Nodal Disease Burden for Early-Stage Oral Cancer

Key Points

Question  Does prophylactic neck dissection provide a survival advantage in patients with clinically node-negative (cN0) oral cancer?

Findings  In this population-based retrospective study, patients who received prophylactic neck treatment had no survival advantage compared with those who received local excision only, and a subgroup of patients at who developed neck disease at follow-up received salvage neck treatment and had the worst survival. Tumor depth of invasion of at least 4 mm was significant but had poor sensitivity and specificity in association with nodal disease.

Meaning  New objective markers are urgently needed to identify high-risk patients with cN0 oral squamous cell carcinoma who may benefit from prophylactic neck treatment.

Importance  Nodal disease has a significant effect on survival of patients with oral squamous cell carcinoma (OSCC). The decision for elective neck dissection for clinically node-negative (cN0) disease remains elusive.

Objectives  To determine the efficacy of prophylactic neck treatment and to assess the value of commonly used clinicopathologic factors associated with nodal disease for early-stage OSCC.

Design, Setting, and Participants  This retrospective study from a population-based cancer registry included patients diagnosed as having OSCC from January 11, 2001, to December 24, 2007, who were identified from the British Columbia Cancer Agency Registry. Comprehensive clinicopathologic data, treatment information, and time to outcome were collected. Five-year overall survival, disease-specific survival, and cumulative incidence of regional failure (RF) were analyzed. Receiver operating characteristic curve analysis with sensitivity and specificity was used to determine the association of these covariates with RF during follow-up. Data were analyzed from January 16 to June 30, 2015.

Interventions  Follow-up of patients with cN0 OSCC with or without prophylactic neck treatment (elective neck dissection [END] and or radiotherapy).

Main Outcomes and Measures  Patient demographic characteristics, clinicopathologic data, treatment data, and time from the initial surgery to last follow-up, the development of RF, or death due to oral cancer or other causes.

Results  Of the 469 patients with cN0 primary OSCC who underwent intent-to-cure surgery for the intraoral lesion, 447 received local excision (LE) for the primary tumor (256 men [57.3%] and 191 women [42.7%]; mean [SD] age, 63.3 [14.7] years). Patients who received prophylactic treatment of the neck (n = 125) compared with LE only (n = 322) had no survival advantage. The estimated 5-year overall and disease-specific survival rates were 61.9% (95% CI, 56.5%-67.8%) and 80.8% (95% CI, 76.1%-85.6%), respectively, for the LE-only group; 54.4% (95% CI, 45.9%-64.5%) and 73.1% (95% CI, 65%-82.3%), respectively, for the LE + END ± radiotherapy group; and 61.7% (95% CI, 52.3%-72.8%) and 80.3% (95% CI, 72%-89.4%), respectively, for the LE + END group. Among the patients with cN0 disease receiving LE only, 89 (27.6%; 95% CI, 23%-33%) developed RF at a median time of 10.8 months, and 71 of the RFs (79.8%) developed within 30 months. Tumor depth of invasion of at least 4 mm and tumor grade of 2 or 3 showed an association with RF but had poor sensitivity and specificity.

Conclusions and Relevance  Commonly used pathologic factors to decide neck dissection for cN0 OSCC are not effective and can cause overtreatment or undertreatment. The need for identification of new objective approaches for risk assessment of RF is urgent.

Each year, 274 000 new cases of oral squamous cell carcinoma (OSCC) occur, accounting for 145 000 deaths worldwide.¹ The reported 5-year survival rates range from 30% to 60% and are among the worst for all cancer types. Despite continued advances in therapeutic techniques, the prognosis of OSCC has not changed significantly during the past 50 years.² The presence of neck node metastasis at presentation or during follow-up is the most important prognostic factor in assessing patient survival.^3-11

Patients with clinically evident neck metastasis (cN+) commonly undergo neck dissection with or without adjuvant radiotherapy (RT), whereas patients with clinically node-negative (cN0) disease may be offered prophylactic neck dissection to prevent regional failure (RF) and associated decreased survival. However, the sequelae of surgery may result in a significant reduction in patients’ well-being and quality of life owing to scarring, depression, or other health indicators that are not often captured by survival analysis.^12-14 Therefore, correct identification of patients with the highest risk for the development of RF is critical to minimize unnecessary morbidity.

