Biographical Information

Dr. Victor Kipnis is the Mathematical Statistician in the Biometry Research Group, Division of Cancer Prevention, at the National Cancer Institute of the United States. He received a M.S. in Mathematics in 1971 and a Ph.D. in Statistics in 1978 from the Moscow State University, Russia, and has been teaching and conducting statistical research in the US beginning in 1986.

One of the most important aspects of Dr. Kipnis’ work since he joined NCI in 1992 has involved statistical issues surrounding the design and analysis of nutritional studies. He has become a widely recognized international leader in the field, has given numerous invited talks and keynote addresses at the national and international conferences, and has authored and co-authored a number of pivotal publications examining the structure of dietary measurement error, its effects on study results, and methods of adjusting for it in nutritional epidemiology and surveillance.

Dr. Kipnis played a leading role in the design and analysis of the OPEN (Observing Protein and Energy Nutrition) biomarker study carried out at NCI in 2001-2002. Using unbiased reference biomarkers, doubly labeled water to measure total energy expenditure and urinary nitrogen to measure protein intake, the OPEN study has provided critical information on the extent and structure of measurement error in commonly used dietary instruments, the food frequency questionnaire (FFQ) and 24-hour recall.

Most recently, Dr. Kipnis has been extensively involved in the development of new statistical approaches to estimating the population distribution of usual intake of episodically consumed foods based on survey data, and predicting individual intake of those foods for relating them to health outcomes. The developed methodology has become an important part of the National Health and Nutrition Examination Survey (NHANES).

Dr. Kipnis is an active member of the NCI “Measurement error working group” and “Surveillance measurement error group”. He is also a chief statistician on the “NIH-AARP Diet and Health cohort study”.

Abstract

Modeling data with excess zeros and measurement error: application to evaluating relationships between episodically consumed foods and health outcomes

Victor Kipnis and NCI Surveillance Measurement Error Working Group

Reported intake of episodically consumed foods using short-term dietary-assessment instruments gives rise to nonnegative data that have excess zeros and measurement error. We describe a general statistical approach for modeling such food intakes reported on two or more 24-hour recalls (24HRs) and demonstrate its use in two scenarios. The first one relates to more typical epidemiologic studies with a food frequency questionnaire (FFQ) as the main dietary-assessment instrument and multiple 24HRs as a reference instrument in a calibration substudy. Following the regression calibration approach for measurement error correction, the information in the calibration study is used to predict individual usual intake given the FFQ and other covariates in the disease model and to evaluate the relationships of usual intakes with health outcomes. The second scenario is related to dietary studies with multiple 24HRs as the main instrument, with a possible addition of a FFQ, as in some dietary surveys. Individual usual intake in this case is generally predicted as the conditional mean intake given reported intake and other covariates in the health model. One feature of the proposed method is that additional covariates potentially related to usual intake may be used to increase the precision of estimates of usual intake and of diet-health outcome associations. Applying the method to data from the Eating at America’s Table Study, we quantify the increased precision obtained from including reported frequency of intake on a FFQ as a covariate in the calibration model. We then demonstrate the method in evaluating the linear relationship between log blood mercury levels and fish intake in women by using data from the National Health and Nutrition Examination Survey, and show increased precision when including the FFQ information. Finally, we present simulation results evaluating the performance of the proposed method in this context.