The relationship between obesity, low back pain and lumbar disc degeneration when genetics and the environment are considered: a systematic review of twin studies

The Spine Journal 15 (2015) 1106–1117

Review Article

The relationship between obesity, low back pain, and lumbar discdegeneration when genetics and the environment are considered:

a systematic review of twin studies

Amabile B. Dario, MSca,*, Manuela L. Ferreira, PhDb, Kathryn M. Refshauge, PhDa,Thais S. Lima, MScc, Juan R. Ordo~nana, PhDd,e, Paulo H. Ferreira, PhDa

aDiscipline of Physiotherapy, Faculty of Health Sciences, The University of Sydney, PO Box 170, 75 East Street Lidcombe, Sydney, NSW, Australia 2141bThe George Institute for Global Health and Institute of Bone and Joint Research, Kolling Institute, Sydney Medical School, The University of Sydney,

Level 13, 321 Kent Street, Sydney, NSW, Australia 2141cBiomechanics and Motor Control Research Group, Science and Technology Faculty-Universidade Estadual Paulista, Presidente Prudente,

Sao Paulo, Brazil 19060-900dMurcia Twin Registry, Department of Human Anatomy and Psychobiology, University of Murcia, Spain 30100

eIMIB-Arrixaca, Department of Human Anatomy and Psychobiology, Murcia, Spain 30100

Received 15 September 2014; revised 13 January 2015; accepted 1 February 2015

Abstract BACKGROUND CONTEXT: The relationsh

FDA device/drug

Author disclosure

close. KMR: Nothing

to disclose. PHF: No

Funding sources:

ence without Borders

http://dx.doi.org/10.10

1529-9430/� 2015 E

ips between obesity and low back pain (LBP) andlumbar disc degeneration (LDD) remain unclear. It is possible that familial factors, including genet-ics and early environment, affect these relationships.PURPOSE: To investigate the relationship between obesity-related measures (eg, weight, bodymass index [BMI]) and LBP and LDD using twin studies, where the effect of genetics and earlyenvironment can be controlled.STUDY DESIGN: A systematic review with meta-analysis.METHODS: MEDLINE, CINAHL, Scopus, Web of Science, and EMBASE databases weresearched from the earliest records to August 2014. All cross-sectional and longitudinal observationaltwin studies identified by the search strategy were considered for inclusion. Two investigators inde-pendently assessed the eligibility, conducted the quality assessment, and extracted the data. Metaa-nalyses (fixed or random effects, as appropriate) were used to pool studies’ estimates of association.RESULTS: In total, 11 articles met the inclusion criteria. Five studies were included in theLBP analysis and seven in the LDD analysis. For the LBP analysis, pooling of the five studiesshowed that the risk of having LBP for individuals with the highest levels of BMI or weight wasalmost twice that of people with a lower BMI (odds ratio [OR] 1.8; 95% confidence interval[CI] 1.6–2.0; I250%). A dose-response relationship was also identified. When genetics andthe effects of a shared early environment were adjusted for using a within-pair twin case-control analysis, pooling of three studies showed a reduced but statistically positive associationbetween obesity and prevalence of LBP (OR 1.5; 95% CI 1.1–2.1; I250%). However, the asso-ciation was further diminished and not significant (OR 1.4; 95% CI 0.8–2.3; I250%) whenpooling included two studies on monozygotic twin pairs only. Seven studies met the inclusioncriteria for LDD. When familial factors were not controlled for, body weight was positively as-sociated with LDD in all five cross-sectional studies. Only two cross-sectional studies investi-gated the relationship between obesity-related measures and LDD accounting for familialfactors, and the results were conflicting. One longitudinal study in LBP and three longitudinalstudies in LDD found no increase in risk in obese individuals, whether or not familial factorswere controlled for.

status: Not applicable.

s: ABD: Nothing to disclose. MLF: Nothing to dis-

to disclose. TSL: Nothing to disclose. JRO: Nothing

thing to disclose.

The author ABD is supported by the program ‘‘Sci-

,’’ Brazil.

Conflict of interest: The authors declare that they have no competing

interests.

* Corresponding author. Faculty of Health Sciences, University of Syd-

ney, PO Box 170, Lidcombe 1825, Australia. Tel.: (61) 293-519-562; fax:

(61) 293-519-601.

E-mail address: [email protected] (A.B. Dario)

16/j.spinee.2015.02.001

lsevier Inc. All rights reserved.

1107A.B. Dario et al. / The Spine Journal 15 (2015) 1106–1117

CONCLUSIONS: Findings from this review suggest that genetics and early environment are pos-sible mechanisms underlying the relationship between obesity and LBP; however, a direct causallink between these conditions appears to be weak. Further longitudinal studies using the twin designare needed to better understand the complex mechanisms underlying the associations between obe-sity, LBP, and LDD. � 2015 Elsevier Inc. All rights reserved.

Keywords: Obesity; Body mass index; Body weight; Low back pain; Lumbar disc degeneration; Genetics; Twins

Introduction

Low back pain (LBP) is a major health problem globally[1], being the largest contributor to the number of years thatpeople live with disability [2]. Although decades of re-search have been dedicated to identifying the etiology ofLBP, the factors that trigger an episode of LBP remain un-clear [3], limiting the possibility of designing effective pre-ventative strategies. A variety of factors have inconsistentlybeen found to be associated with LBP, and the increasedrisk has been small. One of these factors, obesity, is a po-tential target for prevention strategies, and therefore, ithas been the focus of several studies in the field [4,5].

Obesity is recognized as a major public health problem,and its prevalence is increasing rapidly in westernizedcountries [6,7]. Obese individuals are at higher risk of de-veloping a wide spectrum of chronic diseases such as dia-betes, cardiovascular disease, cancer, and musculoskeletaldisorders, such as spinal problems [8]. Body weight, an im-portant factor related to spinal loading, has been associatedwith several signs of lumbar disc degeneration (LDD),including disc space narrowing [9] and decreased signalintensity of the lumbar intervertebral discs [10]. Despitecontroversy [11–13], LDD has been proposed as one ofthe main risk factors of LBP [10,14].

