Review Article

Bullying Victimization at School and Headache:A Meta-Analysis of Observational Studies

Gianluca Gini, PhD; Tiziana Pozzoli, PhD; Michela Lenzi, PhD; Alessio Vieno, PhD

Background and objectives.—Being bullied at school is a risk factor for a variety of negative consequences, includingsomatic problems. The purpose of this meta-analysis is to determine the association between peer victimization and headachein the school-age population.

Methods.—A systematic literature search was conducted in September 2013 to identify observational studies that exam-ined the association between being bullied and headache in children and adolescents. Odds ratios (OR) were pooled by usinga random-effects model. Moderator and sensitivity analyses were conducted.

Results.—Twenty studies, including a total of 173,775 participants, satisfied the pre-stated inclusion criteria. Fourteenstudies reported data on the prevalence of headache, which was on average 32.7% (range: 9.1-71.7%) in the bullied group and19.1% (range: 5.3-46.1%) in the control group. Two separate meta-analyses of the association between being bullied andheadache were performed on 3 longitudinal studies (OR = 2.10, 95% confidence interval = 1.19-3.71) and 17 cross-sectionalstudies (OR = 2.00, 95% confidence interval = 1.70-2.35), respectively. Results showed that bullied children and adolescentshave a significantly higher risk for headache compared with non-bullied peers. In the cross-sectional studies, the magnitude ofeffect size significantly decreased with the increase of the proportion of female participants in the study sample. No furthermoderators were statistically significant.

Conclusions.—The positive association between bullying victimization and headache was confirmed. Further research onthe environmental factors that may influence this symptom is needed.

Key words: bullying, victimization, headache, physical health, meta-analysis

Abbreviations: 95% CI 95% confidence interval, MOOSE Meta-analysis of Observational Studies in Epidemiology, OR oddsratio, TTH tension-type headache, SES socioeconomic status

(Headache 2014;54:976-986)

Recurrent headache is the most frequent neuro-logical symptom during school age and one of themost frequent manifestations of pain in childhoodand adolescence.1-3 A recent systematic review4

showed that headache is very common across theworld with about 60% of children and adolescentsreporting this symptom over at least a 3-month

period. Moreover, epidemiological studies pointedout that the prevalence of headache has increasedover the last decades in the school-age population.5-8

Quite recently, studies on the potential risk factors foryouth’s headache have drawn attention to the role ofpsychological and social factors, including negativeexperiences at school.9-12 For example, stressors in theschool environment, such as schoolwork pressure,13

negative feelings about school,14,15 perception ofbeing treated badly or unfairly by teachers,11-13 fear offailure,16 and harassment by peers13,16 turned out to be

From the Department of Developmental and Social Psychol-ogy, University of Padua, Padua, Italy.

Address all correspondence to G. Gini, Department of Devel-opmental and Social Psychology, via Venezia 8, 35131 Padova,Italy.

Accepted for publication January 20, 2014.

Conflict of Interest: None.

Funding Source: No external funding was secured for this study.

ISSN 0017-8748doi: 10.1111/head.12344

Published by Wiley Periodicals, Inc.Headache© 2014 American Headache Society

976

associated with higher levels of headache in childrenand adolescents.

A serious and frequent source of concern in chil-dren’s and adolescents’ school life is bullying, that is, arepetitive physical or psychological abuse by a stron-ger schoolmate or group on a weaker peer.17,18 Epide-miological studies across countries indicate that10-20% of students are frequently bullied by school-mates.18,19 Importantly, pediatric and psychologicalresearch is increasingly demonstrating the adverseconsequences of being bullied at school for child-ren’s and adolescents’ psychosocial adjustment,20,21

health,22,23 and medicine use.24 The well-establishedcognitive appraisal model of stress and coping devel-oped by Lazarus25 and the “sustained activationhypothesis”26 can help to understand stress reactionsto bullying.These models suggest that repeated bully-ing experiences in children’s life might cause a state ofemotional distress that can lead to adverse healthoutcomes such as recurrent headache.

To date, 2 meta-analyses22,23 have shown thatbullied students can be affected by poor physicalhealth and that these youths are about 2 times morelikely than non-bullied agemates to report a variety ofsymptoms, such as headache, backache, abdominalpain, skin problems, vomiting, etc. However, boththese meta-analytic reviews only reported an overallrisk estimate for victims’ health problems and did notspecifically focus on headache. The current meta-analysis aims at (1) estimating the risk for headachein children and adolescents who are bullied by peers(ie, victims) compared with non-bullied peers; (2) per-forming separate meta-analyses of longitudinal andcross-sectional studies; (3) testing for potential mod-erators of variation in the magnitude of effect sizes,that is, testing whether certain study features explaindifferences in the strength of the effect sizes.

