术前超声预测腹腔镜胆囊切除术难度的受试者工作特征曲线分析

Evaluation of ultrasonic predicting difficulty of laparoscopic cholecystectomy by receiver operator characteristic curve

目的建立术前超声预测腹腔镜胆囊切除术(laparoscopiccholecystectomy,LC)难度评分表,并评价其科学性。方法连续选择118例接受LC患者,随机分为两组:训练样本100例,验证样本18例。采用自身前后对照试验方案,术前应用超声检测胆囊大小(长x胆囊底宽)mm2,胆囊壁厚度(mm),胆囊颈有无结石嵌顿,单个胆囊结石数目及最大长径(mm),脐孔及胆囊周围有无粘连;术后登记手术时间(min),术中出血量(mL),有无中转开腹,手术并发症以及术后住院时间(d)。根据100例训练样本LC实际难度分为容易和困难两级,应用t检验和χ2检验统计筛选出有统计学意义的超声检测指标,建立术前超声预测LC难度评分表,进行受试者工作特征曲线(receiveroperatorcharacteristiccurve,ROC)分析。结果胆囊大小、胆囊壁厚度、胆囊颈有无结石嵌顿、脐孔粘连及胆囊周围粘连5项超声检测指标在LC容易和困难两级之间的差异有显著性,P均<0.05。应用5项指标建立术前超声预测LC难度评分表,经ROC分析,曲线下面积为0.922,与完全随机情况下获得的曲线下面积(0.5)相比,差异有显著性,P<0.05。经18例检验样本前瞻性误判概率评估,结果显示术前超声预测LC难度误判率约5.6%。结论术前超声预测LC难度评分表可以正确预测LC难度。

[Objective] To estabhsh and evaluate a scoring system for ultrasonic predicting difficulty of laparoscopic cholecystectomy （LC）. [Methods] With self-controlled trial, 118 consecutive patients undergoing LC were divided into two groups： 100 in training group, 18 in testing group. Preoperative ultrasonic parameters including the size of gallbladder, thickness of gallbladder wall, impacted stone in neck of gallbladder, the number and the max diameter of stone, and the adhesions in umbilicus or peri-gallbladder were collected, postoperative variables including operating time, bleeding loss, conversion to open cholecystectomy, postoperative complication, the days postoperative stay in hospital were also collected. According to the real difficulty of 100 training samples, Chi-square test was applied to choose statistically significant ultrasonic preoperative variables which formed a scoring system for ultrasonic predicting difficulty of LC. Receiver operator characteristic curve （ROC） was then applied to analyze the scoring system. [Results] The variables of preoperative ultrasonic parameters with statistically significant effect were the size of gallbladder, thickness of gallbladder wall, impacted stone in neck of gallbladder and the adhesions in umbihcus or peri-gallbladder, which built up the scoring system for ultrasonic predicting difficulty of LC. The scoring system ROC Az was 0.922, which was statistically higher than the randomizing scheme ROC Az=0.5, P〈0.05. The appraisal of prospective misjudge possibihty was applied to the scoring system for ultrasonic predicting difficulty of LC in 18 testing samples, the misjudge rate about 5.6% was found. [Conclusions] The scoring system for ultrasonic predicting difficulty of LC could be applied preoperatively to estimate the difficulty in patients undergoing LC.