UPPER VERSUS LOWER LIP CANCER: ARE THERE REALLY DIFFERENCES?

Transcription

1 STAGE WETENSCHAP UPPER VERSUS LOWER LIP CANCER: ARE THERE REALLY DIFFERENCES? Nienke Suzanne Pietersma December April 2012 Integraal Kankercentrum Nederland (IKNL), Groningen Directe begeleider: Dr. BAC van Dijk; Epidemiologie, IKNL Facultair begeleider: Prof. Dr. GH de Bock; Oncologische Epidemiologie, UMCG Prof. Dr. JLN Roodenburg, Hoofd-Hals oncologie, UMCG Dr. JGAM de Visscher, Hoofd-Hals oncologie, MCL 1

2 Summary Introduction Lip cancer is a rare malignancy of the head and neck region in the Netherlands. Generally, the impression exists that upper lip (UL) cancer has a worse prognosis than lower lip (LL) cancer. However, this is based on few studies and conflicting results have been reported. Objectives The main objective was to investigate the difference in survival between UL and LL cancer. Additionally, we investigated which factors are associated with a risk of dying. Materials and methods Data for were derived from the Netherlands Cancer Registry of the North-Eastern region. Only primary squamous cell carcinoma (SCC) of the vermilion of the lip were included. Information on patient characteristics, tumour characteristics and treatment was collected. The crude 5-year overall survival (OS) rates and Hazard Ratios (HR) were calculated. Results In this study 1126 SCCs of the vermilion border of the lip, observed in 1105 patients, were included. Most tumours were found in men (71.9%). Almost 90% presented on the lower lip (predominantly in men) and 10% on the upper lip (predominantly in women). Between the upper and lower lip SCCs, there were no significant differences in age distribution, tumour grade, tumour stage or involved surgical margins. The 5-year OS rate was the same for both tumour sites: 74%. In sexstratified analysis, also no statistically significant differences in survival between UL and LL SCCs were found. Prognostic factors associated with a worse outcome were: male sex (HR 1.46; 95%CI ), increasing age (per 10 years, HR 2.07; 95%CI ), and high tumour stage (HR 2.60; 95%CI ). Conclusions Our results suggest that there is no difference in survival between upper and lower lip SCC. List of definitions and abbreviations SCC NCR NNCR ICD-O TNM-classification PA reports Vermilion border lip UL LL SL OS HR CI Squamous Cell Carcinoma Netherlands Cancer Registry North-Eastern region of the Netherlands Cancer Registry International Classification of Diseases for Oncology Cancer staging system that describes the extent of cancer in a patients body. T (Tumour), N (LymphNode), M (Metastasis). Pathology reports Red of the lip ( lipstick zone ), oral mucosa not included. Vermilion (border) of the upper lip (ICD-O C00.0) Vermilion (border) of the lower lip (ICD-O C00.1) Skin of lip, skin surrounding the red of the lips. (ICD-O C44.0) Overall survival Hazard Ratio Confidence interval 2

3 Table of contents SUMMARY... 1 LIST OF DEFINITIONS AND ABBREVIATIONS... 2 TABLE OF CONTENTS INTRODUCTION An introduction to lip cancer Aim of the study Hypothesis Overview of this MD thesis LITERATURE Research strategy Literature review Incidence Anatomy and Histology Risk factors Treatment Prognosis and Survival MATERIALS AND METHODS Cancer registration Patient population and inclusion criteria Data collection and Database check Variables and Endpoints Statistical Analysis RESULTS Database check for tumour site Characteristics of original data Sex and age Tumour grade, metastasis at presentation and tumour stage Treatment Clinical outcome: Overall survival Prognostic factors DISCUSSION Summary of results Preliminary conclusion Results in relation to literature Characteristics Survival Prognostic factors Strengths and Weaknesses Conclusions Implications for future research REFERENCE LIST APPENDIX A: DUTCH SUMMARY APPENDIX B: EVIDENCE TABLE APPENDIX C: KAPLAN-MEIER CURVE FOR UL, LL AND SL APPENDIX D: TABLE OF CHARACTERISTICS FOR REFERENCES APPENDIX E: PUBMED SEARCH STRATEGY

