearson χ2 test were used for all statistical comparisons; P < .05 was considered statistically significant. All statistical analyses were performed using SPSS software, version 12.0 (SPSS Inc).
A total of 90 patients were enrolled in this study. The MRND and SOND groups contained 50 cases and 40 cases, respectively. The mean age of the MRND group was 53.8 years, and the mean age of the SOND group was 50.4 years. Men made up the majority in both groups (MRND, 72% [n = 36]; SOND, 63% [n = 25]). The oropharynx (48% [n = 24]) was the most common primary tumor site in the MRND group and the oral cavity (70% [n = 28]) was the most common primary tumor site in the SOND group.
The pathologic TNM stage was analyzed according to a review of the surgical specimen. Pathologic T2 and T1 were the most frequent tumor stages in the MRND and the SOND groups, respectively. Pathologic N2b and N0 were the most frequent nodal stages in the MRND and the SOND groups, respectively. In the MRND group, a level I-V dissection was performed in 26% of patients (n = 13) and a level II-V dissection was performed in 74% of patients (n = 37). The RA incision for ND was used in 72% of MRND patients (n = 36) and in 75% of the SOND patients (n = 30). Bilateral ND was performed in 10% (n = 5) and 8% (n = 3) of the MRND and the SOND groups, respectively (Table 1).
We analyzed the number of the RANDs performed per month and found that the number of MRND cases performed per month was increasing, with an average of 1.4 cases per month. The number of SOND cases performed per month remained relatively constant throughout the study period at an average of 1.1 cases per month (Figure 1).
There was a gradual decrease over the course of the study in operation time for the MRND group. The mean operation time for MRND subgroup 5 (the last subgroup, with surgeon most experienced) was 212.4 minutes, which was 29% shorter than the average time of subgroup 1 (the first subgroup, surgeon least experienced) (298.1 minutes; P = .001) (Figure 2A). There was also a gradual decrease in operation time for the SOND group; we saw a decrease from 226.5 minutes (subgroup 1) to 106.1 minutes (subgroup 4), which is a 53% decrease (P = .001) (Figure 3A). Other operative parameters including mean hospital stay, mean duration of drainage, and mean amount of drainage are shown in Figure 2 and Figure 3; there were no statistically significant differences in these parameters between the subgroups of either the MRND or the SOND groups.
All postoperative complications are listed in Table 2. In MRND subgroup 1, there was 1 case of Horner syndrome. It developed after the sympathetic ganglion was misidentified during dissection around the carotid sheath. There was 1 case of permanent marginal palsy after MRND. We presumed that the palsy developed as a result of heat damage from the Harmonic curved shears, which were used near the lower border of the mandible. There was 1 case of hematoma that necessitated reoperation to evacuate it. The bleeding from the hematoma was coming from a branch of the internal jugular vein, and it was successfully ligated during the reoperation through the previous RA incision. The other 2 cases of hematoma were controlled with conservative measures like needle aspiration and local compression. There were 2 cases of chyle leakage, both of which were left-side MRND cases. The source of the chyle leakage was identified and ligated using a hemoclip. Eight cases of transient marginal weakness, 1 case of wound dehiscence, 13 cases of earlobe numbness, and 8 cases of seroma developed in the MRND group. Also in the MRND group, 2 cases of hematoma that were managed with local compression and 3 cases of chyle leakage that were resolved after supraclavicular compression and administration of a fat-free diet developed. Overall, there were no significant differences in postoperative complications between the MRND subgroups.
In the SOND group, there were no major complications. Minor complications included 2 cases of transient marginal weakness, 8 cases of earlobe numbness, and 1 hematoma. There were no significant differences in postoperative complications between the SOND subgroups.
