下载文档原格式
性,对正常组织细胞相对低毒[7]㊂贝伐珠单抗通过抑制血管内皮生长因子,阻断肿瘤血管新生,从而抑制肿瘤生长[8]㊂此外,还可降低组织间隙压,增强肿瘤组织局部药物浓度,强化化疗药物的作用效果㊂贝伐珠单抗可作为维持治疗的理想药物,多项研究证实,一线化疗结束后继续应用贝伐珠单抗维持治疗可使患者获益[9]㊂曾有研究将贝伐珠单抗与卡培他滨联合应用于大肠癌一线治疗后的维持治疗,患者的无进展生存期明显延长,且毒副作用无显著增加,证实了该治疗模式的合理性[10]㊂消化道肿瘤标志物CEA㊁CA199㊁CA724在结肠癌的诊断及疗效评价中具有重要参考价值,本研究发现,两组患者血清CEA㊁CA199㊁CA724均较治疗前降低,且观察组显著低于对照组,表明贝伐珠单抗联合卡培他滨维持治疗可改善疗效㊂贝伐珠单抗无典型的化疗药物毒性作用,卡培他滨的毒性作用主要体现在手足综合征㊁皮肤色素沉着等方面,且多为Ⅰ~Ⅱ级,对生活状态影响较小㊂本研究中,两组不良反应主要包括手足综合征㊁粒细胞减少㊁胃肠道反应㊁肝功能损害㊁乏力,均以Ⅰ~Ⅱ级并发症为主,经对症治疗后均得以缓解,无患者因无法耐受而放弃治疗,且两组患者各毒副反应发生率差异无统计学意义,提示与卡培他滨单药维持治疗相比,贝伐珠单抗联合卡培他滨维持治疗并不会增加毒副作用,证实了其安全性㊂根据本研究结果,两组在维持治疗后的生存获益存在显著差异,从生存指标来看,观察组总生存率高于对照组,证实贝伐珠单抗联合卡培他滨维持治疗可使患者获得更好的疾病控制和生存获益㊂综上所述,晚期结肠癌一线治疗至疾病控制后,采用贝伐珠单抗联合卡培他滨维持治疗可改善疗效和远期生存情况,同时毒副作用轻微㊂本文受限于样本量,结果可能存在一定偏倚,未来尚需开展前瞻性随机对照研究以验证结论㊂ʌ参考文献ɔ[1]㊀卢元丽,张志国,张颖,等.贝伐珠单抗联合XELOX化疗方案治疗老年晚期结直肠癌肝转移患者的临床观察[J].中西医结合肝病杂志,2021,31(9):806-812.[2]㊀刘传亮.腹腔热灌注化疗对结肠癌术后患者CEACA19-9及临床疗效的影响[J].河北医学,2016,22(9):1472-1474.[3]㊀Cavaletti G,Frigeni B,Lanzani F,et al.The total neuropathyscore as an assessment tool for grading the course of chemo-therapy-induced peripheral neurotoxicity:comparison withthe national cancer institute-common toxicity scale[J].Pe-ripher Nerv Syst,2010,12(3):210-215.[4]㊀席向艳,王有广,张念杰,等.贝伐珠单抗联合FOLFOX6治疗直肠癌的效果及对患者免疫功能的影响[J].结直肠肛门外科,2021,27(2):7-8.[5]㊀李剑萍,高学仁,张晓艳,等.卡培他滨联合贝伐珠单抗或卡培他滨单药在晚期结直肠癌维持治疗中的疗效与安全性[J].昆明医科大学学报,2021,42(9):119-126. [6]㊀邓文静,余更生,刘爱,等.贝伐珠单抗联合mFOLFOX6治疗转移性结直肠癌的临床疗效及左右半结肠癌的疗效差异[J].现代肿瘤医学,2020,28(5):94-98. [7]㊀Elshenawy M A,Badran A,Aljubran A,et al.Survival benefitof surgical resection after first-line triplet chemotherapy andbevacizumab in patients with initially unresectable metastaticcolorectal cancer[J].World J Surg Oncol,2020,18(1):163-165.[8]㊀Cutsem E V,Danielewicz I,Saunders M P,et al.Trifluri-dine/tipiracil plus bevacizumab in patients with untreatedmetastatic colorectal cancer ineligible for intensive therapy:the randomized TASCO1study[J].Ann Oncol,2020,31(9):1160-1168.[9]㊀Bang Y H,Kim J E,Ji S L,et al.Bevacizumab plus capecit-abine as later-line treatment for patients with metastaticcolorectal cancer refractory to irinotecan,oxaliplatin,and flu-oropyrimidines[J].Sci Rep,2021,11(1):227-229. [10]㊀王婷,胡玉海,白琳,等.贝伐珠单抗联合伊立替康和卡培他滨治疗大肠癌的疗效分析[J].现代消化及介入诊疗,2022,27(4):508-512.ʌ文章编号ɔ1006-6233(2024)01-0163-06美他多辛治疗对急性酒精中毒患者苏醒时间症状消失时间和氧化应激反应的影响王小银,㊀左㊀爽,㊀徐礼友,㊀李海山(安徽省合肥市第二人民医院急诊科,㊀安徽㊀合肥㊀230000)㊃361㊃ʌ基金项目ɔ2022年度安徽省中医药传承创新科研项目,(编号:2022CCYB28)ʌ摘㊀要ɔ目的:探究美他多辛治疗对急性酒精中毒患者的影响㊂方法:选取2020年1月至2022年12月我院急诊收治的104例急性酒精中毒患者为研究对象,采用随机数字表法分为常规组(n=52)和美他多辛组(n=52)㊂对照组给予常规治疗,观察组常规治疗基础上给予美他多辛治疗㊂记录两组患者临床指标(苏醒时间㊁症状消失时间),比较治疗前和治疗后,两组患者血气指标[血氧饱和度(SaO2)㊁氧分压(PO2)]及呼气酒精浓度(BrAC)水平,比较治疗前和治疗后,两组患者氧化应激[超氧化物歧化酶(SOD)]状态,治疗前和治疗后,两者患者肝肾功能[丙氨酸氨基转移酶(ALT)㊁天冬氨酸转氨酶(AST)㊁血肌酐(Scr)㊁尿素氮(BUN)]㊁凝血功能[凝血酶原时间(PT)㊁凝血酶时间(TT)㊁活化部分凝血酶时间(APTT)㊁纤维蛋白原(FIB)]㊂结果:美他多辛组苏醒时间㊁症状消失时间均短于常规组(P均<0.05);治疗后,两组患者SaO2㊁PO2㊁SOD㊁FIB水平均较治疗前升高,美他多辛变化水平高于常规组(P均<0.05);BrAC㊁肝肾功能指标(ALT㊁AST㊁Scr㊁BUN)㊁部分凝血功能指标(PT㊁TT㊁APTT)水平较治疗前降低,美他多辛变化水平高于常规组(P均<0.05)㊂结论:美他多辛能够缩短急性酒精中毒患者苏醒时间及症状消失时间,改善患者血气状态,缓解患者氧化应激反应,改善患者肝肾功能及凝血功能㊂ʌ关键词ɔ㊀急性酒精中毒;㊀美他多辛;㊀苏醒时间;㊀症状消失时间;㊀氧化应激反应ʌ文献标识码ɔ㊀A㊀㊀㊀㊀㊀ʌdoiɔ10.3969/j.issn.1006-6233.2024.01.032Effect of Metadoxine Treatment on Time to Awakening Time to Symptom Disappearance and Oxidative Stress in Patients withAcute Alcohol IntoxicationWANG Xiaoyin,ZUO Shuang,XU Liyou,et al(Hefei Second People's Hospital,Anhui Hefei230000,China)ʌAbstractɔObjective:To explore the effects of Metadoxine treatment in patients with acute alcoholism. Methods:A total of104patients with acute alcoholism admitted to the emergency department of our hospital from January2020to December2022were selected as the study subjects,and were randomly divided into the conventional group(n=52)and the Metadoxine group(n=52)by using the method of randomized numerical table.The control group was given conventional treatment,and the observation group was given Metadoxine on the basis of conventional treatment.Clinical indexes(awakening time,symptom disappearance time)were re-corded in the two groups,blood gas indexes[blood oxygen saturation(SaO2),partial pressure of oxygen (PO2)]and breath alcohol concentration(BrAC)levels were compared between the two groups before and after treatment,oxidative stress[superoxide dismutase(SOD)]status was compared between the two groups before and after treatment,and liver and kidney function[alanine aminotransferase(ALT),aspartate amin-otransferase(AST),blood creatinine(Scr),urea nitrogen(BUN)],and coagulation function[prothrombin time(PT),prothrombin time(TT),activated partial thromboplastin time(APTT),fibrinogen(FIB)].Re-sults:The awakening time and symptom disappearance time of the Metadoxine group were shorter than that of the conventional group(P all<0.05);after treatment,the levels of SaO2,PO2,SOD and FIB of the two groups increased compared with those before treatment,and the level of change of Metadoxine was higher than that of the conventional group(P all<0.05);the levels of BrAC,liver and renal function indexes(ALT, AST,Scr,BUN),and some of the indicators of coagulation function(PT,TT,APTT)levels were lower than before treatment,and the level of change of Metadoxine was higher than that of the conventional group(all P <0.05).Conclusion:Metadoxine can shorten the awakening time and symptom disappearance time of patients with acute alcohol intoxication,improve the blood gas status of patients,alleviate the oxidative stress reaction of patients,and improve the liver and kidney function and coagulation function of patients.ʌKey wordsɔ㊀Acute alcohol intoxication;Metadoxine;Wake-up time;Symptom disappearance time; Oxidative stress㊀㊀急性酒精中毒是由酒精导致机体肝脏㊁神经系统功能紊乱的疾病,临床表现为呕吐㊁意识模糊㊁心跳加㊃461㊃速㊁呼吸缓慢㊁昏迷等,甚至呼吸衰竭㊁死亡[1]㊂急性酒精中毒病情进展急骤,若不及时治疗,短时间内可诱发脑功能损伤,治疗需解除中枢神经抑制及机体代谢损伤,保护重要器官[2-3]㊂临床上,急性酒精中毒主要采用解毒㊁催吐㊁洗胃㊁补液等治疗方式,大多患者苏醒后仍会出现头晕㊁恶心呕吐等不良反应,且呕吐物易反流进入器官,导致误吸㊁窒息[4]㊂美他多辛可加速乙醇㊁乙醛代谢,加速其排出体外,避免乙醇㊁乙醛大量堆积,导致肝功能损伤[5]㊂为探究美他多辛治疗对急性酒精中毒患者的影响,本文做以下研究㊂1㊀资料与方法1.1㊀一般资料:纳入2020年1月至2022年12月我院门诊收治的104例急性酒精中毒患者㊂纳入标准:①符合诊断标准,明确有过量饮酒史,昏迷㊁易激惹㊁感觉迟钝㊁肢体不协调等,血液酒精含量>200mg/dL;②年龄ȡ18岁;③均适合所用治疗方案;④患者知情同意㊂排除标准:①其他原因导致昏迷者;②恶性肿瘤者;③对所用药物过敏者;④合并慢性基础病者;⑤血液㊁免疫系统疾病者;⑥精神系统疾病者;⑦严重器官功能障碍者㊂采用随机数字表法分为常规组和美他多辛组,各52例,两组患者一般资料差异无统计学意义(P>0.05)㊂见表1㊂表1㊀患者一般资料比较[ xʃs,n(%)]组别n性别男㊀㊀㊀㊀㊀女年龄(岁)就诊时间(h)饮酒量(mL)意识状态昏睡㊀㊀㊀浅昏迷美他多辛组5242(80.77)10(19.23)33.75ʃ3.42 1.18ʃ0.23435.84ʃ51.2928(53.85)24(46.15)常规组5244(84.62)8(15.38)34.03ʃ3.51 1.20ʃ0.26441.37ʃ52.3325(48.08)27(51.92)χ2/t0.2690.4120.4150.5440.346P0.6040.6810.6790.5870.5561.2㊀方法:对照组:给予洗胃㊁利尿㊁供氧㊁补液㊁大量补充维生素C㊁保肝等常规治疗,同时患者取平卧位,将头偏向一侧,及时清理口腔呕吐物㊂观察组:常规治疗基础上给予美他多辛(浙江震元制药有限公司;国药准字H20130022;规格:0.3g/支)治疗,将0.9g美他多辛加入500mL生理盐水,静脉滴注,单次给药㊂1.3㊀观察指标:①临床指标:记录患者症状消失时间㊁苏醒时间,苏醒时间指开始治疗至意识完全恢复,唤之能醒,且能准确回答问题㊂②血气指标及呼气酒精浓度(breath alcoho1concentration,BrAC):治疗前和治疗后,采用HL-6502型血液分析仪(百特生物)检测患者血氧饱和度(blood oxygen saturation,SaO2)㊁氧分压(oxygen pressure,PO2)水平,Alcotest6810型酒精检测仪(德尔格)检测患者BrAC㊂③氧化应激:治疗前和治疗后,采集血清样本,采用ELISA检测超氧化物歧化酶(superoxide dismutase,SOD)水平,试剂盒购自上海酶联生物㊂④肝肾功能:治疗前和治疗后,采集患者血清样本,采用7180型全自动生化分析仪(日立)检测患者丙氨酸氨基转移酶(alanine aminotransferase, ALT)㊁天冬氨酸转氨酶(aspartate aminotransferase, AST)㊁血肌酐(serum creatinine,Scr)㊁尿素氮(blood u-rea nitrogen,BUN)水平㊂⑤凝血功能:治疗前和治疗后,采用XL3600型全自动凝血功能仪(众驰伟业)检测凝血酶原时间(prothrombin time,PT)㊁凝血酶时间(thrombin time,TT)㊁活化部分凝血酶时间(thrombin time,APTT)㊁纤维蛋白原(fibrinogen,FIB)水平㊂1.4㊀统计学分析:采用SPSS19.0软件进行数据处理㊂计量资料年龄㊁饮酒量㊁就诊时间㊁临床指标㊁血气指标㊁BrAC㊁氧化应激指标㊁肝肾功能指标及凝血功能指标以( xʃs)表示,组间比较行独立样本t检验,组内比较行配对样本t检验,计数资料性别㊁意识状态以n (%)表示,采用χ2检验,P<0.05表示差异有统计学意义㊂2㊀结㊀果2.1㊀临床指标比较:美他多辛组苏醒时间㊁症状消失时间均短于常规组(P均<0.05)㊂见表2㊂2.2㊀血气指标及BrAC比较:治疗前,两组患者SaO2㊁PO2及BrAC水平差异均无统计学意义(P均>0.05);治疗后,两组患者SaO2㊁PO2水平较升高,BrAC水平较治疗前降低,美他多辛组变化水平大于常规组(P均<0.05)㊂见表3㊂㊃561㊃表2㊀临床指标比较( xʃs)组别n苏醒时间(min)症状消失时间(h)美他多辛组5271.34ʃ6.85 2.84ʃ0.75常规组5285.76ʃ8.43 3.27ʃ0.88t9.573 2.682P<0.0010.009表3㊀血气指标及BrAC比较( xʃs)组别时间n SaO2(%)PO2(mmHg)BrAC(mg/dL)美他多辛组治疗前5278.54ʃ3.4684.35ʃ6.89101.45ʃ10.24治疗后97.87ʃ1.08∗96.66ʃ9.12∗57.74ʃ5.81∗差值19.33ʃ1.3112.31ʃ1.1443.71ʃ4.67常规组治疗时5278.77ʃ3.6284.65ʃ6.93100.97ʃ10.05治疗1d后96.38ʃ1.04∗93.27ʃ17.13∗66.48ʃ6.91∗差值17.61ʃ1.258.62ʃ1.0234.49ʃ3.62t差值 6.85017.39511.252P差值<0.001<0.001<0.001㊀㊀注:与同组治疗前比较,∗P<0.052.3㊀氧化应激比较:治疗前,两组患者SOD水平差异无统计学意义(P>0.05);治疗后,两组患者SOD水平升高,美他多辛组变化水平大于常规组(P均<0.05)㊂见表4㊂表4㊀氧化应激比较( xʃs,U/mL)组别n治疗前治疗后差值美他多辛组5276.58ʃ8.1496.45ʃ10.13∗19.87ʃ2.11常规组5277.03ʃ8.2191.62ʃ9.87∗14.59ʃ1.64t0.281 2.46314.247P0.7800.015<0.001㊀㊀注:与同组治疗前比较,∗P<0.052.4㊀肝肾功能比较:治疗前,两组患者ALT㊁AST㊁Scr㊁BUN水平差异均无统计学意义(P均>0.05);治疗后,两组患者ALT㊁AST㊁Scr㊁BUN水平均降低,美他多辛组变化水平大于常规组(P均<0.05)㊂见表5㊂表5㊀肝肾功能比较( xʃs)组别时间n ALT(U/L)AST(U/L)Scr(μmoL/L)BUN(mmoL/L)美他多辛组治疗前5230.25ʃ3.3393.18ʃ9.46148.65ʃ15.3619.74ʃ2.17治疗后21.43ʃ2.07∗36.25ʃ3.72∗85.44ʃ8.75∗ 