Adjustment for bias due to errors in exposure assessments in multicenter cohort studies on diet and cancer: a calibration approach

Does adjustment for non-differential misclassification of dichotomous exposure induce positive bias if there is no true association?

This article is a response to an off-the-record discussion that I had at an international meeting of epidemiologists more than decade ago. It centered on a concern, perhaps widely spread, that adjustment for exposure misclassification can induce a false positive result. I trace the possible history of this supposition and test it in a simulated case-control study under the assumption of non-differential misclassification of binary exposure, in which a Bayesian adjustment is applied. Probabilistic bias analysis is also briefly considered. The main conclusion is that adjustment for the presumed non-differential exposure misclassification of dichotomous does not "induce" positive associations, especially if the focus of the interpretation of the result is taken away from the point estimate. The misconception about positive bias induced by adjustment for exposure misclassification, if more clearly explained during the training of epidemiologists, may promote appropriate (and wider) use of the adjustment techniques. The simple message that can be derived from this paper is: "Exposure misclassification as a tractable problem that deserves much more attention than just a typical qualitative throw-away discussion".

Meal occasion, overweight, obesity and central obesity in children and adults: a cross-sectional study based on a nationally representative survey. Colombia, 2015

OBJECTIVE

To establish the association of the number of meals/day with overweight (Ow), obesity (Ob) and central obesity (CO).

DESIGN

Cross-sectional, nationally representative surveys.

SETTING

Colombia.

PARTICIPANTS

A total of 6985 children aged 5-17 years and 7846 adults aged 18-64 years were included.

MAIN OUTCOMES AND MEASURES

According to the WHO, Ow was defined in children as a body mass index (BMI)-for-age Z-score between >1 and ≤2 and in adults as a BMI between ≥25 and <30 (kg/m²). Ob was defined as a Z-score >2 in children and as a BMI ≥30 in adults. CO in children was established by sex and age using cut-off points equivalent to those of adults established by the International Diabetes Federation: ≥90 and ≥80 cm in males and females, respectively. The number of meals/day was estimated with a Food Frequency Questionnaire. Meals/day were grouped into three categories: (reference ≤3, 4 and 5+ meals/day). Crude and adjusted relative prevalence ratios (PRs) and their 95% CIs were calculated. The adjustments included usual energy intake/day and physical activity.

RESULTS

In children, 18.5% had Ow, 6.7% had Ob and 4.0% had CO. The adjusted PRs for five or more meals/day versus three or fewer meals/day were 1.10 (95% CI 0.79 to 1.55) for Ow, 0.95 (95% CI 0.57 to 1.59) for Ob and 1.06 (95% CI 0.72 to 1.55) for CO. In adults, 32.3% had Ow, 13.1% had Ob and 44.8% had CO. The adjusted PRs for five or more meals/day versus three or fewer meals/day were 0.58 (95% CI 0.45 to 0.76) for Ow, 0.51 (95% CI 0.36 to 0.72) for Ob and 0.70 (95% CI 0.54 to 0.92) for CO.

CONCLUSIONS

In children, meals/day were not associated with Ow, Ob or CO. In adults, this inverse relationship exists regardless of energy intake/day, whether physical activity goals are met, sex, age and other potentially confounding sociodemographic and environmental variables.

Timing of eating across ten European countries - results from the European Prospective Investigation into Cancer and Nutrition (EPIC) calibration study

OBJECTIVE

To examine timing of eating across ten European countries.

DESIGN

Cross-sectional analysis of the European Prospective Investigation into Cancer and Nutrition (EPIC) calibration study using standardized 24 h diet recalls collected during 1995-2000. Eleven predefined food consumption occasions were assessed during the recall interview. We present time of consumption of meals and snacks as well as the later:earlier energy intake ratio, with earlier and later intakes defined as 06.00-14.00 and 15.00-24.00 hours, respectively. Type III tests were used to examine associations of sociodemographic, lifestyle and health variables with timing of energy intake.

SETTING

Ten Western European countries.

SUBJECTS

In total, 22 985 women and 13 035 men aged 35-74 years (n 36 020).

RESULTS

A south-north gradient was observed for timing of eating, with later consumption of meals and snacks in Mediterranean countries compared with Central and Northern European countries. However, the energy load was reversed, with the later:earlier energy intake ratio ranging from 0·68 (France) to 1·39 (Norway) among women, and from 0·71 (Greece) to 1·35 (the Netherlands) among men. Among women, country, age, education, marital status, smoking, day of recall and season were all independently associated with timing of energy intake (all P<0·05). Among men, the corresponding variables were country, age, education, smoking, physical activity, BMI and day of recall (all P<0·05).

CONCLUSIONS

We found pronounced differences in timing of eating across Europe, with later meal timetables but greater energy load earlier during the day in Mediterranean countries compared with Central and Northern European countries.