4 1 Introduction In this first chapter we will describe the outline of this MD thesis. To this purpose, a brief introduction to the subject of lip cancer is given in paragraph 1.1. In paragraph 1.2, the aim of the study, research questions, and hypothesis are described, followed by an overview of the chapters and appendices of this thesis in paragraph An introduction to lip cancer Carcinoma of the lip is a rare malignancy of the head and neck region. The incidence rate is 0.65 per per year for the Netherlands a. The incidence of lip cancer varies widely throughout the world. The highest incidence rates are found in Australia and Canada, while it s virtually unknown in parts of Asia 1. Over 90% of lip cancers are squamous cell carcinoma s (SCCs). The vast majority, about 90%, of SCCs are located on the vermilion of the lower lip ( lipstick zone ) 2-5. The percentage of lower lip cancers has decreased significantly over the last 30 years, while the percentage of external upper lip cancer has increased over this period 1,6. Men, fair-skinned individuals and people over 50 years of age have a higher risk of developing lip cancer. Additional risk factors significant for the development of lip cancers are cumulative life-time exposure to sunlight, smoking habits, outdoor occupation, rural residency, socioeconomic status and immune status 1-3,6-12. Most of these factors are correlated to each other, in particular to the exposure to sunlight. SCC of the lip carries a good prognosis which is attributed to early detection and the relative infrequent occurrence of regional metastases 3,6,9. The 5-year survival rate is 92% in the Netherlands a. The prognosis for cure depends on the extent of the disease at the time of presentation. The most powerful predictor of survival is the presence or absence of cervical lymph node metastases 13. SCCs of the vermilion have a low risk for regional metastases, but when metastases develop they are mostly found in levels IA and IB in the neck 14. Distant metastases may occur if there is a high stage and regional metastases in the neck are present. Prognostic factors for local recurrence are large tumour size and surgical margins containing SCC, whereas regional metastases are statistically associated with increasing tumour thickness, an infiltrative invasion pattern, and perineural invasion 13,19. Some studies have reported a lower 5-year survival rate for upper lip cancer than for a comparable lesion of the lower lip 3,15. However, other studies reported the opposite 6 or couldn t find a difference at all 16. In our opinion, worse prognosis for upper lip cancer is not inconclusively proved. The aim of this study is to investigate the differences between upper and lower lip cancer. The main focus is survival and determining prognostic factors associated with these two anatomical sites of lip cancer. 1.2 Aim of the study Investigating the difference in survival between upper and lower lip cancer and determining prognostic factors associated with these anatomical sites of lip cancer. Despite the vast amount of data and studies on oral cavity cancer, relatively few researches have been performed regarding lip cancer. Few large population studies on survival of lip cancer have been performed and in these studies a distinction between upper and lower lip cancer is rarely made. a National Cancer Registry Netherlands, period ; ( consulted December 2011) 4

5 Aim of this study is to determine if a difference in survival between upper lip and lower lip cancer exists. Further we want to investigate the prognostic factors associated with these two anatomical sites of lip cancer. Research questions: 1. Is there a difference in survival between SCC of the upper and lower lip? 2. What are the prognostic factors associated with the survival of upper lip and lower lip SCCs? 1.3 Hypothesis The survival for patients with upper lip cancer is lower than the survival for patients with lower lip cancer. In this project, the difference in survival between upper lip and lower lip cancer was studied. Few studies on this subject have been performed and although conflicting data on the subject exist, the medical world generally has the belief that upper lip cancer has a worse prognosis than cancer of the lower lip 2,4,17 Most articles and books on oncology refer to the study performed by Zitsch et al. 3. This study however, only included 39 UL cases on a total of 1047 patients (3.7%). Despite the conflicting results that have been published, we still expect the survival of upper lip cancer to be lower than cancer of the lower lip. This hypothesis is based on the reports in oncology books and earlier studies. 1.4 Overview of this MD thesis In chapter 1 of this MD thesis, an introduction to lip cancer will be given and the aim of the study is described. The research questions are formulated, as well as our hypothesis. In chapter 2 a review of the literature is presented together with our literature research methods. The most important findings are highlighted. Incidence, histologic features, risk factors, treatment and prognosis for lip cancer are discussed. The materials and methods used to answer the research questions are defined in chapter 3. The inclusion criteria, data collection, studied variables and statistical analysis are explained. Chapter 4 contains the results of this study. First, the characteristics of data will be addressed, followed by results on survival and prognostic factors. The tables and survival curves will illustrate and clarify the results. The research questions of this study are answered in chapter 5. The results will be discussed in relation to literature. Finally, conclusions and implications for future research are formulated. Appendix A contains a Dutch summary of this thesis. Appendix B is an evidence table on prognostic variables and survival of lip carcinoma. Seven studies have been reviewed, and results were entered systematically in this table. The table was used in chapter 2 and chapter 5. Appendix C shows the Kaplan-Meier curve for three groups: patients with cancer of the upper lip, lower lip, or skin of the lip. Appendix D contains a table of characteristics of a selection of articles that have been used in this study. Appendix E explains the strategy for the literature search. 5

6 2 Literature The literature and results of earlier studies are reviewed in this chapter. In paragraph 2.1 the search strategy for studies on lip carcinoma is described, followed by a review of articles in paragraph 2.2. The literature is covered methodically, paragraph 2.2 is divided in sections; incidence, anatomy and histology, risk factors, treatment, and prognosis and survival. 2.1 Research strategy A systematic research was done in Pubmed for articles on lip cancer. Lip cancer-related and prognosis-related search terms were selected. Also, the terms upper lip and lower lip were added to the search, so articles that mentioned the tumour site in title or abstract were included. The complete search strategy is mentioned in appendix E. An evidence table for prognostic factors and survival is given in appendix B. Seven articles are included in this table 3,6,13,15,16,18,19. Four of these make a distinction between upper and lower lip cancer. The other three studies focus on poor prognostic factors for lip cancer in general. While odds ratios and p-values were preferable, we decided not to add these as an inclusion criterion for the evidence table, since studies on prognostic factors and survival of lip cancer are limited. Articles were excluded when study populations contained 50 patients or less. The evidence-table is used and referred to in paragraph Prognosis and Survival. 2.2 Literature review Incidence Lip cancer is a rare malignancy of the head and neck region, incidence rates vary widely across the world. Carcinoma s of the lip (ICD C00.1-C00.9) 20 are a rare malignancy of the head and neck region, accounting for about 10% of all head and neck tumours with an average incidence rate of 0.65 per per year in the Netherlands (225 diagnoses annually) b. The incidence of lip cancer varies widely throughout the world. In some countries the lip is the most prevalent site of oral cancer, for example in the Australian population where over 50% of all oral cavity cancers are located on the lip 7,21. The highest reported incidence rates are 13.5 per c per year for South Australia and 12.7 per per year c for the East of Canada 1,5. Lip cancer is virtually unknown in Asia, with rates of 0.1 per per year in China, Japan and Singapore 1. Interestingly, incidence rates for oral cavity cancer in Asia are among the highest of the world 22,23. A decreasing incidence trend for lip cancer has been observed in most parts of the world 1,7,24, but while incidence rates are generally stable or falling among males worldwide, they are rising in many female populations 1,5,6,10,24. Over 90% of lip cancers present on the lower lip 1-5. The percentage of lower lip cancers has decreased significantly over the last 30 years, while the percentage of upper lip cancer has increased over this period 1,6. b National Cancer Registry Netherlands, period ; ( consulted December 2011) c Average age standardized rates per population (Standard World Population) 6

