指導教授：賴美淑臺灣大學：流行病學與預防醫學研究所廖立人Liao, Li-JenLi-JenLiao2014-11-272018-06-292014-11-272018-06-292014http://ntur.lib.ntu.edu.tw//handle/246246/262364目的: 對於臨床上頸部轉移零期之頭頸癌病患，頸部淋巴理想的處置目前學界並沒有共識。過去大部份的耳鼻喉頭頸外科醫師多採取預防性頸淋巴結擴清手術處置，如此可能因而導致過度的治療。近年來有越來越多新的方法可以用來診斷臨床上頸部轉移零期之頭頸癌，包括電腦斷層(CT)、磁核共振(MRI)、超音波(US)、超音波導引細針穿刺(US-FNA)、正子攝影(PET)以及前哨淋巴結切片(SNB)檢查，本論文的目地在於詳細評估這些可以用於頸淋巴節轉移的工具。 方法: 本論文第一部份包括系統性的文獻回顧、統合分析，評估單一診斷工具的表現; 第二部份則透過假設性評估分析，比較結合多個診斷工具不同策略之陰性預測值; 第三部份聚焦於結合多個診斷工具之不同策略的經濟評估; 最後部份提出一份關於台灣結合前哨淋巴結切片檢查手術，於臨床上頸部轉移零期口腔癌病人的可行性報告。 結果: 在系統性的文獻回顧共收集了 73 篇文獻，CT 有10篇，MRI 有7篇，PET 有12篇，US 有 9篇，US-FNA 有 5篇，而SNB　則有 55篇。在統合分析顯示 US-FNA的敏感度為56% (95% 信賴區間 45%~67%)，SNB的敏感度為85% (82~87%)。CT, MRI, US 及 PET的敏感度分別為47% (38.2%~56.0%), 56.6% (39.8~71.9%), 63.3% (54.0~71.7%), 及48.3% (30.9~66.1%)，特異度分別為88.9% (82.0%~93.3%), 82.5% (39.8~71.9%), 79.1% (73.4~83.8%) 及86.2 % (76.9~92.1%)。CT, MRI, PET, US, US-FNA 及SNB　的 AUC 分別為0.81(0.56~1), 0.79(0.66~0.93), 0.83(0.69~0.96), 0.81(0.74~0.87), 0.97(0.85~1) 及0.98(0.96~0.99。假設性評估分析發現，採取 CT或 MRI加上SNB的策略，即使檢查前的頸淋巴轉移率高達60%，陰性預測值都可以高於85%。在決策分析方面，如不考慮成本，若檢查前的頸淋巴轉移率高於10%，CT或MRI 加上SNB的策略會有較高的效益值。若考量成本效益，CT-FNA 會有最低的成本花費，但是CT加上SNB仍有最佳的成本效益。在前哨淋巴結切片可行性研究，從2013年六月份至2014年三月份，共有11名口腔癌病患接受結合前哨淋巴結切片檢查的手術，所有的病人都可以發現前哨淋巴結（100%），所有的前哨淋巴結都位於頸部I~III　區域，四名病患前哨淋巴結術中發現有轉移，有一名病患追蹤一年頸部復發，因此前哨淋巴結切片檢查手術的敏感度估計為80%。 結論: 前哨淋巴結切片檢查是一項可靠的診斷工具; 結合前哨淋巴結切片檢查的手術於臨床上頸部轉移零期頭頸癌病人的處置未來可能有進一步發展的希望。Background: The optimum management over the neck of clinical negative (cN0) head and neck cancer has been a debate for a long time. In the past, most head and neck surgeons did prophylactic neck dissection and potentially leaded to over-treatment. Since more and more modern diagnostic technologies are developing, the possibility of conservative treatment is increasing. The aim of this dissertation is to comprehensively assess multiple diagnostic tests, including traditional CT, MRI image, high resolution ultrasound (US), ultrasound guided fine needle aspiration(US-FNA), positron emission tomography (PET) and sentinel node biopsy (SNB). Materials and Methods: While the first part focuses on systematic review of individual diagnostic test, meta-analysis is also done to assess the diagnostic performance of each test. The second part uses hypothetical estimation of serial tests, which compares multiple diagnostic strategies, to evaluate the negative predictive value, over- and under treatment of each strategy. The third part uses decision modeling, which is under view point of patient and health care system, to assess the efficiency of the proposed diagnostic strategies. At the end, a feasibility study of sentinel node navigation surgery (SNNS) in Taiwan is presented. Results: In systematic review, total 73 studies were recruited. Ten studies fulfilled all inclusion criteria for CT, 7 studies for MRI, 12 studies for PET, 9 studies for US, 5 studies for US-FNA, 55 studies for SNB. In meta-analysis, the pooled estimates for sensitivity are 56% (95% confidence interval [CI], 45%~67%) and 85% (82~87%) for US-FNA and SNB respectively. The pooled estimates for sensitivity were 47% (95% confidence interval [CI], 38.2%~56.0%), 56.6% (39.8~71.9%), 63.3% (54.0~71.7%), 48.3% and (30.9~66.1%) for CT, MRI, US and PET respectively. The pooled estimates for specificity were 88.9% (82.0%~93.3%), 82.5% (39.8~71.9%), 79.1% (73.4~83.8%) and 86.2 % (76.9~92.1%) for CT, MRI, US and PET respectively. The AUC are 0.81(0.56~1), 0.79(0.66~0.93), 