A simple novel prognostic model for early stage oral tongue cancer

Abstract

The prognostication of patient outcome is one of the greatest challenges in the management of early stage oral tongue squamous cell carcinoma (OTSCC). This study introduces a simple histopathological model for the prognostication of survival in patients with early OTSCC. A total of 311 cases (from Finland and Brazil) with clinically evaluated early stage OTSCC (cT1–T2cN0cM0) were included in this multicentre retrospective study. Tumour budding (B) and depth of invasion (D) were scored on haematoxylin–eosin-stained cancer slides. The cut-off point for tumour budding was set at 5 buds (low <5; high ≥5) and for depth of invasion at 4 mm (low <4 mm; high ≥4 mm). The scores of B and D were combined into one model: the BD predictive model. On multivariate analysis, a high risk score (BD score 2) correlated significantly with loco-regional recurrence ( P = 0.033) and death due to OTSCC ( P < 0.001) in early stage OTSCC. The new BD model is a promising prognostic tool to identify those patients with aggressive cases of early stage OTSCC who might benefit from multimodality treatment.

Oral tongue squamous cell carcinoma (OTSCC) is the most aggressive type of oral SCC and its management may comprise multimodality treatment even in the early stages. Many histomorphological grading models have been introduced for the prognostication of OSCC. The main role of histopathological prognostication is to complement the shortcomings of the TNM staging system for OSCC and thus to provide the clinician with the tools to develop a precise treatment plan. In addition to the complexity of previous models, a number of studies have reported their inability to predict the outcome of patients with early stage OSCC or early OTSCC. This indicates the critical need for a comprehensive predictive model in order to make the appropriate choices in multimodality treatment of early OTSCC. Such choices may entail adjuvant therapy with severe toxic effects, and should consequently be applied only to patients at high risk.

Despite the role of epithelial mesenchymal transition (EMT) in cancer cell migration and metastasis, none of the previous prognostic models has included histopathological parameters to consider EMT. Thus, it would appear advantageous for a prognostic model to include tumour budding, which correlates with EMT in many carcinomas, including tongue SCC. In addition, tumour budding is known to reflect the biological activity of cancer and has been recognized increasingly as a significant, independent, and reproducible prognostic feature.

The depth of tumour invasion or tumour thickness (these two terms have been used synonymously by some authors ), as measured from the surface of the tumour to the deepest point of invasion, has been established as a valuable histopathological indicator representing the malignant potential of early OSCC. Furthermore, it has been reported by many recent and well-documented studies as an important prognostic parameter in early OTSCC.

With this background, the aim of the present study was to introduce a practical and simple model by combining the scores for tumour budding and depth of invasion (BD scoring from 0 to 2) to predict the behaviour of early stage OTSCC. Using this model we aim to be able to select those OTSCC patients who could be at high risk of recurrence and/or distant dissemination of the disease and subsequently dying of the disease.

Patients and methods

A total of 311 cases diagnosed with OTSCC between 1979 and 2010 and clinically staged as cT1cN0cM0 or cT2cN0cM0 were included in this study. Of these cases, 224 were from the Finnish university central hospitals (Helsinki, Oulu, Turku, Tampere, Kuopio) and had comprised part of a previously studied cohort of 233 Finnish patients. The nine patients whose data on recurrence were incomplete were excluded from this latter group. The remaining 87 patients were diagnosed and treated at A.C. Camargo Hospital, São Paulo, Brazil.

Clinical data and follow-up records including cause of death and dates of diagnosis, recurrence, and/or death were retrieved. Surgical resection of the primary tumour was carried out for all patients. With regard to clinical staging, 124 cases (39.9%) were classified as cT1cN0cM0 and 187 cases (60.1%) were cT2cN0cM0. One hundred and sixty five (53.1%) were males and 146 (46.9) were females. Their median age at diagnosis was 63 years (range 10–95 years) ( Table 1 ).