Coffee and Tea Consumption and the Contribution of Their Added Ingredients to Total Energy and Nutrient Intakes in 10 European Countries: Benchmark Data from the Late 1990s

BACKGROUND

Coffee and tea are among the most commonly consumed nonalcoholic beverages worldwide, but methodological differences in assessing intake often hamper comparisons across populations. We aimed to (i) describe coffee and tea intakes and (ii) assess their contribution to intakes of selected nutrients in adults across 10 European countries.

METHOD

Between 1995 and 2000, a standardized 24-h dietary recall was conducted among 36,018 men and women from 27 European Prospective Investigation into Cancer and Nutrition (EPIC) study centres. Adjusted arithmetic means of intakes were estimated in grams (=volume) per day by sex and centre. Means of intake across centres were compared by sociodemographic characteristics and lifestyle factors.

RESULTS

In women, the mean daily intake of coffee ranged from 94 g/day (~0.6 cups) in Greece to 781 g/day (~4.4 cups) in Aarhus (Denmark), and tea from 14 g/day (~0.1 cups) in Navarra (Spain) to 788 g/day (~4.3 cups) in the UK general population. Similar geographical patterns for mean daily intakes of both coffee and tea were observed in men. Current smokers as compared with those who reported never smoking tended to drink on average up to 500 g/day more coffee and tea combined, but with substantial variation across centres. Other individuals' characteristics such as educational attainment or age were less predictive. In all centres, coffee and tea contributed to less than 10% of the energy intake. The greatest contribution to total sugar intakes was observed in Southern European centres (up to ~20%).

CONCLUSION

Coffee and tea intake and their contribution to energy and sugar intake differed greatly among European adults. Variation in consumption was mostly driven by geographical region.

Meal patterns across ten European countries - results from the European Prospective Investigation into Cancer and Nutrition (EPIC) calibration study

OBJECTIVE

To characterize meal patterns across ten European countries participating in the European Prospective Investigation into Cancer and Nutrition (EPIC) calibration study.

DESIGN

Cross-sectional study utilizing dietary data collected through a standardized 24 h diet recall during 1995-2000. Eleven predefined intake occasions across a 24 h period were assessed during the interview. In the present descriptive report, meal patterns were analysed in terms of daily number of intake occasions, the proportion reporting each intake occasion and the energy contributions from each intake occasion.

SETTING

Twenty-seven centres across ten European countries.

SUBJECTS

Women (64 %) and men (36 %) aged 35-74 years (n 36 020).

RESULTS

Pronounced differences in meal patterns emerged both across centres within the same country and across different countries, with a trend for fewer intake occasions per day in Mediterranean countries compared with central and northern Europe. Differences were also found for daily energy intake provided by lunch, with 38-43 % for women and 41-45 % for men within Mediterranean countries compared with 16-27 % for women and 20-26 % for men in central and northern European countries. Likewise, a south-north gradient was found for daily energy intake from snacks, with 13-20 % (women) and 10-17 % (men) in Mediterranean countries compared with 24-34 % (women) and 23-35 % (men) in central/northern Europe.

CONCLUSIONS

We found distinct differences in meal patterns with marked diversity for intake frequency and lunch and snack consumption between Mediterranean and central/northern European countries. Monitoring of meal patterns across various cultures and populations could provide critical context to the research efforts to characterize relationships between dietary intake and health.

A multilevel model to estimate the within- and the between-center components of the exposure/disease association in the EPIC study

In a multicenter study, the overall relationship between exposure and the risk of cancer can be broken down into a within-center component, which reflects the individual level association, and a between-center relationship, which captures the association at the aggregate level. A piecewise exponential proportional hazards model with random effects was used to evaluate the association between dietary fiber intake and colorectal cancer (CRC) risk in the EPIC study. During an average follow-up of 11.0 years, 4,517 CRC events occurred among study participants recruited in 28 centers from ten European countries. Models were adjusted by relevant confounding factors. Heterogeneity among centers was modelled with random effects. Linear regression calibration was used to account for errors in dietary questionnaire (DQ) measurements. Risk ratio estimates for a 10 g/day increment in dietary fiber were equal to 0.90 (95%CI: 0.85, 0.96) and 0.85 (0.64, 1.14), at the individual and aggregate levels, respectively, while calibrated estimates were 0.85 (0.76, 0.94), and 0.87 (0.65, 1.15), respectively. In multicenter studies, over a straightforward ecological analysis, random effects models allow information at the individual and ecologic levels to be captured, while controlling for confounding at both levels of evidence.

