鼻咽癌多重篩檢工具的評估

Evaluation of Multiple Detection Modalities for Nasopharyngeal Carcinoma

曾智傑(Chih-Chieh Tseng)

指導教授 ： 陳秀熙

共同指導教授 ： 嚴明芳

國立臺灣大學/公共衛生學院/預防醫學研究所/碩士(2007年)

https://doi.org/10.6342/NTU.2007.01556

摘要

從篩檢實務面來看，利用單一檢測工具篩檢疾病時雖然有方法簡便和價格便宜的優點，但是卻有敏感度和精確度不足的問題。為了增加篩檢敏感度和精確度，我們就需要合併多重檢測工具的方式，然而多重檢測工具之使用在方法學上我們會遇到一些問題，是單一檢測工具無法解決的問題，例如多重檢測工具合併後的敏感度和精確度、檢測工具多早前可以偵測出疾病，以及如何應用多重篩檢工具敏感度及精確度於多重檢測工具的成本效益問題。以鼻咽癌而言，利用單一工具EBV VCA IgA篩檢鼻咽癌時，會有精確度不足及偽陽性太高的問題。為了同時增加篩檢工具的敏感度和精確度，就需要合併多重鼻咽癌的檢測工具。然而合併了多重檢測工具後，篩檢的成本也隨之提高，因此就需要各種篩檢方法經濟效益的分析。本論文主要研究的目的為：(1)利用貝氏固定和隨機模式進行單一檢測工具敏感度和精確度的整合分析模式。(2)多重檢測工具的條件相依模式。(3)利用多階段模式估計多重檢測工具的臨床偵測前期與敏感度之關係。(4)利用(3)的結果進行鼻咽癌的多重檢測工具的成本效益分析。
本研究以發展多重篩檢工具的方法學並將之應用於鼻咽癌早期發現，其架構如下：
1. 以系統性文獻回顧方式，搜尋Medline和PubMed兩大外文資料庫，計收錄關鍵字含screening, Ebstein-barr virus (EBV)和Nasopharyngeal carcinoma文章計36篇，由於各研究間存在著異質性，因此我們分別以貝氏固定效應模式和貝氏隨機效應模式進行整合性分析，得到鼻咽癌篩檢單一工具的敏感度及精確度。
2. 進一步進行同時考慮兩種工具的整合性分析，由於兩種測量工具存在生物相似性，因此我們引入貝氏條件相依模式進行整合性分析。
3. 利用敏感度和臨床偵測前期之間的關係求算利用不同檢測工具進行篩檢時，由該工具最早可偵測時間點到臨床症狀發生之時間間隔。
4. 考慮利用兩種工具進行檢測時，族群內的個體可能存在何種工具較早可偵測到的順序不一致的情況，因此我們發展一混合性多階段疾病自然史模式，利用電腦模擬方式產生虛設世代，並進行參數估計及模式驗證。
5. 利用(4)的結果進行僅用單一工具(EBV VCA IgA或Nasopharyngoscope)及併用兩種工具相較於不篩檢時的經濟評估。
本研究貝氏隨機效應模式整合分析的結果發現敏感度最高的是EBV BamHI-W[中位數為91% (95%CI=67-97%)]，其次為EBV VCA IgA (cross-sectional)、Nasopharyngeal swab、EBV VCA IgA、EBV EA IgA、Nasopharyngoscope及EBV EBNA。精確度最高的是Nasopharyngoscope [中位數為98% (95%CI=66-99%)]，其次為EBV EA IgA、EBV VCA IgA (cross-sectional)、EBV BamHI-W、EBV EBNA、Nasopharyngeal swab及EBV VCA IgA。貝氏固定效應模式亦有相似的發現，但由模式的DIC值顯示大部份隨機模式均較固定模式來的好，除了鼻咽拭子之外。再利用多重檢測工具的敏感度、轉移速率和臨床偵測前期的關連，可得利用EBV BamHI-W檢測的臨床症前期約為2.79年，鼻咽拭子約為2.77年，EBV VCA IgA約為2.68年，EBV EA IgA、EBV EBNA及鼻咽鏡為2.34年，鼻咽細胞學為2.22年，合併鼻咽內視鏡和EBV VCA IgA、合併EBV EA IgA和EBV VCA IgA及合併EBV EBNA和EBV VCA IgA為2.97年。合併EBV BamHI-W和EBV VCA IgA 3.01年。
利用混合性多階段疾病自然史模式所得估計結果接近提供的真值，利用估計值得到僅使用EBV VCA IgA、僅使用鼻咽內視鏡及合併兩種工具的敏感度分別為87.6%, 72.4%及97.4%，與之前整合式分析結果相近。進一步利用以上參數比較未篩檢，利用EBV VCA IgA及鼻咽鏡每一年或每兩年篩檢一次，結果發現和未篩檢比較，每2年EBV VCA IgA篩檢的增加成本效益比例(ICER)是NT365,095、每年EBV VCA IgA篩檢的增加成本效益比例(ICER)是NT442,145，而每年合併EBV VCA IgA和鼻咽鏡篩檢的增加成本效益比例(ICER)是NT1,091,533。和未篩檢比較，每2年EBV VCA IgA篩檢17,857人可以減少一個鼻咽癌死亡。每年EBV VCA IgA篩檢23,622人可以減少一個鼻咽癌死亡。每年合併EBV VCA IgA和鼻咽鏡篩檢21,739人可以減少一個鼻咽癌死亡。
本研究最大新穎想法是將多階段模式疾病隨時間進展之概念與篩檢敏感度之相關進行聯結及互轉，此方法若應用於篩檢實務，將可提供做篩檢政策之參考，如篩檢間隔的決定及成本效益分析。