Clinicians rely on a combination of clinical, radiological, and pathologic features to provide a management plan for each patient. Despite advances in medical imaging, including high-resolution computed tomographic and magnetic resonance imaging scans and functional positron emission tomographic imaging, a significant amount of occult neck disease cannot be identified by these methods.^5,15-18 Therefore, alternative predictors of nodal disease have been investigated, such as tumor depth of invasion (DOI) and sentinel lymph node biopsy, to help identify those patients with cN0 disease at increased risk for nodal disease. A greater DOI has been associated with an increased risk for nodal metastases, but the value of DOI to rationalize the potential morbidity associated with elective neck dissection (END) remains controversial.^19-24 Sentinel lymph node biopsy has been reported to have high positive and negative predictive values for occult nodal metastasis.^25-33 However, the procedure has not gained widespread use for OSCC because it requires surgeons and pathologists experienced in the technique and results in prolonged surgery time with increased treatment cost.³² Thus, identification of decisive variables associated with nodal disease remains elusive, and the decision for treatment of the neck in cN0 disease rests on the surgeon’s judgment and patient’s circumstances.

This study retrospectively describes the incidence of nodal disease among patients with cN0 OSCC in British Columbia, Canada. We investigate (1) the efficacy of prophylactic neck treatment for early-stage OSCC, (2) the frequency and timing of the development of RF at follow-up, and (3) the efficacy of commonly used clinicopathologic factors associated with RF for early-stage OSCC.

Data of patients diagnosed as having OSCC from January 11, 2001, to December 24, 2007, were retrospectively retrieved from the British Columbia Cancer Agency Registry. Data were analyzed from January 16 to June 30, 2015. Eligible OSCC included anatomical sites represented by International Statistical Classification of Diseases and Related Health Problems, Tenth Revision codes C02.0 to C06.9, all tumor stages, and cN0 status at the time of initial disease presentation. We excluded patients whose anatomical sites were the base of tongue, tonsillar area, and lips. Information on age, sex, and tumor anatomical sites were extracted directly from the Registry. Medical record review was performed to collect history of tobacco use, initial clinical tumor stage, treatment types, tumor grade, and tumor DOI.

Patients with cN0 disease who underwent local excision (LE) with curative intent, with or without neck dissection, were included in the study. Elective neck dissection was defined as neck dissection performed at the time of LE of the primary lesion or after the pathologic evaluation of the primary tumor. Outcome events included in this study were (1) 5-year overall survival (OS), calculated from the time of initial treatment to the date of death due to any cause or the last known date alive; (2) 5-year disease-specific survival (DSS), calculated from the time of initial treatment to the date of death due to oral cancer or the last known date alive; and (3) 5-year cumulative incidence rate of RF, calculated from the time of initial treatment to the development of pathologically proven squamous cell carcinoma in regional neck lymph nodes.

This study was approved by the institutional review board of the University of British Columbia, who waived the need for informed consent for this retrospective registry review.

Statistical analysis was performed with R software (version 3.2.3; https://cran.r-project.org/bin/windows/base/old/3.2.3/). Patient demographic, clinical, and tumor characteristics were summarized in descriptive absolute and relative frequencies and as means with SDs. Categorical variables were compared using the χ² or the Fisher exact test. Continuous variables were compared using the 2-tailed t test.

We used Kaplan-Meier survival analysis to estimate the proportion of survival with log-rank tests for between-group comparison. Cumulative incidence rate of RF was calculated after adjusting for death due to any cause, a competing risk factor. We used the Cox proportional hazards regression analysis to estimate the effect of variables (age, sex, history of tobacco use, lesion site, tumor size, presence of nodal disease, or initial treatment types) on survival. Univariate and multivariate logistic regression analyses were performed to ascertain the effects of these variables on the likelihood that patients with cN0 disease developed RF. Receiver operating characteristics analysis was performed to determine the association of DOI with true RF vs true RF-free outcomes in patients with cN0 disease. Tests were 2-tailed, with P < .05 considered significant.

We identified 3054 eligible patients from the British Columbia Cancer Agency Registry. Among them, 861 patients (28.2%) had primary OSCC, of whom 617 (71.7%) had cN0 disease. Figure 1 illustrates the study population with nodal status at the initial presentation, initial treatment types, and nodal status at the initial treatment and during follow-up.

