Transoral robotic surgery for the benefit of patients with head and neck cancer of unknown primary: our experience at St George’s University Hospital, London

Abstract

Introduction

Head and neck carcinoma of unknown primary represents 1–5% of all head and neck cancers and presents a diagnostic and therapeutic dilemma. In approximately 40% of cases, a primary tumour location remains unknown despite investigation. With advancements in our understanding of the role of high-risk human papilloma virus in head and neck cancer, transoral robotic surgery presents an option for diagnosis and therapy.

Materials and methods

This is a retrospective case series from a single centre. Case notes were reviewed for 28 patients who had transoral robotic surgery for head and neck carcinoma of unknown primary between May 2015 and July 2019.

Results

Transoral robotic surgery identified an oropharyngeal primary tumour in 19 of 28 (67.8%) patients. All oropharyngeal primaries were p16 positive. The base of tongue identification rate was 63.2%. Median length of inpatient stay postoperatively was 1.0 day. Normal oral intake resumed within 48 hours in 96% (27/28) of patients. Three patients (10.3%) suffered minor postoperative bleeds that were all managed conservatively.

Discussion

The base of tongue primary identification rate (63.2%) in this series is consistent with that previously reported (43–63%; 95% confidence interval). Primary tumour identification rate if a patient is p16 positive is 86.3% (19/22), with 100% of these being oropharyngeal. We suggest future investigation into p16 status as a means of stratifying patients with head and neck carcinoma of unknown primary for transoral robotic surgery.

Conclusion

Transoral robotic base of tongue mucosectomy (or lingual tonsillectomy) is a promising technique that offers a high yield of positive identification for the primary tumour. It is well tolerated with minimal associated morbidity. Our findings are comparable with those in the current literature.

Introduction

Traditional teaching has held that the major risk factors for head and neck squamous cell carcinoma (HNSCC) in the UK are tobacco smoking and alcohol consumption. The classical epidemiological picture of HNSCC was a disease predominantly affecting those of middle age and beyond, male, and heavy consumers of tobacco and alcohol. However, the advent of the new millennium has seen an upsurge in an entirely different phenotype of disease: one affecting patients in their fourth decade or younger, lifelong abstainers from tobacco and alcohol, and presenting with regionally advanced disease. Subsequent investigation has highlighted a strong association between this phenotype and high-risk human papilloma virus (HPV), particularly subtypes 16 and 18.^1,2

p16 immunostaining is used as a surrogate marker for high-risk HPV infection.² Patients with oropharyngeal cancer who demonstrate p16 positivity present with regionally advanced disease (ie bulky nodal metastases), have comparatively favourable clinical outcomes in contrast to the ‘classical’ patient with HNSCC.^2,3 p16 staining has consequently been useful in predicting radiotherapy response and prognosis in tonsillar carcinoma patients.⁴ It is thought that this reflects differences in tumour biology that make HPV-associated disease more susceptible to chemoradiotherapy, as well as favourable clinical factors such as an increased likelihood of identification of the primary tumour.^3,4 Consequently, the American Joint Committee of Cancer Tumour Staging Manual eighth edition reflects the differences in disease phenotype with two entirely distinct staging systems for oropharyngeal cancer: one for HPV-positive disease and one for HPV-negative disease.⁵

A patient who presents with a neck lump that is suspicious for squamous cell carcinoma undergoes diagnostic workup according to the British Association of Head and Neck Oncologists (BAHNO) and National Institute for Health and Care Excellence guidance.^6,7 This involves a thorough history and full ear, nose and throat examination including flexible nasopharyngolaryngoscopy, followed by neck ultrasound with guided cytology. Should cytology prove suspicious for malignancy, the patient undergoes further investigation by means of imaging; computed tomography (CT), magnetic resonance imaging (MRI), positron emission tomography CT (PET-CT), and endoscopic examination of the upper aerodigestive tract with biopsies of suspected mucosal lesions.^6,7 However, despite extensive investigations, it is occasionally not possible to identify the primary tumour and the search for it must begin.