5.83ʃ1.04∗㊃661㊃差值8.82ʃ1.6456.93ʃ6.1363.21ʃ6.8213.91ʃ1.09常规组治疗时5229.76ʃ3.1593.03ʃ9.34148.17ʃ15.1319.95ʃ2.24治疗后22.87ʃ2.24∗40.66ʃ4.12∗89.28ʃ9.34∗7.15ʃ1.15∗差值 6.89ʃ1.1752.37ʃ5.8758.89ʃ6.2112.80ʃ1.03t差值 6.908 3.874 3.377 5.337P差值<0.001<0.0010.001<0.001㊀㊀注:与同组治疗前比较,∗P<0.052.5㊀凝血功能比较:治疗前,两组患者PT㊁TT㊁APTT 及FIB水平差异均无统计学意义(P均>0.05);治疗后,两组患者PT㊁TT㊁APTT水平均降低,FIB水平升高,美他多辛组变化水平大于常规组(P均<0.05)㊂见表6㊂表6㊀凝血功能比较( xʃs)组别时间n PT(s)TT(s)APTT(s)FIB(g/L)美他多辛组治疗前5221.24ʃ2.0724.95ʃ2.6346.75ʃ4.88 1.86ʃ0.52治疗后13.59ʃ1.42∗15.43ʃ1.59∗35.28ʃ3.64∗ 3.61ʃ1.12∗差值7.65ʃ0.979.52ʃ1.1811.47ʃ1.13 1.75ʃ0.46常规组治疗时5221.15ʃ2.0325.12ʃ2.7146.92ʃ4.95 1.84ʃ0.51治疗后15.37ʃ1.58∗18.16ʃ1.83∗37.55ʃ3.81∗ 3.07ʃ1.01∗差值 5.78ʃ0.84 6.96ʃ1.039.37ʃ1.07 1.23ʃ0.37t差值10.50911.7869.731 6.352P差值<0.001<0.001<0.001<0.001㊀㊀注:与同组治疗前比较,∗P<0.053㊀讨㊀论酒精属中枢神经抑制剂,酒精进入体内后,主要经肝脏代谢酶系统转化为乙醛,然后由乙醛脱氢酶转化为乙酸,最后水解成水和二氧化碳,极少数可通过肾脏和呼吸排出,其中间产物乙醛能够促进机体去甲肾上腺素和肾上腺素分泌,使患者心跳加快,面部潮红,适当饮酒能够活血化瘀,促进血液循环,过量饮酒则引发酒精中毒[6]㊂急性酒精中毒患者短期饮用大量酒精,肝脏无法将其分解,导致大量酒精及其代谢产物在体内堆积,此时,酒精可经血脑屏障进入脑组织,还可使体内β-内啡肽水平升高,抑制神经系统功能,诱发脑血管平滑肌收缩,引发血管痉挛,甚至抑制呼吸系统,导致窒息㊂美他多辛能够加速体内酒精及其代谢产物排出体外,缓解酒精中毒状态,已被用于急性酒精中毒治疗[7],但其相关报道较少,本研究主要探究美他多辛治疗对急性酒精中毒患者的影响㊂既往研究表明,急性酒精中毒患者机体产生大量氧自由基,造成生物膜脂质过氧化损伤,SOD能够清除自由基,维持氧化应激平衡,过量自由基可导致SOD活性降低[8]㊂另外,由于酒精浓度过高患者呼吸中枢被抑制,造成机体缺血㊁缺氧,出现异常呼吸状态,其血氧饱和度下降,血气指标异常[9]㊂本研究显示,美他多辛组苏醒㊁症状消失时间均短于常规组,且SaO2㊁PO2㊁SOD水平高于常规组,BrAC水平低于常规组,提示美他多辛能够显著促进急性酒精中毒患者体内酒精代谢,加快患者康复,缓解氧化应激状态,改善血气状态㊂分析原因可能是因为美他多辛可使肝脏能量物质浓度升高,加速酒精代谢,进而保护肝脏细胞,缓解肝脏损伤,抑制病情进展;此外,美他多辛还具有抗氧化作用,能够促进细胞内谷胱甘肽作用,清除体内氧化自由基,㊃761㊃减轻机体氧化应激,增强机体免疫力,改善患者酒精中毒状态,促进患者康复[10]㊂过量酒精㊁乙醛刺激能够导致肝细胞发生脂肪变性,甚至坏死㊂本研究发现,治疗后,美他多辛组肝肾功能指标㊁凝血功能指标均优于常规组,表明美他多辛能够改善急性酒精中毒患者肝肾功能及凝血功能㊂这可能是因为美他多辛能够激活乙醇脱氢酶和乙醛脱氢酶,促进患者肝脏细胞内乙醇转化及乙醛清除,减轻酒精对肝脏㊁肾脏细胞及血管内皮损伤,缓解炎症反应,改善肝肾功能,美他多辛还能提高辅酶I 活性,加快脂肪氧化分解,降低机体脂肪含量,还可通过稳定谷胱甘肽活性防止脂肪过度氧化,避免肝细胞损伤[11]㊂此外,美他多辛还可抑制NF -κB 介导的炎症系统,抑制炎症因子分泌,减轻炎症反应,进而减轻肝脏细胞损伤,改善肝功能㊂肝功能的改善能够促进脂质的分解转运和凝血因子生成,改善患者凝血功能[12]㊂综上所述,美他多辛能够降低急性酒精中毒患者酒精水平,促进患者康复,缓解患者氧化应激状态,改善患者血气状态㊁肝肾功能及凝血功能㊂ʌ参考文献ɔ[1]㊀狄才英,连美英,李东旭.院前气管插管术治疗急性酒精中毒合并呼吸衰竭患者的临床效果[J ].广西医学,2021,43(2):175-178.[2]㊀周菁,田丽晓,张丽娜.醒脑静注射液联合盐酸纳美芬对急性重度酒精中毒患者氧化应激反应及神经递质水平的影响[J ].世界临床药物,2022,43(11):1470-1474.[3]㊀Caputo F ,Testino G.Orthotopic liver transplantation for pa-tients with end -stage alcohol -related liver disease and severeacute alcohol -related hepatitis [J ].Minerva Surg ,2021,76(5):444-449.[4]㊀陆丽,赵磊,张德桂,等.双管喉罩在急性重度酒精中毒患者呼吸道管理中的应用价值[J ].安徽医学,2020,41(8):957-959.[5]㊀Suresh Babu K ,Paradesi D.Investigation of related impuritiesin metadoxine by a reversed phase high performance liquid chromatography technique [J ].Anal Sci ,2021,37(4):581-584.[6]㊀Sinenchenko AG ,Lodyagin AN ,Savello VE ,et al.Acute se-vere oral poisoning with 1,4-butandiol and ethanol with the development of coma [J ].Zh Nevrol Psikhiatr Im S S Korsak-ova ,2020,120(3):77-81.[7]㊀Mirijello A ,Addolorato G.Treatment of acute alcohol intoxi-cation :The role of metadoxine [J ].Eur Intern Med ,2023,110(1):128.[8]㊀颜建辉,何鹤彬,王敏.纳洛酮联合舒血宁对急性乙醇中毒患者SODGSHiNOS 的影响[J ].河北医学,2020(6):936-940.[9]㊀马君媛.酒毒内蕴型急性酒精中毒患者应用醒脑静治疗对呼气酒精浓度水平的影响[J ].四川中医,2020,38(8):111-114.[10]㊀孙红菊,王衍颜,徐丛聪,等.硫普罗宁联合美他多辛治疗酒精性肝病患者疗效初步研究[J ].实用肝脏病杂志,2020,23(1):54-57.[11]㊀蒋雯,韦亦霖,温清,等.美他多辛通过抑制巨噬细胞和中性粒细胞向肝脏浸润缓解急性酒精性肝损伤[J ].中国药学(英文版),2022,31(1):47-54.[12]㊀黄帅,陈文玲.美他多辛联合硫普罗宁对于酒精性肝病肝纤维化及肝功能的影响[J ].中国医刊,2020,55(3):263-268.文献综述ʌ文章编号ɔ1006-6233(2024)01-0168-04老年动脉粥样硬化性心脏病诊断及临床治疗进展周柯妤,㊀陈㊀茜,㊀廖㊀行(四川大学华西医院,㊀四川㊀成都㊀610041)ʌ关键词ɔ㊀动脉粥样硬化;㊀心脏病;㊀诊㊀断;㊀治㊀疗ʌ文献标识码ɔ㊀A㊀㊀㊀㊀㊀ʌdoi ɔ10.3969/j.issn.1006-6233.2024.01.033㊀㊀动脉粥样硬化(atherosclerosis ,AS )作为心血管疾病的病理基础,常常伴随脂质沉积㊁内皮细胞受损㊁血细胞(血小板㊁白细胞)黏附侵袭㊁平滑肌细胞和胶原纤维增生和泡沫细胞生成等状况发生㊂冠状动脉发生㊃861㊃ʌ基金项目ɔ四川省自然科学基金项目,(编号:2023NSFSC1632)。
MAC4LDFpis Rev 04/221Product InformationMetaPolyzyme, DNA freeSuitable for Microbiome researchMAC4LDFSynonym: Multilytic Enzyme Mix Storage Temperature –20 °CProduct DescriptionMetagenomics investigates all DNA that has been isolated directly from given single samples, such as environmental samples or biological organisms.1,2Metagenomics allows for the investigation of microbes that exist in extreme environments, and which have been historically difficult to isolate, culture, andstudy.3 Metagenomics has revealed the existence of novel microbial species.4 Applications ofmetagenomics work include public health dataanalysis,5,6 discovery of novel proteins, enzymes and natural products,7,8 environmental studies,9,10 and agricultural investigations.11,12Microbes are difficult to disrupt because the cell walls may form capsules or resistant spores. DNA can be extracted by using lysing enzymes such as lyticase, chitinase, zymolase, and gluculase to induce partial spheroplast formation. Spheroplasts are subsequently lysed to release DNA.MetaPolyzyme products (Cat. Nos. MAC4L, MAC4LDF) are based on a multi-lytic enzyme mixture, originally developed by Scott Tighe, for use in microbiome and DNA extraction efficiency studies, and formulated for effective lysis of microbiome samples from extreme environments. MetaPolyzyme was originally evaluated and developed in consultation and collaboration with the Association of Biomolecular Resource Facilities (ABRF) Metagenomics and Microbiome ResearchGroup (MMRG; formerly the Metagenomics Research Group, MGRG).13-16Studies of microbial communities have beenenhanced by the use of culture-independent analytical techniques such as 16S rRNA gene sequencing and metagenomics. DNA contamination during sample preparation is a major problem of sequence-based approaches. Extraction reagents free of DNA contaminants are thus essential. MetaPolyzyme, DNA free was developed to address the need for DNA-free reagents, to minimize microbial DNA contamination from reagents. This productundergoes strict quality control testing to ensure the absence of detectable levels of contaminatingmicrobial DNA using 35 cycles PCR amplification of 16S and 18S rDNA using universal primer sets.Precautions and DisclaimerFor R&D use only. Not for drug, household, or other uses. Please consult the Safety Data Sheet for information regarding hazards and safe handling practices.ReagentThe enzymes in MetaPolyzyme, DNA free are:• Mutanolysin • Achromopeptidase • Lyticase • Chitinase • Lysostaphin •LysozymeAll the enzymes are individually tested for theabsence of contaminating DNA using 16S and 18S PCR amplification.Mutanolysin (from Streptomyces globisporus )Mutanolysin is a muralytic enzyme (muramidase) that cleaves the β-N -acetylmuramyl-(1→4)-N -acetylglucosamine linkage of the bacterial cell wall peptidoglycan-polysaccharide, particularly the β(1→4) bond in MurNAc-GlcNAc.17 Mutanolysin particularly acts on many Gram-positive bacteria, where the enzyme’s carboxy -terminal moietiesparticipate in the recognition and binding of unique cell wall structures.MAC4LDFpis Rev 04/22AchromopeptidaseAchromopeptidase (known also as β-lytic protease 18) has potent bacteriolytic activity on many Gram-positive aerobic bacteria 19 with high lytic activity, against bacterial strains with the A1α chemotype (such as Aerococcus viridans ), and the A3αchemotype (such as Staphylococcus epidermidis ) for cell wall peptidoglycan structures. The enzyme has been reported to have particular recognition for Gly-X sites in peptide sequences, and for Gly-Gly and ᴅ-Ala-X sites in peptidoglycans.20Lyticase (from Arthrobacter luteus )Lyticase is useful in digestion of linear glucosepolymers with β(1→3) linkages, of yeast glycan coats and for spheroplast formation, and of the cell wall of active yeast cells.Chitinase (from Streptomyces griseus )Chitinase degrades chitin by enzymatic hydrolysis to N-acetyl-D-glucosamine. Degradation occurs via two consecutive enzyme reactions: •Chitodextrinase-chitinase, apoly(1,4-β-[2-acetamido-2-deoxy-D-glucoside])-glycanohydrolase, removes chitobiose units from chitin.•N-acetylglucosaminidase-chitobiase cleaves the disaccharide to its monomer subunits, N-acetyl-D-glucosamine (NAGA).Lysostaphin (from Staphylococcus staphylolyticus )Lysostaphin is a lytic enzyme with activity against Staphylococcus species, including S. aureus . Lysostaphin has hexosaminidase, amidase, and endopeptidase activities. It cleaves polyglycine crosslinks in the cellular wall of Staphylococcus species, which leads to cell lysis.21,22Lysozyme (from chicken egg white)Lysozyme hydrolyzes β(1→4) linkages betweenN -acetylmuraminic acid and N -acetyl-D-glucosamine residues in peptidoglycan, and betweenN -acetyl-D-glucosamine residues in chitodextrin. Lysozyme lyses the peptidoglycan cell wall of Gram-positive bacteria.23Storage/StabilityThis product ships at cooler temperature conditions. Long-term storage at –20 °C is recommended. Reconstituted solutions of MetaPolyzyme, DNA free may be stored at –20 °C, but long-term solution stability has not been examined.Preparation InstructionsBecause of the great diversity of samples formetagenomics studies, it will be necessary for each researcher to work out particular conditions for optimal sample preparation and treatment. It is recommended to reconstitute MetaPolyzyme, DNA free in sterile PBS buffer, pH 7.5 (no EDTA, calcium or magnesium present in solution). The following is a sample procedure, to be scaled appropriately:1. Add 100 µL of sterile PBS (pH 7.5) to 1 vial ofMetaPolyzyme, DNA free.1.1. Resuspend by gentle agitation or pipetting. 