CHAPTER 10 216 10 dietary factors or medication. Logically, when a study is subjected to confounding the results do not faithfully reflect the relationship between the variables under study. From a research perspective it is therefore sensible to account for such factors as much as possible, but it is impossible and perhaps even undesirable to take all into account. Factors that can influence the metabolome can be divided into two; 1) influenceable factors, such as storage of samples and time of day of sample collection; 56-59 2) uninfluenceable factors, such as sex and age.60, 61 Sampling is an important step in research into metabolites, as discussed in the General Introduction (Chapter 1), and should be planned and carried out meticulously as changes in handling of a sample can have a large effect on the outcome, already because of the (sometimes) extreme variability in stability of certain metabolites and lipids.58 A way to control the data for these changes is protocolized sampling, as we have done in our metabolomics studies. Unfortunately, not in all studies collection and storing of samples is done in such a protocolized fashion or even described in the respective papers62-66. Demographic variables such as age, sex, ethnicity, smoking behaviour and obesity are frequent confounders in omics association studies. It is known that sex and age are often the principal factors explaining metabolome variability.60, 67 Migraine prevalence differs between men and women and lifestyle characteristics, such as smoking and alcohol intake, differ from those without migraine.68-70 Given these differences in prevalence they are important conceivable confounding factors that need to be considered in migraine research to avoid a false-positive (Type I) error. It is possible to either account for confounding factors through experimental design, by means of randomization, restriction and matching before data gathering or after the data gathering process by means of statistical analyses, via stratification or multivariate models.71, 72 Endocannabinoids for instance, are involved in a large number of pathways and processes.73 Several endocannabinoids have been associated with body mass index (BMI), sex, age, weekly alcohol consumption and smoking.73-75 In addition, there is also a well-known relation between depression and endocannabinoids.76-81 Energy homeostasis is one of the processes best known to be controlled by the endocannabinoid system.73, 82 Thus, the endocannabinoid system affects the food intake and indirectly the BMI.73, 83 Although the pathways underlying sex differences in the endocannabinoid system are not fully elucidated, it is generally accepted that differences exist between the male and female endocannabinoid system.84-86 Moreover, studies in mice have found that chronic exposure to ethanol leads to increased levels of AEA and 2-AG.87-89 Smoking is expected to also play a role in the endocannabinoid system as it is involved in the reward circuits of the brain. There is some evidence on a decrease in the expression of cannabinoid receptor 1 (CB1) molecules and an increase in the AEA level in smokers. 90, 91 Given that these factors could influence both the exposure and the outcome, all these factors were taken into account when studying endocannabinoid levels in Chapter 3 when we were solely interested on the endocannabinoid-migraine relationship. Comparing our study with similar studies there was a considerable clinical and methodological heterogeneity across studies. In previous studies on endocannabinoid levels in patients there is only mention of age-matching in the design of one
RkJQdWJsaXNoZXIy MTk4NDMw