PET-CT is an essential imaging modality for evaluating patients with malignant cervical adenopathy from an unknown primary.⁸ It aids detection of an occult primary in approximately one-third of cases with a sensitivity of up to 97% and specificity of 68%.^6,9 However, it has limited ability to detect subcentimetre tumours, and has a high false positive rate.⁹

The location of the primary tumour can thus remain unknown in over 40% of patients with head and neck cancer who present with no obvious primary lesion on clinical or radiological assessment.^10–12 The vast majority (90%) of these unknown primaries are squamous cell carcinomas, while the remainder consist of adenocarcinomas, melanomas and other rarer types.^1,2 When a primary tumour is located, the literature suggests that HPV-positive disease has a predilection for the oropharynx, with the primary location usually in the tonsils (44.7%) or base of the tongue (43.9%).⁹ In contrast, HPV-negative disease can arise from anywhere in the upper aerodigestive tract.¹³

Identifying the site of the primary tumour is oncologically preferable, as it allows targeted management, resulting in a more favourable clinical outcome, as well as reduced morbidity and adverse effects from wide-field radiation and radical surgical intervention.^2–4,9,15 In a retrospective matched-pair and cohort comparison study of 136 patients, identification of the primary tumour demonstrated significantly improved overall, cause-specific and disease-free survival.¹⁴

Transoral robotic surgery (TORS) is a potential strategy to maximise the diagnostic yield while also providing therapy (complete resection) in some cases. TORS for head and neck cancer of unknown primary is performed when conventional diagnostic workup (clinical assessment, PET-CT, examination under anaesthesia and biopsy) has failed to identify a primary tumour. Figure 1 shows the local diagnostic protocol at St George’s University Hospital, London.⁹ Initial reports have highlighted the potential for TORS to act as an additional weapon in the diagnostic armoury when searching for an unknown primary.^9,15 A 2019 meta-analysis of tongue-base mucosectomy TORS in patients with head and neck cancer of unknown primary demonstrated an additional primary identification rate of 74%, with specific identification rates for lingual tonsillectomy and palatine tonsillectomy of 53% and 31% respectively.¹⁵ However, evidence for the utility of TORS for this specific application is limited by the relative scarcity of head and neck cancer of unknown primary, the novelty of the procedure and the availability of the relevant technology and operator expertise. Current evidence is in the form of case series of varying quality, with relatively small numbers of participants using heterogenous diagnostic pathways.¹⁵

Figure 1.

Diagnostic pathway for patients with head and neck cancer of unknown primary at St George’s University Hospital, London

Surgical procedure

TORS is used to perform a tongue-base mucosectomy (or lingual tonsillectomy) and/or palatine tonsillectomy. Depending on local protocol, the procedure may be unilateral or bilateral with current literature indicating that malignancy in the ipsilateral lingual or palatine tonsil is found in 97% of cases.¹⁵ The surgical technique involves a da Vinci^© robot (Intuitive Surgical Inc.) docked at the patient’s left side and a 30° upward facing endoscope introduced into the oropharynx. A full description of the surgical strategy has previously been published (Fig 2).⁹

Figure 2.

Tongue base in the ‘operative position’ with resection margins of the lingual tonsils of base of tongue mucosectomy outlined using a solid black line as displayed by Ofo et al⁹

Materials and methods

This is a retrospective case series from a single centre. Case notes were reviewed for patients who underwent TORS for head and neck cancer of unknown primary between May 2015 and July 2019. Twenty-eight patients underwent TORS for head and neck cancer of unknown primary. Two of these patients underwent two TORS procedures each and therefore there were 30 TORS procedures in total.

Patients were investigated according to BAHNO guidance, and head and neck cancer of unknown primary was determined by strict multidisciplinary team agreement: metastatic neck disease confirmed on cytology, with negative clinical examination (including nasopharyngolaryngoscopy) and negative imaging, including PET-CT.⁶ We examined the case series for primary detection rate, length of stay, complications, time to resume oral intake and p16 status.

Results

Of the 28 patients who underwent TORS, 22 (78.6%) were male and 6 (21.4%) were female. The mean age of patients at the time of robotic procedure was 59.4 years. Twenty-four of twenty-eight (85.7%) patients were p16 positive. Eight patients were current smokers, ten were ex-smokers and ten had never smoked. Eighteen patients currently drink alcohol, five patients previously drank alcohol and two had never drunk alcohol before. For three patients the data for alcohol intake was not available.

