Prognostic impact of perineural invasion in oral cancer: a systematic review.

Introduction: Numerous studies have evaluated the prognostic significance of perineural invasion (PNI) in oral cancer; however, the results are inconclusive. Purpose: To identify the prognostic value of PNI in oral cancer through a metanalysis.
Methods: A literature review was carried out, searching the MedLine databases via Pubmed, Scielo, Lilacs, Cochrane and Websco.
Results: A total of 56 studies were included. The results indicate that PNI in oral cancer has an incidence of 28% (95% confidence interval (CI) 24-31%); 5-year survival with relative risk (RR) 0.67 (0.59-0.75); 5-year disease-free survival RR 0.71 (0.68-0.75); locoregional recurrence with RR 2.09 (1.86-2.35). Conclusions: PNI is a negative prognostic factor in oral cancer.

Introduction

Among head and neck tumours, lesions of the upper aerodigestive tract stand out, of which 40% are oral cavity tumours [1,2]. Approximately 95% of these are squamous cell carcinoma (SCC) [3,4]. SEER [5] data (Surveillance, Epidemiology and End Results Program) indicate that the estimated incidence of oral cancer for 2020 is 53,260 new cases in the USA (2.9%), with 10750 estimated deaths (1.8%). Its mortality in the period from 2013 to 2017 was 2.5/100,000. Over the past 20 years it has remained stable in terms of incidence and mortality.

The prognosis of oral cavity cancer patients depends mainly on the size of the lesion, level of local invasion, regional lymphatic dissemination and presence of distant metastases [6]; other predictive factors are race, sex, vascular embolisation, perineural invasion (PNI) and histological grade [7]. However, SCC is not the most marked tendency to PNI, such as, for example, adenoid cystic carcinoma. Assessment of prognostic factors acts as a guide for treatment and follow-up decisions [8].

PNI is a process predominantly characterised by neoplasm invasion of the nerves, which can occur in the absence of lymphatic or vascular invasion. Its propagation does not occur through lymphatic dissemination, but through molecular mediators that guide these cells through neural invasion.

PNI was first described in head and neck cancer by Cruveiheir [9] in 1835, being defined more than a century later as the invasion of one of the three nerve layers or the involvement of at least one third of its circumference [10]. Its incidence varies from 2 to 30%, with some reports of 82%. Such variation occurs when there are studies that specifically study PNI and studies where PNI is an evaluated cofactor, as well as discrepancies when there is a slide review with an active search for PNI [11].

The growth patterns of PNI are diverse, and can occur as: intraneural invasion, increasing formation, circular formation and onion-peel formation [12]. Neoplastic cells tend to be concentrated in the perineurium, which is hypovascularised, and can extend up to 12 cm beyond the surgical margin through skip lesions [13]. Tumour growth via the neural pathway can occur in two ways: i) perineural invasion, usually in the small nerves, identified under microscopy; invasion of minor nerves is associated with an increased risk of local recurrence and cervical metastases, and is a predictor of survival, regardless of the risk of capsular rupture [14]; ii) perineural dissemination, where there is gross invasion of the nerve.

The purpose of this study is to analyse the prognostic impact of PNI in oral cancer patients through a systematic review, regarding locoregional recurrence, disease-free survival and mortality.

Methods

Studies that evaluated PNI in patients with oral cavity SCC were included. There was no restriction on the study design, year of publication; all articles in English, Spanish and Portuguese were reviewed. Overall survival, disease-free survival, and locoregional recurrence were evaluated. The MedLine databases were consulted via Pubmed, Scielo, Lilacs, Cochrane and Websco, with the keywords: “mouth cancer” OR “oral cancer” AND “perineural invasion” AND “prognosis”.

Studies where there was no analytical distinction among the multiple sites of primary disease, non-surgical treatment, studies that primarily evaluated surgical technique, studies evaluating a specific head and neck cancer population, duplicate studies or studies with replicated series, in vitro studies, and those that primarily assessed toxicity or quality of life were excluded.