7 2.2.2 Anatomy and Histology The lips mark the transition zone between skin and oral mucosa, and show distinct histological features. The lips consist of an external pars cutanea (skin), a pars intermedia (the vermilion) and an internal pars mucosa 25,26. The vermilion, or vermilion border, is the surface covering both lips between the skin-vermilion and the vermilion-oral mucosa junction (figure 1). Figure 2 shows the histological features of this transition zone. The pars cutanea (skin of lip) consists of a stratified and keratinized epidermis and, underneath, a dermal layer that contains hair follicles and subaceous glands. The vermilion has a thin, partlykeratinized stratified squamous layer of epithelium. Subaceous glands can be found but hair follicles are usually not present. Rete ridges d are more marked than in the cutaneous zone. Capillaries are abundant and give the vermilion its red color, because of the translucent epitheliocytes. The vermilion doesn t have a dermal layer. The dermis of the pars cutanea transforms to the lamina propria mucosae of the vermilion. The internal pars mucosa consists of moist, non-keratinized epithelium and contains salivary glands. The musculus orbicularis oris forms, together with a layer of connective tissue, the core of the lips 25,26. In this study, the histological features were of importance because they were used to make a distinction between the skin, the vermilion and the oral mucosa (chapter 3). Fig 1. Vermilion, white roll, commissure and skin of the lips. * Skin of lip begins at skin-vermilion junction. Fig 2. Histology of the lip 26 In case of large lesions, difficulties arise with determining the precise site of origin. In literature, the classification of lip cancer is often ambiguous and not uniform across the world, making it difficult to compare the results of studies to each other. However, most studies use the following definition: No other cancer than that arising in the red of the lip should be regarded as genuine lip cancer. This definition is used for this study as well. In 1979 the International Classification of Diseases for Oncology 20 (ICD-O) was introduced, benefiting the comparison between studies. While most of the lip carcinomas will be classified as lip cancer, there is a possibility that some skin cancers will be, unjustly, classified as lip cancer 1,27,28. This potential classification problem arises especially in the case of large lesions, since it may be impossible to determine the exact site of origin of the tumour when it involves both the lip and the surrounding skin. Figure 3 shows a couple of photographs to illustrate this classification problem. When skin tumours are incorrectly classified as lip tumours, it might influence the results of studies, because some distinct differences between the tumour sites have been found in incidence rates, patient and clinical characteristics, and survival rates. The epidemiology of lip cancer supports the proposal that the lip should be considered as a distinct cancer site, rather than being included with other forms of intraoral cancer or skin cancer 1, d Rete ridges are epidermal thickenings that extend downwards between dermal papillae 7

8 Fig 3. Pictures of four patients with SCC of the lip, treated in the Medical Centre in Leeuwarden (MCL). The majority of lip malignancies are squamous cell carcinomas, followed by basal cell carcinomas. The proportion of basal cell carcinoma on the upper lip is higher than for the lower lip. About 90% of the lip malignancies are squamous cell carcinomas (SCCs), with a range of 50% - 96% in earlier studies 3,6,8-10,16. SCC is most frequent on both the upper and lower lip, but in some studies a larger proportion of basal cell carcinomas (BCCs) is found on the upper lip 3,10,16. The BCCs mainly occur in female patients 32. Zitsch et al. found a statistically significant difference in site predilection between different tumour types in their series of patients. In the study by Geraud et al. an association between histology and gender of the patients was detected neither for the upper nor the lower lip Risk factors Men, fair-skinned individuals and people over 50 years of age have a higher risk of developing lip cancer. Risk factors for the development of lip cancers are in particular cumulative life-time exposure to UV radiation and smoking. Age The risk of lip cancer increases with age, presenting predominantly in the 6 th and 7 th decade 1-3,6-12. It can be the result of accumulation of molecular changes as a result of variables such as prolonged sun exposure and smoking, or as part of the biological ageing process 1,7,33. Sun exposure A major risk factor for lip carcinoma has been shown to be DNA mutations caused by UV radiation. This conclusion stems from the observations that there is an increased incidence of the disease in people with fair complexion 34, those with outdoor occupations 35,36 and those who live in rural areas 9,36. Sunlight affects the lower lip much more than the upper lip, catching more sunlight due to the prominent location. UVB is assumed to be the major carcinogen. The fact that males are 8

