nonalcoholic fatty liver disease and risk of incident type ...nafld was based on either imaging or...

Nonalcoholic Fatty Liver Diseaseand Risk of Incident Type 2Diabetes: A Meta-analysisDiabetes Care 2018;41:372–382 | https://doi.org/10.2337/dc17-1902

OBJECTIVE

Several studies have explored the impact of nonalcoholic fatty liver disease (NAFLD)on risk of incident type 2 diabetes. However, the extent to which NAFLDmay conferrisk of incident diabetes remains uncertain. We performed a meta-analysis of rele-vant studies to quantify themagnitude of the association betweenNAFLD and risk ofincident diabetes.

RESEARCH DESIGN AND METHODS

We collected data using PubMed, Scopus, and Web of Science from January 2000 toJuly 2017. We included only large (n ‡500) observational studies with a follow-upduration of at least 1 year in which NAFLD was diagnosed on imaging methods.Eligible studies were selected according to predefined keywords and clinical out-comes. Data fromselected studieswere extracted, andmeta-analysiswas performedusing random-effects modeling.

RESULTS

A total of 19 observational studies with 296,439 individuals (30.1% with NAFLD) andnearly 16,000 cases of incidentdiabetes over amedianof 5 yearswere included in thefinal analysis. Patients with NAFLD had a greater risk of incident diabetes than thosewithout NAFLD (random-effects hazard ratio [HR] 2.22, 95% CI 1.84–2.60; I2 = 79.2%).Patients with more “severe” NAFLD were also more likely to develop incident di-abetes; this risk increased across the ultrasonographic scores of steatosis (n = 3 stud-ies), but it appeared to be even greater among NAFLD patients with advanced highNAFLD fibrosis score (n = 1 study; random-effects HR 4.74, 95% CI 3.54–5.94). Sen-sitivity analyses did not alter these findings. Funnel plot and Egger test did not revealsignificant publication bias. Study limitations included high heterogeneity, varyingdegrees of confounder adjustment across individual studies, and lack of studies usingliver biopsy.

CONCLUSIONS

NAFLD is significantly associated with a twofold increased risk of incident diabetes.However, the observational design of the eligible studies does not allow for provingcausality.

Nonalcoholic fatty liver disease (NAFLD) has become themost common liver disease inhigh-income countries (affecting up to one-third of adults in Europe and the U.S.), andits prevalence is expected to rise further in the near future (1,2).NAFLD has been traditionally considered to be the simple “hepaticmanifestation” of

themetabolic syndrome (2,3).Moreover, it is alsowell known that patientswith type 2

1Section of Endocrinology, Diabetes and Metab-olism, Department of Medicine, University andAzienda Ospedaliera Universitaria Integrata ofVerona, Verona, Italy2Nutrition and Metabolism, Faculty of Medicine,University of Southampton, Southampton, U.K.3Southampton National Institute for Health Re-search Biomedical Research Centre, UniversityHospital Southampton, Southampton, U.K.

Corresponding author: Giovanni Targher,[email protected].

Received 12 September 2017 and accepted 6November 2017.

This article contains Supplementary Data onlineat http://care.diabetesjournals.org/lookup/suppl/doi:10.2337/dc17-1902/-/DC1.

© 2018 by the American Diabetes Association.Readers may use this article as long as the workis properly cited, the use is educational and notfor profit, and the work is not altered. More infor-mation is available at http://www.diabetesjournals.org/content/license.

Alessandro Mantovani,1

Christopher D. Byrne,2,3 Enzo Bonora,1

and Giovanni Targher1

372 Diabetes Care Volume 41, February 2018

META-ANALYSIS

https://doi.org/10.2337/dc17-1902http://crossmark.crossref.org/dialog/?doi=10.2337/dc17-1902&domain=pdf&date_stamp=2018-01-05mailto:[email protected]://care.diabetesjournals.org/lookup/suppl/doi:10.2337/dc17-1902/-/DC1http://care.diabetesjournals.org/lookup/suppl/doi:10.2337/dc17-1902/-/DC1http://www.diabetesjournals.org/content/licensehttp://www.diabetesjournals.org/content/license

diabetes have a high prevalence of NAFLD(up to 70–75%) and that these patientsare also at higher risk of developing non-alcoholic steatohepatitis (NASH) andhave a twofold to fourfold higher risk ofdeveloping serious liver-related complica-tions (cirrhosis, liver failure, and hepato-cellular carcinoma) (4–6).However, it is now becoming increas-

ingly clear that the link between NAFLDand type 2 diabetes is more complex thanpreviously believed (7–9). NAFLD andtype 2 diabetes sharemultiple cardiomet-abolic risk factors and pathophysiological(proinflammatory and profibrotic) path-ways. In addition, increasing epidemio-logical evidence suggests that there is abidirectional relationship betweenNAFLDand type 2 diabetes and that NAFLD mayprecede and/or promote thedevelopmentof type 2 diabetes (3,4,7–9).To our knowledge, there are only two

previously published meta-analyses thathave shown thatNAFLD is associatedwithan increased risk of incident diabetes(10,11). However, both of these meta-analyses (published in 2011 and 2016, re-spectively) also included a large numberof observational studies in which the di-agnosis of NAFLD was based on abnormalserum liver enzyme levels, which arethought to be only surrogate markers ofNAFLD (12). Currently, there is intense de-bate about the prognostic role of NAFLDper se on the long-term risk of incidentdiabetes. As will be discussed in detailbelow, over the past year, numerous largeobservational studies have been publishedinwhich the diagnosis of NAFLDwas basedon ultrasonography or computed tomo-graphy, which are the most widely usednoninvasive methodologies to diagnoseNAFLD in clinical practice (1–3,12).We herein report the results of a com-

prehensive systematic review and meta-analysis of observational cohort studiesthat has investigated the association be-tween NAFLD (as detected by ultrasonog-raphy or other imaging methods) and therisk of incident diabetes. Our aim was togauge precisely the nature and magnitudeof the association betweenNAFLD and riskof incident diabetes. We have also investi-gated whether the severity of NAFLD (instudies using either ultrasonographic scor-ing systems or noninvasive fibrosismarkers)is associated with an even greater risk ofincident diabetes. Clarification of themagni-tude of risk of incident diabetes associatedwith the different stages of liver disease

within the spectrum of NAFLD may haverelevant clinical implications for the diag-nosis, prevention, and treatment of type 2diabetes.

RESEARCH DESIGN AND METHODS

Registration of Review ProtocolThe protocol for this systematic reviewwas registered in advance with PROSPERO(international prospective register of sys-tematic reviews, no. CRD42017072305).

Data Sources and SearchesStudies were included if they were obser-vational cohort studies that reported theincidence rates of type 2 diabetes in adultindividuals (.18 years old)withNAFLD ascompared with those without NAFLD.Study participants were of either sex withno restrictions in terms of ethnicity and co-morbidities.We includedonly large (n$500)observational studies with a follow-up ofat least 1 year in which the diagnosis ofNAFLD was based on either imaging orhistology in the absence of competingcauses of hepatic steatosis. Based ondata from the eligible studies, “severe”NAFLD was defined either by presenceof increasing ultrasonographic steatosisscores or by high NAFLD fibrosis score(NFS), which is a reliable noninvasivemarker of advanced NAFLD fibrosis (12).In these eligible studies, the diagnosis ofincident diabetes was based on a self-reported history of disease or use of hypo-glycemic drugs, and in most cases, it wasalso based on a fasting plasma glucose level$7.0 mmol/L or an HbA1c level $6.5%($48 mmol/mol).

