基于标志性时间断点估计的分段线性时变效应 Cox 模型

索珊珊 ,  边文明 ,  蔡嘉鑫 ,  王杨 ,  陈涛 ,  李超 ,  邵方

中国卫生统计 ›› 2026, Vol. 43 ›› Issue (5) : 682 -687.

PDF (1390KB)
中国卫生统计 ›› 2026, Vol. 43 ›› Issue (5) : 682 -687. DOI: 10.11783/j.issn.1002-3674.2026.05.007
论著

基于标志性时间断点估计的分段线性时变效应 Cox 模型

作者信息 +

A Segmented Linear Time-varying Effect Cox Model with Landmark Time Breakpoint Estimation

Author information +
文章历史 +
PDF (1422K)

摘要

目的 为简洁刻画Cox 模型中的时变效应,本研究介绍了一种分段线性时变效应Cox 模型,并引入标志性时间断点(landmark time breakpoint, LTB)。研究旨在将此方法应用于收缩压干预试验(systolic blood pressure interven⁃ tion trial, SPRINT)数据,以分析时变风险特征。方法 模型拟合分两步完成:①通过对加权舍恩菲尔德(Schoenfeld)残差进行分段线性回归来估计LTB;②以估计的LTB为断点,拟合具有时变效应的分段线性Cox回归模型。结果 在SPRINT数据分析中, LTB 估计为2.66 年(95% CI:1.76~3.57)。基于此断点,风险比呈先降低后上升的分段变化趋势。模拟实验提示,该方法在特定条件下估计可能存在偏差,但在SPRINT数据分析中仍能提供临床意义明确的断点及直观可解释的时变效应曲线。结论 分段线性时变效应Cox模型结合LTB的应用,为生存分析中时变效应的可解释估计提供了一种解释性强的建模选择。

Abstract

Objective To concisely characterize time-varying effects in Cox models, this study introduces a piecewise linear time-varying effect Cox model and incorporates the landmark time breakpoint (LTB). The research aimed to apply this method to data from the systolic blood pressure intervention trial (SPRINT) to analyze time-varying risk characteristics. Methods Model fitting was performed in two steps: ①Estimating the LTB by performing piecewise linear regression on weighted Schoenfeld residuals; ② Fitting a piecewise linear Cox regression model with time-varying effects, using the estimated LTB as the breakpoint.Results In the SPRINT data analysis, the LTB was estimated at 2. 66 years (95% CI: 1. 76~ 3. 57). Based on this breakpoint, the hazard ratio (HR) exhibited a piecewise trend of initially decreasing and then increasing. Simulation experiments indicated that the method might produce biased estimates under specific conditions, but in the SPRINT data analysis, it was still able to provide breakpoints with clear clinical significance and intuitively interpretable time-varying effect curves.Conclusion The application of the segmented linear time-varying effect Cox model combined with LTB provides an efficient and intuitive framework for the interpretable estimation of time-varying effects in survival analysis.

关键词

Cox模型 / 时间变化效应 / 分段回归 / 标志性时间断点 / 收缩压干预试验

Key words

Cox model / Time-varying effects / Segmented regression / Landmark time breakpoint / Systolic blood pressure intervention trial

引用本文

引用格式 ▾
索珊珊,边文明,蔡嘉鑫,王杨,陈涛,李超,邵方. 基于标志性时间断点估计的分段线性时变效应 Cox 模型[J]. 中国卫生统计, 2026, 43(5): 682-687 DOI:10.11783/j.issn.1002-3674.2026.05.007

登录浏览全文

4963

注册一个新账户 忘记密码

参考文献

[1]

Bellera CA, MacGrogan G, Debled M, et al. Variables with time—varying effects and the Cox model: some statistical concepts illus— trated with a prognostic factor study in breast cancer[J].BMC Medical Research Methodology, 2010,10:1-12.

[2]

Austin PC, Giardiello D. The impact of violation of the propor— tional hazards assumption on the calibration of the Cox proportional hazards model[J].Statistics in Medicine, 2025, 44(13—14): e70161.

[3]

Austin PC, Fang J, Lee DS. Using fractional polynomials and restricted cubic splines to model non—proportional hazards or time—varying covariate effects in the Cox regression model[J].Statistics in Medicine, 2022, 41(3):612-624.

[4]

Royston P, Altman DG. Regression using fractional polynomials of continuous covariates: parsimonious parametric modelling[J]. Jour— nal of the Royal Statistical Society Series C: Applied Statistics,1994, 43(3):429-453.

[5]

Benner A, Ambler G. mfp: Multivariable fractional polynomials [J].R News, 2005, 5(2):20-23.

[6]

Royston P, Sauerbrei W. Multivariable model—building: a pragmatic approach to regression anaylsis based on fractional polynomials for modelling continuous variables[M]. Hoboken: John Wiley & Sons,2008.