9 affected more than females is probably due to increased frequency of outdoor occupation 3,6,36 and, potentially, to the protective effects of lipstick for women 37. Fair complexion The remarkable difference in incidence between black and white people is due to the photo protective effects of high levels of melanin (skin pigment) in black people 12. People with fair complexion show more skin reaction to sun exposure and are therefore more prone to developing lip cancer 34. Tobacco Until now, no intimate relationship between lip cancer and smoking has been demonstrated. However, several case-control studies found that joint exposure to outdoor work (and thus sunlight) and smoking significantly increased the risk of lip cancer, demonstrating a synergistic effect 35,38. Patients with cancer of the lip also have an increased risk of second primary cancers in the respiratory track, suggesting smoking as a common risk factor 39,40. Viruses A link between SCC of the lip and potentially oncogenic viruses has been suggested, especially for herpes simplex virus type I (HSV) and human papilloma virus (HPV) 35,41. However, a clear association could not be found 34,38. At the moment, it seems unlikely that viruses are an independent risk factor. They supposedly need additional promotors, e.g. smoking and ultraviolet light, to induce dysplasia or carcinoma of the lip. Immune status The immune system is known to be of importance for protecting the host against cancers. There is little reported evidence concerning an individual s immune status and lip cancer, but some studies have shown an increased risk of SCC of the lip in renal transplant patients 42,43. The higher risk of lip and skin cancer is attributed to the immunosuppressive therapy given to these patients. Outdoor occupation, rural residency, socioeconomic status, familial predisposition These four factors have been cited as risk factors for lip cancer 7,9, The relative significance of these associations are confounded by the fact that each of these variables are correlated with one another, and with additional risk factors, especially to prolonged exposure to sunlight and heavy smoking. Other factors Other factors such as diet, nutrition, and poor oral health may be linked to the pathogenesis of lip cancer, but since no extensive controlled studies have been performed, their role in the etiology of lip cancer remains controversial. Alcohol use alone does not promote the development of lip cancer, but it does work synergistically with smoking and exposure to UV radiation 34. The definitive pathogenic pathway remains unclear, but cancer of the lip is evidently a result of multiple factors acting either independently or synergistically. The etiologic factors also explain the large variation in incidence rates between countries. The high rates in Australia and Canada, for example, are generally explained by a combination of fair skin and high exposure to solar radiation 1,5. The fall in incidence rates among males is probably due to the public awareness of the dangers of sunlight exposure, a decrease in smoking habits in industrialized countries, a decline of outdoor employment and migration from rural to urban areas during the last decades. The increased incidence rates among women may be due to sun bathing, which has been more popular in the last decades, especially among women, the post war increase in smoking, and additional, yet unknown factors 9. The majority of the lesions originate on the lower lip, mainly as a result of sunlight. Lower lip cancers are more common in men, upper lip cancers are more common in women. The lower lip, that is more prominent than the upper lip and therefore receives more direct exposure to solar radiation, is far more frequently affected. Over 90% of lip carcinoma originates on the lower lip 2-5. This seems to be due to the fact that the lower lip is more prominent than the upper lip and is therefore more exposed to UV radiation. Lower lip cancers are more common among men, while upper lip cancers are more common among women 3,6,10,44. An explanation for this difference has not been found yet. It may be due to a larger proportion of men having outdoor occupations, and therefore a higher cumulative exposure to UV radiation. A protective effect of women s lipstick has also been suggested 37. Not many other differences in characteristics between the tumour sites have been reported. Only a difference in size was reported by Zitsch et al. They report 2.5% of cancers of the lower lip to be 3 cm or larger, whereas 11% of the upper lip cancers were 3 cm or larger, this was a statistically significant difference 3, but this would not explain the higher incidence of upper lip cancer in women. 9

10 2.2.4 Treatment Lip tumours are mainly treated by surgical excision and radiotherapy. The common initial symptoms of lip cancer are ulceration, encrustation and soreness. Surgery and radiotherapy are currently the treatments of choice for SCC of the lip 45. The choice of treatment is based on TNM staging. In general, surgical excision is used when possible, in combination with frozen section analysis of the surgical margins. When frozen-section analysis is performed, a 3-5 mm margin with excision seems to be appropriate The excisional defects can be closed primarily or reconstruction can be performed using local skin flaps or the radial forearm flap. For large lesions, radiotherapy can be chosen. Radiotherapy is thought to offer better cosmetic and functional results than extensive surgery 48. If there are regional lymph node metastases, a modified radical neck dissection is performed 14,45. Chemotherapy is rarely used. Level of compliance to the regional guideline for treatment of lip cancer 45 is low 49. However, both surgery and radiotherapy produce equally high cure rates 3,48,50. Therefore, the choice for either form of treatment modality may be determined by other, non- tumour related, factors such as age and condition of the patient, time and costs, cosmetic and functional results, and preference of the patient and clinician Prognosis and Survival Cancer of the lip carries a good prognosis with a relative 5-year survival approaching 92% in the Netherlands. Relative survival rates that have been reported range from 73% to 100% 8,9,12,13,23,24. High survival rates of lip tumours compared with that reported of other oral cavity SCCs may be the result of easier recognition by patients and healthcare providers, leading to early diagnosis, and the relative infrequency of associated regional metastases 3,6,913. This assumption can be ascertained by tumour stage, being predominantly stage I or II at presentation 3,11,13,49. Because of the prominent location, lip carcinomas are well accessible for treatment. Long-term survival rates diminish greatly when local or regional recurrence is witnessed. Other negative prognostic factors are large tumour size, high tumour stage and inadequate surgical margins. Local recurrence In several studies, local recurrence was found to be an adverse prognostic factor 3,13,18. The recurrence rate of lip cancers is quite low, we found a range of 4.9% % 3,13,16,18,48. The 39.7% recurrence rate however, comes from a study with a high percentage of stage III and IV tumours 16. Most studies have found a recurrence rate around 7%. De Visscher et al. found local recurrence to be associated with a tumour size > 16 mm and involved surgical margins 13. Lymph node involvement The development of lymph node metastases is infrequent in lip cancer patients. The metastatic rates range from 7.1% to 13.7% 3,18,51. Patients with cervical lymph node metastases have a significant lower survival rate, we found 5-year overall survival rates of 37,2% 18, 52% 3, and 88% 15. Frierson et al. observed that carcinomas that had metastasized were more likely to be stage III or IV tumours, had a thickness of at least 6 mm, were poorly differentiated and more often showed perineural invasion than the non-metastasizing tumours 19. Each of these differences were significantly different. De Visscher et al. agreed, finding regional metastases is strongly associated with increasing tumour thickness, an infiltrative invasion pattern and perineural invasion 13. Califano et al. found that upper lip carcinoma s metastasized more often than lower lip or commissural tumours 15. A meta-analysis comparing SCCs of the lip, facial skin and ears found a higher metastatic rate for lip SCC (13.7%) than for the other groups (facial skin: 5.2%, ears: 8.8%) 18. However, when survival was calculated for metastatic SCCs in these groups, 5-year survival rates were more favorable for lip metastasis (lip: 36.2%; facial skin: 26.8%, ear: 27.8%) 18. Tumour size Large tumour size is a poor prognostic factor 13,16,19. Patients with tumours >3 cm had a significantly lower survival than patients with tumours < 3cm (Resp. 64% vs 92%) 3. Luna-Ortiz et al. found, of all variables studied, only tumoural size to be statistically significant as a prognostic factor

