基于贝叶斯多变量联合模型的体检人群脑卒中发病风险因素的纵向研究

背景 脑卒中是目前影响人类健康的主要公共卫生问题之一;健康体检纵向数据累积了大量的健康信息,由于缺失数据多、样本量小等诸多问题,导致其利用率低、重要信息未能得到充分挖掘,进而对常见慢性病的有效防控等工作带来一定困难。目的 基于贝叶斯多变量联合模型,探讨体检人群脑卒中发病风险因素,为慢性病风险因素分析提供新的方法。方法 本研究使用空军军医大学西京医院健康医学中心2008—2015年的体检资料。随访情况：以首次发生脑卒中为结局事件,发生结局事件立即停止随访;若未发生,到2015年体检信息收集完成后结束随访;体检间隔时间为1年。依据随访过程中是否发生脑卒中分为脑卒中组和非脑卒中组。纵向观察变量包括总胆固醇（TC）、三酰甘油（TG）、低密度脂蛋白胆固醇（LDL-C）、高密度脂蛋白胆固醇（HDL-C）、体质指数（BMI）和收缩压（SBP）。采用多因素Cox回归模型分析基线情况对脑卒中结局事件的影响;采用贝叶斯多变量联合模型,分析随访过程中TC、TG、LDL-C、HDL-C、BMI和SBP的纵向变化轨迹对脑卒中发病的影响。结果 本研究共纳入234例研究对象,1 581条纵向随访记录,平均随访时间为...     

Latent variable discovery in classification models

The naive Bayes model makes the often unrealistic assumption that the feature variables are mutually independent given the class variable. We interpret a violation of this assumption as an indication of the presence of latent variables, and we show how latent variables can be detected. Latent variable discovery is interesting, especially for medical applications, because it can lead to a better understanding of application domains. It can also improve classification accuracy and boost user confidence in classification models.     

2011-2018年中国手足口病发病的时空特征及影响因素研究

目的 分析我国手足口病的时空特征,探讨社会、经济、人口和卫生服务等因素对我国手足口病发病的影响,为手足口病疫情防控提供参考依据。方法 采用贝叶斯时空模型对手足口病数据进行拟合,评估手足口病的时空变化,并识别手足口病发病风险与社会、经济、人口和卫生服务等因素的潜在关联。结果 2011-2018年,我国共报告手足口病例17 118 050例,死亡2 283例,2011-2014年报告发病率呈波动上升趋势,2014-2018年报告发病率呈波动下降趋势,报告死亡率一直呈波动下降趋势。手足口病的发病具有空间聚集性,报告发病率最高的地区为华南地区且为热点区域及高风险区域,报告发病率最低的为西北地区,冷点区域及低风险区域也集中在其部分区域。手足口病发病风险与人均地区生产总值（RR=3.54）、每万人规模以上工业企业单位数（RR=1.61）、城市化率（RR=3.00）、人口出生率（RR=2.36）、每万人医疗机构床位数（RR=3.40）和人均公园绿地面积（RR=0.57）有关。结论 2011-2018年我国手足口病防控重点区域为东南沿海地区,在加快城市化进程的同时需考虑增加人均公园绿地面积,以降低手足口病的发病率。     

The geographic distribution of un-immunized children in Ontario,Canada: Hotspot detection using Bayesian spatial analysis