Previous studies have suggested that familial factors (ie,early environmental and genetic influences) play an impor-tant role on obesity, LBP, and LDD. According to twinstudies, the estimated contribution from heritability for to-tal body fat ranges between 70% and 80% [15], for LBP be-tween 30% and 46% [16], and for LDD the contributionranges between 47% and 60% [14], suggesting a major ge-netic component in these conditions. However, most studiesthat investigated the relationship between obesity, LBP, andLDD did not account for genetic or early environmentalfactors, which might explain their conflicting findings.

Twin studies represent a unique and powerful design forinvestigating risk factors for health conditions as they allowcontrolling for various confounders, including genetic fac-tors, consequently providing more precise estimates of risk.To our knowledge, there has been no published systematicreview specifically investigating the relationship betweenobesity, LBP, and LDD in twin studies. Therefore, thissystematic review aimed to investigate whether there isan association between obesity and LBP and obesity andLDD, and whether this association is influenced by geneticsand early environment.

Methods

A review protocol was registered in the ‘‘Internationalprospective register of systematic reviews’’ under theregistration number CRD42014005747. We used the Meta-analysis of Observational Studies in Epidemiology guide-lines to lead each section of this systematic review [17].

Search strategy

MEDLINE, CINAHL, Scopus, Web of Science, andEMBASE databases were searched using a combinationof key words related to obesity, LBP, and LDD. The searchwas conducted from the earliest records to August 2014 toidentify cross-sectional and longitudinal observational twinstudies that investigated the obesity-LBP and obesity-LDDrelationships. Additionally, citation tracking was conductedof the reference list of included studies and relevant publi-cations in the field. If additional clarification or data wererequired, authors were contacted by email.

Selection of studies

All articles identified by the search strategy were inde-pendently screened by two investigators (ABD and TL),with a third independent investigator (PHF) resolving anydisagreement. The assessment involved three stages:screening of titles, abstracts, and full text. The number ofstudies identified was recorded for all screening stages.

Inclusion and exclusion criteriaWe included cross-sectional and longitudinal observatio-

nal studies that investigated the relationship between obesityand LBP and obesity and LDD using twins, where the ge-netic and early shared environment components were orwere not adjusted for (case-control studies and studies thatrecruited twin samples, respectively). Twins needed to ac-count for at least 90% of the total sample, with no restrictionon age, gender, or zygosity. No restriction was applied onthe year of publication or language. Studies were excludedif they investigated specific spinal pathologies (fracture,cancer, and systemic diseases) or pregnancy-related LBP.

Exposure factorsThe exposure factors were obesity or a measure of obe-

sity such as body mass index (BMI), percent fat mass, orweight.

1108 A.B. Dario et al. / The Spine Journal 15 (2015) 1106–1117

OutcomesThe outcomes of interest were present occurrence (prev-

alence) of LBP or LDD in cross-sectional studies and futureoccurrence (incidence) of LBP and LDD in longitudinalstudies. All definitions for LBP were accepted, as these var-ied considerably among studies. For LDD, studies were in-cluded if the outcome was a pathoanatomical finding basedon imaging, such as disc space narrowing or changes in discsignal.

Data extraction

Data were extracted from all included studies regardingparticipants, sampling methods, response rates, length offollow-up, and information on exposure factors (obesity-re-lated measures) and potential confounders (eg, gender,age). A standardized form developed for this systematic re-view was used to extract data. When studies performed lon-gitudinal and cross-sectional analyses, data from bothanalyses were extracted. When studies reported more thanone cross-sectional analysis, estimates were extracted fromthe analysis with the largest sample size. We extracted es-timators such as odds ratios (ORs) and measurements ofvariability for the associations between obesity and bothLBP and LDD. To investigate if genetics and early sharedenvironment factors affected these associations, data fromstudies reporting results from the total sample and fromcase-control analyses were extracted separately. In the totalsample analysis, no adjustment for genetics or early sharedenvironment factors was performed and twins were ana-lyzed as individuals rather than pairs, irrespective of dis-cordance for LBP within twin pairs. The case-controlanalysis included only complete twin pairs who were dis-cordant for LBP status, that is, one twin reported LBP,whereas the other did not.

This approach enabled control of various confounders,including genetic factors and twins’ early shared environ-mental factors. It was assumed that the case-control designallowed clear identification of a relationship between anoutcome (eg, LDD) and an exposure factor (eg, BMI) be-cause it controls for genetic factors and early shared envi-ronment. Theoretically, when the magnitude of theassociation between two variables (eg, LDD and BMI) in-creases from the total sample analysis (no adjustment forgenetic factors or early shared environment) to a monozy-gotic (MZ) case-control analysis (adjustment for earlyshared environment and approximately 100% of geneticfactors), the relationship between the two variables is moredirect and possibly more consistent with a direct causalpath [18].

Methodologic quality

The quality of included studies was assessed using astandardized checklist based on the recommendations forpublishing a systematic review [19,20] and the STROBE

(Strengthening the Reporting of Observational Studies inEpidemiology Statement) guidelines [21]. The checklistcomprised eight criteria: representative sample, definedsample, blinding of assessors to the predictor, blinding ofassessors to the outcome, follow-up rate greater than85%, defined method of assessment, reporting on outcomedata, and statistical adjustment for potential confounders.Two members (ABD and TL) of the research team con-ducted the critical appraisal independently. Results werecompared and disagreements were resolved by the third in-dependent investigator (PHF).

