Metabolic profiling and population screening of analgesic usage in nuclear magnetic resonance spectroscopy-based large-scale epidemiologic studies

Abstract

The application of a (1)H nuclear magnetic resonance (NMR) spectroscopy-based screening method for determining the use of two widely available analgesics (acetaminophen and ibuprofen) in epidemiologic studies has been investigated. We used samples and data from the cross-sectional INTERMAP Study involving participants from Japan (n = 1145), China (n = 839), U.K. (n = 501), and the U.S. (n = 2195). An orthogonal projection to latent structures discriminant analysis (OPLS-DA) algorithm with an incorporated Monte Carlo resampling function was applied to the NMR data set to determine which spectra contained analgesic metabolites. OPLS-DA preprocessing parameters (normalization, bin width, scaling, and input parameters) were assessed systematically to identify an optimal acetaminophen prediction model. Subsets of INTERMAP spectra were examined to verify and validate the presence/absence of acetaminophen/ibuprofen based on known chemical shift and coupling patterns. The optimized and validated acetaminophen model correctly predicted 98.2%, and the ibuprofen model correctly predicted 99.0% of the urine specimens containing these drug metabolites. The acetaminophen and ibuprofen models were subsequently used to predict the presence/absence of these drug metabolites for the remaining INTERMAP specimens. The acetaminophen model identified 415 out of 8436 spectra as containing acetaminophen metabolite signals while the ibuprofen model identified 245 out of 8604 spectra as containing ibuprofen metabolite signals from the global data set after excluding samples used to construct the prediction models. The NMR-based metabolic screening strategy provides a new objective approach for evaluation of self-reported medication data and is extendable to other aspects of population xenometabolome profiling.

Figure 1

Schematic diagram showing the study design for the generation of prediction models for acetaminophen.

Figure 2

Box-plot of minimum and maximum number of errors in predicting ¹H NMR urine spectra, also the mean number of errors ± one standard deviation, after 1000 re-sampling iterations for (A) spectra containing acetaminophen related metabolites (n = 175); (B) spectra not containing acetaminophen related metabolites (n = 275); (C) percentage of overall model error rates, calculated based on the mean number of spectra incorrectly predicted for spectra with and without acetaminophen metabolites.

Figure 3

Histograms of number of errors in predicting specimens with acetaminophen related metabolites (red bars) and specimens without acetaminophen related metabolites (blue bars) in each of the 1000 re-samplings for spectra with the presence or absence of acetaminophen visually verified for (A) model PCT10, based on mean-centred data; (B) model PCT11, based on unit-variance scaling and (C) model PCT12, based on pareto scaling. Results for all other models, see Supplementary Figure 4 and Supplementary Figure 5.

Figure 4

Histograms of overall distribution, number of errors for model IBU1, based on mean-centred data, for predicting (A) spectra with ibuprofen and (B) without ibuprofen related metabolites.