[A systematic review of worldwide natural history models of colorectal cancer: classification, transition rate and a recommendation for developing Chinese population-specific model].

Objective: To review the worldwide studies on natural history models among colorectal cancer (CRC), and to inform building a Chinese population-specific CRC model and developing a platform for further evaluation of CRC screening and other interventions in population in China. Methods: A structured literature search process was conducted in PubMed and the target publication dates were from January 1995 to December 2014. Information about classification systems on both colorectal cancer and precancer on corresponding transition rate, were extracted and summarized. Indicators were mainly expressed by the medians and ranges of annual progression or regression rate. Results: A total of 24 studies were extracted from 1 022 studies, most were from America (n=9), but 2 from China including 1 from the mainland area, mainly based on Markov model (n=22). Classification systems for adenomas included progression risk (n=9) and the sizes of adenoma (n=13, divided into two ways) as follows: 1) Based on studies where adenoma was risk-dependent, the median annual transition rates, from ' normal status' to ' non-advanced adenoma', 'non-advanced' to ' advanced' and ' advanced adenoma' to CRC were 0.016 0 (range: 0.002 2-0.020 0), 0.020 (range: 0.002-0.177) and 0.044 (range: 0.005-0.063), respectively. 2) Median annual transition rates, based on studies where adenoma were classified by sizes, into <10 mm and ≥10 mm (n=7), from ' normal' to adenoma <10 mm, from adenoma <10 mm to adenoma ≥10 mm and adenoma ≥ 10 mm to CRC, were 0.016 7 (range: 0.015 0-0.037 0), 0.020 (range: 0.015-0.035) and 0.040 0 (range: 0.008 5-0.050 0), respectively. 3) Median annual transition rates, based on studies where adenoma, were classified by sizes into diminutive (≤5 mm), small (6-9 mm) and large adenoma (≥10 mm) (n=6), from ' normal' to diminutive adenoma,'diminutive' to ' small','small' to ' large', and large adenoma to CRC were 0.013 (range: 0.009-0.019), 0.043 (range: 0.020-0.085), 0.044 (range: 0.020-0.125) and 0.033 5 (range: 0.030-0.040), respectively. Staging system of CRC mainly included LRD (localized/regional/distant, n=10), Dukes' (n=7) and TNM (n=3). When using the LRD classification, the median annual transition rates from ' localized' to ' regional' and ' regional' to 'distant' were 0.28 (range: 0.20-0.33) and 0.40 (range: 0.24-0.63), respectively. Under the Dukes' classification, the median annual transition rates appeared as 0.583 (range: 0.050-0.910), 0.656 (range: 0.280-0.720) and 0.830 (range: 0.630-0.865) from Dukes' A to B, B to C and C to Dukes' D, respectively. Again, when using the TNM classification, very limited transition rate was reported. Serrated pathway was only described in one study. Conclusions: Studies on the natural history model of colorectal cancer was still limited worldwide. Adenoma seemed the most common status setting for precancer model, and the risk-dependent classification for adenoma was consistent with the most commonly used system in clinical practice as well as major cancer screening programs in China. Since the staging systems of cancers varied, and shortage of transition rates based on TNM classification (commonly used in China), there will be a challenge for building Chinese population-specific natural history model of colorectal cancer, information from other classification systems could be conditionally applied.目的：系统了解全球结直肠癌疾病自然史模型研究，为构建我国人群结直肠癌疾病自然史模型及开展相关干预方案提供参考。 方法：检索PubMed 1995-2015年结直肠癌自然史模型研究，摘录汇总癌前病变及癌症分期的分类系统及对应转归概率，指标主要为1年进展或消退概率的中位数（M）值。 结果：最终纳入24篇文献，其中多数（22篇）采用马尔可夫模型。腺瘤分类系统包括按风险高低（9篇）和腺瘤大小（13篇，细化为两种）。（1）基于风险分类系统的研究显示，从健康发展为低风险腺瘤其1年概率的M＝0.016 0 (0.002 2~ 0.020 0），低风险进展为高风险腺瘤和高风险进展为结直肠癌概率的M值分别为0.020 （0.002~ 0.177）和0.044 (0.005～0.063）。（2) 7篇以腺瘤10 mm为界的模型文献提示，由"健康"发展为<10 mm腺瘤的1年概率M＝0.016 7 （0.015 0~ 0.037 0），<10 mm腺瘤发展为≥10 mm的概率M＝0.020 (0.015~ 0.035）。（3) 6篇以腺瘤≤5、6~ 9及≥10 mm（微小、小及大）为分界的文献中，由"健康"发展为微小腺瘤概率的M＝0.013 （0.009~ 0.019），微小腺瘤成为小腺瘤和小腺瘤成为大腺瘤概率的M值分别为0.043 （0.020～ 0.085）和0.044 （0.020~ 0.125）。结直肠癌分期系统主要包括癌灶范围分类（LRD, 10篇）和Dukes’分期（7篇），TNM分期研究仅3篇，其参数更有限。个别文献提供了"锯齿状腺瘤路径"及其参数。 结论：目前全球结直肠癌疾病自然史模型研究文献有限，且多将"腺瘤"设置为癌前病变类型，而按"腺瘤风险"的分类与我国临床及大型癌症筛查项目一致，文献常见的癌症分期系统较难与我国主要使用的TNM系统数据对接，后期参数确定还需结合结直肠癌其他分期系统进行转换。.