Exclusion criteria of the meta-analysiswere as follows: 1) reviews, editorials, ab-stracts, case reports, practice guidelines,and cross-sectional studies; 2) studies thatused only serum liver enzyme levels, fattyliver index, or other surrogate markers todiagnose NAFLD; 3) studies with a samplesize of,500 individuals or with a follow-up duration ,1 year; 4) studies con-ducted in the pediatric population (,18years old); and 5) studies that did not re-port any hazard ratio (HR) and 95% CI forthe outcome of interest (incident diabetes).

Included and excluded studies werecollected following the Preferred Report-ing Items for Systematic Reviews andMeta-Analyses (PRISMA) flow diagram.Additionally, because included studieswere observational in design, we followedthe Meta-analysis Of Observational Stud-ies in Epidemiology (MOOSE) guidelines

for the meta-analysis of observationalstudies.

Data Extraction and QualityAssessmentRelevant studies were identified by sys-tematically searching PubMed, Scopus,and Web of Science from 1 January 1990to 15 July 2017 (date last searched) us-ing the free text terms “fatty liver” (OR“NAFLD” OR “nonalcoholic fatty liver dis-ease” OR “nonalcoholic steatohepatitis”)AND “diabetes risk” OR “diabetes inci-dence” OR “incident diabetes.” No lan-guage restriction was applied. Referencelists of relevant articles and previous re-view articles were hand searched forother relevant studies. Two investigators(A.M. and G.T.) independently examinedall titles and abstracts and obtained fulltexts of potentially relevant articles.Work-ing independently and in duplicate, weread the articles and determined whetherthey met inclusion criteria. Discrepancieswere resolved by consensus, referringback to the original article, in consultationwith a third author. For all studies, we ex-tracted information on study design, studysize, source of data, population character-istics, duration of follow-up, outcome ofinterest, matching, and confounding fac-tors. Additionally, in the case of multiplepublications, we included the most up-to-date or comprehensive information.

Two authors (A.M. and G.T.) assessedthe risk of bias independently. Since allthe included studies were nonrandomizedand had a cohort design, the Newcastle-Ottawa Scale (NOS) was used to judgestudy quality, as recommended by theCochrane Collaboration (13). This scaleuses a star system (with a maximum ofnine stars) to evaluate a study in three do-mains: selectionofparticipants, comparabil-ity of study groups, and the ascertainmentof outcomes of interest. We judged studiesthat received a score of nine stars to be atlow risk of bias, studies that scored seven oreight stars to be at medium risk, and thosethat scored six or less to be at high risk.

Data Synthesis and AnalysisThe outcomemeasure of this meta-analysiswas the occurrence of incident diabetesamong individuals with NAFLD comparedwith incidence of diabetes among thosewithoutNAFLD.Whenpossible,wepooledadjusted HRs (or odds ratios [ORs]), withtheir 95% CIs. In the case of studies report-ingHRswith varying degrees of adjustment,

care.diabetesjournals.org Mantovani and Associates 373

http://care.diabetesjournals.org

we always used the fully adjusted HR esti-mate. Visual inspection of the forest plotswas used to investigate the possibility ofstatistical heterogeneity. Statistical hetero-geneity was assessed by the I2 statistic,which provides an estimate of the per-centage of variability across studies thatis due to heterogeneity rather than chancealone. According toHiggins andThompson(14), a rough guide to interpretation is asfollows: I2 values of;25% represent lowheterogeneity; ;50% represent mediumheterogeneity; and;75% represent highheterogeneity.The results of the eligible studies were

pooled, and an overall estimate of effectsize (ES) was calculated using a random-effects model, as this methodology takesinto account any differences betweenstudies even if there is no statistically sig-nificant heterogeneity (15). Publicationbias was evaluated using the funnel plotand Egger regression test (16).Given the expected heterogeneity of

the eligible studies, sensitivity analyseswere also carried out to relate the primaryoutcome (i.e., incident diabetes) with theindividual study design characteristics. Inparticular, based on data from the eligiblestudies, the prognostic impact of NAFLDon risk of incident diabetes was assessedby stratifying the studies according to theduration of follow-up, the study country,the study design, the “severity” of NAFLD(based on ultrasonographic scoring sys-tems or the NFS), or whether the studieshad eight or nine stars on the NOS (i.e.,the “high-quality” studies) and whetherthe studies had full adjustment for covar-iates (i.e., those studies adjusting at leastfor age, sex, BMI [orwaist circumference],family history of diabetes, fasting glucoselevels [or impaired fasting glycemia], lip-ids, hypertension [or blood pressure val-ues], smoking, and physical activity).Additionally, we tested for possibly exces-sive influence of individual studies using ameta-analysis influence test that elimi-nated each of the included studies at atime. All statistical tests were two sidedand used a significance level of P, 0.05.We used STATA 14.2 (StataCorp, CollegeStation, TX) for all statistical analyses.

RESULTS

Characteristics of Included StudiesBased on the titles and abstracts of 2,072citations, we identified 36 potentiallyrelevant studies. Of these, we excluded

17 studies for the reasons specified inthe flow diagram (Supplementary Fig.1). Thus, 19 unique observational cohortstudies were eligible for inclusion in themeta-analysis and were assessed forquality.

As summarized in Table 1, all the eligi-ble studies had an observational retro-spective or prospective design (17–35).The eligible studies recruited participantsfrom approximately general populationsin which NAFLD was diagnosed by imag-ing methods (mainly ultrasonography)and incident diabetes was diagnosed bybiochemistry (fasting glucose levels orHbA1c), clinical history, or drug treatment.No studies with biopsy-proven NAFLDwere available for the analysis.

Overall, in the 19 observational cohortstudies included in the meta-analysis,there were 296,439 adult individuals(30.1% with NAFLD; n = 89,123) with15,751 cases of incident diabetes over amedian follow-up period of 5 years (inter-quartile range 4.6–9.1). Most of thesestudies were carried out in Asia (China,Taiwan, South Korea, Sri Lanka, and Ja-pan); two community-based studieswere carried out in the U.S.Most of thesestudies included middle-aged subjectspredominantly of male sex. Of the 19 in-cluded studies, 10 studies received eightstars at the NOS, 7 studies received six orseven stars, and 2 studies received fewerthan six stars, indicating an overall me-dium risk of bias (Supplementary Table 1).

NAFLD and Risk of Incident DiabetesThe distribution of studies by estimate ofthe association betweenNAFLD and risk ofincident diabetes is plotted in Fig. 1. Six-teen studies provided data suitable forthe pooled primary analysis (n = 214,805with 10,356 cases of incident diabetes).We excluded three studies from this pri-mary analysis because the authors did notprovide any HR for incident diabetesamong individuals with NAFLD pooled to-gether (21,24,32); these three studieswere used in a secondary analysis for ex-amining the association between NAFLDseverity and diabetes risk (see below).

NAFLD was significantly associated withan increased risk of incident diabetes(random-effects HR 2.22, 95% CI 1.84–2.60; I2 = 79.2%). Notably, since we al-ways used the fully adjustedHR estimatesfor each eligible study (as specified in Ta-ble 1), this random-effects HR was inde-pendent of a relatively large number of

common risk factors and potential con-founders. As also shown in Fig. 1, whenthe comparison was stratified by thestudy country, the association of NAFLDwith risk of incident diabetes was signifi-cant in all study countries, but it appearedtobe stronger in Japan, China, and Taiwanthan in the U.S. and in other Asian coun-tries (South Korea and Sri Lanka).

As shown in Fig. 2, when the compari-son was stratified by the length of follow-up period, the association of NAFLD withthe risk of incident diabetes appeared tobe stronger in those studies with morethan 5 years of follow-up (n = 16 studies;random-effects HR 2.60, 95% CI 1.92–3.29; I2 = 74.6%).

As shown in Supplementary Fig. 2,when the comparison was stratified bythe studydesign, the associationbetweenNAFLD and the risk of incident diabeteswas consistent in prospective studies (n =3 studies; random-effects HR 2.25, 95% CI1.93–2.58; I2 = 0%) and in retrospectivestudies (n = 13 studies; random-effectsHR 2.26, 95% CI 1.80–2.72; I2 = 81.0%).