METHODSLiterature Search.—Several methods were used to

identify relevant studies. First, electronic searches inPsychInfo, Pubmed, EMBASE, the Cochrane Librarydatabase, the Campbell Collaboration database, andScopus were conducted in September 2013 with thefollowing keywords: “bullying,” or “peer victimiza-tion” and “headache,” “somatic,” “psychosomatic,”

and “physical health.” Second, the “cited by” functionin Scopus was used to retrieve empirical articles thathave cited the 2 meta-analyses22,23 on the associationbetween bullying and health problems. Third, previ-ous issues of the journal “Headache” were searchedfor relevant studies. Finally, review articles regardingconsequences of bullying and the reference sectionsof the collected articles were reviewed for possiblerelevant citations. If a study was not available in full-text, the corresponding author was contacted. Thismeta-analysis was planned, conducted, and reportedin adherence to the Meta-analysis of ObservationalStudies in Epidemiology (MOOSE) guidelines.27

Inclusion Criteria.—A study had to meet thefollowing a priori criteria to be included. The mostbasic requirement was the inclusion of measures ofbullying victimization at school in childhood oradolescence and of headache. Consistent with theinternational literature, bullying was defined as adeliberate, repeated exposure to aggressive acts per-formed by a peer or a group of peers with higherpower or strength than the bullied schoolmate (ie, the“victim”).17,18 These measures could include (1) self-report questionnaires; (2) peer or adult reports; or (3)an interview that resulted in a quantitative ratingof peer victimization and headache frequency. Themajority of studies defined frequent headache asoccurring “about every week/at least weekly,” withfew others using a slightly broader definition (“atleast monthly/several times in the last 12 months”).Second, studies were required to have reported effectsizes and related confidence intervals or enoughinformation to calculate these data – for example, byreporting comparisons between bullied children anda control group (defined as children from the samepopulation of victims who were classified as notbullied). Both cross-sectional and longitudinal studieswere included. We excluded the following types ofstudies: studies that did not include a control group;studies that measured headache with items includedin a larger scale, as this problem could not be clearlydistinguished from other symptoms; studies withduplicated data; studies that did not report analyseson the variables of interest; and studies with adults orpsychiatric patients. Two authors (GG, TP) indepen-dently assessed whether articles met the inclusion

Headache 977

criteria. In the case of disagreement, a consensus wasreached through discussion.

Coding of Studies.—Studies were coded on design(cross-sectional vs longitudinal), length of follow upfor longitudinal studies, type of bullying and of symp-toms measure (self-report questionnaire vs peer/adultreports vs interview), confounding variables (eg, age,gender), type of sampling procedure, sample compo-sition and characteristics, and geographical locationof study. Two authors (GG, TP) independently codedthe studies. Quantitative data were extracted fromtext and tables; for the sake of comparability withthe results of the former meta-analyses,22,23 the dataadjusted for confounders were preferred.

Statistical Analyses.—Analyses were done usingComprehensive Meta-Analysis.28 We extracted oddsratio (OR) and their 95% confidence interval (CI)from each study.With very few exceptions, studies didnot report results for boys and girls separately; there-fore, we were not able to compare effect sizes bygender group. Because most of the studies reportedthe proportion of girls in the sample, we used thisinformation to test for possible moderation by gendercomposition of the sample.

Data from individual studies were pooled using arandom-effects model. Each study was weighted bythe inverse of its variance, which, under the random-effects model, includes the within-study variance plusthe between-studies variance tau-squared (Τ2). The Zstatistic was calculated, and a 2-tailed P value of lessthan .05 was considered to indicate statistical signifi-cance. Statistical heterogeneity was assessed using theQ statistic to evaluate whether the pooled studiesrepresent a homogeneous distribution of effect sizes.Also reported is the I2 statistic, indicating the propor-tion of observed variance that reflects real differencesin effect size.29

To address the possible“publication bias”– that is,the fact that studies with non-significant results are lesslikely to be published – we computed the “fail-safe N”(Nfs) according to the method that Orwin30 proposed,which is more conservative than the traditionalRosenthal’s Nfs.31,32 Orwin’s Nfs determines thenumber of additional studies in a meta-analysis yield-ing null effect sizes that would be needed to yield a“trivial” OR of 1.05. Researchers suggest that meta-

analysts calculate a tolerance level around a fail-safe Nthat is equal to 5 times the number of effects includedin the meta-analysis plus 10 (the “5k + 10” bench-mark).32,33 Moreover, the association between thestandardized effect sizes and the variances of theseeffects was analyzed by rank correlation with use ofthe Kendall tau method. If small studies with negativeresults were less likely to be published, the correlationbetween variance and effect size would be high. Con-versely, a lack of a significant correlation can be inter-preted as the absence of publication bias.34

