With great interest, we read the article “association between imaging characteristics and different molecular subtypes of breast cancer?” (by Wu et al. 2016). In the present article, the authors developed three multivariate regression prediction models (Luminal A, Luminal B, and HER2 overexpressed) to investigate the independent predictive factors associated with different molecular subtypes of breast cancer. We would like to thank the authors for this highly interesting work.
In this study, model predictors included 25 clinical and mammography imaging variables and 15 magnetic resonance imaging variables. The full data set (Luminal A: N = 144, Luminal B: N = 85, and HER2 overexpressed: N = 56) had far fewer events than the recommended number of 10 or more per variable (8,543,964). Too much variables in a multivariable regression prediction model may cause the problem of overfitting . Overfitted models will fail to replicate in future samples, thus creating considerable uncertainty about the scientific merit of the finding . Therefore, we suggest that the research might be improved by variable selection. If the number of variables seems to be too large, background knowledge obtained from analyses of former studies or from theoretical considerations should be applied to prefilter variables to meet the events-per-variable-rule requirements of a problem . The conclusions drawn from this research should be interpreted with caution. A further study with large sample is required.
Supported by Science and Technology Planning Project of Guangdong Province ( 2016KT1071/20151108 ), Medical Scientific Research Foundation of Guangdong Province ( WSTJJ20140103440306198/2015KT1247 ) and Administration of Traditional Chinese Medicine of Guangdong Province ( 2017KT1203 , 2015KT1061/20151248 ).
All authors contributed equally to this work.
References
1. Wu M., Ma J.: Association between imaging characteristics and different molecular subtypes of breast cancer. Acad Radiol 2017; 24: pp. 426-434.
2. Zhang Z.: Too much covariates in a multivariable model may cause the problem of overfitting. J Thorac Dis 2014; 6: pp. E196-E197.
3. Babyak M.A.: What you see may not be what you get: a brief, nontechnical introduction to overfitting in regression-type models. Psychosom Med 2004; 66: pp. 411-421.
4. Heinze G., Dunkler D.: Five myths about variable selection. Transpl Int 2017; 30: pp. 6-10.