Prognosis of oral epithelial dysplasia in individuals with and without oral lichen planus

Objectives: To investigate the role of oral lichen planus (OLP) on the long- term prognosis of oral epithelial dysplasia (OED). Methods: Retrospective single- centre cohort study using the 2007– 2019 database of the Head and Neck Cancer and Oral Medicine units of University College London Hospital. The exposure of interest was the presence of OLP, and the prognostic outcomes included the development of new primary episodes of OED, progression to malignancy and mortality. Cox proportional hazard and Poisson regression models were performed. Results: A total of 299 patients, of whom 144 had OED arising on the background of OLP (OLP/OED) and 155 had OED without underlying OLP (non- OLP/OED), were included. A pre- existing diagnosis of OLP was significantly associated with a twofold increased risk of subsequent primary OED events (HR = 2.02, p = 0.04), which also developed faster (1.46 vs. 2.96 years, p = 0.04) and with more involvement of non-cancer- prone sites ( p = 0.001) than in the non- OLP/OED group. There was no difference between groups in the progression to malignancy or mortality. Conclusions: Oral lichen planus/OED patients are at higher risk of multiple episodes of primary OED, which can develop faster and at non- cancer- prone sites as compared to non- OLP/OED individuals. Further research is needed to clarify the effects of OLP upon progression to OSCC and mortality.

Little is known, however, regarding the prognosis of oral epithelial dysplasia (OED) in patients with OLP. This is a notably relevant issue as 10% to 25% of individuals with OLP or lichenoid disorders are suggested to develop OED at some point during the course of their disease (De Jong et al., 1984;Patil et al., 2015;Shearston et al., 2019).
Epithelial dysplasia represents the most commonly used predictor of an oral potentially malignant disease progressing to OSCC Mehanna et al., 2009) and a fundamental intermediate step in the transition from normal mucosa to OSCC reflecting the accumulation of genetic abnormalities (Califano et al., 1996;Zhou et al., 2016).
Current data suggest that overall, and regardless of the type of underlying clinical disease, approximately 10% of individuals with OED eventually progress to OSCC (Mehanna et al., 2009;Shariff & Zavras, 2015). Interestingly, Rock et al. (2018) reported no notable difference in the progression rate of OED between individuals with and without OLP. Furthermore, Zhang et al. (2000) reported no differences in loss of heterozygosity associated with OED in individuals with and without oral lichen planus/lichenoid disorders.
These preliminary studies seem to suggest that the progression of OED to cancer and the underlying genetic abnormalities might not be notably different between individuals with or without underlying OLP. However, evidence remains limited by the small sample size and the lack of other important prognostic endpoints such as the development of multiple and multifocal episodes of OED/OSCC (field cancerization) and mortality. Here, we report an observational study of a cohort of patients with OED aimed at assessing the effect of underlying OLP upon the (i) development of new dysplastic lesions, (ii) progression to malignancy and (iii) mortality.

| Study design and setting
The present retrospective cohort study used anonymized data collected as part of a service evaluation assessing the outcomes of care provided to patients with OED. The individuals of interest were identified from the Head and Neck Cancer Multidisciplinary Team (MDT)

| Study population, data collection, and definitions
Individuals diagnosed with OED were ascertained through the above databases and reviewed against the study inclusion and exclusion criteria (Table S1). In brief, we identified individuals with at least one episode of OED between November 2007 and February 2019 and no previous history of OSCC. Oral epithelial dysplasia was defined and graded as per WHO criteria (El-Naggar et al., 2017). We defined the first episode of OED (dysplastic event) as the index OED lesion and patients were stratified into two groups on the basis of their index OED lesion: low risk (mild dysplasia) and high risk (moderate/ severe dysplasia; Warnakulasuriya et al., 2008).
A new primary episode of OED was defined as a new dysplastic event occurring after the previous episode of OED and (i) at a different intra-oral anatomical site or (ii) at the same anatomical site as the previous OED but ≥6 months after complete surgical excision with histopathologically confirmed disease-free margins. Subsequently, we identified within the group of OED the individuals who also had a diagnosis of OLP. The diagnosis of OLP was clinically and histologically confirmed based on the modified WHO diagnostic criteria (Van Der Meij & Van Der Waal, 2003). Individuals who were diagnosed with OLP after the index OED lesion or individuals being diagnosed with both OED and OLP at the same time were excluded. Individuals with a diagnosis of OSCC within 6 months after the index OED lesion were also excluded due to the likelihood of their malignant disease being synchronous to the OED (Lumerman et al., 1995). Figure S1 presents a flow chart of the process of database patient identification and selection. Details on the observation period, end of data collection and the list of the data retrieved from the hospital records are provided in the Appendix S1 document.