The introduction of the robotic surgical system brought about new surgical methods for the treatment of head and neck cancer. Progress first came with the application of the robot in the removal of primary head and neck tumors, which are deep-seated tumors that need to be removed with a long incision or other morbidity-bearing approaches. TORS has already been evaluated for its feasibility, oncologic outcome, functional results, and learning curve.1,2,10- 12 However, RAND for head and neck cancer treatment is a newly developed procedure, and few studies about it have been undertaken.3- 5,13 We studied the learning curve for RAND to establish a timeline reference for surgeons learning to perform RAND. To our knowledge, this is the first report of its kind for the RAND procedure.
The mean operation time for ND decreased significantly across the course of the study for both the MRND and the SOND groups, but the rate of decrease in the mean operation time was greater in the SOND group. Based on this finding, we suggest that SOND is a more suitable procedure for surgeons who are beginning to perform RAND as part of their head and neck cancer treatment repertoire. The omission of level IV and V tumor dissections can have a significant impact on the effort and skills required to perform RAND. Once SOND techniques have been mastered, progressive trials using MRND techniques are recommended for surgeons in institutions with no prior experience with robot-assisted head and neck surgery.
The MRND groups were heterogeneous with regard to the neck levels dissected. Some MRND cases did not involve level I neck dissections, which means that these cases did not receive the true MRND procedure; this might have led to a bias in the group comparison results. It is not clear whether the inclusion of level I dissections in some groups can account for the significantly different operation times between the early and late cases; therefore, we analyzed the differences between the MRND subgroups based on the inclusion of level I dissections. This revealed that there were no significant differences between subgroups with levels I-V NDs and subgroups with levels II-V NDs. We suggest that the mean operation time difference between the MRND subgroups I and V is not significantly biased by the proportion of the level I dissections included in the subgroups.
In terms of mean length of hospital stay, mean duration of drainage, and mean amount of drainage, there were decreasing trends for all parameters in both the MRND and the SOND groups; however, as the case numbers increased, the differences lost statistical significance. The mean length of hospital stay after conventional ND is relatively long in Korea because patients prefer to stay longer in the hospital and actively participate in decisions of discharge timing. Accordingly, the hospital stay in this study was also long in both groups.
Regarding complications, there were no differences between the subgroups of either the MRND or the SOND groups. During the initial period, 1 case of Horner syndrome developed.
In the robotic surgical system, the surgical field and structures are magnified 10 times, and the magnified sympathetic ganglion can be mistaken for a lymph node. Therefore, it is important for surgeons to be aware that the structures in the posterior area of the carotid sheath can easily be mistaken for lymph nodes, so the area should be cautiously scanned from a superior and inferior point of view to achieve a clear and accurate view of nervous attachment. There was 1 case of permanent marginal palsy in the MRND group. One of the greatest advantages of the RAND procedure is that the incision location is shifted from the anterior neck to the hidden RA area, and consequently, better cosmesis is achieved compared with that from conventional open ND. However, the cosmetic advantage of RAND is counterbalanced by the greater risk for permanent postoperative marginal palsy. Permanent nerve damage would eliminate other cosmetic concerns; therefore, every effort should be made to avoid this complication. The surgeon must always know the location of the marginal branch of the facial nerve during the operation, and immoderate retraction by an assistant of the nerve near the mandible angle should be avoided.
There were 3 cases of hematoma in the MRND group and 1 case in the SOND group. We reoperated for 1 hematoma in the MRND group because it was rapidly expanding in an urgent setting immediately after the operation. We had a hard time finding and controlling the bleeding branch of the internal jugular vein compared with similar postoperative bleeding situations in open neck dissections. However, since we could easily elevate the skin flap and apply the self-retaining retractor in the reoperation setting through the previously elevated plane, it was not difficult to access the bleeding focus. Because of the large dissection area from the RA incision to the neck, there is more space to hold larger amounts of blood, and consequently the detection of hematoma might be delayed. However, the area from the RA incision to the anterior border of the sternocleidomastoid muscle can be well collapsed immediately after the operation by applying negative pressure with a closed-suction drain. Therefore, we speculate that the RA approach itself will not be the primary cause of delay in the hematoma detection.