11 Tumour stage and differention grade The TNM classification 52 is supposed to say something about prognosis. TNM stage III or IV reduces the survival significantly in comparison with tumour stage I or II 19. Poorly differentiated tumours show lower survival rates than tumours with a good or moderate differentiation grade 3,19. For example, poorly differentiated tumours had a survival of 71% and good/moderately differentiated tumours had a survival of 90% in the study by Zitsch et al. 3. Surgical margins Local recurrence and regional metastasis are associated with surgical margins containing SCC 3,13. Having involved surgical margins, and thus leaving cancer cells behind, reduced the 5-year survival of patients from 94% to 76% in the study by Zitsch et al. 3. This is why it is important to check the surgical margins histologically during the excisional procedure, so re-excision can be performed when necessary. There seem to be differences in pathogenesis between upper and lower lip cancer. Some studies report a significant difference in survival rates between upper and lower lip SCC. However, worse prognosis for upper lip cancer is not inconclusively proved. Carcinoma of the upper lip and commissure are said to have a 5-year survival rate between 10-20% lower than a comparable lesion of the lower lip, but there still remains discussion on this subject. Zitsch et al. and Califano et al. state that upper lip cancer is an adverse prognostic factor 3,15. They hypothesize upper lip cancer to belong to a more aggressive class of disease, because of the lower survival rate. Zitsch et al. found a 5-year overall survival rate of 77% for the upper lip and 90% for the lower lip 3. In the article by Califano et al. the 5-year overall survival rate for the upper lip was 91% and for the lower lip 98% 15. The tumours of the upper lip were significantly larger 3 and metastasized more often 15 than lower lip cancers. No articles have mentioned a possible cause or difference in pathogenesis between upper lip and lower lip cancer. The medical world has adapted the conclusion that cancer of the upper lip has a worse prognosis, even though conflicting results have been published. For example, Czerninski et al. claim that cancer of the upper lip revealed a higher cumulative survival rate than the lower lip (81.4% and 77%, resp.) 6. Yet another study couldn t find a difference in survival at all between the two tumour sites 16. In our opinion, worse prognosis for upper lip cancer is not inconclusively proved. There have been few large population studies on the subject and articles published conflicting results. 11

12 3 Materials and Methods In this chapter the materials and methods used for this research are described. In paragraph 3.1 and 3.2 the procedure of cancer registration and the criteria for patient inclusion are outlined. For this study, an existing database was used. A database check for tumour site was performed (upper lip, lower lip or skin of lip), the methods for this check are defined in paragraph 3.3. The variables are specified in paragraph 3.4. In the last section, paragraph 3.5, the statistical analyses are described. 3.1 Cancer registration The Netherlands Cancer Registry (NCR) records new malignant cancer Table 1. Anatomical sites of the cases (basal cell carcinoma of the skin excluded). Signaling sources for lip according to the ICD-O the registry are the automated pathology databank (cytology and classification, histology reports), medical registrations in hospitals, hematologic C00.0 External upper lip departments and hospital discharge data banks. After signaling, trained C00.1 External lower lip registry clerks collect a standard set of data using the medical files of C00.2 External lip, NOS patients. C00.3 Mucosa of the upper lip Data are collected in agreement with international classifications, using C00.4 Mucosa of the lower lip standard definitions in order to allow for comparisons between different C00.5 Mucosa of the lip, NOS registries, countries and studies. The tumour sites for lip cancer are C00.6 Commissure of the lip registered according to the International Classification of Diseases for Oncology 20 C00.9 Lip, NOS (ICD-O codes no C , C44.0). These codes are C44.0 Skin of lip explained in table 1. The tumour stage is registered following the TNMclassification e,52 (table 2). Small adjustments were made in the 6 th edition of the TNM-classification f. Carcinoma are graded into well, moderately or poorly differentiated tumours. Lymph node metastases and distant metastases are only registered when existing at the time of presentation or at the time of treatment of the primary lip tumour. Metastases presenting later in follow-up are not registered in the NCR database. Every year, the NCR updates the cancer database with information on the vital status (alive, dead, or emigrated) of patients by linking their data to the municipal records. The records are checked for vital status and date of death or emigration is collected. This study was approved by the NCR. 3.2 Patient population and inclusion criteria We analyzed data on lip cancers registered between January 1989 and December 2009 in the North- Eastern region of the Netherlands Cancer Registry (NNCR). The NNCR covers four provinces of the Netherlands and includes 23 hospitals. Only squamous cell carcinoma s (SCCs) were included in this study (N = 1993), because over 95% of all lip cancers in our database were SCCs and other histologies have been shown to be related to survival. The NCR only registers primary tumours, recurrences are not included in this study. Second primary tumours that have the same histology and originate from the same site (upper lip, lower lip, or commissure) as the previous tumour, are regarded recurrences, but patients who developed a second primary tumour at a different site of the lip than the first tumour were included. The 1993 SCCs occurred in 1876 patients, in 36 patients more than one primary SCC developed. Only SCCs arising on the vermilion of the upper and lower lip and SCCs of the skin of the lip were included. e TNM-classification used for registration during periods of time: th edition 53 ; th edition revised 54 ; th edition 55 ; th edition 52. f Adjustment in 6th edition: a distinction was made between T4a; tumour invades through cortical bone, inferior alveolar nerve, floor of mouth, or skin (chin or nose), and T4b; tumour invades masticator space, pterygoid plates, skull base, or encases internal carotid artery. 12