RESULTSStudy Identification and Characteristics.—After

the removal of duplicates, a list of 137 potentiallyeligible studies was generated (Fig. 1). Based ontitles and abstracts, 57 articles were excluded at thefirst screening because they were qualitative studies,reviews or commentaries, or studies that did notmeasure school bullying. Seven studies35-41 were notavailable in full text. Full-text copies of the remaining73 potentially relevant studies were obtained. Thirty-seven studies were excluded because they did notmeet the inclusion criteria (eg, they did not have acontrol group). Fifteen studies did not report enoughdata to compute effect sizes or confidence intervals.As a result, the remaining 20 studies were includedfor this meta-analysis.Three studies were longitudinalstudies, and 17 employed a cross-sectional design.

The Table summarizes the characteristics of thestudies included in this meta-analysis, includingsample-size and response rate, age and gender compo-sition of the sample, type of measures, study design,and type of sampling. A total of 173,775 children andadolescents participated in the 20 studies. Across the17 studies that provided information about the sam-ple’s gender composition, 51.3% (range: 32.8-62.4%)of the participants were girls. Fourteen studiesreported data on the prevalence of headache, whichwas on average 32.7% (range: 9.1-71.7%) in thebullied group and 19.1% (range: 5.3-46.1%) in thecontrol group.

Five studies were from Norway,42-45 2 of whichwere from the same publication; 2 respectivelyfrom India,46,47 the Netherlands,48,49 Turkey,50,51 andthe United States;52,53 and 1 respectively from

978 June 2014

China,54 Finland,44 Greenland,55 Italy,56 the UnitedKingdom,57 and Russia.58 One article reported datafrom multiple countries.18 Information about race/ethnicity and socioeconomic status (SES) of the par-ticipants was not systematically reported in allstudies. Overall, the heterogeneity of racial and SESclassification within and across the studies was suchthat it precluded any analysis by race/ethnicity orSES.

Meta-Analysis of Longitudinal Studies.—Threestudies used a longitudinal design. The follow-upduration ranged between 9 months and 11 years.Across the 3 samples, bullied children were foundto have a significantly higher risk for headachethan non-bullied agemates were (OR = 2.10, 95%CI = 1.19-3.71, Z = 2.57, P = .01). Figure 2 shows theforest plot for this meta-analysis. Studies were notcompletely homogeneous (Q = 4.11, P = .13,I2 = 51.37%).

Meta-Analysis of Cross-Sectional Studies.—Across the 17 samples that were included in the cross-sectional studies, bullied children were found to have

a significantly higher risk for headache than werenon-bullied peers (OR = 2.00, 95% CI = 1.70-2.35,Z = 8.43, P < .001). Figure 3 shows the forest plot forthis meta-analysis. Effect sizes within this group ofstudies were not homogeneous (Q = 65.64, P < .001,I2 = 75.63%).

Moderator Analyses.—Moderator analyses withgender composition of the sample, number of con-founders, and geographical location were performedto explore possible explanations for heterogeneity inthe effect sizes across cross-sectional studies.The pro-portion of girls in the sample was available for 15 outof the 17 cross-sectional studies, and it was used as acontinuous predictor in a weighted mixed-effectsmeta-regression. Results indicated that the magni-tude of the effect size significantly decreased withthe increase of the number of female participantsin the study sample (B = −.06, 95% CI: −.07 to −.04,P < .001). Conversely, the number of confoundersconsidered in the study (range: 0-6) did not moderatethe magnitude of the effect (B = .005, 95% CI: −.04 to.05, P = .82). Also the study’s geographical location

Records identified through database searching (n = 153)

Scre

enin

g In

clud

ed

Elig

ibili

ty

Iden

tifi

cati

on

Records after duplicates removed (n = 137)

Records screened (n = 137)

Records excluded by title/abstract screened

(reviews, commentaries, qualitative studies, studies not on bullying) (n = 57), full-text

not available (n = 7)

Full-text articles assessed for eligibility

(n = 73)

Full-text articles excluded, not meeting inclusion criteria (n = 37), not including enough data

to calculate ES (n = 15)

Studies included in meta-analysis (n = 20)

Fig 1.—Flow diagram of study inclusion. ES = effect size.