All patients (100%) underwent diagnostic workup as per St George’s University Hospital local protocol (Fig 1). All patients underwent a thorough endoscopic examination of the mucosal surfaces of the upper aerodigestive tract prior to TORS. Prior to TORS, five patients underwent bilateral palatine tonsillectomy and one patient underwent ipsilateral palatine tonsillectomy through traditional surgical techniques.

Clinical nodal status of patients prior to TORS was: N1 (14%), N2 (7%), N2a (52%), N2b (17%), N3 (7%). One case was not given a documented nodal status prior to TORS.

The procedures undertaken are displayed in Table 1. Average patient positioning and access preparation time was 9.3 minutes, average robot docking time was 9.2 minutes, average robotic resection time was 28 minutes and average total operating time was 46.6 minutes. There were no intraoperative complications for any of the patients.

Table 1.

Transoral robotic surgeries undertaken in this case series

Case	Treatment
1	Left tongue-base mucosectomy and left modified radical neck dissection
2	Left tongue-base mucosectomy
3	Right tongue-base mucosectomy
4	Left tongue-base mucosectomy and remnant tonsillectomy
5	Panendoscopy + tongue-base mucosectomy + bilateral tonsillectomy
6	Right tongue-base mucosectomy + bilateral tonsillectomy
7	Lateral oropharyngectomy + left modified radical neck dissection
8	Panendoscopy + left tongue-base mucosectomy + excision of left lateral tongue lesion
9	Right tongue-base mucosectomy + right radical neck dissection
10	Panendoscopy + tongue-base mucosectomy
11	Panendoscopy + tongue-base mucosectomy + bilateral tonsillectomy
12	Tongue-base mucosectomy + bilateral tonsillectomy
13	Panendoscopy + tongue-base mucosectomy + bilateral tonsillectomy
14	Panendoscopy + tongue-base mucosectomy + bilateral tonsillectomy
15	Left tongue-base mucosectomy + left tonsillectomy + left neck dissection
16	Panendoscopy + right lingual tonsillectomy
17	Panendoscopy + tongue-base mucosectomy
18	Left tongue-base mucosectomy + left modified radical neck dissection
19	Panendoscopy + tongue-base mucosectomy + bilateral tonsillectomy
20	Panendoscopy + tongue-base mucosectomy + bilateral tonsillectomy
21	Panendoscopy + tongue-base mucosectomy + biopsy left epiglottic fold
22	Panendoscopy + tongue-base mucosectomy + bilateral tonsillectomy
23	Panendoscopy + bilateral tonsillectomy + tongue-base mucosectomy
24	Tongue-base mucosectomy + right modified radical neck dissection
25	Panendoscopy + bilateral tonsillectomy + tongue-base mucosectomy
26	Tongue-base mucosectomy + bilateral tonsillectomy
27	Panendoscopy + tongue-base mucosectomy + bilateral tonsillectomy
28	Panendoscopy + tongue-base mucosectomy + bilateral tonsillectomy
29	Panendoscopy + tongue-base mucosectomy + bilateral tonsillectomy
30	Tongue-base resection + bilateral selective neck dissection

Findings

TORS identified an oropharyngeal primary tumour in 19 of 28 (67.8%) patients. There were 23 positive biopsy results. All identified oropharyngeal primaries were p16 positive and all were pathologically staged as T1 or T2. Of the oropharyngeal primaries identified, 16/19 had solitary primaries and 3 had synchronous tumours (Table 2). Nine (47%) had tumours of the base of the tongue, seven (37%) were located in the palatine tonsils and three (16%) were base of the tongue and tonsil. In one patient (5%) the primary lesion was on the contralateral side to initial presentation.

Table 2.

Location of identified primary tumour

Location	Total
Base of tongue	9
Palatine tonsils	7
Base of tongue + palatine tonsil	3
Total	19

All oropharyngeal cases (100%) were squamous cell carcinoma. The mean maximum dimension of primary tumour reported on biopsy histology report was 12.6mm (range 2–30mm). Tumour margins were clear in 2 of 20 (10%) biopsies. Of the other 18 biopsies, the margins were positive in 7 (35%) cases and less than 5mm in 11 (55%) cases.

