超早期内镜检查是否总是有益？一项关于非静脉曲张性上消化道出血中时间依赖性结局的系统评价和荟萃分析。
全文（开放获取） · 中文译文
摘要 背景：非静脉曲张性上消化道出血（NVUGIB）是一种常见的急症，具有显著的发病率和死亡率。指南推荐在24小时内进行内镜检查，但超早期内镜检查（≤6小时）的获益仍不明确，尤其是在高危患者中。本研究旨在系统评价与早期内镜检查（>6小时）相比，超早期内镜检查（≤6小时）能否改善成人非静脉曲张性上消化道出血患者的临床结局，包括死亡率、再出血、医疗资源利用以及高危亚组的结局。 方法：开展了一项系统评价和荟萃分析，纳入比较成人非静脉曲张性上消化道出血患者接受超早期与早期内镜检查的研究。检索了PubMed、Embase和Web of Science数据库，检索时间截至2026年1月，以获取相关研究。超早期内镜检查定义为在就诊、入院或临床诊断后6小时内进行的内镜检查；高危患者根据研究层面的标准定义，最常用的标准为格拉斯哥-布拉奇福德评分>12或Rockall评分>5。使用随机效应模型计算合并比值比（OR）或均数差（MD）及其95%置信区间（CI），并采用预先设定的亚组分析和敏感性分析评估结果的稳健性。 结果：荟萃分析共纳入9项研究，包含10,785例患者（2,638例超早期，8,147例早期），并对其数据进行了分析。研究结果表明，超早期内镜检查并未显著降低死亡率（OR 0.88，95% CI [0.58-1.34]）或再出血率（OR 1.23，95% CI [0.90-1.67]），也未影响患者的输血、外科干预、重复内镜检查或住院时间。然而，超早期组入住重症监护病房（ICU）更为频繁（OR 1.41，95% CI [1.20-1.66]）。这些发现在高风险亚组患者中也是一致的。 结论：在患有非静脉曲张性上消化道出血（NVUGIB）的成人中，就关键临床结局而言，超早期内镜检查与早期内镜检查相比并未提供明确优势，尽管ICU入住率更高。这表明处理应侧重于在适当的治疗时间窗内提供充分的复苏和及时的内镜检查。需要进一步的前瞻性研究来评估超早期内镜检查在特定高危患者中的潜在获益。 引言 非静脉曲张性上消化道出血（NVUGIB）是临床实践中最常见的消化道急症之一，也是急诊内镜评估的主要适应证 (; )。NVUGIB 占急性上消化道出血事件的大多数，这些事件在全球范围内仍与显著的发病率和死亡率相关 ()。来自大型人群研究的估计表明，NVUGIB 患者的院内死亡率通常在约 5%–10% 之间，在老年患者和合并症负担显著的患者中观察到明显更高的病死率 (; )。 尽管包括质子泵抑制剂在内的辅助药物治疗、内镜止血技术的改进以及围急诊期整体救治路径的持续优化取得了实质性进展，NVUGIB 的临床影响仍然相当大，常需要住院、输血及大量医疗资源的使用 (; ; )。这些观察结果强调了持续优化 NVUGIB 管理策略以改善患者结局并减轻医疗系统负担的重要性 (; ; )。 内镜在非静脉曲张性上消化道出血（NVUGIB）的诊断、风险分层和治疗干预中起着核心作用（；）。当前国际指南推荐在患者病情稳定后24小时内进行内镜检查，以避免不必要的延误，同时允许充分的复苏治疗（；）。然而，这一24小时的时间窗相对宽泛，并未明确更早的干预是否能带来额外的临床获益（）。随着急诊内镜可及性和24小时服务可用性的提高，一些中心已开始探索提前内镜检查的时机，特别是在就诊或诊断后6小时内进行检查，这被视为超早期内镜（；）。 超早期内镜检查的理论依据是，及时识别和处理高危出血病变可能减少持续出血、再出血及相关并发症()。然而，迄今为止评估该策略的研究报告结果不一致：一些研究提示可能降低死亡率或再出血，而另一些研究则未发现明确获益，在某些情况下甚至增加了重症监护病房（ICU）入住或医疗资源利用(; )。这种差异可能反映不同研究中患者基线风险、复苏充分性、研究设计及残余混杂问题的差异。此外，“超早期内镜检查”的定义在各研究中有所不同，从严格的6小时截断值到更宽的早期时间间隔，使证据合成和解释复杂化。另外，高危患者亚组中的证据仍然有限且无定论，阻碍了就超早期内镜检查的常规应用制定明确指导建议()。鉴于这些不确定性，有必要开展一项更新的系统评价和荟萃分析，专门针对非静脉曲张性上消化道出血（NVUGIB）成人患者中≤6小时的时间阈值。因此，我们比较了超早期内镜检查（≤6小时）与早期内镜检查（>6小时），并评估了死亡率、再出血、医疗资源利用及高危亚组中的结局。本综合旨在明确在适当稳定后，常规加速内镜检查是否比符合指南的早期内镜检查提供额外的临床价值。 材料与方法 本系统综述和荟萃分析按照《Cochrane干预系统综述手册》进行，并遵循《系统综述和荟萃分析优先报告条目》（PRISMA）声明进行报告(; )。该研究方案在研究开始前已在PROSPERO数据库前瞻性注册（CRD420251273617）。 文献检索 在PubMed、Embase和Web of Science数据库中进行了系统性文献检索，检索时间从建库至2026年1月1日，旨在识别比较非静脉曲张性上消化道出血（NVUGIB）患者超早期与早期内镜检查的研究。检索策略结合了与NVUGIB、消化内镜及干预时机相关的医学主题词和自由词。核心检索词包括（“非静脉曲张性上消化道出血”或“NVUGIB”）AND（“消化内镜”或endoscop*或EGD）AND（“时间因素”或时机或早期或“≤6”或“6小时”或“6小时内”）。未对语言或发表状态施加限制。相关文章和既往系统综述的参考文献列表均经过人工筛选，以识别其他符合条件的研究。完整的检索策略见.。检索策略由Cai-Juan Liu和Logen Liu独立设计和实施。任何分歧均通过讨论解决；若无法达成共识，则由Guo-Qing Li担任仲裁并做出最终决定。 纳入标准 研究若纳入成年患者（≥18岁）患有非静脉曲张性上消化道出血（NVUGIB），并直接比较超早期内镜检查与早期内镜检查，且提供可提取的时机分组结局数据，则视为符合纳入标准。“超早期内镜检查”定义为在患者就诊、入院或临床诊断后6小时内进行的上消化道内镜检查，而“早期内镜检查”指在上述时间点超过6小时之后进行的操作，包括6–24小时等特定时间间隔，或在明确界定>6小时组时的更宽泛对照时间窗。高危状态（如有报告）按研究层面标准定义，通常为格拉斯哥-布拉奇福德出血评分（GBS）>12或Rockall评分>5。诊断性和治疗性内镜操作均被纳入。符合条件的研究设计包括随机对照试验以及前瞻性或回顾性观察性队列研究，且至少报告一项相关临床结局。 如果研究仅涉及食管或胃静脉曲张出血患者、包含无法分离成人数据的儿科人群，或设计为病例报告、病例系列、综述、述评、评论或缺乏足够可提取数据的会议摘要，则予以排除。未直接比较超早期内镜与早期内镜的研究也被排除。此外，如果根据预设的时间分组无法独立提取非静脉曲张性上消化道出血（NVUGIB）结局数据，则报告混合性静脉曲张和非静脉曲张出血的研究也被排除。 研究筛选与数据提取 研究筛选与数据提取由两名研究者使用预先制定的标准化数据提取表独立进行。首先对研究标题和摘要进行初步筛选以排除明显不相关的研究，随后对可能符合条件的研究进行全文审查。提取的数据包括研究特征、患者人口学特征和基线风险分层（如Rockall评分或GBS评分）、内镜干预的定义和时间安排、关注的结局指标，以及可获得的调整后效应估计值和调整模型中包含的变量。关注的结局包括死亡率、再出血、输血需求、住院时间、重症监护病房（ICU）入住以及需要外科手术或重复内镜干预。分歧通过讨论或咨询第三名研究者解决。 两名评价员独立评估了纳入研究的方法学质量和偏倚风险。随机对照试验采用 Cochrane 偏倚风险工具第 2 版（）进行评估，非随机研究采用非随机干预研究偏倚风险评估工具（ROBINS-I）（）进行评估。任何分歧均通过共识或由第三名评价员裁决解决。 统计分析 首先对纳入研究进行定性综合。当存在足够的临床和方法学同质性时，进行荟萃分析。