Headache 979

Tabl

e.—

Cha

ract

eris

tics

ofth

eSt

udie

sIn

clud

edin

the

Met

a-A

naly

sis

Firs

tAut

hor

(Yea

rof

Pub

licat

ion)

Sam

ple

Size

(Res

pons

eR

ate)

Age

Ran

ge(%

ofG

irls

)B

ully

ing

Mea

sure

Hea

dach

eM

easu

reA

djus

tmen

tfo

rC

onfo

unde

rsSt

udy

Des

ign

Type

ofSa

mpl

ing

Ars

lan

etal

(201

2)50

1315

(97%

)11

-19

(53.

4)Se

lf-r

epor

tqu

esti

onna

ire

Self

-rep

ort

ques

tion

nair

eSE

S,sc

hool

grad

e,ge

nder

Cro

ss-s

ecti

onal

Clu

ster

rand

omsa

mpl

ing

Bie

blet

al(2

011)

5265

(n/a

)12

-20

atti

me

3(5

2.9)

Tim

e1:

play

sess

ion;

tim

e3:

self

-rep

ort

ques

tion

nair

eSe

lf-r

epor

tqu

esti

onna

ire

Gen

der

Lon

gitu

dina

lC

onve

nien

cesa

mpl

e

Due

etal

(200

5)18

123,

227

(>90

%)

11-1

5(5

1)Se

lf-r

epor

tqu

esti

onna

ire

Self

-rep

ort

ques

tion

nair

eA

ge,f

amily

afflu

ence

,co

untr

yC

ross

-sec

tion

alC

lust

erra

ndom

sam

plin

g

Fekk

eset

al(2

004)

4827

66(1

00%

)9-

12(5

0)Se

lf-r

epor

tqu

esti

onna

ire

Self

-rep

ort

ques

tion

nair

eG

ende

rC

ross

-sec

tion

alU

nkno

wn

Fekk

eset

al(2

006)

4911

18(7

0%)

9-11

(50.

3)Se

lf-r

epor

tqu

esti

onna

ire

Self

-rep

ort

ques

tion

nair

eG

ende

r,ag

e,ha

ving

frie

nds

Lon

gitu

dina

lU

nkno

wn

Gin

i(20

08)56

565

(94%

)8-

11(5

2.9)

Self

-rep

ort

ques

tion

nair

eSe

lf-r

epor

tqu

esti

onna

ire

Gen

der,

age

Cro

ss-s

ecti

onal

Sim

ple

rand

omsa

mpl

ing

Haa

vet

etal

(200

4)42

8316

(88%

)15

(54.

4)Se

lf-r

epor

tqu

esti

onna

ire

Self

-rep

ort

ques

tion

nair

eN

one

Cro

ss-s

ecti

onal

Popu

lati

onst

udy

Hes

keth

(201

0)54

2191

(80%

)9-

12(4

4)Se

lf-r

epor

tqu

esti

onna

ire

Self

-rep

ort

ques

tion

nair

eG

ende

r,ag

e,re

side

nce,

pare

ntal

educ

atio

nC

ross

-sec

tion

alSi

mpl

era

ndom

sam

plin

g

Kar

atas

and

Ozt

urk

(201

1)51

92(8

2%)

10-1

2(5

1.1)

Self

-rep

ort

ques

tion

nair

eP

aren

t-re

port

ques

tion

nair

eN

one

Cro

ss-s

ecti

onal

Sim

ple

rand

omsa

mpl

ing

Ksh

irsa

gar

(200

7)46

500

(100

%)

8-12

(62.

4)Se

mi-

stru

ctur

edin

terv

iew

Sem

i-st

ruct

ured

inte

rvie

wN

otsp

ecifi

edL

ongi

tudi

nal

Sim

ple

rand

omsa

mpl

ing

Lie

net

al(2

009)

(sam

ple

1)67

3790

(88%

)15

-16

(49.

3)Se

lf-r

epor

tqu

esti

onna

ire

Self

-rep

ort

ques

tion

nair

eFa

mily

stru

ctur

e,SE

S,et

hnic

ity,

expo

sure

tovi

olen

ce,h

avin

gcl

ose

frie

nds

Cro

ss-s

ecti

onal

Popu

lati

onst

udy

Lie

net

al(2

009)

(sam

ple

2)67

3790

(80%

)18

-19

(55.

