結合糞便血紅蛋白及基因模式的大腸直腸癌個人化篩檢

Abstract

背景 由於糞便血紅蛋白濃度（f-Hb）已被證明是大腸直腸癌發生及死亡的預測因子。因此，如何利用 f-Hb 結合已開發的個人化基因資料來運用糞便免疫測試 (FIT)來發展個人化大腸直腸癌是值得研究的。 研究目的 鑑於有限的醫療資源和大腸鏡檢查的專業人力，我們的目標是開發精準的大腸直腸癌 FIT 篩檢，以優化 FIT 和大腸鏡檢查的使用。 研究方法 我們首先建立了一個具有特定狀態相關多因子多階段疾病進展模型，利用此模型發展一動態個人化風險預測模型，用於從小腺瘤到大腺瘤和臨床前可偵測大腸直腸癌( CRC )直至臨床期 CRC 的轉變。我們使用透過文獻回顧可獲得的資料來估計每個特定疾病狀態中相關因子的臨床比重。我們依照個人化風險預測模式使用數位雙胞胎方法創建兩個虛擬組別，分別為兩年篩檢與精準個人化篩檢。我們使用馬可夫指數回歸模式來估計臨床權重以建立多重疾病狀態風險分數。 結果 我們呈現多重疾病狀態風險分數並衍生動態個人化風險曲線。發展每十分位數分布的風險分數來預測大腸直腸癌的風險。更依據此風險分數將篩檢間隔分為1-6年形成個人化篩檢依據。在這一百萬人口中，我們所設計的個人化篩檢間隔結果發現精準大腸直腸癌FIT可減少12% FIT檢測量，在 7% 的陽性率下，產生840251次FIT 和 58818支大腸鏡檢查的減少。 結論 透過發展個人化不同間隔的篩檢(不管有無結合糞便DNA測試)，我們證實糞便血紅蛋白濃度合併基因訊息的最佳使用可以減少FIT測試及大腸鏡檢查。

Background As fecal hemoglobin concentration (f-Hb) has been demonstrated as a predictor for incident colorectal cancer and its mortality, it is of worthy of being investigated how to make use of f-Hb combined with the already developed personalized genetic profiles to develop personalized colorectal cancer screening with fecal immunological test (FIT). Objective We aimed to develop precision colorectal cancer FIT screening for the optimal use of FIT and colonoscopies given limited medical resources and professional manpower for colonoscopy. Methods We first built up a multistate disease progression model with the superimposition of state-specific correlates for dynamic personalized risk prediction model for each transition from small adenoma, through large adenoma and pre-clinical detectable CRC until clinical CRC. Each clinical weight corresponding to each state-specific correlate was estimated by using available empirical tabular data from literature review. A digital twin approach was used to create two virtual groups guided by the proposed individualized multistate risk prediction model including biennial screening regime versus precision screening regime. The Markov exponential regression model was applied to estimate clinical weights for building multistate risk scores. Results Multistate risk scores are presented for deriving dynamic personalized risk curves. The decile distribution of risk score was developed for predicting the risk of CRC. Various inter-screening intervals were further assigned for each decile of risk score ranging from 1 to 6 years. Given personalized inter-screening interval assigned for one million population, precision CRC FIT reduced 12% of FIT test, yielding the reduction of 840251 FIT and 58818 colonoscopies given 7% positive rate. Conclusions The optimal use of f-HB with or without combing genetic profiles for reducing FIT test and colonoscopies has been demonstrated by developing personalized inter-screening with and without stool DNA test.