Study design and subjects
The ScenoProt intervention study, conducted at the University of Helsinki, included 20- to 69-year-old healthy adult participants. Their body mass index (BMI) ranged between 18.5 and 35.0 kg/m2, and they volunteered to follow any of the three diets for the 12-week controlled trial in a parallel design. We recruited participants through newspaper advertisements, social media, and the university´s mailing lists. Exclusion criteria were fasting plasma glucose > 6.9 mmol/l and total cholesterol > 6.5 mmol/l, use of medication for diabetes or hypercholesterolemia, disorders of the intestinal or endocrine systems or of lipid metabolism, renal, or liver diseases, eating disorder, any malignant illness within the past five years, recent use of antibiotics (during the last three months), food allergies, extreme sports, smoking, and current pregnancy or lactation. Participant enrolment is shown in Fig. 1.
The screening was implemented from December 2016 to the beginning of March 2017, and the intervention periods began between January and March 2017. Altogether, 146 participants passed the screening and were stratified by sex and age and randomly allocated (allocation ratio 1:1:1) by the study PI to one of the three diet groups: ANIMAL, an animal-source protein diet representing an average Finnish diet and containing 70% animal- and 30% plant-source protein; 50/50, containing equal amounts of animal- and plant-source proteins; and PLANT, a plant-source protein diet containing 30% animal- and 70% plant-source protein. Participants were unaware of their diet group until they had completed baseline measures, after which their diet group was introduced using a colour code for each diet. The colour codes were used throughout the study. Participants were advised to discontinue use of dietary supplements and herbal or other natural remedies 2 weeks prior to the intervention period. Power calculation and sample size estimation of this research has been described in detail elsewhere [23, 24]. No power calculations regarding vitamin or mineral intakes or their biomarkers presented in this paper were carried out.
Intervention diets
The intervention diets and compliance have been described in detail elsewhere (23, 24). In short, the three diets were designed to contribute, on average, 17% of energy from protein. The amount of red meat, poultry, and dairy products differed among the diet groups, being highest in the ANIMAL group, whereas the amounts of eggs and fish were the same in each diet group. In the 50/50 and PLANT diets, red meat, poultry and dairy products were partially replaced with plant protein sources such as cereal products, vegetable dishes including peas, lentils, chickpeas, tofu, faba beans, nuts, almonds, seeds, and plant-based dairy substitutes (Table 1). The food items supplied by the study provided on average 80% of the daily energy intake in all diet groups [23]. Participants were allowed to consume fruits, berries, dietary fats and oils, confectionery products, and other than plant-based dairy or vegetable substitute beverages as they usually did; they acquired these foods themselves.
Table 1 Consumption frequencies and daily consumption of specific foods and food groups in the intervention diets based on the delivered food items and diet instructions The participants visited the research unit weekly, when they were provided with most of their protein sources by the study to be consumed at home: meat, poultry, ready-made plant-protein products, bread, nuts and seeds, pulses (such as pea flower and dried pea groats), and fish. None of the ready-made plant-based foods or plant-based dairy substitutes supplied in this intervention were fortified with vitamin B-12, folic acid, iodine, iron, or zinc. The breads supplied contained iodized salt. The participants were instructed on how to implement their diet at a food level, and they received recipes for cooking.
Assessment of baseline characteristics
The background data on age, sex, education, and previous use of dietary supplements (the type of the supplement but not the dose) were collected by a questionnaire. Multivitamins were assumed to include folic acid, vitamin B-12, iron, zinc, and iodine. In additional analysis, the data were dichotomized to previous users and non-users of vitamin B-12, folate, iodine, iron, and zinc supplements. BMI was calculated as weight (kg)/height (m2).
Assessment of dietary intake
Four-day food records, including three weekdays and one weekend day, were collected prior to the start of the intervention and during the last week of the intervention. The participants were instructed to record portion sizes using weighing, package labels, or household measures. The baseline food records from two participants were not available. Food records were reviewed by nutritionists, and missing information was requested if needed. Vitamin and mineral intakes were calculated by the AivoDiet software (version 2.2.0.1, Aivo Oy, Turku, Finland), including the Fineli® Food Composition Database Release 16 (2013), maintained by the Finnish Institute for Health and Welfare [25]. Vitamin and mineral intakes at baseline and end point were calculated as means of daily intake. Each food item or mixed dish in the dataset was assigned to a food group, and retention factors (EuroFIR) [26] of folate, vitamin B-12, vitamin C, iron, and zinc were applied with a single factor per nutrient per food group. Total, animal-derived, plant-derived, and other sources of iron intake were calculated separately. All composite dishes and single food items were classified into 14 food categories based on a modified classification of the Fineli® database, and the proportions of the sources for nutrient intakes were calculated for each category. For analysing animal- and plant-derived iron intakes and the share of iodized salt of the total iodine intake, the data were disaggregated into ingredients of composite dishes and single food items and these were classified to 17 different ingredient categories, also based on a modified classification of the Fineli® database.
