Introduction

Predicting the risk of lymph node metastases in clinically N0 patients with oral cancer is very important in treatment planning for the neck. Tumour thickness has been shown to be one of the most important features in predicting lymph node metastases in oral cancer. Since 1986 several studies have focused on tumour thickness [1–3]. In 2005, Pentenero et al. [4] showed in a large review cut-off points as prognostic factors ranging from 2 mm to 10 mm. Table 1 shows cut-off values for all studies used in this article and studies performed after the review of Pentenero et al. [4]. Tumour thickness is not the only important predictive factor; other important characteristics of the tumour are: epidemiologic parameters such as age, sex, race, alcohol, and/or tobacco intake, clinical parameters (TNM stage, site) and histopathological parameters such as the tumour border, being either infiltrative or more pushing, perineural invasion and vascular invasion [4–6]. Recently, it has also been shown that gene expression of the primary tumour correlates with the risk of cervical metastases [7].

Table 1 Reviewed articles: from 1986 to 2009, displayed with number of patients, TNM stage, studied site, and cut-off values found by ultrasound for N0 versus pathological N+ (FOM floor of mouth, N+ tumour thickness significant for nodal involvement)
Full size table

Separate from predicting the development of lymph node metastases, the measurement of tumour thickness may have implications for the treatment of the primary tumour as well. Limited oral cancer in general is treated surgically, using postoperative radiotherapy or chemoradiation in the case of unfavourable histopathological features. In superficial oral cancer, apart from local excision, CO2 laser resections and photodynamic therapy (PDT) have also been shown to yield excellent results [8].

However, the results of these alternative treatments are very much dependent on the amount of tumour infiltration (i.e. tumour thickness). Although depth of infiltration is predictive for the chance of local recurrence as well [6], it is not routinely used as an indication for postoperative radiotherapy in most clinics. A recent study showed that tumour thickness is also predictive for the risk of contralateral cervical metastasis in tongue carcinoma [9].

To measure tumour thickness, different techniques are available. Several studies have compared intra-oral ultrasound (US) with magnetic resonance imaging (MRI) or computed tomography (CT) [10, 11]. In 1989, Iro and Nitsche [12] showed that using a miniaturised transducer enables the depiction of tumours of the anterior tongue and the floor of the mouth without artefacts. They found this method to be superior to conventional US for this region.

Since June 2007 we have been using a new intra-operative transducer (Philips IU-22 L 15-7; 7–15 MHz). In this study, we present our results with respect to the feasibility of US measurement of intra-oral tumour thickness, correlation with the presence of lymph node metastases as well as correlation with MRI, with histopathology as the “gold standard”.

Patients and methods

Ethical considerations

This was a retrospective study with anonymous evaluation of data; therefore, written informed consent was not required.

Selection criteria

All patients with stage T1–T2 oral cancer planned for transoral excision or photodynamic therapy (PDT) in our department were included (n = 65). All patients were seen between 2004 and 2010.

Participants

Ages ranged from 42 to 89 years (mean 65 years) and the sex ratio was 34 men to 31 women. The cancers were localised in the tongue (n = 38), the floor of the mouth (n = 22) and elsewhere in the mouth (n = 5).

Clinical T-stages of the 65 patients were as follows: T1 (n = 50), T2 (n = 15); N-stages were: N0 (n = 59), N1 (n = 4) and N2 (n = 2).

Intra-oral US was used in all 65 patients, and MRI was performed in 36 cases. MRI was only performed as part of the preoperative treatment planning, in patients possibly applicable for transoral excision. In 13 cases the lesion was not found on MRI, because of superficial growth. These lesions were interpreted as minimal lesions of less than 1-mm thick.

Patients underwent transoral excision (n = 30), CO2 laser excision (n = 3) or photodynamic therapy (PDT) (n = 29), or a combination of surgery with postoperative radiotherapy (n = 3). Within the group of patients treated with PDT (n = 29) the localisations were as follows: 13 tumours of the tongue, 13 in the floor of the mouth and three elsewhere in the mouth.

Twelve patients underwent a primary neck dissection.

Histopathological data on thickness were available in all 33 surgically treated patients.

Current treatment protocol

US measurement of tumour thickness is currently used to select patients treatable with PDT. In tumours of less than 5-mm thick, PDT treatment can be used as light penetration up to 1 cm. We chose PDT especially in cases where surgery carried a risk of Warthin’s duct obstruction or a marginal mandibulectomy. For thicker tumours, we always perform a local excision. The neck is staged using US-guided aspiration cytology (USgFNAC) and a sentinel node procedure in the case of a negative USgFNAC and treatment of the primary with transoral excision. For negative sentinel nodes, a wait-and-see policy is the standard. In the case of positive sentinel nodes or a positive USgFNAC, a neck dissection is carried out.