13 3.3 Data collection and Database check All SCCs of the lip were retrieved from the NNCR database (N = 1908). Figure 4 shows a flow chart for data collection. Patients identity in the NNCR database was protected by patient-specific numbers. These numbers correspond to hospital ID-numbers, only retraceable for selected employees of the NNCR. As mentioned before in chapter 2, determining the precise site of origin comes with difficulties. A check was performed because it is important to know that localization is registered correctly since tumour site is the most defining variable in this study. We checked the data of the patients diagnosed with lip cancer in the Medical Centre in Leeuwarden (MCL, N = 341) and the University Medical Centre in Groningen (UMCG, N = 244). The medical files of the patients were studied and tumour site was re-registered according to the ICD-O 20 (table 1). The differentiation between tumour sites was made with information found in the pathology reports (PA reports), patient letters, and photographs or pictures when available. PA reports are available in case a biopsy and/or surgical excision is performed. In paragraph the histology of the lips was discussed. We used this information on histology to distinguish the skin from the vermilion border. PA reports describe the macroscopic and microscopic features of the entire surgical excision and usually mention which section of the excision is infiltrated with atypical cells (e.g. dermal layer or mucosa). When a distinction could not be made with the information from the medical files, the cases were marked as unknown. Regular registration by NCR employees is also based on information in the medical files and the PA reports but less detail is paid to the microscopic features in the PA reports. The investigation was done by one person. Data were entered in a Stata database and compared to the data of the two hospitals in the original database. With the results of this comparison, we could determine if registration of sub site has been performed correctly and if the original database gives a good reflection of reality. Table 2. TNM-classification for cancer of the Lip and Oral cavity according to the UICC, T Primary Tumour Tis Carcinoma in situ T1 Tumour 2 cm or less in greatest dimension T2 Tumour more than 2 cm but not more than 4 cm in greatest dimension T3 Tumour more than 4 cm in greatest dimension T4 Tumour invading adjacent structures (e.g.cortical bone, inferior alveolar nerve, floor of mouth, skin (chin or nose), masticator space) N Regional lymph nodes N0 No evidence for regional lymph node metastasis N1 Metastasis in a single ipsilateral lymph node, 3 cm or less in greatest dimension N2 Metastasis in single ipsilateral lymph node, more than 3 cm but not more than 6 cm in greatest dimension; or in multiple ipsilateral lymph nodes, none more than 6 cm in greatest dimension; or in bilateral or contralateral lymph nodes, none more than 6 cm in dimension N3 Metastasis in a lymph node, more than 6 cm in greatest dimension M Distant metastasis M0 No distant metastasis M1 Distant metastasis 3.4 Variables and Endpoints Variables regarding patient characteristics, tumour characteristics, treatment and follow-up for vital status were selected from the NNCR database. All tumours in this study are of the same histological type: squamous cell carcinoma. Patient characteristics were gender and age at diagnosis. Tumour characteristics included year of incidence, localization of the tumour, histological grade, stage of the disease according to TNM classification 52 (table 2), and presence of lymph node metastases and distant metastases at presentation. Treatment characteristics included surgical margin status (involved/clear) and type of treatment: surgery and/or radiotherapy. Vital status and the date of assessment are used, in combination with date of incidence, to calculate the follow-up period for the patient. Consequently, the follow-up period is defined as the number of days between the incidence date and the date of linkage to municipal records or date of death or emigration. The follow-up time was used to calculate survival. Endpoint in this study was 5-year overall survival (OS) for patients. 13

14 3.5 Statistical Analysis To assess the implications of our database check, statistical analysis was performed for three groups: upper lip vermilion cancer (UL), lower lip vermilion cancer (LL) and cancer of the skin surrounding the lips (SL). To determine prognostic factors and differences in survival, only the UL and LL group were used. A flow chart for inclusion and statistical analysis is given in figure 4. The differences in clinical features between nominal or ordinal variables and localization were studied using the Chi-square or Fisher s exact test. Differences in age between the tumour site groups were tested using the Kruskal-Wallis test, because more than two groups were involved and age was not normally distributed. Overall survival was calculated using the Kaplan-Meier, followed by a log rank test to test whether the Kaplan-Meier curves differed from each other. The Kaplan-Meier analysis was executed for the UL and LL group, and for the two tumour sites stratified by gender. Resulting 5-year OS rates are presented. In survival analysis only one record per patient can be included. When patients in the database had multiple primary tumours, we chose to include 1) the first tumour or 2) the tumour with the highest stage when simultaneous primaries were present. To determine the potential prognostic value of clinical and histologic parameters on survival, Cox regression analyses were performed. We included all the variables separately in a univariate Cox regression test. The variables that resulted in a significant outcome, and factors that showed significance in earlier studies (e.g. tumour site) were subsequently included in a multivariate regression test. All tests were assessed at a confidence interval of 95%. The values were considered to be significant at p The statistical analyses were conducted using the Stata program g. Fig 4. Flow chart for inclusion and statistical analysis for tumours and patients; C-numbers correspond to the ICD-O classification 20 in table 1. g StataCorp Stata Statistical Software: Release 12. College Station, TX: StataCorp LP. 14