1.2. Set solution aside at 2-8 °C until Step 7. 2. Thoroughly suspend sample in sterile PBS(pH 7.5). 3. Add 200 µL of sample in PBS to a 2 mLpolypropylene microcentrifuge tube. 4. Optional pellet wash:4.1. To sample tube, add 1 mL of PBS (pH 7.5). 4.2. Vortex, centrifuge and remove supernatant. 4.3. Repeat pellet wash two more times ifneeded. 5. Resuspend pelleted sample in 150 µL of PBS(pH 7.5). Vortex thoroughly.6. Optional: if solution will sit for more than 4 hours,sodium azide may be added to 0.02%. 7. Add 20 µL (*) of MetaPolyzyme, DNA free tosample solution. 8. Incubate at 35 °C for 2-24 hours.9. Continue with standard DNA extraction protocol. (*) The optimal volume and concentration of MetaPolyzyme, DNA free may vary in different experiments.References1. Gilbert, J.A., and Dupont, C.L., Ann. Rev. MarineSci., 3, 347-371 (2011). 2. Kang, H.S., and Brady, S.F., J. Am. Chem. Soc.,136(52), 18111-18119 (2014). 3. Ufarté, L. et al., Biotechnol. Adv., 33(8),1845-1854 (2015). 4. Davison, M. et al., Photosynth. Res., 126(1),135-146 (2015). 5. Afshinnekoo, E. et al., Cell Syst., 1(1), 72-87(2015).The life science business of Merck operatesas MilliporeSigma in the U.S. and Canada.Merck and Sigma-Aldrich are trademarks of Merck KGaA, Darmstadt, Germany or its affiliates.All other trademarks are the property of their respective owners. Detailed information on trademarks is available via publicly accessible resources.© 2022 Merck KGaA, Darmstadt, Germany and/or its affiliates. All Rights Reserved.MAC4LDFpis Rev 04/22 DK,DT,GCY,TJ,RBG,SBC,MAM36.The MetaSUB International Consortium,Microbiome, 4, 24 (2016). [Erratum inMicrobiome, 4, 45 (2016).]7.Trinidade, M. et al., Front. Microbiol., 6, 890(2015).8.Coughlan, L.M. et al., Front. Microbiol., 6, 672(2015).9.Palomo, A. et al., ISME J., 10(11), 2569-2581(2016).10.Pold, G. et al., Appl. Environ. Microbiol., 82(22),6518-6530 (2016).11.Mitra, N. et al., J. Gen. Virol., 97(8), 1771-1784(2016).12.Theuns, S. et al., Infect. Genet. Evol., 43,135-145 (2016).13.Baldwin, D.A. et al., "Life at the Extreme", ABRFMetagenomics Research Group Poster 2015,presented at the ABRF 2015 Conference, St.Louis, MO, USA, March 28-31, 2015.14.Baldwin, D.A. et al., "Implementing NewStandards in Metagenomics and the ExtremeMicrobiome Project", ABRF MetagenomicsResearch Group Poster 2016, presented at theABRF 2016 Conference, Fort Lauderdale, FL, USA, February 20-23, 2016.15.McIntyre, A. et al., "Life at the Extreme: TheABRF Metagenomics Research Group", ABRFMetagenomics Research Group Poster 2017,presented at the ABRF 2017 Conference, SanDiego, CA, March 25-28, 2017.16.Tighe, S. et al., J. Biomol. Tech., 28(1), 31-39(2017).17.Gründling, A., and Schneewind, O., J. Bacteriol.,188(7), 2463-2472 (2006).18.Li, S.L. et al., J. Bacteriol., 172(11), 6506-6511(1990).19.Ezaki, T., and Suzuki, S., J. Clin. Microbiol.,16(5), 844-846 (1982). 20.Li, S. et al., J. Biochem., 124(2), 332-339(1998).21.Browder, H.P. et al., Biochem. Biophys. Res.Commun., 19, 383-389 (1965).22.Robinson, J.M. et al., J. Bacteriol., 137(3),1158-1164 (1979).23.Vocaldo, D.J. et al., Nature, 412(6849), 835-838(2001).NoticeWe provide information and advice to our customers on application technologies and regulatory matters to the best of our knowledge and ability, but without obligation or liability. Existing laws and regulations are to be observed in all cases by our customers. This also applies in respect to any rights of third parties. Our information and advice do not relieve our customers of their own responsibility for checking the suitability of our products for the envisaged purpose. The information in this document is subject to change without notice and should not be construed as a commitment by the manufacturing or selling entity, or an affiliate. We assume no responsibility for any errors that may appear in this document.Technical AssistanceVisit the tech service page at/techservice.Standard WarrantyThe applicable warranty for the products listed in this publication may be found at /terms. Contact InformationFor the location of the office nearest you, go to /offices.。
３８０６Ｍｅｔａ分析在宫颈癌循证医学中的应用贺红英李力宫颈癌是全球女性中仅次于乳腺癌的最常见的妇科恶性肿瘤之一，在一些发展中国家其发病率居首位。
宫颈癌的流行病学研究及合理有效的诊治显得尤为重要。
近２０年来．循证医学及Ｍｅｔａ分析得到了迅速的推广和应用…。
以其严谨的理念与方法．解决了人们面对眼花缭乱，甚至自相矛盾的医学信息的困惑。
本文对近年来宫颈癌流行病学及其诊治相关问题的Ｍｅｔａｌ分析进行回顾。
主要综述Ｍｅｔａ分析在宫颈癌痛因、诊断、治疗的循证医学中应用。
１Ｍｅｔａ分析在宫颈癌病因及流行病学中的应用近ｌＯ余年来。
以大规模人群为基础的流行病学资料显示。
在９９％以上的宫颈癌中存在人乳头状瘤病毒（ｈｕｍａｎｐａｐｉｌｌｏｍａｖｉｒｕｓ。
ＨＰＶ）ＤＮＡ。
ＨＰＶ感染是宫颈癌发生的首要因素且为始动因素【２４】。
根据基因组的同源性可将其分为１１０多种型剐，其中感染人生殖道的ＨＰＶ有３５个型别。
根据其对生殖系统的致瘤性不同分为低危型（非癌相关型）和高危型（癌相关型）两大类。
前者包括ＨＰＶ６、１１、４０、４２等。
主要引发良性增生，如尖锐湿疣。
也可导致轻度宫颈上皮内癌变；后者与宫颈癌的发生密切相关．共有１５型ＨＰｖ从宫颈癌中分离出来，包括ＨＰＶ１６、１８、３１、３３、３５等，约８０％的宫颈癌与４个型剔的ＨＰＶ感染有关．分别是ＨＰＶ１６、１８、３１和４５型，其中５０％的宫颈癌与ＨＰｖ１６感染有关ｆ４－５］。
ＨＰＶＤＮＡ整合人宿主细胞基因组中在ＨＰＶ诱导形成宫颈癌的过程中起关键作用【６ｌ，另宫颈癌还与生殖道其他微生物感染（如ＨＳＶ．２型、沙眼衣原体、生殖道支原体、淋球茵、真茵、阴道滴虫等）、口服避孕药、性因素、社会因素等有关。
国际癌症研究中心（ＩＡＲＣ）进行了一项有关宫颈癌与ＨＰｖ的Ｍｅｔａ分析【７ｌ。
对８５个公开发表的文献中１００５８例宫颈癌患者纳入评价．发现与宫颈癌相关的ＨＰＶ类型为：１６、１８、４５、３ｌ、３３、５８、５２、３５、５９、５６、６、５ｌ、６８、３９、８２、７３、６６和７０，其中有２／３的宫颈癌病例与ＨＰｖ１６（５１％）、１８（１６．２％）密切相关。
动态网络分析（DNA）介绍介绍动态网络分析是一个新的科学领域，综合了基于网络科学和网络理论的社会网络分析（SNA），链接分析（LA）和多Agent系统（MAS）研究。
主要有两个方面的研究：1.动态网络（DNA）数据的统计分析；2.网络动态性的仿真。
在网络中，当你孤立了网络中的主角（关键节点）并不意味着整个网络就不稳定和不能做出反应了，这恰好是忽略了网络的动态性。
比如“孤立”一个“主角”，而又有新的“主角”产生。
我们需要理解动态网络的演化过程，并且我们在面对一个演化网络或信息的丢失必须对“孤立策略”进行评估，（重点是网络的动态性和信息的丢失）。
DNA中Agent角色是通过其过程来考虑的而不是其位置，这就意味着Agent能通信，储存信息，学习。
而且网络随Agents的改变而动态的改变。
连接是个概率，网络的多颜色，多元始由一系列网络组合而成一个复杂网络，其中一个网络的改变都促进或限制其他网络的变化，经常导致错误的级联产生[1]。
我们无法预测变化，但能快速探测变化的发生和对实时的变化做出一些推理；当研究无法预测网络行为，但能提供一种更精确探测到变化的发生和在什么时间将发生的方法，这也是很重要的[2]。
DNA的产生[1]传统SNA关注小的、有边界的网络，在一种类型的节点（人）之间建立两三种链接，在一个时间点我们可能得到完整的信息。
之后一些研究进行了扩展，研究大的网络，或两种类型的节点（人和事件），或者是无边界网络。
动态网络分析中的网络在网络的复杂性，动态性，多状态性，多元性和多个层次上的不确定性不同于传统的社会网络。
在SNA中节点是静态的，而在DNA模型中节点有学习的能力，属性随着时间而改变，节点不断适应；DNA考虑了网络演化的要素研究和在某种环境中可能发生哪些变化。
在一定程度上DNA有点像定量分析，它同概率论相关，但是又不像定量分析，因为DNA中的节点具有主动学习的能力。
目前有一些关于SNA的前沿的研究都扩展动态分析和多颜色（multi-color）网络的领域，主要有三个方面：元矩阵；把关系作为一种概率；社会网络同认知科学和M-Agent系统的结合。
多水平Ｍｅｔａ回归分析是多水平分析方法在Ｍｅｔａ分析中的应用。
对多水平Ｍｅｔａ回归分析及其在流行病学研究中的应用进行介绍，为流行病学资料的Ｍｅｔａ分析提供参考。
１Ｍｅｔａ回归分析概述１．１Ｍｅｔａ分析简介Ｍｅｔａ分析最早由英国教育心理学家Ｇｌａｓｓ于１９７６年命名并将其定义为：“Ｔｈｅｓｔａ－ｔｉｓｔｉｃａｌａｎａｌｙｓｉｓｏｆｌａｒｇｅｃｏｌｌｅｃｔｉｏｎｏｆａｎａｌｙｓｉｓｒｅｓｕｌｔｓｆｒｏｍｉｎｄｉｖｉｄｕａｌｓｔｕｄｉｅｓｆｏｒｔｈｅｐｕｒｐｏｓｅｏｆｉｎｔｅｇｒａｔｉｎｇｔｈｅｆｉｎｄｉｎｇｓ”。
此后，不少统计学家都对Ｍｅｔａ分析下过定义，但都倾向于“Ｍｅｔａ分析是对以往的研究结果进行系统定量综合的统计学方法”这一含义〔１～４〕。
１．２流行病学研究与Ｍｅｔａ回归分析流行病学研究方法通常分为四大类：描述性研究、分析性研究、实验性研究和理论性研究。
前两类均属观察性研究，是流行病学最常用的研究方法。
观察性研究容易受到混杂偏倚和选择偏倚的影响，各项研究的对象选择、研究方法等的不同都会导致研究间的异质性，对异质性较大的资料进行传统的Ｍｅｔａ分析可能会导致错误的结论，从而误导读者。
因此，在对流行病学研究资料进行Ｍｅｔａ分析时，需分析各研究间的异质性，并对异质性的来源进行评估〔５〕。
Ｍｅｔａ回归分析可评价研究间异质性的大小及来源。
根据统计模型的不同，可将Ｍｅｔａ回归分析分为固定效应的Ｍｅｔａ回归分析和随机效应的Ｍｅｔａ回归分析两大类。
基于固定效应模型的Ｍｅｔａ回归分析假设多项研究具有一个共同的效应尺度，各项研究效应尺度存在的差异主要是因为随机误差造成；随机效应模型则假设各项研究不具有共同的效应尺度，而是每项研究都有自己的效应尺度，并将其定义为多水平Ｍｅｔａ回归分析及其在流行病学研究中的应用王安伟１，黄文丽２（１．大理学院公共卫生学院，云南大理６７１０００；２．云南省地方病防治所，云南大理６７１０００）［摘要］目的：介绍多水平Ｍｅｔａ回归分析方法及其在流行病学研究中的应用。