Measurement of tumour thickness

For US measurement of tumour thickness, an intra-operative transducer (Philips IU-22, L15-7; 7–15 MHz) was used (n = 44). Before June 2007, a small parts transducer probe (5–7 MHz) was used in 21 patients. Tumour thickness was measured by placing the probe directly on the lesion. The same radiologist performed all intra-oral measurements on the outpatient clinic, 1 or 2 weeks before surgery without anaesthesia.

Using an US probe or MRI, the thickness of the tumour is measured and not the depth of infiltration from an imaginary line at the level of the mucosal surface. So, in this study we measured maximal tumour thickness also with histopathological sections, and not the depth of infiltration.

On histological sections, the tumour thickness was measured using the ocular micrometer. MRI examinations were performed at 3.0 T (Philips Achieva release 3.2.1, Philips Medical Systems, Best, The Netherlands) using a dedicated 16-channel SENSE neurovascular coil.

The following series were acquired: STIR TSE COR, TR (repetition time), IR (inversion time), TE (echo time) 3,880/180/20 ms, ETL: 12, FOV 300/228/40 mm, matrix: 320/320, 2 nex, slice thickness 4 mm; STIR TSE TRA, TR/IR/TE 4,228/180/20, ETL: 12, FOV: 180/200/80 mm, matrix 300/312, 2 nex, SW 3.5 mm, T1 TSE TRA, TR/TE: 780/10, ETL: 5, FOV 180/180/80, matrix 384/384, 2 nex, slice thickness: 3.5 mm; T1 3D Thrive [performed after intravenous injection of 15 cc gadoterate meglumine(Dotarem)], TR/TE: 5/2,22, ETL:90, TA: 10, FOV 230/272/220, matrix 288/288, 2 nex, slice thickness: 0.8 mm; T1 TSE COR (post contrast): TR/TE: 812/10, ETL: 6, FOV: 180/150/96 mm, matrix: 320/320, 3 nex, slice thickness 3.5 mm.

The US findings of the patients treated with PDT and CO2 laser could not be correlated with histopathological measurements, for obvious reasons. Figure 1 shows the measurement of a tumour using the three techniques.

Fig. 1
figure 1

Tumour thickness on histology, US and MRI. a Histological section showing a tumour of the tongue. Tumour thickness is 8 mm (A indicates diameter). b US picture of the same tumour of the tongue. c Tumour thickness measured on MRI

Full size image

Statistical analysis

The correlation between tumour thickness as measured by US, MRI and in the histopathological sections was evaluated. Pearson’s product correlation analysis was performed on all data in order to determine the overall correlation between them. The correlation between tumour thickness and the presence of lymph node metastases was analysed using the chi-squared test.

Results

Tumour thickness

For all 65 patients, US measurement of the oral lesion was obtained. For 11 cases, intra-oral measurement was not optimal due to limited mouth opening (n = 2) or the impossibility of depicting the lesion (n = 9). There was no difference between the old and the new transducer probe (four versus seven cases). Tumour thickness ranged from 1 mm to 35 mm, with an average of 8 mm.

MRI was performed on 36 patients. Tumour thickness ranged from 1 mm to 24 mm, with an average of 5 mm.

On histopathology (n = 33), tumour thickness ranged from 1 mm to 35 mm with an average of 8 mm.

Correlation with histopathology

Correlation between US and histopathology could be made in 31 cases. For two surgically treated patients, US could not be performed. Correlation between MRI and histopathology could be made in 22 cases. Pearson’s product correlation (Figs. 2, 3) between US and MRI imaging was significantly different. With a correlation product of R = 0.87, US correlated better with histopathology than MRI (R = 0.54). Comparing the old transducer used before 2007 and the new intra-oral probe, the Pearson’s product was respectively 0.05 versus 0.93 (Fig. 4)

Fig. 2
figure 2

Correlation between tumour thickness on histological section and by US; n = 32, Pearson’s product R = 0.87

Full size image
Fig. 3
figure 3

Correlation between tumour thickness on histological section and by MRI; n = 22, Pearson’s product R = 0.54

Full size image
Fig. 4
figure 4

Correlation between tumour thickness on histological section and by US. Sub-selection between a Philips small-parts transducer probe (5–7 MHz) and the intraoperative transducer probe, Philips IU-22)

Full size image

Neck node metastases

In total, 12 patients (N0 = 6, N1 = 4, N2 = 2) initially underwent 12 neck dissections, six therapeutic and six elective. Of these 12 neck dissections, nine were pathological N+ (pN+). In total, 53 clinically N0 patients did not undergo a neck dissection initially. During follow-up, seven patients in this wait-and-see group developed a neck node metastasis. Of these seven patients, three also had local recurrence. Thus, a total of 16 of the 65 (25%) patients were pN+. Of these 16 patients, US could be performed in 13 cases.