[Estimate of corrected values and the effect of correction for measurement error in dietary data obtained by the Food Frequency Questionnaire for Adolescents (AFFQ)]

The scope of this study was to estimate calibrated values for dietary data obtained by the Food Frequency Questionnaire for Adolescents (FFQA) and illustrate the effect of this approach on food consumption data. The adolescents were assessed on two occasions, with an average interval of twelve months. In 2004, 393 adolescents participated, and 289 were then reassessed in 2005. Dietary data obtained by the FFQA were calibrated using the regression coefficients estimated from the average of two 24-hour recalls (24HR) of the subsample. The calibrated values were similar to the the 24HR reference measurement in the subsample. In 2004 and 2005 a significant difference was observed between the average consumption levels of the FFQA before and after calibration for all nutrients. With the use of calibrated data the proportion of schoolchildren who had fiber intake below the recommended level increased. Therefore, it is seen that calibrated data can be used to obtain adjusted associations due to reclassification of subjects within the predetermined categories.

Intake estimation of total and individual flavan-3-ols, proanthocyanidins and theaflavins, their food sources and determinants in the European Prospective Investigation into Cancer and Nutrition (EPIC) study

Epidemiological studies suggest health-protective effects of flavan-3-ols and their derived compounds on chronic diseases. The present study aimed to estimate dietary flavan-3-ol, proanthocyanidin (PA) and theaflavin intakes, their food sources and potential determinants in the European Prospective Investigation into Cancer and Nutrition (EPIC) calibration cohort. Dietary data were collected using a standardised 24 h dietary recall software administered to 36 037 subjects aged 35-74 years. Dietary data were linked with a flavanoid food composition database compiled from the latest US Department of Agriculture and Phenol-Explorer databases and expanded to include recipes, estimations and retention factors. Total flavan-3-ol intake was the highest in UK Health-conscious men (453·6 mg/d) and women of UK General population (377·6 mg/d), while the intake was the lowest in Greece (men: 160·5 mg/d; women: 124·8 mg/d). Monomer intake was the highest in UK General population (men: 213·5 mg/d; women: 178·6 mg/d) and the lowest in Greece (men: 26·6 mg/d in men; women: 20·7 mg/d). Theaflavin intake was the highest in UK General population (men: 29·3 mg/d; women: 25·3 mg/d) and close to zero in Greece and Spain. PA intake was the highest in Asturias (men: 455·2 mg/d) and San Sebastian (women: 253 mg/d), while being the lowest in Greece (men: 134·6 mg/d; women: 101·0 mg/d). Except for the UK, non-citrus fruits (apples/pears) were the highest contributors to the total flavan-3-ol intake. Tea was the main contributor of total flavan-3-ols in the UK. Flavan-3-ol, PA and theaflavin intakes were significantly different among all assessed groups. This study showed heterogeneity in flavan-3-ol, PA and theaflavin intake throughout the EPIC countries.

Bias in protein and potassium intake collected with 24-h recalls (EPIC-Soft) is rather comparable across European populations

Purpose

We investigated whether group-level bias of a 24-h recall estimate of protein and potassium intake, as compared to biomarkers, varied across European centers and whether this was influenced by characteristics of individuals or centers.

Methods

The combined data from EFCOVAL and EPIC studies included 14 centers from 9 countries (n = 1,841). Dietary data were collected using a computerized 24-h recall (EPIC-Soft). Nitrogen and potassium in 24-h urine collections were used as reference method. Multilevel linear regression analysis was performed, including individual-level (e.g., BMI) and center-level (e.g., food pattern index) variables.

Results

For protein intake, no between-center variation in bias was observed in men while it was 5.7% in women. For potassium intake, the between-center variation in bias was 8.9% in men and null in women. BMI was an important factor influencing the biases across centers (p < 0.01 in all analyses). In addition, mode of administration (p = 0.06 in women) and day of the week (p = 0.03 in men and p = 0.06 in women) may have influenced the bias in protein intake across centers. After inclusion of these individual variables, between-center variation in bias in protein intake disappeared for women, whereas for potassium, it increased slightly in men (to 9.5%). Center-level variables did not influence the results.

Conclusion

The results suggest that group-level bias in protein and potassium (for women) collected with 24-h recalls does not vary across centers and to a certain extent varies for potassium in men. BMI and study design aspects, rather than center-level characteristics, affected the biases across centers.

Electronic supplementary material

The online version of this article (doi:10.1007/s00394-011-0279-z) contains supplementary material, which is available to authorized users.

Dietary reporting errors on 24 h recalls and dietary questionnaires are associated with BMI across six European countries as evaluated with recovery biomarkers for protein and potassium intake

Whether there are differences between countries in the validity of self-reported diet in relation to BMI, as evaluated using recovery biomarkers, is not well understood. We aimed to evaluate BMI-related reporting errors on 24 h dietary recalls (24-HDR) and on dietary questionnaires (DQ) using biomarkers for protein and K intake and whether the BMI effect differs between six European countries. Between 1995 and 1999, 1086 men and women participating in the European Prospective Investigation into Cancer and Nutrition completed a single 24-HDR, a DQ and one 24 h urine collection. In regression analysis, controlling for age, sex, education and country, each unit (1 kg/m2) increase in BMI predicted an approximately 1·7 and 1·3 % increase in protein under-reporting on 24-HDR and DQ, respectively (both P < 0·0001). Exclusion of individuals who probably misreported energy intake attenuated BMI-related bias on both instruments. The BMI effect on protein under-reporting did not differ for men and women and neither between countries on both instruments as tested by interaction (all P>0·15). In women, but not in men, the DQ yielded higher mean intakes of protein that were closer to the biomarker-based measurements across BMI groups when compared with 24-HDR. Results for K were similar to those of protein, although BMI-related under-reporting of K was of a smaller magnitude, suggesting differential misreporting of foods. Under-reporting of protein and K appears to be predicted by BMI, but this effect may be driven by ‘low-energy reporters’. The BMI effect on under-reporting seems to be the same across countries.

