Predictive value of tumor thickness for cervical lymph-node involvement in squamous cell carcinoma of the oral cavity

A meta-analysis of reported studies


  • Shao Hui Huang MSc,

    1. Radiation Medicine Program, Princess Margaret Hospital, Toronto, Ontario, Canada
    2. Department of Radiation Oncology, University of Toronto, Toronto, Ontario, Canada
    Search for more papers by this author
  • David Hwang MB,

    1. Department of Radiation Oncology, University of Toronto, Toronto, Ontario, Canada
    Search for more papers by this author
  • Gina Lockwood MMath,

    1. Department of Biostatistics, Princess Margaret Hospital, Toronto, Ontario, Canada
    Search for more papers by this author
  • David P. Goldstein MD,

    1. Department of Otolaryngology-Head & Neck Surgery, University of Toronto, Toronto, Ontario, Canada
    2. Department Surgical Oncology, Princess Margaret Hospital, Toronto, Ontario, Canada
    Search for more papers by this author
  • Brian O'Sullivan MD

    Corresponding author
    1. Department of Radiation Oncology, University of Toronto, Toronto, Ontario, Canada
    2. Department of Otolaryngology-Head & Neck Surgery, University of Toronto, Toronto, Ontario, Canada
    • Room 5-624, Department of Radiation Oncology, Princess Margaret Hospital, 610 University Avenue, Toronto, ON, Canada. M5G 2M9===

    Search for more papers by this author
    • Fax: (416) 946-6566



Tumor thickness (TT) appears to be a strong predictor for cervical lymph-node involvement in squamous cell carcinoma of the oral cavity (OSCC), but a precise clinically optimal TT cutoff point has not been established. To address this question, the authors conducted a meta-analysis.


All relevant articles were identified from MEDLINE and EMBASE as well as from cross-referenced publications cited in relevant articles. Lymph-node involvement was confirmed and identified as positive lymph-node declaration (PLND) by either pathologic positivity on immediate neck dissection or by neck recurrence identified after follow-up ≥2 years. Odds ratios (OR) were calculated to quantify the predictive value of TT. Negative predictive values (and the percentage of patients falsely predicted to not have PLND [FN-PLND]) were compared to determine the optimal TT cutoff point.


Sixteen studies were selected from 72 potential studies, yielding a pooled total of 1136 patients. Data were examined for the following TT cutoff points: 3 mm (4 studies, 387 patients), 4 mm (9 studies, 778 patients), 5 mm (6 studies, 367 patients), and 6 mm (4 studies, 488 patients). The OR (95% CI) was 7.3 (5.3-10.1) for the overall group. The proportion of FN-PLND was 5.3% (95% CI, 2.0-11.2), 4.5% (2.6-7.2), 16.6% (11.5-22.8), and 13.0% (9.7-16.9) for TT<3, <4, <5, and <6 mm, respectively. There was a statistically significant difference between the 4-mm and 5-mm TT cutoff points (P = .007).


TT was a strong predictor for cervical lymph-node involvement. The optimal TT cutoff point was 4 mm. Cancer 2009. © 2009 American Cancer Society.

The presence of cervical lymph-node metastasis is consistently a strong determinant of survival in patients with squamous cell carcinoma of the oral cavity (OSCC).1-5 The incidence of occult lymph-node metastasis in early-stage tumors (primary site T-categorization T1 or T2) has been reported to be between 27%-40%.6-8 In general, management of the neck, especially using elective neck dissection (END) is advocated when the risk of cervical lymph-node involvement is greater than 15%-20%.9, 10 END may be both diagnostic and therapeutic. It provides pathological information on the status of neck nodes, thus helping to determine the need for adjuvant therapies. As well, it removes any clinically undetectable metastases. While the rate of occult lymph-node metastases in early stage OSCC warrants consideration of END, a significant number of patients undergoing END will have no evidence of regional lymph-node metastases and may be subjected to the potential morbidity of a neck dissection. Efforts to identify reliable parameters that predict the risk of cervical lymph-node involvement should improve elective neck management.