Limiting the analysis to “high-quality”studies and to studies with adjustmentfor multiple covariates provided overallestimates consistent with the pooledprimary analysis (n = 10 studies; random-effects HR 1.85, 95% CI 1.47–2.22; I2 =68.3%). Finally, eliminating each of theincluded studies from the analysis hadno effect on the overall risk of incidentdiabetes (data not shown).

As shown in Supplementary Fig. 3, theEgger regression test did not show statis-tically significant asymmetry of the funnelplot (P = 0.31), thus suggesting that pub-lication bias was unlikely.

Severe NAFLD and Risk of IncidentDiabetesFour cohort studies reported data onpatients with “severe”NAFLD, defined ei-ther by ultrasonographic severity ofsteatosis or by high NFS. The distributionof studies by estimate of the associationbetween severe NAFLD and risk of inci-dent diabetes is plotted in Fig. 3. Com-pared with the non-NAFLD group, thepresence of more “severe” NAFLD wassignificantly associated with an increasedrisk of incident diabetes (n = 4 studies;random-effects HR 2.63, 95% CI 1.57–3.70; I2 = 82.4%). This risk increasedacross the ultrasonographic scores of he-patic steatosis (n = 3 studies; random-effects HR 2.15, 95% CI 1.72–2.58;

374 NAFLD and Risk of Incident Diabetes Diabetes Care Volume 41, February 2018

http://care.diabetesjournals.org/lookup/suppl/doi:10.2337/dc17-1902/-/DC1http://care.diabetesjournals.org/lookup/suppl/doi:10.2337/dc17-1902/-/DC1http://care.diabetesjournals.org/lookup/suppl/doi:10.2337/dc17-1902/-/DC1http://care.diabetesjournals.org/lookup/suppl/doi:10.2337/dc17-1902/-/DC1http://care.diabetesjournals.org/lookup/suppl/doi:10.2337/dc17-1902/-/DC1

Tab

le1—

Principal

observational

(retrosp

ective

orprosp

ective

)co

hort

studiesoftheasso

ciationbetwee

nNAFL

D(detec

tedbyim

agingtech

niques)an

dtherisk

ofinciden

ttype2

diabetes

(ordered

bypublic

ationye

ar)

Autho

rs,

year

(ref.)

Studydesign;sam

plesize

andpopulation;

follow-up;andNAFLDdiagn

ostictool

Diagnosisofinciden

tdiabetes

Number

inciden

tcases

ofdiabetes;%in

non-

NAFLDvs.NAFLDcases

(when

available)

Adjustmen

tsconsidered

Mainfindings

NOS

Okamoto

etal.,

2003

(17)

Retrospective

cohortstudy;n=

840(14.3%

withNAFLD)Japanese

subjectswitho

utdiabetes;10years;

liver

ultrason

ograph

y

Fastingglucose

.6.1mmol/L

orHbA

1c$6.5%

n=82

incident

cases;

7.6%

vs.22.5%

Age,sex,BMI,family

history

of

diabetes,fastingglucose,HbA1c,

alcoho

lintake,frequencyof

check-

ups,changesof

BMId

uringfollow-

up

NAFLDwas

associated

with

incidentdiabetesin

univariate

analysis(OR2.62,95%

CI1.6–4.3).

Thisassociationdisapp

earedafteradjusting

forpotentialconfounding

factors(aOR1.83,

95%CI0.9–3.5)

6

Shibataet

al.,

2007

(18)

Retrospective

cohortstudywithanested

case-controlanalysis;n

=3,189(33.6%

withNAFLD)maleJapanese

workers

withno

rmalglucosetolerancewitho

utknow

nchronicliver

diseases;4

years;

liver

ultrason

ograph

y

Fastingglucose

$7.0mmol/L

or2-hglucose

$11.1mmol/L

on75-g

OGTT

n=109incident

cases;

1.8%

vs.8.1%

Age

andBMI(inthewhole-cohort

analysis),age,

BMI,sm

oking

history,b

loodpressure,p

hysical

activity,follow-upduration,

metabolic

syndrome(inthenested

case-controlanalysis)

NAFLDwas

indep

enden

tlyassociated

with

inciden

tdiabetes

both

inthewholecohort

(aHR5.50,95

%CI3.6–8.5)and

inthenested

case-con

trolanalysis(aHR4.60,95%

CI3.0–

6.9)

4

Kim

etal.,

2008

(19)

Retrospective

cohortstudy;n=5,372

(33.3%

withNAFLD)SouthKo

rean

subjectswitho

utdiabetes

witho

utknow

nchronicliver

diseases;5

years;

liver

ultrason

ograph

y

Fastingglucose

$7.0mmol/L,

clinicalhistory,or

drug

treatm

ent

n=234incident

cases;

2.3%

vs.8.5%

Age,sex,BMI,family

history

of

diabetes,smoking,fastingglucose,

HDLcholesterol,triglycerides,

serum

ALT

NAFLDwas

indep

enden

tlyassociated

with

inciden

tdiabetes

(aHR1.51,95

%CI1

.04–

2.2).M

oderate/severe

NAFLDhadhigher

HRsvs.m

ildNAFLD.Exclusion

ofdrinkers

didno

tattenu

atethisassociation

8

Bae

etal.,

2011

(20)

Retrospective

cohortstudy;n=7,849

(29.2%

withNAFLD)SouthKo

rean

subjectswitho

utdiabetes;5

years;liver

ultrason

ograph

y

Fastingglucose

$7.0mmol/L,

HbA

1c$6.5%

,clinical

history,or

drug

treatm

ent

n=435incident

cases;

3.7%

vs.9.9%

Age,sex,BMI,triglycerides,HDL

cholesterol,systolic

blood

pressure,smoking,physical

activity,alcoholintake,IFGstatus

NAFLDwas

indep

enden

tlyassociated

with

inciden

tdiabetes

(aHR1.33,

95%CI1

.1–

1.7).Thisassociationwas

muchstronger

inpre-existing

IFG

8

Sung

etal.,

2012

(22)

Retrospective

cohortstudy;n=12,853

(27.6%

withNAFLD)SouthKo

rean

subjectswitho

utdiabetes;5

years;liver

ultrason

ograph

y

Fastingglucose

$7.0mmol/L,

clinicalhistory,or

drug

treatm

ent

n=223incident

cases;

0.8%

vs.4.3%

Age,sex,BMI,ed

ucationalstatus,

smoking,physicalactivity,alcohol

intake,HOMA-IR,serum

triglycerides,serum

ALT

TheclusteringofincreasedHOMA-IR,

overw

eigh

t/obesity,andNAFLDmarkedly

increasestheoddsofdeveloping

diabetes,

with

effectsindependentofeachotherandof

confoundingfactors.NAFLDwas

associated

with

incident

diabetes

(aOR2.42,95%

CI1.7–3.4)

7

Park

etal.,

2013

(21)

Prospective

cohortstudy(healthcheck-

up);n=25,232

(35%

withNAFLD)Sou

thKo

rean

men

witho

utdiabetes

witho

utknow

nchronicliver

diseases;5

years;

liver

ultrason

ograph

y

Fastingglucose

$7.0mmol/L,

HbA

1c$6.5%

,clinical

history

n=2,108incident

cases;

7%inno

-steatosisvs.