Meta-analysis

Extracted estimates of risk and confidence intervals (CIs)were synthesized in a meta-analysis, when the data reportedwere sufficiently homogenous. Where studies providedresults for more than one description of LBP, we chose thedefinition that involved longer andmore disabling symptoms(eg, chronic instead of acute LBP). When predictors werepresented in incremental categories, we selected the cate-gory with higher levels of exposure for the meta-analysis.Dose-response relationship was calculated when studiesprovided estimated risks for different levels of exposure(eg, overweight and obesity). For those studies with differentdegrees of control for confounders, we used the model thatadjusted for the greatest number of variables. We used thelowest available anthropometric level for weight or the nor-mal category for BMI as the reference category. Data werepooled using Comprehensive Meta-Analysis software, ver-sion 2.2.064 (Biostat, Englewood, USA, 2008). Study heter-ogeneity was analyzed using visual inspection of graphs andthe I2 statistic. True homogeneity was considered to be I250,low heterogeneity lower than 30%, moderate 30% to 49%,substantial 50% to 74%, and considerable heterogeneitygreater than 75% [22]. In case of heterogeneity equal to orhigher than substantial, a random effects model was usedto calculate the pooled OR estimates and their variances.

Results

Included studies

The systematic search identified 822 publications, 769were removed after screening for duplicates and ineligibletitles and abstracts (Fig. 1). Fifty-three studies were identi-fied as potentially eligible and, after full-text screening, 11publications met our inclusion criteria and were included inthe review [4,5,14,23–30]. One study reported data for LBPand LDD [26]. The included studies were published be-tween 1999 and 2011. The total number of participantsfrom studies assessing LBP and LDD was 45,784 and4,205, respectively. Included studies recruited twins fromregistries in the United States [5], Finland [14,23–25,27],Australia [28], United Kingdom [26,28,30], and Denmark[4,29]. Comprehensive descriptions are provided in

Fig. 1. Selection of the included studies. LBP, low back pain; LDD, lumbar disc degeneration.

1109A.B. Dario et al. / The Spine Journal 15 (2015) 1106–1117

Table 1 for studies investigating the relationship betweenobesity and LBP symptoms and in Table 2 for obesityand LDD [31].

Methodologic quality

A summary of the methodologic quality of includedstudies is shown in Table 3. All five (100%) studies inves-tigating LBP had a representative and well-defined sample,well-defined method of assessment of predictors and out-comes, and reported outcome data and conducted adjust-ment for potentially confounding factors. One study didnot include blinded assessors for predictors and outcomes[26], and the only included longitudinal study had afollow-up rate below 85% [29]. For LDD, six (86%) studieshad a representative sample and all seven (100%) definedthe sample and the method of assessment of predictorsand outcomes. Assessors were blinded for predictors insix (86%) and for outcomes in five (71%) studies. Four(57%) of the studies reported outcome data and adjustedthe analysis for potentially confounding factors. Two of

the three longitudinal studies for LDD had a follow-up ratebelow 85%.

Assessment and definition of obesity-related measures

The most common measure of obesity in the includedstudies was BMI [4,5,26,28,29]. The second most commonmeasure was body weight [14,24–27,30], followed by per-centage of body fat [24] and intrinsic disc loading (esti-mated by body weight divided by the axial spinal discarea) [23]. The definition and cutoff points used in eachstudy are provided in Table 1 and Table 2 (LBP and LDDstudies, respectively).

Assessment and definition of LBP and LDD

Low back pain was assessed by questionnaires with abody chart in all [4,26,29,30] except one study [5]. Lowback pain was defined as pain, stiffness, and discomfortin the lumbar area accompanied [26,30] or not [4,5,29]by disability. Duration of pain and disability ranged from

Table 1

Characteristics of the included studies for low back pain

Study Design Study population Obesity measure Low back pain measure Results: total sample

Results: case-control* or genetic

analysis

Leboeuf-Yde

et al. [4],

1999

Cross-sectional Danish twins aged 12 to 41 y

n529,424 (3,751 complete MZ

pairs)

Sex: 52% female

BMI (underweight,!20 kg;

normal weight, 20–24 kg;

overweight, 25–29 kg; heavily

overweight,O29 kg)

Yes/no: questionnaire with body

chart. LBP previous y lasting

1–7; 8–30; andO30 d.

Overweight positively associated

with LBP. Higher ORs for LBP

O30 d were found when

compared with LBP 1–7 d in

overweight and heavily

overweight. In heavily

overweight twins, the OR rose

from 0.8 in LBP 1–7 d to 1.7 in

LBPO30 d. Dose response

found for LBPO30 d: 0.7

underweight; 1.6 overweight;

and 1.7 heavily overweight.

The significant association

between BMI and LBP was not

found in MZ twin pairs of

dissimilar body weight:

overweight OR 1.1 (95% CI

0.8–1.5); heavily overweight

OR 1.1 (95% CI 0.5–2.0).

MacGregor

et al. [30],

2004

Cross-sectional English twins aged 45 to 72 y

n51,064 (181 MZ and 351 DZ

pairs)

Sex: 100% female

Weight Yes/no: questionnaire with body

chart. Persistent LBP: pain

with a total duration ofO30 d

associated with disability.

Heavier twins have 2.36 times

(95% CI 1.47–3.76) greater

chance of having LBP than

lighter twins (p!.01). Dose

response found with ORs

increasing according to the

increase in quartiles of weight

(1 � quartile OR 1; 2 � quartile

OR 1.94, 95% CI 1.22–3.09;

3 � quartile OR 2.25, 95% CI

1.42–3.56; 4 � quartile OR

2.36, 95% CI 1.47–3.76).

Weight-LBP association is

explained mostly by shared

genetic factors rather than

shared familial environmental

factors.

Hestbaek

et al. [29],

2006

Cross-sectional

and longitudinal

(8-y follow-up)

Danish twins aged 12 to 22 y

n59,569 (413 MZ pairs

discordant for LBP at the

baseline)

Sex: 51% female

BMI (underweight!17 kg;

normal weight 17–23.9 kg;

overweight 24–28.9 kg; obesity

O29 kg/BMI dichotomized

O24 kg overweight)

Yes/no: questionnaire with body

chart. LBP (at all): at least 1 d

during the previous y.

Persistent LBP: at least 30 d

during the previous y.

Cross-sectional: Overweight

positively associated with

persistent LBP (OR 1.38; 95%

CI 1.06–1.79). Dose response

not found: overweight OR 1.41

(95%CI 0.82–2.43); Obese OR

1.01 (95% CI 0.41–2.49). Sex:

overweight associated with

LBP only for girls (OR 1.7;

95% CI not reported).