US-China Education Review A, ISSN 2161-623XJanuary 2013, Vol. 3, No. 1, 46-50 Learning Goals and Strategies in the Self-regulation of LearningMartha Leticia Gaeta GonzálezUniversidad Popular Autónoma del Estado de Puebla, Puebla, MéxicoIn order to self-regulate their learning, students need to use different strategies to plan, monitor, and evaluate theirlearning activities (meta-cognitive strategies), as well as to control their motivation and emotion (volitionalstrategies). Students’ effectiveness in their self-regulated learning process also varies depending on the academicenvironment and students’ personal goal orientations. In this study, the author analyzed the interactions betweenthese cognitive, volitional, and motivational variables in late adolescence. To achieve this goal, the author proposeda model by means of SEM (Structural Equation Modeling). The investigation was developed with 268 4th-gradesecondary school students, from public and private schools, in a northwestern city in Spain. Analysis of theproposed model showed the following results: the perception of a classroom learning goal structure relatessignificantly to a personal learning goal orientation, and the latter relates positively to the use of meta-cognitivestrategies, the use of volitional strategies has a mediating effect between a learning goal orientation and the use ofmeta-cognitive strategies. Results are discussed in detail in the document.Keywords: learning goals, meta-cognitive strategies, SRL (self-regulated learning)IntroductionIn the academic context, teachers face the challenge of promoting students’ integral development, through the acquisition of knowledge and skills that can be adapted throughout the different stages of their life. For which education is viewed as a process, in which students must become more self-regulated as learners. SRL (Self-Regulated Learning) should not be viewed as a mental ability or an academic performance skill, but rather as a self-directed process in which students transform their mental abilities into academic skills. It refers to self-generated thoughts, affect and behavior that are oriented towards the achievement of their goals, with the interaction of environmental conditions (Zimmerman, 2002).In this context, meta-cognitive processes, such as planning, monitoring, and evaluation promote students’ SRL. Conceptually, meta-cognition consists of the personal awareness, knowledge, and regulation of one’s cognitive processes (Brown, 1987). While, cognitive strategies are used to help an individual achieve a particular goal (e.g., solving a problem), meta-cognitive strategies are used to ensure that the goal has been reached (e.g., evaluating one’s understanding of that problem).Moreover, between the intention of achieving a goal and implementing activities to achieve it, there are a number of cognitive and meta-cognitive factors, related to the control of these activities, which may facilitate or impede its implementation. So, students’ abilities to use strategies that help them to direct their motivation towards action, in the set-goal direction, are a central aspect of SRL (Wolters, Pintrich, & Karabenick, 2003). Specifically, volitional strategies for maintaining motivation and effort towards goals, as well as for controllingMartha Leticia Gaeta González, Ph.D., professor, researcher, Universidad Popular Autónoma del Estado de Puebla.ll Rights Reserved.LEARNING GOALS AND STRATEGIES IN THE SELF-REGULATION OF LEARNING 47negative emotions, are interrelated and jointly involved in the self-regulation of learning (Boekaerts, 1995).Also, students’ effectiveness in the process of SRL varies depending on the academic environment and their personal goal orientations. Specifically, perceptions of a learning-oriented classroom structure arepositively related to more adaptive learning patterns, such as the use of effective learning strategies, as well asto involvement in the class, motivation, effort, affective states, and eventually academic achievement (Sideridis,2005). In contrast, a performance-oriented classroom structure has been associated with negative learningpatterns (Ryan, Gheen, & Midgley, 1998).Based on the above, this paper proposes a model, using SEM (Structural Equation Modeling) to examine the interactions between the classroom goal structure, personal goal orientation, and the use of volitional andmeta-cognitive strategies in 4th-grade secondary school students.MethodParticipantsA total of 268 4th-grade secondary school students, ranging in age from 15 to 16 years, from public (n =129) and private (n = 139) schools, participated in this investigation. Stratified random sampling was used inthe study.InstrumentsStudents’ perceptions of their classroom goal structure and their goal orientation were assessed by means of the corresponding questionnaire sections from the PALS (Patterns of Adaptive Learning Survey) (Midgley etal., 2000). This instrument contains three subscales that measure students’ perceptions of the meaning ofacademic tasks and achievement that are emphasized in the classroom. The questionnaire also provides an ll Rights Reserved.evaluation of three general types of personal academic goals.Volitional variables were measured by means of the AVSI (Academic Volitional Strategy Inventory) (McCann & Turner, 2004). This instrument measures the extent to which students engage in motivationalregulation strategies for controlling their motivation and emotional states, as they initiate and attempt tomaintain action on academic requirements.The use of meta-cognitive strategies was evaluated through the corresponding scale from the MSLQ (Motivated Strategies for Learning Questionnaire) (Pintrich, Smith, García, & McKeachie, 1991). This scalemeasures the extent to which students use strategies to control and regulate their own cognition.Procedure and Data AnalysesAll the assessment instruments were administered to the students in their classroom, in one session, during the normal academic schedule. Students were assured that their answers would be kept confidential.SEM was utilized to determine how well the proposed theoretical model fit the research data. For the analyses, the author used the LISREL (Linear Structural Relations) 8.80 computer program (Jöreskog &Sörbom, 2006).Model to Be InvestigatedThe proposed model and the relationships between the corresponding variables are displayed in Figure 1.From a general perspective, in the research model, it is hypothesized that:(1) Classroom performance-approach goal structure and classroom performance-avoid goal structurewould positively relate to performance goal orientation;LEARNING GOALS AND STRATEGIES IN THE SELF-REGULATION OF LEARNING48(2) Performance goal orientation would be positively related to volitional strategies and to meta-cognitivestrategies;(3) Classroom mastery goal structure would be positively associated to mastery goal orientation;(4) Mastery goal orientation would positively relate to meta-cognitive strategies;(5) Volitional strategies would mediate the relationship between mastery goal orientation andmeta-cognitive strategies.Figure 1. Graphic representation of the proposed model.Resultsll Rights Reserved.Based on the fit indices, the hypothesized model fit the data quite well. The RMSEA (Root Mean Square Error of Approximation) = 0.053(0.034, 0.069) shows an appropriate value. Data provided by other indices alsooffer support for the acceptance of the model proposed in this study: NNFI (Non-Normed Fit Index) = 0.93;CFI (Comparative Fit Index) = 0.97; GFI (Goodness-of-Fit Index) = 0.94; AGFI (Adjusted Goodness-of-FitIndex) = 0.90. Furthermore, the SRMR (Standardised Root Mean Square Residual) = 0.06 indicates anacceptable mean residual correlation. Finally, the x2/df = 1.74, that provides information on the parsimony ofthe model, shows an excellent level.Despite the good fit of the tested model, the results suggested that there was room for improvement. A close examination of the estimated parameters’ significances and the hypothetical relevance of those notestimated (observed through modifying indexes and standardized residuals) led us to modify the proposedmodel: the path showing the hypothetical association between the classroom performance-approach goalstructure and the performance goal orientation was deleted (γ = 0.05; t = 0.74), also the path showing therelationship between the performance goal orientation and volitional strategies (β = 0.01; t = 0.15) waseliminated, since they both were not significant. Also, a path showing the link between volitional strategies andthe performance goal orientation was included for its estimation (β = 0.18; t = 2.23).The new tested model was both conceptually meaningful and provided good results on the model fit(RMSEA = 0.052(0.035, 0.068), NNFI = 0.96; CFI = 0.97; GFI = 0.94; AGFI = 0.91; SRMR = 0.06; x2/df = 1.71)and at the specific parameters’ estimation level. The obtained results (standardized data) regarding the specificrelations between the different variables are shown in Figure 2.LEARNING GOALS AND STRATEGIES IN THE SELF-REGULATION OF LEARNING 49Figure 2. Path coefficients of the proposed relationships in the model (standardized results).The obtained results confirm totally or partially the hypotheses used for the construction of the model.First, classroom performance-avoid goal structure significantly relates to performance goal orientation (γ = 0.70;t = 8.47), however, this is not the case for the classroom performance-approach goal structure and performancegoal