9)Se

lf-r

epor

tqu

esti

onna

ire

Self

-rep

ort

ques

tion

nair

eFa

mily

stru

ctur

e,SE

S,et

hnic

ity,

expo

sure

tovi

olen

ce,h

avin

gcl

ose

frie

nds

Cro

ss-s

ecti

onal

Popu

lati

onst

udy

Løh

reet

al(2

011)

4341

9(1

00%

)7-

16(n

/a)

Mul

ti-i

nfor

man

t(se

lf-,

teac

her-

,par

ent-

repo

rts)

Self

-rep

ort

ques

tion

nair

eG

ende

r,gr

ade

Cro

ss-s

ecti

onal

Con

veni

ence

sam

ple

Lun

tam

oet

al(2

012)

4422

15(9

1%)

13-1

8(5

0)Se

lf-r

epor

tqu

esti

onna

ire

Self

-rep

ort

ques

tion

nair

eN

one

Cro

ss-s

ecti

onal

Popu

lati

onst

udy

Nat

vig

etal

(200

1)45

856

(83.

7%)

13-1

5(5

0.6)

Self

-rep

ort

ques

tion

nair

eSe

lf-r

epor

tqu

esti

onna

ire

Gen

der,

age,

scho

olC

ross

-sec

tion

alU

nkno

wn

Ram

yaan

dK

ulka

rni(

2011

)4750

0(n

/a)

8-14

(32.

8)In

terv

iew

Inte

rvie

wN

one

Cro

ss-s

ecti

onal

Sim

ple

rand

omsa

mpl

ing

Schn

ohr

and

Nic

lase

n(2

006)

5589

1(n

/a)

11-1

5(n

/a)

Self

-rep

ort

ques

tion

nair

eSe

lf-r

epor

tqu

esti

onna

ire

Gen

der,

age

Cro

ss-s

ecti

onal

Clu

ster

rand

omsa

mpl

ing

Srab

stei

net

al(2

006)

5315

,305

(83%

)11

-15

(53.

5)Se

lf-r

epor

tqu

esti

onna

ire

Self

-rep

ort

ques

tion

nair

eG

ende

r,ag

e,ra

ce,

over

wei

ght/

obes

ity,

mat

erna

ledu

cati

on

Cro

ss-s

ecti

onal

Clu

ster

rand

omsa

mpl

ing

Stic

kley

etal

(201

3)58

2892

(96%

)13

-17

(57.

6%)

Self

-rep

ort

ques

tion

nair

eSe

lf-r

epor

tqu

esti

onna

ire

Age

,par

enta

led

ucat

ion,

fam

ilyst

ruct

ure

Cro

ss-s

ecti

onal

Sim

ple

rand

omsa

mpl

ing

Will

iam

set

al(1

996)

5729

62(9

3.1%

)7-

10(n

/a)

Sem

i-st

ruct

ured

inte

rvie

wSe

mi-

stru

ctur

edin

terv

iew

Not

spec

ified

Cro

ss-s

ecti

onal

Popu

lati

ons

stud

y

SES,

soci

oeco

nom

icst

atus

.

980 June 2014

(coded as Europe vs other countries) was not asignificant moderator (k = 11, OR = 2.03, 95% CI:1.59-2.60, and k = 5, OR = 2.00, 95% CI: 1.32-3.02,respectively; Q = .48, P = .79).

Sensitivity Analysis.—Finally, consistent with theMOOSE guidelines27 and to the former meta-analyses,22,23 a sensitivity analysis was performedbased on 2 aspects of study quality (beyond thoserequired as inclusion criteria): (1) the use of a ran-domized sampling design or a whole population of

students; and (2) a good response rate (>80%). Thir-teen cross-sectional studies satisfied both criteria. Wethen performed a separate meta-analysis of this sub-group of studies, and the resulting OR and confidenceinterval was OR = 1.90, 95% CI = 1.61-2.25.

Another sensitivity analysis was performed withthe 13 studies that used only self-report question-naires to gather data from participants. Estimated ORwas 1.87, with a 95% CI ranging from 1.57 to 2.23.

Publication Bias.—No evidence of publication biaswas present. Kendall’s tau was 0.13 with 2-tailedP = .44. An additional 253 studies with null effectsizes would be needed to attenuate the effect size to anegligible value (“5k + 10” benchmark = 110).

DISCUSSIONThe results of this meta-analysis confirmed that

bullied youths are about twice more likely thannon-bullied agemates to suffer from headache. Sameresults were found both in longitudinal and cross-sectional studies. The sensitivity analysis, whichremoved those studies with potential to introducebias, revealed similar results. These findings con-

Fig 2.—Forest plot for random-effects meta-analysis of theassociation between being bullied and headache: longitudinalstudies. Note. Effect sizes are expressed as odds ratios. Studiesare represented by symbols whose area is proportional to thestudy’s weight in the analysis. CI = confidence interval.

Fig 3.—Forest plot for random-effects meta-analysis of the association between being bullied and headache: cross-sectional studies.Note. Effect sizes are expressed as odds ratios. Studies are represented by symbols whose area is proportional to the study’s weightin the analysis. CI = confidence interval.

