激光多普勒成像对烧伤深度诊断价值的荟萃分析

Meta-analysis on the diagnostic value of laser Doppler imaging for burn depth

目的应用荟萃分析方法系统评价激光多普勒成像（LDI）对深、浅Ⅱ度烧伤创面的诊断价值。方法以＂burn、burns、burn depth、laser Doppler imaging、laser Doppler perfusion imaging、LDI、LDPI＂为检索词检索《PubMed》《Cochrane Library》数据库，以＂烧伤、烧伤深度、激光多普勒、激光多普勒血流灌注成像、激光多普勒成像仪＂为检索词检索《中国期刊全文数据库》《万方数据库》《维普数据库》，搜索各数据库自建库起至2015年10月公开发表的有关LDI诊断深、浅Ⅱ度烧伤创面的文献。采用诊断准确性研究质量评价工具2评估纳入研究的偏倚风险及临床适用性。采用Meta－DiSc 1.4统计软件对数据进行分析，评估纳入研究间的异质性；计算纳入研究中LDI诊断深、浅Ⅱ度烧伤创面的合并效应量，包括敏感度、特异度、阳性似然比、阴性似然比、诊断比值比及综合受试者工作特征曲线下面积与Q^＊指数；采用敏感性分析验证纳入研究总体合并效应量结果的稳定性；采用亚组分析寻找纳入研究间异质性来源。绘制Deek漏斗图评估纳入研究结果引起的发表偏倚。结果共纳入5篇符合入选标准的文献，包括138例患者、179个创面。纳入研究中仅1项偏倚风险低、其余4项偏倚风险高/不确定；在临床适用性方面，仅1项临床适用风险度高/不确定、其余4项临床适用风险度低。纳入研究间不存在阈值效应引起的异质性（受试者工作特征平面图中散点分布未见呈＂肩臂＂状，βs＝0.01，P〉0.05），纳入研究间存在非阈值效应引起的高度异质性（敏感度I2＝54%）。LDI诊断深、浅Ⅱ度烧伤创面的纳入研究总体合并敏感度、特异度、阳性似然比、阴性似然比、诊断比值比分别为95%（95%置信区间为88%～99%）、96%（95%置信区间为90%～99%）、9.75（95%置信区间为4.35～21.81）、0.11（95%置信区间为0.05～0.22）、257.93（95%置信区间为58.96～1 128.41），综合受试者工作特征曲线下面积为0.98、Q^＊指数为94%。纳入研究总体合并效应量结果稳定性好，异质性来源可能为病例选择的偏倚风险、LDI设备型号以及对照诊断方法的选择。Deek漏斗图显示，不存在纳入研究结果引起的发表偏倚（P〉0.05）。结论LDI诊断深、浅Ⅱ度烧伤创面的敏感度、特异度、阳性似然比、诊断比值比及Q^＊指数均较高，综合受试者工作特征曲线下面积大，阴性似然比较低，具有较高的诊断价值。

ObjectiveTo evaluate the value of laser Doppler imaging （LDI） in diagnosing deep or superficial partial-thickness burn wound with meta-analysis.MethodsDatabases including PubMed, Cochrane Library were searched using key words ＂ burn, burns, burn depth, laser Doppler imaging, laser Doppler perfusion imaging, LDI, LDPI＂ , and Chinese Journals Full-text Database, Wanfang Database, VIP Database were searched using Chinese key words ＂烧伤,烧伤深度,激光多普勒,激光多普勒血流灌注成像,激光多普勒成像仪＂ to obtain the published trials of LDI in diagnosing deep or superficial partial-thickness burn wound from the establishment of each database to October 2015. The risk of bias and applicability concerns of the included studies were evaluated by Quality Assessment of Diagnostic Accuracy Studies-2. Data were processed with Meta-DiSc 1.4 statistical software. The heterogeneity among the included studies was evaluated. The pooled estimates of LDI in diagnosing deep or superficial partial-thickness burn wound in the included studies, including sensitivity, specificity, positive likelihood ratio, negative likelihood ratio, diagnostic odds ratio, as well as the area under the curve of summary receiver operating characteristic and index of Q^＊ were calculated. The stability of results of the overall pooled estimates of the included studies was validated by sensitivity analysis. The sources of heterogeneity among the included studies were sought through subgroup analysis. The publication bias caused by the results of the included studies was evaluated by drawing the Deek′s funnel plot.ResultsA total of 5 articles conforming to the inclusion criteria, involving 138 patients and 179 burn wounds, were included. One of the included studies had low risk of bias, while the other 4 studies had high/unclear risk of bias. In applicability concern aspect, one of the included studies had high/unclear concerns regarding applicability, while the other 4 studies had low concerns regarding applicability. There was no heterogeneity caused by threshold effects among the included studies （no ＂ shoulder arm＂ plot of the scattered-point distribution was observed in the space of the receiver operating characteristic, βs=0.01, P〉0.05）. There was a high heterogeneity caused by non-threshold effects among the included studies （sensitivity I^2=54%）. The overall pooled sensitivity, specificity, positive likelihood ratio, negative likelihood ratio, and diagnostic odds ratio of the included studies on LDI in diagnosing deep or superficial partial-thickness burn wound were 95% （with 95% confidence interval 88%-99%）, 96% （with 95% confidence interval 90%-99%）, 9.75 （with 95% confidence interval 4.35-21.81）, 0.11 （with 95% confidence interval 0.05-0.22）, and 257.93 （with 95% confidence interval 58.96-1 128.41）, respectively. The area under the curve of summary receiver operating characteristic was 0.98, with index of Q^＊ 94%. The results of the overall pooled estimates of the included studies was stable. The risk of bias of the selection of the patients, equipment type of LDI, and the selection of diagnostic method in control might be sources of the heterogeneity of the included studies. The Deek′s funnel plot indicated that there was no publication bias caused by the results of the included studies （P〉0.05）.ConclusionsLDI has high sensitivity, specificity, positive likelihood ratio, diagnostic odds ratio and index of Q^＊, large area under the curve of summary receiver operating characteristics, and low negative likelihood ratio in diagnosing deep or superficial partial-thickness burn wound, which has relatively high diagnostic value.