通过粪便隐血试验进行结直肠癌筛查的有效性与安全性：系统评价和荟萃分析。
全文（开放获取） · 中文译文
摘要 目的：本系统综述评估了针对45-75岁成人采用粪便隐血试验（gFOBT或FIT）的人群结直肠癌（CRC）筛查项目的有效性。 方法：我们纳入了将人群结直肠癌筛查项目参与者与未筛查个体进行比较的研究。结局指标为全因死亡率、结直肠癌特异性死亡率、结直肠癌发病率和IV期结直肠癌发病率。两名 reviewers 独立进行研究筛选、数据提取和偏倚风险评估。通过 meta 分析计算合并 RR，并使用 GRADE 评估证据确定性。 结果：全因死亡率未见差异（RR = 1.00，95% CI 0.99-1.01；5项研究，1,474,576名参与者；中等确定性证据）。评估 gFOBT 的 RCT 显示结直肠癌特异性死亡率降低（RR = 0.89，95% CI 0.84-0.94；5项研究，1,474,576名参与者；中等确定性证据），相当于每100,000名接受筛查的个体中结直肠癌死亡减少38例（95% CI，减少55例至减少21例）。评估 FIT 的非 RCT 报告结直肠癌特异性死亡率降幅更大（RR = 0.21，95% CI 0.20-0.21；5项研究，7,435,956名参与者），但证据确定性较低。RCT 显示对结直肠癌发病率或IV期结直肠癌发病率无显著影响，而评估 FIT 的非 RCT 提示这两种结局均有所降低；然而，证据确定性同样较低。 结论：基于粪便隐血试验的筛查可降低结直肠癌特异性死亡率，但不能降低全因死亡率。基于FIT的筛查可能通过降低结直肠癌发病率和IV期疾病发生率而带来额外获益；然而，这些发现得到低确定性证据支持，尚需在随机对照试验（RCTs）中加以证实。 引言 结直肠癌（CRC）是全球第三大常见恶性肿瘤，约占所有癌症病例的10%，同时也是全球癌症相关死亡的第二大原因[]。2022年，全球估计新发CRC病例190万例，死亡超过90万例，凸显其巨大的全球疾病负担。CRC主要影响年龄≥50岁的人群，但年轻成人中的发病率正在上升[]。 CRC特别适合筛查，因为它通常经历一个漫长的临床前期阶段，往往从腺瘤性息肉和其他癌前病变进展为浸润性癌[]。检出并切除癌前病变可预防CRC，而在更早阶段确诊可提高治愈性治疗的可能性并降低发病率和死亡率[]。分期越低，治愈性治疗的可能性越大。相比之下，IV期结直肠癌的5年生存率约为13–18%[]。 结直肠癌（CRC）筛查可采用多种方法，包括结肠镜检查、软式乙状结肠镜检查、CT结肠成像、粪便DNA检测和粪便隐血检测[]。其中，结肠镜检查被认为是参考标准，既可用于筛查，也可用于诊断，而无创检查的阳性结果通常需要结肠镜检查来确认。结肠镜检查具有极佳的准确性，灵敏度为95%，特异度为98%；然而，在幅员辽阔、人口众多且经济尚不发达的国家，结肠镜检查难以实施[]。相比之下，粪便隐血检测价格低廉、可及性广、无创，更适合大规模人群筛查项目。 检测粪便隐血有两种可用的方法：愈创木脂法（gFOBT）和免疫化学法（FIT）[]。前者是一种比色法，基于存在于血红蛋白中的过氧化物酶所催化的愈创木脂氧化反应，产生蓝色。FIT使用针对人血红蛋白或其他血液成分的特异性抗体。与gFOBT法相比，FIT的主要优点包括：检查前无需准备（无需饮食和用药限制）；对上消化道来源的出血不敏感（因为它仅与未降解的珠蛋白反应）；并且每克粪便中血红蛋白的检出限更好[]。 基于粪便的结直肠癌筛查随机试验主要评估了gFOBT，并确定了其可降低结直肠癌特异性死亡率[–]。然而，在当代筛查项目中，FIT已在很大程度上取代了gFOBT，因为FIT具有实用和分析方面的优势，而关于其对死亡率影响的证据主要来自非随机的人群研究[–]。因此，目前支持基于粪便筛查的证据包括两个不同但互补的证据体系：确立基于gFOBT筛查死亡率获益的随机证据，以及评估FIT在当代实践中有效性的较新的观察性证据。 因此，本系统评价旨在对针对45–75岁成人使用粪便隐血试验（gFOBT或FIT）的人群结直肠癌筛查项目的效果提供最新且全面的评估。 方法 本系统评价按照Cochrane协作网的推荐进行[]，并根据PRISMA（系统评价和荟萃分析优先报告条目）声明[]进行报告。其方案已在国际前瞻性系统评价注册库（PROSPERO）注册——注册号CRD42020169541。 纳入标准 本系统综述纳入随机对照研究（RCTs），其中患者被随机分为两组（有干预组和无干预组）。由于缺乏通过FIT方法筛查评估死亡率结局的RCTs，因此纳入了使用FIT方法并评估死亡率的前瞻性人群队列研究。纳入的研究遵循下述“PICO”结构。 研究对象（P） 研究对象被定义为年龄在45至75岁之间、具有结直肠癌（CRC）标准风险、对该癌症无症状并接受了该肿瘤人群筛查的成年个体。“标准风险”这一术语适用于无CRC易感倾向的人群。 干预类型（I） 干预组个体通过gFOBT或FIT进行粪便隐血试验。 对照（C） 对照组个体被定义为未参与CRC人群筛查者。 结局指标（O） 全因死亡率和CRC特异性死亡率被视作主要结局；其他结局包括CRC发病率及其IV期（远处转移）发病率、伴或不伴息肉切除术的结肠镜检查相关不良事件（出血或肠穿孔）、生活质量、五年内生存情况以及化疗和放疗需求。 排除标准 我们排除了筛查方法不是gFOBT或FIT的研究、研究对象为有症状患者或结直肠癌（CRC）发生高风险个体的研究、非基于人群的筛查研究、未将死亡率作为结局指标的研究，以及评估gFOBT的非随机对照试验（non-RCTs）。 研究识别——电子数据库 在MEDLINE（通过PubMed）、Embase、LILACS和Cochrane对照试验中心注册库（CENTRAL）中进行了全面的文献检索。检索策略结合了受控词汇（如MeSH和Emtree词）以及与结直肠癌、结直肠肿瘤、粪便隐血试验、愈创木脂粪便隐血试验、结直肠癌筛查和队列研究相关的自由文本关键词。酌情使用运算符（“AND”和“OR”）组合检索词。对于MEDLINE和Embase，应用了Cochrane协作网推荐的用于识别随机对照试验（RCT）的验证过的检索过滤器。此外，还将与队列研究相关的MeSH和Emtree词纳入检索策略，以捕获符合条件的观察性研究。对语言、出版日期或出版状态未作任何限制。初次检索于2019年10月18日进行，并于2024年4月19日更新。所有数据库的详细检索策略见附录。 使用EndNote X9文献管理软件下载参考文献并删除重复条目。通过免费网络应用程序Rayyan QCRI []进行摘要和标题的初步筛选。 文献筛选 标题和摘要由成对的审稿人独立进行双人双筛，以确定潜在符合条件的研究。三名审稿人（LTS、SNS 和 VSNN）筛查了截至 2019 年发表的研究，而两名审稿人（MCJF 和 GCP）筛查了 2019 年至 2024 年间发表的研究。入选进行全文审查的研究随后根据预先定义的 PICO 框架评估其一致性。分歧通过讨论解决，并在审稿人之间达成共识。 数据提取与管理 三名审稿人（LTS、MCJF 和 GCP）使用标准表格从入选研究中提取以下数据：发表年份、国家、随机或非随机对照试验、筛查频率、干预和对照类型、随机分配至干预组和对照组的患者人数、主要结局以及参与者的平均年龄。 为确保审稿人之间的一致性，我们在开始审查前进行了校准练习。如果原始研究存在重复发表或多篇报告，则利用同一研究中所有项目可获得的最佳信息来优化数据提取。 纳入研究的偏倚风险评估 使用修订版 Cochrane 偏倚风险评估工具（RoB 2）[]评估偏倚风险，该工具评估五个领域：随机化、偏离预期干预措施、结局数据缺失、结局测量和选择性报告。每个领域均由两名审稿人独立评估为低偏倚风险、存在一定担忧或高偏倚风险。 