In addition to the oropharyngeal primaries identified, one patient was found to have a metastatic adenocarcinoma of lung on histological examination of the modified radical neck dissection specimen that was performed synchronously with TORS. Another patient had a left lateral tongue lesion identified on synchronous panendoscopy, which was biopsied and returned positive for squamous cell carcinoma.

Postoperatively

Three patients (10.3%) suffered minor postoperative bleeds, all of which were treated medically. All bleed-related complications were Clavien–Dindo grade 2 or below.¹⁶

Median length of inpatient stay postoperatively was 1.0 day. Normal oral intake was resumed within 48 hours in 27 patients with dietician and speech and language therapy input. One of these patients was discharged home but then returned seven days later presenting with odynophagia and dysphagia. This patient did not return to normal oral intake during the subsequent admission and had a surgically guided gastrostomy inserted. The only patient who did not return to normal oral intake within 48 hours of surgery had a radiologically inserted gastrostomy during the same admission. That patient was found to have metastatic lung adenocarcinoma, was found to suffer from mild dementia, identified postoperatively, which made engagement with swallowing rehabilitation difficult. The patient died during the same admission. Consequently, the true gastrostomy rate as a direct effect of swallowing difficulty following TORS was 3.5% (1/28).

Discussion

The advantages of TORS over traditional surgical techniques are superior visualisation and access to the entire tongue-base mucosa or lingual tonsils, allowing precise tissue resection with minimal associated morbidity in comparison with traditional surgical methods.⁹ Although transoral laser microsurgery is an alternative to TORS, with comparable results (primary identification rate for lingual tonsillectomy via transoral laser microsurgery 57%; range 43–72%; 95% confidence interval), current literature also describes the advantages of TORS over transoral laser microsurgery in many ways, including superior three-dimensional high-definition visualisation, wider range of instrument motion, reduction in hand tremors and much improved surgeon ergonomics.^9,15

Despite the reported total (lingual tonsillectomy and palatine tonsillectomy) transoral laser microsurgery identification rate in the 2019 meta-analysis, only 17% (n = 81) of patients definitely had transoral laser microsurgery, compared with 83% who had TORS.¹⁵ Focusing on lingual tonsillectomy alone, 64 patients in total underwent transoral laser microsurgery compared with 371 patients undergoing TORS.¹⁵ More data directly comparing transoral laser microsurgery with TORS are therefore required to fully evaluate the two methods.

Recent literature has also highlighted the importance of careful patient selection in TORS with the ‘eight Ts’ of endoscopic access: teeth, trismus, transverse dimensions (mandibular), tori, tongue, tilt, treatment (prior radiation) and tumour.¹⁷ The main risk of TORS, however, is a postoperative haemorrhage rate of at least 5%, which is similar to that of traditional palatine tonsillectomies performed with diathermy.¹⁵

Comparison with other studies

Patient demographics for patients with head and neck cancer of unknown primary in our population (79.6% male, 21.4% female; mean age 59.4 years) were similar to those reported in the wider literature. A systematic review and meta-analysis of 556 patients from 21 studies investigating transoral tongue-base mucosectomy for the identification of the primary site in the workup of head and neck cancer of unknown primary found 88% of subjects were male with a mean age of 56.6 years. HPV positivity in the same paper was reported at 70.5%, while in our study it was 82.6%¹⁵ (Table 3).

Table 3.

Comparison of key findings of this case series to 2019 meta-analysis of transoral robotic surgery for head and neck cancer of unknown primary¹⁵

Key finding	Farooq et al (2019)15	St George’s University Hospital case series
Demographics:
Mean age	56.6 years	59.4 years
Sex	88% male, 12% female	79.6% male, 20.4% female
Diagnostic workup	Heterogenous between studies	100% concordance to BAHNO guidance
	Timing of TORS most effective at end of diagnostic pathway	100% of TORS procedures at the end of the diagnostic pathway
P16 positive status	70.5%	82.6%
Primary identification rate:
Oropharyngeal	74%	67.8%
Tongue-base/lingual tonsillectomy	53% (95% CI 43–63%)	63.2%
Haemorrhage rate	4.9%	10.7% (n = 3)
Length of stay	Range 1.4–6.3 days	Median 1.0 days

The clear diagnostic pathway that has been established within our department ensured that all patients were investigated exhaustively with the aim of identifying the primary tumour. However, Farooq et al demonstrated heterogenous preoperative workup worldwide.¹⁵ Panendoscopy and biopsy were performed in 85.7% of studies, conventional diagnostic imaging (CT/MRI or both) was only used in 42.9% of studies and PET-CT was used as initial workup in 85.7% of studies.¹⁵ The efficacy of TORS tongue-base mucosectomy was found to be of highest yield and most cost effective when at the end of the diagnostic pathway (ie when imaging, examination under anaesthesia and tonsillectomy were all negative).¹⁵ This was true for all cases within our series and we would concur with the findings of the meta-analysis.