Reliability of anthropometric measures in a longitudinal cohort of patients initiating ART in West Africa

Background

Anthropometric measurements are a non invasive, inexpensive, and suitable method for evaluating the nutritional status in population studies with relatively large sample sizes. However, anthropometric techniques are prone to errors that could arise, for example, from the inadequate training of personnel. Despite these concerns, anthropometrical measurement error is seldom assessed in cohort studies. We describe the reliability and challenges associated with measurement of longitudinal anthropometric data in a cohort of West African HIV+ adults .

Methods

In a cohort of patients initiating antiretroviral treatment in Mali, we evaluated nutritional status using anthropometric measurements(weight, height, mid-upper arm circumference, waist circumference and triceps skinfold). Observers with no prior experience in the field of anthropometry were trained to perform anthropometrical measurements. To assess the intra- and inter-observer variability of the measurements taken in the course of the study, two sub-studies were carried out: one at the beginning and one at the end of the prospective study. Twelve patients were measured twice on two consecutive days by the same observer on both study occasions. The technical error of measurement (TEM) (absolute and relative value), and the coefficient of reliability (R) were calculated and compared across reliability studies.

Results

According to the R and relative TEM, inter-observer reliabilities were only acceptable for height and weight. In terms of intra-observer precision, while the first and second anthropometrists demonstrated better reliability than the third, only height and weight measurements were reliable. Looking at total TEM, we observed that while measurements remained stable between studies for height and weight, circumferences and skinfolds lost precision from one occasion to the next.

Conclusions

Height and weight were the most reliable measurements under the study's conditions. Circumferences and skinfolds demonstrated less reliability and lost precision over time, probably as a result of insufficient supervision over the entire length of the study. Our results underline the importance of a careful observer's selection, good initial preparation, as well as the necessity of ongoing training and supervision over the entire course of a longitudinal nutritional study. Failure to do so could have major repercussions on data reliability and jeopardize its utilization.

Region-specific nutrient intake patterns exhibit a geographical gradient within and between European countries

Until recently, the study of nutrient patterns was hampered at an international level by a lack of standardization of both dietary methods and nutrient databases. We aimed to describe the diversity of nutrient patterns in the European Prospective Investigation into Cancer and Nutrition (EPIC) study at population level as a starting point for future nutrient pattern analyses and their associations with chronic diseases in multi-center studies. In this cross-sectional study, 36,034 persons aged 35-74 y were administered a single, standardized 24-h dietary recall. Intake of 25 nutrients (excluding intake from dietary supplements) was estimated using a standardized nutrient database. We used a graphic presentation of mean nutrient intakes by region and sex relative to the overall EPIC means to contrast patterns within and between 10 European countries. In Mediterranean regions, including Greece, Italy, and the southern centers of Spain, the nutrient pattern was dominated by relatively high intakes of vitamin E and monounsaturated fatty acids (MUFA), whereas intakes of retinol and vitamin D were relatively low. In contrast, in Nordic countries, including Norway, Sweden, and Denmark, reported intake of these same nutrients resulted in almost the opposite pattern. Population groups in Germany, The Netherlands, and the UK shared a fatty acid pattern of relatively high intakes of PUFA and SFA and relatively low intakes of MUFA, in combination with a relatively high intake of sugar. We confirmed large variability in nutrient intakes across the EPIC study populations and identified 3 main region-specific patterns with a geographical gradient within and between European countries.

Energy intake and sources of energy intake in the European Prospective Investigation into Cancer and Nutrition

OBJECTIVES

To describe energy intake and its macronutrient and food sources among 27 regions in 10 countries participating in the European Prospective Investigation into Cancer and Nutrition (EPIC) study.

METHODS

Between 1995 and 2000, 36 034 subjects aged 35-74 years were administered a standardized 24-h dietary recall. Intakes of macronutrients (g/day) and energy (kcal/day) were estimated using standardized national nutrient databases. Mean intakes were weighted by season and day of the week and were adjusted for age, height and weight, after stratification by gender. Extreme low- and high-energy reporters were identified using Goldberg's cutoff points (ratio of energy intake and estimated basal metabolic rate <0.88 or >2.72), and their effects on macronutrient and energy intakes were studied.