对于非随机对照试验（non-RCTs），使用ROBINS-I工具 [] 评估偏倚风险，该工具评估七个领域：混杂、参与者选择、干预分类、偏离预期干预、缺失数据、结局测量和选择性报告。每个领域被评定为低、中、严重或关键偏倚风险，随后给出总体偏倚风险判断。 数据分析 使用Stata统计软件（版本16）（Stata Corp LLC，美国德克萨斯州大学城）进行Meta分析。主要效应指标为结直肠癌发病率、结直肠癌死亡率、全因死亡率和不良事件的相对危险度（RRs），并附有相应的95%置信区间（CIs）。采用DerSimonian–Laird随机效应模型计算合并效应估计值，以考虑各研究间潜在的临床和方法学异质性。 亚组分析 按照研究设计将研究分为随机对照试验或非随机对照试验研究。对每个亚组分别进行Meta分析。 敏感性分析 采用留一法敏感性分析，通过依次从Meta分析中排除每项研究并重新计算合并效应估计值来进行。此外，使用固定效应模型进行了敏感性分析，并生成相应的敏感性分析图，以评估结果的稳健性和稳定性。 异质性分析 通过目视检查森林图（各研究效应估计值的置信区间无重叠）以及Higgins或I2统计量（其中I2 > 50%表示存在中等程度的异质性可能性），并使用卡方检验（Chi2）（其中p < 0.10表示存在异质性）[]来确定所纳入研究结果之间的不一致性。 发表偏倚 按照方法学指南的建议 []，当某项荟萃分析至少有10项研究可用时，计划评估发表偏倚。在这种情况下，将目视检查漏斗图的对称性，并使用Egger回归检验进行正式评价。 证据质量 根据推荐分级的评估、制定与评价（GRADE）指南 [] 评估效应量证据的质量。 结果 检索策略在去除重复后识别出4,135条记录。其中，37项研究被选入全文审查（图）。最终，10项研究符合纳入标准并被纳入本综述，共涉及14篇出版物，因为部分研究在不止一篇出版物中进行了报告 [–, –]。 图1 PRISMA研究选择流程流程图 共有23项研究因以下原因被排除：与已纳入研究数据重叠（n = 12）；非随机对照试验（RCT）筛查（n = 4）；未报告死亡率数据（n = 2）；病例对照设计（n = 2）；无原始数据的综述文章（n = 1）；不良事件未按研究组报告（n = 1）；以及非基于人群的筛查（n = 1）。排除研究列表及其相应参考文献见附录。 研究特征 在纳入的10项研究中，5项为随机对照试验，采用粪便潜血试验（gFOBT）作为干预措施，以不筛查作为对照。大多数研究采用双年度的gFOBT筛查频率，仅有一项研究使用了不同的筛查间隔（干预组参与者被邀请参加首轮筛查及一轮或两轮复筛）[]。其余5项研究为前瞻性队列研究，评估基于粪便免疫化学试验（FIT）的筛查与不筛查相比的效果。FIT筛查主要采用双年度频率，筛查期约为4至11年不等，反映了随时间推移的重复筛查；一项研究仅评估了单轮筛查。 结直肠癌筛查在45至60岁之间启动，通常持续至69至80岁。所有纳入研究的主要特征，包括筛查频率和筛查期持续时间，汇总于表中。 表1 纳入研究的特征 380,651/ 327,100 203,670/ 175,778 3,067,853/ 2,349,846 431,190/ 430,755 34,144/ 34,164 15,587/ 15,394 准实验 262,893/ 634,919 Pitkaniemi/Paimela Koskenvuo 180,210/ V 76,466/ 76,384 6,961/ 26,285 FIT：粪便免疫化学试验；gFOBT：愈创木脂粪便隐血试验；RCT：随机对照试验；CRC：结直肠癌；CRC-M：结直肠癌特异性死亡率；CRC-F：结直肠癌频率（发病率）；CRC-IV：IV期结直肠癌频率；All-M：全因死亡率 偏倚风险 在RoB 2工具的任何一个领域，没有RCT被判定为高偏倚风险。总体而言，大多数随机试验在评估的各领域被评估为低偏倚风险或存在一定问题，其详细评估结果见图。根据ROBINS-I工具，四项非RCT被判定为严重偏倚风险，一项为中等偏倚风险，详细评估见。主要问题涉及残余混杂、筛查参与的自选性、健康志愿者偏倚以及比较组之间的污染。相比之下，与结局测量、缺失数据和选择性报告相关的偏倚风险在各研究中普遍较低。 图2 根据RoB 2工具对纳入的随机对照试验进行偏倚风险评估 荟萃分析 关于全因死亡率，没有证据表明结直肠癌筛查降低了死亡风险（RR：1.0，95% CI：0.99–1.01；5项RCT，1,474,576名受试者；证据确定性中等；图；表）。 图3 结直肠癌筛查与不筛查相比对全因死亡率影响的森林图。根据Cochrane偏倚风险2（RoB 2）工具，Low = 低偏倚风险；SC = 有一定问题。注：权重来自随机效应模型 表2 粪便潜血试验筛查结直肠癌的结果总结以及使用GRADE评估的证据确定性 RR 1.00 (0.99 至 1.01) 每100.000例减少0例 （从减少111例到增加111例） ⨁⨁⨁◯ 中等 RR 0.89 (0.84 至 0.94) 每100.000例减少38例 （从减少55例到减少21例） ⨁⨁⨁◯ 中等 RR 0.21 (0.20 至 0.21) 每100.000例减少405例 （从减少410例到减少405例） ⨁⨁◯◯ 低 CI：置信区间；RR：风险比。中等 = 真实效应可能接近估计值，但也可能存在实质性差异。低 = 真实效应可能与估计值存在实质性差异 解释 a. 所有研究在偏离预期干预措施方面被归类为有一定问题 b. 主要问题包括残留混杂、自我选择偏倚以及比较组之间的污染 在结直肠癌（CRC）死亡率方面，评估愈创木脂粪便隐血试验（gFOBT）的随机对照试验（RCTs）报告的合并相对风险（RR）为0.89（95% CI：0.84–0.94；5项研究，1,474,576名受试者；中等确定性证据），相当于每100,000名筛查个体中结直肠癌死亡减少38例（95% CI：减少55例至减少21例；图；表）。这些随机对照试验的留一法敏感性分析显示，没有单项研究实质性影响合并效应估计值，。与这些发现一致，采用固定效应模型和随机效应模型的敏感性分析产生了相似的结果，。 相比之下，在非随机对照试验中，不同模型的合并估计值存在差异，因为Chiu等人（Chiu et al.）的研究贡献了总研究人群的约73%。鉴于该研究的主导地位，固定效应模型被认为是最合适的，因此被用于主要分析。评估FIT筛查的非随机对照试验报告合并RR为0.21（95% CI：0.20–0.21；5项研究；7,435,956名受试者；低确定性证据），相当于每筛查100,000人中结直肠癌（CRC）死亡减少405例（减少410例至405例；Fig. ; Table ）。当应用随机效应模型时，合并效应估计值减弱，且不再具有统计学显著性（RR 0.47，95% CI 0.20–1.13）。 对于随机对照试验，与不筛查相比，CRC筛查与总体CRC发病率的显著降低无关联（RR：0.98，95% CI：0.91–1.06；5项研究，1,474,576名受试者，中等确定性证据 Fig. ）。相比之下，在评估FIT筛查的非随机对照试验中，主要固定效应分析（很大程度上受Chiu et al.研究的影响，该研究占总体研究人群约73%）表明，筛查个体的CRC发病率显著降低（RR：0.34，95% CI：0.33–0.35；6,158,696名受试者）。 图5 结直肠癌筛查与筛查人群和未筛查人群总体结直肠癌发病率之间关联的森林图。注：权重和亚组间异质性检验来自Mantel-Haenszel模型。 关于分期特异性结局，与不筛查相比，在随机对照试验（RCTs）中评估的粪便隐血试验（gFOBT）筛查未与IV期结直肠癌（CRC）发病率的显著降低相关（RR: 0.88, 95% CI: 0.77–1.02; 4项研究,