对于二分类结局，计算合并比值比（OR）及其 95% 置信区间（CI）；连续结局则采用均数差（MD）进行合并。使用 Cochran Q 检验评估统计学异质性，并通过 I2 统计量进行量化，I2 值 0–25%、26%–50% 和 >50% 分别解释为低、中、高度异质性（）。鉴于研究设计、对照组评估时间窗、患者基线风险和结局定义方面预期存在临床异质性，对所有合并结局均采用随机效应模型作为主要分析方法。 根据对照时间窗、基线患者风险和死亡结局类型进行了预设亚组分析。通过依次排除单个研究进行敏感性分析，以评估合并估计值的稳健性。当个别研究报告了调整后的估计值时，优先描述这些值；然而，合并分析主要基于可提取的事件数，因为调整后的效应估计值并非在所有研究和结局中均一致可得。使用漏斗图和Egger回归检验评估发表偏倚 (; )。所有统计分析均使用Review Manager（RevMan）5.4版和Stata 14.0版进行，双侧P值<0.05被认为具有统计学显著性。 结果 基本特征与质量评估 系统检索从PubMed、Embase和Web of Science中识别出920条可能纳入考虑的记录。去除371条重复记录后，根据标题和摘要筛选了549项研究，其中526项因不相关或研究类型不合适而被排除。剩余23篇可能符合纳入条件的文章，并获取其全文以评估是否符合纳入标准。在全文评估后，排除了14项研究，主要原因是内镜检查时机未按6小时截断值进行分层、未报告相关结局，或研究人群包含静脉曲张性与非静脉曲张性出血的混合病例且无法分离数据。最终，9项研究符合纳入标准，被纳入系统评价和荟萃分析（；；；；；；；；）。研究筛选流程见中的PRISMA流程图。 PRISMA流程图展示了本系统评价和荟萃分析的研究筛选过程。 纳入研究的特征汇总于。九项研究发表于2004年至2025年，在美国、韩国和加拿大（各两项研究）以及中国、日本和西班牙（各一项研究）进行。纳入了随机对照试验和观察性队列研究。七项研究比较了超早期内镜检查（≤ 6 h）与在6–24 h内进行的内镜检查（; ; ; ; ; ; ），而两项研究采用了>6 h (6–48 h)的更宽比较窗口（; ）。总计纳入10,785例非静脉曲张性上消化道出血（NVUGIB）患者，其中超早期内镜组2,638例，早期内镜组8,147例。 纳入研究的特征。 注释。 RCS回顾性队列研究 随机对照试验 前瞻性观察队列 格拉斯哥–布拉奇福德评分 非静脉曲张性上消化道出血 未报告 小时 上标*a*表示总体NVUGIB人群。 上标*b*表示高风险NVUGIB亚组人群。 基线风险分层采用 GBS 或 Rockall 评分进行报告。三项研究提供了高危患者的单独数据(; ; )，其中两项研究将高危定义为 GBS > 12 (; )，一项研究采用 Rockall 评分 > 5 ()。七项研究报告了死亡人数(; ; ; ; ; ; )；两项研究未报告具体死亡人数(; )，一项研究报告两组均无死亡()。总体而言，超早期内镜组发生 90 例院内或 30 天死亡，早期内镜组发生 111 例，对应的粗死亡率分别为 3.67% 和 5.73%。 质量评价 使用 ROBINS-I 工具，七项非随机研究被判定为总体偏倚风险中等，一项为总体偏倚风险严重()，主要原因是观察性时机研究固有的残余混杂和选择偏倚。该随机对照试验使用 Cochrane 偏倚风险工具（RoB 2.0）进行评估，在随机化过程、结局数据缺失和结局测量领域被评为低风险，但在对预期干预的偏离和选择性报告方面存在一些担忧。 纳入研究的偏倚风险和方法学质量评价。 注。 对于 ROBINS-I：D1，混杂；D2，参与者选择；D3，干预分类；D4，对预期干预的偏离；D5，数据缺失；D6，结局测量；D7，报告结果的选择；L，低；M，中等；S，严重。 对于RoB 2.0，D1，随机化过程；D2，偏离预期干预措施；D3，结局数据缺失；D4，结局测量；D5，报告结果的选择；RCT，随机对照试验。 死亡率 共纳入九项研究，比较非静脉曲张性上消化道出血（NVUGIB）患者中超早期组与对照组的死亡率（；；；；；；；；）。采用随机效应模型，超早期组与对照组之间的总体死亡率未见显著差异（OR = 0.88，95% CI [0.58–1.34]；P = 0.56；），且存在显著异质性（I2 = 68%）。 NVUGIB患者死亡率的森林图。 （A）总体死亡率；（B）排除研究后的敏感性分析；（C）高危患者中的分析；（D）按内镜检查时机进行的亚组分析；（E）按死亡率结局类型进行的亚组分析。 排除权重最大的研究后的敏感性分析将合并估计值转向超早期组死亡率降低（OR = 0.78，95% CI [0.52–1.17]；），但仍无显著差异（P = 0.23），异质性降至I2 = 39%。 在高危患者中，超早期组与对照组之间的死亡率未检测到显著差异（合并OR = 0.92，95% CI [0.38–2.20]；P = 0.85；），且存在显著异质性（I2 = 72%）。 死亡率亚组分析 根据对照组的内镜检查时机和死亡率结局类型进行了亚组分析。 关于对照组的内镜检查时机，7项研究将超早期组与在6–24 h内接受内镜检查的对照组进行了比较(; ; ; ; ; ; )。合并分析显示两组间死亡率差异无统计学意义（OR = 0.98, 95% CI [0.65–1.48]; P = 0.93; ），且存在明显异质性（I2 = 68%）。相比之下，2项研究将超早期组与>6 h组进行了比较(; )。在该亚组分析中，发现超早期组与较低的死亡风险相关（OR = 0.38, 95% CI [0.15–0.97]; P = 0.04），异质性较低（I2 = 0%）。亚组差异检验未达到统计学显著性（P = 0.07）。 关于死亡结局类型，3项研究报告了院内死亡率(; ; )。合并分析显示，超早期组与对照组的院内死亡率差异无统计学意义（OR = 1.23, 95% CI [0.78–1.93]; P = 0.37; ），异质性较低（I2 = 0%）。相比之下，6项研究报告了30天死亡率(; ; ; ; ; )。合并估计值同样显示，与对照组相比，超早期组的30天死亡率没有显著降低（OR = 0.72, 95% CI [0.38–1.36]; P = 0.31），且存在明显异质性（I2 = 79%）。院内死亡率与30天死亡率之间的亚组差异检验未达到统计学显著性（P = 0.18）。 再出血 共有七项研究比较了非静脉曲张性上消化道出血（NVUGIB）患者中超早期组与对照组的再出血情况（；；；；；；）。使用随机效应模型，与对照组相比，超早期内镜检查与再出血风险的显著变化无关（OR = 1.23，95% CI [0.90–1.67]；P = 0.19；），存在显著异质性（I2 = 55%）。 NVUGIB患者再出血的森林图。 （A）总体再出血；（B）排除研究（；）的敏感性分析；（C）高危患者中的分析；（D）根据内镜检查时机进行的亚组分析。 排除权重最大（31.8%）的研究（；）后的敏感性分析显示，对再出血的影响相似且无显著差异（OR = 1.24，95% CI [0.76–2.03]；P = 0.39；），异质性增加至I2 = 62%。 在高危患者中，两组之间的再出血未检测到显著差异（合并OR = 1.29，95% CI [0.95–1.75]；P = 0.10；），异质性较低（I2 = 0%）。 再出血的亚组分析。 根据对照组内镜时机进行了亚组分析。五项研究将超早期组与6–24小时内行内镜检查的对照组进行了比较（；；；；），合并分析显示两组再出血差异无统计学意义（OR = 1.17，95% CI [0.76–1.79]；P = 0.47；），存在显著异质性（I2 = 68%）。两项研究将超早期组与>6小时组进行比较（；），合并估计同样发现再出血差异无统计学意义（OR = 1.36，95% CI [0.89–2.07]；P = 0.15），异质性低。
全文（开放获取） · 英文原文