The fact that every case of chyle leakage developed only during left-side dissections alerted us to the importance of secure handling of the thoracic duct during the operation. Overall, the MRND groups had a higher incidence of postoperative complications than the SOND groups, but in both groups the last subgroups had lower complication rates than the initial subgroups, though these results are not statistically significant.
Olofsson and Tytor14 detailed the number of complications that developed after the neck dissection and reported that there were 8 neck dissections with major surgical complications (4.3%) and 52 with minor complications (28.0%) following 186 neck dissections in 171 patients. In our study, the major complications developed in 5 cases (10%), and the minor complications in 35 cases (70%) among a total of 50 MRND cases. This discrepancy is mainly owing to the differences in the definition of major and minor complications. In our study, we defined major complications as a permanent sequela, including Horner syndrome; permanent marginal nerve palsy; phrenic nerve palsy; or a situation necessitating reoperation, including hematoma and chyle leakage. In contrast, Olofsson and Tytor14 defined major complications as a situation leading to the death or to interference with the normal diet as pharyngocutaneous fistula. According to their definition, there were no major complications in our patients.
This study has several limitations. The operating surgeon might have been too experienced at the onset of the study: he had completed over 300 endoscopic head and neck procedures, including endoscopic thyroidectomies, submandibular gland excisions, and branchial cleft cyst excisions. The generalizability of this study might be compromised by the bias introduced by a highly experienced surgeon. The robotic surgical system is based on the endoscopic system but has the additional advantage of delivering 3-dimensional images to the surgeon. Therefore, surgeons who are not experienced in endoscopic surgical procedures would face unfamiliar visual information and the novel sensation of a constrained environment for handling instruments with a remotely placed motion joint. Poor transmission of tactile information to the surgeon is another limitation of the robotic surgical system. Meticulously handling tissue with Maryland forceps can be a valuable opportunity to reduce unnecessary bleeding, but this sensation can only be mastered through practice with endoscopic surgery before performing a RAND.
We demonstrated a decrease in mean operation time with the accumulation of experience performing robot-assisted MRND and SOND, without a change in postoperative complications. This study can serve as a timeline reference for any institution intending to adopt RAND as an alternative procedure. A comparison study of learning curves between endoscopically experienced and inexperienced surgeons would make for a particularly useful extension to our work.
Submitted for Publication: June 2, 2014; final revision received August 11, 2014; accepted September 8, 2014.
Corresponding Author: Yoon Woo Koh, MD, PhD, Department of Otorhinolaryngology, Yonsei University College of Medicine, 50 Yonsei-ro, Seodaemun-gu, Seoul 120-752, South Korea (firstname.lastname@example.org).
Published Online: November 13, 2014. doi:10.1001/jamaoto.2014.2830.
Author Contribution: Drs W. S. Kim and Koh had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: W. S. Kim, Koh, Choi.
Acquisition, analysis, or interpretation of data: W. S. Kim, Ban, Chang, Byeon, H. Kim, Han, Koh.
Drafting of the manuscript: W. S. Kim, Ban, Chang, H. Kim, Han.
Critical revision of the manuscript for important intellectual content: Byeon, Koh, Choi.
Statistical analysis: W. S. Kim, Ban, Chang, H. Kim, Han.
Administrative, technical, or material support: Byeon, Koh.
Study supervision: Koh, Choi.
Conflict of Interest Disclosures: None reported.
Previous Presentation: This study was presented at the Fifth World Congress of the International Federation of Head and Neck Oncologic Societies and the Annual Meeting of the American Head & Neck Society; July 29, 2014; New York, New York.
Additional Information: Drs Chang and Koh contributed equally to this work.
Correction: This article was corrected on December 3, 2014, to add the Additional Information to the end matter specifying that Drs Chang and Koh contributed equally to this work.