For the 447 patients with cN0 disease who received LE for the primary tumor and constituted the study population (256 men [57.3%] and 191 women [42.7%]; mean [SD] age, 63.3 [14.7] years), 322 (72%) received LE only and 125 (28%) received LE with additional prophylactic neck treatment, including END with (n = 25) or without (n = 100) adjuvant RT (Figure 1). Of the 469 patients, 22 (4.7%) received radiotherapy for positive margins or adjuvant intent after LE. This may confound the outcome of nodal disease. Thus, they were excluded from the downstream analysis. Among the 125 patients, 15 (12%; 95% CI, 6%-18%) had pathologically positive nodes (pN+) at the time of END, whereas 12 (9.6%; 95% CI, 4.3%-15%) developed RF during follow-up. For 322 patients with cN0 disease who underwent LE only, 89 (27.6%; 95% CI, 23%-33%) developed RF at follow-up.

To examine whether survival was improved for patients with prophylactic neck treatment, we compared the LE-only group (n = 322) with the prophylactic treatment group (LE + END ± RT) (n = 125) or the LE + END group (no RT) (n = 100) (eTable1 in the Supplement). The estimated 5-year OS and DSS rates were 61.9% (95% CI, 56.5%-67.8%) and 80.8% (95% CI, 76.1%-85.6%), respectively, for the LE-only group; 54.4% (95% CI, 45.9%-64.5%) and 73.1% (95% CI, 65%-82.3%), respectively, for the LE + END ± RT group; and 61.7% (95% CI, 52.3%-72.8%) and 80.3% (95% CI, 72%-89.4%), respectively, for the LE + END group. Compared with the LE-only group, the LE + END ± RT group showed hazard ratios of 1.3 (95% CI, 0.9-1.8) and 1.4 (95% CI, 0.9-2.3) in 5-year OS and DSS rates, respectively. Similarly, the LE + END group showed hazard ratios of 0.99 (95% CI, 0.7-1.7) and 1.0 (95% CI, 0.7-1.5) in 5-year OS and DSS rates, respectively. We found no statistical difference between the LE-only and the LE + END ± RT groups in 5-year OS and DSS rates (Figure 2 and eTable2 in the Supplement).

To avoid any confounding factors caused by the delivery of prophylactic neck treatment, a subset of patients with cN0 disease who received LE only (n = 322) were included in the examination of the incidence of RF during follow-up and its possible predictors. Patient demographics and clinicopathologic characteristics are summarized in the Table. For convenience, we used cN0→cN0 to represent the group of patients with cN0 disease throughout follow-up and cN0→pN+ to represent the group of patients with cN0 disease at the initial presentation and pN+ disease during follow-up. Among the 89 patients with cN0→pN+, the median time from LE to pN+ disease was less than 1 year (mean, 10.8 months; 95% CI, 9.5-15.1 months), and 71 (79.8%) of them developed pN+ disease within 30 months (Figure 3). In addition, the 5-year DSS was significantly reduced in this group compared with the cN0→cN0 group (32 patients [56.6%] vs 122 patients [91%]).

Using univariate logistic regression analysis, we observed that high-risk lesion sites (tongue and the floor of the mouth), tumor grade of 2 or 3, and DOI of at least 4 mm were associated with RF. In multivariate analysis, only tumor grade 2 or 3 and DOI of at least 4 mm were independent factors. Compared with a DOI of less than 4 mm, patients with a DOI of at least 4 mm showed an odds ratio of 2.0 (95% CI, 1.2-3.6) in developing RF (eTable3 in the Supplement).

When assessing the association of the DOI with RF, the receiver operating characteristics curve analysis showed an area under the curve of 63.2% (95% CI, 55.8%-70.6%). Setting a DOI of 4 mm as the cutoff resulted in poor sensitivity, specificity, and positive and negative predictive values (54.2%, 67.3%, 45%, and 74.8%, respectively). Tumor grade of 2 or 3 had a better sensitivity of 72.3% but poorer specificity of 48.8%. For a high-risk lesion site, sensitivity and specificity were 86.7% and 26.8%, respectively (eTable 4 in the Supplement).

Using the conjunctive consolidation technique^34-36 for various combinations of categories of DOI (4-mm cutoffs) and tumor grades, we have identified 4 new categories with gradients of proportion of pN+ at follow-up from 15.9% to 60%. Compared with the group with a DOI of less than 4 mm and grade 1, the odds ratios were 2.0 (95% CI, 1.0-4.1) for the group with DOI of less than 4 mm and grade 2 or 3 or a DOI of at least 4 mm and grade 1; 4.3 (95% CI, 2.0-9.5) for the group with a DOI of at least 4 mm and grade 2; and 8.0 (95% CI, 2.0-35.1) for the group with a DOI of at least 4 mm and grade 3. For multivariate analysis, only the group with a DOI of at least 4 mm and grade 2 was selected, showing low sensitivity (64.3%) and specificity (18%) (eTable 3 in the Supplement).