關鍵字

篩檢 ； 鼻咽癌 ； 敏感度 ； 精確度

並列摘要

Disease screening with single detection modality is convenient and cheap, but it may be less effectiveness in terms of sensitivity and specificity. In order to solve this problem, multiple detection modalities may be needed. However, the methodology of combining multiple detection modalities is difficult, such as the combined sensitivity and specificity, detection ability, and cost-effectiveness analysis. Take screening of nasopharyngeal carcinoma (NPC) with EBV VCA IgA for example. The specificity of EBV VCA IgA is not good enough that the false positive rate is too high. Although multiple detection modalities can raise sensitivity and specificity at the same time, it will also increase the cost of screening. Therefore, the cost-effectiveness analysis of multiple detection modalities ensures. The purpose of this study is to complete (1) the meta-analysis of sensitivity and specificity of each detection modality using Bayesian fixed-effect and random-effect model, (2) conditional dependency model of multiple detection modalities, (3) the relation between sensitivity and sojourn time using multi-state model, (4) the cost-effectiveness of multiple detection modalities for NPC.
The methodology part of multiple detection modalities is described as following: (1) We systemically searched Medline and PubMed with the key words of screening, Ebstein-barr virus(EBV) and nasopharyngeal carcinoma. The meta-analysis of the included 36 literature was performed with Bayesian fixed-effect and random-effect model to deal with the heterogeneity between each study. The results of sensitivity and specificity of each detection modality were obtained. (2) We used Bayesian conditional dependency model to estimate the combined sensitivity and specificity of detection modalities. This model adjusted with the biological correlation between combined detection modalities. (3)With the relation of sensitivity and pre-clinical detectable phase (PCPD, sojourn time), we estimated the earliness of detectable phase of each modalities. (4)We proposed a multi-state model for a mixture type of tumor progression. We also simulated the parameters for the mixture type model with the method of Monte Carlo simulation. Validation of the parameters was also performed. (5) We completed the cost-effectiveness analysis of NPC screening with different screening strategies using the parameters we generated in (4).
The results of meta-analysis of the sensitivity for different detection modalities for nasopharyngeal carcinoma with Bayesian random-effect model are described as following. Serology of EBV BamHI-W region gene is the most sensitive detection modality (median value of 91%), followed by nasopharyngeal swab , EBV VCA IgA , EBV EA IgA , EBV EBNA, and nasopharyngoscope. In the same way, the most specific detection modality is nasopharyngoscope (median value of 98%), followed by EBV EA IgA, nasopharyngeal swab, EBV BamHI-W, EBV VCA IgA and EBV EBNA. The results of Bayesian fixed-effect model were similar to that of random-effect model. Using DIC (Deviation information criteria) to compare the two models, we found the random-effect model is better than fixed-effect model.
The results of PCDP of each detection modalities were obtained with the relation between sensitivities, transitional rates, and PCDP. The PCDP of EBV BamHI-W is 2.79 year, nasopharyngeal swab is 2.77 year, EBV VCA IgA is 2.68 year, EBV EA IgA, EBV EBNA and nasopharyngoscope are 2.34 year, and nasopharyngeal cytology is 2.22 year. The PCDP of combining nasopharyngoscope and EBV VCA IgA, EBV EA IgA and EBV VCA IgA, and EBV EBNA and EBV VCA IgA is the same, 2.97 year. The PCDP of combining EBV BamHI-W and EBV VCA IgA is longest, 3.01 year.
The estimated value is close to the true value in the model of mixture type tumor by simulation method. The sensitivity of simulation model is also close to that of meta-analysis. In cost-effectiveness analysis, the ICER (incremental cost-effectiveness ratio) is 365,095 when the strategy of no screening and 2-year screening with EBV VCA IgA is compared. The ICER is 442,145 with no screening and annually screening with EBV VCA IgA. The ICER is 1,091,533 with no screening and annually screening with combining EBV VCA IgA and nasopharyngoscope. Compared with no screening, 2-year screening with EBV VCA IgA can reduce one NPC death per 17,857 screening, annual screening with EBV VCA IgA can reduce one NPC death per 23,622 screening, and annual screening with combining EBV VCA IgA and nasopharyngoscope can reduce one NPC death per 21,739.
The novelty of this study is combining the idea of sensitivity and multi-state model. In this way, the sensitivity of detection modalities has taken the influence of time progression into account. This method can be applied in disease screening in terms of screening interval determination and cost-effectiveness analysis.

並列關鍵字

Detection modalities ； Nasopharyngeal carcinoma ； Sensitivity ； Specificity

參考文獻

◎ Branscum AJ, Gardner IA, Johnson WO. Estimation of diagnostic-test sensitivity and specificity through Bayesian modeling. Prev Vet Med. 2005;68:145-163.

◎ Chang AR, Liang XM, Chan A, Chan M, Teo RM, Johnson PJ. The use of brush cytology and directed biopsies for the detection of nasopharyngeal carcinoma and precursor lesions. Head Neck 2001; 23:637-645.

◎ Chen HH, Duffy SW, Tabar L. A markov chain method to estimate the tumour progression rate from preclinical to clinical phase, sensitivity and positivie predictive value for mammography in breast cancer screening. The Statistician 1996;45:307-317.

◎ Chen MR, Liu MY, Hsu SM, Fong CC, Chen CJ, Chen IH, Hsu MM, Yang CS, Chen JY. Use of bacterially expressed EBNA-1 protein cloned from a nasopharyngeal carcinoma biopsy as a screening test for NPC patients. J Med virol 2001;64:51-57.

◎ Cheng WM, Chan KH, Chen HL, Luo RX, NG S, Luk W, Zheng BJ et al. Assessing the risk of nasopharyngeal carcinoma on the basis of EBV antibody spectrum. Int J Cancer 2002;97:489-492.