In the 1970s, Breslow established a strong link between tumor thickness (TT) and both tumor-free survival and metastasis in patients with cutaneous melanoma.11, 12 Mohit-Tabatabai et al and Spiro et al first applied Breslow's hypothesis regarding the link between lymph-node involvement and TT to OSCC.13, 14 Since then, many studies have been carried out to test this relationship. Most studies have suggested that TT is a strong predictor for lymph-node involvement in OSCC, including observations from our institution.15-18 However, controversy exists about the optimal TT cutoff point for a clinically relevant risk to the neck of harboring microscopic disease.

The objective of this study was to perform a meta-analysis to assess the relationship between TT and the risk of cervical lymph-node involvement and, furthermore, to determine the optimal TT cutoff point for prompting prophylactic neck management.


Search Strategies

We conducted an electronic search for relevant studies examining the relationship between TT and cervical lymph-node involvement in OSCC. Literature searches were performed through MEDLINE (1966 to January, 2007) as well as an inclusive search of EMBASE. Because both “tumor thickness” and “depth of invasion” are commonly used as synonyms in many of the studies,19 both terms were used for the literature search. Search strategies were developed and independently conducted by the first author and an experienced medical librarian. The search query used for the MEDLINE database was ((“mouth neoplasms/[MeSH Terms]” or “oral cavity” or “tongue” or “floor of mouth” or “buccal mucosa”) and (“tumo(u)r thickness” or “thickness of tumo(u)r” or “depth of invasion” or “invasion depth”)). For the term ‘tumor’, both American and British English spelling conventions were used in literature search. The electronic search was supplemented with papers obtained by crosschecking the bibliographies of relevant articles. Only original articles published in English on human subjects were selected for further review. Studies published later than the search date are not included in the actual analysis.

Selection Process

Two reviewers analyzed each selected article independently according to inclusion/exclusion criteria established by the senior author of this study before review. The inclusion criteria were: OSCC treated with primary surgery; all T and N categories at inception; TT measured on the surgical specimen; TT or depth of invasion described in relation to the risk of cervical lymph-node involvement; and a minimum follow-up period of 2 years for untreated necks. Studies were excluded if they included nonoral cavity cancers or histologies other than regular SCC within the cohort (verrucous carcinoma and variants of squamous cell carcinoma including papillary or adenosquamous histologies were not included); patients treated primarily by nonsurgical therapies; adjuvant or neoadjuvant treatment other than surgery for the untreated neck; TT measured by methods other than on pathologic examination, such as in a biopsy sample or by means of imaging modalities; inability to separate regional lymph-node failure from locoregional failure; or studies reporting only statistical summaries without absolute numbers. The 2 reviewers selected articles and retrieved data independently according to the above inclusion/exclusion criteria. Any discrepancies on the eligibility of the specific article or the retrieved data were reconciled by re-reviewing the article. The reasons for exclusion were also recorded and compared. Each reviewer independently appraised the methodological quality of all studies and abstracted the data from the included studies. Both the terms “tumor thickness” and “depth of invasion” have been used in various studies. For this meta-analysis, we adopted the most prevalent convention of using the term “tumor thickness” irrespective of the actual term used by the original authors. This has generally been defined according to the definition proposed by Moore et al20 to be “the deepest invasion of tumor in the tissue from the mucosal surface” (Fig. 1). In the case of any studies that did not define this explicitly, we accepted those that used the term ‘tumor thickness’ without qualification.

Figure 1.

Schematic diaphragm of tumor thickness measurement is measured from the surface of adjacent normal mucosa to the deepest point of tumor invasion. Ca indicates carcinoma.

