Objective: To systematically evaluate the diagnostic value of optical imaging combined with indocyanine green (ICG)-guided sentinel lymph node (SLN) biopsy in gastric cancer, and to identify potential factors that would influence diagnostic accuracy. Methods: Study was carried out by searching the electronic database of PubMed, Embase, Medline, Web of Science, and the Cochrane Library with keywords as "gastric/stomach" and "cancer/carcinoma/tumor/tumour/adenocarcinoma/neoplasm" and "sentinel lymph node" and "near-infrared/NIR or fluorescent imaging" and "indocyanine green/ICG" . Literature inclusion criteria: (1) gastric cancer clinical stage was cT0-3; (2) clinical stage determined by at least 2 kinds of imaging modalities; (3) optical imaging (near-infrared or fluorescence imaging) combined with ICG-guided SLN biopsy; (4) prospective study to predict lymph node metastasis; (5) intraoperative or postoperative pathology for all lymph nodes removed; (6) patients number in the literature >10 cases. Exclusion criteria: (1) patients with a history of ICG allergy or chemoradiotherapy; (2) previous history of endoscopic mucosal resection or endoscopic submucosal dissection; (3) patients with a variety of gastrointestinal tumor; (4) case reports, conference abstracts, clinical guidelines, editorials, reviews, meta-analysis and correspondence letters; (5) in vitro or animal experiments; (6) insufficient diagnostic efficacy data. The meta-analysis was performed in the Stata12.0 software using the "bivariate mixed-effects model" combined with the "midas" command to pool the data. Information such as true positive value, false positive value, false negative value, and true negative value of each included articles were extracted. The literature quality assessment map was drawn to describe the overall quality of the articles; the heterogeneity analysis was performed with the forest map, with P<0.01 considered as statistical significance; the funnel plot was used to describe publication bias, with P<0.1 considered as statistically significant. Area under curve (AUC) of summary receiver operator characteristic (SROC) was used to describe the diagnostic accuracy and the AUC closer to 1 indicated higher diagnostic accuracy. If there was heterogeneity (I(2)>50%) among studies, regression analysis and subgroup analysis were performed. P<0.05 was considered as statistically significant. Results: A total of 15 studies (1020 patients) were included. The optical imaging contained near-infrared (NIR) and fluorescent imaging (FI). The diagnostic value of optical imaging combined with ICG-guided SLN biopsy in gastric cancer was as follows: the pooled sensitivity (Sen) was 0.95 (95% CI: 0.82 to 0.99), specificity (Spe) was 1.00 (95% CI: 0.92 to 1.00), positive likelihood ratio (PLR) was 30.39 (95% CI: 9.14 to 101.06), negative likelihood ratio (NLR) was 0.05 (95% CI:0.01 to 0.20), diagnostic odds ratio (DOR) was 225.54 (95% CI: 88.81 to 572.77), AUC was 1.00 (95% CI: 0.99 to 1.00), threshold value was sensitivity=0.95 (95% CI: 0.82 to 0.99) and specificity=1.00 (95% CI: 0.92 to 1.00). Deeks method revealed DOR funnel plot of SLN biopsy was not asymmetrical obviously with significant difference (P=0.01), which indicated remarkable publishing bias. Meta-subgroup analysis showed that compared to FI, NIR imaging had higher sensitivity (0.98 vs. 0.73); compared to 0 minutes, optical imaging performed 20 minutes after ICG injection had higher sensitivity (0.98 vs. 0.70); compared to mean detected number of SLN of 4, mean detected number≥4 had higher sensitivity (0.96 vs. 0.68); compared to HE stain, immunohistochemistry + HE had higher sensitivity (0.99 vs. 0.84); compared to subserous injection of ICG, submucosa injection of ICG had higher sensitivity (0.98 vs. 0.40); compared to injection of 5 g/L ICG, 0.5 g/L and 0.05 g/L had higher sensitivity (0.98 vs. 0.83); compared to cT2-3 tumor, early stage (cT1) tumor had higher sensitivity (0.96 vs. 0.72); compared to ≤ enrolled 26 cases in the study, > 26 cases had higher sensitivity (0.96 vs. 0.65); compared to papers before 2010, papers after 2010 had higher sensitivity (0.97 vs. 0.81); whose differences were all significant. Sensitivity differences between mean tumor diameter of ≤30 cm and >30 cm, open surgery and laparoscopic surgery, lymph node regional dissection and retrieved dissection were not significant (all P>0.05). Conclusions: Optical imaging combined with ICG-guided SLN biopsy is clinically feasible, and especially suitable for early gastric cancer. However, the ICG being used in current studies may be overdosed. Higher sensitivity may be achieved from NIR imaging when compared with FI method.