【Abstract】 BACKGROUND: Nonvariceal upper gastrointestinal bleeding (NVUGIB) is a common emergency with substantial morbidity and mortality. Guidelines recommend endoscopy within 24 hours, but the benefit of ultra-early endoscopy (≤6 hours) remains unclear, particularly in high-risk patients. This study aimed to systematically evaluate whether ultra-early endoscopy (≤6 hours) improves clinical outcomes compared with early endoscopy (>6 hours) in adult patients with nonvariceal upper gastrointestinal bleeding, including mortality, rebleeding, healthcare utilization, and outcomes in high-risk subgroups. METHODS: A systematic review and meta-analysis of studies comparing ultra-early and early endoscopy in adults with NVUGIB was conducted. The PubMed, Embase, and Web of Science databases were searched up to January 2026 for relevant studies. Ultra-early endoscopy was defined as endoscopy performed within 6 hours of presentation, admission, or clinical diagnosis, and high-risk patients were defined according to study-level criteria, most commonly a Glasgow-Blatchford score >12 or Rockall Score >5. Pooled odds ratios (ORs) or mean differences (MDs) with 95% confidence intervals (CIs) were calculated using random-effects models, and prespecified subgroup and sensitivity analyses were used to assess the robustness of the findings. RESULTS: Nine studies, comprising 10,785 patients (2,638 ultra-early, 8,147 early), were included in the meta-analysis, and their data analyzed. The findings suggest that ultra-early endoscopy did not significantly reduce mortality (OR 0.88, 95% CI [0.58-1.34]) or rebleeding (OR 1.23, 95% CI [0.90-1.67]), nor affect transfusion, surgical intervention, repeat endoscopy, or the length of hospital stay of the patients. However, intensive care unit (ICU) admission was more frequent in the ultra-early group (OR 1.41, 95% CI [1.20-1.66]). These findings were also consistent for patients in high-risk subgroups. CONCLUSION: Among adults with NVUGIB, ultra-early endoscopy provides no clear advantage over early endoscopy in terms of the key clinical outcomes, although the rate of ICU admission is higher. This suggests that management should focus on providing adequate resuscitation and timely endoscopy within an appropriate therapeutic window. Further prospective studies are warranted to evaluate the potential benefit of ultra-early endoscopy in selected high-risk patients. 【Introduction】 Nonvariceal upper gastrointestinal bleeding (NVUGIB) is one of the most common gastrointestinal emergencies encountered in clinical practice and represents a leading indication for urgent endoscopic evaluation (; ). NVUGIB accounts for the majority of acute upper gastrointestinal bleeding episodes, which continue to be associated with significant morbidity and mortality worldwide (). Estimates from large population-based studies indicate that the rate of in-hospital mortality among patients with NVUGIB generally ranges from approximately 5%–10%, with a markedly higher fatality rate observed in older patients and those with significant comorbidity burdens (; ). Despite substantial improvements in adjunctive medical treatments, including proton pump inhibitors, refinements in endoscopic hemostasis techniques, and the ongoing optimization of overall peri-emergency care pathways, the clinical impact of NVUGIB remains considerable, often requiring hospital