全文（开放获取） · 英文原文
【Abstract】 PURPOSE: This systematic review evaluated the effectiveness of population-based colorectal cancer (CRC) screening programs using fecal occult blood tests (gFOBT or FIT) among adults aged 45-75 years. METHODS: We included studies comparing participants in population-based CRC screening programs with unscreened individuals. Outcomes were all-cause mortality, CRC-specific mortality, CRC incidence, and stage IV CRC incidence. Two reviewers independently performed study selection, data extraction, and risk-of-bias assessment. Pooled RRs were calculated by meta-analysis, and certainty of evidence was assessed using GRADE. RESULTS: No difference was observed in all-cause mortality (RR = 1.00, 95% CI 0.99-1.01; 5 studies, 1,474,576 participants; moderate-certainty evidence). RCTs evaluating gFOBT demonstrated a reduction in CRC-specific mortality (RR = 0.89, 95% CI 0.84-0.94; 5 studies, 1,474,576 participants; moderate-certainty evidence), corresponding to 38 fewer CRC deaths per 100,000 individuals screened (95% CI, 55 fewer to 21 fewer). Non-RCTs evaluating FIT reported larger reductions in CRC-specific mortality (RR = 0.21, 95% CI 0.20-0.21; 5 studies, 7,435,956 participants), although the certainty of evidence was low. RCTs showed no significant effect on CRC incidence or stage IV CRC incidence, whereas non-RCTs evaluating FIT suggested reductions in both outcomes; however, the certainty of evidence was also low. CONCLUSION: Fecal occult blood test-based screening reduces CRC-specific mortality but not all-cause mortality. FIT-based screening may provide additional benefits by reducing CRC incidence and stage IV disease; however, these findings are supported by low-certainty evidence and require confirmation in RCTs. 【Introduction】 Colorectal cancer (CRC) is the third most commonly diagnosed cancer worldwide, accounting for approximately 10% of all cancer cases, and the second leading cause of cancer-related death globally []. In 2022, an estimated 1.9 million new cases of CRC and more than 900,000 deaths occurred worldwide, highlighting its substantial global burden. It predominantly affects individuals aged ≥ 50 years, although incidence is increasing among younger adults []. CRC is particularly amenable to screening because it commonly develops through a prolonged preclinical phase, often progressing from adenomatous polyps and other precursor lesions to invasive cancer []. Detection and removal of premalignant lesions can prevent CRC, while diagnosis at an earlier stage increases the likelihood of curative treatment and reduces morbidity and mortality []. The lower the stage is, the greater the likelihood of curative treatment. In contrast, Stage IV colorectal cancer is associated with a 5-year survival rate of approximately 13–18% []. Screening for CRC can be performed using several methods, including colonoscopy, flexible sigmoidoscopy, CT colonography, stool DNA testing, and fecal occult blood testing []. Among these, colonoscopy is considered the reference standard and can be used as both a screening and diagnostic procedure, whereas positive results from noninvasive tests generally require confirmation by colonoscopy. Colonoscopy has excellent