Longitudinal: 8-y follow-up did

not confirm obesity as a risk

factor for persistent LBP (OR

1.01; 95% CI 0.9–1.43).

However, OR for LBP for

smokers in relation to

nonsmokers increased with

increased BMI (1.5/1.6/2.6/

11.3).

Cross-sectional: Overweight not

associated with present LBP

(at all) (OR 1.75; 95% CI

0.82–3.90)

Longitudinal: Overweight not

associated with future LBP (at

all) (OR 0.89; 95% CI 0.30–

2.60)

1110

A.B.Dario

etal./TheSpineJournal15(2015)1106–1117

Wright

etal.[5],

2010

Cross-sectional

NorthAmerican

twins,meanage

of31y

n53,471

(66%

MZ)

Sex:62%

female

BMI(overw

eight25to

29.9

kg;

obeseO30kg;underweight

excluded)

Yes/no:questionnaire.Lifetim

e

history

ofchronic

LBP,atleast

3mo.

Overw

eightpositively

associated

withLBP.

OR

foroverweight

was

1.51(95%

CI1.25–1.84);

obeseORwas

1.94(95%

CI

1.53–2.46)when

adjusted

to

age,sex,anddepression.W

hen

adjusted

toageandsex,the

OR

ofoverweightwas

1.55

(95%

CI1.28–1.87)andobese

OR

was

2.11(95%

CI1.67–

2.67).

Alloftheassociationswere

dim

inished,andLBPremained

significantlyassociated

just

withobesity(O

R1.60;95%CI

1.01–2.53).

Livshits

etal.[26],

2011

Cross-sectional

English

twinsaged

18to

84y

n52,256

(371and698ofMZ

andDZtwin

pairs)

Sex:100%

female

Weight;BMI

Yes/no:questionnaire

withbody

chart.LBPassociated

with

disabilitylastingmorethan

1

mo.

Weight

andBMIwerestrongly

correlated

withLBP(r50.91,

p!.0001).Com

paringupper

versuslower

percentilesof

weightsuggested

that

heavier

twinshave1.87times

(95%

CI

1.16–2.99)greater

chance

of

havingLBPthan

lightertwins

(p!.009).Weight

explained

someofthevariance

ofLBP(b

weight0.028

kg;SE0.005).

NA

LBP,

low

backpain;BMI,bodymassindex;

OR,oddsratio;CI,confidence

interval;SE,standarderror;MZ,monozygotic;

DZ,dizygotic;

NA,notanalyzed.

*Twin

pairdissimilar

forbodyweight.

1111A.B. Dario et al. / The Spine Journal 15 (2015) 1106–1117

1 day in the previous year [26] to a lifetime total of 3months [5]. Magnetic resonance imaging was used in allstudies to identify LDD through qualitative assessment[14,24,26,28] or a combination of qualitative and quantita-tive assessments [23,25,27]. The phenotype of LDD wascharacterized by decreased disc height [14,23–28] or discsignal intensity [14,24–28], disc bulging [14,23,25–28], os-teophytes [14,26,28], disc irregularity [14,23], disc hernia-tion [14], or a combination of different parameters [26,28].

Association between obesity-related measures and LBPsymptoms

Results of studies reporting total sample analysesFive studies investigated the relationship between

obesity-related measures (ie, BMI and body weight) andLBP using total samples of twins with no adjustment forfamilial factors [4,5,26,29,30]. All studies reported suffi-cient and similar data to be pooled in a meta-analysis. Pool-ing of the data (Fig. 2) revealed that twins classified in thehighest level of BMI or weight had 1.8 times increasedodds of having LBP (OR 1.8; 95% CI 1.6–2.0; p5.001;I250%) compared with those with normal or lighter bodyweight. Pooling of longitudinal data was not possible as on-ly one study was identified. This study, conducted in Den-mark, followed participants for 8 years, finding no effect ofBMI on LBP incidence (OR 1.0; 95% CI 0.9–1.4) [29].

Obesity-LBP dose-response relationshipA possible dose-response relationship between obesity-

related measures (weight or BMI) and LPB was investi-gated in four studies [4,5,29,30]. Pooling of the four studies(Fig. 2) revealed that the prevalence of LBP in obese twins(OR 1.8; 95% CI 1.6–2.0; p5.001; I250%) was higher thanthe prevalence of LBP in overweight twins (OR 1.5; 95%CI 1.3–1.7; p5.001; I2555%). In addition, one study foundthat twins who were underweight had a lower prevalence ofLBP than twins who had normal values of BMI (OR 0.7;95% CI not reported) [4].

Results of studies reporting analyses that accounted for ge-netic and early environmental factors

A total of four studies investigated the effect of geneticfactors and early shared environment on the LBP-obesityrelationship [4,5,29,30]. Three [4,5,29] studies conducteda within-pair case-control analysis, where twin pairs dis-similar for body weight (one twin classified as normalweight and the other as overweight or obese) were ana-lyzed. Pooling of these twin studies showed a statisticallysignificant positive association between obesity/overweightand prevalence of LBP (OR 1.5; 95% CI 1.1–2.1; p5.02;I250%). However, when pooling included case-controlstudies with MZ twins only [4,29], the association was nolonger statistically significant (OR 1.4; 95% CI 0.8–2.3;p5.26; I250%). The only longitudinal study that used awithin-pair case-control design did not identify a significant

Table 2

Characteristics of the included studies for lumbar disc degeneration

Study Design Study population Obesity measure

Lumbar disc degeneration

measure* Results: total sample

Results: case-control or

genetic analysis

Videman et al. [14], 2006y Longitudinal (5-y follow-

up)

Finish twins aged 35 to 69 y

at baseline

n5140 (70 MZ twin pairs)

Sex: 100% male

Weight Disc height

Disc bulging

Disc herniation

Osteophytes

High-intensity zones

Body weight not associated

with the progression of

LDD over 5 y.

NA

Battie et al. [25], 2008 Cross-sectional Finish twins aged 35 to 70 y

n5600 (152 MZ and 148

DZ twin pairs)

Sex: 100% male

Weight Disc height

Disc bulging

Disc signal intensity

Body weight associated

with signs of LDD, at

least at one spinal level.

