Adipokines in anorexia nervosa: A systematic review and meta-analysis.

Objective: The association between adipokine dysregulation and weight loss of patients with anorexia nervosa (AN) has been long investigated, in search of a causal relationship. We sought to: a) synthesize the available evidence on potential differences between AN patients and controls with regards to adipokine measurements (namely, leptin, adiponectin, resistin, soluble leptin receptor, visfatin, vaspin and omentin), b) estimate the potential differences between constitutionally thin (CT) subjects and AN patients, and c) present the available evidence with regards to biomarker efficacy of adipokines in AN.

Methods: A structured literature search, last updated in 2/2019, was conducted in the following databases: MEDLINE, clinicaltrials.gov, PsycINFO, PSYNDEX and WHO Registry Network. The primary outcome was the standardized mean difference of each adipokine between AN patients and controls of normal BMI. Secondary outcomes included the correlation of leptin with BMI and bone mineral density among AN patients. The study protocol is published in PROSPERO (CRD42018116767).

Results: In a total of 622 screened studies, after exclusion of non-relevant articles and duplicates, 84 reports on leptin, 31 reports on adiponectin, 12 on resistin, 10 on soluble leptin receptor, 5 on visfatin, 3 on vaspin and omentin were finally included in the meta-analysis. Publication bias assessment underlined the possibility of non-significant studies being underrepresented; still, significant heterogeneity renders this statement inconclusive. Leptin [ELISA: SMD (95% CI): -3.03 (-4, -2.06)], radioimmunoassay [RIA: -3.84 (-4.71, -2.98)] and resistin [-1.67 (-2.85, -0.48)] were significantly lower in patients with AN compared with controls, whereas visfatin decrease did not reach significance (-2.03 (-4.38, 0.3). Mean adiponectin, vaspin and soluble leptin receptor levels were significantly higher. In subgroup analysis, a significantly attenuated SMD was reported in ELISA studies compared with RIA studies. Leptin was significantly lower in AN patients compared to CT subjects and BMI marginally did not appear to confound the result. In all analyses, except for the correlation of leptin with BMI in AN patients, high heterogeneity was present. Meta-regression analysis indicated a potential confounding action of controls' BMI and age on leptin SMD and between-assay differences. Publication bias assessment underlined the possibility of nonsignificant studies being underrepresented; still, further investigation did not corroborate this and significant heterogeneity renders this statement inconclusive.

Conclusion: A distinct profile of adipokine dysregulation is apparent in AN patients, following the anticipated pattern of low BMI. A precise estimation of the magnitude is hindered by heterogeneity, partly caused by varying assays and methodologies. Interestingly, while mean leptin levels are lower in AN subjects compared with constitutionally thin women, there is an overlap in individual levels between the two groups and therefore, they cannot be used to differentiate between these states.