accuracy, with 95% sensitivity and 98% specificity; however, it is difficult to implement in countries with continental dimensions, a large population, and a developing economy []. In contrast, fecal occult blood testing is inexpensive, widely accessible, noninvasive, and more feasible for large-scale population-based screening programs. There are two methods available to search for fecal occult blood: guaiac (gFOBT) and the immunochemical method (FIT) []. The first is a colorimetric method, which is based on the oxidation of guaiac catalyzed by the action of peroxidase present in hemoglobin, resulting in a blue color. FIT uses specific antibodies against human hemoglobin or other blood components. The main advantages of FIT over the gFOBT method include the fact that there is no need for preparation before the exam (food and medication restrictions); it is not sensitive to bleeding originating from the upper gastrointestinal tract (because it only reacts with undegraded globin); and it has a better detection limit of hemoglobin per gram of feces []. Randomized trials of stool-based CRC screening have predominantly evaluated gFOBT and have established a reduction in CRC-specific mortality [–]. However, FIT has largely replaced gFOBT in contemporary screening programs because of its practical and analytical advantages, while evidence regarding its effect on mortality derives predominantly from non-randomized population-based studies [–]. Consequently, the evidence supporting stool-based screening currently comprises two distinct but complementary bodies of evidence: randomized evidence establishing the mortality benefit of gFOBT-based screening and more recent observational evidence evaluating the effectiveness of FIT in contemporary practice. Therefore, this systematic review aimed to provide an updated and comprehensive assessment of the effects of population-based CRC screening programs using fecal occult blood tests (gFOBT or FIT) among adults aged 45–75 years. 【Methods】 This systematic review was performed according to the recommendations of the Cochrane Collaboration [] and reported according to the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analysis) Statement []. Its protocol was registered on the International Prospective Register of Systematic Reviews (PROSPERO) - Registration number CRD42020169541. 【Eligibility Criteria】 For this systematic review, randomized controlled studies (RCTs) in which patients were randomly allocated into two groups (with and without intervention) were included. Owing to the lack of RCTs that evaluated mortality outcomes with screening via the FIT method, prospective population cohort studies that used the FIT method and evaluated mortality were included. The included studies followed the “PICO” structure described below. 【Participants (P)】 Participants were defined as adult individuals aged between 45 and 75 years, with a standard risk for CRC, who were asymptomatic for this cancer, and who underwent population screening for this neoplasm. The term “standard risk” applies to people who are not predisposed to CRC. 【Types of Interventions (I)】 Individuals in the intervention group underwent fecal occult blood tests via gFOBT or FIT. 