We further examined the proportion of patients with cN0→cN0 and cN0→pN+ who would be overtreated or undertreated at their corresponding DOIs (Figure 4). We observed that if an END was performed on all patients who had a DOI of at least 4 mm, only 45 (45%) would have occult metastatic neck disease treated, and 38 (25.2%) would have occult neck disease not addressed as the primary tumor had a DOI of less than 4 mm.

The presence of regional nodal disease is indeed a burden for patients with OSCC. In our study cohort, nearly half of the patients (363 of 847 [42.9%]) had nodal disease at the time of initial presentation (n = 244), at the time of END (n = 15), or within the 5-year follow-up (n = 104). This finding echoes previously published data from other groups.^4,37,38 The present study presents, to our knowledge, the largest population-based cohort of patients with cN0 disease to help address the efficacy of prophylactic neck treatment and to examine the natural history of the disease burden without neck treatment.

Several randomized clinical trials have suggested benefits from elective or prophylactic neck treatment for cN0 disease.^39-42 A Cochrane meta-analysis,⁴³ including 4 small sample-size randomized clinical trials with a pooled total of 283 patients, stated that END reduced the risk for disease-specific death by approximately 40%. The authors admitted that the observed benefits of END in patients with cN0 disease were mostly contributed from older studies. Although the advancement in imaging technology improved the early detection of clinically occult disease at follow-up and the success of salvage neck treatment, 1 newer study⁴¹ did not observe the benefits in DSS between END and no-END groups. This observation concurred with our finding of a lack of significant improvement in survival with the deliverance of prophylactic neck treatment.

The most recently published randomized clinical trial investigating the efficacy of END⁴² enrolled 500 patients with early-stage (T1 and T2) OSCC and suggested a significant improvement in OS and disease-free survival in patients. Notably, the report was premature, with 25% of patients having less than 16 months of follow-up. In addition, disease-free survival, defined as the interval from the date of randomization to the first documented evidence of any disease or death due to any cause, is clinically less meaningful than DSS and can lead to false conclusions. For example, for patients in the control group who developed nodal disease at follow-up, disease-free survival does not capture the effect on survival because the first incidence of nodal disease was counted as an event before death could happen. For those in the experimental group (ie, END), the high incidence of clinically occult disease at the time of END (24%) was not counted as an event until the reoccurrence of nodal disease, with significant time-to-event delay in this scenario. Therefore, the statistically significant difference in disease-free survival may inflate the benefits of the experimental group. Clinical equipoise remains in this area.^40,44,45

The presence of nodal disease at the time of surgery or at follow-up is a significant independent prognostic factor on survival. With just more than one-fourth of patients with cN0 disease (89 [27.6%]) developing nodal diseases during follow-up, performing prophylactic neck treatment would result in overtreating 233 patients (72.4%) with cN0 disease. Unnecessary treatment may involve significant cost and morbidity. In a previous report on resource use in surgical procedures for early-stage oral cancer,⁴⁶ LE only required significantly shorter operative time (mean [SD], 4.3 [2.2] vs 1.0 [0.3] hours) and shorter postoperative hospitalization (36 vs 4 hours) compared with procedures that included neck dissection.

With the combination of DOI of at least 4 mm and tumor grade 3, our data showed a significantly high odds ratio. Although this combination seemed to have high sensitivity (92.9%), owing to the low case number (n = 10), the specificity was minimal (2%, meaning high false-negative rates) and may not be clinically useful. In addition, the decision of tumor grade has its inherited problems, including sampling bias, intrinsic interobserver variability, subjectivity, and low specificity (most cases were grade 2), making it an unreliable prognostic factor.^47,48 Preoperative means to measure a true DOI are only approximations at best, and surgical samples are formalin fixed and, in most cases, only available after tumor excision. Therefore, DOI, like tumor grade, cannot be used as a real-time clinical decision-making aid for preoperative or intraoperative planning. In addition, the DOI threshold for nodal metastatic risk may vary across oral cavity subsites.^49-51 Without level 1 evidence, the use of 4-mm DOI as a single indicator to decide the need for END should be of paramount caution. In this scenario, any additional treatment for patients who will never develop nodal disease would result in increased hospitalization, associated costs, patient morbidity, and patient convalescence.

