陳秀熙臺灣大學：預防醫學研究所詹麗雪Chan, Li-SheueLi-SheueChan2007-11-282018-06-292007-11-282018-06-292005http://ntur.lib.ntu.edu.tw//handle/246246/59211前言： 雖然大規模乳癌篩檢在降低乳癌死亡率上的效益己經被證實，且國際乳癌篩檢網絡組織(International Breast Screening Network, IBSN)在乳房攝影術品質管制的努力也獲得肯定，但大規模乳癌篩檢在降低乳癌死亡率上與乳房攝影術品質管制的關係該如何聯結則鮮少被提出。此外，因應大規模乳癌篩檢的醫師人力供需平衡的問題也缺乏實證研究。 目的：　本論文主要目的有三，一、了解IBSN羅列之乳癌篩檢品質保證元素與主要結果之間的相關；二、利用問卷調查方式了解臺灣地區實證資料；三、利用調查所知醫師全職人力資料推估在大規模乳癌篩檢的需求下，臺灣地區醫師人力的供需平衡。 材料與方法：　由文獻查證的方式得到與乳癌篩檢相關的篩檢前期、篩檢期及篩檢後期的相關指標，並利用反向相關分析(Backward correlation analysis)了解其結構面、過程面及結果面的元素與篩檢成效指標之關聯性。為能了解臺灣地區乳房攝影的品質保證指標，我們針對85家國民健康局認為有能力執行乳房攝影術的醫療機構之乳房外科及放射科醫師做調查。此外，在全國乳癌篩檢的政策下，醫師人力的需求則利用馬可夫鏈樹狀圖進行推算。 結果：　在調整乳癌發生率後，影響乳癌死亡率的最重要結構面因子為乳房外科醫師執照(迴歸係數＝-0.23，標準差=0.08，p=0.007)，影響乳癌死亡率的最重要過程面因子為乳房攝影術為是否雙面(two views)照影(迴歸係數＝-0.28，標準差=0.026， p<0.001)，而最重要的結果面指標則是敏感度(迴歸係數＝-1.813，標準差=0.9241， p=0.0498)。 逐步反向分析顯示淋巴結陽性比例會被篩檢間隔個案相對於發生率的比例所影響，而此比例又被敏感度及是否有標準化報告的選項所影響，敏感度則被乳房攝影術為是否雙面及重覆判讀所影響。 　　實際調查資料結果可以預測臺灣地區乳癌死亡率為每十萬人14.4例，非常接近現在臺灣地區的實際乳癌死亡率每十萬人11.32例。 　　在假設70%的參與率下，臺灣地區於六年期間乳房攝影術的需求量約為125萬張，依據臺灣地區現有醫師人力發現乳房外科約可承受110萬張而放射科醫師約可承受410萬張的需求量。若考慮全職人力(乳房外科醫師現有一人力約相當於貢獻於乳癌病人10%全職人力)，則有約99名全職乳房外科人力於六年期間服務222名乳癌個案。 結論：　本研究針對乳房攝影術品質保證指標與死亡率降低之相關性做一連串探索，並利用此資料結合五階段乳癌自然史應用於臺灣地區實施乳房攝影術乳癌篩檢政策下人力供需平衡的推算。Introduction. While mass screening for breast cancer with mammography in the reduction of mortality has been demonstrated and quality assurance mammography programme has been sponsored by International Breast Screening Network (IBSN) how the components of quality assurance mammography programme affect mortality or relevant outcomes has been rarely addressed. Demand and supply of manpower involved in breast cancer screening is also lacking given this programme. Objective. This thesis aimed (1) to investigate the relationships of components of quality assurance programme of breast cancer screening presented in the IBSN to relate outcome; (2) to conduct an empirical survey in Taiwan following the design inherent from the components of (1); and (3) to assess the balance between supply and demand in association with population-based breast cancer screening following information on full-time equivalent (FTE) and capacity of clinical manpower obtained from (2). Materials and Methods. Data on three dimensions related to pre-screening phase, screening phase and post-screening phase were extracted from literatures. The relationships of structural components, process components, and outcome components to main outcome were analyzed by ecological correlation study based on information collected from previous literatures by using backward correlation analysis. To make a better understanding of the components of quality assurance in Taiwan, we conducted a empirical survey aimed at target population from 85 institutions with mammography screening approved by Bureau of Health Promotion. In order to calculate the demand for mammography in national screening program, and disease burden, we used Markov cycle tree underpinning the five-sate Markov model for natural course of breast cancer to predict the required number. When dependent variables were of incidence of rare event, Poisson regression was conducted. For variables not of rare event, such as sensitivity, specificity, and prediction of positive