impact of imputation of missing data on estimation of survival rates: an example in breast cancer

background: multifactorial regression models are frequently used in medicine to estimate survival rate of patients across risk groups. however, their results are not generalisable, if in the development of models assumptions required are not satisfied.  missing data is a common problem in pathology. the aim of this paper is to address the danger of exclusion of cases with missing data, and to highlight the importance of imputation of missing data before development of multifactorial models. methods: this study was performed on 310 breast cancer patients diagnosed in shiraz (southern iran). performing a complete-case cox regression model, a prognostic index was calculated so as to categorise the patients into 3 risk groups. then, applying the multivariate imputation via chained equations (mice) method, missing data were imputed 10 times. using imputed data sets, modelling was performed to assign patients into risk groups. estimated actuarial overal survival (os) rates corresponding to analysis of complete-case and imputed data sets were compared. results: cases with at least one missing datum experienced a significantly better survival curve. estimates derived analysing complete-case data, relative to imputed data sets, underestimated the os rate in all risk groups. in addition confidence intervals were wider indicating loss in precision due to attrition in sample size and power. conclusion: results obtained highlighted the danger of exclusion of missing data. imputation of missing data avoids biased estimates, increases the precision of estimates, and improves genralisability of results to other similar populations. key words: missing data; multiple imputation; breast neoplasm; overall survival, iran references 1. cancer research uk. uk cancer incidence statistics. http://info cancerresearchuk org/ cancerstats/ incidence/ ?a=5441 2007 january [cited 2007 feb 26];available from: url: http:// info. cancerresearchuk. org/ cancerstats/ incidence/?a=5441 2. mcpherson k, steel cm, dixon jm. abc of breast diseases. breast cancer-epidemiology, risk factors, and genetics. bmj 2000 sep 9; 321(7261):624-8. 3. naghavi m. iranian annual of national death registration report. iran ministry of health and medical education; 2005. 4. concato j, feinstein ar, holford tr. the risk of determining risk with multivariable models. ann intern med 1993 feb 1; 118(3):201-10. 5. wyatt jc, altman dg. prognostic models: clinically useful or simply forgotten. british medical journal 1995; 311:1539-41. 6. burton a, altman dg. missing covariate data within cancer prognostic studies: a review of current reporting and proposed guidelines. br j cancer 2004 jul 5; 91(1):4-8. 7. altman dg, bland jm. missing data. bmj 2007 feb 24; 334(7590):424. 8. altman dg, lyman gh. methodological challenges in the evaluation of prognostic factors in breast cancer. breast cancer res treat 1998; 52(1-3):289-303. 9. van buuren s, boshuizen hc, knook dl. multiple imputation of missing blood pressure covariates in survival analysis. stat med 1999 mar 30; 18(6):681-94. 10. rajaeefard ar, baneshi mr, talei ar, mehrabani d. survival models in breast cancer. iranian red crescent medical journal 2009; 11(3):295-300. 11. ayatollahi sm ghasa. menstrual-reproductive factors and age at natural menopause in iran. international journal of gynaecology and obstetrics 2003; 80(3):311-3. 12. cox dr. regression models and life tables. journal of royal statistical society 1972; 34:187-220. 13. schafer jl. analysis of incomplete multivariate data. florida: chapman and hall; 1997. 14. schafer jl. multiple imputations: a primer. stat methods med res 1999 mar; 8(1):3-15. 15. moons kg, donders ra, stijnen t, harrell fe, jr. using the outcome for imputation of missing predictor values were preferred. j clin epidemiol 2006 oct; 59(10):1092-101. 16. r: a language and environment for statistical computing [computer program]. 2007. 17. mice: multivariate imputation by chained equations [computer program]. 2007. 18. design: design package [computer program]. 2008. 19. donders ar, van der heijden gj, stijnen t, moons kg. review: a gentle introduction to imputation of missing values. j clin epidemiol 2006 oct; 59(10):1087-91. 20. fairclough dl. patient reported outcomes as endpoints in medical research. stat methods med res 2004 apr; 13(2):115-38. 21. greenland s, finkle wd. a critical look at methods for handling missing covariates in epidemiologic regression analyses. am j epidemiol 1995 dec 15; 142(12):1255-64. 22. baneshi mr. statistical models in prognostic modelling of many skewed variables and missing data: a case study in breast cancer (phd thesis submitted at edinburgh university) 2009. 23. harrell fe. regression modelling strategies with application to linear models, logistic regression, and survival analysis. new york: springer-verlag; 2001.