Assessment of nutrient biomarkers
Blood samples were collected after overnight fast (10–12 h) at screening, baseline, and end point visits, and stored at − 70 °C until analysis. Plasma ferritin and transferrin receptor (TfR) concentrations, serum folate and holotranscobalamin II (holoTC), haemoglobin and high-sensitive C-reactive protein (hs-CRP) were analysed at Helsinki University Hospital Laboratories, Helsinki, Finland. All samples were analysed according to accredited standard methods. All biomarkers used were considered to most reliably represent the status of each nutrient [27,28,29,30,31]. In this study, no biomarker for zinc was examined because of the lack of sensitive clinical criteria to evaluate marginal or low zinc status [32].
Serum folate and holoTC were measured by immunochemiluminometric methods and plasma ferritin by photometric methods using the Abbott Architect iSR2000. The intra-assay coefficient of variation percentage (CV%) for folate was 8%, for holoTC 7%, and for ferritin 5%. Inter-assay CV% for folate was < 10% within the range 6–18 nmol/l, for holoTC < 7% within the range of 15–50 pmol/l, and for ferritin < 5% within the range of 20–323 µg/l. Plasma TfR was measured by immunoturbidimetric assay performed on the Abbott Architect c16000. For TfR, intra-assay CV% was 5.8% and inter-assay CV% < 8% within the range of 0.5–1.1 mg/l. Blood haemoglobin was analysed using photometric methods with inter-assay CV% ≤ 1.0% within the whole measurement range.
The participants collected 24-h urine samples at the beginning and at the end point of the intervention. 24-h urine iodine excretion (U-I) at the end point of the study was calculated based on urinary volume and urinary iodide concentration analysed at Vita laboratories in Helsinki by inductively coupled plasma - mass spectrometry with the Agilent 7700 ICP-MS. The CV% was 1.7% for intra-assay and 8.1% for inter-assay variation.
Statistical analysis
Data are presented as means ± SD. Daily intakes are expressed also as energy-adjusted (µg or mg/MJ). Education was dichotomized to upper secondary or less and Bachelor’s degree or higher. For the analysis regarding folate and iron intake, the female participants were dichotomized according to Finnish mean age of menopause [33]: to ˂ 51 years (reproductive age) and ≥ 51 years (peri- and postmenopausal age). Differences among the diet groups in the intakes of vitamins and minerals at the end point were compared by one-way analysis of variance (ANOVA). Energy intake (MJ) was tested as a covariate for plant and animal-derived iron, but no change in the significance of the results was observed. The differences between nutrient biomarkers among dietary supplement users and non-users were tested by independent samples t tests. We contrasted nutrient biomarker concentration among the diet groups by analysis of covariance (ANCOVA), with adjustments made with the covariate: baseline concentrations, followed by post hoc comparisons, adjusted by the Bonferroni method. In addition, BMI, energy intake, age, sex, history of dietary supplement use for all biomarkers, and hs-CRP as indicator of iron status were verified as covariates, but no change in the significance of the results was observed. Multivariate significance test for indicators of iron status was done by using the multivariate linear model, where baseline values were used as covariates. Folate status was analysed separately for folic acid supplement users and non-users. We analysed biomarkers of iron status separately for male and female participants. The baseline value of TfR was missing for one subject in the 50/50 group and another subject in the PLANT group, as was the end point value of ferritin for one subject in the ANIMAL group. As an exception to other nutrient biomarker analyses, U–I was analysed by one-way ANOVA because baseline data were unavailable. The associations between dietary intake and nutrient biomarkers were determined with Spearman correlation.
P values < 0.05 were considered significant. Data analyses were done with R software (version 3.5.1 and Studio) (R Core Development Team) and with SPSS version 24 (25) (IBM).