NA

Videman et al. [23], 2008ay Longitudinal (5-y follow-

up).

Finish twins aged 35 to 69 y

at baseline

n5134 (67 MZ twin pairs)

Sex: 100% male

Weight; BMI Disc height

Disc bulging

Body weight not associated

with the progression of

LDD over 5 y.

NA

Videman et al. [27], 2008b Cross-sectional Finish twins aged 35 to 70 y

at baseline

n5519 (234 MZ and 285

DZ twin pairs)

Sex: 100% male

Intrinsic disc loading

parameter (body weight/

axial disc area)

Disc height

Disc bulging

Disc signal

Disc irregularity

The variance explained by

the intrinsic disc loading

in the L1–L4 discs was

3% in disc height

narrowing, 8% in

anterior bulging, 5% in

posterior bulging, and

7% in the adjusted disc

signal.

NA

Videman et al. [24], 2010 Cross-sectional Finish male twins aged 36

to 69 y

n588 (44 MZ pairs with 8

kg or more weight

discordant)

Sex: 100% male

Weight; BMI; body fat Disc signal intensity

variation

Disc height

NA Heavier MZ twins had

5.4% (p5.005) higher

disc signal variation in

L1–L4 compared with

the lighter co-twin.

Greater body mass

appears to delay L1–L4

disc desiccation slightly.

Body weight not

associated with the disc

height.

Livshits et al. [26], 2011 Cross-sectional English twins aged 18 to 84

y

n52,256 (371 and 698 MZ

and DZ pairs)

Sex: 100% female

Weight; BMI Disc signal intensity

Disc height

Disc bulging

Anterior osteophytes

Overall summary score

(LSUM)

Weight explained some of

the variance of lumbar

degeneration signs

(LSUM) (b weight 0.086

kg; SE 0.02).

NA

1112

A.B.Dario

etal./TheSpineJournal15(2015)1106–1117

William

set

al.[28],2011

Longitudinal

(10.7-y

follow

-up).

English

andAustralians

twinsaged

32to

70yat

thebaseline

n5468(90MZand144

same-sexDZpairs)

Sex:95%

female

BMI

Discsignal

intensity

Discheight

Discbulging

Anteriorosteophytes

OverallLDD

summary

score

(LSUM)

Cross-sectionalbaselinez:

BMIpositively

associated

withLDD

(OR1.23;

95%

CI1.05–

1.44;

p5.011)

Cross-sectionalfollow

-upz :

BMIpositively

associated

withLDD

(OR1.33;

95%

CI1.15–

1.53;

p5.001)

Longitudinal:BMInot

associated

withthe

progressionofLDD

over

10.7

y.

Cross-sectionalbaselinez:

BMInotassociated

with

LDD

inMZtwins(O

R

0.89;

95%

CI0.61–1.29;

p5.52),justin

DZtwins

(OR1.43;

95%

CI1.23–

1.65;

p!.001).

Cross-sectionalfollow-upz :

BMIassociated

with

LDD

inMZ(O

R1.56;

95%

CI1.08–2.26;

p5.02)andDZ(O

R

1.25;

95%

CI;p!

.001)

twins.

LDD,lumbar

discdegeneration;LSUM,Lumbar

discdegenerationsummaryscore;BMI,bodymassindex;SE,standard

error;MZ,monozygotic;

DZ,dizygotic;

NA,notanalyzed;

OR,oddsratio;CI,

confidence

interval.

*Im

aged

bymagnetic

resonance

imaging.

yArticleshaveusedasimilar

sample.

zSupplementary

dataprovided

byauthors.

1113A.B. Dario et al. / The Spine Journal 15 (2015) 1106–1117

association between obesity and incidence of LBP (OR 0.9;95% CI 0.3–2.6) [29] (Fig. 3).

Association between obesity and LDD

Despite seven studies [14,23–28] having investigated theassociation between obesity-related measures and signs ofLDD, a meta-analysis was not feasible because of the het-erogeneity in study estimates of association or lack of out-come data.

Results of studies reporting total sample analysesSigns of LDD such as disc height narrowing, bulging,

and signal intensity were associated with increased bodyweight [23,25,28]. Body weight and intrinsic load in theL1–L4 discs explained between 1.4% [26] and 17% [23]of the variance in signs of lumbar degeneration. Three lon-gitudinal studies investigating the effect of obesity in theprogression of LDD did not find any increase in risk of de-veloping LDD over 5 [14,27] or 10 years [28].

Results of studies reporting analyses that accounted for ge-netic and early environmental factors

Only two studies, both cross-sectional, that investigatedthe relationship between obesity-related measures and LDDaccounted for genetic and early environmental factors[24,28]. One study used a within-pair MZ case-controlanalysis and found that heavier twin had 5.4% higher discsignal variation (a measure of water concentration in inter-vertebral discs) in L1–L4 intervertebral discs comparedwith the lighter co-twin. [24].

In pairs of same-sex twins [28], a positive associationbetween BMI and LDD was found for the total twin sample(OR 1.2; 95% CI 1.1–1.4) and for dizygotic (DZ) twins(OR 1.2; 95% CI 1.2–1.7), but not for MZ twins (OR0.9; 95% CI 0.6–1.3) [28]. A cross-sectional analysis ofthe 10-year follow-up data showed that BMI was associatedwith LDD in the total sample analysis (OR 1.3; 95% CI1.2–1.5) and in both DZ (OR 1.3; 95% CI 1.0–1.5) andMZ (OR 1.6; 95% CI 1.1–2.3) twins.