BackgroundIn Ontario, Canada, little is currently known about the extent to which un-immunized children may cluster geographically. Our objectives were to: describe the geographic distribution of fully un-immunized children; identify geographic clusters (hotspots) of un-immunized children; and to characterize the contribution of spatial effects and covariates on hotspots, where found.MethodsOur analytic cohort consisted of Ontario students aged 7–17 years in the 2016–2017 school year. We defined students as un-immunized if they had zero doses of any vaccine and a non-medical exemption recorded in Ontario’s registry. We calculated unadjusted proportions of un-immunized students by Census Subdivision (CSD) and then used a sequential approach to identify hotspots starting first with hotspot identification at the CSD level and then probed identified hotspots further by Dissemination Area (DA) and including covariates. Hotspots were identified using the Besag-York-Mollie Bayesian spatial model and were defined as areas with >95% probability of having two times the proportion of un-immunized students, relative to the province overall.ResultsWe identified 15,208 (0.94%) un-immunized children within our cohort consisting of more than 1.61 million students. Unadjusted proportions of un-immunized students varied greatly by geography, ranging from 0% to 21.5% by CSD. We identified 16 hotspot CSDs which clustered in five distinct areas, all of which were located in southern Ontario. The contribution of covariates and spatial effects on the risk of having un-immunized students varied greatly across hotspot areas.ConclusionsAlthough the provincial proportion (0.94%) of un-immunized students is small, geographical clustering of such students is evident in Ontario and in some areas presents an important risk for future outbreaks. Further qualitative work within these hotspot areas would be a helpful next step to better characterize the factors associated with vaccine refusal in these communities.     

Molecular evolution of SARS-CoV-2 from December 2019 to August 2022

Severe acute respiratorysyndrome coronavirus-2 (SARS-CoV-2) pandemic spread rapidly and this scenario is concerning worldwide, presenting more than 590 million coronavirus disease 2019 cases and 6.4 million deaths. The emergence of novel lineages carrying several mutations in the spike protein has raised additional public health concerns worldwide during the pandemic. The present study review and summarizes the temporal spreading and molecular evolution of SARS-CoV-2 clades and variants worldwide. The evaluation of these data is important for understanding the evolutionary histories of SARSCoV-2 lineages, allowing us to identify the origins of each lineage of this virus responsible for one of the biggest pandemics in history. A total of 2897 SARS-CoV-2 whole-genome sequences with available information from the country and sampling date (December 2019 to August 2022), were obtained and were evaluated by Bayesian approach. The results demonstrated that the SARS-CoV-2 the time to the most recent common ancestor (tMRCA) in Asia was 2019-12-26 (highest posterior density 95% [HPD95%]: 2019-12-18; 2019-12-29), in Oceania 2020-01-24 (HPD95%: 2020-01-15; 2020-01-30), in Africa 2020-02-27 (HPD95%: 2020-02-21; 2020-03-04), in Europe 2020-02-27 (HPD95%: 2020-02-20; 2020-03-06), in North America 2020-03-12 (HPD95%: 2020-03-05; 2020-03-18), and in South America 2020-03-15 (HPD95%: 2020-03-09; 2020-03-28). Between December 2019 and June 2020, 11 clades were detected (20I [Alpha] and 19A, 19B, 20B, 20C, 20A, 20D, 20E [EU1], 20F, 20H [Beta]). From July to December 2020, 4 clades were identified (20J [Gamma, V3], 21 C [Epsilon], 21D [Eta], and 21G [Lambda]). Between January and June 2021, 3 clades of the Delta variant were detected (21A, 21I, and 21J). Between July and December 2021, two variants were detected, Delta (21A, 21I, and 21J) and Omicron (21K, 21L, 22B, and 22C). Between January and June 2022, the Delta (21I and 21J) and Omicron (21K, 21L, and 22A) variants were detected. Finally, between July and August 2022, 3 clades of Omicron were detected (22B, 22C, and 22D). Clade 19A was first detected in the SARS-CoV-2 pandemic (Wuhan strain) with origin in 2019-12-16 (HPD95%: 2019-12-15; 2019-12-25); 20I (Alpha) in 2020-11-24 (HPD95%: 2020-11-15; 2021-12-02); 20H (Beta) in 2020-11-25 (HPD95%: 2020-11-13; 2020-11-29); 20J (Gamma) was 2020-12-21 (HPD95%: 2020-11-05; 2021-01-15); 21A (Delta) in 2020-09-20 (HPD95%: 2020-05-17; 2021-02-03); 21J (Delta) in 2021-02-26 (2020-11-02; 2021-04-24); 21M (Omicron) in 2021-01-25 (HPD95%: 2020-09-16; 2021-08-08); 21K (Omicron) in 2021-07-30 (HPD95%: 2021-05-30; 2021-10-19); 21L (Omicron) in 2021-10-03 (HPD95%: 2021-04-16; 2021-12-23); 22B (Omicron) in 2022-01-25 (HPD95%: 2022-01-10; 2022-02-05); 21L in 2021-12-20 (HPD95%: 2021-05-16; 2021-12-31). Currently, the Omicron variant predominates worldwide, with the 21L clade branching into 3 (22A, 22B, and 22C). Phylogeographic data showed that Alpha variant originated in the United Kingdom, Beta in South Africa, Gamma in Brazil, Delta in India, Omicron in South Africa, Mu in Colombia, Epsilon in the United States of America, and Lambda in Peru. The COVID-19 pandemic has had a significant impact on global health worldwide and the present study provides an overview of the molecular evolution of SARS-CoV-2 lineage clades (from the Wuhan strain to the currently circulating lineages of the Omicron).     