orientation (γ = 0.05; t = 0.74). Second, performance goal orientation is significantly associated tometa-cognitive strategies (β = 0.14; t = 2.14), but not to volitional strategies (β = 0.01; t = 0.15). Third,classroom mastery goal structure is significantly related to mastery goal orientation (γ = 0.55; t = 5.39). Fourth,mastery goal orientation significantly relates to meta-cognitive strategies (β = 0.19; t = 2.08). Fifth, volitional ll Rights Reserved.strategies have a significant mediating effect between mastery goal orientation and meta-cognitive strategies(standardized indirect coefficient = 0.33, p 0.05). Additionally, volitional strategies influence performancegoal orientation (β= 0.18; t = 2.23).ConclusionsThe analyses of the relationships between the model variables reveal the following results: classroom mastery goal structure predicts mastery goal orientation; classroom performance-avoid goal structure andvolitional strategies explain performance goal orientation; mastery goal orientation and performance goalorientation explain meta-cognitive strategies; volitional strategies mediate the relationship between masterygoal orientation and meta-cognitive strategies.From the above, it can be concluded that students’ perception of the classroom structure is an important factor for the development of their personal goal orientation (Ames, 1992). Goal orientation, in turn, appears todefine the strategies that students use to take responsibility (or not) for persevering towards their goalsattainment, by controlling their motivation and emotion (Wolters & Rosenthal, 2000). This effort andpersistence for goal achievement has a positive effect on the use of strategies to control and direct their mentalprocesses for the SRL.Contrary to the author’s expectation, a classroom performance goal structure does not influence students’performance goal orientations, which indicates that the students in this study perceive that the goal for engagingin academic work is not to prove competence (for example, get good grades), but to avoid demonstrating lackLEARNING GOALS AND STRATEGIES IN THE SELF-REGULATION OF LEARNING50of competence (for example, not to be the worst in class), which leads them to compare themselves to othersand to avoid demonstrating any lack of ability (performance orientation).According to the proposed model, volitional strategies have an important mediating role between mastery goal orientation and meta-cognitive strategies. This indicates that learning-oriented students are more likely tofind a link between their efforts and their results, and work to reduce or avoid both internal and externaldistractions (Pintrich & Schunk, 2006), showing higher levels of persistence, compared toperformance-oriented students. This use of motivational and emotional control strategies will produce, as aresult, a greater commitment to learning and to the use of cognitive control strategies.Based on the results of this study, the author emphasizes the importance of helping adolescents in the acquisition of a greater sense of independence and self-confidence, through building classroom environmentsthat empower learners to regulate their learning experience. As it is seen, students’ perceptions of alearning-oriented classroom structure are positively related to a greater academic involvement (Sideridis, 2005),through a mastery goal orientation. Moreover, it can be emphasized that teachers should promote the use ofvolitional strategies to help students maintaining their interest and focus on learning, as well as their emotionalbalance in order to become cognitive engaged.ReferencesAmes, C. (1992). Classrooms: Goals, structures, and student motivation. Journal of Educational Psychology, 84, 261-271.Boekaerts, M. (1995). Self-regulated learning: Bridging the gap between meta-cognitive and meta-motivational theories.Educational Psychologist, 30(4), 195-200.Brown, A. (1987). Meta-cognition, executive control, self-regulation and other more mysterious mechanisms. In F. E. Weinert, & R. H. Kluwe (Eds.), Metacognition, motivation and understanding (pp. 65-116). Hillsdale, N. J.: Erlbaum.Jöreskog, K. G., & Sörbom, D. (2006). LISREL 8,80. Chicago, I. L.: Scientific Software International Inc..ll Rights Reserved.McCann, E. J., & Turner, J. E. (2004). Increasing student learning through volitional control. Teachers College Record, 106(9), 1695-1714.Midgley, C., Maehr, M. L., Hicks, L., Roeser, R., Urdan, T., Anderman, E. M., & Kaplan, A. (2000). The patterns of adaptive learning survey (PALS). Ann Arbor: University of Michigan.Pintrich, P. R., & Schunk, D. H. (2006). Motivación en contextos educativos (M. Limón Trans.). Madrid: Prentice Hall (Trabajo original publicado en 2002).Pintrich, P. R., Smith, D., García, T., & McKeachie, W. J. (1991). A manual for the use of the motivated strategies for learning questionnaire (MSLQ). Ann Arbor: University of Michigan.Ryan, A. M., Gheen, M. H., & Midgley, C. (1998). Why do some students avoid asking for help? An examination of the interplay among students’ academic efficacy, teachers’ social-emotional role, and the classroom goal structure. Journal of EducationalPsychology, 90(3), 528-535.Sideridis, G. D. (2005). Goal orientations, classroom goal structures, and regulation in students with and without learning disabilities: Should we alter student’s motivation, a classroom’s goal structure, or both? In G. D. Sideridis, & T. A. Citro(Eds.), Research to practice: Effective interventions in learning disabilities (pp. 193-219). Boston, M. A.: LearningDisabilities Worldwide.Wolters, C. A., Pintrich, P. R., & Karabenick, S. A. (2003). Assessing self-regulated learning. Paper presented at the Conference on Indicators of Positive Development: Definitions, Measures, and Prospective Validity, March 12-13, 2003.Wolters, C. A., & Rosenthal, H. (2000). The relation between students’ motivational beliefs and their use of motivational regulation strategies. International Journal of Educational Research, 33, 801-820.Zimmerman, B. J. (2002). Becoming a self-regulated learner: An overview. Theory into Practice, 41(2), 64-70.。
Meta-Level Control in Multi-Agent SystemsAnita Raja and Victor LesserDepartment of Computer Science,University of Massachusetts,Amherst,MA01003-4610,USAaraja,lesser@Ph:413-545-3444Fax:413-545-1249AbstractSophisticated agents operating in open environmentsmust make complex real-time control decisions onscheduling and coordination of domain actions.Thesedecisions are made in the context of limited resourcesand uncertainty about outcomes of actions.The ques-tion of how to sequence domain and control actionswithout consuming too many resources in the process,isthe meta-level control problem for a resource-boundedrational agent.Our approach is to design and build ameta-level control framework with bounded computa-tional overhead.This framework will support decisionson when to accept,delay or reject a new task,when it isappropriate to negotiate with another agent,whether torenegotiate when a negotiation task fails and how mucheffort to put into scheduling when reasoning about anew task.IntroductionSophisticated agents operating in complex environmentsmust reason about their local problem solving actions,in-teract with other agents,plan a course of action and carryit out.All these have to be done in the face of limited re-sources and uncertainty about action outcomes in real-time.Furthermore,new tasks can be generated by existing or newagents at any time,thus an agent’s deliberation must be inter-leaved with execution.The planning,scheduling and coor-dination of tasks are non-trivial,requiring either exponentialwork,or in practice,a sophisticated scheme that controls thecomplexity.In this paper,we describe a framework whichwill provide effective allocation of computation resulting inimproved performance of individual agents in a cooperativemulti-agent system.In this framework,agent actions are broadly classifiedinto three categories-domain,control,and meta-level con-trol actions.Domain actions are executable primitive ac-tions which achieve the various high-level tasks.Control ac-tions are of two types,scheduling actions which choose thehigh level tasks,set constraints on how to achieve them andsequence the detailed domain level actions which achievethe selected tasks;and coordination actions which facilitateate amount of processor and other resources at appropriate times.To do this an agent would have to know the effect of all combinations