[7]

Ambler G, Royston P. Fractional polynomial model selection proce— dures: investigation of Type I error rate[J].Journal of Statistical Computation and Simulation, 2001, 69(1):89-108.

[8]

Harrell FE. Regression modeling strategies: with applications to lin— ear models, logistic regression, and survival analysis[M]. New York: Springer,2001.

[9]

Muggeo VMR. Estimating regression models with unknown break—points[J].Statistics in Medicine, 2003, 22(19):3055-3071.

[10]

Muggeo VMR. Interval estimation for the break point in segmented regression: a smoothed score—based approach[J]. Australian&New Zealand Journal of Statistics,2017, 59(3): 311-322.

[11]

Wang Y, Nonzee NJ, Zhang H, et al. Interpretation of coefficients in segmented regression for interrupted time series analyses[J].BMC Medical Research Methodology, 2025,25(1): 98.

[12]

Muggeo VMR. Segmented: an R package to fit regression models with broken—line relationships[J].R News, 2008, 8(1):20-25.

[13]

Cleveland WS. Robust locally weighted regression and smoothing scatterplots[J].Journal of the American Statistical Association, 1979, 74(368):829-836.

[14]

Kim HJ, Yu B, Feuer EJ. Inference in segmented line regression: a simulation study[J].Journal of Statistical Computation and Simula— tion, 2008, 78(11):1087-1103.

[15]

Lerman PM. Fitting segmented regression models by grid search [J]. Journal of the Royal Statistical Society Series C: Applied Statis— tics,1980, 29(1):77-84.

[16]

Yu B, Barrett MJ, Kim HJ, et al. Estimating joinpoints in continu— ous time scale for multiple change—point models[J]. Computational Statistics & Data Analysis,2007, 51(5):2420-2427.

[17]

Gillis D, Edwards BPM. The utility of joinpoint regression for esti— mating population parameters given changes in population structure [J].Heliyon, 2019, 5(11): e02515.

[18]

Dhamnetiya D, Patel P, Jha RP, et al. Trends in incidence and mor— tality of tuberculosis in India over past three decades: a joinpoint and age—period—cohort analysis[J].BMC Pulmonary Medicine, 2021, 21:1-14.

[19]

Liu S, Wang B, Fan S, et al. Global burden of musculoskeletal dis— orders and attributable factors in 204 countries and territories: a sec— ondary analysis of the Global Burden of Disease 2019 study[J].BMJ Open, 2022, 12(6):e062183.

[20]

Reif de Paula T, Haas EM, Keller DS. Colorectal cancer in the 45—to—50 age group in the United States: a National Cancer Database (NCDB) analysis[J].Surgical Endoscopy, 2022, 36(9): 6629-6637.

[21]

SPRINT Research Group. A randomized trial of intensive versus standard blood—pressure control[J].New England Journal of Medi— cine, 2015, 373(22):2103-2116.

[22]

R Core Team. R: A language and environment for statistical com— puting[CP/OL]. Vienna: R Foundation for Statistical Computing.2022.

[23]

Therneau T. A package for survival analysis in R[J].R package version, 2015, 2(7):2014.

[24]

Therneau TM. Extending the Cox model[C].Proceedings of the first Seattle symposium in biostatistics: survival analysis. New York, NY: Springer US, 1997:51-84.

[25]

Hothorn T, Hornik K, Van De Wiel MA, et al. Implementing a class of permutation tests: the coin package[J].Journal of Statistical Software, 2008, 28:1-23.

[26]

Uno H, Tian L, Horiguchi M, et al. Package ‘survRM2’[CP].2022.

[27]

Fleming TR, Harrington DP. Counting processes and survival analysis[M]. Hoboben: John Wiley & Sons,2013.

[28]

Brilleman SL, Wolfe R, Moreno—Betancur M, et al. Simulating sur— vival data using the simsurv R package[J].Journal of Statistical Software, 2021, 97:1-27.

[29]

Friedman JH. Multivariate adaptive regression splines[J].The Annals of Statistics, 1991, 19(1):1-67.

[30]

Chen T, Shao F, Chen K, et al. Time to clinical benefit of intensive blood pressure lowering in patients 60 years and older with hyper— tension: a secondary analysis of randomized clinical trials[J].JAMA Internal Medicine, 2022, 182(6):660-667.

[31]

Chen KY, Nie Z, Shi R, et al. Time to benefit of sodium—glucose cotransporter—2 inhibitors among patients with heart failure[J].JAMA Network Open, 2023, 6(8):e2330754.

基金资助

国家自然科学基金(82322060)

国家自然科学基金(82473732)

AI Summary AI Mindmap
PDF (1390KB)

0

访问

0

被引

详细

导航
相关文章

AI思维导图

/

〈 〉