A New Statistical Method for Dose-Finding Based on Efficacy and Toxicity in Early Phase Clinical Trials

Most statistical methods for dose-finding in phase I clinical trials determine a maximum tolerable dose based on toxicity while ignoring efficacy. Most phase II designs assume that an acceptable dose has been determined and aim to estimate treatment efficacy, possibly with early stopping rules for safety monitoring. The purpose of this paper is to describe a new statistical strategy for dose-finding in single-arm clinical trials where patient outcome is characterized in terms of both response and toxicity. The strategy, which may be considered a phase I/II hybrid, was first proposed by Thall and Russell [1] and subsequently modified by Thall [2]. The underlying mathematical model expresses the probabilities of response and toxicity as interdependent functions of dose. The method is based on fixed standards for the minimum probability of response and the maximum probability of toxicity appropriate for the particular trial. The best acceptable dose is chosen for each successive patient cohort adaptively, based on the fixed standards and the dose-outcome data from patients treated previously in the trial. The scientific goals are to select one best acceptable dose for future patients and to estimate the response and toxicity probabilities at that dose, or to stop the trial early if it becomes sufficiently unlikely that any dose is both safe and efficacious. An application of the method to a trial of donor lymphocyte infusion as salvage therapy for chemo-refractory AML/MDS patients is described. To illustrate the method's flexibility and potential breadth of application, two additional examples are provided, including a hypothetical trial in which a 5% response rate is of interest.     

A Bayesian dosing method for carboplatin given by continuous infusion for 120 h

Carboplatin (CBDCA), an analogue of cisplatin, exhibits reduced toxicity but wide interpatient variability of its pharmacokinetic parameters. Individualization of the CBDCA dose is therefore necessary. Although various formulas have been developed for this purpose, major side effects have been reported on CBDCA administration by short-term infusion (0.5 or 1h). We therefore propose a new schedule of CBDCA administration. Instead of a dosing method based on the estimation of renal function when a classic administration schedule is used, we propose a pharmacokinetic dosing method (Bayesian method), whereby CBDCA is given by continuous infusion for 120 h. First, CBDCA was given to 21 patients to determine the population pharmacokinetic parameters of carboplatin. Then, on the basis of total platinum plasma concentration measurements and Bayesian estimation of pharmacokinetic parameters, it was possible to individualize the CBDCA dose within the first 24 h of the infusion. This new protocol for CBDCA administration was evaluated in 36 new patients (60 courses). Three theoretical end points at the end of the infusion were considered. For a given theoretical end point, 20 courses were taken into account. The theoretical end points (i.e., 1, 1.5, and 1.8 mg/l) were compared with the concentrations measured at the end of the infusion, which were 0.99 ± 0.10, 1.41 ± 0.13, and 1.72 ± 0.20 mg/l, respectively. This Bayesian dosing method can easily be used in clinical practice, and the determination of predictive performances has shown that the method is precise and unbiased. With no more toxicity or practical difficulties than those produced by other methods, and with acceptable tolerance, it was possible to reach a median dose that was 20% higher than the usual dose (484 ± 190 mg/m² as compared with 400 mg/m²). In conclusion, this new schedule of CBDCA administration appears to be interesting in terms of tolerance. However, new studies are required to confirm that this new scheme leads to equal or better efficacy than the classic protocol. Received: 10 December 1995 / Accepted: 15 December 1996     