9.8%

inmild

steatosis

vs.17.8%

inmod

erate-severe

steatosis

Age,waistcircumference,HDL

cholesterol,triglycerides,systolic

bloodpressure,C-reactiveprotein,

HOMA-IR,serumcreatinine,family

history

ofdiabetes,p

hysical

activity,m

etabolic

syndrome

NAFLDwas

indep

enden

tlyassociated

with

inciden

tdiabetes;theHRswereincreased

inmild

steatosis(1.09,95

%CI0.8–1.5)and

inmod

erate/severe

steatosis(1.73,95%

CI1.0–3.0)vs.n

o-steatosis

8

Kasturiratne

etal.,

2013

(23)

Retrospective

cohortstudy;n=2,276

(40.7%

withNAFLD)SriLankan

individu

alswitho

utdiabetes

witho

utknow

nchronicliver

diseases;3

years;

liver

ultrason

ograph

y

Fastingglucose

$7.0mmol/L,

clinicalhistory,or

drug

treatm

ent

n=242incident

cases;

10.5%vs.19.7%

Age,sex,family

history

ofdiabetes,

BMI,waistcircumference,

hypertension,serum

ALT,

dyslipidem

ia,IFG

status

NAFLDwas

indep

enden

tlyassociated

with

inciden

tdiabetes

(aHR1.64,

95%CI1

.2–

2.2).N

AFLDwas

theon

lyindepend

ent

predictorof

incident

diabetes

amon

gthose

withIFGat

baseline

5

Continuedon

p.376



Tab

le1—

Continued

Autho

rs,

year

(ref.)

Studydesign;sam

plesize

andpopulation;


ostictool

Diagnosisofinciden

tdiabetes

Number

inciden

tcases

ofdiabetes;%in

non-

NAFLDvs.NAFLDcases

(when

available)

Adjustmen

tsconsidered

Mainfindings

NOS

Changet

al.,

2013

(24)

Retrospective


(30.4%

withNAFLD)SouthKo

rean

subjectswitho

utdiabetes

witho

utknow

nchronicliver

diseases;5

years;

liver

ultrason

ograph

y

Fastingglucose

$7.0mmol/L,

HbA

1c$6.5%

,ordrug

treatm

ent

n=2,025incident

cases;

3.5%

inno

-NAFLDvs.

7.4%

inNAFLDwith

lowNFS

vs.15.3%

inNAFLDwith

interm

ediate

orhigh

NFS

Age,sex,sm

oking,alcoholintake,

physicalactivity,fam

ilyhistory

ofdiabetes,totalcholesterol,

triglycerides,HDLcholesterol,

HOMA-IR,C

-reactiveprotein

TheaH

Rsforinciden

tdiabetes

inNAFLDwith

lowNFS

andNAFLDwithinterm

ediate

or

highNFS

vs.noNAFLDwere2.01,

95%CI

1.8–2.2,and

4.74,95%

CI3.7–6.1,

respectively.Thisassociationremained

significant

insubjectswithfastingglucose

levels,100mg/dL

orwithHbA

1c,5.8%

8

Choiet

al.,

2013

(25)

Retrospective

cohortstudy;n=7,849

(29%

withNAFLD)SouthKo

rean

witho

utdiabetes

subjectswitho

utknow

nchronicliver

diseases;4

years;

liver

ultrason

ograph

y

Fastingglucose

$7.0mmol/L,

HbA

1c$

6.5%

,ordrug

treatm

ent

n=435incident

cases;

3.5%

incontrolsvs.

4.6%

intheincreased

ALT

vs.7.3%inthe

steatosisvs.11.8%

inthecombined

abno

rmalitygrou

p

Age,sex,BMI,systolic

bloodpressure,

triglycerides,HDLcholesterol,IFG

status,physicalactivity,smoking,

alcoholintake

TheHRsand95

%CIo

finciden

tdiabetes

progressivelyincreasedacross

the

elevated

ALT,thehep

aticsteatosis,andthe

combined

abnorm

alitygroups.Subjectsin

thecombined

abnorm

alitygrouphad

the

highestrisk

ofinciden

tdiabetes

(aHR1.64

,95

%CI1.3–2.1)

8

Yamazaki

etal.,

2015

(26)

Retrospective

cohortstudy;n=3,074

(23.7%

withNAFLD)Japanese

subjects

witho

utdiabetes

witho

utknow

nchronicliver

diseases;11.3years;liver

ultrason

ograph

y

Fastingglucose

$7.0mmol/L,

HbA

1c$6.5%

,clinical

history,or

drug

treatm

ent

n=189incident

cases;

3.1%

vs.16.1%

Age,sex,family

history

ofdiabetes,

BMI,IFGstatus,dyslipidem

ia,

hypertension,p

hysicalactivity

NAFLDwas

indep

enden

tlyassociated

with

inciden

tdiabetes

(aOR2.37,

95%CI1.6–

3.5).N

AFLDimprovem

entwas

associated

witharedu

ctionof

incident

diabetes

(aOR0.27,95%

CI0.1–0.6)

8

Minget

al.,

2015

(27)

Retrospective

cohortstudy;n=

508(19.1%

withNAFLD)Chinese

subjectswitho

utdiabetes

witho

utknow

nchronicliver

diseases;5

years;

liver

ultrason

ograph

y

Fastingglucose

$7.0mmol/L,

2-hglucose$11.1mmol/L

on75-g

OGTT,ordrug

treatm

ent

n=20

incident

cases;

2.4%

vs.10.3%

Age,sex,ed

ucationallevel,sm

oking,

alcoholintake,p

hysicalactivity,

family

history

ofdiabetes,B

MI,

bloodpressure,fastingglucose,2-h

glucose,triglycerides,HDL

cholesterol

NAFLDwas

indep

enden

tlyassociated

with

inciden

tdiabetes

(aHR4.46,

95%CI1

.9–

10.7)bu

tno

twithincident

prediabetes

(aHR1.64,95%

CI0.97–2.8)

6

Liet

al.,

2015

(28)

Retrospective

cohortstudy;n=4,736

(29.8%

withNAFLD)Chinese

subjects

witho

utdiabetes

witho

utknow

nchronicliver

diseases;4

years;liver

ultrason

ograph

y

Fastingglucose

$7.0mmol/L,

clinicalhistory,or

drug

treatm

ent

n=380incident

cases;

4.1%

vs.17.4%

Age,sex,bloodpressure,lipids,serum

ALT,u

ricacid,creatinine

NAFLDwas

indep

enden

tlyassociated

with

inciden

tdiabetes

(aHR3.37,

95%CI

2.4–4.3)

7

Shah

etal.,

2015

(29)

Prospective

cohortstudy;n=3,153

(24.9%

withNAFLD)U.S.ind

ividuals

witho

utdiabetes

from

theMulti-Ethnic

Stud

yofAtherosclerosiswitho

utknow

nchronicliver


compu

tedtomograph

y

Fastingglucose

$7.0mmol/L,

clinicalhistory,or

drug

treatm

ent

n=216incident

cases

Age,sex,race,fam

ilyhistory

of

diabetes,B

MI,waist

circumference,systolic

blood

pressure,triglycerides,H

DL

cholesterol,fastingglucose,

C-reactiveprotein,e

xercise,

statin

use

NAFLD(defi

ned

asfirstquartile

ofhep

atic

attenuationoncomputedtomograp

hy)

was

indep

enden

tlyassociated

with

inciden

tdiabetes

(aHR2.06,

95%CI

1.5–2.8;P

,0.001)

8

Continuedon

p.377


Tab

le1—

Continued

Autho

rs,

year

(ref.)

