The impact of handling missing data on alcohol consumption estimates in the UK women cohort study

  • U. NurEmail author
  • N. T. Longford
  • J. E. Cade
  • D. C. Greenwood


We discuss methods for dealing with incomplete-data in the United Kingdom Women’s Cohort Study. We demonstrate by example how important it is to address the issues related to missing data with statistical integrity, illustrate the deficiencies of a data-reduction and a single-imputation method, and discuss how the method of multiple imputation overcomes them. Although the method entails some complexity, the computational activities can be organized in such a way that efficient analyses can be conducted by analysts who are not acquainted with all the details of the imputation method and who wish to rely on software they use and regard as standard.


Alcohol Complete-data analysis Food frequency questionnaire Missing data Multiple imputation 



UK women cohort study


Food frequency questionnaire


World cancer research fund


Missing at random


  1. 1.
    Gronbaek M, et al. Type of alcohol consumed and mortality from all causes, coronary heart disease, and cancer. Ann Intern Med. 2000;133:411–9.PubMedGoogle Scholar
  2. 2.
    IARC. IARC monographs on the evaluation of carcinogenic risks to humans; alcoholic beverage consumption and ethyl carbamate (Urethane). Lyon: IARCPress; 2007. p. 149–156.Google Scholar
  3. 3.
    Boffetta P, Hashibe M. Alcohol and cancer. Lancet Oncol. 2006;7:149–56.CrossRefPubMedGoogle Scholar
  4. 4.
    Paton A. ABC of alcohol. BMJ Publishing Group; 1994.Google Scholar
  5. 5.
    Wood AM, et al. Comparison of imputation and modelling methods in the analysis of a physical activity trial with missing outcomes. Int J Epidemiol. 2005;34:89–99.CrossRefPubMedGoogle Scholar
  6. 6.
    Nur U, et al. Dealing with incomplete data in questionnaires of food and alcohol consumption. Stat Transit. 2005;7:111–34.Google Scholar
  7. 7.
    Cade J, et al. Costs of a healthy diet: analysis from the UK Women’s cohort study. Public Health Nutr. 1999;2:505–12.CrossRefPubMedGoogle Scholar
  8. 8.
    Pollard J, et al. Lifestyle factors affecting fruit and vegetable consumption in the UK Women’s cohort study. Appetite. 2001;37:71–9.CrossRefPubMedGoogle Scholar
  9. 9.
    Greenwood DC, et al. Seven unique food consumption patterns identified among women in the UK Women’s cohort study. Eur J Clin Nutr. 2000;54:314–20.CrossRefPubMedGoogle Scholar
  10. 10.
    Holland B, et al. McCance and Widdowson’s the composition of foods. London: The Royal Society of Chemistry and MAFF; 1991.Google Scholar
  11. 11.
    Rubin DB. Multiple imputation for nonresponse in surveys. New York: Wiley; 1987.CrossRefGoogle Scholar
  12. 12.
    Clark TG, Altman DG. Developing a prognostic model in the presence of missing data: an ovarian cancer case study. J Clin Epidemiol. 2003;56:28–37.CrossRefPubMedGoogle Scholar
  13. 13.
    Van Buuren S, Boshuizen HC, Knook DL. Multiple imputation of missing blood pressure covariates in survival analysis. Stat Med. 1999;18:681–94.CrossRefPubMedGoogle Scholar
  14. 14.
    Royston P. Multiple imputation of missing values: update. Stata J. 2005;5:188–201.Google Scholar
  15. 15.
    Carlin JB, Galati JC, Royston P. A new framework for managing and analyzing multiply imputed data in Stata. Stata J. 2008;8:49–67.Google Scholar
  16. 16.
    Rubin DB. Multiple imputation after 18+ years. J Am Stat Assoc. 1996;91:473–89.CrossRefGoogle Scholar
  17. 17.
    Statacorp. STATA statistical software. [8.0]. 2004. College Station, TX: Stata Corporation; 2009.Google Scholar

Copyright information

© Springer Science+Business Media B.V. 2009

Authors and Affiliations

  • U. Nur
    • 1
    Email author
  • N. T. Longford
    • 2
  • J. E. Cade
    • 3
  • D. C. Greenwood
    • 3
  1. 1.Cancer Research UK Cancer Survival Group, Non-Communicable Disease Epidemiology UnitLondon School of Hygiene and Tropical MedicineLondonUK
  2. 2.Departament d’Economia i EmpresaUniversitat Pompeu FabraBarcelonaSpain
  3. 3.Centre for Epidemiology and BiostatisticsUniversity of LeedsLeedsUK

Personalised recommendations