of actions ahead of time,which is in-tractable for any reasonably sized problem.The question of how to approximate this ideal of sequencing domain and control actions without consuming too many resources in the process,is the meta-level control problem for a resource bounded rational agent.Our solution to this problem is to construct a MDP-based meta-level controller which uses reinforcement learn-ing(RL)to learn the utility of control actions and decision strategies in resource-bounded contexts.In order to con-struct such a controller,it is necessary to identify the fea-tures which affect the decisions.The paper is structured as follows:wefirst enumerate the assumptions made in our ap-proach and describe the agent architecture which will pro-vide meta-level control.We then present and evaluate a case-base of hand-generated heuristics for meta-level con-trol.These heuristics will determine the features necessary for the RL meta-level controller to make accurate decisions. Preliminary experimental results illustrating the strength of meta-level control in agent reasoning and the effectiveness of the heuristics are provided.AssumptionsThe following assumptions are made in the framework de-scribed in this paper:The agents are cooperative and will prefer alternatives which increase social utility even if it is at the cost of decreasing local utility.Since the environment is cooperative,we can safely assume that the cumulative of meta-level control decisions at the individual agent level rep-resent the meta-level control reasoning process for a multi-agent system.An agent may concurrently pursue multiple high-level goals and the completion of a goal derives util-ity for the system or agent.The overall goal of the system or agent is to maximize the utility generated over somefi-nite time horizon.The high-level goals are generated by ei-ther internal or external events being sensed and/or requests by other agents for assistance.These goals must often be completed by a certain time in order to achieve any utility. It is not necessary for all high-level goals to be completed in order for an agent to derive utility from its actions.The partial satisfaction of a high-level goal is sometimes permis-sible while trading-off the amount of utility derived for de-crease in resource usage.The agent’s scheduling decisions involve choosing which of these high-level goals to pursue and how to go about achieving them.There can be non-local and local dependencies between tasks and methods. Local dependencies are inter-agent while non-local depen-dencies are intra-agent.These dependencies can be hard or soft precedence relationships.Coordination decisions in-volve choosing the tasks which require coordination and also which agent to coordinate with and how much effort must be spent on coordination.Scheduling and coordination actions do not have to be done immediately after there are requests for them and in some cases may not be done at all.There are alternative ways of completing scheduling and coordination activities which trade-off the likelihood of these activities re-sulting in optimal decisions versus the amount of resources used.We also make the simplifying assumption that negoti-ation results are binding and we assume that the agents will not decommit from their contract at later stages.Agent ArchitectureIn this section,we provide an overview of our architecture which provides effective meta-level control for bounded ra-tional agents.Figure1describes the controlflow within this proposed architecture.The number sequences describe the steps in a singleflow of control.At the heart of the system is the Domain Problem Solver(DPS).It receives tasks and other external requests from the environment(Step1).When an exogenous event such as arrival of a new task occurs,the DPS sends the corresponding task set,resource constraints as well constraints of other tasks which are being executed, and performance criteria to the meta-level controller(Step 2).The controller computes the corresponding state and de-termines the best action prescribed by the hand-generated heuristic policy for that particular task environment.The best action can be one of the following:to call one of the two domain schedulers on a subset of tasks;to gather more information to support the decision process;to drop the new task or to do nothing.The meta-level controller then sends the prescribed best action back to the DPS(Step2a).The DPS,based on the exact nature of the prescribed ac-tion,can invoke the complex scheduler,simple scheduler or coordination component(Step3)and receives the appro-priate output(Step3a).If the action is to invoke the complex scheduler,the scheduler component receives the task struc-ture and objective criteria as input and outputs the best satis-ficing schedule as a sequence of primitive actions.The com-plex scheduler can also be called to determine the constraints on which a coordination commitment is established.If the meta-level or the domain scheduler prescribe an action that requires establishing a commitment with a non-local agent, then the coordination component is invoked.The coordina-tion component receives a vector of commitments that have to be established and outputs the status of the commitments after coordination completes.The simple scheduler is in-voked by the DPS and receives the task structure and goal criteria.It uses pre-computed abstract information of the task to select the appropriate schedule whichfits the criteria. The DPS can invoke the execution component either to execute a single action prescribed by the meta-level con-troller or a schedule prescribed by the domain-level sched-uler(Step4).The execution results are sent back to the DPS(Step4a)where they are evaluated and if the execu-tion performance deviates from expected performance,the necessary measures are taken by the DPS.This work accounts for the cost at all three levels of the decision hierarchy-domain,control and meta-level con-trol activities.The cost of domain activities is modeled di-rectly in the task structures which describe the tasks.They are reasoned about by control activities like negotiation and scheduling.The cost of control activities are reasoned about by the meta-level control activities.Negotiation costs are reasoned about explicitly in this framework since they can be mod-eled as part of the domain activities needed to complete aFigure1:Control-flow in a bounded rational agent high-level goal.The negotiation tasks are split into an in-formation gathering phase and a negotiating phase,with the outcome of the former enabling the latter.The negotiation phase can be achieved by choosing a member from a fam-ily of negotiation protocols(7).The information gathering phase is modeled as a MetaNeg method in the task structure and the negotiation methods are modeled as individual prim-itive actions.Thus,reasoning about the costs of negotiation is done explicitly,just as it is done for regular domain-level activities.The MetaNeg method belongs to a special class of domain actions which request an external agent for a cer-tain set of information and it does not use local processor time.It queries the other agent and returns information on the agent’s expected utility from its tasks,expected comple-tion time of its tasks andflexibility of its schedule.This in-formation is used by the meta-level controller to determine the relevant control actions.However,reasoning about the cost associated with scheduling activities is implicit.Afixed cost is associated with each of the two schedulers and these costs affect the subsequent choice of domain activities made by the control activities.The earliest start time of domain activities are determined by the latestfinish times of their corresponding control activities.Meta-level control activities in this framework are mod-eled as inexpensive activities.The cost for meta-level con-trol in this framework are incurred by the computation of state features which facilitate the heuristic decision-making process.The state features and their functionality are de-scribed in greater detail later on in this section.The domain level scheduler depicted in the architecture will be an extended version of the Design-to-Criteria(DTC) scheduler(6).Design-to-Criteria(DTC)scheduling is the soft real-time process offinding an execution path through a hierarchical task network such that the resultant sched-ule meets certain design criteria,such as real-time dead-lines,cost limits,and