admission, blood transfusions, and intensive healthcare resource utilization (; ; ). These observations underscore the continued importance of optimizing management strategies for NVUGIB to improve patient outcomes and reduce the burden to healthcare systems (; ; ). Endoscopy plays a central role in the diagnosis, risk stratification, and therapeutic intervention of NVUGIB (; ). Current international guidelines recommend performing endoscopy within 24 h of patient stabilization to prevent unnecessary delays while allowing adequate resuscitation (; ). However, this 24-hour window is relatively broad and does not address whether earlier intervention could confer additional clinical benefits (). With improvements in emergency endoscopy accessibility and 24-hour service availability, some centers have explored advancing the timing of endoscopy, particularly performing procedures within 6 h of presentation or diagnosis, considered ultra-early endoscopy (; ). The theoretical rationale for ultra-early endoscopy is that the prompt identification and management of high-risk bleeding lesions may reduce ongoing bleeding, rebleeding, and associated complications (). Nevertheless, studies to date evaluating this strategy have reported inconsistent results: some suggest potential reductions in mortality or rebleeding, whereas others have found no clear benefit and, in some cases, even increased intensive care unit (ICU) admission or healthcare resource utilization (; ). Such discrepancies may reflect differences in the baseline patient risk, the adequacy of resuscitation, study design, and residual confounding issues in different studies. Additionally, the definition of “ultra-early endoscopy” varies across studies, ranging from strict 6-hour cutoffs to broader early time intervals, complicating evidence synthesis and interpretation. Additionally, evidence in high-risk patient subgroups remains limited and inconclusive, preventing firm guidance being developed on the routine application of ultra-early endoscopy (). Given these uncertainties, an updated systematic review and meta-analysis focused specifically on the ≤6-hour threshold in adults with NVUGIB is warranted. We therefore compared ultra-early endoscopy (≤6 h) with early endoscopy (>6 h) and evaluated mortality, rebleeding, healthcare utilization, and outcomes in high-risk subgroups. This synthesis aimed to clarify whether routine acceleration of endoscopy provides additional clinical value beyond guideline-concordant early endoscopy after appropriate stabilization. 【Materials and Methods】 This systematic review and meta-analysis was conducted in accordance with the Cochrane Handbook for Systematic Reviews of Interventions and is reported following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement (; ). The study protocol was prospectively registered in the PROSPERO database (CRD420251273617) prior to initiation of the study. 