0.83(0.69~0.96), 0.81(0.74~0.87), 0.97(0.85~1) and 0.98(0.96~0.99) for CT, MRI, PET, US, US-FNA and SNB respectively. For hypothetical estimation, the NPV of CT/MRI then SNB strategies will still be higher than 85%, even the pre-test occult rate up to 60%. In decision analysis, without considering the cost, combined CT-SNB or MRI-SNB will have a higher expected utility if the pre-test occult metastasis rate is greater than10%. In cost-effectiveness analysis, the strategy of CT followed by US-FNA had the lowest price. The strategy of CT followed by SNB would be the most cost-effectiveness strategies. In feasibility study, between June 2013 and March 2014, eleven patients were recruited. At least one sentinel lymph node was identified in all patients (100%). All sentinel nodes were located at level I~ level III. The sensitivity of SNNS is 80% (4/5). Conclusions: SNB is reliable in evaluation of cN0 neck. SNNS could be a promising diagnostic and management strategy for cN0 HN cancer patients.Acknowledgements………………………………………………………………..I 中文摘要.......................................................................................………………..II English Abstract…………………………………………………………….…….IV Abbreviations………………………………………………………………………1 Chapter 1 Background…………………………………………………………. …2 1.1 The importance of head and neck cancers and clinically N0 neck………….…2 1.2 Nature history of clinical N0 neck with potential occult neck metastasis……..4 1.3 Current two main management strategies of cN0 neck………………………. 6 1.3.1 Elective neck dissection (END): mostly over-treated…………….................6 1.3.2Watchful waiting (WW) policy: some under-treatment……………………...9 1.4 Overview of modern diagnostic tools in cN0 neck with/without occult neck nodal metastasis staging…………………………………………………………………..9 1.4.1 Computed Tomography (CT) and Magnetic Resonance Imaging (MRI)……11 1.4.2 Ultrasonography (US) and US-guided fine needle aspiration (US-FNA)……13 1.4.3 Positron emission tomography (PET)………………………………………..15 1.4.4 Sentinel Node Biopsy (SNB)…………………………………………………16 1.4.5 Molecular Diagnosis………………………………………………….............17 1.5 Research gaps…………………………………………………………………..19 1.6 Study purposes……………………………………………………………….....22 Chapter 2 Materials and Methods…………………………………………………..24 2.1 Systematic Review, Meta-analysis for the Diagnostic Accuracy of Individual Diagnostic Tools in Staging cN0 HNSCC with and without Occult Neck Nodal Metastasis………………………………………………………………….…………24 2.1.1 Population………………………………………………………….…………..25 2.1.2 Index tests…………………………………………………………….………..25 2.1.3 Inclusion criteria……………………………………………………..………...25 2.1.4 Exclusion criteria………………………………………………….…….……..26 2.1.5 Study design……………………………………………………………..……..26 2.1.6 Pubmed database searching……………………………………………...……..26 2.1.7 Quality assessment ……………………………………………………..………27 2.1.8 Meta-analysis ……………………………………………………………….….28 2.1.8.1 Univariate meta-analysis: US-FNA and SNB………………………….…….29 2.1.8.2 The SROC model for all modalities……………………………………….…29 2.1.8.3 Bivariate models: CT, MRI, US & PET………………………………….…..30 2.1.8.4 Validation of results…………………………………………………….…….30 2.2 Comparison of Negative Predictive Value among Different Combined Diagnostic Tests under Hypothetic Estimation…………………………………………………..31 2.2.1 Sequential test model…………………………………………………………..32 2.2.2 Diagnostic performance (sensitivity & specificity) of individual test…….