Overall, our oropharyngeal primary tumour identification rate (67.8%) was similar to that reported in the latest meta-analysis (74%). However, the added value of TORS in detecting tonsil squamous cell carcinoma is questionable, as these tumours can be easily detected using much cheaper traditional palatine tonsillectomy techniques such as bipolar diathermy. Therefore, the tongue-base primary tumour detection rate for TORS when conventional imaging and surgical techniques have failed is 63.2% (12/19; Fig 3). The recent meta-analysis demonstrated a total pooled rate of 53% (43–63%; 95% confidence interval) for TORS lingual tonsillectomy or tongue-base mucosectomy in identification of a primary.¹⁵ This places our current data set at the uppermost range for identification of base of tongue primary (Table 4).

Figure 3.

TORS-specific tongue-base primary tumour detection rate when conventional imaging and surgical techniques have failed

Table 4.

Patient numbers and respective identification rates from 2019 meta-analysis of transoral robotic surgery for head and neck cancer of unknown primary to this case series¹⁵

Studies reported in meta-analysis15	Patients in case series (n)	Lingual tonsillectomy identification rate (%)
Mehta 201324	10	90
Channir 201525	13	54
Patel 201326	41	51
Winter 201727	32	53
Durmus 201328	14	28.5
Geltzeiler 201729	50	64
Hatten 201730	60	47
Byrd 201431	22	73
Blanco 201332	4	0
Krishnan 201733	7	71
Patel 201734	35	42.9
Wolford 201135	9	44
Abuzeid 201636	1	100
Total	298	53a
Mistry et al (this series)	28	63.2

^aTotal pooled rate (95% CI 43–63%)

CI, confidence interval.

Given the current clinical equipoise about the stratification of HPV-positive and -negative patient groups as two separate disease entities, we provide our p16 data as an area for future research and discussion. p16 status was analysed in our patients using immunohistochemistry. While we recognise this is approximately 80% sensitive, it is the preferred method for the majority of centres.^15,18 When analysing the impact of p16 status on primary identification our results show an 86.3% (19/22) primary tumour identification rate if a patient is p16 positive. If a patient is p16 negative, the head and neck primary tumour identification rate was 0% (0/6). We therefore suggest an area for future research and discussion would be the use of p16 status to stratify patients with head and neck cancer of unknown primary for diagnostic TORS (Fig 4).

Figure 4.

p16 disease status and primary oropharyngeal and base of tongue identification rates

Postoperative haemorrhage rates (10.7%) were modestly higher than that reported (4.9%) although this may reflect the relatively modest sample size of our study, or under reporting in other studies. The same must also be considered for the gastrostomy rate (3.5%) compared with the wider literature for tongue-base mucosectomy (0.7%). Given the experience of gastrostomy insertion with the patient in whom a lung adenocarcinoma and mild dementia was found, we have now adjusted our protocol to avoid TORS tongue-base mucosectomy in elderly patients with pre-existing cognitive impairment. Median length of stay (1.0 day) was shorter than that described in the wider literature (range 1.4–6.3 days) for those requiring TORS tongue-base mucosectomy. Contributing to this favourable outcome was input from speech and language therapy and a dietician in the preoperative planning as well as the immediate postoperative period, to ensure that oral intake was resumed within 48 hours for all our patients.