A limited-sampling method for evaluation of the area under the curve of ultrafilterable carboplatin in children

Carboplatin is a widely used cytotoxic agent in numerous solid tumors of children. Since there is a large degree of interpatient variability in the area under the curve of free carboplatin for a given dose of the drug, the current tendency is to adjust the carboplatin dose so as to reach a target area under the curve rather than to determine a carboplatin dose on the basis of the body surface area. A limited-sampling method was developed for estimation of the ultrafilterable carboplatin area under the curve and for adjustment of the carboplatin dose on subsequent treatments. Population parameters were obtained from 16 children (reference group). We used the maximum a posteriori (MAP) Bayesian approach on 15 children with complete carboplatin pharmacokinetic data (test group). Two blood samples were sufficient to obtain reliable prediction of the area under the curve. The best sampling times were: (a) 30 min after the end of the infusion and (b) 5 h after the end of the infusion. On the basis of these data it is possible to prescribe prospectively a target area under the curve for free carboplatin given in a fractionated daily infusion and to adapt the carboplatin dose directly to ultrafilterable carboplatin measurements. Received: 8 June 1997 / Accepted: 9 January 1998     

Methodology in meta–analysis: a study from Critical Care meta–analytic practice

Methodological aspects of meta-analytic practice, heterogeneity, publication bias, metaregression and effect metric, were investigated in 14 meta-analyses reflecting major therapeutic concern in Critical Care practice. Compared with the standard Q test, the exact Zelen test was more sensitive in identifying heterogeneity. Assessment of heterogeneity impact by the I ² statistic was consistent with inferences afforded by both the Q and Zelen test. Publication bias was subject to test and metric determination: funnel plots exhibited variable asymmetry across studies and between metrics; the regression asymmetry test appeared more sensitive than the rank correlation test; the “trim and fill” method was the most sensitive, but suggested, on the basis of quantification of the effects of potentially missing studies, that meta-analyses may be resistant to such missingness. Metaregression of treatment effect against control risk using Bayesian hierarchical regression in all metrics (log odds ratio, log risk ratio and RD) suggested that naïve linear regression approaches over-diagnosed significant relationships and exhibited regression dilution. Heterogeneity, publication bias and risk related treatment effects all demonstrate estimator and metric dependence; the RD metric would appear the most capricious in this regard.     

贝叶斯法估算丙戊酸药动学参数及个体化给药

目的:采用癫痫患者丙戊酸群体药动学参数结合贝叶斯(Bayesian)法估算癫痫患者丙戊酸的个体药动学参数;制定或优化欲达预期血药浓度所应实施的给药方案,使癫痫患者丙戊酸给药有效、合理、毒副作用小.方法:癫痫患者口服丙戊酸达稳态,取其每天早晨服药前10 min血样57人次,用荧光偏振免疫法(Fluorescence Polarization Immunoassay,FPIA)测得血清中丙戊酸和游离丙戊酸血药浓度谷值.用Bayesian法估算其药动学参数,并用逐步回归法分析个体的性别、年龄等18种因素对其药动学参数的影响.结果:按口服一房室一级吸收和消除的开放模型用Bayesian法估算得丙戊酸药动学参数清除率CL,平均值为(8.7±0.6)mL·h-1·kg-1,逐步回归方程:CL(mL·h-1·kg-1)=3.972 1.631X5 1.608X11;稳态血药浓度谷值Cp0逐步回归方程:Cp0(mg·L-1)=57.58-0.71X4-12.145X5 2.705X7 0.403x17.式中X4表示体重(kg),X5表示身高(cm)体重(kg)比,X7表示每日每千克体重的给药剂量(mg·d-1·kg-1),x11表示当合并用苯妥英钠时系数为1,否则为0,x17表示血清肌酐(μmol·L-1).实测丙戊酸稳态血药浓度谷值(52.4±5.4)mg·L-1与估算值(53.5±5.2)mg·L-1间差异无统计学意义,P＞0.05.结论:可采用Bayesian法估算癫痫患者丙戊酸动力学参数和进行个体化给药的拟定或调整给药方案.