Epidemiology and Biostatistics Section, School of Population Health, University of Auckland, New Zealand.
Department of Population Science and Experimental Medicine, Institute of Cardiovascular Sciences, University College London, UK.
Am J Hypertens. 2017 Dec 8;31(1):53-62. doi: 10.1093/ajh/hpx134.
The relationships of many factors with cardiovascular autonomic function (CVAF) outcome parameters may not be uniform across the entire distribution of the outcome. We examined how demographic and clinical factors varied with different subgroups of CVAF parameters.
Quantile regression was applied to a cross-sectional analysis of 4,167 adults (56% male; age range, 50-84 years) from 4 ethnic groups (3,419 New Zealand European, 303 Pacific, 227 Maori, and 218 South Asian) and without diagnosed cardiac arrhythmia. Pulse rate variability (root mean square of successive differences (RMSSD) and SD of pulse intervals) and baroreflex sensitivity were response variables. Independent variables were age, sex, ethnicity, brachial and aortic blood pressure (BP) variables, body mass index (BMI), and diabetes.
Ordinary linear regression showed that age, sex, Pacific and Maori ethnicity, BP variables, BMI, and diabetes were associated with CVAF parameters. But quantile regression revealed that, across CVAF percentiles, the slopes for these relationships: (i) varied by more than 10-fold in several cases and sometimes changed direction and (ii) noticeably differed in magnitude often (by >3-fold in several cases) compared to ordinary linear regression coefficients. For instance, age was inversely associated with RMSSD at the 10th percentile of this parameter (β = -0.12 ms/year, 95% confidence interval = -0.18 to -0.09 ms/year) but had a positive relationship at the 90th percentile (β = 3.17 ms/year, 95% confidence interval = 2.50 to 4.04 ms/year).
The relationships of demographic and clinical factors with CVAF parameters are, in many cases, not uniform. Quantile regression provides an improved assessment of these associations.
许多因素与心血管自主功能（CVAF）结果参数的关系可能在整个结果分布中并不一致。我们研究了人口统计学和临床因素如何随 CVAF 参数的不同亚组而变化。
对来自 4 个种族（3419 名新西兰欧洲人、303 名太平洋人、227 名毛利人和 218 名南亚人）且无诊断性心律失常的 4167 名成年人（56%为男性；年龄范围为 50-84 岁）进行横断面分析，采用分位数回归。脉搏率变异性（均方根差的连续差异（RMSSD）和脉搏间隔的标准差）和压力反射敏感性是反应变量。自变量为年龄、性别、种族、肱动脉和主动脉血压（BP）变量、体重指数（BMI）和糖尿病。
普通线性回归显示，年龄、性别、太平洋和毛利族裔、BP 变量、BMI 和糖尿病与 CVAF 参数相关。但分位数回归显示，在 CVAF 百分位数中，这些关系的斜率：（i）在几种情况下变化超过 10 倍，有时甚至改变方向；（ii）与普通线性回归系数相比，在幅度上明显不同（在几种情况下相差超过 3 倍）。例如，年龄与 RMSSD 参数的第 10 百分位呈负相关（β=-0.12ms/年，95%置信区间=-0.18 至-0.09 ms/年），但在第 90 百分位呈正相关（β=3.17ms/年，95%置信区间=2.50 至 4.04 ms/年）。
人口统计学和临床因素与 CVAF 参数的关系在许多情况下并不一致。分位数回归提供了对这些关联的更好评估。
Zotero 插件