Table 1
Clinicopathological characteristics of the 311 study patients.
Characteristic Number of patients %
Age, years
Median: 63
≤63 157 50.5
>63 154 49.5
Gender
Male 165 53.1
Female 146 46.9
Clinical stage
T1N0M0 124 39.9
T2N0M0 187 60.1
Grade (differentiation)
I (well differentiated) 105 33.8
II (moderately differentiated) 131 42.1
III (poorly differentiated) 75 24.1
Worst pattern of invasion
Superficial 78 25.1
Deep 233 74.9
Lymphocytic host response
Superficial 169 54.3
Deep 142 45.7
Perineural invasion
Absent 269 86.5
Present 42 13.5
Recurrence
No 222 71.4
Yes 89 28.6
Status
Alive 153 49.2
Died of OTSCC 63 20.3
Died of other causes 95 30.5
Tumour budding
Superficial (<5 buds) 215 69.1
Deep (≥5 buds) 96 30.9
Depth of invasion
Superficial (<4 mm) 116 37.3
Deep (≥4 mm) 195 62.7
BD score
Low risk (score 0) 103 33.1
Intermediate risk (score 1) 125 40.2
High risk (score 2) 83 26.7

OTSCC, oral tongue squamous cell carcinoma; BD, tumour budding (B) and depth of invasion (D).

Tumour budding is defined as the presence of a single cancer cell or small cluster of <5 cancer cells at the invasive front. During the scoring of tumour budding, the whole tumour area was scanned at low magnification (×4), then the highest number of tumour budding was counted at a higher magnification (×20) and used as the score for budding. Depth of invasion (or tumour thickness) was measured from the tumour surface to the deepest point of invasion. The cut-off point for tumour budding was set at 5 buds (low <5; high ≥5), and the cut-off point for depth of invasion was set at 4 mm (low <4 mm; high ≥4 mm). In addition, other previously examined parameters, including worst pattern of invasion, lymphocytic host response, and perineural invasion, were also analyzed. The scoring of all parameters was carried out on haematoxylin–eosin-stained sections by an independent observer (A.A.) and reviewed by an experienced head and neck pathologist (I.L.). Both observers were blinded to the patient data and clinical outcome.

Tumour budding ‘B’ and depth of invasion ‘D’ were combined in one model, the BD model, as follows ( Table 2 ):

Table 2
Description of BD scores.
BD score Histological description
Score 0 Tumour with <4 mm depth of invasion and <5 buds at the invasive front
Score 1 This tumour should have only one of the following features:
Tumour with ≥4 mm depth of invasion and <5 buds at the invasive front, OR
Superficial tumour (<4 mm), but with high activity of tumour budding at the invasive front (≥5 buds)
Score 2 Tumour with ≥4 mm depth of invasion and with high activity of tumour budding at the invasive front (≥5 buds)

BD, tumour budding (B) and depth of invasion (D).

Score 0: neither budding nor depth of invasion is higher than the cut-off ( Fig. 1 A, B).

Fig. 1
Scores of the BD model. (A) Score 0: superficial tumour with only large islands at the IF (×20); (B) high magnification of the region inside the rectangle in A showing the absence of tumour budding at the IF (×100). (C) Score 1: presence of tumour budding at the IF of a superficial tumour (×20); (D) high magnification of the region inside the rectangle in C demonstrating the presence of more than 5 buds at the IF (×200). (E) Score 1: deep tumour without any small clusters at the IF (×4); (F) magnification of the rectangle region in E (×10). (G) Score 2: presence of tumour budding at the IF of a deep tumour (×40); (H) high magnification of the rectangle in G showing tumour budding at the IF (×200). A–F: haematoxylin–eosin-stained sections; G and H: pan-cytokeratin-stained sections. IF = invasive front.

Score 1: only one of the parameters (budding or depth of invasion) is higher than the cut-off ( Fig. 1 C–F).

Score 2: both parameters (budding and depth of invasion) are higher than the cut-off ( Fig. 1 G, H).

In accordance with the BD model, each case was assigned a score between 0 and 2. Cases with a score of 0 were classified as low risk, those with a score of 1 as intermediate risk, and those with a score of 2 as high risk.

Statistical analysis

The χ 2 test was used to assess possible relationships between clinicopathological variables and the BD score. The Kaplan–Meier method was used to obtain the estimated time to recurrence and time to disease-specific death. The statistical significance of the differences in Kaplan–Meier curves between the BD groups was determined by log rank test. Univariate analysis was done using the Cox proportional hazard regression model to evaluate the relationship between the BD model scores and recurrence or death from OTSCC. Multivariate analysis (using Cox regression) was performed to evaluate the independent prognostic strength of BD scores when adjusted for other clinicopathological covariates (age, gender, tumour stage, and tumour grade) and worst pattern of invasion. The likelihood ratio statistic was used to test the significance of variables in the estimated models. Statistical significance was set at P < 0.05. Harrell’s C -statistic was used to evaluate the predictive power of survival models. Since the BD score is a linear derivative of both tumour budding (B) and depth of invasion (D), it is not possible to include all three variables in a single model. Thus, two separate models (BD included vs. B and D included) were estimated for both disease-free survival (DFS) and disease-specific survival (DSS) to show their relationship with B, D, and BD variables.