RESULTS

Low-energy reporting was more prevalent in women than in men. The exclusion of extreme-energy reporters substantially lowered the EPIC-wide range in mean energy intake from 2196-2877 to 2309-2866 kcal among men. For women, these ranges were 1659-2070 and 1873-2108 kcal. There was no north-south gradient in energy intake or in the prevalence of low-energy reporting. In most centres, cereals and cereal products were the largest contributors to energy intake. The food groups meat, dairy products and fats and oils were also important energy sources. In many centres, the highest mean energy intakes were observed on Saturdays.

CONCLUSIONS

These data highlight and quantify the variations and similarities in energy intake and sources of energy intake among 10 European countries. The prevalence of low-energy reporting indicates that the study of energy intake is hampered by the problem of underreporting.

Dietary intake of the water-soluble vitamins B1, B2, B6, B12 and C in 10 countries in the European Prospective Investigation into Cancer and Nutrition

OBJECTIVES

To describe the intake of vitamins thiamine (B1), riboflavin (B2), B6 (pyridoxine), B12 (cobalamine) and C (ascorbic acid) and their food sources among 27 centres in 10 countries participating in the European Prospective Investigation into Cancer and Nutrition (EPIC) study.

METHODS

Between 1995 and 2000, 36 034 persons aged between 35 and 74 years were administered a standardized 24-h dietary recall using a computerized interview software programme (EPIC-SOFT). Intakes of the four B vitamins and vitamin C were estimated using the standardized EPIC Nutrient Database (ENDB). Mean intakes were adjusted for age and weighted by season and day of recall.

RESULTS

Intake of B vitamins did not vary considerably between centres, except in the UK health-conscious cohort, in which substantially higher intakes of thiamine and lower intakes of vitamin B12 were reported compared with other centres. Overall, meat was the most important contributor to the B vitamins in all centres except in the UK health-conscious group. Vitamin C showed a clear geographical gradient, with higher intakes in the southern centres as compared with the northern ones; this was more pronounced in men than in women. Vegetables and fruits were major contributors to vitamin C in all centres, but juices and potatoes were also important sources in the northern centres.

CONCLUSIONS

This study showed no major differences across centres in the mean intakes of B vitamins (thiamine, riboflavin, B6, B12), whereas a tendency towards a north-south gradient was observed for vitamin C.

A Bayesian multilevel model for estimating the diet/disease relationship in a multicenter study with exposures measured with error: the EPIC study

In a multicenter study, the overall relationship between diet and cancer risk can be broken down into: (a) within-center relationships, which reflect the relationships at the individual level in each of the centers, and (b) a between-center relationship, which captures the association between exposure and disease risk at the aggregate level. In this work, we propose the use of a Bayesian multilevel model that takes into account the within- and between-center levels of evidence, using information at the individual and aggregate level. Correction for measurement error is performed in order to correct for systematic between-center measurement error in dietary exposure, and for attenuation biases in relative risk estimates within centers. The estimation of the parameters is carried out in a Bayesian framework using Gibbs sampling. The model entails a measurement, an exposure, and a disease component. Within the European Prospective Investigation into Cancer and Nutrition (EPIC) the association between lipid intake, assessed through dietary questionnaire and 24-hour dietary recall, and breast cancer incidence was evaluated. This analysis involved 21 534 women and 334 incident breast cancer cases from the EPIC calibration study. In this study, total energy intake was positively associated with breast cancer incidence at the aggregate level, whereas no effect was observed for fat. At the individual level, height was positively related to breast cancer incidence, whereas a weaker association was observed for fat. The use of multilevel models, which constitute a very powerful approach to estimating individual vs aggregate levels of evidence should be considered in multicenter studies.

Validity and calibration of food frequency questionnaires used with African-American adults in the Jackson Heart Study

OBJECTIVE

To examine the relative validity of two food frequency questionnaires (FFQs) developed for use in investigating diet and disease relationships within the adult African-American population in the southern United States.

DESIGN

Cross-sectional analyses of dietary nutrient intake data, comparing four 24-hour dietary recalls with an FFQ developed by the Lower Mississippi Delta Nutrition Intervention Research Initiative, and its shorter version adapted for use in the Jackson Heart Study.

SUBJECTS

A representative subset of participants (n=499, aged 35 to 81 years) from the baseline Jackson Heart Study cohort (N=5,302) was selected for this study. Data collection took place between winter 2000 and spring 2004.

STATISTICAL ANALYSES

Pearson's correlation coefficients (energy adjusted and de-attenuated) for 26 nutrients estimates from each of the FFQs, comparing them with the mean of four 24-hour dietary recalls. The ability of the FFQs to rank individuals based on nutrient intakes was compared to that of the mean of four 24-hour dietary recalls and attenuation coefficients were also calculated.

