Application of Bayesian additive regression tree model in the evaluation of individualized efficacy of hypertension drugs
ZHAO Hong-yu and FENG Yi-ping contributed equally to this article
More Information
-
摘要:
目的 以高血压合并糖尿病人群为例，应用贝叶斯累加回归树(Bayesian additive regression tree, BART)模型评价卡托普利及其联合尼群地平用药对血压控制的个性化疗效，并识别高收益患者亚组特征。 方法 纳入2011年1月至2015年7月山东省胶南市“全人群高血压、糖尿病综合防治项目”中的高血压合并糖尿病患者，按不同用药方式分为三组，采用倾向性得分随机化匹配形成可比样本后，建立BART模型探索用药的个性化疗效。 结果 在单用卡托普利与不用药、联合用药与不用药，联合用药与单用卡托普利三组对比中，三组模型曲线下面积(area under curre, AUC)及其95% CI分别为0.710(0.686~0.734)、0.796(0.754~0.838)、0.768(0.725~0.812)。对于大多数患者，联合用药效果优于单用卡托普利和不用药，其中，相比不用药者，SBP较低且有良好运动习惯是单用卡托普利和联合用药疗效更佳者的共同特征。 结论 构建的BART模型用于预测高血压合并糖尿病患者的高血压药物个性化疗效效果良好，并且能够用于总结高收益亚组特征，为精准医疗在高血压治疗中的实践提供依据。 Abstract:Objective To evaluate the personalized efficacy of captopril and its drug combination with nitrendipine on blood pressure control by Bayesian additive regression tree (BART) model, and to identify subgroup characteristics of high-yield patients, taking people with hypertension and diabetes as samples. Methods According to different medication modes, the patients with hypertension and diabetes who were included in the "Comprehensive Prevention and Control Project of Hypertension and Diabetes for All Populations" in Jiaonan, Shandong Province from January 2011 to July 2015 were divided into three groups. The propensity score was used to randomize the matching formation and to get comparable samples, then BART model was established to explore the personalized efficacy of medication. Results In the comparison between captopril alone and no medication, combination medication and no medication, combined medication and captopril alone, the area under curre (AUC) of the three groups of models and their 95% CI were 0.710 (0.686-0.734), 0.796 (0.754-0.838) and 0.768 (0.725-0.812), respectively. And for most patients, the combined medication effect was better than the single captopril and no medicine. Among them, lower systolic blood pressure and good exercise habits were the common characteristics of those who used captopril alone and combination drugs that had better efficacy compared with those who did not use drugs. Conclusion The constructed BART model is used to predict the personalized efficacy of hypertension drugs in patients with hypertension and diabetes, which can be used to summarize the characteristics of high-yield subgroups, and also provide the basis for the practice of precision medicine in the treatment of hypertension. -
Key words:
- Bayesian additive regression tree /
- Hypertension /
- Captopril /
- Precision medicine
-
图 1 2011年1月至2015年7月胶南市三组患者高血压控制概率差值分布
注：图A：组1患者单用卡托普利减去不用药的高血压控制概率差值分布；图B：组2患者联合用药减去不用药的高血压控制概率差值分布；图C：组3患者联合用药减去单用卡托普利的高血压控制概率差值分布。
Figure 1. Distribution of probability of hypertension control difference among the three groups of patients in Jiaonan from January 2011 to July 2015
图 2 2011年1月至2015年7月胶南市三组患者BART模型的ROC曲线
注：图A：组1单用卡托普利与不用药的ROC曲线；图B：组2联合用药与不用药的ROC曲线；图C：组3联合用药与单用卡托普利的ROC曲线。
Figure 2. ROC curves of BART model of three groups of patients in Jiaonan from January 2011 to July 2015
图 3 2011年1月至2015年7月胶南市三组患者在不同治疗方案下的预测血压控制概率分布