Statistical Analysis

Cervical lymph-node involvement was confirmed based on 1) pathological lymph-node metastasis (PNM) in regional lymph nodes for cases that underwent a neck dissection as part of their initial treatment, ie, pathological positive lymph node(s); or 2) the development of delayed lymph-node metastasis (DNM) for cases that did not undergo a neck dissection as part of their initial treatment but developed a recurrence in the neck after a minimum 2 years of follow-up. The designation “Positive Lymph node Declaration” (PLND) was used as the aggregate observation derived from the detection of either pathologic positivity on immediate neck dissection or alternatively delayed lymph-node recurrence at follow-up. The absolute number of cases with or without PLND for a specific TT cutoff point from each eligible study was recorded. Odds Ratios (OR) for PLND with 95% Confidence Intervals (CI) were calculated for individual studies using the Peto estimates method21 to detect heterogeneity of each study. OR for the pooled data were calculated using the DerSimonian and Laird method21 to determine the predictive value of TT. The pooled OR represents the effect of comparing small TT to large TT where the underlying continuous TT parameter has been dichotomized with some variability because the TT cutoff points between studies varied from 3 mm to 5 mm. TT cutoff points of 3, 4, 5, and 6 mm were assessed using negative predictive values (NPV). The percentage of cases with PLND below the TT cutoff point (ie, percentage falsely predicted negative [FN-PLND, = 1-NPV]) and 95% confidence interval (CI) were calculated and compared between TT cutoff points using Logistic Regression with Generalized Estimating Equations to account for study clusters. Positive predictive values were not compared because of variability in inclusion criteria across studies. Specifically, the outer limit of tumor thickness varied between studies but was not sufficiently described to allow comparison. As a result the interpretation of positive predictive value is problematic for these data. For all analyses statistical significance was defined as P ≤ .05.


Study Selection

After we combined electronic searching results from the 2 electronic searching databases as well as cross-referencing citations from relevant articles, 72 full-text original studies were selected for further review. Sixteen studies (15 retrospective and 1 prospective) that satisfied all the inclusion/exclusion criteria were chosen for this meta-analysis.15, 16, 22-36 Among them, 9 studies included clinical T1-2 before surgery, 10 contained clinical N0, and 7 addressed clinical stage I&II (T1-2N0). Twelve studies clearly defined TT following the recommendation of Moore et al to measure “from a theoretical reconstructed normal mucosal line to the deepest portion of tumor”.20 Two studies stated that TT was measured from the surface of the tumor excluding keratin, parakeratin, and inflammatory exudates. The remaining 2 did not specify how the TT was measured. Eight studies used an ocular micrometer to measure TT with an accuracy of 0.1 mm while the remaining 8 studies did not indicate this. In all studies, TT was measured on formalin-fixed, paraffin-embedded, tissue samples obtained from primary surgery. The details of the eligible studies are listed (Table 1).

Table 1. Eligible Studies Chosen for This Meta-analysis*
NamePLND CategoryDisease SiteStageTotal No.TT cutoff point
  • PLND indicates positive lymph node declaration; PNM, pathological nodal metastasis; DNM, delayed nodal metastasis; FOM, floor of mouth; TT, tumor thickness.

  • *

    TT cutoff point indicated in this table was derived from source data but not necessarily the only cutoff point used in the original study. Some of eligible studies divided source data into ‘<=’ vs. ‘>’ categories, while others ‘<’ vs. ‘>=’. Since the possibility of TT measured exactly ‘=’ 3.0 or 4.0 or 5.0 or 6.0 mm is rare, we have grouped studies “<=” and “<” certain TT cutoff point together as “<=”; and grouped ‘>’ and ‘>=’ together as ‘>’. In this meta-analysis, 4 studies were used twice for 2 different TT cutoff points and 1 study was used 4 times for 4 different TT cutoff points.