…..32 2.2.3 Conditional probability of sequential multiple tests…………………………..33 2.2.4 Based case simulation…………………………………………………………34 2.2.5 Sensitivity analysis……………………………………………………………..34 2.3 Decision Modeling of Different Diagnostic Strategies in Managing cN0 Neck with and without Occult Lymph Node Metastasis………………………………………...36 2.3.1 Decision Analysis for Expected Utility Based on Weiss’s Model 38 2.3.1.1Decision tree………………………………………………………………….38 2.3.1.2 Parameters and outcome evaluation…………………………………………39 2.3.2 Cost-effectiveness analysis: view point from health care system 40 2.3.2.1 Decision model structure……………………………………………………..40 2.3.2.2 Patients Population…………………………………………………………...41 2.3.2.3 Probability……………………………………………………………………42 2.3.2.4 Utility………………………………………………………………………...42 2.3.2.5 Costs………………………………………………………………………….42 2.3.2.6 Outcome measurement……………………………………………………….43 2.4 Sentinel Node Navigation Surgery in Oral Squamous Cell Carcinoma: a Feasibility Study in Taiwan……………………………………………………….....45 2.4.1 Patient Selection…………………………………………………………….....47 2.4.2 Sentinel node Navigation surgery- Preoperative Dynamic Lymphoscintigraphy Technique……………………………………………………………………………48 2.4.3 Intra-operative SNB Surgical Technique………………………………………48 2.4.4 Histopathological protocol and further management……………………….....49 Chapter 3 Results………………………………………………………………….....50 3.1 Systematic Review, Meta-analysis for the Diagnostic Accuracy of Individual Diagnostic Tools in Staging cN0 HNSCC with and without Occult Neck Nodal Metastasis……………………………………………………………………………50 3.1.1 Univariate meta-analysis: US-FNA and SNB……………………………….....50 3.1.2 Bivariate models Analysis: CT, MRI, US & PET……………………………...51 3.1.3 SROC model analysis for all modalities…………………………………….....51 3.1.4 Quality assessment with QUDAS-1 and QUDAS-2…………………………...51 3.2 Comparison of Negative Predictive Value among Different Combined Diagnostic Tests Scenario ………………………………………………………………………..52 3.2.1 Simulation with sensitivity analysis……………………………………………52 3.2.2 Based case analysis………………………………………………………….....53 3.2.3 Sensitivity analysis after exclusion of studies with potential bias ……….…....53 3.3 Decision Modeling of Different Diagnostic Strategies in ManagingcN0 Neck with and without Occult Lymph Node Metastasis ………………………………………..54 3.3.1 Perspective of patient’s preference……………………………………………..54 3.3.2 Perspective of the health care system…………………………………………..55 3.3.2.1 Patients…………………………………………………………………….....55 3.3.2.2 Deterministic cost-effectiveness analysis…………………………………....55 3.3.2.3 Probabilistic cost-effectiveness analysis……………………………………..56 3.4 Role of Intra-operative Sentinel Lymph Node Biopsy in Oral Squamous Cell Carcinoma: a Feasibility Study in Taiwan……………………………………………57 3.4.1 Patients ………………………………………………………………………..57 3.4.2 Result of sentinel node navigation surgery………………………………….....57 3.4.3 Pathological examination………………………………………………………59 Chapter 4 Discussion…………………………………………………………………60 4.1 CT and MRI is basic in neck nodal evaluation…………………………………..60 4.2 PET has limited sensitivity to guide cN0 neck management…………….