Kostens inverkan på halten fungicidrester i kroppen : En Bayesiansk multilevelanalys

Sammanfattning: Today a growing number of people are transitioning to a more plant based diet. At the same time plant basedproducts often contain higher residue levels of pesticides compared to products derived from animals. Consumingmore plant based food could therefore be seen as a potential health risk. The pilot study The ClimateFriendly and Ecological Food on Microbiota (CLEAR) from 2017 examined the relationship between diet andpesticides. The 30 participants were assigned to one of three groups: control, climate-friendly and ecological.During the study the participants were at three occasions asked to fill in a questionnaire concerning, amongother things, eating habits. They were also asked to leave a urine sample, which was analyzed for residuesof four pesticides. This thesis analyzed the two pesticide residues associated with fungicides: propamocarband ethylenethiourea. These two residues were combined to a single variable. The thesis aims to answer thetwo following questions: Are there any differences in residue levels of fungicides between people who follow aconventional, climate friendly or ecological diet? What is the relationship between the consumption of plantbased food (vegetables, fruit, cereals etc.) and residue levels of fungicides? These questions were previouslyexplored by Berg (2018) with an ANOVA-model and by examining correlations. Berg found no significantdifferences between the groups and no strong correlations between residues of fungicides and food intake.In this thesis we instead attempt to answer the questions using Bayesian multilevel models, which are moresuited to the structure of the data compared to methods that have been used previously. The analysis showsthat there are no differences in residue levels between those who were prescribed a conventional, climatefriendly or ecological diet. There are also no unequivocal relationships between residue levels and food intake.However, posterior probabilities show that it is likely that the residue levels are lower for the ecological groupcompared to the control group. Posterior probabilities also show that there are some more likely relationshipsfor the food groups cereals and rice, fruits and berries, as well as for the products wine, coffee and chocolate.To reach any conclusions regarding these relationships further studies are required. The results in this thesisare to a large degree deemed to depend on the uncertainty that follows from a small sample size. The Bayesianmultilevel models have however given a more thorough understanding of the relationships than was achievedwith previously used methods.