Asakage 199816DNMTongueT1-2N0444 mm
Fukano 199722DNMTongueT1-4N0305 mm
Iwai 200223DNMTongueT1-3 all N304 mm, 6 mm
Kane 200615PNMOral cavityT1-2N0485 mm
Kurokawa 200224DNMTongueT1-2N0504 mm
Lim 200425PNMTongueT1-2N0564 mm
Mishra 199926DNMBuccal mucosaT1-3N01763 mm, 4 mm, 5 mm, 6 mm
O'Brien 200327DNMOral cavityT1-2 all N1454 mm
Rodolico 200428DNMLower lipT1-3 all N953 mm, 6 mm
Sheahan 200336DNMOral cavityT1-2N0215 mm
Shintani 199730DNMTongueT1-2 all N443 mm, 5 mm
Sparano 200431PNMTongueT1-2N0454 mm
Steinhart 199332PNMFOMall T all N485 mm
de Visscher 199833DNMLower lipT1-4 all N1684 mm, 6 mm
Williams 199434PNMOral cavityT1-3N0644 mm
Yuen 200235DNMTongueT1-2N0723 mm
TotalPNM: 5  1136 
DNM: 11

Study Population

The selected studies yielded a pooled total of 1136 patients for this meta-analysis. The disease subsites were as follows: 47% (537 of 1136) oral tongue, 17% (192 of 1136) buccal mucosa, 12% (137 of 1136) floor of mouth, and 23% (263 of 1136) lower lip. The pooled data were further examined by varying the TT cutoff points. This provided 4 studies for TT cutoff point at 3 mm, 9 at 4 mm, 6 at 5 mm, and 4 at 6 mm yielding a pooled subtotal of 387, 778, 367, and 488 data observations, respectively. Although 1136 patients were involved in the study, a total of 2020 data observations were used in the analysis because of the nature of the source data: 4 studies23, 28, 30, 33 were used twice for 2 different TT cutoff points and 1 study26 was used 4-times for 4 different TT cutoff points.

Odds Ratio

The odds ratio (OR) with 95% confidence interval (CI) for PLND based on TT for the individual studies are shown in Figure 2. The OR varied considerably from 3.2 (95% CI, 0.8-12.1) to 57.1 (13.1-248.2). For the overall group, T1-2, N0 and the stage I and II (T1-2N0) subgroup, the OR (95% CI) of PLND were 7.3 (95% CI, 5.3-10.1), 6.2 (4.1 – 9.3), 8.8 (5.3 – 14.7) and 7.1 (4.3-11.8), respectively (Table 2).

Figure 2.

Forest plot for all studies in the meta-analysis is depicted. The width of a rectangle indicating the odds ratio (OR) on the Forest plot is proportional to the size of the study. ORs ranged from 3.2 (95% confidence interval [CI], 0.8-12.1) to 57.1 (95% CI, 13.1-248.2) among individual studies.

Table 2. Odds Ratio and 95% Confidence Interval of Cervical Lymph-Node Involvement Based on Tumor Thickness for the Overall Group and Various Subgroups
GroupNo. of StudyOR95% CI
  1. OR indicates odds ratio; CI, confidence interval.

  2. OR was obtained by including all studies using cutoff points provided in each study (see Fig. 2 for cutoff points).

T1-2 N077.114.3-11.8

Optimal Tumor-Thickness Cutoff Point

The NPV and FN-PLND for the different TT (≤ 3 mm, ≤ 4 mm, ≤5 mm, and ≤6 mm) were calculated (Table 3). There was a significant increasing trend in the FN-PLND as the TT cutoff point increased (P = .03). The incidence of FN-PLND for a TT cutoff point of 4 mm and 5 mm was 4.5% (95%CI: 2.6-7.2) and 16.6% (11.5-22.8), respectively, which was statistically significant (P = .007).

Table 3. Negative Predictive Values (NPV) and Percentage of Falsely Predicted Negative (1-NPV) for Tumor Thickness Cutoff Points of 3 mm, 4 mm, 5 mm, and 6 mm*
TT Cutoff PointNo. of StudiesNo. of Observations at Lower Range of TT Cutoff PointNo. of PLNDNPVFN-PLND (1-NPV)FN-PLND L 95%FN-PLND U 95%
  • NPV indicates negative predictive value; PLND, positive lymph node declaration.

  • *

    FN-PLND, percentage of patients with PLND who fall below the TT cutpoint; FN-PLND L 95% and FN-PLND U 95%, lower and upper limit of 95% confidence interval for FN-PLND respectively. FN-PLND represents the percentage of patients with lymph node metastasis at the given TT cutoff. There was a significant trend for FN-PLND as the TT cutoff point increased (test for trend, P = .03).