| Study objective and outcomes
The main objective was to assess the number of outcomes relevant to the prognosis of OED. The outcomes of interest included (i) the development, after the index OED, of subsequent new primary OEDs and related time to onset; (ii) the total number of new primary episodes of OED, their site and degree; (iii) the progression to OSCC | 3 PIMOLBUTR et al. and related time to onset; (iv) the total number of OSCCs and (v) mortality related to OSCC and other causes.

| Exposure of interest
The exposure of interest was the presence of OLP. The study population was divided into two groups based on the exposure of interest (exposed and non-exposed to pre-existing OLP).

| Study covariates
Study covariates obtained at the baseline (at diagnosis of the index OED) were considered in the analyses as potential confounding factors including age, gender, smoking (no/yes) and alcohol consumption (no/yes) and the site and grade of index OED lesion. Sites of dysplastic lesions as a variable were grouped into cancer-prone (floor of mouth and tongue) and non-cancer-prone sites (buccal mucosa, gingivae, alveolar mucosa, labial mucosa and palate).

| Statistical analysis
Statistical analyses were performed using STATA 15.1 (Stata Corporation). Demographics and clinicopathological data were described using mean and standard deviation or median and interquartile range (IQR) when data were not normally distributed. Baseline characteristics of OED patients with and without OLP were compared using chi-squared tests (for categorical variables), Student's t-test (for continuous variables) and Wilcoxon rank-sum test (for non-normally distributed continuous variables) as appropriate.
To investigate the risk of developing new primary OEDs and the risk of progression to OSCC after the index OED between subgroups (OED patients with OLP compared to those without OLP), Kaplan-Meier curves by pre-existing OLP were estimated and statistical significance was tested using log-rank tests. Univariate and multivariate Cox proportional hazard regression models were constructed and hazard ratios (HRs) and 95%CI for the risk of having additional new primary OEDs and the risk of progression to OSCC were calculated.
The proportional hazards assumption was verified using Schoenfeld residual tests (Schoenfeld, 1982). In case of the proportional hazard assumption being violated for exposure of interest (OLP) or any of the covariates of the models (hazard ratio associated with that variable not being constant over time), we planned to split data into periods in which the hazard ratio remains constant (where the proportional hazard assumption holds) and fit separate Cox models for each period separately (Bellera et al., 2010;Koletsi & Pandis, 2017).
Univariate and multivariate Poisson regression models were utilized to investigate the relationship between OLP and the total number of additional new primary OEDs as well as the number of new primary OSCCs. To account for differences in observation periods for each individual, the Poisson models were adjusted using the "exposure" option within STATA (Long & Freese, 2014).
Both crude and adjusted incidence rate ratios (IRR) and 95%CI were calculated.
To evaluate the effect of OLP on cause-specific mortality, competing risk analyses were carried out to appropriately account for a competing risk (death due to other causes) as recommended (Läärä et al., 2017). Patients who were found to have died during the study period were grouped according to causes of death -OSCC and all other causes (competing risks). Since Kaplan-Meier curves are not valid when the competing risks are present, cumulative incidence function curves were, therefore, estimated for each cause of death.
Cox proportional hazards regression on the cause-specific hazards of death from OSCC and other causes were fitted and cause-specific hazard ratios and 95%CI were calculated.
All multivariate models were adjusted using a stepwise approach with a view to control for the following potential confounding factors: age at diagnosis, gender, smoking and alcohol consumption, sites and degree of index OED. We planned to use multiple imputations by chain equation (MICE) in order to handle missing values and 40 imputed datasets were generated (m = 40) for variables with missing values assuming missing at random (MAR) mechanism, and imputed values were combined using Rubin's rule (Pedersen et al., 2017). All statistical tests were two-tailed and p-value ≤0.05 was considered statistically significant.