【Comparison (C)】 Individuals in the control group were defined as those who did not participate in population screening for CRC. 【Outcomes (O)】 All-cause mortality and CRC-specific mortality were considered main outcomes, and the other outcomes included the incidence of CRC and its incidence at stage IV (distant metastases), adverse events related to colonoscopy with and without polypectomy (bleeding or intestinal perforation), quality of life, survival within five years and the need for chemotherapy and radiotherapy. 【Exclusion Criteria】 We excluded studies whose screening method was not gFOBT or FIT, studies whose population corresponded to symptomatic patients or individuals at high risk for CRC development, non-population-based screening studies, studies that did not assess mortality as an outcome, and non-RCTs evaluating gFOBT. 【Identification of Studies - Electronic Databases】 A comprehensive literature search was conducted in MEDLINE (via PubMed), Embase, LILACS, and the Cochrane Central Register of Controlled Trials (CENTRAL). The search strategy combined controlled vocabulary terms (e.g., MeSH and Emtree terms) and free-text keywords related to colorectal cancer, colorectal neoplasms, fecal occult blood testing, guaiac fecal occult blood testing, colorectal cancer screening, and cohort studies. Operators (“AND” and “OR”) were used to combine search terms as appropriate. For MEDLINE and Embase, validated search filters recommended by the Cochrane Collaboration for identifying RCTs were applied. Additionally, MeSH and Emtree terms related to cohort studies were incorporated into the search strategy to capture eligible observational studies. No restrictions were imposed on language, publication date, or publication status. The initial search was performed on October 18, 2019, and updated on April 19, 2024. Detailed search strategies for all databases are provided in . EndNote X9 citation management software was used to download references and remove duplicate entries. The initial screening of abstracts and titles was performed via the free web application Rayyan QCRI []. 【Study Selection】 Titles and abstracts were independently screened in duplicate by pairs of reviewers to determine potential eligibility. Three reviewers (LTS, SNS and VSNN) screened studies published up to 2019, while two reviewers (MCJF and GCP) screened studies published between 2019 and 2024. Studies selected for full-text review were subsequently assessed for consistency with the predefined PICO framework. Disagreements were resolved through discussion, with consensus reached among the reviewers. 【Data Extraction and Management】 The three reviewers (LTS, MCJF and GCP) used a standard form to extract the following data from the selected studies: year of publication, country, randomized or non-RCT, screening frequency, type of intervention and comparison, number of patients randomized to the intervention and comparison groups, main outcomes and mean age of participants. To ensure consistency between reviewers, we performed a calibration exercise before beginning the review. In the case of duplicate publications or multiple reports from the primary study, data extraction was optimized using the best information available for all items in the same study. 