value, etc., simple linear regression was used. Results. After controlling the underlying incidence rate, the most significant structural components responsible for mortality was licensure of breast surgeon (regression coefficient=-0.23 (SE=0.08), P=0.007). The most significant process components was the number of view to be taken on mammography examination (regression coefficient=-0.28 (SE=0.026), P < 0.001). The most important outcome components was sensitivity (regression coefficient= -1.813 (SE=0.9241), P=0.0498). Step-by-step backward analysis showed the proportion of node positive was determined by high interval cancer as percentage of the underlying incidence rate, which was, in turn, affected by sensitivity and nomenclature. Sensitivity was influenced by number of view to be taken and double reading. In addition, double reading and number of view to be taken play important role in situ detection rate. Based on the results from the empirical survey on components of quality assurance and the developed prediction equation, we predicted 27% proportion of node positive and 14.4/100,000 mortality given significant components for both radiologist and breast surgeon, which is closed to, but slightly higher, than ,11.32/100,000 mortality as observed in Taiwan now. Given 70% attendance rate, 1252544 mammograms are requested during six year period. The number of supply of manpower was 1120940 from breast surgeon and 4099788 from radiologists in six years. The demand of breast cancer for breast surgeon in treating disease was 21837 given 991 breast surgeons. Taking full-time equivalent time (FTE with only 10% breast surgeon get involved in breast cancer into account), approximately 99 FTE from breast surgeons will be involved in treating breast. This implies each FTE of breast surgeon may serve 222 breast cancers during six years. In conclusion, a series of quantitative models pertaining to the relationships of components of quality assurance mammography programme to mortality or related outcome. These models together with the five-state Markov model were further applied to analysis of demand and supply of mass screening for breast cancer with mammography in Taiwan.CONTENT Acknowledgment vii Abstract（in Chinese） ix Abstract xii Part I Introduction 1.1 Quality Assurance for Screening Mammography in the IBSN 1 1.2 Significance and Application of Quality Assurance programme 2 1.3 Objective 3 Part II Literature Review 2.1 Population-based randomized controlled trials on breast cancer screening – experience from western countries 5 2.2 International Breast Cancer Screening Network, IBSN 9 2.3 Components of quality assurance 12 2.4 Reliability of mammography 17 2.5 Data quality control 18 Part III Materials and Methods 3.1 Meta-Analysis for Quality Assurance Programme of Mammography Screening 20 3.2 Empirical Survey on Components of Quality Assurance 22 3.3 Decision Tree for Estimating Demand of Mammography Screening 22 3.4 Statistical Methods 24 Part IV Results 4.1 Descriptive Findings of Main Outcomes 26 4.2 Meta-analysis for the relationships between components of quality assurance and main outcomes 27 4.3 Results of Survey 31 4.4 Demand and Supply of Population-based Screening with