The impact of lymphovascular invasion on survival in oral carcinoma

Introduction 

In the search for prognostic factors in oral carcinoma many variables have been identified and can broadly be placed in the categories of tumour related, patient related and treatment related factors.1

Prognosis in oral cancer is largely determined by tumour related factors. The tumour-node-metastasis (TNM) system of cancer classification recognises this fact.2 However, this staging system does not incorporate histopathological factors at a time when they are increasingly recognised as being significant to prognosis.

The study aim was to relate histopathological features of primary oral carcinomas and neck dissection specimens to survival in oral carcinoma.

Materials and methods 

The management principle of the two units from which the data for this study were sourced is identical. All operable patients with primary intraoral carcinomas are treated with surgery and concomitant neck dissection. Adjuvant radiotherapy is prescribed on pathological analysis of the surgical specimens, if there is evidence of close margins at the primary site (<5mm), greater than one node positive, extracapsular spread or bulky metastatic disease in the neck dissection specimens.

Seventy-two consecutive patients were identified from pathology records. The patients had been treated by one of two consultant maxillofacial surgeons between 1999 and 2003.

All patients had received primary surgical tumour resection with simultaneous neck dissection. The pathological specimens where reported by two pathology departments based at Wythenshawe Hospital and Manchester Royal Infirmary.

All the participating pathologists agreed to use The Royal College of Pathologists Head and Neck Carcinoma Dataset guidelines for reporting the presence of lymphovascular invasion. Tumours were scored positive, if there was obvious carcinoma in endothelial-lined vessels on medium power observation.

For inclusion all patients had to meet the following criteria.

Patient confidentiality was maintained throughout the study. Only hospital record numbers were used for identification and the database was only accessible through a hospital workstation to those with a valid authorisation password.

There were 72 patients included in the final study analysis, of these 51 (71%) were male, and 21 (29%) were female. The mean age of the study group was 60.7 years (SD 13.1, range 32–87). In 74% of cases the primary site was the anterior two thirds of tongue, and floor of mouth. Primaries of the anterior two thirds of the tongue represented 32% of the total group and those of the floor of mouth 42%, respectively. In the other 19 patients, seven (10%) had their tumour sited in the buccal mucosa, six (8%) in the retromolar trigone, five (7%) in the lower alveolus and the remaining single case was in the hard palate. Twenty-one (29%) of the patients were staged as having T1 tumours, 28 (39%) were T2, five (7%) were T3, and the remaining 18 (25%) patients had T4 tumours.

Histopathological features of the primary tumour were ascertained from the pathology records. For all 72 cases specifically the presence or otherwise of residual disease, pattern of invasion, lymphovascular invasion, perineural invasion, dysplasia at the margins, tumour thickness, tumour size, mandibular invasion and the histological grade of the tumour were recorded.

All patients had a documented nodal status established from the pathology reports. Of the 72 cases, 37 (51%) patients were node negative, 13 (18%) had N1 disease and 22 (31%) had N2 disease. Two patients had N2a disease; the remaining cases were N2b. Extracapsular spread of involved nodes was also evaluated. None of the patients had distant metastases at diagnosis. Staging information is given in Table 1.

Table 1. TNM and Group staging details of patients studied according to the AJCC system

		T1	T2	T3	T4
	N0	16	16	0	5
	N1	1	7	0	5
	N2a	1	0	0	1
	N2b	3	5	5	6
	N2c	0	0	0	1
	N3	0	0	0	0

STAGE 1: 16; STAGE 2: 16; STAGE 3: 8; STAGE 4: 32.

There were 12 bilateral neck dissections and as such 78 sides of neck dissection to examine. The total number of nodes harvested was 1699. A mean of 21.5 nodes per neck dissection (range 8–53).

Statistical methods 

The Cox regression method was used to investigate the effect of several variables at any given time. A forward stepwise Cox regression was performed for analysis of multiple variables which may have a prognostic impact on survival in this cohort of patients. Initial screening for prognostic variables was done by univariate analysis. Variables shown on univariate analysis to be prognostically statistically significant were entered into a multivariate analysis by the Cox proportional hazards method. The candidate variable with the highest regression coefficient value squared was chosen. The remaining candidate variables were added on the basis of which increased the regression coefficient greatest, until the point were there was no longer any influence discernible. If the regression coefficient has a positive value then the hazard function is higher and the prognosis worse.

Values of the regression coefficient, β. The standard error, SE. The relative risk, eβ and the P value for the final cox model were determined.