Headache 981

firm and complement those of 2 previous meta-analyses22,23 on the health consequences of peer vic-timization.The present study differs from those meta-analyses in important ways. First, only this reviewreports meta-analytic effects specifically for the asso-ciation between bullying victimization and headache.Second, it includes 11 new studies that were notincluded in the first meta-analysis (% of overlap ofthe 2 meta-analytic databases: 40.9%), and 2 newstudies that were not included in the latter one (% ofoverlap of the 2 meta-analytic databases: 42.8%).Third, another moderator that was not consideredbefore was tested (ie, number of confounders).Finally, only in this work a sensitivity analysiswith studies that used self-report measures wasperformed.

Similarly to what have been reported on a formermeta-analysis,23 a meta-regression analysis showedthat the strength of the relationship between beingbullied and suffering from headache is higher whensamples contain proportionally more boys.A possibleexplanation of this finding might deal with the factthat a school/classroom environment with a higherproportion of male students is a social context inwhich bullying behavior is more likely to happen, andwhere supportive and helping behaviors in favor ofthe bullied pupils are less frequent.59 This couldincrease the negative impact of being bullied on chil-dren’s health status. However, given the explorativenature of the moderator analysis, a significant findingis not to be considered definitive, but it does suggest adirection for additional research. The influence of theschool environment’s gender composition on peervictimization and its consequences on children’s well-being is a topic that warrants further research.

The current meta-analytic findings confirm forheadache what have been reported for health prob-lems in general. A possible explanation of this signifi-cant association may be related to the well-knownpsychological vulnerability of bullied students, giventhat research has shown a strong association betweenemotional difficulties and somatic complaints.60 Thefeelings associated with being bullied represent aform of social pain, a term used to describe the feel-ings of pain that follow the experiences of peer rejec-tion, ostracism, or loss.61 Neuroscience research is

increasingly showing that social pain is experienced ina similar way to physical pain, at least as far as thebrain is concerned. Indeed, recent studies have shownthat social pain and physical pain rely on similarneurobiological and neural substrates and are expe-rienced physiologically in a similar manner.61,62 Simi-larly, peer victimization is linked to dysregulation ofthe hypothalamic-pituitary-adrenal axis, the body’sstress response system.63-65 From a physiological per-spective, peer victimization represents a relativelyextreme and/or persistent stressor, which ultimatelyleads to lower cortisol levels compared with that ofnon-bullied peers.64,65

Of course, other psychological characteristics ofbullied youth may influence the relationship betweenbullying and health problems. For example, one mayhypothesize that students who lack adequate copingskills, as well as have low self-esteem or lack asser-tiveness, in front of victimization experiences are atincreased risk for negative outcomes compared topeers who possess more developed psychological andsocial competencies. This is certainly an interestinghypothesis that should be tested in future longitudi-nal studies.

Strengths and Limitations.—This is the first meta-analytic study that estimated the relationshipbetween being bullied and headache. Strengths of thismeta-analysis include the large overall sample sizeand the wide geographic distribution of the samples,which support the generalizability of the overall find-ings. Moreover, the large majority of the studiesincluded in the meta-analysis were characterized bygood methodological quality, as defined, for example,by the use of a random sampling design. Furthermore,we did not find evidence of publication bias that mayhave led to overestimating the association betweenbullying experiences and headache. Finally, we wereable to perform separate meta-analyses of longitudi-nal and cross-sectional studies, which yielded thesame results, even though the lack of large longitudi-nal studies is still a limit of the literature in thisfield.

The results of this meta-analysis should be inter-preted in the context of the study limitations.The fact that the available studies neither explicitlycompared male and female samples, nor reported

982 June 2014

separate effects for different ethnic groups limitedthe possibility for more detailed analyses. In particu-lar, youths’ cultural background could influence howbullying victimization is experienced, as well as theirability to cope with it and the negative consequencesthat may arise from this socially adverse experience.Moreover, much variability exists in the methodsand instruments used to assess the prevalence ofheadache and peer victimization experiences. Themajority of studies used a variety of self-reportquestionnaires, both for peer victimization and forchildren’s health complaints. In some cases, thesemeasures were reduced to a single-item question-naire. Self-report measures are very common in bul-lying research and are usually considered to be validand reliable.66 However, possible problems withthese instruments are that they require a good levelof respondents’ self-consciousness and that somebullied children may tend to deny their condition. Toavoid these problems, future studies should collectinformation about youths’ bullying experiencesthrough multiple independent informants, such aschildren themselves, their peers within the class, andtheir teachers or parents. Moreover, future investi-gations should also cover the issue of cyberbullyingvictimization, that is, victimization experiencedthrough information technologies (ie, cell phones,Internet). Also, the assessment of children’s healthcomplaints must be improved. For example, none ofthe available studies included independent objectiveinformation, such as children’s school absenteeismextracted from school attendance records or theirvisits to the school nurse office; further improvementon the accuracy of headache reports in these agegroups would profit from the use of prospectivemeasurement in diaries, instead of only retrospectiverecalls. Moreover, studies in this field do not reportinformation on the type of headache (migraine vstension-type headache [TTH]) suffered by bulliedyouth. It is important that future research worksaddress this limitation by comparing the specificeffects of bullying as a stressor on both migraine andTTH. Finally, our meta-analysis shares the samelimitations of all meta-analyses of observationalstudies. Because individuals cannot be randomlyallocated to groups, the influence of confounding