Studydesign;sam

plesize

andpopulation;


ostictool

Diagnosisofinciden

tdiabetes

Number

inciden

tcases

ofdiabetes;%in

non-

NAFLDvs.NAFLDcases

(when

available)

Adjustmen

tsconsidered

Mainfindings

NOS

Fukuda

etal.,

2016

(30)

Retrospective

cohortstudy;n=4,629

(38.4%

withNAFLD)Japanese

subjects

witho

utdiabetes

witho

utknow

nchronicliver


ultrason

ograph

y

Fastingglucose

$7.0mmol/L,

HbA

1c$6.5%

,ordrug

treatm

ent

n=351incident

cases;

3.2%

inno

noverw

eight,

no-NAFLDvs.14.4%

inno

noverw

eight,

NAFLDvs.8.0%in

overweight,no

-NAFLD

vs.26.4%

inoverweight,NAFLD

Age,sex,family

history

ofdiabetes,

alcoholintake,smoking,regular

exercise,H

bA1c

aHRsforinciden

tdiabetes

vs.nonoverw

eigh

twithoutNAFLDgroupwere3.59

,95

%CI

2.1–5.8,intheno

noverw

eightwithNAFLD

grou

p,1.99,95%

CI1.5–2.7,inthe

overweightw

itho

utNAFLDgrou

p,and6.77,

95%CI

5.2–8.9,intheoverweightwith

NAFLDgrou

p,respectively

7

Chen

etal.,

2016

(31)

Prospective

cohortstudy;n=6,542(3.2%

withNAFLD)Chinese

subjectswitho

utdiabetes

witho

utknow

nchronicliver

diseases;6

years;liver

ultrason

ograph

y

Fastingglucose

$7.0mmol/L,

HbA

1c$6.5%

,ordrug

treatm

ent

n=368incident

cases

Age,BMI,triglycerides,fasting

glucose,IFG

status

NAFLDwas

indep

enden

tlyassociated

with

inciden

tdiabetes

(aHR2.17,

95%CI

1.6–3.0)

7

Liet

al.,

2017

(32)

Prospective


(31.8%

withNAFLD)Chinese

subjects

witho

utdiabetes

witho

utknow

nchronicliver


ultrason

ograph

y

Fastingglucose

$7.0mmol/L,

clinicalhistory,or

drug

treatm

ent

n=1,262incident

cases;

4.6%

inno

-NAFLDvs.

10.6%inmild

NAFLD

vs.18.1%

inmod

erate-severe

NAFLD

Age,sex,BMI,waistcircumference,

alcoholintake,smoking,exercise,

family

history

ofdiabetes,fasting

glucose,triglycerides,total

cholesterol

aHRsforinciden

tdiabetes

vs.those

without

NAFLDgroupwere1.88,

95%CI1.6–2.2,

inthemild

NAFLDgrou

pand2.34,95%

CI1.9–3.0,inthemod

erate-severe

NAFLD

grou

p,respectively

8

Maet

al.,

2017

(33)

Retrospective

cohortstudy;n=1,051

(17.8%

withNAFLD)U.S.ind

ividuals

witho

utdiabetes

witho

utknow

nchronicliver


compu

tedtomograph

y

Fastingglucose

$7.0mmol/L,

clinicalhistory,or

drug

treatm

ent

n=64

incident

cases

Age,sex,sm

oking,exercise,alcohol

intake,fastingglucose,systolic

bloodpressure,B

MI,visceral

adipose

tissue,andchangesinBMI,

visceraladipose

tissue,andliverfat

duringfollow-up

NAFLDwas

indep

enden

tlyassociated

with

inciden

tdiabetes

(aOR2.66,

95%CI1.2–

5.7)

8

Chen

etal.,

2017

(34)

Prospective


(32%

withNAFLD)Taiwanesesubjects

witho

utdiabetes

witho

utknow

nchronicliver

diseases;18years;liver

ultrason

ograph

y

Fastingglucose

$7.0mmol/L,

clinicalhistory,or

drug

treatm

ent

n=6,555incident

cases

Age,sex,BMI,hypertension,fam

ilyhistory

ofdiabetes,smoking,

alcoholintake,e

xercise,

triglycerides,H

DLcholesterol,total

cholesterol,serum

AST,ALT,G

GT,

andALP

levels

NAFLDwas

indep

enden

tlyassociated

with

inciden

tdiabetes

(aHR2.38,

95%CI

1.6–2.5,for

thewho

lesample;aH

R2.08,

95%CI

1.9–2.2,formen

andaH

R2.65,

95%CI

1.4–2.9,forwom

en)

8

Liuet

al.,

2017

(35)

Retrospective


(18.8%

withNAFLD)Chinese

elderly

men

witho

utdiabetes

witho

utknow

nchronicliver

diseases;5

years;liver

ultrason

ograph

y

Fastingglucose

$7.0mmol/L,

2-hglucose$11.1mmol/L

on75-g

OGTT,clinical

history,or

drug

treatm

ent

n=453incident

cases;

2.1%

vs.3.7%

Age,BMI,sm

oking,maritalstatus,

alcoholintake,h

ypertension,

dyslipidem

ia

NAFLDwas

indep

enden

tlyassociated

with

inciden

tdiabetes

(aHR1.67,

95%CI

1.4–2.1)

7

ALP,alkalineph

osph

atase;ALT,alanine

aminotransferase;A

ST,aspartate

aminotransferase;aHR,adjustedHR;aOR,adjustedOR;G

GT,g-glutamyltransferase;HOMA-IR,H

OMAof

insulin

resistance;IFG

,impaired

fastingglycem

ia;O

GTT,oralglucose

tolerancetest.



I2 = 0%), but it appeared to be even greateramong NAFLD patients with high NFS (n =1 study; random-effects HR 4.74, 95% CI3.54–5.94). However, only a single studyassessed NAFLD severity by using NFS, andno studies involving non-Asian individualswere available for this analysis, thus limit-ing the generalizability of the finding.

CONCLUSIONS

Ourmeta-analysis provides evidence for asignificant association between imaging-diagnosed NAFLD and the long-term riskof incident diabetes. Indeed, this meta-analysis involves a total of 19 unique ob-servational studies with aggregate dataon 296,439 adult individuals (30.1% withNAFLD) and nearly 16,000 cases of inci-dent diabetes followed up over a medianperiod of 5 years.We found that the presence of imaging-

diagnosed NAFLD conferred a HR of 2.2for incident diabetes, a risk that appearedto increase furtherwith greater “severity”of NAFLD (assessed in four observationalstudies using either the ultrasonographic

severity of steatosis or the NFS) and re-mained statistically significant in thosestudies where analysis was fully adjustedfor potentially confounding factors (n = 10studies; random-effects HR 1.85, 95% CI1.47–2.22; I2 = 68.3%). In addition, whenthe analysis was stratified either by follow-up duration or by study country, the asso-ciation between NAFLD and diabetes riskappeared to be stronger in studies with afollow-up duration longer than 5 years andin studies performed in Japanese and Chi-nese populations compared with thoseconducted in the U.S. and in other Asiancountries.

Unfortunately, most of the publishedstudies that used liver biopsies to diag-nose NAFLD (i.e., the “gold standard”method for diagnosing and stagingNAFLD) did not have a control group andcannot, therefore, be included in thismeta-analysis. To date, however, verylittle is known about how long-termNAFLD or its histologic features may af-fect risk of incident diabetes. A retro-spective cohort study of 396 patients

with biopsy-confirmed NAFLD, who didnot have diabetes at baseline, reportedthat a significantly higher proportion ofpatients with fibrosis stages 3–4 devel-oped incident diabetes than those withfibrosis stages 0–2 (51% vs. 31%) over amean follow-up of 18.4 years (36). Inter-estingly, for patients with fibrosis stages0–2, fat score was also independently as-sociated with incident diabetes (36). Theissue of whether the increased diabetesrisk is confined to patients with more se-vere NAFLD or applies to all patients withNAFLD is particularly relevant in view ofthe disease burden that NAFLD repre-sents and might impact on the healthcare resources needed to survey andmanage these patients adequately. Theresults of our meta-analysis (based exclu-sively on studies using ultrasonographicscoring systems or the NFS) suggest thatit is more advanced NAFLD that carries agreater diabetes risk. This is also consis-tent with the conclusion of another com-prehensive meta-analysis supporting alink between NAFLD severity and risk offatal and nonfatal cardiovascular events(37). However, this question remainslargely unsolved, and further follow-upstudies in larger cohorts of both Asianand non-Asian individuals with biopsy-confirmed NAFLD (with an adequategroup of control individuals) are neededin order to definitely establish whetherNAFLD severity differentially affects riskof incident diabetes.