Results 

Univariate analysis 

Twenty variables were assessed in the univariate analysis of which seven variables showed significance at the 5% level for survival. Survival was negatively influenced by six tumour related factors, increasing T stage (P=0.039), increasing N stage (P=0.004), greater than two nodes histologically positive nodal disease (P=0.017), tumour size >4cm (P=0.022), residual disease at the primary site (P=0.012), extracapsular nodal spread (P=0.01) and the one treatment related factor analysed, adjuvant radiotherapy (P=0.039). Following the univariate analysis nodal status was chosen as the candidate variable for the multivariate analysis. Nodal status was the first predictor to emerge on regression analysis. Eleven variables were taken forward for multivariate analysis, chosen to have a significance level below the 20% level on univariate analysis. These included the seven variables shown to be statistically significant in the univariant analysis as well as, smoking status (P=0.17), presence of mandibular invasion (P=0.063), pattern of invasion (P=0.16), dysplasia at the margins (P=0.18) and lymphovascular invasion (P=0.118). There were only two variables which made a significant difference (P<0.05) to the multivariate model. The presence of lymphovascular invasion (P=0.015) and histological evidence of mandibular invasion (P=0.047). The Cox regression was performed via a forward stepwise method though no difference was found in the results on performing backward stepwise regression. The distributions of cases of lymphovascular invasion in relation to nodal status are shown in Table 2.

Table 2. Relationship of lymphovascular invasion to nodal status in cases studied

			Lymphovascular invasion			Total
			No	Yes	Unknown
	Nodal stage	0	29	7		36
		1	3	10		13
		2	13	7		20
	Total		45	24		69

The relative risk of death for patients with evidence of two or more cervical metastases was 3.74 (P=0.005) and the addition of lymphovascular invasion to the model increased the relative risk by 2.99 (P=0.015). These results are relative to patients with no evidence of cervical metastases and those with a single node positive (N0 and N1) which represented the baseline survival curve (Table 3). These relative risks translate into a 78% five year survival rate for those with no cervical metastases and no evidence of lymphovascular spread, compared to 15% in those patients with evidence of lymphovascular spread and two or greater cervical metastases at diagnosis (Fig. 1).

Table 3. Final Cox model

		β	SE	eβ	P
	N0 and N1	−0.172	0.678	0.842
	N2	1.319	0.460	3.741	0.005
	LVI	1.095	0.451	2.990	0.015

N0 denotes no nodal disease; N1 denotes one positive node only (<3cm diameter); N2 denotes one positive node (>3cm diameter), 2 or more positive nodes, or contralateral nodes; LVI denotes lymphovascular invasion; β denotes regression coefficient; SE denotes standard error; eβ denotes relative risk.

View Large Image Download to PowerPoint

Figure 1. Group 1=N0N1/LVI –ve, Group 2=N0N1/LVI +ve, Group 3=N2/LVI –ve, and Group 4=N2/LVI +ve.

Discussion 

The natural history of tumours and the three key events which characterise malignancy are reflected in the TNM system.3 These are T is the ability to grow and locally invade, N is the ability to metastasize to regional draining lymphatics, and M is the ability to metastasize to distant sites.2

In relation to oral cavity tumours these three criteria are independent indicators of prognosis although they are intimately inter-related. Increasing size by T stage leads to an increase in the rate of occult metastases.4, 5 Increasing N stage is associated with the development of distant metastases, particularly if there are multiple neck levels involved, multiple positive nodes, levels IV and V involved or if there is evidence of extracapsular spread.6, 7, 8

The strict inclusion criteria of the study led to a smaller sample size, though this prevented the potential effects that previous radiotherapy, chemotherapy or surgery may have had on pathological features of the primary tumour and neck dissection specimens and also on survival.

The inclusion of only surgically treated cases with concomitant neck dissection in the study, led to the exclusion of smaller tumours (T1) from the overall analysis, as these cases may have been treated with laser excision and observation of the neck from the outset. Patients with large primary tumours (T4b and T4c), those with large nodal metastases (N3), or distant metastases (M1) on presentation, may have been excluded from the study as they did not undergo surgery as primary treatment, if at all.

In this study there was a high incidence of lymphovascular invasion identified in the primary tumour of patients with cervical metastases (51%). The detection of lymphovascular invasion is notoriously difficult in routine histological sections, which has led to it being excluded from some histological grading systems.9 The high incidence of lymphovascular invasion identified at the primary site translated into a statistically significant relationship between lymphovascular invasion and the presence of cervical metastases. These findings are similar to those of others.10, 11 However, not all studies have universally demonstrated a relationship between lymphovascular invasion and the presence of cervical metastases.

It would seem reasonable that the presence of lymphovascular invasion at the primary site would predict for cervical metastases as invasion of the lymphatics is the first step in the development of a metastatic focus.