variables cannot be fully evaluated. Although manystudies controlled for important confounding vari-ables, such as parental education and SES, otherunknown confounders could be partially responsiblefor the effect observed.

CONCLUSIONBullied youths are about 2 times more likely than

non-bullied agemates to report frequent headache.This meta-analysis complements the growing body ofresearch that documents the poor personal adjust-ment of bullied children and adolescents, in terms ofboth internalizing and externalizing problems, whichother recent meta-analyses12,13 on the psychosocialconsequences of peer victimization have summarized.It is important that pediatricians, school nurses, andother professionals be ready to identify children whoare at risk of being bullied at school because thepotential negative health, psychological, and educa-tional consequences of bullying experiences are farreaching.

STATEMENT OF AUTHORSHIP

Category 1(a) Conception and Design

Gianluca Gini; Tiziana Pozzoli; Michela Lenzi;Alessio Vieno

(b) Acquisition of DataGianluca Gini; Tiziana Pozzoli

(c) Analysis and Interpretation of DataGianluca Gini; Tiziana Pozzoli

Category 2(a) Drafting the Manuscript

Gianluca Gini; Tiziana Pozzoli; Michela Lenzi;Alessio Vieno

(b) Revising It for Intellectual ContentGianluca Gini; Tiziana Pozzoli; Michela Lenzi;Alessio Vieno

Category 3(a) Final Approval of the Completed Manuscript

Gianluca Gini; Tiziana Pozzoli; Michela Lenzi;Alessio Vieno

Headache 983

REFERENCES

1. Kroner-Herwig B, Gassmann L, van Gessel H, VathN. Multiple pains in children and adolescents: A riskfactor analysis in a longitudinal study. J PediatrPsychol. 2011;36:420-432.

2. Fichtel A, Larsson B. Psychosocial impact of head-ache and comorbidity with other pains amongSwedish school adolescents. Headache. 2002;42:766-775.

3. Perquin CW, Hazebroek-Kampschreur AAJM,Hunfeld JAM, van Suijlekom-Smit LWA, PasschierJ, van der Wouden JC. Chronic pain among childrenand adolescents: Physician consultation and medica-tion use. Clin J Pain. 2000;16:229-235.

4. Abu-Arafeh I, Razak S, Sivaraman B, Graham C.Prevalence of headache and migraine in childrenand adolescents: A systematic review of population-based studies. Dev Med Child Neurol. 2010;52:1088-1097.

5. Rhee H. Prevalence and predictors of headaches inUS adolescents. Headache. 2000;40:528-538.

6. Bandell-Hoekstra IE, Abu-Saad HH, Passchier J,Frederiks CM, Feron FJ, Knipschild P. Prevalenceand characteristics of headache in Dutch school-children. Eur J Pain. 2001;5:145-153.

7. Laurell K, Larsson B, Eeg-Olofsson O. Prevalenceof headache in Swedish schoolchildren, with a focuson tension-type headache. Cephalalgia. 2004;24:380-388.

8. Sillanpaa M, Anttila P. Increasing prevalence ofheadache in 7-year-old schoolchildren. Headache.1996;36:466-470.

9. Torsheim T, Wold B. School-related stress, schoolsupport, and somatic complaints: A general popula-tion study. J Adolesc Res. 2001;16:293-303.

10. Nicholson RA, Houle TT, Rhudy JL, Norton PJ.Psychological risk factors in headache. Headache.2007;47:413-426.

11. Santinello M, Vieno A, De Vogli R. Primary head-ache in Italian early adolescents: The role of per-ceived teacher unfairness. Headache. 2009;49:366-374.

12. Lenzi M, Vieno A, De Vogli R, et al. Perceivedteacher unfairness and primary headache in earlyadolescence: A cross-national comparison. Int JPublic Health. 2013;58:227-235.

