A multi-marker integrative analysis reveals benefits and risks of bariatric surgery

Abstract

Bariatric surgery (BS) is an effective intervention for severe obesity and associated comorbidities. Although several studies have addressed the clinical and metabolic effects of BS, an integrative analysis of the complex body response to surgery is still lacking. We conducted a longitudinal data study with 36 patients with severe obesity who were tested before, 6 and 12 months after restrictive BS for more than one hundred blood biomarkers, including clinical, oxidative stress and metabolic markers, peptide mediators and red blood cell membrane lipids. By using a synthetic data-driven modeling based on principal component and correlation analyses, we provided evidence that, besides the early, well-known glucose metabolism- and weight loss-associated beneficial effects of BS, a tardive, weight-independent increase of the hepatic cholesterol metabolism occurs that is associated with potentially detrimental inflammatory and metabolic effects. Canonical correlation analysis indicated that oxidative stress is the most predictive feature of the BS-induced changes of both glucose and lipids metabolism. Our results show the power of multi-level correlation analysis to uncover the network of biological pathways affected by BS. This approach highlighted potential health risks of restrictive BS that are disregarded with the current practice to use weight loss as surrogate of BS success.

Figure 1

Flow diagram of the study. For each time of the study, the number of patients and the range (min–max) of patients’ BMI are displayed. The number of observations included in PCA analysis for each biomarker dataset is indicated in parentheses.

Figure 2

Principal components analysis. The main PCs extracted from each data set are presented in terms of percentage of explained variance and biological significance as inferred by component loading matrix (see Supplemental material for details).

Figure 3

Early and late effects of bariatric surgery as assessed by PCs temporal trends. PC values at baseline (T0), six (T6) and twelve (T12) months from BS are reported. Values are means with 95% CI. *p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.001, repeated measures ANOVA with contrasts. ECM extracellular matrix.

Figure 4

Graphical representation of the canonical correlation analysis. Each oval box represents a significant correlation (r) between linear combinations of PCs from two data set, with lines indicating the PCs from each data set significantly contributing to the combination. Thick lines indicate prominent contributions, i.e. the PC with a coefficient at least 1.5 times greater in modulus than that of any other PC in the same combination. PC box size is proportional to the variance explained by that PC within the dataset. ****p < 0.0001; ***0.0001 ≤ p < 0.001; **0.001 ≤ p < 0.01; *0.01 ≤ p < 0.05. Dashed lines indicate not statistically significant (0.05 < p < 0.15) though very strong (r > 0.7) correlations.