15 4 Results In this chapter the results of the study are described. The chapter is divided in two main parts. The first paragraph focuses on the results of the database check. In this part we decide if the data in the original NNCR database are representing the distribution of lip cancer accurately. We also look at the potential effect of the database check on our endpoint (survival). In paragraphs 4.2 to 4.4, results in relation to our research questions are given. 4.1 Database check for tumour site We performed the check for tumour site in two hospitals using records of 585 patients (28.3% of total number of records in the database, N=2069). Table 3 describes the distribution of lip cancers to tumour site before and after correction. Unfortunately 71 tumours (12.1%) could not be re-registered ( X in table 3), because PA reports weren t conclusive or medical files were no longer available. Since 80% of the missing data occurred in incidence years , the distribution of tumour site is also mentioned for the period After the check, the tumour sites of 106 out of 514 patients (20.6%), not including X, had been corrected to a different localization. Comparing the two databases, results are rather similar, the most dominant findings being a decline of SL from 29.2% to 11.1% and an increase of LL from 58.8% to 67.7%. The proportion of UL tumours declined with 1.2%. Of the 106 corrected tumour sites, 77 tumours (72.6%) originally came from SL. These were mainly corrected to the LL (92.2%) and UL (5.2%). Six of the corrected tumours (5.7%) came from the UL localization and 7 (6.6%) of the LL tumours were corrected. The UL:LL ratio declined from 1:12 to 1:19, so the share of UL cancer was weakened. Table 3. Localization of lip tumours in patients treated in the Medical Centre of Leeuwarden (MCL) and the University Medical Centre Groningen (UMCG) before and after the re-registration. NNCR database n (%) Re-registration n (%) Tumour site Incidence years Incidence years Incidence years Incidence years External upper lip (UL)* 28 (4.8%) 19 (5.0%) 21(3.6%) 14 (3.7%) 00.1 External lower lip (LL)* 344 (58.8%) 247 (64.3) 397 (67.7%) 274 (71.4%) 00.3 Mucosa of the upper lip 8 (1.4%) 7 (1.8) 9 (1.5%) 8 (2.1%) 00.4 Mucosa of the lower lip 16 (2.7%) 10 (2.6) 8 (1.4%) 6 (1.6%) 00.5 Mucosa of the lip, NOS 2 (0.3%) 2 (0.5) 1 (0.2%) 1 (0.3%) 00.6 Commissure of the lip 8 (1.4%) 8 (2.1) 13 (2.2%) 13 (3.4%) 00.9 Lip, NOS 8 (1.4%) 8 (2.1) 0 (0.0%) 0 (0.0%) 44.0 Skin of lip (SL) 171 (29.2%) 83 (21.6) 65 (11.1%) 50 (13.0%) X** 0 (0.0%) 0 (0.0%) 71 (12.1%) 15 (3.9%) Total NOS: not otherwise specified; * External is defined as vermilion border of the lip; **X: could not be assessed or medical files no longer available 15

16 4.2 Characteristics of original data In this study a total of 1908 SCCs was included. Of these, 1126 SCCs occurred on the lip. The most frequently involved site was the LL, consisting of 998 tumours (52.3%), 128 originated on the UL (6.7%). A total of 782 SCCs (41.0%) occurred on the SL, these are not included for survival analysis but patient characteristics and clinical features of the three groups are shown in table 4. Table 4. Patient characteristics and clinical features of lip cancer cases (all SCC) in the Northern Netherlands (N = 1908); comparison between cancer of the upper and lower lip, and skin of the lip Upper Lip Lower Lip Skin of Lip Total n (%) n (%) n (%) n (%) Gender* Male 52 (41.3%) 749 (76.5%) 547 (71.0%) 1348 (71.9%) Female 74 (58.7%) 230 (23.5%) 224 (29.0%) 528 (28.1%) Age at presentation* Median Percentile 5/95 45/88 47/87 46/88 46/87 Tumour grade Well diff 37 (28.9%) 356 (35.7%) 256 (32.8%) 649 (34.0%) Moderately/poorly 34 (26.6) 265 (26.6%) 138 (17.6%) 437 (22.9%) Unknown 57 (44.5%) 377 (37.7%) 388 (49.6%) 822 (43.1%) Lymph node metastasis** N0 128 (100%) 986 (98.8%) 778 (99.5%) 1892 (99.2%) N (1.2%) 4 (0.5%) 16 (0.8%) Stage Low (I+II) 125 (97.7%) 981(98.3%) 690 (88.2%) 1796 (94.1%) High (III+IV) 3 (2.3%) 17 (1.7%) 7 (0.9%) 27 (1.4%) Unknown (10.9%) 85 (4.5%) Treatment*** Surgery 125 (97.7%) 888 (89.0%) 673 (86.1%) 1686 (88.4%) Radiotherapy 4 (3.1%) 118 (11.8%) 106 (13.6%) 228 (12.0%) No treatment 1 (0.8%) 19 (1.9%) 26 (3.3%) 46 (2.4%) Surgical margins Clear 60 (46.9%) 270 (27.1%) 114 (14.6%) 444 (23.3%) Tumour residue 2 (1.6%) 20 (2.0%) 12 (1.5%) 34 (1.8%) Unknown 63 (49.2%) 597 (59.8%) 547 (69.9%) 1207 (63.2%) No surgery 3 (2.3%) 111 (11.1%) 109 (14.0%) 223 (11.7%) Total * Only tumours with the highest stage/first primary tumours included (N = 1876); ** at presentation, data on lymph node metastasis later in follow-up were not available; ***Some patients received both surgery and radiotherapy, this explains the total adding up >100% Sex and age A male-to-female predominance was found, with 1348 (71.9%) male and 528 (28.1%) female patients. The highest male:female ratio was found in the LL group, about 3.3:1. A slightly lower ratio of 2.4:1, was found for the SL. For the UL group, an opposing predominance was found; 74 (58.7%) of UL cases were found in women, resulting in a 1:1.4 ratio. The differences in sex predominance between the groups were statistically significant (P < 0.001). 16