Vascular invasion does not seem to have had as much attention placed upon it in the literature in relation to oral carcinoma and the development of lymph node metastases and prognosis. The importance of lymphovascular invasion to prognosis in other solid tumours is well recognised.12, 13, 14 Most of the published data would seem to indicate a greater role for perineural invasion in relation to the development of metastases and survival in oral carcinoma. Still there is no consensus view as to which holds the greatest predictive value. Close et al.10 found a highly significant correlation between the presence of lymphovascular invasion in the primary tumour and the development of cervical metastasis, no other histopathological parameter predicted for the development of cervical metastasis in their series.10 The authors published follow up data on this cohort of patients two years later outlining the influence of histopathological factors on local control, neck recurrence, distant metastasis and overall survival. Of all the histopathological factors studied only lymphovascular invasion had an impact on survival on univariate and multivariate analysis.15 Some authors have questioned the value of lymphovascular invasion in the development of cervical metastasis, though often the studies performed have been poorly designed or had small sample sizes.16

In common with other studies the results of this study demonstrate that the greater the number of involved nodes in the neck, the worse the prognosis. Kalnins et al.17 showed a five year survival rate of 75% in patients who were lymph node negative. This figure fell to 49%, 30% and 13% in groups with one, two or three positive nodes, respectively.17 These findings have been confirmed by several other authorities.18, 19 There is no doubt of the prognostic influence of node positive disease on outcome. However, there is gathering evidence that patients with a single node positive in the cervical chain, without evidence of extracapsular spread may have a similar five year survival outcome to those without evidence of nodal disease.20

The local recurrence rate of 14% seen in this study compares favourably with other large series which reported local recurrence rates of between 16 and 20%.21, 22, 23

However, the nodal recurrence rate of 19% was slightly higher than that reported in the same series and others were nodal recurrence rates ranged between 6 and 13%.21, 22, 24

On univariate analysis seven factors were shown to be predictive of survival; T stage (P=0.039), N stage (P=0.004), tumour size (P=0.022), greater than two nodes positive (P=0.017), extracapsular spread (P=0.01) and adjuvant radiotherapy (P=0.039). Residual disease at the primary site was also shown to be predictive of survival on univariate analysis in this study (P=0.012). The finding of residual disease at the primary site was also found to be predictive of survival in the univariate analysis by Woolgar et al. 2003 and analogous to this study was not on multivariate analysis.20 In this study group 53% of the patients had a course of radiotherapy postoperatively. This is similar to that seen in other studies.25 Analysis of survival in relation to the prescription of radiotherapy was performed and revealed a statistically significant difference in survival between the two groups. This is not all that surprising in that the patients receiving radiotherapy are a selected group with adverse prognostic signs from the outset. The addition of variables below the 20% significance level into the multivariate analysis demonstrated that lymphovascular invasion at the primary site was the covariate which contributed greatest to the final Cox regression model. The results of this study show that on univariate analysis lymphovascular invasion was not significant at the 5% level to survival (P=0.118).

The results of the Cox regression analysis were unexpected. There are a number of retrospective reviews in the literature which analyse survival in oral carcinoma. These studies have statistically related numerous clinical and histopathological parameters to survival in patients treated primarily with surgery, followed by adjuvant radiotherapy (if felt to be required on analysis of the surgical specimen).8, 20, 21, 26, 27, 28 Of these studies which used Cox regression analysis, only one study showed lymphovascular invasion to be significant and this was only in the univariate analysis.28 Similar to the findings in this study most published reports find nodal status to be the parameter to have the greatest influence on survival, followed by the presence of extracapsular spread.8, 20, 28

However, none of these studies apart from that performed by Chandu et al.28 had lymphovascular invasion as a predictor of survival in their univariate analysis; presumably they did not include this parameter in their multivariate analysis. Certainly, it would not appear to have been added to the multivariate analysis on examination of their statistical methods and it is unclear from the studies quoted whether lymphovascular invasion was evaluated in their univariate analysis.8, 20, 26, 27

In this study the presence of extracapsular spread ranked second to nodal status on univariate analysis though it had no additional influence on survival in the multivariate analysis and subsequent Cox model.

The influence of lymphovascular invasion in other solid tumours on the development of distant metastases and prognosis has been well documented.12, 13, 29, 30 However, these are tumours with a well recognised tendency for distant metastases.

The final Cox survival model in this study, and the influence of lymphovascular invasion on survival in the presence of higher stages of nodal disease has not been previously reported in oral cancer patients. This study raises the question of whether patients with evidence of lymphovascular invasion following primary surgical management should be selected for aggressive post operative adjuvant therapy. Even in the baseline hazard group (nodal negative and single node positive) there was an influence demonstrated on survival in the presence of lymphovascular invasion. The small sample size of this study coupled with its retrospective nature limit the inferences which can be taken from it. A larger multicentered prospective study is needed to examine the influence of lymphovascular invasion on survival in oral cancer patients. If such a clinical trial can confirm the findings of this study as regards the influence of lymphovascular invasion on survival, it may allow for the selection of patients for adjuvant chemoradiotherapy and targeted molecular therapy against tumours with evidence of lymphovascular invasion in the first instance.

Conflict of Interest Statement 

None declared.