OBJECTIVES:To compare machine learning (ML) to traditional models to predict radiographic progression in patients with early axial spondyloarthritis (axSpA).METHODS:We carried out a prospective... Show moreOBJECTIVES:To compare machine learning (ML) to traditional models to predict radiographic progression in patients with early axial spondyloarthritis (axSpA).METHODS:We carried out a prospective French multicentric DESIR cohort study with 5 years of follow-up that included patients with chronic back pain for <3 years, suggestive of axSpA. Radiographic progression was defined as progression at the spine (increase of at least 1 point of mSASSS scores/2 years) or at the sacroiliac joint (worsening of at least one grade of the mNY score between 2 visits). Statistical analyses were based on patients without any missing data regarding the outcome and variables of interest (295 patients). Traditional modelling: we performed a multivariate logistic regression model (M1); then variable selection with stepwise selection based on Akaike Information Criterion (stepAIC) method (M2), and Least Absolute Shrinkage and Selection Operator (LASSO) method (M3). ML modelling: using “SuperLearner” package on R, we modelled radiographic progression with stepAIC, LASSO, random forest, Discrete Bayesian Additive Regression Trees Samplers (DBARTS), Generalized Additive Models (GAM), multivariate adaptive polynomial spline regression (polymars), Recursive Partitioning And Regression Trees (RPART) and Super Learner. Accuracy of these models was compared based on their 10-fold cross-validated AUC (cv-AUC).RESULTS:10-fold cv-AUC for traditional models were 0.79 and 0.78 for M2 and M3, respectively. The three best models in the ML algorithms were the GAM, the DBARTS and the Super Learner models, with 10-fold cv-AUC of: 0.77, 0.76 and 0.74, respectively.CONCLUSIONS:Two traditional models predicted radiographic progression as good as the eight ML models tested in this population. Show less
Background: Strong evidence indicates that venous thromboembolism is a presenting symptom of cancer. Cancer is a known cause of pulmonary hypertension; however, it remains unknown whether pulmonary... Show moreBackground: Strong evidence indicates that venous thromboembolism is a presenting symptom of cancer. Cancer is a known cause of pulmonary hypertension; however, it remains unknown whether pulmonary hypertension is a marker of occult cancer. We examined the association between a pulmonary hypertension diagnosis and cancer risk in a cohort study using population-based data from the Danish health system. Patients and Methods: Using Danish nationwide registries, we identified 6335 patients with a pulmonary hypertension diagnosis and without a previous cancer diagnosis between 1995 and 2017. We computed the age-, sex-, and calendar year-standardized incidence ratio (SIR) as the ratio of observed to expected number of cancers using national incidence rates as the reference. We performed a subgroup analysis among patients with chronic thromboembolic pulmonary hypertension in the period in which a specific ICD-10 code was available (2006-2017). Results: We identified 212 cancers within the first year of follow-up and 796 cancers thereafter. The one-year risk of cancer was 3.3% and the one-year SIR was 1.96 (95% confidence interval [CI]: 1.70-2.23). In the second and subsequent years, the SIR remained elevated (SIR: 1.15 [95% CI: 1.08-1.24]). In patients with chronic thromboembolic pulmonary hypertension, the one-year SIR was Conclusion: Cancer risk was clearly higher in patients with pulmonary hypertension compared with the general population. The association was particularly strong in the first year of follow-up, but remained elevated thereafter. However, absolute risks were low, limiting the clinical relevance of pursuing early cancer detection in these patients. Show less
Objective: We aimed to validate a food-frequency questionnaire (FFQ) for Dutch pregnant women, against three 24 h-recalls and blood concentrations of B-vitamins and fatty acids, using the method of... Show moreObjective: We aimed to validate a food-frequency questionnaire (FFQ) for Dutch pregnant women, against three 24 h-recalls and blood concentrations of B-vitamins and fatty acids, using the method of triads. Methods: We included 83 pregnant women from the general population of Rotterdam, the Netherlands, at a median gestational age of 15.6 weeks. Participants completed three non-consecutive 24 h-recalls, and subsequently filled out the 293-item FFQ. Participants provided blood samples from which we analyzed serum folate and vitamin B12, as well as red blood cell folate, linoleic acid, and total saturated, monounsaturated, and polyunsaturated fatty acids. Results: Estimated energy intake did not differ between the FFQ and 24 h-recalls. Deattenuated Pearson's correlation coefficients, between energy-adjusted nutrient intake estimates from the FFQ and the 24 h-recalls, ranged from 0.41 (fat) to 0.88 (fiber) for macronutrients, and were around 0.6 for most micronutrients, except for vitamin E (0.27). Using the triad method, we obtained validity coefficients of 0.86 (95% Confidence Interval (CI) 0.36, 1.00) for serum folate, 0.86 (95% CI 0.18, 1.00) for red blood cell folate, and 1.00 (95% CI 0.42, 1.00) for vitamin B12. Validity coefficients for serum fatty acids ranged from 0.22 to 0.67. Conclusion: This FFQ is a reliable tool for estimating intake of energy, macronutrients, folate and vitamin B12 among women in mid-pregnancy. Show less
Neeland, I.J.; Boone, S.C.; Mook-Kanamori, D.O.; Ayers, C.; Smit, R.A.J.; Tzoulaki, I.; ... ; Mutsert, R. de 2019
Background-Identifying associations between serum metabolites and visceral adipose tissue (VAT) could provide novel biomarkers of VAT and insights into the pathogenesis of obesity-related diseases.... Show moreBackground-Identifying associations between serum metabolites and visceral adipose tissue (VAT) could provide novel biomarkers of VAT and insights into the pathogenesis of obesity-related diseases. We aimed to discover and replicate metabolites reflecting pathways related to VAT.Methods and Results-Associations between fasting serum metabolites and VAT area (by computed tomography or magnetic resonance imaging) were assessed with cross-sectional linear regression of individual-level data from participants in MESA (Multi-Ethnic Study of Atherosclerosis; discovery, N=1103) and the NEO (Netherlands Epidemiology of Obesity) study (replication, N=2537). Untargeted H-1 nuclear magnetic resonance metabolomics profiling of serum was performed in MESA, and metabolites were replicated in the NEO study using targeted H-1 nuclear magnetic resonance spectroscopy. A total of 30 590 metabolomic spectral variables were evaluated. After adjustment for age, sex, race/ethnicity, socioeconomic status, smoking, physical activity, glucose/lipid-lowering medication, and body mass index, 2104 variables representing 24 nonlipid and 49 lipid/lipoprotein subclass metabolites remained significantly associated with VAT (P=4.88 x 10(-20)-1.16 x 10(-3)). These included conventional metabolites, amino acids, acetylglycoproteins, intermediates of glucose and hepatic metabolism, organic acids, and subclasses of apolipoproteins, cholesterol, phospholipids, and triglycerides. Metabolites mapped to 31 biochemical pathways, including amino acid substrate use/metabolism and glycolysis/gluconeogenesis. In the replication cohort, acetylglycoproteins, branched-chain amino acids, lactate, glutamine (inversely), and atherogenic lipids remained associated with VAT (P=1.90 x 10(-35)-8.46 x 10(-7)), with most associations remaining after additional adjustment for surrogates of VAT (glucose level, waist circumference, and serum triglycerides), reflecting novel independent associations.Conclusions-We identified and replicated a metabolite panel associated with VAT in 2 community-based cohorts. These findings persisted after adjustment for body mass index and appear to define a metabolic signature of visceral adiposity. Show less