Discussion

Main findings

It is known that familial factors (genetic and early envi-ronment) play a significant role in obesity, LBP, and LDD[15,16,32]. However, there is limited evidence to examinewhether familial factors affect a potential relationship be-tween obesity and both LBP and LDD [33]. Twin studies,particularly using a within-pair twin case-control design,have the potential to provide less biased estimates of riskfor a condition by controlling for possible confoundingfrom genetic factors and early shared environment [34].We aimed to review and summarize the evidence from twinstudies that investigated the effect of obesity-related

Table 3

Methodologic quality assessment of included studies

Study

Sources of bias

Appropriate measure

of variables Confounding

Representative

sample*

Defined

sampley

Assessor

blinded

to predictorz

Assessor

blinded

to outcomexFollow-up

rateO85%jjMethods of

assessment{

Outcome

data

reported#Statistical

adjustment**

Low back pain

Leboeuf-Yde et al. [4], 1999 Y Y Y Y N/A Y Y Y

MacGregor et al. [30], 2004 Y Y Y Y N/A Y Y Y

Hestbaek et al. [29], 2006 Y Y Y Y N Y Y Y

Wright et al. [5], 2010 Y Y Y Y N/A Y Y Y

Livshits et al. [26], 2011 Y Y N N N/A Y Y Y

Lumbar disc degeneration

Videman et al. [14], 2006 Y Y Y Y Y Y N N

Battie et al. [25], 2008 Y Y Y Y N/A Y N N

Videman et al. [23], 2008a Y Y Y Y N Y N N

Videman et al. [27], 2008b Y Y N N N/A Y Y Y

Videman et al. [24], 2010 N Y Y N N/A Y Y Y

Livshits et al. [26], 2011 Y Y Y Y N/A Y Y Y

Williams et al. [28], 2011 Y Y Y Y N Y Nyy Y

Y, yes; N, no; N/A, not applicable.

* Participants were selected as consecutive or random cases.y Description of participant source and inclusion and exclusion criteria.z Assessor unaware of the predictor of the study.x Assessor/patient (self-reported questionnaire) unaware of at least one outcome of the study.jj Outcome data were available for at least 85% of participants at one follow-up point.{ Standardized and fully defined method to assess the predictor and outcome.# Raw data, percentages, p value, risk estimates reported, and confidence interval.

** Multivariate analysis conducted with adjustment for potentially confounding factors.yy Authors provided supplementary data after requested.

1114 A.B. Dario et al. / The Spine Journal 15 (2015) 1106–1117

measures on LBP and LDD when familial influences wereor were not controlled for. The results of this systematic re-view suggest that individuals who are obese or overweightare more likely to have LBP and LDD than those who are inthe normal weight range or underweight. However, aftercontrolling for familial factors, the associations betweenobesity-related measures and LBP appear to diminish andare no longer evident after full adjustment in MZ twins. Re-sults from the longitudinal studies showed evidence thatobesity-related measures do not increase the risk for LBPor LDD, irrespective of adjustment for familial effects.

Obesity and LBP

The magnitude of association between obesity and LBPfound in this review was weak (OR51.8) according to thebenchmarks used for observational studies [35]. However, re-sults tended to be consistent in all five twin studies with sam-ples from four different countries and participants’ agesranging from 12 to 84 years. Our meta-analysis also identifieda dose-response relationship between obesity and LBP. Thesefindings are consistent with results from previous metaanaly-ses, in which a similar effect sizewas found for the associationbetweenobesity and prevalenceofLBP in individuals from thegeneral population [34,36] and a similar dose-response rela-tionship [36]. Although these results tend to support the poten-tial for a causal relationship between obesity and LBP, the

criteria for the temporal relationship required to demonstratecausation has not been fulfilled in our review.We identifiedon-ly one longitudinal study and that study did not identify any in-creased incidence of LBP in obese individuals. Similarly, noeffect was found of obesity on LPB incidence in samples fromthe general population [37,38].

The obesity-LBP association is only apparent in cross-sectional studies, however, the inverse relationship, thatis, LBP leading to obesity, should not be disregarded. Evi-dence from a nontwin study suggests that individuals withLBP, particularly chronic pain, tend to gain more weightthan those with no symptoms of pain [39]. Therefore, tobetter understand the interaction between obesity andLBP, further investigation of the direction of the associa-tion should be conducted in longitudinal studies using atwin design.

Another interesting finding of our review was the smalland nonsignificant association (OR 1.4; 95% CI 0.8–2.3;p5.26) between obesity and LBP observed in the MZ anal-ysis when compared with the total sample analysis. Thispattern of reduced association was consistent across all in-cluded studies that adjusted for genetic factors and with ourown unpublished data from our research group in 156 MZSpanish twins dissimilar for LBP status. Although we can-not rule out the possibility of this finding being a reflectionof smaller samples and larger CIs, the imprecision of thedata is unlikely to be the reason. The pooled MZ analysis

Fig. 2. Pooled ORs of all included cross-sectional studies that investigated the relationship between obesity and low back pain without adjustment for genetic

factors or shared early environment. Pooling is stratified by incremental levels of exposure. Squares represent each individual study. Diamonds represent the

pooled effect. Weight % represents the influence of each study in the overall meta-analysis. OR, odds ratio; CI, confidence interval; I2, heterogeneity of studies.

1115A.B. Dario et al. / The Spine Journal 15 (2015) 1106–1117

included a large sample of 469 twin pairs (938 individuals).These twin pairs were matched for age, sex, and geneticfactors, in addition to twins’ early shared environmentalfactors. By controlling for these confounders, the MZ anal-ysis potentially provides a more precise estimate than thetotal sample analysis. In sum, the trend toward progressivereduction in the association across the phases (total sample:OR 1.8; DZ/MZ twins together: OR 1.5; MZ twins: OR 1.4)suggests that genetic factors and early environment sharedby twins are possibly confounding the association betweenobesity and LBP.

Obesity and LDD

Positive associations between body weight and LDDwere consistently present in the total sample analyses ofall four cross-sectional studies included in our review[23,25,26,28]. However, there was no temporal effect ofbody weight on the progression of LDD; the risk of LDDwas not increased in overweight or obese individuals inany of the three longitudinal studies [14,27,28]. Our results

Fig. 3. Pooled ORs of all included cross-sectional within-pair case-control studie

represent each individual study. Diamonds represent the pooled effect of obesity

overall meta-analysis. OR, odds ratio; CI, confidence interval; I2, heterogeneity

are partially in agreement with observational studies thatused samples from the general population. Although onelongitudinal study with a 20-year follow-up concluded thatbody weight did not increase the risk of LDD measured us-ing plain X-ray (OR 1.1; 95% CI 0.3–3.6; p5.90) [40], an-other study with a four-year follow-up showed thatoverweight individuals (BMIO25 kg/m2) at the age of 25years had higher risks of developing LDD 4 years later(OR 4.3; 95% CI 1.3–14.3) [10].