【Literature search】 A systematic literature search was performed in the PubMed, Embase, and Web of Science databases from their inception to January 1, 2026, to identify studies comparing ultra-early and early endoscopy in patients with NVUGIB. The search strategy incorporated both Medical Subject Headings and free-text terms related to NVUGIB, gastrointestinal endoscopy, and the timing of intervention. Core search terms included (“nonvariceal upper gastrointestinal bleeding” or “NVUGIB”) AND (“Endoscopy, Gastrointestinal” or endoscop* or EGD) AND (“Time Factors” or timing or early or “≤6” or “6-hour” or “within 6 hour”). No restrictions were applied with respect to language or publication status. The reference lists of relevant articles and previous systematic reviews were manually screened to identify additional eligible studies. The full search strategy is outlined in . The search strategy was designed and implemented independently by Cai-Juan Liu and Logen Liu. Any disagreements were resolved through discussion; if consensus could not be reached, Guo-Qing Li acted as the referee and made the final decision. 【Inclusion criteria】 Studies were considered eligible if they enrolled adult patients (≥18 years old) with NVUGIB and directly compared ultra-early endoscopy with early endoscopy using extractable outcome data for the timing groups. “Ultra-early endoscopy” was defined as upper gastrointestinal endoscopy performed within 6 h of patient presentation, hospital admission, or clinical diagnosis, while “early endoscopy” referred to procedures performed more than 6 h after these time points, including specific intervals such as 6–24 h or broader comparator windows beyond 6 h when the >6-hour group was clearly defined. High-risk status, when reported, was defined by study-level criteria, typically a Glasgow-Blatchford bleeding score (GBS) >12 or Rockall Score >5. Both diagnostic and therapeutic endoscopic procedures were considered. Eligible study designs included randomized controlled trials and prospective or retrospective observational cohort studies reporting at least one relevant clinical outcome. 【Exclusion criteria】 Studies were excluded if they exclusively involved patients with esophageal or gastric variceal bleeding, included pediatric populations without separable adult data, or were designed as case reports, case series, reviews, editorials, commentaries, or conference abstracts lacking sufficient extractable data. Studies without a direct comparison between ultra-early and early endoscopy were also excluded. In addition, studies reporting mixed variceal and nonvariceal bleeding were excluded if outcome data for NVUGIB could not be independently extracted according to the prespecified timing groups. 【Study selection and data extraction】 Study selection and data extraction were independently conducted by two investigators using a predefined standardized data extraction form. Study titles and abstracts were initially screened to exclude clearly irrelevant studies, followed by a full-text review of the potentially eligible articles. The extracted data included study characteristics, patient demographics and baseline risk stratification (e.g., Rockall or GBS scores), definitions and timing of endoscopic interventions, outcome measures of interest, and, when available, adjusted effect estimates and variables included in the adjustment models. The outcomes of interest included mortality, rebleeding, transfusion requirements, length of hospital stay, ICU admission, and need for surgical or repeat endoscopic intervention. Disagreements were resolved through discussion or consultation with a third investigator. 