Mammography 32 Part V Discussion 34 References 39 附錄一：乳房外科醫師乳房篩檢品質保證及人力調查問卷 94 附錄二：放診科醫師乳房篩檢品質保證及人力調查問卷 97 LIST OF TABLE（表目錄） Table 2.1.1 Breast cancer screening projects: randomised clinical trial 49 Table 2.2.2 Breast cancer screening projects: non-randomised clinical trial 51 Table 4.1.1. Indicators of outcome of mammography breast cancer screening from 22 countries in IBSN_1 52 Table 4.1.2. Indicators of outcome of mammography breast cancer screening from 22 countries in IBSN_2 53 Table 4.1.3. Indicators of outcome of mammography breast cancer screening from 22 countries in IBSN_3 54 Table 4.2.1. The correlation between components of mammography and breast cancer mortality rate in IBSN 55 Table 4.2.2. The correlation between components of mammography and breast cancer abnormal recall rate in IBSN 57 Table 4.2.4.3 The correlation between components of mammography and breast cancer nodal positive in IBSN 59 Table 4.2.3.4 The correlation between components of mammography and breast cancer interval cancer rate in IBSN 61 Table 4.2.3.1 The correlation between components of mammography and breast cancer sensitivity in IBSN 63 Table 4.2.3.2 The correlation between components of mammography and breast cancer specificity in IBSN 65 Table 4.2.3 The correlation between components of mammography and breast cancer positive predictive value in IBSN 67 Table 4.2.4.1 The correlation between components of mammography and breast cancer in situ cancer detection rate in IBSN 69 表4.3.1 研究對象與單位資源 71 表4.3.2 乳房外科與放診科專科醫師人力分佈特性 72 表4.3.3 乳房攝影篩檢經費補助來源-組織面 73 表4.3.4 乳房攝影篩檢人員在職訓練需求-結構面 74 表4.3.5 乳房攝影品質-過程面執行情形（一） 75 表4.3.6 醫師對於不正常X光攝影檢查結果之告知方式- 過程面執行情形（二） 76 表4.3.7 乳房攝影品質-結果面執行情形 77 表4.3.8 醫師人力常態性工作時間分配 78 表4.3.9 醫師人力使用於乳房腫瘤疾病個案時間分配情形/ (時/週) 79 表4.3.10 判讀正常與異常乳房X光攝影檢查Screen film時間 80 表4.3.11 醫師人力非常態工作時間分配 (時/月) 81 表4.3.12 醫師非常態工作之時間比例 82 表4.3.13 乳房X光攝影檢查給付情形（一） 83 表4.3.14 執行乳房X光攝影檢查給付情形（二） 83 表4.3.15 問卷反應測試結果 84 Table 4.4.1 Mammography demands and disease burden in six years for women aged 50-69 and 40-49 years by different attendance rates 85 Table 4.4.2 Capacity for mammography taken by breast breast surgeon and radiologist 86 LIST OF FIGURE（圖目錄） Figure 3.1 The process of study design（研究設計流程） 87 Figure 3.3.2 Decision tree for demand of mammography screening 88 Figure 3.3.1 Five-state natural history model for breast cancer 89 Figure 4.2.5.1 Significant components affecting mortality 90 Figure 4.2.5.2 Interval cancer and lymph node positive 90 Figure 4.2.5.3 Sensitivity and standard report and interval cancer 91 Figure 4.2.5.4 Process component and sensitivity 91 Figure 4.2.5.5 Double reading and two view and in situ detection rate 92 Figure 4.2.5.6 Minimum number of examination and recall rate and positive predictive value 92 Figure 4.2.5.7 Recall rate and specificity 93 Figure 4.3.1 Prediction of performance of screening, prognostic factors, and mortality for mammography screening by radiologist and breast surgeon in Taiwan 933197964 bytesapplication/pdfen-US乳癌篩檢乳房攝影術品質保證逐步反向分析國際乳癌篩檢網絡mass screening for breast cancermammographyquality assurancestep-by-step backward analysisInternational Breast Screening Network (IBSN)[SDGs]SDG3thesishttp://ntur.lib.ntu.edu.tw/bitstream/246246/59211/1/ntu-94-R92846002-1.pdf