Supplemental material - Report on the comparison of laboratory measurements

Bias in protein and potassium intake collected with 24-h recalls (EPIC-soft) is rather comparable across European populations. European Journal of Nutrition.

Sandra P. Crispim[a],[b], Anouk Geelenb, Jeanne H. M. de Vriesb, Heinz Freislingc, Olga Souvereinb, Paul J. M. Hulshofb, Marga C. Ocke[c], Hendriek Boshuizenb,d, Lene F. Andersen[d], Jiri Ruprich[e], Willem De Keizer[f],[g], Inge Huybrechtsc,h, Lionel Lafay[h], Maria Santucci de Magistris[i], Fulvio Ricceri[j], Rosario Tumino[k], Vittorio Krogh[l], H. Bas Bueno-de-Mesquitad,[m], Joline WJ Beulens[n], Marie-Christine BoutronRuault[o], Androniki Naska[p], Francesca L Crowe[q], Heiner Boeing[r], Alison McTaggart[s], Rudolf Kaaks[t], Pieter van’t Veerb, Nadia Slimanic

With the purpose of harmonizing biomarker data used in the paper ‘Bias in protein and potassium intake collected with 24-h recalls (EPIC-Soft) is rather comparable across European populations’, a calibration study was conducted among laboratories that performed chemical analyses of urine samples from the EPIC and EFCOVAL studies.

Nutrients and specimens in the calibration study: The amount of urinary nitrogen and urinary potassium of 45 samples from the EPIC cohort were determined at both the MRC Dunn Clinical Nutrition Centre in Cambridge (EPIC) in the early 2000’s and at the Division of Human Nutrition at Wageningen University (EFCOVAL) in 2008.

Methodology: Agreement between the two biomarker measurements determined by the Cambridge and Wageningen laboratories was assessed using Bland-Altman plots and Pearson’s coefficients of correlation. The paired t-test was used to test for significant differences (p<0.05) between the mean biomarker measurements from the two laboratories. In addition, comparability of laboratory methods used in EPIC and EFCOVAL labs was further substantiated by evaluating standard reference materials and quality control procedures (e.g., inter-laboratory proficiency tests) of the lab measurements.

Linear regression analyses were carried out to generate calibration equations between the biomarker measurements from the two laboratories. Protein and potassium values which were +/- 3SD from the mean were considered outliers (n = 1 for nitrogen and for potassium) and excluded from the regression analyses.

Results: A good agreement and correlation was seen between the measurements from the two laboratories based on the Bland-Altman plots (Figure S1 and S2) and Pearson’s correlation coefficients (r>0.97), respectively (Table S1). No significant differences were seen between the mean nitrogen (p=0.10) and potassium (p=0.68) measurements from the two laboratories.

Based on those results, we judged that calibration of data between EFCOVAL and EPIC studies was not necessary and original biomarker data of the two studies was used in the main analyses of the paper.

Because the calibration study was limited by the rather small sample size, we still estimated the calibration equations and performed a sensitivity analysis using the calibrated data in the multilevel analysis. The EFCOVAL data was calibrated using the following regression formulas:

EFCOVAL_N_calibrated = 0.92 x EFCOVAL_N+ 0.03

EFCOVAL_K_calibrated = 0.92 x EFCOVAL_K + 1.87

Standard reference materials and inter-laboratory proficiency tests of the laboratory procedures in the two laboratories indicated no discrepancies between the analyses performed. Thus, there was no evidence to opt for having one or the other laboratory as reference in the calibration procedure, and the Cambridge lab (from the EPIC study) was chosen.

Table S1 – Comparison of means (SD) and Pearson’s correlation coefficients between biomarker measurements determined by the Wageningen and Cambridge laboratories in the calibration study

Biomarker / n / Wageningen / Cambridge / p[u] / r[v]
Urinary nitrogen (g/100ml) / 45 / 0.70 (0.3) / 0.68 (0.3) / 0.10 / 0.97
Urinary potassium (mmol/L) / 45 / 42.1 (17.3) / 40.6 (16.2) / 0.68 / 0.99

Figure S1 - Bland-Altman plot for nitrogen measurements from the Cambridge and Wageningen laboratories.

Figure S2 - Bland-Altman plot for potassium measurements from the Cambridge and Wageningen laboratories.

[a]Division of Human Nutrition, WageningenUniversity, Wageningen, The Netherlands.

[b]International Agency for Research on Cancer (IARC), Lyon, France.

[c]National Institute for Public Health and the Environment (RIVM), Bilthoven, The Netherlands.

[d]Department of Nutrition, Institute of Basic Medical Sciences, University of Oslo, Oslo, Norway.

[e]National Institute of Public Health, Department for Food Safety and Nutrition, Brno, Czech Republic.

[f]Department of Nutrition and Dietetics, UniversityCollegeGhent, Ghent, Belgium.

[g]Department Public Health, GhentUniversity, Ghent, Belgium.

[h]French Agency for Food, Environmental and Occupational Health Safety (ANSES), Paris, France.

[i]Department of Public Clinical and Experimental Medicine, Federico II, Naples, Italy.

[j]Human Genetics Foundation (HUGEF), Turin, Italy.

[k]Cancer Registry and Histopathology Unit, "Civile M.P.Arezzo" Hospital,Ragusa, Italy.

[l]Nutritional Epidemiology Unit, Fondazione IRCCS Istituto Nazionale Tumori, Milan, Italy.

[m]Department of Gastroenterology and Hepatology, University Medical Centre Utrecht (UMCU), Utrecht, The Netherlands.

[n]JuliusCenter for Health Sciences and Primary Care, UniversityMedicalCenterUtrecht, Utrecht, The Netherlands.

[o]Inserm, Centre for Research in Epidemiology and Population Health, Villejuif, France.

[p]WHO Collaborating Center for Food and Nutrition Policies, Department of Hygiene, Epidemiology and Medical Statistics, University of Athens Medical School, Athens, Greece.

[q]Cancer Epidemiology Unit, Nuffield Department of Clinical MedicineUniversity of Oxford, Oxford, United Kingdom.

[r]Department of Epidemiology, German Institute of Human Nutrition, Potsdam-Rehbrücke, Germany

[s]Department of Public Health and Primary Care, University of Cambridge, Cambridge, United Kingdom.

[t]Division of Cancer Epidemiology, GermanCancerResearchCenter, Heidelberg, Germany.

[u] Paired t-test

[v] Pearson correlation