RESULTS

Median nutrient intake estimates tended to be higher on the long and lower on the short FFQ compared to the median for the mean of four 24-hour dietary recalls. Energy adjusted and deattenuated correlations of FFQ intake estimates with recalls ranged from 0.20 for sodium to 0.70 for carbohydrate for the short FFQ and from 0.23 for polyunsaturated fat to 0.75 for dietary fiber and magnesium for the long. Attenuation coefficients for men on average were 0.42 for the short and 0.49 for the long FFQ. For women, these were 0.31 for the short and 0.42 for the long FFQ.

CONCLUSIONS

Both FFQs appear to be reasonably valid for assessment of dietary intake of adult African Americans in the South. The Lower Mississippi Delta Nutrition Intervention Research Initiative FFQ exhibited higher intake estimates and stronger correlations with recalls than the Jackson Heart Study FFQ for most nutrients analyzed, more so for women than men.

The evaluation of the diet/disease relation in the EPIC study: considerations for the calibration and the disease models

BACKGROUND

International multicentre studies on diet and cancer are relatively new in epidemiological research. They offer a series of challenging methodological issues for the evaluation of the association between dietary exposure and disease outcomes, which can both be quite heterogeneous across different geographical regions. This requires considerable work to standardize dietary measurements at the food and the nutrient levels.

METHODS

Within the European Prospective Investigation into Cancer and Nutrition (EPIC), a calibration study was set up to express individual dietary intakes according to the same reference scale. A linear regression calibration model was used to correct the association between diet and disease for measurement errors in dietary exposures. In the present work, we describe an approach for analysing the EPIC data, using as an example the evaluation of the association between fish intake and colorectal cancer incidence.

RESULTS

Sex- and country-specific attenuation factors ranged from 0.083 to 0.784, with values overall higher for men compared with women. Hazard ratio estimates of colorectal cancer for a 10 g/day increase in fish intake were 0.97 [95% confidence interval (CI): 0.95-0.99] and 0.93 (0.88-0.98), before and after calibration, respectively.

CONCLUSIONS

In a multicentre study, the diet/disease association can be evaluated by exploiting the whole variability of intake over the entire study. Calibration may reduce between-centre heterogeneity in the diet-disease relationship caused by differential impact of measurement errors across cohorts.

The EPIC nutrient database project (ENDB): a first attempt to standardize nutrient databases across the 10 European countries participating in the EPIC study

OBJECTIVE

This paper describes the ad hoc methodological concepts and procedures developed to improve the comparability of Nutrient databases (NDBs) across the 10 European countries participating in the European Prospective Investigation into Cancer and Nutrition (EPIC). This was required because there is currently no European reference NDB available.

DESIGN

A large network involving national compilers, nutritionists and experts on food chemistry and computer science was set up for the 'EPIC Nutrient DataBase' (ENDB) project. A total of 550-1500 foods derived from about 37,000 standardized EPIC 24-h dietary recalls (24-HDRS) were matched as closely as possible to foods available in the 10 national NDBs. The resulting national data sets (NDS) were then successively documented, standardized and evaluated according to common guidelines and using a DataBase Management System specifically designed for this project. The nutrient values of foods unavailable or not readily available in NDSs were approximated by recipe calculation, weighted averaging or adjustment for weight changes and vitamin/mineral losses, using common algorithms.

RESULTS

The final ENDB contains about 550-1500 foods depending on the country and 26 common components. Each component value was documented and standardized for unit, mode of expression, definition and chemical method of analysis, as far as possible. Furthermore, the overall completeness of NDSs was improved (>or=99%), particularly for beta-carotene and vitamin E.

CONCLUSION

The ENDB constitutes a first real attempt to improve the comparability of NDBs across European countries. This methodological work will provide a useful tool for nutritional research as well as end-user recommendations to improve NDBs in the future.

An approach to estimate between- and within-group correlation coefficients in multicenter studies: plasma carotenoids as biomarkers of intake of fruits and vegetables

In a multicenter study, the overall correlation between two variables can be broken down into a within- and a between-group correlation reflecting associations at the individual and aggregate levels, respectively. A random-effects model is used to estimate variance components of nutrition-related variables and the within- and between-group correlation coefficients. Using data from the European Prospective Investigation into Cancer and Nutrition (EPIC), the authors analyzed the association between levels of three plasma carotenoids (alpha-carotene, beta-cryptoxanthin, and lycopene) and dietary intake of three items (total fruits, carrots, and tomatoes), assessed through dietary questionnaire and single 24-hour dietary recall measurements, in a cross-sectional study involving 3,089 subjects from nine European countries. Intraclass correlation coefficients were 0.178 for alpha-carotene, 0.216 for beta-cryptoxanthin, and 0.299 for lycopene. The between-group correlation coefficients were higher than the within-group coefficients for all three carotenoids. For beta-cryptoxanthin and fruit intake, the between-group versus the within-group correlations were 0.78 and 0.26 for the dietary questionnaire and 0.85 and 0.19 for the 24-hour dietary recall. Results indicate that variability of exposure is driven mainly by the individual compared with the aggregate variation and that biomarker levels perform fairly accurately in discriminating population-level consumption of fruits and vegetables.

