A New Proposal for Multivariable Modelling of Time‐Varying Effects in Survival Data Based on Fractional Polynomial Time‐Transformation

The Cox proportional hazards model has become the standard for the analysis of survival time data in cancer and other chronic diseases. In most studies, proportional hazards (PH) are assumed for covariate effects. With long‐term follow‐up, the PH assumption may be violated, leading to poor model fit. To accommodate non‐PH effects, we introduce a new procedure, MFPT, an extension of the multivariable fractional polynomial (MFP) approach, to do the following: (1) select influential variables; (2) determine a sensible dose‐response function for continuous variables; (3) investigate time‐varying effects; (4) model such time‐varying effects on a continuous scale. Assuming PH initially, we start with a detailed model‐building step, including a search for possible non‐linear functions for continuous covariates. Sometimes a variable with a strong short‐term effect may appear weak or non‐influential if ‘averaged’ over time under the PH assumption. To protect against omitting such variables, we repeat the analysis over a restricted time‐interval. Any additional prognostic variables identified by this second analysis are added to create our final time‐fixed multivariable model. Using a forward‐selection algorithm we search for possible improvements in fit by adding time‐varying covariates. The first part to create a final time‐fixed model does not require the use of MFP. A model may be given from ‘outside’ or a different strategy may be preferred for this part. This broadens the scope of the time‐varying part. To motivate and illustrate the methodology, we create prognostic models from a large database of patients with primary breast cancer. Non‐linear time‐fixed effects are found for progesterone receptor status and number of positive lymph nodes. Highly statistically significant time‐varying effects are present for progesterone receptor status and tumour size. (© 2007 WILEY‐VCH Verlag GmbH & Co. KGaA, Weinheim)

[1]  N. Mantel.  Why Stepdown Procedures in Variable Selection , 1970 .

[2]  David R. Cox,et al.  Regression models and life tables (with discussion , 1972 .

[3]  D. Cox.  Regression Models and Life-Tables , 1972 .

[4]  K. Gabriel,et al.  On closed testing procedures with special reference to ordered analysis of variance , 1976 .

[5]  David A. Schoenfeld,et al.  Partial residuals for the proportional hazards regression model , 1982 .

[6]  J. Anderson,et al.  A Two-step Regression Model for Hazard Functions , 1982 .

[7]  A. Karr,et al.  Nonparametric Survival Analysis with Time-Dependent Covariate Effects: A Penalized Partial Likelihood Approach , 1990 .

[8]  D. Harrington,et al.  Counting Processes and Survival Analysis , 1991 .

[9]  P. Sen,et al.  Time-dependent coefficients in a Cox-type regression model , 1991 .

[10]  Robert Gray,et al.  Flexible Methods for Analyzing Survival Data Using Splines, with Applications to Breast Cancer Prognosis , 1992 .

[11]  R. Tibshirani,et al.  Varying‐Coefficient Models , 1993 .

[12]  Z. Ying,et al.  Checking the Cox model with cumulative sums of martingale-based residuals , 1993 .

[13]  P. Royston,et al.  Regression using fractional polynomials of continuous covariates: parsimonious parametric modelling. , 1994 .

[14]  K R Hess,et al.  Assessing time-by-covariate interactions in proportional hazards regression models using cubic spline functions. , 1994, Statistics in medicine.

[15]  B. Silverman,et al.  Nonparametric regression and generalized linear models , 1994 .

[16]  P. Grambsch,et al.  Proportional hazards tests and diagnostics based on weighted residuals , 1994 .

[17]  B. Silverman,et al.  Nonparametric Regression and Generalized Linear Models: A roughness penalty approach , 1993 .

[18]  K R Hess,et al.  Graphical methods for assessing violations of the proportional hazards assumption in Cox regression. , 1995, Statistics in medicine.

[19]  Mike Clarke,et al.  EFFECTS OF RADIOTHERAPY AND SURGERY IN EARLY BREAST-CANCER - AN OVERVIEW OF THE RANDOMIZED TRIALS , 1995 .

[20]  H C van Houwelingen,et al.  Time-dependent effects of fixed covariates in Cox regression. , 1995, Biometrics.

[21]  Michal Abrahamowicz,et al.  Time-Dependent Hazard Ratio: Modeling and Hypothesis Testing with Application in Lupus Nephritis , 1996 .

[22]  N H Ng'andu,et al.  An empirical comparison of statistical tests for assessing the proportional hazards assumption of Cox's model. , 1997, Statistics in medicine.

[23]  H. Heinzl,et al.  Gaining more flexibility in Cox proportional hazards regression models with cubic spline functions. , 1997, Computer methods and programs in biomedicine.

[24]  Willi Sauerbrei,et al.  The Use of Resampling Methods to Simplify Regression Models in Medical Statistics , 1999 .

[25]  P. Royston,et al.  Building multivariable prognostic and diagnostic models: transformation of the predictors by using fractional polynomials , 1999 .

[26]  H. Boshuizen,et al.  Multiple imputation of missing blood pressure covariates in survival analysis. , 1999, Statistics in medicine.

[27]  N. Brünner,et al.  The urokinase system of plasminogen activation and prognosis in 2780 breast cancer patients. , 2000, Cancer research.

[28]  P. Grambsch,et al.  Modeling Survival Data: Extending the Cox Model , 2000 .

[29]  E Biganzoli,et al.  Time-dependent relevance of steroid receptors in breast cancer. , 2000, Journal of clinical oncology : official journal of the American Society of Clinical Oncology.

[30]  P. Royston,et al.  Fractional polynomial model selection procedures: investigation of type i error rate , 2001 .

[31]  Ib M. Skovgaard,et al.  Efficient Estimation of Fixed and Time‐varying Covariate Effects in Multiplicative Intensity Models , 2002 .

[32]  Zongwu Cai,et al.  Local Linear Estimation for Time-Dependent Coefficients in Cox's Regression Models , 2003 .

[33]  Juliane Schäfer,et al.  Dynamic Cox modelling based on fractional polynomials: time‐variations in gastric cancer prognosis , 2003, Statistics in medicine.

[34]  Daniel Krewski,et al.  Flexible Modeling of Exposure-Response Relationship between Long-Term Average Levels of Particulate Air Pollution and Mortality in the American Cancer Society Study , 2003, Journal of toxicology and environmental health. Part A.

[35]  P. Royston,et al.  Stability of multivariable fractional polynomial models with selection of variables and transformations: a bootstrap investigation , 2003, Statistics in medicine.

[36]  P. Royston,et al.  A new approach to modelling interactions between treatment and continuous covariates in clinical trials by using fractional polynomials , 2004, Statistics in medicine.

[37]  S. Hilsenbeck,et al.  Time-dependence of hazard ratios for prognostic factors in primary breast cancer , 2004, Breast Cancer Research and Treatment.

[38]  P. Royston,et al.  Is treatment with interferon-α effective in all patients with metastatic renal carcinoma? A new approach to the investigation of interactions , 2004, British Journal of Cancer.

[39]  P. Royston.  Multiple Imputation of Missing Values , 2004 .

[40]  L. J. Wei,et al.  On the Cox Model With Time-Varying Regression Coefficients , 2005 .

[41]  H Putter,et al.  Long‐term survival with non‐proportional hazards: results from the Dutch Gastric Cancer Trial , 2005, Statistics in medicine.

[42]  M. Abrahamowicz,et al.  Joint estimation of time‐dependent and non‐linear effects of continuous covariates on survival , 2007, Statistics in medicine.