This is the largest and most updatedmeta-analysis aimed at investigating theprognostic role of imaging-diagnosedNAFLD on the long-term risk of incidentdiabetes. Collectively, our findings con-firm and extend (with a sample size atleast three times larger) the results oftwo previousmeta-analyses that incorpo-rated studies using both abnormal serumliver enzymes and imaging techniquesto diagnose NAFLD (10,11). In the firstmeta-analysis, Musso et al. (10) in 2011reported that ultrasound-diagnosed NAFLD(only three studies included) was associ-ated with an increased risk of incident di-abetes (random-effects OR 3.51, 95% CI2.28–5.41; I2 = 70%). In the second meta-analysis, Ballestri et al. (11) confirmedthat NAFLD (defined by ultrasonography;only nine studies included in total) sig-nificantly increased the risk of incidentdiabetesover a follow-upofnearly 5 years(random-effects OR 1.86, 95% CI 1.76–1.95; I2 = 86%).

Figure 1—Forest plot and pooled estimates of the effect of NAFLD on the risk of incident diabetes in16 eligible studies, stratified by study country. ES, effect size.


Notably, our meta-analysis is the firstto show that the association betweenNAFLD and the risk of incident diabetesis stronger in some ethnic groups (espe-cially in Japanese individuals) and in stud-ies with longer follow-up duration andthat the more “severe” forms of NAFLDseem to be associated with an evengreater risk of developing diabetes. Thislatter finding is also indirectly supportedby the results of the studyofMaet al. (33)demonstrating that among the Framing-ham Heart Study participants, baselinehepatic fat content (per SD increase)was independently associated with in-creased odds of incident diabetes over;6 years of follow-up.There is now convincing evidence of

biological plausibility that NAFLD mayincrease risk of incident type 2 diabetes(3,4,7,8). Indeed, NAFLD, especially NASHwith varying levels of hepatic fibrosis,exacerbates hepatic insulin resistanceand causes the release of multiple proin-flammatory mediators and prodiabeto-genic hepatokines (e.g., fetuin-A, fetuin-B,fibroblast growth factor 21, retinol bindingprotein 4, and selenoprotein P) that maypromote the development of diabetes(38–40). Among the hepatokines, dataabout fetuin-Aaction inmiceand inhumans

in particular supports a causative relation-ship of NAFLD with incident diabetes(38,41,42). However, whether improve-ment or resolution of NAFLD could de-crease risk of incident diabetes remainsuncertain. Some evidence suggests thatthe risk of diabetes appears to diminishover time following the resolution orimprovement of NAFLD (26,43). However,as these two studies are not randomizedcontrolled trials of NAFLD management,these results should be interpretedcautiously.

Although our meta-analysis of obser-vational studies provides support forthe existence of a significant associationbetween NAFLD and increased risk ofincident diabetes, it remains to be def-initely proven that improving the livercondition in NAFLD decreases risk of de-veloping diabetes. It should also benoted that there may be a dissociationbetween NAFLD and insulin resistancein humans carrying some geneticvariants, such as the patatin-like phos-pholipase 3 gene (44). Experimental evi-dence (mainly derived from animalstudies) also indicates that specific ma-nipulation of liver fat is insufficient to af-fect insulin sensitivity/glycemia (45–47).In addition, it is known that there are

ethno-racial differences in liver fat content(as well as in the amount of lipid accumu-lated in skeletal muscle and abdominalcavity) and risk of diabetes. For example,compared with white individuals, obeseblack individuals exhibit a lower preva-lence of NAFLD but similar type 2 diabe-tes prevalence (48–50). That said, ourdata strongly emphasize that there is areal need now to include outcomes suchas incident diabetes and changes in HbA1cand insulin sensitivity in randomizedplacebo-controlled trials focused on testing theefficacy of novel therapies for liver dis-ease in NAFLD. Improved understandingof these features and precise phenotyp-ing of NAFLD could help to improve strat-ification of cardiometabolic risk. Thismight also have important implicationsfor future strategies in the preventionand treatment of type 2 diabetes andother cardiometabolic diseases in clinicalpractice (51,52).

Our meta-analysis has some importantlimitations (strictly inherent to the na-ture of the included studies) that shouldbementioned.Althoughweuseda random-effects model, the interpretation of theresults of this meta-analysis requiressome caution, given the (expected) highheterogeneity (I2 .75%) observed in theoverall primary analysis. It is plausible toassume that this high heterogeneity likelyreflects differences in the demographicand ethno-racial characteristics of studypopulations, in the length of follow-up, inthe design of the study, and in the sever-ity of NAFLD. We systematically exploredand identified all these possible sourcesof statistical heterogeneity using strati-fied analyses and sensitivity analyses(as detailed in the RESULTS section). Al-though we found significant heterogene-ity between studies when investigatingassociations in the overall analysis, it isnoteworthy that there was very low het-erogeneity between studies, as well asstronger associations between NAFLDand diabetes risk, when we restrictedthe statistical analyses to studies withonly the more “severe” forms of NAFLDon ultrasonography, studies with a pro-spective design (compared with retro-spective ones), or studies performed inthe U.S. population. However, we believethat more detailed analyses of the causesof heterogeneity will require collabora-tive pooling of individual participantdata from large prospective studies asthese become available over time.

Figure 2—Forest plot and pooled estimates of the effect of NAFLD on the risk of incident diabetes in 16eligible studies, stratifiedby lengthof follow-up (FU) (basedon themedian follow-upof theeligible studies).ES, effect size.



Another potential limitation of themeta-analysis (strictly inherent to theobservational nature of the includedstudies) is that information about thetemporal changes of some important var-iables (e.g., medication use and lifestylechanges) that may impact NAFLD andincident diabetes is often missing andthat the varying degree of confounderadjustment across the individual studieshampered a systematic assessment of theimpact of known risk factors on the out-come of interest. As shown in Table 1,some studies reported incomplete adjust-ments for established risk factors andpotential confounders (e.g., waist circum-ference or insulin resistance); as such, itwas not possible to combine models instudies that adjusted for the same set ofpotential confounding factors. Anotherlimitation of the meta-analysis was thatnone of the eligible studies used liver bi-opsy for the diagnosis of NAFLD. Con-versely, most of the eligible studies usedultrasonography, which is the recom-mended first-line imaging method fordetecting NAFLD in clinical practice andenables a reliable and accurate detectionof mild-to-moderate hepatic steatosiscompared with liver histology. Further-more, the results regarding the associ-ation between the severity of NAFLDand diabetes risk derived from veryfew studies, and only a single study as-sessed NAFLD severity by using the NFS.

Additionally, since the diagnosis of diabe-tes was not always consistent among theincluded studies, some inaccuracy in theestimated incidence of diabetes and inthe identification of diabetes subtypesmay not be excluded (although the vastmajority of diabetes cases were likely tobe type 2). Despite the fact that both fast-ing plasma glucose and HbA1c levels wereavailable for the majority of the eligiblestudies, the diagnosis of diabetes wasbased on HbA1c or fasting glucose mea-surements, without further systematicconfirmation by a second determinationon a separate day; however, this is an in-trinsic limitation of all large observationalstudies, in which the confirmation of di-abetes diagnosis, on at least two separateoccasions, has been never made. Finally,in none of the published studies, exceptfor the studies by Shibata et al. (18) andMing et al. (27), was the diagnosis of di-abetes based on 2-h postload plasma glu-cose levels.

