應用隨機性成本效益於評估免疫化學糞便潛血法大腸直腸癌篩檢：分析基隆社區闔家歡健康篩檢資料

Abstract

背景 大腸直腸癌（Colorectal cancer ;CRC）篩檢一般使用傳統化學性糞便潛血法(Fecal Occult Blood Test; FOBT)，近年來免疫法興起許多人建議使用，但仍缺乏實證成本效益分析。而且影響篩檢成效因素尚包篩檢參與率及轉介順從率，這些因素會因為不同方法及個人特性的差異而有所不同。此外，成本計算上常會發生成本分布呈現偏態的問題及參數不確定性等問題，這些問題在族群決策分析上均扮演著重要的角色。 研究目的 本研究為隨機成本效益分析，主要目的包括四部份： 1. 利用社區族群篩檢實證資料探討社經人口學、生活型態、個人病史及家族病史等因素與大腸直腸腫瘤產生之相關性。 2. 利用社區族群進行免疫法糞便潛血檢查大腸直腸癌篩檢實證資料以探討人口學、生活型態、個人病史及家族病史如何影響第一次參檢率、再參檢率及轉介順從率。 3. 利用上述社區資料所得到大腸直腸癌個案探討使用二階段模式(two-stage model/hurdle model)調整偏態樣本分佈後，推估有關大腸直腸癌篩檢個案之疾病成本(Cost-of-Illness, COI)。 4. 利用上述1及2所得參數進行決定性及隨機機率模式針對不同間隔免疫法糞便潛血檢查、傳統化學法糞便潛血檢查及大腸直腸鏡檢之族群及個人層次之成本效益分析。 材料與方法 本研究用基隆市整合式篩檢1999-2004年間資料，將大腸直腸腺腫或癌之發生作為依變項，以人口學變項、生活型態、個人病史及家族病史等視為自變項，以Logistic regression function去探討發生大腸直腸腺腫或癌症之危險因子。此外，亦利用相同的自變項及分析方法，將篩檢率及轉介順從率視為依變項，探討其間的關係。最後將上述依個人特性去預到的發生腫瘤的危險性及參檢率、轉介順從率等應用於第四部分成本效益分析。 其次，由於篩檢大腸直腸癌個案成本含有設限及偏態的問題，本論文應用Hurdle model 進行成本的效正。 成本效益分析分別使用決定性模式及隨機模式。在隨機模式中使用貝氏隨機模式將相關參數依貝氏共軛對方式指定進行隨機成本效益分析，以計算增加成本效益比及達到成本效益機率之接受曲線。 主要研究結果 以決定模式作評估，並以社會的觀點來看時，相對於未篩檢，各項篩檢策略均具絕對之優勢，均為成本節約（cost-saving）的方案。所累積的淨效益以每十年一次大腸直腸鏡檢最多，為12.7993人年命，其次為每年一次免疫法的FOBT，可得12.7988人年。若比較各組之增加成本效果比(incremental cost-effectiveness ratio;ICER)時，則每年一次免疫法的FOBT為最佳策略。 同樣的方法進行次族群分析時，除50-59歲女性以外，其它的次族群均以每10年一次的大腸直腸鏡檢為最佳策略。 以個人的觀點評估55歲男性，在不同的大腸直腸的危險因子的影響下，具大腸直腸癌家族病史者之最佳策略為每10年做一次大腸直腸鏡檢；無家族史者則以每年做一次iFOBT為最佳篩檢策略。 當考慮到個人因素異質性，以隨機模式進行評估時以每10年一次的大腸直腸鏡檢具成本效益的機率最高，尤其是針對60-69歲男性而言。 結論 本研究應用隨機成本效益貝氏分析基隆地區以不同間隔提供免疫法糞便潛血以進行大腸直腸癌篩檢時發現，即使考慮二次不確定變異下，每年篩檢一次在族群層次上具成本效益。透過個人對大腸直腸癌的易感受性及參檢率與順從率等影響因素之探討，及修正成本偏態後，可顯示個人層次成本效益之結果，此種方法可應用於未列入國家篩檢政策之疾病。

Background Chemical Fecal Occult Blood Test (FOBT) has been used to establish mass colorectal cancer (CRC) screening program traditionally. However, since Immunochemical FOBT has a higher predicting power, many people start to advocate using it for CRC screening. But so far there are not enough evidences to prove if it is cost-effectiveness. In addition, determinants of the effectiveness of screening also include attendance rate and compliance rate. Both of them will vary among people because of different personal characteristics. Besides, calculating cost often suffers the problems such as skewed data and uncertainty of parameters. Such problems play important roles for population decision-making. Research purpose This study was designed as a probabilistic cost-effectiveness analysis with four purposes: 1. To use community population screening data to investigate the association between covariates and tumor detected, those covariates including factors of socio-demographic, life style, personal disease history and family history. 2. To apply community population screening data to investigate the association between the attendance rate and referral rate and the same set of covariates used in 1. 3. To use cost data in community population screening to calculate the cost of illness of CRC. The Hurdle model was used to adjust the skewed cost. 4. To apply those parameters estimated above to establish deterministic and stochastic models, to compare different intervals of immunochemical FOBT screenings, and also for comparison the chemical FOBT and colonoscopy through cost-effectiveness analysis. Materials and methods The main data source is Keelung Community-based Integrated Screening (KCIS) data from 1999 to 2004.We investigated the association between colorectal tumor and a set of covariates, including socio-demography, life style, personal disease history and family history, through logistic regression function. Then we used the same method to predict compliance behavior. Finally, incorporating those parameters estimated by above methods into cost-effectiveness analysis. Besides, in order to account for censored and skewed cost problems in CRC screening, we adopted the Hurdle model to adjust the CRC medical cost. In terms of the analysis of cost-effectiveness, two approaches were used, namely deterministic and stochastic. In the stochastic model, Bayesian probabilistic estimation was used through Bayesian Conjugate distribution. The comparing indicators were incremental cost –effectiveness ratios (ICER) and acceptability curve. Results From a societal viewpoint through deterministic approach, comparing to no-screen, the other screening strategies are dominant and cost-saving. Discounted cumulative effectiveness shows colonoscopy every 10 years is the most-effective one, which will save 12.7993 life years. The following one is to have iFOBT (immunochemical Fecal Occult Blood Test) annually, which will save 12.7988 life years. Comparing the ICERs of all strategies to no-screen, iFOBT annually is the best strategy. Processing sub-groups analysis under the same approach, we found that undergoing colonoscopy every 10 years was the best strategy among all subgroups except the subgroup constituted of females aged 50-59. When evaluating males aged 55 affected by different risk factors of colorectal tumor, for those with family history of CRC, the best strategy is colonoscopy every 10 years. On the other hand, for those without family history of CRC, iFOBT annually is the best strategy. Taking personal heterogeneity into account, the economical evaluation through probabilistic model showed colonoscopy every 10 years is the best strategy, especially for the subgroup constituted of males aged 60-69. Conclusion This study applied Bayesian probabilistic cost-effectiveness to evaluate community-based CRC screening program by iFOBT. Under the consideration about second uncertainty, iFOBT annually shows cost-effectiveness on the population-level. Since we had been incorporated the covariates about personal susceptibility and compliance, and modified the skewed cost data, then we did the personal-level cost- effectiveness analysis. Such method could be applied to national screening policy for those diseases without any screening policy now.