In the present review, the effects of familial factors onthe obesity-LDD relationship were only examined in twocross-sectional studies [24,28]. While one study found thatthe heavier MZ twin (at least 8 kg heavier than the co-twin)had a lower prevalence of LDD, the other study [28] foundthe body weight-LDD association to be present in the totalsample and in DZ twins, but the association disappeared inMZ twins at a younger age. It is plausible to suggest thatthe effects of genetic factors on the obesity-LDD relation-ship are stronger earlier in life. Interestingly, the heritabilityof progression of LDD has been found to be mainly influ-enced by genetic factors at younger age [28].

s investigating the relationship between obesity and low back pain. Squares

on low back pain. Weight % represents the influence of each study in the

of studies; MZ, monozygotic twin pairs; DZ, dizygotic twin pairs.

1116 A.B. Dario et al. / The Spine Journal 15 (2015) 1106–1117

Interpretation and implications for clinical practice andresearch

Currently there is uncertainty regarding the signifi-cance of obesity as a risk factor for LBP and LDD.Therefore, a recommendation to intervene to reduce obe-sity for the purpose of reducing LBP is not yet warrantedin clinical guidelines. Inadequate control for familial fac-tors is a possible explanation for conflicting results inthis field. We found evidence that familial factors poten-tially influence the obesity-LBP association. The resultsof this review provide a different perspective on the rela-tionship between obesity and LBP. The identification, infuture research, of specific genes or early shared environ-mental factors (eg, diet, engagement in physical activity)that influence both obesity and LBP might reveal newmechanisms underlying this relationship and could, inturn, lead to effective preventative strategies. We advo-cate that future high-quality longitudinal research, pref-erably using a within-pair twin case-control design, isan ideal method to understand this relationship moreprecisely.

Strengths and limitations

One of the main strengths of this review is the inclusionof twin studies that facilitates insights into causal relation-ships between variables. The case-control analysis oftwin studies allows the investigation of a more direct rela-tionship between obesity and both LBP and LDD bycontrolling for possible genetic and early environmentalconfounding. This design also has the potential to providemore precise estimates of risks for a disease. The inclusionof a dose-response analysis was a unique feature in our re-view, which provides further insights into a possible causalrelationship between obesity and LBP. Overall, the qualityof included studies was high (mean: 83.3%) for LBP andLDD. We were also able to provide an efficient summaryof results of twin studies by pooling the estimates of therelationship between obesity-related measures and LBP.Unfortunately, included studies were too heterogeneousfor the measures of LDD and this precluded pooling ofdata. Also, the effect of familial factors was mostly avail-able in cross-sectional studies that assessed LBP as theoutcome.

Conclusion

Although obesity is commonly reported to be a risk fac-tor for LBP, our results do not support a direct causal rela-tionship between obesity and LBP. Genetic factors andearly environment are possible factors influencing this rela-tionship. Further longitudinal studies using the twin designare needed to better understand the complex mechanismsunderlying the association between obesity and LBP andobesity and LDD.

References

[1] Hoy D, Bain C, Williams G, March L, Brooks P, Blyth F, et al. A sys-

tematic review of the global prevalence of low back pain. Arthritis

Rheum 2012;64:2028–37.

[2] Murray CJL, Vos T, Lozano R, Naghavi M, Flaxman AD, Michaud C,

et al. Disability-adjusted life years (DALYs) for 291 diseases and in-

juries in 21 regions, 1990-2010: a systematic analysis for the Global

Burden of Disease Study 2010. Lancet 2012;380:2197–223.

[3] Talmage JB, Hyman MH, Melhorn JM, Ackerman WE, American

Medical A. AMA guides to the evaluation of disease and injury cau-

sation. Chicago, Illinois: American Medical Association, 2013.

[4] Leboeuf-Yde C, Kyvik KO, Bruun NH. Low back pain and lifestyle.

Part II-obesity. Information from a population-based sample of

29,424 twin subjects. Spine 1999;24:779–83; discussion 783–774.

[5] Wright LJ, Schur E, Noonan C, Ahumada S, Buchwald D, Afari N.

Chronic pain, overweight, and obesity: findings from a community-

based twin registry. J Pain 2010;11:628–35.

[6] Low S, Chin MC, Deurenberg-Yap M. Review on epidemic of obe-

sity. Ann Acad Med Singapore 2009;38:57–9.

[7] James PT. Obesity: the worldwide epidemic. Clin Dermatol 2004;22:

276–80.

[8] Guh DP, Zhang W, Bansback N, Amarsi Z, Birmingham CL, Anis AH.

The incidence of co-morbidities related to obesity and overweight: a

systematic review and meta-analysis. BMC Public Health 2009;9:88.

[9] Urquhart DM, Kurniadi I, Triangto K, Wang Y, Wluka AE,

O’Sullivan R, et al. Obesity is associated with reduced disc height

in the lumbar spine but not at the lumbosacral junction. Spine

2014;39:E962–6.

[10] Liuke M, Solovieva S, Lamminen A, Luoma K, Leino-Arjas P,

Luukkonen R, et al. Disc degeneration of the lumbar spine in relation

to overweight. Int J Obes 2005;29:903–8.

[11] Boden SD, Davis DO, Dina TS, Patronas NJ, Wiesel SW. Abnormal

magnetic-resonance scans of the lumbar spine in asymptomatic subjects.

A prospective investigation. J Bone Joint Surg Am 1990;72:403–8.

[12] Jarvik JJ, Hollingworth W, Heagerty P, Haynor DR, Deyo RA. The

longitudinal assessment of imaging and disability of the back (LAID-

Back) study: baseline data. Spine 2001;26:1158–66.

