以線性混合模型探討有機磷中毒病人血清膽鹼酯酶與時間之動態關係

Abstract

簡介 — 有機磷在世界各地廣泛地被使用於農業及環境衛生上，而中毒事件也層出不窮。其主要機轉為身體膽鹼酯酶受有機磷抑制導致乙醯膽鹼過度刺激結後神經元或器官受器產生如腺體分泌物增加、肌肉無力、甚至昏迷等種種症狀。有機磷中毒病患血清膽鹼酯酶(SChE)在中毒後隨時間的變化相當程度代表了病患復元的情況，但每位病人的SChE具重覆測量的特性，相同病人檢驗值間將會出現相關性問題。線性混合模型為醫學上常用來處理重覆測量資料的方法之一，最主要是在模型中加入隨機效應來克服組內資料之間的相關性，並由隨機誤差中加以分離。本研究主旨為利用統計混合模型來探討有機磷中毒病患血清膽鹼酯酶(SChE)的恢復情形。 方法 —本研究之假說為有機磷中毒病患血清膽鹼酯酶與時間呈某種程度的相關性；不同病人血清膽鹼酯酶的初值與上升速度存在著個體差異；以及部份相關臨床因子會對膽鹼酯酶上升趨勢有影響。利用回溯性研究設計，收集自2000年至2010年因急性有機磷中毒至某醫學中心急診就診之成人病患共212位，由排除條件刪除81位餘下131位病患相關資料及共859筆SChE濃度資料。並收集病患個人基本資料、第一次生命徵象及嚴重度評估、各項檢查檢驗數據、中毒相關因素、接受的治療、各次SChE檢驗時間與結果、及預後等變項作分析。統計方法使用混合模型中的隨機係數模型，加入個人隨機截距效應及時間的隨機斜率，再放入臨床相關因子，並與一般線性回歸模型作比較。 結果 —有機磷中毒病患不論中毒物質種類、過去病史、就醫時間、症狀等等均存在很大的差異性，但多數病人的初次SChE濃度均有明顯的降低。SChE的分析呈現右偏分佈，經過對數轉換後(LSChE)分佈趨於對稱。LSChE的相關性矩陣顯示出各次測量值間確有正相關，且相關性由相鄰測量開始依測量間隔時間增加而逐漸遞減。混合模型的模型配適度顯著地較固定效應模型改善許多，可見中毒病患的嚴重度及復原過程的確存在很大的個人差異。混合模型配適的結果顯示病人LSChE濃度高低與時間、有機磷種類、初次測量值有關，而其隨著時間上升的斜率則與有機磷種類、初次測量值、及疾病嚴重度有關。在有機磷種類方面怖飛松(profenophos)有較低的SChE濃度、陶斯松(chlorpyrifos)的SChE上升趨勢較緩慢、達馬松(methamidophos)則上升趨勢較快。疾病嚴重程度較高、初次測量值較高的病患有較平緩的上升趨勢。性別、前24小時解毒劑2-PAM使用劑量、延遲就醫時間、與時間二次方均與上升趨勢相關性不顯著。 結論 — 本論文探討了有機磷中毒病患血清膽鹼酯酶濃度在復元過程的上升趨勢，研究結果可作為未來建構有機磷中毒病患復元模式的基礎。

Introduction—Organophosphate poisoning (OPP) occurs worldwide, accounting for 200,000 deaths annually in developing countries. Organophosphorus compounds inhibit the activity of cholinesterase and result in the over-activation of acetylcholine on autonomic ganglia and end organs. Serum cholinesterase (SChE) is of diagnostic value in OPP patients and often checked repeatedly during the treatment course. Most of the previous studies on OPP patients focused on mortality and disease severity, while the recovery pattern of SChE has been rarely addressed. To deal with the repeated measured SChE, correlation within the same patient must be taken into account. The linear mixed model is one of the most commonly used methods in repeated measures of normal data. It adds random effects in order to capture variance resulting from between-cluster level as well as intra-cluster correlation. This study aimed to investigate the dynamic relationship between SChE and time in OPP patients using linear mixed models. Material and Methods—This study is a retrospective cohort study, utilizing medical records from a medical center in Taoyuan, Taiwan. A total of 212 adult patients who visited the emergency room between 2000 and 2010 due to acute OPP were included, and 131 patients were analyzed after exclusion of 81 patients by criteria. Information regarding basic personal characteristics, first vital signs and severity scores, lab data, type and ingestion amount of organophosphate, treatment, and serial SChE value were collected. Random coefficient model with random intercept and random slope of time were added to address the dynamic relationships of SChE with time and associated factors. Results—Organophosphate-poisoned patients differ significantly in baseline characteristics, but most patients had lowered initial SChE levels. The histogram of SChE revealed a positively skewed distribution and thus log-transformed SChE (LSChE) was used as the dependent variable. Serial LSChE showed a positive correlation with decreasing magnitude as the time gap increased. The goodness of fit significantly improved after random intercepts and random slopes of time were added to a linear mixed model. Time, type of organophosphate, and first LSChE level were independently related with LSChE level, while type of organophosphate, first LSChE level, and disease severity score were significantly related with the slope of time. Profenophos has lowest LSChE levels among all types of organophosphate. Chlorpyrifos and methamidophos had significant slower and faster rates of LSChE recovery compared with other organophosphates, respectively. Sex, dose of 2-PAM during the initial 24 hours, delay of medical assistance, and quadratic form of time did not significantly affect the recovery of LSChE. Conclusion—This study constructed a model for recovery of SChE in OPP patients. Several major determinants responsible for the recovery of SChE were also identified. It can be used as a prototype model in further studies on OPP. Clinicians may also make use of such information to guide clinical decisions during initial period of treatment.