目的： 系统评价光学成像结合吲哚菁绿（ICG）引导胃癌前哨淋巴结（SLN）活检的诊断价值。 方法： 以"gastric/stomach" and "cancer/carcinoma/tumor/tumour/adenocarcinoma/neoplasm" and "sentinel lymph node" and "near-infrared/NIR or fluorescent imaging" and "indocyanine green/ICG"为关键词，检索Pubmed、Embase、Medline、Web of science和Cochrane Library等电子数据库，纳入光学成像结合ICG引导SLN活检的前瞻性诊断试验。文献纳入标准：（1）患有可手术切除的胃癌（cT(0~3)）；（2）肿瘤临床分期至少由两种影像学检查来确定；（3）光学成像（近红外成像或荧光成像）结合ICG引导胃癌SLN活检的诊断准确性试验；（4）预测胃癌淋巴结转移情况的前瞻性研究；（5）对术中切下的所有淋巴结进行术中或术后病理活检；（6）文献中统计分析的患者例数>10例。排除标准：（1）患者有相关用药过敏史或放化疗史；（2）既往接受过内镜下黏膜切除或内镜黏膜下剥离术；（3）研究纳入人群合并多种消化道肿瘤疾病；（4）病例报告、会议摘要、临床指南、社论、综述、Meta分析及书信；（5）体外实验及动物实验；（6）诊断效能数据不足。应用Stata12.0软件以"双变量混合效应模型"结合"midas"命令进行分析。提取各纳入文献的真阳性值、假阳性值、假阴性值、真阴性值等信息。绘制文献质量评估图描述总体纳入文献质量；用森林图进行异质性分析，P<0.01认为差异有统计学意义；用漏斗图描述文献发表偏倚，P<0.1认为差异有统计学意义；集成受试者工作曲线法（SROC）计算曲线下面积（AUC）描述诊断准确性，（SROC）AUC越接近于1，表示诊断准确性越高；若研究间存在异质性（I(2)>50%），则进行Meta回归分析及亚组分析，P<0.05认为差异有统计学意义。 结果： 共纳入15篇文献，1 020例患者，光学成像涉及近红外成像和荧光成像两种方法。光学成像结合ICG引导SLN活检诊断价值为：合并灵敏度为0.95（95%CI：0.82~0.99），合并特异度为1.00（95%CI：0.92~1.00），阳性似然比为30.39（95%CI：9.14~101.06），阴性似然比为0.05（95%CI：0.01~0.20），诊断比值比为225.54（95%CI：88.81~572.77），SROC（AUC）为1.00（95%CI：0.99~1.00），临界值为灵敏度=0.95（95%CI：0.82~0.99）、特异度=1.00（95%CI：0.92~1.00）。Deeks法发现胃癌SLN活检的"诊断比值比"漏斗图明显不对称，差异有统计学意义（P=0.01），提示存在明显的发表偏倚。进一步的Meta亚组分析表明：相比荧光成像，近红外成像可获得更高的灵敏度（0.98比0.73）；相比ICG注射后立即光学成像，20 min后光学成像可获得更高的灵敏度（0.98比0.70）；相比SLN平均检出数<4枚，检出数≥4枚可获得更高的灵敏度（0.96比0.68）；相比苏木精-伊红（HE）染色，免疫组织化学染色IHC（+HE）可获得更高的灵敏度（0.99比0.84）；相比浆膜下注射ICG，黏膜下注射获得更高的灵敏度（0.98比0.40）；相比注射浓度为5 g/L的ICG，0.5和0.05 g/L的ICG可获得更高的灵敏度（0.98比0.83），相比肿瘤临床分期cT(2~3)，cT(1)期可以获得更高的灵敏度（0.96比0.72）；相比研究病例数≤26例，>26例可获得更高的灵敏度（0.96比0.65）；相比发表在2010年前的文献，2010年后文献可获得更高的灵敏度（0.97比0.81），差异均有统计学意义（均P<0.05）。而肿瘤平均直径≤30 mm与>30 mm、开腹手术与腹腔镜手术、淋巴结区域清扫与捡出切除合并的灵敏度比较，差异均无统计学意义（均P>0.05）。 结论： 光学成像结合ICG引导胃癌SLN活检是临床可行性良好的诊断方法，尤其适用于早期胃癌。目前胃癌SLN活检研究中，使用的ICG浓度可能过高；而且近红外成像优于荧光成像可能可以获得更高的灵敏度。.
Keywords: Fluorescence imaging; Indocyanine green (ICG); Meta-analysis; Near-infrared (NIR); Sentinel lymph node (SLN); Stomach neoplasms.