【Assessment of Risk of Bias in Included Studies】 Risk of bias was assessed using the revised Cochrane Risk of Bias tool (RoB 2) [], which evaluates five domains: randomization, deviations from intended interventions, missing outcome data, outcome measurement, and selective reporting. Each domain was independently assessed by two reviewers as having low risk of bias, some concerns, or high risk of bias. For non-RCTs, risk of bias was assessed using the ROBINS-I tool [], which evaluates seven domains: confounding, participant selection, intervention classification, deviations from intended interventions, missing data, outcome measurement, and selective reporting. Each domain was rated as low, moderate, serious, or critical risk of bias, and an overall risk-of-bias judgment was subsequently assigned. 【Data Analysis】 Meta-analyses were conducted using Stata Statistical Software, version 16 (Stata Corp LLC, College Station, TX, USA). The primary effect measures were the relative risks (RRs) of colorectal cancer incidence, colorectal cancer mortality, all-cause mortality, and adverse events, with corresponding 95% confidence intervals (CIs). Pooled effect estimates were calculated using the DerSimonian–Laird random-effects model to account for potential clinical and methodological heterogeneity across studies. 【Subgroup Analysis】 The studies were divided according to their design into RCT or non-RCT studies. The meta-analyses were performed separately for each subgroup. 【Sensitivity Analysis】 A leave-one-out sensitivity analysis was performed by sequentially excluding each study from the meta-analysis and recalculating the pooled effect estimate. Additionally, sensitivity analyses using a fixed-effects model were conducted, and the corresponding sensitivity analysis plots were generated to evaluate the robustness and stability of the findings. 【Heterogeneity Analysis】 Inconsistencies between the results of the studies included were ascertained by visual inspection of forest plots (no overlap of CIs around the effect estimates of the individual studies) and by Higgins or I2 statistics, in which I2 > 50% indicated a moderate probability of heterogeneity, and by chi-square tests (Chi2), where p < 0.10 indicated heterogeneity []. 【Publication Bias】 Publication bias was planned to be assessed when at least 10 studies were available for a given meta-analysis, as recommended by methodological guidelines []. In such cases, funnel plot asymmetry would be visually inspected and formally evaluated using Egger’s regression test. 【Quality of Evidence】 The quality of the evidence of the effect size was assessed according to the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) guidelines []. 【Results】 The search strategy identified 4,135 records after duplicate removal. Of these, 37 studies were selected for full-text review (Fig. ). Ultimately, 10 studies met the eligibility criteria and were included in the review, comprising a total of 14 publications, as some studies were reported in more than one publication [–, –]. Fig. 1PRISMA flow diagram of the study selection process A total of 23 studies were excluded for the following reasons: overlapping data with an included study (n = 12); non-RCT screening (n = 4); no mortality data reported (n = 2); case–control design (n = 2); review article without primary data (n = 1); adverse events not reported by study group (n = 1); and non-population-based screening (n = 1). The list of excluded studies and their corresponding references are provided in the . 