Misclassification of Human Papillomavirus Infection in Epidemiological Studies: Nature and Consequences

人類乳突瘤病毒研究之測量誤差：誤差本質及其後果

孫建安(Chien-An Sun) ； 楊智如(Chih-Ju Yang) ； 朱堂元(Tang-Yuan Chu) ； 謝長堯(Chang-Yao Hsieh) ； 游山林(San-Lin You) ； 余慕賢(Mu-Hsien Yu)

《Journal of Medical Sciences》 20卷7期 (2000/08) Pp. 333-341

摘要

測量誤差所導致暴露於危險因子狀態的錯誤分類是傳統和分子流行病學研究中主要的偏差之一。例如，使用特異度不高之第一代C型肝炎病毒抗體的酵素免疫檢驗方法來進行流行病學研究，可能會產生過多的假陽性個案而導致C型肝炎病毒感染盛行率的高估以及對於此一病毒感染與罹患肝癌危險性之相關性評估的偏差。事實上，整個人類乳突瘤病毒感染之流行病學研究領域係隨著人類乳突瘤病毒檢測技術之進展而發展，此一領域亦是提供在分子流行病學中測量誤差所導致嚴重錯誤研究結果之最明顯的研究案例。造成人類乳突瘤病毒檢測產生誤差的主要原因係來自於使用不同敏感度之實驗方法及此一病毒所呈現感染之動態變化。此篇論文旨在引用文獻報告中的研究數據來探討人類乳突瘤病毒感染狀態之錯誤分類對於婦女族群感染盛行率及此一病毒感染與婦女罹患子宮頸癌病變相關性等估計值的影響。分析結果發現即使是低度的錯誤分類（例如，在檢測人穎乳突瘤病毒感染之檢驗方法的敏感度和特異度為99%），亦能對於婦女族群人額乳突瘤病毒感染盛行率及此一病毒感染所導致子宮頸癌病變危險之相關性等流行病學估計值造成嚴重的影響。本文作者認為致力於人類乳突瘤病毒實驗檢測程序之標準化和不同實驗室之間的比較是提供可以比較研究數據的基本要求；同時，採用長期追蹤、重覆收取研究資料、檢體之流行病學世代追蹤研究設計才能增加人類乳突瘤病毒感染累積暴露量評估之正確性及降低研究其所導致臨床疾病之研究偏差。

關鍵字

人類乳突瘤病毒 ； 錯誤分類 ； 子宮頸癌變

並列摘要

Measurement error in exposure assessment is one of the major sources of bias in traditional and molecular epidemiological studies. For example, use of a relatively low specificity of the enzyme immunoassay tests for antibodies to hepatitis C virus (HCV) could result in a large number of false postive individuals, which would bias the studies of HCV prevalence and the relationship between this virus infection and risk of liver cancer. The entire field of epidemiology of genital tract human papillomavirus (HPV) infection in female population has made progress in parallel with advances in HPV-testing methods and has provided one of the clearest practical examples of the devastating effect of measurement error in molecular epidemiology. Misclassification of HPV infection status is mainly caused by laboratory techniques with widely varying degrees of sensitivities for the detection of HPV DNA and the transience of HPV infection. This report examines effects of misclassification of HPV infection status in biasing the estimations of HPV prevalence and the association between this virus infection and risk of cervical neoplasia in epidemiological studies. Some simple numeric examples have been derived from published study results in the literature. As shown in the present report, even low levels of misclassification (such as the sensitivity and specificity of 99% for detection of HPV) can considerably distort the presumed HPV prevalence as determined by epidemiological surveys of the female population. Furthermore, the same amounts of measurement errors can dramatically affect epidemiological association between HPV infection and risk of cervical neoplasia. It is conceivable that efforts in standardizing assay procedures and interlaboratory comparisons are essential to obtain comparability of information across studies. Additionally, a longitudinal, repeated-assessment epidemiological cohort study is the design of choice to increase the accuracy and to reduce bias in the evaluation of cumulative HPV exposure and disease outcome.

並列關鍵字

human papillomavirus ； misclassification ； cervical neoplasia