【Quality assessment】 The methodological quality and risk of bias of the included studies were independently assessed by two reviewers. Randomized controlled trials were evaluated using the Cochrane Risk of Bias tool version 2 (), and non-randomized studies were assessed using the Risk Of Bias in Non-randomized Studies of Interventions (ROBINS-I) tool (). Any disagreements were resolved by consensus or by adjudication by a third reviewer. 【Statistical analysis】 A qualitative synthesis of the included studies was initially performed. Meta-analysis was conducted when sufficient clinical and methodological homogeneity was present. For dichotomous outcomes, pooled odds ratios (ORs) with 95% confidence intervals (CIs) were calculated, while continuous outcomes were synthesized using mean differences (MDs). Statistical heterogeneity was assessed using Cochran’s Q test and quantified with the I2 statistic, with I2 values of 0–25%, 26%–50%, and >50% interpreted as indicating low, moderate, and substantial heterogeneity, respectively (). Given the anticipated clinical heterogeneity in study design, comparator timing windows, baseline patient risk, and outcome definitions, random-effects models were used as the main analytical approach for all pooled outcomes. Prespecified subgroup analyses were performed according to comparator timing window, baseline patient risk, and mortality outcome type. Sensitivity analyses were conducted by sequentially excluding individual studies to evaluate the robustness of the pooled estimates. Adjusted estimates were preferentially described when reported by individual studies; however, pooled analyses were primarily based on extractable event counts because adjusted effect estimates were not consistently available across studies and outcomes. Publication bias was assessed using funnel plots and Egger’s regression test (; ). All the statistical analyses were performed using Review Manager (RevMan) version 5.4 and Stata version 14.0, with two-sided P values <0.05 considered statistically significant. 【Results】 【Basic characteristics and quality assessment】 The systematic search identified 920 records from PubMed, Embase, and Web of Science with potential for consideration. After the removal of 371 duplicate records, 549 studies were screened based on their titles and abstracts, of which 526 were excluded as irrelevant or involving an inappropriate study type. This left 23 potentially eligible articles, for which the full texts were retrieved to assess their eligibility. Following this full-text evaluation, 14 studies were excluded, mainly because the endoscopy timing was not stratified using a 6-hour cutoff, relevant outcomes were not reported, or the study population included mixed variceal and nonvariceal bleeding without separable data. Ultimately, nine studies met the inclusion criteria and were included in the systematic review and meta-analysis (; ; ; ; ; ; ; ; ). The study selection process is illustrated in the PRISMA flow diagram in . 【PRISMA flowchart presenting the study selection process for this systematic review and meta-analysis.】 The characteristics of the included studies are summarized in . The nine studies were published between 2004 and 2025 and were conducted in the United States, Korea, and Canada (two studies each), as well as China, Japan, and Spain (one study each). Both randomized controlled trials and observational cohort studies were included. Seven studies compared ultra-early endoscopy (≤ 6 h) with endoscopy performed within 6–24 h (; ; ; ; ; ; ), while two studies used a broader comparison window of >6 h (6–48 h) (; ). In total, 10,785 patients with NVUGIB were included, with 2,638 patients in the ultra-early endoscopy group and 8,147 in the early endoscopy group. 