All statistical analyses were done using IBM SPSS Statistics for Windows, version 20.0 (IBM Corp., Armonk, NY, USA) and Stata 13 (StataCorp LP, College Station, TX, USA).

Statistical analysis

The χ 2 test was used to assess possible relationships between clinicopathological variables and the BD score. The Kaplan–Meier method was used to obtain the estimated time to recurrence and time to disease-specific death. The statistical significance of the differences in Kaplan–Meier curves between the BD groups was determined by log rank test. Univariate analysis was done using the Cox proportional hazard regression model to evaluate the relationship between the BD model scores and recurrence or death from OTSCC. Multivariate analysis (using Cox regression) was performed to evaluate the independent prognostic strength of BD scores when adjusted for other clinicopathological covariates (age, gender, tumour stage, and tumour grade) and worst pattern of invasion. The likelihood ratio statistic was used to test the significance of variables in the estimated models. Statistical significance was set at P < 0.05. Harrell’s C -statistic was used to evaluate the predictive power of survival models. Since the BD score is a linear derivative of both tumour budding (B) and depth of invasion (D), it is not possible to include all three variables in a single model. Thus, two separate models (BD included vs. B and D included) were estimated for both disease-free survival (DFS) and disease-specific survival (DSS) to show their relationship with B, D, and BD variables.

All statistical analyses were done using IBM SPSS Statistics for Windows, version 20.0 (IBM Corp., Armonk, NY, USA) and Stata 13 (StataCorp LP, College Station, TX, USA).

Results

Clinicopathological characteristics of the patients are summarized in Table 1 . The median follow-up time was 57 months (range 1–278 months). Sixty-three patients (20.3%) died of OTSCC, 95 patients (30.5%) died of causes other than OTSCC, and 153 patients (49.2%) were alive at the end of follow-up or at the last visit on record.

Loco-regional recurrence (LRR) was reported in 89 patients (28.6%), and 48 (53.9%) of them died from OTSCC. The presence of LRR was significantly associated with poor DSS ( P < 0.001).

Of all histopathological parameters evaluated in this study, only the worst pattern of invasion and tumour budding were associated with both DFS and DSS in the unadjusted univariate analysis ( Table 3 ). Depth of invasion was related to DSS, but not to DFS.

Table 3
Association of disease-free survival and disease-specific survival with the clinical and histopathological parameters (unadjusted univariate analysis).
Variable Disease-free survival (DFS) Disease-specific survival (DSS)
HR (95% CI) P -value a HR (95% CI) P -value a
Age, years 0.003 0.014
≤63 1 1
>63 1.89 (1.23–2.89) 1.87 (1.13–3.11)
Gender 0.715 0.441
Male 1 1
Female 1.08 (0.71–1.64) 1.22 (0.74–1.99)
TNM stage 0.527 0.141
T1N0M0 1 1
T2N0M0 0.87 (0.57–1.33) 1.48 (0.87–2.54)
Grade (differentiation) 0.739 0.199
I (well differentiated) 1 1
II (moderately differentiated) 1.11 (0.68–1.82) 1.68 (0.92–3.07)
III (poorly differentiated) 1.25 (0.72–2.16) 1.58 (0.79–3.16)
Worst pattern of invasion 0.045 0.001
Superficial 1 1
Deep 1.71 (0.98–2.99) 3.29 (1.42–7.62)
Lymphocytic host response 0.512 0.889
Superficial 1 1
Deep 0.89 (0.57–1.33) 0.97 (0.59–1.59)
Perineural invasion 0.199 0.502
Absent 1 1
Present 1.46 (0.84–2.55) 1.27 (0.65–2.49)
Tumour budding 0.005 <0.001
Superficial 1 1
Deep 1.85 (1.21–2.82) 2.59 (1.58–4.26)
Depth of invasion 0.109 <0.001
Superficial 1 1
Deep 1.44 (0.91–2.28) 3.15 (1.60–6.18)
BD score 0.025 <0.001
Low risk 1 1
Intermediate risk 1.50 (0.87–2.58) 3.79 (1.57–9.15)
High risk 2.14 (1.22–3.74) 6.24 (2.59–15.04)
Only gold members can continue reading. Log In or Register to continue

Jan 17, 2018 | Posted by in Oral and Maxillofacial Surgery | Comments Off on A simple novel prognostic model for early stage oral tongue cancer
Premium Wordpress Themes by UFO Themes