We also observed RF in patients with smaller tumors, whereas patients with thicker tumors remained free of nodal disease (Figure 4). This observation evidently shifts the paradigm that smaller tumors are less aggressive and patients are at lower risk for developing nodal disease. Our data also showed near-plateau numbers of undertreatment and overtreatment at various DOI cutoffs (Figure 4), further indicating the limitation of DOI alone in capturing and justifying neck dissection. The need for objective real-time biomarkers to predict nodal disease for treatment planning at initial diagnostic biopsy is urgent.

Cancer metastasis to regional lymph nodes has been recognized as one of the hallmarks of carcinogenesis and involves a complicated, multistep process.⁵² Because we found that approximately 80% of patients with cN0 disease who developed RF did so within 30 months after initial LE in our cohort, we hypothesize that the tumors with nodal involvement are molecularly different from those without. Moreover, among those with cN0 disease who underwent END, we found a subset of patients with disease refractory to prophylactic neck treatment as evidenced by 18 patients, including 6 of 15 patients with pN+ disease (40%; 95% CI, 15%-65%) and 12 of 110 patients with pN0 disease (10.9%; 95% CI, 5%- 17%) at the time of neck dissection who developed pN+ disease during follow-up despite having received neck treatment. This observation, combined with the unreliability of DOI, may reflect the phenomenon of skip metastasis¹¹ or suggest that subgroups within OSCC are biologically and behaviorally different. Other explanations, such as the presence of micrometastases not identified in the pathologic specimen^53,54 or of insufficient nodal yield at the first surgery,⁵⁵ are also possible. Ongoing efforts from our group and others using advanced molecular and optical technologies may provide promising leads for better understanding.^56,57 If we can identify a higher-risk cohort of patients, then END should confer a survival advantage over “wait and treat,” but if END is practiced in a less discriminating fashion, no such survival advantage is demonstrated, as discussed earlier.

Several inherent limitations of this retrospective study include potential selection bias and a lack of consensus as to when an END was performed. Information on alcohol use or comorbidities that are potential risk factors for survival were missing from most of the patient medical records and are not available for analysis. However, through the funding support from the Terry Fox Research Institute, the COOLS (Canadian Optically-guided Approach for Oral Lesions Surgical) study, a pan-Canadian phase 3 randomized clinical trial for early-stage oral cancer, has completed the accrual of 400 surgically treated patients who are currently being observed.⁵⁸ The results from this trial will be able to confirm findings with direct and indirect cost reports in the near future. In addition, fresh-frozen plasma samples collected in this trial, along with comprehensive annotation and long-term follow-up information, would be submitted for next-generation molecular studies and should definitely enhance our understanding of the biology and clinical behavior of this disease.

Prophylactic neck treatment may not improve patient survival, and although reliance on a DOI of at least 4 mm is associated with a higher incidence of RF, it has poor sensitivity and specificity for nodal involvement among patients with early-stage cN0 disease. Because these current standards of care are unreliable, the need to identify other effective standards, perhaps biomarkers with greater specificity and sensitivity in determining a high-risk group of nodal disease, is alarming. Only after identification of these other effective standards will we be able to accurately investigate whether END could in fact confer a survival advantage.

Corresponding Author: Catherine F. Poh, DDS, PhD, Department of Oral Medical and Biological Sciences, Faculty of Dentistry, University of British Columbia, 2199 Wesbrook Mall, Vancouver, BC V6T 1Z3, Canada (cpoh@dentistry.ubc.ca).

Accepted for Publication: June 24, 2016.

Published Online: August 25, 2016. doi:10.1001/jamaoto.2016.2241

Author Contributions: Ms Liu and Dr Poh had full access to all 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: Liu, Prisman, Poh.

Acquisition, analysis, or interpretation of data: Liu, Durham, Wu, Anderson, Poh.

Drafting of the manuscript: Liu, Durham, Prisman, Poh.

Critical revision of the manuscript for important intellectual content: Liu, Wu, Anderson, Poh.

Statistical analysis: Liu, Poh.

Obtained funding: Poh.

Administrative, technical, or material support: Liu, Durham, Poh.

Study supervision: Durham, Wu, Prisman, Poh.

Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest and none were reported.

Funding/Support: The study was supported by grant 20336 from the Canadian Cancer Society Research Institute, grant 2009-24 from the Terry Fox Research Institute, and the BC Cancer Foundation.

Role of the Funder/Sponsor: The funding sources had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Additional Contributions: Jack J. Lee, DDS, PhD, Anderson Cancer Institute, provided guidance on data analysis. Tony Han, BS, provided editorial assistance. Both received compensation for their roles.