17 The median patient age was 70 years, with a 5-95 percentile ranging from years. Over 90% of lip cancer patients were diagnosed above 52 years of age. Age was not normally distributed. No significant age differences were found between the tumour sites. Also, no significant age differences were found between males and females Tumour grade, metastasis at presentation and tumour stage In 649 SCC s (34%) the tumour grade distribution for patients with lip cancer was well-differentiated and 437 SCC s (22.9%) had a moderate or poor differentiation. Unfortunately, the tumour grades of 822 (43.1%) tumours were unknown. Unknown tumour grades were equally found in all incidence years. Of the tumours located on the skin of the lip, 138 (17.6%) were moderately/poorly differentiated, this proportion was significantly lower than for the upper lip and lower lip (both 26.6%, P < 0.001). However, the proportion unknown was larger in this group. When excluding the unknown grades, a significance remained (P = 0.021), implying a better differentiation grade for tumours of the skin of the lip. Only 16 patients (0.8%) were diagnosed with lymph node metastases at the incidence or treatment date of the primary lip malignancies; 12 lymph node metastases (1.2%) were found in the LL group and 4 (0.5%) in the SL group. There were no positive lymph nodes present at time of presentation in the UL group. No distant metastasis were present at the time of diagnosis of the primary lip SCC. Tumour stage at presentation was generally low, 1796 of 1908 (94.1%) was a stage I or II SCC. Only 27 cases (1.4%) were high stage tumours (stage III or IV). There were no statistically significant differences in tumour stage between groups Treatment Of all patients, the vast majority of lip cancer cases (88.4%) is treated by surgical excision of the SCC. There was a statistically significant difference for choice of treatment between tumour sites (P = 0.001). A higher percentage of patients in the UL group (97.7%) was treated with surgery in comparison to the LL group (89.0%) (P=0.002). Also, radiotherapy is less frequently used for UL SCCs (3.1%) than for LL SCCs (11.8%) (P=0.003). 4.3 Clinical outcome: Overall survival The overall survival (OS) curves of both groups (UL and LL), are given in figure 4. Table 5 presents the 5-year OS rates for tumour site and gender. The 5-year OS rate was the same for both tumour sites: 74%. Because the proportion of females in the UL group was significantly higher, a survival analysis for males and females separately was performed as well. The Kaplan Meier curves for male and female patients are given in figures 5 and 6. The female UL patients have a higher 5-year OS (81%) than female LL patients (76%). For male patients it was the opposite, male patients with UL cancer had a lower 5-year OS rate than LL cancer patients, 65% and 74% respectively. However, the survival curves for UL versus LL cancer in male and female patients were not significantly different. 17

18 Table 5. Crude 5-year survival rates for patients with UL or LL cancer; comparison between tumour sites and gender. Number of 5-year 95% CI patients survival Male % Upper lip 52 65% Lower lip % Female % Upper lip 74 81% Lower lip % Total % Upper lip % Lower lip % Fig 5. Overall survival curves (Kaplan-Meier) by tumour site among 801 male patients with primary SCC of the lip. Fig 6. Overall survival curves (Kaplan-Meier) by tumour site among 304 female patients with primary SCC of the lip. 4.4 Prognostic factors Prognostic factors were evaluated in 1126 lip SCCs. In this analysis, three factors were an independent prognostic factor for death: male patients, increasing age, and high tumour stage. Tumour site was found not to be a prognostic factor. Hazard Ratio s are presented in table 6. For hazard for death, the HR for men compared to women was 1.46 (P = 0.001; CI ). For age, the HR increased with 2.07 (P < 0.001; CI ) every 10 years. A tumour stage III or IV had a HR of 2.60 (P < 0.001; CI ) in comparison to stage I or II tumours. The choice for either surgical treatment or radiotherapy showed no significant HR in multivariate analysis. In the Cox model the HR for death for LL patients compared to UL patients was 1.09 (P = 0.56 ; CI ). We also calculated the HR for involved surgical margins versus clear margins (23 vs 324 tumours). This is not included in the table as it overlaps with surgery and a lot of data are missing. Univariate and bivariate analysis showed involved surgical margins to be a poor prognostic factor (bivariate HR 2.36; P = 0.017; CI ). This significance disappeared when multivariate analysis, excluding treatment, was performed. 18