………61 4.3 Role of Ultrasound and US-FNA in cN0 neck management……………………..62 4.4 Sentinel node biopsy procedure………………………………………………….63 4.5 Intra-operative molecular biology and histologic tumor thickness………………65 4.6 Anatomic criteria in lymph node evaluation……………………………………..66 4.7 The necessary of combined tests…………………………………………………67 4.8 Sentinel node navigation surgery for early oral cancer…………………………..68 4.9 Health technology assessment by diagnostic meta-analysis……………………..70 4.10 Health technology assessment by economic evaluation ………………………..74 4.11 Perspective of surgeon’s preference in sentinel node navigation surgery………77 4.12 Other limitations………………………………………………………………...78 4.13 Future perspective………………………………………………………………80 4.14 Conclusions..........................................................................................................81 Figures Figure 1 Weiss decision analysis model………………………………………….…..82 Figure 2 The scar of END over neck…………………………………………………84 Figure 3 Summary of the different diagnostic tools in staging of non-palpable cN0 neck nodes……………………………………………………………………………85 Figure 4 The criteria used by ultrasound to different metastatic from benign lymph node …………………………………………………………………………….........86 Figure 5 The images show MRI exam and US-FNA………………………………...87 Figure 6 The image of PET scan and US-FNA……………………………………...88 Figure 7 Photos show the portable detector and marking……………………….…..89 Figure 8 Conceptualized sketch of the proposed serial diagnostic strategies ….……90 Figure 9 Proposed strategy of sequential tests………………………………….……91 Figure 10 Probability estimation of sequential test…………………………….……92 Figure 11 Decision tree modified from Weiss’s model………………………………93 Figure 12 Decision tree for cost-effectiveness analysis……………………………..94 Figure 13 Markov model and the transitional probability…………………………...95 Figure 14 Overall survival curve of 218 cN0 oral cancer with and without neck metastasis in our institute……………………………………………………………96 Figure 15 Preoperative dynamic lymphoscintigraphy………………………………97 Figure 16 Case presentation during operation………………………………………98 Figure 17 Literatures search and selection of studies…………………………….....99 Figure 18 Graph summarization (forest plots and SROC curves) for various diagnostic tools in cN0 neck evaluation………………………………………………………..100 Fig 18.1 The result of recruited study about USFNA in cN0 neck staging…………100 Fig 18.2 The result of recruited study about SNB in cN0 neck staging ……………101 Fig 18.3 The result of recruited study about CT in cN0 neck staging ……………103 Fig 18.4 The result of recruited study about MRI in cN0 neck staging ……………104 Fig 18.5 The result of recruited study about PET in cN0 neck staging……………..105 Fig 18.6 The result of recruited study about US in cN0 neck staging……………...106 Figure 19 QUDAS-2 all included studies…………………………………………..107 Figure 20 Simulation results on conditional probability analysis……………….....108 Figure 21 Hypothetical over- and under-treatment under different management strategies……………………………………………………………..109 Figure 22 Analysis based on Weiss decision model………………………………..110 Figure 23 Cost-effectiveness plane for twelve staging strategies for cN0 neck……112 Figure 24 Probabilistic sensitivity analysis with cost-effectiveness scatter plot…...113 Figure 25 Acceptability curve comparing the twelve strategies……………………114 Figure 26 Summary of recruited eleven oral cancer patients received SNNS……...115 Figure 27 Post operative view of SNNS and elective neck dissection…….