Outcomes were treated as categorical and analysed with relative risk (RR) including 95% confidence interval (95% CI). Significant heterogeneity (occurs when different studies have different designs, for example) was defined as I[2] > 50%. A random effect model was used, except when statistical heterogeneity was not significant. The funnel plot was used to assess heterogeneity. Analyzes developed in RevMan 5.4 and R software, in the “Meta-Analysis” package.

Results

A total of 112 studies were retrieved; after reading all the articles, 56 studies potentially eligible for inclusion in the review were selected.

The incidence of PNI ranged between 3.35 and 63.15%. The risk of bias was assessed using the Cochrane scale, which highly value blinding. Through this scale, we identified a high potential for bias. In our study, the combined incidence of PNI in oral cavity SCC, using the random effect, was 28% (Fig. 1).

Figure 1.

Combined incidence of PNI using a random effect model. An incidence of 28% was identified.

Of the total 5969 patients evaluated for overall 5-year survival, the RR was 0.67 (95% CI 0.64-0.74), with a high rate of heterogeneity when assessing the fixed model, so that the random evaluation model was used, with RR 0.67 (95% CI 0.59-0.75) (Fig. 2). The funnel plot shows publication bias regarding overall survival (Fig. 3).

Figure 2.

Meta-analysis of 5-year survival using a model with a random effect.

Figure 3.

Funnel plot of the 5-year survival meta-analysis, indicating publication bias.

Regarding disease-free survival at 5 years, the outcome was binary (recurrence/total number of subjects observed in the study). A total of 5508 patients were evaluated; the RR was 0.71 (95% CI 0.68-0.75), with low heterogeneity (Fig. 4). Publication bias was also identified (Fig. 5).

Figure 4.

5-year disease free-survival meta-analysis using a fixed effect model. Low heterogeneity.

Figure 5.

Funnel plot of the 5-year disease free-survival meta-analysis, indicating publication bias.

As for locoregional recurrence, a total of 2593 patients were evaluated; the RR was 2.09 (95% CI 1.86-2.35), with a high heterogeneity rate when assessing the fixed model. For evaluation of the random model, the RR was 2.2 (95% CI 1.6-3.01) (Fig. 6). Publication bias was identified (Fig. 7).

Figure 6.

Meta-analysis of locoregional recurrence using a random effect model.

Figure 7.

Funnel plot of the meta-analysis of locoregional recurrence, with publication bias.

Discussion

The search for prognostic factors serves the purpose of better understanding the natural history of cancer, prediction of therapeutic interventions, identification of homogeneous groups of patients, comparison of results of different treatments, identification of groups with unfavorable evolution and planning of follow-up strategies. It also allows for individualisation of treatment, with more aggressive strategies in groups with adverse variables and deintensification of treatment to those with more favourable variables [15].

One of the limitations in the literature about PNI is the lack of standardisation or method for its detection. Despite the definition disseminated by Liebig et al. [15], many authors use broader definitions, with PNI being contact of the nerve with neoplastic cells. Thus, studies of similar methodologies, with similar populations, tend to have different results, since they start from a different principle of PNI, justifying the combined incidence of PNI in oral cancer ranging from 3.35 to 63.15% in our study.

Furthermore, the technique used to detect PNI in studies is not standardised. Most pathologists report PNI as present or absent, without specifying the location, extent, or size of the nerve involved, which is not enough to identify characteristics that may actually cause a change in the impact of the disease (the simple contact of the nerve with tumour cells does not worsen prognosis; however, multiple foci of invasion lead to increased local recurrence) [16].

Thus, there is difficulty in diagnosing PNI. A false-negative result can occur in biopsies [17], more frequently when pathologists are not specifically looking for PNI. Thus, molecular studies are being carried out in order to facilitate the detection of PNI [18]; the expression of N-CAM demonstrates a relation with the presence of PNI [19], but it is not yet investigated in routine clinical practice. In studies where only PNI was evaluated, it has a higher incidence than in studies looking for multiple risk factors [11].