Meat, fish, and colorectal cancer risk: the European Prospective Investigation into cancer and nutrition

BACKGROUND

Current evidence suggests that high red meat intake is associated with increased colorectal cancer risk. High fish intake may be associated with a decreased risk, but the existing evidence is less convincing.

METHODS

We prospectively followed 478 040 men and women from 10 European countries who were free of cancer at enrollment between 1992 and 1998. Information on diet and lifestyle was collected at baseline. After a mean follow-up of 4.8 years, 1329 incident colorectal cancers were documented. We examined the relationship between intakes of red and processed meat, poultry, and fish and colorectal cancer risk using a proportional hazards model adjusted for age, sex, energy (nonfat and fat sources), height, weight, work-related physical activity, smoking status, dietary fiber and folate, and alcohol consumption, stratified by center. A calibration substudy based on 36 994 subjects was used to correct hazard ratios (HRs) and 95% confidence intervals (CIs) for diet measurement errors. All statistical tests were two-sided.

RESULTS

Colorectal cancer risk was positively associated with intake of red and processed meat (highest [>160 g/day] versus lowest [<20 g/day] intake, HR = 1.35, 95% CI = 0.96 to 1.88; Ptrend = .03) and inversely associated with intake of fish (>80 g/day versus <10 g/day, HR = 0.69, 95 % CI = 0.54 to 0.88; Ptrend<.001), but was not related to poultry intake. Correcting for measurement error strengthened the associations between colorectal cancer and red and processed meat intake (per 100-g increase HR = 1.25, 95% CI =1.09 to 1.41, Ptrend = .001 and HR = 1.55, 95% CI = 1.19 to 2.02, Ptrend = .001 before and after calibration, respectively) and for fish (per 100 g increase HR = 0.70, 95% CI = 0.57 to 0.87, Ptrend<.001 and HR = 0.46, 95% CI = 0.27 to 0.77, Ptrend = .003; before and after correction, respectively). In this study population, the absolute risk of development of colorectal cancer within 10 years for a study subject aged 50 years was 1.71% for the highest category of red and processed meat intake and 1.28% for the lowest category of intake and was 1.86% for subjects in the lowest category of fish intake and 1.28% for subjects in the highest category of fish intake.

CONCLUSIONS

Our data confirm that colorectal cancer risk is positively associated with high consumption of red and processed meat and support an inverse association with fish intake.

European Prospective Investigation into Cancer and Nutrition (EPIC) calibration study: rationale, design and population characteristics

The European Prospective Investigation into Cancer and Nutrition (EPIC), which covers a large cohort of half a million men and women from 23 European centres in 10 Western European countries, was designed to study the relationship between diet and the risk of chronic diseases, particularly cancer. Information on usual individual dietary intake was assessed using different validated dietary assessment methods across participating countries. In order to adjust for possible systematic over- or underestimation in dietary intake measurements and correct for attenuation bias in relative risk estimates, a calibration approach was developed. This approach involved an additional dietary assessment common across study populations to re-express individual dietary intakes according to the same reference scale. A single 24-hour diet recall was therefore collected, as the EPIC reference calibration method, from a stratified random sample of 36 900 subjects from the entire EPIC cohort, using a software program (EPIC-SOFT) specifically designed to standardise the dietary measurements across study populations. This paper describes the design and populations of the calibration sub-studies set up in the EPIC centres. In addition, to assess whether the calibration sub-samples were representative of the entire group of EPIC cohorts, a series of subjects' characteristics known possibly to influence dietary intakes was compared in both population groups. This was the first time that calibration sub-studies had been set up in a large multi-centre European study. These studies showed that, despite certain inherent methodological and logistic constraints, a study design such as this one works relatively well in practice. The average response in the calibration study was 78.3% and ranged from 46.5% to 92.5%. The calibration population differed slightly from the overall cohort but the differences were small for most characteristics and centres. The overall results suggest that, after adjustment for age, dietary intakes estimated from calibration samples can reasonably be interpreted as representative of the main cohorts in most of the EPIC centres.

Soy product consumption in 10 European countries: the European Prospective Investigation into Cancer and Nutrition (EPIC) study

OBJECTIVE

The aim of this study was to describe the variation of soy product intake in 10 European countries by using a standardised reference dietary method. A subsidiary aim was to characterise the pattern of soy consumption among a sub-group of participants with a habitual health-conscious lifestyle (HHL), i.e. non-meat eaters who are fish eaters, vegetarians and vegans.

DESIGN

A 24-hour dietary recall interview (24-HDR) was conducted among a sample (5-12%) of all cohorts in the European Prospective Investigation into Cancer and Nutrition (EPIC). Study participants totalled 35 955 after exclusion of subjects younger than 35 or older than 74 years of age. Soy products were subdivided into seven sub-groups by similarity. Distribution of consumption and crude and adjusted means of intake were computed per soy product group across countries. Intake of soy products was also investigated among participants with an HHL.