| Patient characteristics
This study included a total of 299 patients, of whom 144 (48.16%) were patients with OED arising on the background of OLP (OLP/ OED) and 155 (51.84%) had OED without underlying OLP (non-OLP/OED). There was a significant difference in gender distribution between OLP/OED and non-OLP/OED patients (p < 0.001), with the percentage of females in OLP/OED and non-OLP/OED groups being 65.97% and 45.16% respectively. Patients in the OLP/ OED group were significantly older than those in the non-OLP/ OED group (mean age 63.49 vs. 59.72, p = 0.02). Moreover, smoking and alcohol consumption were significantly less predominant among OLP/OED individuals compared to those with OED alone (p < 0.001 and p = 0.04, respectively). With respect to the oral mucosal sites of the index OED, non-cancer-prone sites were significantly more common among OLP/OED patients (63.19% vs. 50.32%, p = 0.03). Regarding the grade and treatment of the index OED and subsequent OED, no statistically significant difference was found between OLP/OED and non-OLP/OED groups (Table 1).
Full baseline demographics and clinicopathological characteristics are reported in Table 1. There was no difference in the median follow-up between the OLP/OED and non-OLP/OED group (4.54 vs. 3.77 years, p = 0.08).       individuals were early stage (TNM stages I and II). The difference in the stages of primary OSCCs between the two groups was not statistically significant (p = 0.05).

| DISCUSS ION
The results of the present study show that, in individuals with OED, a pre-existing diagnosis of OLP is associated with a two times Of note, the higher risk of developing new primary OED does not seem to translate into a statistically significant higher total number of OED episodes, although we did observe a tendency towards multiple OED lesions in the OLP/OED group (up to seven metachronous episodes of primary OED as opposed to a maximum of four episodes). It is possible that the limited duration of follow-up and the relatively small number of OLP/OED individuals developing multiple OED lesions may have led to insufficient statistical power to detect a significant difference.
With respect to the sites of subsequent OED episodes, the OLP/ OED group showed a statistically significant higher involvement of non-cancer-prone sites. This is in keeping with the findings reported by Mignogna et al. (2007) and further confirms the possibility of field cancerization (Brinkmann & Wong, 2011;Mohan & Jagannathan, 2014) in patients with OLP.
The present study found no convincing evidence of a greater risk of progression of OED to OSCC in individuals with underlying OLP. This is consistent with the findings of other studies (Rock et al., 2018).
Interestingly, the prevalence of oral cancer development in the OLP/ Bold values indicate statistically significant (p < 0.05). a Adjusted for age, gender, smoking and alcohol consumption, site of primary OED and degree of primary OED.
OED subgroup was 13.19% (19/144), which is notably higher than the 8% (6/73) reported by Rock et al. (2018). This discrepancy is, however, not unexpected as the present study included all degrees of dysplasia, whereas Rock et al. (2018) only included patients with mild and moderate OED at baseline. Furthermore, the present study found no significant difference in the total number of OSCCs or in the time to OSCC onset between groups. However, we observed that none of the 155 individuals in the non-OLP/OED experienced more than one OSCC episode during their follow-up, whereas 7 of 144 patients with OLP-associated OED (4.86%) experienced up to two OSCC events after the index OED, which may suggest a higher tendency towards multiple OSSCs in the OLP/OED group. Of note, our multivariate analyses showed that the degree of index OED was a statistically significant predictor of both subsequent progression to OSCC and the total number of OSCCs, which is in keeping with previous literature (Liu et al., 2011;Mehanna et al., 2009;Warnakulasuriya et al., 2011). This is the first study assessing the impact of OLP up mortality rates in a sample of patients with OED.
Although not statistically significant, our data suggest a trend of nearly fourfold increased mortality related to OSCC in OLP/OED patients (8/10) as compared to the non-OLP/OED group (2/10), as well as a notable difference in 5-year cumulative OSCC-related mortality (5.78% vs. 1.64%). As expected, older age was a significant predictor of OSCC-related mortality and mortality due to other causes (Extermann, 2000;Søgaard et al., 2013 there were only a small number of OED patients who developed further OEDs or progressed to OSCC. Therefore, there is a possibility of type II error (false negative) as a result of underpowered statistics, especially when the confidence interval was wide.

| CON CLUS ION
The results of the present single-centre study differ from previous studies as they suggest that the prognosis of OED in patients with background OLP differs from that of OED patients with no underlying OLP. We found convincing evidence of a higher risk of developing new episodes of primary OED in the OLP/OED group occurring at non-cancer-prone sites. The trend towards a high number of multifocal metachronous OEDs, multiple OSCCs and increased OSCC-related mortality in the OLP/OED group was not statistically significant and therefore not fully convincing. We suggest that further long-term prospective studies would be required in order to clarify the effects of underlying OLP upon OED prognosis, in particular progression to OSCC and mortality. Should future studies confirm the hypothesis of a worse prognosis of OED in patients with background OLP, pragmatic changes in the treatment of this sub-group of OED patients may be considered, including a more extensive surgery where feasible, and possibly the use of systemic chemopreventative agents where available (McCarthy, Fedele,