Notwithstanding these limitations, thepresent meta-analysis has several impor-tant strengths. As discussed previously,this meta-analysis provides the mostcomprehensive assessment to date onthe independent prognostic impact ofNAFLD on the long-term risk of incidentdiabetes. These results, obtained by ana-lyzing more than 15,000 new cases of in-cident diabetes among nearly 300,000individuals (incorporating data from

observational cohort studies that arelikely to be an accurate reflection ofNAFLD patients commonly seen in rou-tine clinical practice), provide clear evi-dence that diabetes risk of individualswith NAFLD is significantly higher thanthat of individuals without NAFLD(with a high level of heterogeneity forthe pooled primary analysis and a medium-low quality of the mainly retrospectiveavailable studies). Moreover, it is impor-tant to underline that we employed stan-dardized risk estimates from all eligiblestudies to allow a consistent combinationof estimates across studies. The largenumber of incident cases of diabetes pro-vided high statistical power to quantita-tively assess the association betweenNAFLD and diabetes risk. Finally, selectivereporting bias of studies was not a con-cern in our analyses, as our comprehen-sive search and contact with investigatorsmade it unlikely that any published reportwas missed and visual inspection of plotsand formal tests demonstrated no statis-tical evidence of publication bias.

Currently, there are no approved phar-macological agents for the treatment ofNAFLD. Most interventions evaluated forNAFLD treatment are those commonlyused for the treatment of type 2 diabetesand exert a rather indirect effect on theliver through improvements in both insu-lin sensitivity and insulin action, decreasesin free fatty acid levels, and improvementsin glucose uptake (4,8,53). These pharma-cological interventions have also been todate the most effective treatments forNAFLD, which is perhaps not surprising,considering the high degree of interplaybetween these two diseases.

In conclusion, this largest andmost com-prehensive meta-analysis to date showedthat imaging-diagnosed NAFLD is associ-ated with an approximate doubling of riskof incident diabetes and that this risk seemsto be even greater in presence of more“severe” liver disease (in the fewavailablecohort studies using ultrasonographicscoring systems or noninvasive fibrosismarkers). Because no studieswith biopsy-proven NAFLD were available for theanalysis, the findings of thismeta-analysispave the way for future large, prospec-tive, histologically based studies. It re-mains uncertain whether NAFLD causallyincreases diabetes risk or is a marker ofother shared risk factors. Further studiesare also needed in non-Asian populations,as most of the published studies have

Figure 3—Forest plot and pooled estimates of the effect of the severity of NAFLD (defined byultrasonography [US] or high NFS) on the risk of incident diabetes in four eligible studies. ES, effectsize.


been conducted in Asian populations,where large populations undergo regularhealth checkups, including liver ultraso-nography. Finally, additional studies arealso required to establish whether addingNAFLD (or the different components ofliver disease in NAFLD) to the currentlyavailable algorithms will improve risk pre-diction for diabetes. Despite the above-mentioned caveats, current clinical guidelinesdo recommend routine screening for diabe-tes in patients with NAFLD (53), and there-fore there is a need to clarify themagnitudeof risk of incident diabetes that is associ-ated with the stages of liver disease inNAFLD.

Funding. C.D.B. is supported in part by theSouthampton National Institute for Health Re-search (NIHR) Biomedical Research Centre. G.T.is supported inpart bygrants fromtheUniversitySchool of Medicine of Verona, Verona, Italy.Duality of Interest. No potential conflicts of in-terest relevant to this article were reported.Author Contributions. A.M. and G.T. contrib-uted to study concept and design, acquisition ofdata, and drafting of the manuscript. A.M.contributed to statistical analysis of data.A.M., C.D.B., E.B., and G.T. contributed to anal-ysis and interpretation of data. C.D.B. and E.B.contributed to critical revision of themanuscriptfor important intellectual content. G.T. is theguarantor of this work and, as such, had fullaccess to all the data in the study and takesresponsibility for the integrityof thedataand theaccuracy of the data analysis.

References1. Lonardo A, Bellentani S, Argo CK, et al.; Non-alcoholic Fatty Liver Disease Study Group. Epide-miological modifiers of non-alcoholic fatty liverdisease: focus on high-risk groups. Dig Liver Dis2015;47:997–10062. Younossi ZM, Koenig AB, Abdelatif D, Fazel Y,Henry L, Wymer M. Global epidemiology of non-alcoholic fatty liver diseasedmeta-analytic as-sessment of prevalence, incidence, and outcomes.Hepatology 2016;64:73–843. Anstee QM, Targher G, Day CP. Progression ofNAFLD to diabetes mellitus, cardiovascular dis-ease or cirrhosis. Nat Rev Gastroenterol Hepatol2013;10:330–3444. Bril F, Cusi K. Management of nonalcoholicfatty liver disease in patients with type 2 dia-betes: a call to action. Diabetes Care 2017;40:419–4305. Wild SH, Morling JR, McAllister DA, et al.; Scot-tish and Southampton Diabetes and Liver DiseaseGroup; Scottish Diabetes Research Network Epi-demiologyGroup. Type 2 diabetes and risk of hos-pital admission or death for chronic liver diseases.J Hepatol 2016;64:1358–13646. Baffy G, Brunt EM, Caldwell SH. Hepatocellularcarcinoma in non-alcoholic fatty liver disease: anemerging menace. J Hepatol 2012;56:1384–13917. Byrne CD, Targher G. NAFLD: a multisystemdisease. J Hepatol 2015;62(Suppl.):S47–S64

8. Targher G, Marchesini G, Byrne CD. Risk oftype 2 diabetes in patients with non-alcoholicfatty liver disease: causal association or epiphe-nomenon? Diabetes Metab 2016;42:142–1569. Adams LA, Anstee QM, Tilg H, Targher G. Non-alcoholic fatty liver disease and its relationshipwith cardiovascular disease and other extrahe-patic diseases. Gut 2017;66:1138–115310. Musso G, Gambino R, CassaderM, Pagano G.Meta-analysis: natural history of non-alcoholicfatty liver disease (NAFLD) and diagnostic accu-racy of non-invasive tests for liver disease sever-ity. Ann Med 2011;43:617–64911. Ballestri S, Zona S, Targher G, et al. Nonalco-holic fatty liver disease is associated with an al-most twofold increased risk of incident type 2diabetes and metabolic syndrome. Evidencefrom a systematic review and meta-analysis.J Gastroenterol Hepatol 2016;31:936–94412. Machado MV, Cortez-Pinto H. Non-invasivediagnosis of non-alcoholic fatty liver disease.A critical appraisal. J Hepatol 2013;58:1007–101913. Higgins JPT, Green S, Eds. Cochrane Hand-book for Systematic Reviews of Interventions [In-ternet]. Version 5.1.0 [updated March 2011]. TheCochrane Collaboration. Available from http://handbook.cochrane.org. Accessed 31 May 201714. Higgins JP, Thompson SG. Quantifying het-erogeneity in a meta-analysis. Stat Med 2002;21:1539–155815. Furukawa TA, Guyatt GH, Griffith LE. Can weindividualize the ‘number needed to treat’? Anempirical study of summary effect measures inmeta-analyses. Int J Epidemiol 2002;31:72–7616. EggerM, SmithGD, Phillips AN.Meta-analysis:principles and procedures. BMJ 1997;315:1533–153717. Okamoto M, Takeda Y, Yoda Y, Kobayashi K,Fujino MA, Yamagata Z. The association of fattyliver and diabetes risk. J Epidemiol 2003;13:15–2118. Shibata M, Kihara Y, Taguchi M, Tashiro M,Otsuki M. Nonalcoholic fatty liver disease is a riskfactor for type2diabetes inmiddle-aged Japanesemen. Diabetes Care 2007;30:2940–294419. Kim CH, Park JY, Lee KU, Kim JH, KimHK. Fattyliver is an independent risk factor for the devel-opment of type 2 diabetes in Korean adults. Dia-bet Med 2008;25:476–48120. Bae JC, Rhee EJ, Lee WY, et al. Combinedeffect of nonalcoholic fatty liver disease and im-paired fasting glucose on the development oftype 2 diabetes: a 4-year retrospective longitudi-nal study. Diabetes Care 2011;34:727–72921. Park SK, Seo MH, Shin HC, Ryoo JH. Clinicalavailability of nonalcoholic fatty liver disease as anearly predictor of type 2 diabetes mellitus in Ko-rean men: 5-year prospective cohort study. Hep-atology 2013;57:1378–138322. Sung KC, JeongWS, Wild SH, Byrne CD. Com-bined influence of insulin resistance, overweight/obesity, and fatty liver as risk factors for type 2diabetes. Diabetes Care 2012;35:717–72223. Kasturiratne A, Weerasinghe S, DassanayakeAS, et al. Influence of non-alcoholic fatty liver dis-ease on the development of diabetes mellitus.J Gastroenterol Hepatol 2013;28:142–14724. Chang Y, JungHS, YunKE, Cho J, Cho YK,RyuS.Cohort study of non-alcoholic fatty liver disease,NAFLD fibrosis score, and the risk of incident di-abetes in a Koreanpopulation. Am JGastroenterol2013;108:1861–1868