【Study Characteristics】 Among the 10 studies included, five were RCTs that used gFOBT as an intervention and no screening as a comparator. Most studies applied gFOBT with biennial frequency, and only one study used different screening intervals (intervention group participants were invited to the first screening round and to one or two rescreening rounds) []. The five remaining studies were prospective cohort studies evaluating FIT-based screening compared with no screening. FIT was predominantly performed biennially, with screening periods ranging from approximately 4 to 11 years, reflecting repeated screening over time; one study evaluated a single screening round. CRC screening was initiated between 45 and 60 years of age and generally continued until 69 to 80 years of age. The main characteristics of all included studies, including screening frequency and duration of the screening period, are summarized in Table . Table 1Characteristics of the included studies 380,651/ 327,100 203,670/ 175,778 3,067,853/ 2,349, 846 431,190/ 430,755 34,144/ 34,164 15,587/ 15,394 Quasi- experimental 262,893/ 634,919 Pitkaniemi/Paimela Koskenvuo 180,210/ V 76,466/ 76,384 6,961/ 26,285 FIT: fecal immunochemical test; gFOBT: guaiac-based fecal occult blood test; RCT: randomized controlled trial; CRC: colorectal cancer; CRC-M: colorectal cancer–specific mortality; CRC-F: frequency (incidence) of colorectal cancer; CRC-IV: frequency of stage IV colorectal cancer; All-M: all-cause mortality 【Risk of Bias】 No RCT was judged to be at high risk of bias in any domain of the RoB 2 tool. Overall, most randomized trials were assessed as having low risk of bias or some concerns across the evaluated domains, and their detailed assessments are presented in Fig. . According to the ROBINS-I tool, four non-RCTs were judged to be at serious risk of bias and one at moderate risk of bias, with detailed evaluations provided in the . The main concerns involved residual confounding, self-selection into screening participation, healthy-volunteer bias, and contamination between comparison groups. In contrast, the risk of bias related to outcome measurement, missing data, and selective reporting was generally low across studies. Fig. 2Risk of bias assessment of the included randomized controlled trials according to the RoB 2 tool 【Meta-Analysis】 Regarding all-cause mortality, there was no evidence that CRC screening reduced mortality risk (RR: 1.0, 95% CI: 0.99–1.01; 5 RCTs, 1,474,576 participants; moderate certainty of evidence; Fig. ; Table ). Fig. 3Forest plot of the effect of colorectal cancer screening on all-cause mortality compared with no screening. Low = low risk of bias; SC = some concerns, according to the Cochrane Risk of Bias 2 (RoB 2) tool. Note: Weights are from random-effects model Table 2Summary of findings for colorectal cancer screening with fecal occult blood testing and certainty of evidence assessed using GRADE RR 1.00 (0.99 to 1.01) 0 fewer per 100.000 (from 111 fewer to 111 more) ⨁⨁⨁◯ Moderate RR 0.89 (0.84 to 0.94) 38 fewer per 100.000 (from 55 fewer to 21 fewer) ⨁⨁⨁◯ Moderate RR 0.21 (0.20 to 0.21) 405 fewer per 100.000 (from 410 fewer to 405 fewer) ⨁⨁◯◯ Low CI: confidence interval; RR: risk ratio. Moderate = the true effect is likely close to the estimate but may be substantially different. Low = the true effect may be substantially different from the estimate Explanations a. All studies were classified as some concerns on deviations from intended interventions b. The main concerns were residual confounding, self-selection bias, and contamination between comparison groups In terms of CRC mortality, RCTs evaluating gFOBT reported a pooled RR of 0.89 (95% CI: 0.84–0.94; 5 studies, 1,474,576 participants; moderate-certainty evidence), corresponding to 38 fewer CRC deaths per 100,000 individuals screened (95% CI: 55 fewer to 21 fewer; Fig. ; Table ). Leave-one-out sensitivity analyses of the RCTs showed that no single study materially influenced the pooled effect estimate, . Consistent with these findings, sensitivity analyses using fixed- and random-effects models yielded similar results, . Fig. 4Forest plot showing the effect of colorectal cancer (CRC) screening on CRC-specific mortality compared with no screening. Low = low risk of bias; SC = some concerns, according to the Cochrane Risk of Bias 2 (RoB 2) tool. Results from the random-effects model, as well as the leave-one-out sensitivity analyses, are presented in the . Note: Weights and between-subgroup heterogeneity test are from Mantel-Haenszel model In contrast, among the non-RCTs, pooled estimates differed between models because the study by Chiu et al. contributed approximately 73% of the total study population. Given the predominance of this study, the fixed-effect model was considered the most appropriate and was therefore used for the primary analysis. Non-RCTs evaluating FIT screening reported a pooled RR of 0.21 (95% CI: 0.20–0.21; 5 studies; 7,435,956 participants; low-certainty evidence), corresponding to 405 fewer CRC deaths per 100,000 individuals screened (410 fewer to 405 fewer; Fig. ; Table ). When random-effects model was applied, the pooled effect estimate was attenuated and no longer statistically significant (RR 0.47, 95% CI 0.20–1.13). For RCTs, CRC screening was not associated with a significant reduction in overall CRC incidence compared with no screening (RR: 0.98, 95% CI: 0.91–1.06; 5 studies, 1,474,576 participants, moderate-certainty evidence Fig. ). In contrast, among non-RCTs evaluating FIT screening, the primary fixed-effect analysis (largely influenced by the study by Chiu et al., which accounted for approximately 73% of the total study population) demonstrated a substantially lower incidence of CRC among screened individuals (RR: 0.34, 95% CI: 0.33–0.35; 6,158,696 participants). Fig. 5Forest plot of the association between colorectal cancer screening and overall colorectal cancer incidence in screened versus unscreened populations. Note: Weights and between-subgroup heterogeneity test are from Mantel-Haenszel model Regarding stage-specific outcomes, gFOBT screening evaluated in RCTs was not associated with a significant reduction in the incidence of stage IV CRC compared with no screening (RR: 0.88, 95% CI: 0.77–1.02; 4 studies,