utility preferences.It is the heart of agent control in agent-based systems such as the resource-Bounded Information Gathering agent BIG(2).Casting the language into an action-selecting-sequencing problem,the process is to select a subset of primitive actions from a set of candidate actions,and sequence them,so that the end result is an end-to-end schedule of an agent’s activities that meets situation specific design criteria.We also introduce a simple scheduler based on the use of abstractions of agent task structures.This will support reac-tive control for highly constrained situations.Abstraction is an offline process where potential schedules and their associ-ated performance characteristics for achieving the high level tasks are discovered for varying objective criteria.This is achieved by systematically searching over the space of ob-jective criteria.Also multiple schedules could potentially be represented by the same abstraction.The abstraction hides the details of these potential schedules and provides only the high level information necessary to make meta-level choices.When an agent has to schedule a task but doesn’t have the resources or time to call the complex domain-level scheduler,the generic abstraction information of the task structure can be used to provide the approximate schedule. Taxonomy of meta-level control decisionsWe now describe a taxonomy of the meta-level decisions in a multi-agent system using a simple example scenario. Consider a multi-agent system consisting of2agents A and B.The discussion will focus only on the various meta-level questions that will have to be addressed by agent A.T0and T1are the top-level tasks performed by agent A.Each top-level task is decomposed into two executable primitive ac-tions.In order to achieve the task,agent A can execute one or both of its primitive actions within the task deadline and the utility accrued for the task will be cumulative(de-noted by the sum function).Methods are primitive actions which can be scheduled and executed and are characterized by their expected utility,cost and duration distributions.For instance,the utility distribution of method M2described as,indicates that it achieves utility value of10 with probability0.9and utility of12with probability0.1. Utility is a deliberately abstract domain-dependent concept that describes the contribution of a particular action to over-all problem solving.There exists an enables relationship from task NX belonging to agent B to method M2belonging to agent A’s task T1.This implies that successful execution of NX by agent B is a precondition for agent A to execute method M2.In the remainder of this section,we enumerate the fea-tures computed when the meta-level control component is invoked.The cost of computing and reasoning about these state features reflect the cost of meta-level control reasoning. We then enumerate the various meta-level control decisions and the case-base of heuristics used to make the decisions. The following are some simple state features which are used in the heuristic decision making process of the meta-level controller.We use qualitative values such as high,medium and low,to represent the various factors which af-fect the heuristic features.The quantitative values such as quality of80versus quality of60were classified into qual-itative buckets(high versus medium quality)initially based on intuitions on the expected and preferred behavior of the system.They are verified by multiple simulation runs of the system on various test cases.F0:Current status of system This feature is represented as a3-tuple representing the NewItems Stack,Agenda and ScheduleStack where each entry in the tuple contains the number of items on the corresponding stack.The new items are the tasks which have just arrived at the agent from the environment.The agenda stack is the set of tasks which have arrived at the agent but whose reasoning has been delayed and they have not been scheduled yet.The schedule stack is the set of tasks currently being scheduled.Eg.means there are two new items which have arrived from the environment and there is one task being scheduled.F1:Relation of utility gain per unit time of a par-ticular task to that of currently scheduled task set: The of a task is the ratio ofto of that task.This feature compares the utility of a particular task to that of the existing task set and helps determine whether the new task is very valuable,moderately valuable or not valuable in terms of utility to the local agent.F2:Relation of deadline of a particular task to that of currently scheduled task set:This feature compares the deadline of a particular task to that of the existing task set and helps determine whether the new task’s deadline is very close,moderately close or far in the future.F3:Relation of priority of items on agenda to that of currently scheduled task set:This feature compares the average priority of the existing task set to the priority of the new task and helps determine whether the new task is very valuable,moderately valuable or not valuable in terms of utility to the local agent.Priority is a function of the utility and deadlines of the puting the average priority of a task set is a more complicated function than computing the priority of a single tasks since it involves recognizing dominance of individual tasks.The experiments described in this paper use the above four features.There are other features,simple and complex, which are enumerated below but yet to be implemented..F4:Percent of slack in local schedule:This feature is used to make a quick evaluation of theflexibility in the local schedule.The amount of slack in the local schedule allows the agent to accept new tasks and schedule them in the free slots as well as deal with unexpected meta-level control ac-tivities.F5:Percent of slack in other agent’s schedule:This feature is used to make a quick evaluation of theflexibility on the other agent’s schedule.The computation of feature F5is inexpensive since it is done after an information gather-ing phase,represented by a primitive action called MetaNeg which when executed will gather information on non-local agents which are potential coordination partners for the local agent.F6:Relation of utility gain per unit time of non-local task to non-local agent’s current task set:This feature compares the utility of a particular task to that of the existing task set of a non-local agent and helps determine whether the new task is very valuable,moderately valuable or not valuable with respect to the utility of the other agent.The computation of feature F6is inexpensive since it too is done after the information gathering phase.F7:Expected utility of current schedule item at cur-rent time:This is the expected utility of the current schedule item at time t as determined by the domain-level scheduler which uses expected value computations.F8:Actual utility of current schedule item at current time:This is the actual quality of the current schedule item at run time t.This feature is compared to F7in order to determine whether schedule execution is proceeding as ex-pected.If it is not proceeding as expected,a reschedule is initiated to prevent the agent from reaching a failure point from which recovery is not possible.Features F7and F8 will be computed at specified monitoring points.F9:Expected Rescheduling Cost with respect to a task set:This feature estimates the cost of rescheduling a task set and it depends on the size and quality accumulation factor of the task structure.It also depends on the horizon and effort parameters specified to the domain-level scheduler.F10:Expected DeCommitment Cost with respect to a particular task:This is a complex feature which estimates the cost of decommiting from a method/task by consider-ing the local and non-local down-stream effects of such a decommit.The domain-level scheduler could be invoked a number of times to compute this feature making it expensive to compute.F11:Relation of slack fragmentation in local schedule to new task:This is a complex feature which determines the feasibility offitting a new task given the detailed fragmen-tation of slack in a particular schedule.It involves resolving detailed timing and placement issues.F12:Relation of slack fragments in non-local agent to non-local task:This is a complex feature which deter-mines the feasibility offitting a new task given the detailed fragmentation of slack in a particular non-local schedule.It involves resolving