The location of primary tumours once found in our head and neck cancer of unknown primary cohort (45% base of the tongue, 35% tonsils alone, 15% both base of the tongue and tonsil) is similar to the current literature that demonstrated identification of a base of the tongue primary occurs in 53% of patients after TORS tongue-base mucosectomy and 31% in the tonsils.¹⁵ One lesion was in the contralateral tongue-base to the cervical node with metastatic carcinoma. Excluding the metastatic lung adenocarcinoma (n = 1), lateral tongue lesion (n = 1) and true ‘unknown’ primaries (n = 7), we had an ipsilateral oropharyngeal identification rate of 95% (18/19). In comparison, other studies have shown an ipsilateral primary tumour identification rate of 97%, 1.85% in the contralateral tongue-base, 0.23% in the contralateral palatine tonsil and 0.69% in bilateral palatine tonsils. Comparing data with the UK experience of TORS tongue-base mucosectomy, a retrospective case review of 17 patients identified tongue-base tumours in 53% of patients, with 88% of these being ipsilateral and 12% contralateral. Given the appreciable rate of contralateral oropharynx primary tumours, this data provides further evidence in support of bilateral lingual tonsillectomy.¹⁹

Margin status

Tumour margins were clear in 2 of 20 (10%) of TORS tongue-base mucosectomy biopsies. The two tumours were 21mm and 9mm maximum diameter. The mean diameter for those biopsies in which the margin was greater than 1mm but less than 5mm was 10.0mm, and for biopsies in which the margin was positive was 14.7mm.

The indication for TORS in our series of patients with head and neck cancer of unknown primary was diagnostic, and therefore achieving clear margins was not the primary aim, especially as these tumours are not readily appreciable at the time of surgery in most cases. However, we recognise the possibility of resecting a slightly deeper margin on the ipsilateral side during TORS tongue-base mucosectomy in future cases to aid with achieving clear margins which may subsequently impact on the choice of definitive therapy.

Paraffin sections were used for histopathological analysis in all cases. Although intraoperative frozen sections are a recognised option to assess surgical excision margins when there is clinical doubt, frozen sections are of a lower quality and can lead to false negative or false positive results, particularly with small pieces of tissue.⁶ The additional risk of increased length of general anaesthetic with frozen sections must also be considered.

Cost effectiveness

Given the current resource constraints within the UK healthcare system, some evaluation of cost effectiveness is appropriate. In 2012, The Royal College of Surgeons of England reported capital costs of a da Vinci^© robot of £1.7 million, approximately £1,000 per patient for disposables and £140,000 a year for maintenance.²⁰ This significant capital cost is not only restricted to head and neck surgeons, with the majority of robotic procedures being conducted by urological surgeons.²⁰ In addition, with the evolution of technology the range of possible robotic procedures is likely to increase with an improvement in technical skills. In 2016, NHS England published its clinical commissioning policy proposition for TORS in head and neck cancer of unknown primary.²¹ It described how in a systematic review of the cost effectiveness modelling of TORS over a 10-year time horizon, without taking capital cost into account, the cost of TORS compared with the cost of chemoradiotherapy would result in a cost saving to the society of US$1,366 per patient treated. This was with an incremental effectiveness of 0.25 quality of added life years per patient.²² Since then, another paper comparing TORS with radiation therapy found an incremental cost effectiveness ratio of US$82,190 quality of added life years gained from TORS.²³ However, the limited research into this aspect of TORS has primarily been conducted in North America and subsequently there is need for further thorough evaluation in the context of the NHS. In addition, the NHS England report notes that cost effectiveness reduces progressively as adjunct therapy is added to the treatment plan.²¹ The future of head and neck cancer of unknown primary therapy, in particular with regard to the current clinical equipoise surrounding the de-escalation of therapy in HPV-positive patients with head and neck cancer of unknown primary may have an important impact in any cost effectiveness analysis in the future.

Strengths and limitations

This study provides one of the largest case reviews for TORS in head and neck cancer of unknown primary reported within the UK and adds to the literature for a novel diagnostic and therapeutic tool.

Limitations of the study are that this report includes only 28 patients, so is of small size, and is retrospective in nature. Further information will be sought on patient experience and oncological outcomes.

Conclusion

The data from our series confirm that TORS is a safe and effective tool that can be used as an adjunct to traditional diagnostic avenues in the investigation of patients with head and neck cancer of unknown primary.

TORS tongue-base mucosectomy or lingual tonsillectomy is an effective technique that offers a high yield of positive identification for the primary tumour in head and neck cancer of unknown primary. It is well tolerated with minimal associated morbidity. Long-term follow-up is required to assess its impact on oncological outcomes and survival.