【Characteristics of the included studies.】 Notes. RCSRetrospective cohort study Randomized controlled trial Prospective observational cohort Glasgow–Blatchford score Nonvariceal upper gastrointestinal bleeding Not reported hour The superscript *a* indicates the overall NVUGIB population. The superscript *b* indicates the high-risk NVUGIB subgroup population. Baseline risk stratification was reported using either the GBS or the Rockall score. Three studies provided separate data for high-risk patients (; ; ), including two studies defining high risk as a GBS greater than 12 (; ) and one study using a Rockall score greater than 5 (). Seven studies reported the number of deaths (; ; ; ; ; ; ); two studies did not report specific numbers of deaths (; ), and one study reported there were no deaths in either group (). Overall, 90 in-hospital or 30-day deaths occurred in the ultra-early endoscopy group and 111 in the early endoscopy group, corresponding to crude mortality rates of 3.67% and 5.73%, respectively. 【Quality assessment】 Using ROBINS-I, seven non-randomized studies were judged as having a moderate overall risk of bias and one as having a serious overall risk of bias (), mainly because of residual confounding and selection bias inherent to observational timing studies. The RCT by was assessed using the Cochrane Risk of Bias tool (RoB 2.0) and was rated as low risk in the domains of the randomization process, missing outcome data, and outcome measurement, with some concerns regarding deviations from intended interventions and selective reporting. 【Risk-of-bias and methodological quality assessment of the included studies.】 Notes. For ROBINS-I, D1, confounding; D2, selection of participants; D3, classification of interventions; D4, deviations from intended interventions; D5, missing data; D6, measurement of outcomes; D7, selection of reported results; L, low; M, moderate; S, serious. For RoB 2.0, D1, randomization process; D2, deviations from intended interventions; D3, missing outcome data; D4, measurement of the outcome; D5, selection of reported results; RCT, randomized controlled trial. 【Mortality】 A total of nine studies were included to compare the mortality rate between the ultra-early group and the control group in patients with NVUGIB (; ; ; ; ; ; ; ; ). Using a random-effects model, no significant difference in overall mortality was observed between the ultra-early group and the control group (OR = 0.88, 95% CI [0.58–1.34]; P = 0.56; ), with substantial heterogeneity (I2 = 68%). 【Forest plots of mortality in patients with NVUGIB.】 (A) Overall mortality; (B) sensitivity analysis excluding the study by ; (C) analysis among high-risk patients; (D) subgroup analysis according to endoscopy timing; (E) subgroup analysis according to the type of mortality outcome. Sensitivity analysis excluding the study by , which had the greatest weight, shifted the pooled estimate toward reduced mortality in the ultra-early group (OR = 0.78, 95% CI [0.52–1.17]; ), but this remained non-significant (P = 0.23), with heterogeneity decreasing to I2 = 39%. Among the high-risk patients, no significant difference in mortality was detected between the ultra-early group and the control group (pooled OR = 0.92, 95% CI [0.38–2.20]; P = 0.85; ), with substantial heterogeneity (I2 = 72%). 