注：图A：组1在不同处方假设时患者高血压控制的收益预测；图B：组2在不同处方假设时患者高血压控制的收益预测；图C：组3在不同处方假设时患者高血压控制的收益预测。
Figure 3. Probability distribution of predicted blood pressure control in three groups under different treatment options in Jiaonan from January 2011 to July 2015
表 1 倾向性评分匹配后各组特征比较[n(%)]
Table 1. Comparison of characteristics of each group after propensity score matching [n(%)]
变量 组1 组2 组3 不用药(n=1 302) 卡托普利(n=1 302) P值 不用药(n=267) 尼群地平+卡托普利(n=267) P值 卡托普利(n=267) 尼群地平+卡托普利(n=267) P值 男性 424(32.57) 427(32.80) 0.900 82(30.71) 75(28.09) 0.506 67(25.09) 74(27.72) 0.492 年龄(岁) 67.71±11.38 67.74±10.40 0.940 69.37±11.45 69.27±9.79 0.916 68.67±10.25 69.26±9.81 0.501 SBP(x±s, mm Hg) 149.80±10.87 150.30±10.36 0.257 152.60±12.63 151.40±12.58 0.270 150.90±10.85 151.70±12.65 0.448 DBP(x±s, mm Hg) 87.37±8.04 87.42±8.06 0.882 88.00±8.86 87.99±9.10 0.981 87.61±8.04 88.05±9.18 0.558 BMI(x±s, kg/m2) 25.22±3.33 25.15±3.29 0.585 25.18±4.09 25.37±2.86 0.532 25.51±3.13 25.38±2.86 0.602 运动 746(57.30) 663(50.92) 0.001 157(58.80) 150(56.18) 0.540 144(53.93) 150(56.18) 0.602 脑卒中 40(3.07) 46(3.53) 0.511 15(5.62) 16(5.99) 0.853 17(6.37) 16(5.99) 0.857 冠心病 239(18.30) 236(18.07) 0.879 47(17.60) 55(20.60) 0.379 63(23.60) 55(20.60) 0.404 使用抗血小板药物 37(2.84) 311(23.89) ＜0.001 9(3.37) 126(47.19) ＜0.001 64(23.97) 126(47.19) < 0.001 
下载: 导出CSV
-
[1] Yang W, Lu J, Weng J, et al. Prevalence of diabetes among men and women in China[J]. N Engl J Med, 2010, 362(12): 1090-1101. DOI: 10.1056/nejmoa0908292. [2] 《中国高血压防治指南》修订委员会. 中国高血压防治指南2018年修订版[J]. 心脑血管病防治, 2019, 19(1): 1-44. https://www.cnki.com.cn/Article/CJFDTOTAL-XIXG201901003.htmRevision Committee of Chinese Guidelines for Prevention and Treatment of Hypertension. 2018 guidelines for prevention and treatment of hypertension in China revised[J]. Prevention and Treatment of Cardio-Cerebrai-Vascuiar Disease, 2019, 19(1): 1-44. https://www.cnki.com.cn/Article/CJFDTOTAL-XIXG201901003.htm [3] Ferrannini E, Cushman WC. Diabetes and hypertension: the bad companions[J]. Lancet, 2012, 380(9841): 601-610. DOI: 10.1016/S0140-6736(12)60987-8. [4] 孙宁玲, 王鸿懿. 高血压合并2型糖尿病患者的血压控制专家指导意见(2013版)[J]. 中华高血压杂志, 2013, 21(6): 522-525. DOI: 10.16439/j.cnki.1673-7245.2013.06.004.Sun NL, Wang HY. Expert guidance on blood pressure control in patients with hypertension complicated with type 2 diabetes (2013 Edition)[J]. Chin J Hypertens, 2013, 21(6): 522-525. DOI: 10.16439/j.cnki.1673-7245.2013.06.004. [5] 国家卫生计生委合理用药专家委员会, 中国医师协会高血压专业委员会. 高血压合理用药指南(第2版)[J]. 中国医学前沿杂志(电子版), 2017, 9(7): 28-126. DOI: 10.12037/YXQY.2017.07-07.Expert Committee on Rational Drug Use of National Health and Family Planning Commission, Hypertension Committee of Chinese Medical Doctor Association. Guidelines for rational use of hypertension (2 Edition)[J]. Chinese Journal of the Frontiers of Medical Science (Electronic Version), 2017, 9(7): 28-126. DOI: 10.12037/YXQY.2017.07-07. [6] Tiptaradol S, Aekplakorn W. Prevalence, awareness, treatment and control of coexistence of diabetes and hypertension in Thai population[J]. Int J Hypertens, 2012, 2012: 386453. DOI: 10.1155/2012/386453. [7] McLean DL, Simpson SH, McAlister FA, et al. Treatment and blood pressure control in 47, 964 