Correlation of tumour thickness and lymph node involvement

In the patients with available histopathological measurement, histopathologically measured tumour thickness was used to correlate with lymph node metastasis. For the 32 cases in which histopathology was not possible, the US-measured thickness was used (only measured with the new intra-oral transducer probe). Therefore in total, correlation was possible in 56 patients (Fig. 5). One metastasis (1/20 = 5%) was found in patients with a histological tumour thickness of 3 mm or less. Four patients had cervical metastasis with tumour thickness between 4 and 7 mm (4/22 = 18%). When the tumour thickness exceeded 7 mm, the cervical metastases rate increased to 57% (8/14). The difference in metastasis for 7-mm tumour thickness was significant at p = 0.001. Table 2 shows sensitivity and specificity by cut-off value.

Fig. 5
figure 5

Relationship between tumour thickness and cervical metastases (n = 56). Tumour thickness measured by intra-oral US probe, if not available measured by histopathology N = 56. Green represents the patients with no neck metastasis within the 2-year follow-up period; blue represents the patients with positive neck nodes during the follow-up period

Full size image
Table 2 Sensitivity, specificity and significance level versus cut-off points for tumour thickness
Full size table

Correlation of tumour thickness and lymph node involvement by site

Within the group of tongue carcinomas (n = 36), none of the patients had cervical metastasis with histological tumour thickness of 3 mm or less (Fig. 6). Three patients had cervical metastasis with tumour thickness between 4 and 7 mm (3/18 = 17%). When the tumour thickness exceeded 7 mm, the cervical metastasis rate increased to 44% (4/9). The difference in metastasis for 7-mm tumour thickness is significant with p = 0.03.

Fig. 6
figure 6

Relationship between tumour thickness and cervical metastases divided by site. Tumour thickness measured by intra-oral US probe, if not available measured by histopathology. Green represents the patients with no neck metastasis within the 2-year follow-up period; blue represents the patients with positive neck nodes during the follow-up period. a The group with tongue carcinoma (n = 36); b the group with floor of mouth carcinoma (n = 15)

Full size image

Within the group of floor of mouth carcinomas (n = 15), none of the patients had cervical metastasis with histological tumour thickness of 7 mm or less. When the tumour thickness exceeded 7 mm, the cervical metastasis rate increased to 80% (4/5). The difference in metastasis for 7-mm tumour thickness is significant at p = 0.001.

Local recurrence versus tumour thickness

Tumour thickness measured by histopathology was used for correlation with local recurrences. In our series, 12 patients (12/65 = 19%) had a local recurrence within the follow-up period. In five cases, recurrences occurred at a thickness of 0–4 mm (5/20 = 25%), none in the 4–7-mm group, two in the 7–10-mm group (2/7 = 22%) and no local recurrence occurred in the group with tumour thickness of >9 mm. There was no significant correlation between chance of local recurrence and tumour thickness.

Discussion

Synopsis of key/new findings

Although this is a retrospective study, all measurements were performed only once and by the same observer, and histology was not available in all patients, this study demonstrates that the measurement of tumour thickness with the new generation of intra-oral US transducers correlates well with histopathological thickness in oral cavity cancer (R = 0.93). Between 2004 and 2007, we used a non-dedicated small-parts transducer for measurement and this yielded inferior results (R = 0.05). This clearly shows that it is necessary to use a dedicated small-parts transducer to obtain reliable measurements.

In our study, MRI correlated less well with histopathology than US (R = 0.54). Worldwide, MRI measurements are regularly performed in patients with oral cancer. In our series, 13 of the 36 patients had lesions that could not be measured by MRI because of superficial growth, and these were treated with PDT. Therefore, histopathological measurement could not be performed. We were able to perform correlation in 26 cases; in those cases whereboth MRI and histopathological measurements were available, the correlation was only R = 0.54.

In this study, the risk of metastasis in the neck with tumour thickness of 6 mm or less was 12% (five out of 42), whereas when tumour thickness was 7 mm or more this risk was 57% (eight out of 14). For tongue carcinoma, the risk of metastasis in the neck with tumour thickness of 6 mm or less was 11% (three out of 27), whereas when tumour thickness was 7 mm or more this risk was 44% (four out of nine). For floor of mouth carcinoma, the risk of metastasis in the neck with tumour thickness of 6 mm or less was 0% (none out of ten), whereas when tumour thickness was 7 mm or more this risk was 80% (four out of five). From these data we conclude that within the overall group 7-mm tumour thickness is the cut-off value that optimally stratifies patients at a low versus high risk of development of neck node metastases (p = 0.001). However, as shown in Table 2, it is a matter of compromise between the sensitivity and specificity of a given cut-off value to make a choice for the optimal value. With a cut-off point of 7 mm, the sensitivity in our study was 88% and the specificity 57% (Table 2). By dividing the group by site, the optimal cut-off value would be 7 mm for both groups, although this was determined within small groups (n = 36 vs n = 15).