  • When the TT cutoff point migrates from 4 mm to 5 mm, the rate of PLND increased from 4.5% to 16.6% (P = .007).

3 mm4113694.75.31.914.0
4 mm93541695.
5 mm61813083.416.69.826.6
6 mm43624787.0133.736.4


This meta-analysis has several unavoidable limitations, including its almost entire reliance on retrospective studies. Diagnostic meta-analysis with retrospective data is relatively new, and its use has only been introduced since 1990.37-41 The statistical methods are not as well defined as those used for prospective randomized trials.42 To ensure data quality, we set up several quality assurance measures in the literature search, selection of eligible studies and retrieving data from eligible studies. In addition, the absence of TT data derived from individual patients in the original studies limited the statistical power for this meta-analysis. Furthermore, vagueness of TT definition in the original studies created uncertainty of source data. In this meta-analysis, we included 2 studies that stated that TT was measured from the surface of tumor excluding keratin, parakeratin, and inflammatory exudates.26, 28 One of these studies stated that for exophytic, papillary lesions, TT was measured from the tip of a papilla to an imaginary line connecting the bases of 2 adjacent pegs.28 As well, the inability to confirm whether regional lymph-node failure was the first site of failure in the original studies creates a further potential confounding of source data. Finally, TT was measured on pathological samples from surgery, which limits the putative pretreatment value of this parameter. However, studies on validating the correlation of TT detected on preoperative ultrasonography and /or MRI are being investigated.23, 30, 43-45

A significant percentage of patients with OSCC present with T1 or T2 disease. Occult metastasis to the neck may occur in up to 40% of these patients with no clinically or radiographically evident regional disease. Neck dissections are performed electively for T1 tumors and frequently for T2 tumors because of concerns about occult regional lymph-node metastases. Management of the clinically negative neck in these patients continues to be controversial. There are 2 commonly used strategies for managing the clinically negative neck: elective neck dissection or observation with treatment reserved for recurrence. Less commonly, the N0 neck is electively managed with radiation. Proponents of elective neck management believe that the incidence of occult lymph-node metastases in all but the very superficial lesions is sufficient to warrant management. Neck dissections should allow for correct staging for prognosis and determination of the need for adjuvant therapies. As well, removal of metastatic lymph nodes would potentially reduce the risk of recurrence. Proponents of an observation strategy believe that the majority of these patients will neither have micrometastases nor develop regional lymph-node metastases, and therefore would be unnecessarily subjected to the potential morbidity of a neck dissection or irradiation.

To reduce the number of therapeutic neck interventions, significant effort has been made in searching for clinical, pathologic and genetic predictors of cervical lymph-node involvement. While commonly used, the T category of the current TNM staging system is not a sensitive indicator for the occurrence of lymph-node involvement in OSCC.20, 46-49 TT, as the “third dimension” of a tumor, has been shown to be a strong independent predictor for cervical lymph-node metastasis in many studies and has also been confirmed in this meta-analysis. The association of TT with lymph-node metastasis is believed to reflect the aggressiveness of tumor growth20 and/or is an objective indicator for the proximity of the tumor to lymphovascular structures.50 Another theory for the positive relationship between TT and the presence of lymph-node involvement is that tumor emboli may be more difficult to form in the small-caliber lymphatics of superficial areas than in the wider lymphatics of deeper tissue.51 Several authors have suggested combining TT with the pathological TNM staging system to obtain a modified pT classification.20, 35, 52, 53

Discrepancy exists over the optimal cutoff point for TT to predict clinical risk of cervical lymph-node involvement. Previous studies have suggested a cutoff point of between 3 mm to 6 mm. In this study, the FN-PLND rate was chosen for comparison to determine the optimal TT cutoff point and should be as low as possible. It is our impression that a high NPV (low FN-PLND) is the most clinically relevant factor for decision making to correctly identify patients who can be spared from unnecessary elective neck management. We found a statistically significant difference in the percentage of FN-PLND between a TT cutoff point of 4 mm and 5 mm. The cutoff point at the 5-mm subgroup had a FN-PLND rate of 16.6% compared with 4.5% for the 4-mm subgroup, which represents both a statistically and clinically significant risk of cervical lymph-node metastases. Thus, we conclude that the optimal cutoff point is 4 mm for consideration of neck management. This coincides with the majority of previous studies.1, 16, 23-27, 31, 33, 34, 48, 54 Patients with TT ≤4 mm can be spared from subsequent END.

