Multiple Sclerosis Journal - October 2017 - 1462

Multiple Sclerosis Journal 23(11)
QC, Canada/Department
of Medicine, McGill
University, Montréal, QC,
Canada/Department of
Twin Research and Genetic
Epidemiology, King's
College London, London, UK

Obesity is associated with chronic, mild inflammation
characterized by abnormal cytokine production and
increased pro-inflammatory signaling. Adipose tissue
is known to produce cytokines known as adipokines;
however, the relative contribution of adipocytederived cytokines to the inflammatory state in obesity
is unknown.10 Interestingly, adiponectin, an adipokine
with known anti-inflammatory properties in both the
innate and adaptive arms of the immune system,10 is
reduced in overweight and obese individuals11,12 and
is negatively correlated with BMI.11
In light of these findings, animal and human studies
alike have been undertaken to better understand adiponectin's role in MS etiology and treatment.
Results from studies using experimental autoimmune encephalomyelitis (EAE) models of MS are
suggestive of a protective role for adiponectin in
rodents.13,14 However, findings from studies in clinical populations are diverse. One study showed
reduced levels of this adipokine in peripheral blood
of MS patients following acute relapse,15 while others demonstrate elevated adiponectin in peripheral
blood and cerebrospinal fluid (CSF) of patients in
remission16,17 or unaltered adiponectin in newly
diagnosed MS patients.18
Observational studies, such as those described above,
represent an important step in the identification of
risk factors in disease. Randomized control trials
(RCTs) and/or MR studies can help to clarify the roles
of identified risk factors in disease outcome, as findings of observational studies may be biased due to
residual confounding and/or reverse causation.
Indeed, numerous RCTs and MR studies have provided strong evidence for the presence of bias in previously reported observational associations (many
examples reviewed in Mokry et al.19). However, these
types of studies can also validate observational associations through demonstration of causality (also
reviewed in Mokry et al.19). One such MR study9 suggested that previously reported observational associations between body weight and MS5-8 are not likely
biased due to confounding or reverse causation.
Nonetheless, no study to date has provided such evidence for the reported observational association
between adiponectin level and MS. Confounding due
to reverse causation is of particular concern in epidemiological studies of MS, as timing of disease onset is
unknown. Therefore, the nature of adiponectin's role
in MS etiology therefore merits further investigation.
In the absence of experimental studies investigating
adiponectin's role in MS clinical populations, MR
studies can be conducted to evaluate adiponectin's

1462

role in disease outcome in a manner that allows for
causal inference. This approach is conceptually similar to a RCT, where instead of randomization to a
pharmaceutical intervention, individuals in the population are naturally randomized at conception to varying levels of an exposure (e.g. adiponectin level) due
to genetic variation.
MR is a technique which uses genetic variants
strongly associated with an exposure (e.g. adiponectin
level) to estimate the exposure's effect on disease risk
(e.g. MS).20 Since genetic variants are randomly allocated at meiosis, they are not influenced by confounding factors that may bias observational associations,
except confounding by ancestry. Furthermore, reverse
causation is overcome since allelic randomization
always precedes MS onset.
To better understand whether adiponectin levels may
influence risk of MS, we undertook an MR study of
adiponectin on MS risk using a two-sample MR
design, deriving the effects of single nucleotide polymorphisms (SNPs) on adiponectin and MS risk from
the largest adiponectin and MS samples available to
date: the ADIPOGen Consortium (N = 45,891),21 the
International Multiple Sclerosis Genetics Consortium
(IMSGC, ncases = 14,498/ncontrols = 24,091),22 and the
IMSGC/Wellcome Trust Case Control Consortium 2
(IMSGC/WTCCC2, ncases = 9772/ncontrols = 17,376).23
Methods
SNP selection, effect sizes, and data sources
Genome-wide significant (p < 5 × 10−8) genetic variants associated with adiponectin levels were obtained
from ADIPOGen.21 For this study, we limited our
selection of SNPs and summary statistics to those
that achieved genome-wide significance in the
European sex-combined discovery phase analyses or
joint analyses (30,708 ⩽ n ⩽ 38,276). The effect of
each SNP on natural-log-transformed adiponectin
levels was adjusted for age, sex, BMI, the principle
components of ancestry, study site (where appropriate), and family structure in cohorts with family
members.21 Corresponding effect estimates of the
adiponectin-associated SNPs on risk of MS were
obtained first from the IMSGC Immunochip study,
the largest genetic association study for MS (14,498
cases and 24,091 controls)22 and then from the second largest study, the IMSGC/WTCCC2 (9772 cases
and 17,376 controls),23 if an adiponectin-associated
SNP was not ascertained in the IMSGC Immunochip
study. We have previously used these datasets to
explore the effects of BMI and vitamin D on risk of
journals.sagepub.com/home/msj
https://journals.sagepub.com/home/msj
Table of Contents for the Digital Edition of Multiple Sclerosis Journal - October 2017