Comparisons with other studies

Our findings, using the dedicated intra-oral/intra-operative probe confirm those published by Iro and Nitsche [12]. Rasgon et al. [3] showed that the percentage of cervical lymph node metastases in lesions less than 5-mm thick was significantly lower compared with those with lesions with a thickness more than 5 mm. Also, Shintani et al. [10], who studied 24 patients, found a strong correlation between tumour thickness measured with US and histology (Pearson’s product R = 0.98). They found that tumours thicker than 20 mm showed discordant results, possibly caused by tissue constriction during fixation with formalin. Lesions smaller than 10 mm showed a significantly better correlation.

Shintani et al. [10] also compared the accuracy of US and MRI in assessing tumour thickness. They found that US and MRI correlated well with histopathology. However, MRI could not identify most of the tumours less than 5.0-mm thick. Preda et al. [11] showed a significant direct correlation between the measured histological and measured MRI tumour thickness (correlation coefficient = 0.80, p < 0.0001).

In the literature, discussion remains as to which cut-off point is optimal [1–4, 10–30], and in the more recent literature, cut-off points range between 3 and 8 mm [2, 13]. In a large clinical review by Pentenero et al. [4], tumour thickness was shown to be an important parameter for predicting nodal metastases and for survival. They showed that in the literature the cut-off thickness predicting neck metastasis and survival varied from 1.5 mm to 10 mm. Thus, it remains difficult to choose a reliable cut-off point for elective neck treatment and/or adjuvant radiotherapy.

The indication for elective neck treatment is mainly based on the expected risk of occult metastases. With respect to this risk assessment, Spiro et al. [1] retrospectively analysed tumour thickness in 92 patients treated with surgery for tongue and floor of mouth carcinomas. They concluded that for clinically N0 oral cancer, elective neck dissection was indicated in patients with depth invasion of more than 2 mm because in these tumours the risk of metastases reached 40%. Fukano et al. [13], on the other hand, showed that tongue tumours exceeding 5 mm carried a risk of 65% for neck metastases, whereas those infiltrating 5 mm or less had a risk of only 6% (significant difference, p = 0.0003).

A meta-analysis by Huang et al. [2] showed an association between tumour thickness and cervical lymph-node involvement and they stated that the optimal cut-off point for tumour thickness is 4 mm. For oral cavity tumours thicker than 4 mm, prophylactic neck management should generally be recommended according to this study.

At present, little is known about the correlation between tumour thickness and local recurrences of oral cavity carcinomas. In our study, there was no significant difference in local recurrences using the cut-off point for tumour thickness of more or less than 7 mm. Yuen et al. [6] showed in 2002 that tumour thickness is prognostic for both nodal and local recurrence in oral carcinomas. With a tumour thickness less than 3 mm, 0% had local recurrences and 8% had nodal metastases; tumour thickness of more than 3 mm and up to 9 mm had 44% subclinical nodal metastasis and 7% local recurrence; tumour thickness of more than 9 mm had 53% subclinical nodal metastasis and 24% local recurrence. In our series, not enough patients with tumour thickness of 10 mm or more were included to compare these results. Extracting the results from Spiro et al. [1], local recurrence occurred in 5% of the group of tumours with thickness <2 mm (2/40), 9% (3/35) within the group 3–8 mm thick and 25% (4/17) for the group with thickness >9 mm. Local recurrence occurred significantly more in the group with tumour thickness of more than 8 mm, at p = 0.035. Yet more factors like perineural invasion, radical operation, postoperative treatment and N-staging should also be considered in relation to local recurrence [6].

Clinical applicability of the study

From this and other studies, it is clear that state-of-the-art US has a high correlation with histopathological thickness [10, 12]. US thus seems to be the optimal technique in patients with no limited mouth opening or base of tongue involvement. US measurement is more reliable than MRI for the measurement of tumour thickness, especially in superficial lesions.

All currently available literature shows depth of invasion or tumour thickness measurements of the primary tumour, most often only in small groups. The identification of a cut-off with adequate utility for clinical decision-making requires large studies and independent validation.

Conclusion

Tumour thickness is an important predictive marker for lymph node metastases. As such, it can help in decision-making with regard to management of the primary tumour and neck. Based upon our findings, a wait-and-see policy is only warranted in superficial lesions with tumour thickness less than 7 mm, but only if regular follow-up using US-guided aspiration of the neck is ensured [31, 32].