RESULTS

In total, 195 men and 486 women reported consuming soy products in the 24-HDR interview. Although soy product intake was generally low across all countries, the highest intake level was observed in the UK, due to over-sampling of a large number of participants with an HHL. The most frequently consumed soy foods were dairy substitutes in the UK and France and beans and sprouts among mid-European countries. For both genders, the sub-group of soy dairy substitutes was consumed in the highest quantities (1.2 g day-1 for men; 1.9 g day-1 for women). Participants with an HHL differed substantially from others with regard to demographic, anthropometric and nutritional factors. They consumed higher quantities of almost all soy product groups.

CONCLUSIONS

Consumption of soy products is low in centres in Western Europe. Soy dairy substitutes are most frequently consumed. Participants with an HHL form a distinct sub-group with higher consumptions of fruit, vegetables, legumes, cereals and soy products compared with the other participants.

European Prospective Investigation into Cancer and Nutrition (EPIC): study populations and data collection

The European Prospective Investigation into Cancer and Nutrition (EPIC) is an ongoing multi-centre prospective cohort study designed to investigate the relationship between nutrition and cancer, with the potential for studying other diseases as well. The study currently includes 519 978 participants (366 521 women and 153 457 men, mostly aged 35-70 years) in 23 centres located in 10 European countries, to be followed for cancer incidence and cause-specific mortality for several decades. At enrollment, which took place between 1992 and 2000 at each of the different centres, information was collected through a non-dietary questionnaire on lifestyle variables and through a dietary questionnaire addressing usual diet. Anthropometric measurements were performed and blood samples taken, from which plasma, serum, red cells and buffy coat fractions were separated and aliquoted for long-term storage, mostly in liquid nitrogen. To calibrate dietary measurements, a standardised, computer-assisted 24-hour dietary recall was implemented at each centre on stratified random samples of the participants, for a total of 36 900 subjects. EPIC represents the largest single resource available today world-wide for prospective investigations on the aetiology of cancers (and other diseases) that can integrate questionnaire data on lifestyle and diet, biomarkers of diet and of endogenous metabolism (e.g. hormones and growth factors) and genetic polymorphisms. First results of case-control studies nested within the cohort are expected early in 2003. The present paper provides a description of the EPIC study, with the aim of simplifying reference to it in future papers reporting substantive or methodological studies carried out in the EPIC cohort.

The role of multi-centre cohort studies in studying the relation between diet and cancer

In spite of important progress made during recent decades in nutritional epidemiology methods, many questions about the role of diet in determining cancer risk remain elusive. One example of an unresolved question is whether a high percentage of energy intake in the form of fat (especially saturated fat) is associated with an increased risk of breast cancer. Observations from international correlation and case-control studies support this hypothesis, while results from prospective cohort studies, generally considered less prone to bias, do not. In this paper, we review the advantages and limitations of these different types of epidemiological study design, and discuss how multi-centre studies may help answer some of the unresolved questions about relations between diet, nutritional status, and cancer risk. Multi-centre cohort studies may have the advantage of increased statistical power because of larger variations in individuals' dietary intake patterns and disease risk (as in international correlation studies), while at the same time offering all the possibilities of individual-level studies to model confounding and/or interaction effects.

Cohort studies of fat intake and the risk of breast cancer--a pooled analysis

BACKGROUND

Experiments in animals, international correlation comparisons, and case-control studies support an association between dietary fat intake and the incidence of breast cancer. Most cohort studies do not corroborate the association, but they have been criticized for involving small numbers of cases, homogeneous fat intake, and measurement errors in estimates of fat intake.

METHODS

We identified seven prospective studies in four countries that met specific criteria and analyzed the primary data in a standardized manner. Pooled estimates of the relation of fat intake to the risk of breast cancer were calculated, and data from study-specific validation studies were used to adjust the results for measurement error.

RESULTS

Information about 4980 cases from studies including 337,819 women was available. When women in the highest quintile of energy-adjusted total fat intake were compared with women in the lowest quintile, the multivariate pooled relative risk of breast cancer was 1.05 (95 percent confidence interval, 0.94 to 1.16). Relative risks for saturated, monounsaturated, and polyunsaturated fat and for cholesterol, considered individually, were also close to unity. There was little overall association between the percentage of energy intake from fat and the risk of breast cancer, even among women whose energy intake from fat was less than 20 percent. Correcting for error in the measurement of nutrient intake did not materially alter these findings.

CONCLUSIONS

We found no evidence of a positive association between total dietary fat intake and the risk of breast cancer. There was no reduction in risk even among women whose energy intake from fat was less than 20 percent of total energy intake. In the context of the Western lifestyle, lowering the total intake of fat in midlife is unlikely to reduce the risk of breast cancer substantially.