【Subgroup analyses of mortality】 Subgroup analyses were conducted according to endoscopy timing in the control group and the type of mortality outcome. Regarding endoscopy timing in the control group, seven studies compared the ultra-early group with the control group undergoing endoscopy within 6–24 h (; ; ; ; ; ; ). The pooled analysis showed no significant difference in mortality between the groups (OR = 0.98, 95% CI [0.65–1.48]; P = 0.93; ), with substantial heterogeneity (I2 = 68%). In contrast, two studies compared the ultra-early group with a >6 h group (; ). In this subgroup analysis, the ultra-early group was found to be associated with a lower mortality risk (OR = 0.38, 95% CI [0.15–0.97]; P = 0.04), with low heterogeneity (I2 = 0%). The test for subgroup differences did not reach statistical significance (P = 0.07). Regarding the mortality outcome type, three studies reported in-hospital mortality (; ; ). The pooled analysis showed no significant difference in in-hospital mortality between the ultra-early group and the control group (OR = 1.23, 95% CI [0.78–1.93]; P = 0.37; ), with low heterogeneity (I2 = 0%). In contrast, six studies reported 30-day mortality (; ; ; ; ; ). The pooled estimate similarly demonstrated no significant reduction in 30-day mortality in the ultra-early group compared with the control group (OR = 0.72, 95% CI [0.38–1.36]; P = 0.31), with substantial heterogeneity (I2 = 79%). The test for subgroup differences between in-hospital mortality and 30-day mortality did not reach statistical significance (P = 0.18). 【Rebleeding】 A total of seven studies compared rebleeding between the ultra-early group and the control group in patients with NVUGIB (; ; ; ; ; ; ). Using a random-effects model, ultra-early endoscopy was not associated with a significant change in rebleeding risk compared with the control group (OR = 1.23, 95% CI [0.90–1.67]; P = 0.19; ), with substantial heterogeneity (I2 = 55%). 【Forest plots of rebleeding in patients with NVUGIB.】 (A) Overall rebleeding; (B) sensitivity analysis excluding the study by ; (C) analysis among high-risk patients; (D) subgroup analysis according to endoscopy timing. Sensitivity analysis excluding the study with the largest weight (31.8%) (), showed a similar non-significant effect on rebleeding (OR = 1.24, 95% CI [0.76–2.03], P = 0.39; ), with heterogeneity increasing to I2 = 62%. Among high-risk patients, no significant difference in rebleeding was detected between the two groups (pooled OR = 1.29, 95% CI [0.95–1.75]; P = 0.10; ), with low heterogeneity (I2 = 0%). 【Subgroup analyses of rebleeding】 Subgroup analyses were performed according to endoscopy timing in the control group. Five studies compared the ultra-early group with the control group undergoing endoscopy within 6–24 h (; ; ; ; ), and the pooled analysis showed no significant difference in rebleeding between the groups (OR = 1.17, 95% CI [0.76–1.79]; P = 0.47; ), with substantial heterogeneity (I2 = 68%). Two studies compared the ultra-early group with the >6 h group (; ), and the pooled estimate likewise found no significant difference in rebleeding (OR = 1.36, 95% CI [0.89–2.07]; P = 0.15), with low heterogen