…….....118 Figure 29 Pathology of the harvested SN…………………………………………...119 Figure 29 Two way sensitivity analysis of pre-test occult rate and recurrence rate after neck dissection……………………………………………………………………...120 Tables Table 1 Incidence of pathologically proven occult neck lymph nodal metastasis for cN0 head and neck cancer…………………………………………………………..121 Table 2 Calculation of diagnostic accuracy for individual study…………………...122 Table 3 Steps in the derivation of a SROC curve…………………………………..123 Table 4 The utility in modified Weiss’s model……………………………………..124 Table 5 Primary tumor sites for 218 patients with cN0 necks in FEMH, New Taipei City, Taiwan…………………………………………………………………………125 Table 6 Transition probability used in Markov model……………………………...126 Table 7 Input and output used in cost-effectiveness analysis………………….……127 Table 8 Summary of recruited literatures…………………………………………..128 Table 9 Univariate random effect model analysis of US-FNA in cN0 staging……..129 Table 10 Univariate random effect model analysis of SNB in cN0 staging………...130 Table 11 Bivariate random effect model analysis of CT in cN0 staging……….…...133 Table 12 Bivariate random effect model analysis of MRI in cN0 staging……….....134 Table 13 Bivariate random effect model analysis of US in cN0 staging………...….135 Table 14 Bivariate random effect model analysis of PET in cN0 staging…………..136 Table 15 Summary of the meta-analysis results…………………………………….137 Table 16 NPV of proposed sequential testing under conditional probability simulation…………………………………………………………………………...138 Table 17 Hypothetical over- and under-treatment with different management strategies………………………………………………………….…..139 Table 18 Ranking of each diagnostic strategy based on Weiss model………….…..140 Table 19 Ranking of each diagnostic strategy by cost from cost-effectiveness analysis……………………………………………………………………………...141 Table 20 Demography and results of sentinel node navigation surgery in Department of Otolaryngology Head and Neck Surgery, FEMH, New Taipei City, Taiwan (2013 Jan ~ 2014 Mar)…………………………………………………………………….142 Table 21 Literatures review about sentinel node navigation surgery………………143 Table 22 Recruited papers used size as a criterion to diagnose cN0 neck……….....144 References…………………………………………………………………………..145 Appendix 1: Figures …………………………………………………………….…155 Figure 1 Distribution of QUDAS-2 for studies about CT…………………………..155 Figure 2 Distribution of QUDAS-2 for studies about MRI………………………...156 Figure 3 Distribution of QUDAS-2 for studies about PET…………………………157 Figure 4 Distribution of QUDAS-2 for studies about US……………….…….........158 Figure 5 Distribution of QUDAS-2 for studies about US-FNA…………………….159 Figure 6 Distribution of QUDAS-2 for studies about SNB………………………...160 Figure 7 Simulation of positive metastasis with positive and negative results after exclusion of studies with potential bias in QUDAS 2…………………..…………..161 Figure 8 Simulation of over and under treatment after exclusion of studies with potential bias in QUDAS 2………………………………………………………….162 Appendix 2: Tables……………………………………………………………….…164 Table 1 Data extraction form…………………………………………………….….164 Table 2 QUDAS-1 assessment form………………………………………………..165 Table 3 QUDAS-2 assessment form………………………………………………..167 Table 4 Meta-analysis after exclusion studies with potential bias in QUDAS 2…...169 Table 5 Simulation of NPV after exclusion of studies with potential bias in QUDAS 2…………………………………………………………………………….………170 Appendix 3 Standards for Reporting of Diagnostic Accuracy (STARD) checklist ..171 Appendix 4: Related publishes……………………………………………………..1735818619 bytesapplication/pdf論文使用權限：同意有償授權(權利金給回饋學校)統合分析成本效益分析頸部轉移零期頭頸癌前哨淋巴結切片[SDGs]SDG3thesishttp://ntur.lib.ntu.edu.tw/bitstream/246246/262364/1/ntu-103-D99849015-1.pdf