25. Choi JH, Rhee EJ, Bae JC, et al. Increased riskof type 2 diabetes in subjects with both elevatedliver enzymes and ultrasonographically diag-nosed nonalcoholic fatty liver disease: a 4-yearlongitudinal study. Arch Med Res 2013;44:115–12026. Yamazaki H, Tsuboya T, Tsuji K, Dohke M,Maguchi H. Independent association between im-provement of nonalcoholic fatty liver disease andreduced incidence of type 2 diabetes. DiabetesCare 2015;38:1673–167927. Ming J, Xu S, Gao B, et al. Non-alcoholic fattyliver disease predicts type 2 diabetesmellitus, butnot prediabetes, in Xi’an, China: a five-year cohortstudy. Liver Int 2015;35:2401–240728. Li WD, Fu KF, Li GM, et al. Comparison ofeffects of obesity and non-alcoholic fatty liver dis-ease on incidence of type 2 diabetes mellitus.World J Gastroenterol 2015;21:9607–961329. Shah RV, Allison MA, Lima JA, et al. Liver fat,statin use, and incident diabetes: theMulti-EthnicStudy of Atherosclerosis. Atherosclerosis 2015;242:211–21730. Fukuda T, Hamaguchi M, Kojima T, et al. Theimpact of non-alcoholic fatty liver disease on in-cident type 2 diabetes mellitus in non-overweightindividuals. Liver Int 2016;36:275–28331. Chen GY, Cao HX, Li F, et al. New risk-scoringsystem including non-alcoholic fatty liver diseasefor predicting incident type 2 diabetes in EastChina: Shanghai Baosteel Cohort. J Diabetes In-vestig 2016;7:206–21132. Li Y, Wang J, Tang Y, et al. Bidirectional asso-ciation between nonalcoholic fatty liver diseaseand type 2 diabetes in Chinese population: evi-dence from the Dongfeng-Tongji cohort study.PLoS One 2017;12:e017429133. Ma J, Hwang SJ, Pedley A, et al. Bi-directionalanalysis between fatty liver and cardiovasculardisease risk factors. J Hepatol 2017;66:390–39734. Chen SC, Tsai SP, Jhao JY, Jiang WK, Tsao CK,Chang LY. Liver fat, hepatic enzymes, alkalinephosphatase and the risk of incident type 2 dia-betes: a prospective study of 132,377 adults. SciRep 2017;7:464935. LiuM,Wang J, Zeng J, Cao X, HeY. Associationof NAFLD with diabetes and the impact of BMIchanges: a 5-year cohort study based on 18,507elderly. J Clin Endocrinol Metab 2017;102:1309–131636. Björkström K, Stål P, Hultcrantz R, HagströmH. Histologic scores for fat and fibrosis associatewith development of type 2 diabetes in patientswith non-alcoholic fatty liver disease. Clin Gastro-enterol Hepatol 2017;15:1461–146837. Targher G, Byrne CD, Lonardo A, Zoppini G,Barbui C. Non-alcoholic fatty liver disease and riskof incident cardiovascular disease: ameta-analysis.J Hepatol 2016;65:589–60038. Meex RCR, Watt MJ. Hepatokines: linkingnonalcoholic fatty liver disease and insulin resis-tance. Nat Rev Endocrinol 2017;13:509–52039. Petersen MC, Shulman GI. Roles of diacylgly-cerols and ceramides in hepatic insulin resistance.Trends Pharmacol Sci 2017;38:649–66540. Tilg H, Moschen AR, Roden M. NAFLD anddiabetes mellitus. Nat Rev Gastroenterol Hepatol2017;14:32–4241. Pal D, Dasgupta S, Kundu R, et al. Fetuin-Aacts as an endogenous ligand of TLR4 to promotelipid-induced insulin resistance. NatMed2012;18:1279–1285



42. Stefan N, Häring HU. Circulating fetuin-A andfree fatty acids interact to predict insulin resis-tance in humans. Nat Med 2013;19:394–39543. Sung KC, Wild SH, Byrne CD. Resolution offatty liver and risk of incident diabetes. J Clin En-docrinol Metab 2013;98:3637–364344. Kantartzis K, Peter A, Machicao F, et al.Dissociation between fatty liver and insulin resis-tance in humans carrying a variant of the patatin-like phospholipase 3 gene. Diabetes 2009;58:2616–262345. Benhamed F, Denechaud PD, Lemoine M,et al. The lipogenic transcription factor ChREBPdissociates hepatic steatosis from insulin resis-tance in mice and humans. J Clin Invest 2012;122:2176–219446. Moon YA, Liang G, Xie X, et al. The Scap/SREBP pathway is essential for developingdiabetic fatty liver and carbohydrate-induced

hypertriglyceridemia in animals. Cell Metab2012;15:240–24647. Minehira K, Young SG, Villanueva CJ, et al.Blocking VLDL secretion causes hepatic steatosisbut does not affect peripheral lipid stores or in-sulin sensitivity inmice. J Lipid Res 2008;49:2038–204448. Guerrero R, Vega GL, Grundy SM, BrowningJD. Ethnic differences in hepatic steatosis: an in-sulin resistance paradox? Hepatology 2009;49:791–80149. Liska D, Dufour S, Zern TL, et al. Interethnicdifferences inmuscle, liver and abdominal fat par-titioning in obese adolescents. PLoS One 2007;2:e56950. Nazare JA, Smith JD, Borel AL, et al. Ethnicinfluences on the relations between abdominalsubcutaneous and visceral adiposity, liver fat,and cardiometabolic risk profile: the International

Study of Prediction of Intra-Abdominal Adiposityand Its Relationship With Cardiometabolic Risk/Intra-Abdominal Adiposity. Am J Clin Nutr 2012;96:714–72651. Stefan N, Fritsche A, Schick F, HäringHU. Phe-notypes of prediabetes and stratification of cardi-ometabolic risk. Lancet Diabetes Endocrinol 2016;4:789–79852. Stefan N, Schick F, Häring HU. Causes, char-acteristics, and consequences ofmetabolically un-healthy normal weight in humans. Cell Metab2017;26:292–30053. European Association for the Study of theLiver (EASL); European Association for the Studyof Diabetes (EASD); European Association for theStudy of Obesity (EASO). EASL-EASD-EASO ClinicalPractice Guidelines for the management of non-alcoholic fatty liver disease. J Hepatol 2016;64:1388–1402


nonalcoholic fatty liver disease and risk of incident type ...nafld was based on either imaging or...

Documents