people with diabetes and hypertension: a systematic review of observational studies[J]. Can J Cardiol, 2006, 22(10): 855-860. DOI: 10.1016/s0828-282x(06)70304-x. [8] Burke JF, Sussman JB, Kent DM, et al. Three simple rules to ensure reasonably credible subgroup analyses[J]. BMJ, 2015, 351: h5651. DOI: 10.1136/bmj.h5651. [9] Jones HE, Ohlssen DI, Neuenschwander B, et al. Bayesian models for subgroup analysis in clinical trials[J]. Clin Trials, 2011, 8(2): 129-143. DOI: 10.1177/1740774510396933. [10] Lipsky AM, Gausche-Hill M, Vienna M, et al. The importance of "shrinkage" in subgroup analyses[J]. Ann Emerg Med, 2010, 55(6): 544-552. e3. DOI: 10.1016/j.annemergmed.2010.01.002. [11] Xu YX, Trippa L, Müller P, et al. Subgroup-based adaptive (SUBA) designs for multi-arm biomarker trials[J]. Stat Biosci, 2016, 8(1): 159-180. DOI: 10.1007/s12561-014-9117-1. [12] Chipman HA, George EI, McCulloch RE. BART: Bayesian additive regression trees[J]. Ann Appl Stat, 2010, 4(1): 266-298. DOI: 10.1214/09-aoas285. [13] Sivaganesan S, Müller P, Huang B. Subgroup finding via Bayesian additive regression trees. [J]. Stat Med, 2017, 36(15): 2391-2403. DOI: 10.1002/sim.7276. [14] 中华医学会糖尿病学分会. 中国2型糖尿病防治指南(2017年版)[J]. 中华糖尿病杂志, 2018, 10(1): 4-67. DOI: 10.3760/cma.j.issn.1674-5809.2018.01.003.Diabetes Society of Chinese Medical Association. Guidelines for the prevention and control of type 2 diabetes in China (2017 Edition)[J]. Chin J Diabetes Mellitus, 2018, 10(1): 4-67. DOI: 10.3760/cma.j.issn.1674-5809.2018.01.003. [15] 朱卫国. 尼群地平与卡托普利联用治疗轻中度原发性高血压的疗效观察[J]. 中国继续医学教育, 2016, 8(33): 204-205. DOI: 10.3969/j.issn.1674-9308.2016.33.118.Zhu WG. The clinical efficacy of combined therapy of nitrendipine and captopril in moderate essential hypertension[J]. China Continuing Medical Education, 2016, 8(33): 204-205. DOI: 10.3969/j.issn.1674-9308.2016.33.118. [16] 余梅. 卡托普利和尼群地平治疗高血压疗效观察[J]. 世界最新医学信息文摘, 2015, 15(86): 112, 114. DOI: 10.3969/j.issn.1671-3141.2015.86.056.Yu M. Observation of curative effect of captopril and nitrendipine in treating hypertension[J]. World Latest Medicine Information, 2015, 15(86): 112, 114. DOI: 10.3969/j.issn.1671-3141.2015.86.056. [17] 高建平. 卡托普利联合尼群地平治疗高血压的临床分析[J]. 实用心脑肺血管病杂志, 2010, 18(6): 789. DOI: 10.3969/j.issn.1008-5971.2010.06.058.Gao JP. Clinical analysis of captopril combined with nitrendipine in the treatment of hypertension[J]. Pract J Cardiac Cereb Pnenm Vasc Dis, 2010, 18(6): 789. DOI: 10.3969/j.issn.1008-5971.2010.06.058. [18] 索萌萌. 运动疗法对高血压治疗效果的影响[J]. 山东医学高等专科学校学报, 2019, 41(6): 475-476. DOI: 10.3969/j.issn.1674-0947.2019.06.043.Suo MM. The effect of exercise therapy on hypertension[J]. Journal of Shandong Medical College, 2019, 41(6): 475-476. DOI: 10.3969/j.issn.1674-0947.2019.06.043. [19] 孙朋, 李世昌, 陈祥和, 等. 有氧运动对肥胖青少年左心室射血分数的急性影响[J/OL]. 中国体育科技: 1-8[2021-06-02]. https://doi.org/10.16470/j.csst.2021027.Sun P, Li SC, Chen XH, et al. Acute effects of aerobic exercise on left ventricular ejection Fraction in obese adolescents[J/OL]. China Sport Science and Technology: 1-8[2021-06-02]. https://doi.org/10.16470/j.csst.2021027. -
点击查看大图
图(3) / 表(1)
计量
- 文章访问数: 527
- HTML全文浏览量: 372
- PDF下载量: 87
- 被引次数: 0