As for clinical diagnosis, between 30 and 40% of patients are symptomatic, and 20% present neurological symptoms when, histologically, the tumour invades the nerve and interferes with its blood supply, causing local oedema, demyelination and segmental infarction [15]. In addition, diagnosis by imaging (MRI) is only performed when PNI occurs in large calibre nerves, when it is possible to identify bone erosion, enlargement of the foramina of the skull base, loss of fat in the pterygopalatine fossa and oedema in the nerve [20].

Neural involvement usually starts with branches smaller than 1 mm, progressing to larger ones. When PNI is present, the surgical margins are no longer controlled by surgery, since tumour progression can occur up to 10 cm beyond its point of origin. Tumour cells tend to be concentrated in the perineurium, a poorly vascularised and relatively hypoxic environment, which leads to a relative radio resistance, corroborating the worst prognosis [14] of PNI.

PNI is associated with an increased risk of lymph node metastases. The presence of PNI is related to the expression of cortactin (a protein that has been suggested that affects the overall aggressiveness of head and neck carcinomas), as well as the presence of lymph node metastasis; it is inferred that their hyperexpression promotes cell migration [21], with the presence of both being related to reduced survival [22]. It also contributes to worse prognosis as PNI is more prevalent in advanced tumours, which in itself is already a negative prognostic factor.

There are studies that associate PNI with increased risk of recurrence and reduced survival [15], while others fail to detect such a relationship [23]. This discrepancy probably results from the way in which the studies are conducted (studies that focus only on detecting PNI, without evaluating other factors such as tumour margin or extension, are more positive) [24].

High analytical heterogeneity of the studies was identified, as well as publication bias. When conducting random evaluations of the sample, we observed that PNI represents a risk factor for lower overall survival, higher mortality, lower disease-free survival and greater risk of locoregional recurrence.

PNI should be considered a marker that indicates a more aggressive tumour behaviour, with a higher rate of cervical metastases [25]; therefore, its presence guides a more aggressive approach. In the absence of lymph node metastasis, PNI leads to greater locoregional recurrence and reduced disease-free survival, as an independent risk factor, despite studies with contradictory conclusions. The divergence in the conclusions of these studies is not due to low sampling, but rather to methodological differences in the studies, ranging from the definition of PNI to its measurement.

Considering that PNI has a negative impact on prognosis of patients with oral SCC, both in relation to overall survival and to disease-free survival, and is also a negative factor for locoregional recurrence, we have a perspective that should be regularly documented in all specimens (with description of both their presence and absence). This histopathological information must be performed in a standardised way, being reported the diameter of the affected nerve, as well as the presence of a single or multiple invasion focus.

Regarding the implications for clinical practice, characterisation of PNI is important for the development of personalised treatment strategies, and should be actively sought by the pathologist in the evaluation of oral cancer. However, including PNI in TNM staging system is not possible at present, due to the difficulty of standardising its identification by anatomopathological analyses. Machine learning methods in pathology may improve its detection on the future, as digital image analysis and the application of artificial intelligence develop furthers. The potential of artificial intelligence (AI) is improve workflow and derive novel insights into disease biology [25] (survival and outcome prediction based on clinicopathological variables); this has been carried out on prostate cancer, breast cancer and cervical cancer with success. Despite the promise of AI models, the translational process to clinical application has been slow [26].

Conclusions

PNI is a negative prognostic factor in oral cancer in terms of overall survival, disease-free survival and locoregional recurrence.

Acknowledgements

Jelis Arenas and Pablo Quintana, by the support at this review.

Conflict of interest statement

The authors declare no conflict of interest.

Funding

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Authors’ contributions

DMVOQ: design of the work, data collection, data analysis and interpretation; drafting the article. RAD: design of the work; data analysis and interpretation; critical revision of the article. LPK: final approval of the version to be published.

Ethical consideration

This study was approved by the Institutional Ethics Committee of Faculdade de Medicina da Universidade de São Paulo (approval number/protocol number 367/17).

The research was conducted ethically, with all study procedures being performed in accordance with the requirements of the World Medical Association’s Declaration of Helsinki.

Written informed consent was obtained from each participant/patient for study participation and data publication.