Although most studies have confirmed a link between TT and cervical lymph-node involvement, the strength of the association appears to be variable among individual studies. This finding probably explains the large variation in OR for individual studies shown in this meta-analysis and may be attributable to several factors.

First, the studies exhibited heterogeneity in terms of treatment offered and length of follow-up. Upon reviewing the literature for the meta-analysis, we noted that some studies included patients receiving adjuvant therapies such as postoperative radiation. While adjuvant therapies and insufficient length of follow-up would not influence the frequency of lymph-node involvement noted in patients undergoing elective neck dissection, both would potentially influence the reported outcome of late development of lymph-node metastases in patients in whom elective neck dissection was not performed. To avoid this confounding effect, such studies were excluded from this meta-analysis.

Second, differences in results could be attributable to the imprecise definition of TT used between studies. Several ways of defining TT have been used in the previous studies and summarized by Pentenero et al50 as follows: a) from surface/base of the ulcer to deepest point of invasion; b) from adjacent intact mucosa to deepest point of invasion; c) from basal membrane to deepest point of invasion. Moore et al20 found that survival correlated better with TT when measured from a theoretical construction of a basement membrane through the tumor than from the surface of the tumor.20 Kane et al has also compared depths measuring from the surface of a tumor versus from adjacent normal mucosa and concluded that the latter (referred to as microscopic depth) had a stronger correlation with the risk of lymph-node involvement.15

Third, studies are limited by their small sample sizes, with most studies including less than 100 cases. Determination of the optimal cutoff point is extremely data dependent and there may be a variable distribution of cases among small studies even when the selection criteria are consistent.55 It has been suggested that the optimal cutoff point is better determined based on the findings from multiple investigations with a larger cohort.55 As well, there is significant variation in the selection of subsites of the oral cavity included in the studies. The relationship of lymph-node involvement as a function of TT may vary between subsites because of either potential differences in biologic behavior or the ease with which depth can be determined pathologically between subsites. Limitation on sample size has prevented us from performing a subanalysis on subsites.

Lastly, while studies may vary according to T and N categories, most are confined to T1 and/or T2 cases. However, considering the impact of various imaging modalities and technical advances over time, we believed limiting the inclusion criteria to early stage clinical N0 cases might introduce bias on the studies reported from different institutions or over different time periods; therefore, we included all clinical T and N categories in this meta-analysis.

This meta-analysis was performed with a publication cutoff date of January 2007. We have also performed an updated electronic literature search for MEDLINE entries up to January 2008. Six additional studies were found. Upon review of the full text of these articles, 2 were considered potentially eligible.46, 56 However, both were small studies (27 cases for 1 study using a 3-mm TT cutoff point,46 and 53 for the other56 using a 5-mm TT cutoff point) and would have had negligible impact on the current results. In addition to these reports, a recent description of sentinel lymph-node biopsy suggests that the 4-mm TT cutoff point is potentially most appropriate for the risk stratification for cervical lymph-node involvement.57


An association between TT and cervical lymph-node involvement is confirmed in this study. The optimal cutoff point for TT is 4 mm. For oral cavity tumors thicker than 4 mm, prophylactic neck management is generally recommended. Standardization of the measuring method of TT is required for using TT as valid predictor in planning neck treatment. A less invasive method of assessment is needed if this parameter is to be incorporated into clinical TNM staging system.

Conflict of Interest Disclosures

Funding was provided by the Bartley-Smith/Wharton Foundation at the Princess Margaret Hospital.