2 CHAPTER 2 36 Demographic characteristics The demographic characteristics of the study population of whom good 1H-NMR data were obtained are shown in Table 1. Migraine patients tended to be younger (p = 0.013) and more often were female than controls (p < 0.001). In addition, lifetime migraine patients more often than controls were smokers (44.6% cases vs. 35.1% controls p = 0.006). No difference in BMI was observed between cases and controls (p = 0.934). Of the 289 lifetime migraine patients, 150 (52%) reported at least one severe migraine attack in the 12 months preceding the interview and were assigned to the group of “active migraine patients”. The active migraine patients consisted of 124 women (83%), had a mean age of 44 (SD ± 11.4), 71 (47.3%) were currently smoking and had a mean BMI of 26 (SD ± 4.9). Next, we assessed the influence of age, sex, BMI and smoking (see Additional file 2) on the weighted metabolite score. All covariates showed to be of influence on the weighted metabolite score and were added to the logistic regression model. Association of metabolites with lifetime migraine diagnosis Elastic net regression analysis of all 289 migraine patients and 1,360 controls for all 100 signals identified six 1H-NMR signals as the best prediction subset. These signals were representative of four different metabolites (isoleucine, methionine, 1,5-anhydrosorbitol and creatine) and one unknown signal (Table 2). Subsequent logistic regression analysis showed support for association (odds ratio (OR) = 2.72; 95% confidence interval (CI) 1.97 - 3.75; p = 1.28 x 10-9) explaining 3.9% of the variance in migraine status (Nagelkerke R2). After correction for age, sex, BMI and smoking the association no longer showed support for association (OR = 1.49; 95% CI 0.99 - 2.23; p = 0.051). Table 1 Demographic characteristics Variable Lifetime migraine patients (N = 289) Controls (N = 1360) p-value Active migraine patientsb (N = 150) p-value Age (years) 46.5 ± 12.1 48.7 ± 14.5 0.013c* 44.0 ± 11.4 < 0.001c* Female sex (%) 220 (76.1) 673 (49.5) < 0.001d* 124 (82.7) < 0.001d* BMI 26.9 ± 5.0 26.8 ± 4.6 0.934c 26.3 ± 4.9 0.219c Smokinga (yes) (%) 129 (44.9) 481 (36.0) 0.006d* 71 (47.3) 0.008d* MO patients 163 (56.4) - - 77 (51.3) - Values are expressed as absolute values and percentage or mean ± SD. Numbers and proportions may not add up to total of 100 due to rounding or missing values. aDefined as currently cigarette smoking. bDefined as having at least one severe migraine attack in the last 12 months. cStudent’s t-test. dChi-square Test. *Significant p-values (p < 0.05). Missing values in lifetime migraine patients for BMI (n = 2), smoking status (n = 2) and in controls for BMI (n = 24) smoking status (n = 27). MO = migraine without aura, BMI = body mass index Association of metabolites with active migraine diagnosis Next, we performed an elastic net regression on all 150 active migraine patients and 1,360 controls for all 100 signals. This analysis identified 22 predictive signals. The subsequent logistic regression analysis was performed on 146 cases and 1,343 controls, as not all subjects had sufficient signal data
RkJQdWJsaXNoZXIy MTk4NDMw