[13] Borenstein DG, O’Mara JW Jr, Boden SD, Lauerman WC,

Jacobson A, Platenberg C, et al. The value of magnetic resonance

imaging of the lumbar spine to predict low-back pain in asympto-

matic subjects: a seven-year follow-up study. J Bone Joint Surg

Am 2001;83-A:1306–11.

[14] Videman T, Batti�eMC, Ripatti S, Gill K,Manninen H, Kaprio J. Deter-

minants of the progression in lumbar degeneration: a 5-year follow-up

study of adult male monozygotic twins. Spine 2006;31:671–8.

[15] Malis C, Rasmussen EL, Poulsen P, Petersen I, Christensen K, Beck-

Nielsen H, et al. Total and regional fat distribution is strongly influ-

enced by genetic factors in young and elderly twins. Obes Res

2005;13:2139–45.

[16] Battie MC, Videman T, Levalahti E, Gill K, Kaprio J. Heritability of

low back pain and the role of disc degeneration. Pain 2007;131:

272–80.

[17] Stroup DF, Berlin JA, Morton SC, Olkin I, Williamson GD,

Rennie D, et al. Meta-analysis of observational studies in epidemiol-

ogy: a proposal for reporting. Meta-analysis of Observational Studies

in Epidemiology (MOOSE) group. JAMA 2000;283:2008–12.

[18] Brian S, Everitt DH. In: Encyclopedia of statistics in behavioral sci-

enceVol. 1, 1st ed. Hoboken, NJ: John Wiley & Sons, 2005.

[19] Pengel LH, Herbert RD, Maher CG, Refshauge KM. Acute low back

pain: systematic review of its prognosis. BMJ 2003;327:323.

[20] Altman DG. Systematic reviews of evaluations of prognostic varia-

bles. BMJ 2001;323:224–8.

[21] von Elm E, Altman DG, Egger M, Pocock SJ, Gøtzsche PC,

Vandenbroucke JP, et al. The strengthening the reporting of observa-

tional studies in epidemiology (STROBE) statement: guidelines for

reporting observational studies. PLoS Med 2007;4:e296.

1117A.B. Dario et al. / The Spine Journal 15 (2015) 1106–1117

[22] Higgins JPT, Green S. Cochrane handbook for systematic reviews of

interventions version 5.0.2 2009. updated September 2009.

[23] Videman T, Gibbons LE, Battie MC. Age- and pathology-specific

measures of disc degeneration. Spine 2008;33:2781–8.

[24] Videman T, Gibbons LE, Kaprio J, Battie MC. Challenging the cu-

mulative injury model: positive effects of greater body mass on disc

degeneration. Spine J 2010;10:26–31.

[25] Battie MC, Videman T, Levalahti E, Gill K, Kaprio J. Genetic and

environmental effects on disc degeneration by phenotype and spinal

level: a multivariate twin study. Spine 2008;33:2801–8.

[26] Livshits G, Popham M, Malkin I, Sambrook PN, MacGregor AJ,

Spector T, et al. Lumbar disc degeneration and genetic factors are

the main risk factors for low back pain in women: the UK Twin Spine

Study. Ann Rheum Dis 2011;70:1740–5.

[27] Videman T, Battie MC, Parent E, Gibbons LE, Vainio P, Kaprio J.

Progression and determinants of quantitative magnetic resonance

imaging measures of lumbar disc degeneration: a five-year follow-

up of adult male monozygotic twins. Spine 2008;33:1484–90.

[28] Williams FM, Popham M, Sambrook PN, Jones AF, Spector TD,

MacGregor AJ. Progression of lumbar disc degeneration over a dec-

ade: a heritability study. Ann Rheum Dis 2011;70:1203–7.

[29] Hestbaek L, Leboeuf-Yde C, Kyvik KO. Are lifestyle-factors in ado-

lescence predictors for adult low back pain? A cross-sectional and pro-

spective study of young twins. BMC Musculoskelet Disord 2006;7:27.

[30] MacGregor AJ, Andrew T, Sambrook PN, Spector TD. Structural,

psychological, and genetic influences on low back and neck pain: a

study of adult female twins. Arthritis Rheum 2004;51:160–7.

[31] Medicaid Access Study G. Access of Medicaid recipients to outpa-

tient care. N Engl J Med 1994;330:1426–30.

[32] Sambrook PN, MacGregor AJ, Spector TD. Genetic influences on

cervical and lumbar disc degeneration: a magnetic resonance imaging

study in twins. Arthritis Rheum 1999;42:366–72.

[33] Flamme CH. Obesity and low back pain—biology, biomechanics and

epidemiology. Orthopade 2005;34:652–7.

[34] Ferreira PH, Beckenkamp P, Maher CG, Hopper JL, Ferreira ML. Na-

ture or nurture in low back pain? Results of a systematic review of

studies based on twin samples. Eur J Pain 2013;17:957–71.

[35] Webb P, Bain C. Essential epidemiology. Cambridge University

Press, 2010.

[36] Shiri R, Karppinen J, Leino-Arjas P, Solovieva S, Viikari-Juntura E.

The association between obesity and low back pain: a meta-analysis.

Am J Epidemiol 2010;171:135–54.

[37] Biering-Sorensen F. Physical measurements as risk indicators for

low-back trouble over a one-year period. Spine 1984;9:106–19.

[38] Symmons DP, van Hemert AM, Vandenbroucke JP, Valkenburg HA.

A longitudinal study of back pain and radiological changes in the

lumbar spines of middle aged women. I. Clinical findings. Ann

Rheum Dis 1991;50:158–61.

[39] Lake JK, Power C, Cole TJ. Back pain and obesity in the 1958 British

birth cohort: cause or effect? J Clin Epidemiol 2000;53:245–50.

[40] Nemoto O, Kitada A, Naitou S, Tsuda Y, Matsukawa K, Ukegawa Y.

A longitudinal study for incidence of low back pain and radiological

changes of lumbar spine in asymptomatic Japanese military young

adults. Eur Spine J 2013;22:453–8.