detailed timing and placement issues.The following are some of the specific meta-level deci-sions that will be addressed by any individual agent.We describe how the heuristics determine the best action when certain exogenous events occur.The description is limited to reasoning about features F0-F4.Current work allows for reasoning about all12features.1.Arrival of a new task from the environment:When a newtask arrives at the agent,the meta-level control compo-nent has to decide whether to reason about it later;drop the task completely;or to do scheduling-related reason-ing about an incoming task at arrival time and if so,what type of scheduling-complex or simple.The decision tree describing the various action choices named A1-A9is shown in Figure2.Each of the meta-level decisions have an associated decision tree.As each exogenous event oc-curs for a particular environment,its corresponding de-cision tree is added incrementally to the parent MDP for[A7][A8][A9] Figure2:Decision tree when a new task arrivesthat environment and the optimal policy will be computed offline.Heuristic Rule:If the new task has very low or negligible priority and high opportunity cost with respect to taking resources away from future higher priority tasks, then it should be discarded.If the incoming task has very high priority,in other words,the expected utility is very high and it has a relatively close deadline,then the agent should override its current schedule and schedule the new task immediately.If the deadline is very tight the agent will uses the abstraction-based simple scheduler;else,it will use the more complex scheduler.If the current sched-ule has average utility that is significantly higher than the new task and the average deadline of the current schedule is significantly closer than that of the new task,then rea-soning about the new task should be postponed till later.If the new task is scheduled immediately,the scheduling action costs time,and there are associated costs of drop-ping established commitments if the previous schedule is significantly revised or completely dropped.These costs are diminished or avoided completely if the task reason-ing is postponed to later or completely avoided if the task is dropped.2.Decision on whether to negotiate:The meta level con-troller will decide to negotiate based on the information returned by the MetaNeg action.It queries the other agent and returns information on the agent’s expected utility from its tasks,expected completion time of its tasks and flexibility of its schedule.We know that method M2in agent A is enabled by task NX belonging to agent B.The benefit from including method M2in agent A’s schedule is that it increases its total utility.However,it also re-quires agent A and B to negotiate over the completion time of task NX by agent B and this negotiation has an associated cost as well as there is a resource cost to theagent which agrees to the contract.Heuristic Rule:If the other agent’s current expected utility is much lower than the results of the negotiation,then the local agent will ini-tiate negotiation.Negotiation is also initiated if the other agent’s tasks have high utility but the deadlines are far enough in the future to permit the other agent to execute the enabling task.If the other agent’s tasks have higher priority than the local task,then the negotiation option is dropped.3.Choice of negotiation protocol:When an agent decidesto negotiate,it should also decide whether to negotiate by means of a single step or a multi-step protocol that may require a number of negotiation cycles tofind an ac-ceptable solution or even a more expensive search for a near-optimal solution.The single shot protocol is quick but has a higher chance of failure where as a more com-plex protocol takes more time and has a higher chance of success Heuristic Rule:If the agent receives high utility from the results of the negotiation,then the agent should choose the more effective albeit more expensive protocol.The protocol which has a higher guarantee of success re-quire more resources,more cycles and more end-to-end time in case of multi-step negotiation and higher compu-tation power and time in case of near-optimal solutions.(The end-to-end time is proportional to the delay in be-ing able to start task executions).If the agent does not have too much resources to expend on the negotiation or if there is a very slight probability that the other agent will accept the contract,then the local agent should choose the single shot protocol.4.Failure of a negotiation to reach a commitment:If thenegotiation between two agents using a particular nego-tiation protocol fails,the initiating agent should decide whether to retry the negotiation;whether to use the same protocol or an alternate protocol with the same agent or alternate agents and how many such retries should take place?Heuristic Rule:If negotiation is preferred(the agent will receive high utility as a result of the negotia-tion),then a more complex negotiation protocol is chosen since it has a higher probability of succeeding.Since re-sources have already been spent onfiguring out a solution to the negotiation,it may be profitable to put in a little more effort and achieve a solution.If there is a very slight or no probability offinding an acceptable commitment, then resources which can be profitably spent on other so-lution paths are being wasted and the agent mightfind itself in a dead-end situation with no resources left for an alternate solution.So the negotiation option should be dropped.Two other meta-level decisions which are being devel-oped determine the parameters for invoking the domain scheduler including scheduler horizon,scheduler effort and slack amount in overall schedule and also determine whether to invoke the domain level scheduler for a reschedule since the performance of the agent is not going as expected.C ontrol Activity is thefixed control action used by the agentExperimental ResultsFor the purposes of this paper,we used the environment in-troduced in the previous section with randomly generated which adheres to the above mentioned characteristics.The maximum possible utility from task T0is23.0and mini-mum is17.0;the maximum from task T1is56.0and mini-mum utility is12.0;NX has a deterministic utility of10.00. We make the simplifying assumption that task NX arrives at agent B only as a result of a successful negotiation with agent A.There are four possible meta-decisions upon arrival of a new task:NTCS,New Task Complex Scheduling in-vokes the complex DTC scheduler on the new task only and has a time cost of2;Drop,this causes the agent to drop the new task and not reason about it ever again has a time cost of0;ATCS,All Task Complex Scheduling invokes the complex DTC scheduler on the new task as well as all other tasks which are on the agenda or in partial execution and has a time cost of3;and SS,Simple Scheduling invokes the sim-ple abstraction based analysis on the new task only and has a time cost of1.There are two possible options for Nego-tiation:NM1,Negotiation Mechanism1which is the sim-ple single-shot protocol and NM2,Negotiation Mechanism 2which is the more complex multi-shot protocol.The design criteria in these experiments is to maximize overall utility over afinite horizon.Individual tasks have hard deadlines associated with them.It is assumed that if a task has not accrued utility by its deadline,it receives a utility of zero.This simple design criteria setting is one that lends itself to meta-level control as the existence of a hard deadlines(in contrast to a soft preference,e.g.,soft dead-line or no deadlines)make processor and other resources valuable commodities requiring a the non-myopic reasoning provided by the meta-level control component.The results for the experiments on agents which have meta-reasoning capabilities are shown in Table1and the results on agents which have no meta-level reasoning capa-bilities are shown in Table2.The above described scenario is used in both cases.All domain,control and meta-level actions have a time cost associated with them which are re-flected in the results.Consider Table1where each row represents a specific task arriving at the specified agent at the associated arrival time with a deadline.The task names are augmented with the arrival count to differentiate between various instances of the same task.For eg.Row4describes task TO arriving at agent A as its fourth task at time55with a deadline of80.。