13. Hjern A, Alfven G, Östberg V. School stressors, psy-chological complaints and psychosomatic pain. ActaPaediatr. 2008;97:112-117.

14. Gordon KE, Dooley JM, Wood EP. Self-reportedheadache frequency and features associated withfrequent headaches in Canadian young adolescents.Headache. 2004;44:555-561.

15. Metsähonkala L, Sillanpää M, Tuominen J. Socialenvironment and headache in 8- to 9-year-oldchildren: A follow-up study. Headache. 1998;38:222-228.

16. Kröner-Herwig B, Morris L, Heinrich M. Bio-psychosocial correlates of headache: What predictspediatric headache occurrence? Headache. 2008;48:529-544.

17. Olweus D. Bullying at School: What We Know, WhatWe Can Do. Oxford, UK: Blackwell; 1993.

18. Due P, Holstein BE, Lynch J, et al. Bullying andsymptoms among school-aged children: Interna-tional comparative cross sectional study in 28 coun-tries. Eur J Public Health. 2005;15:128-132.

19. Vieno A, Gini G, Santinello M. Different forms ofbullying and their association to smoking and drink-ing behavior in Italian adolescents. J Sch Health.2011;81:389-395.

20. Reijntjes AHA, Kamphuis JH, Prinzie P, Telch MJ.Peer victimization and internalizing problems inchildren: A meta-analysis of longitudinal studies.Child Abuse Negl. 2010;34:244-252.

21. Ttofi MM, Farrington DP, Losel F, Loeber R. Do thevictims of school bullies tend to become depressedlater in life? A systematic review and meta-analysisof longitudinal studies. J Aggress Confl Peace Res.2011;3:63-73.

22. Gini G, Pozzoli T. Association between bullying andpsychosomatic problems: A meta-analysis. Pediat-rics. 2009;123:1059-1065.

23. Gini G, Pozzoli T. Bullied children and psychoso-matic problems: A meta-analysis. Pediatrics. 2013;132:720-729.

24. Due P, Hansen EH, Merlo J, Andersen A, HolsteinBE. Is victimization from bullying associated withmedicine use among adolescents? A nationallyrepresentative cross-sectional survey in Denmark.Pediatrics. 2007;120:110-117.

25. Lazarus RS. Psychological Stress and the CopingProcess. New York: McGraw-Hill; 1966.

26. Ursin H. Sensitization, somatization, and subjectivehealth complaints. Int J Behav Med. 1997;4:105-116.

27. Stroup DF, Berlin JA, Morton SC, et al. Meta-analysis of observational studies in epidemiology: A

984 June 2014

proposal for reporting. Meta-analysis Of Observa-tional Studies in Epidemiology (MOOSE) group.JAMA. 2000;283:2008-2012.

28. Borenstein M, Hedges L, Higgins J, Rothstein H.Comprehensive Meta-Analysis [computer pro-gram]. Version 2.2 Englewood, NJ: Biostat; 2006.

29. Higgins J, Thompson S, Deeks J, Altman D. Measur-ing inconsistency in meta-analyses. BMJ. 2003;327:557-560.

30. Orwin RG. A fail-safe N for effect size in meta-analysis. J Educ Stat. 1983;8:157-159.

31. Rosenthal R. Combining results of independentstudies. Psychol Bull. 1978;85:185-193.

32. Rosenthal R. The file drawer problem and tolerancefor null results. Psychol Bull. 1979;86:638-641.

33. Muellerleile P, Mullen B. Sufficiency and stability ofevidence for public health interventions using cumu-lative meta-analysis. Am J Public Health. 2006;96:515-522.

34. Begg CB, Mazumdar M. Operating characteristics ofa rank correlation test for publication bias. Biomet-rics. 1994;50:1088-1101.

35. Brandibas G, Jeunier B, Stilhart C. Somatic and psy-chological consequences of bullying between pupils.Revue Francophone Stress Trauma. 2007;7:143-151.

36. Eder A. Risk factor loneliness. On the interrelationsbetween social integration, happiness and health in11-, 13- and 15-year old schoolchildren in 9 Euro-pean countries. Health Promot Int. 1990;5:19-33.

37. Kim YS. School bullying among Korean adolescentsin a prospective follow-up study:Time trends, trajec-tories, predictors and adverse outcomes of bullying.Diss Abstr Int. 2006;66(8-B):4180.

38. Mazur J, Woynarowska B. Bullying at school andsatisfaction from health and life in pupils aged 11-15[Przemoc w szkole a zdrowie i zadowolenie z zyciauczniów 11-15-letnich]. Pediatr Pol. 2004;79:707-715.

39. Suzuki S, Suzuki S. Counseling for a preparatoryschool (Yobiko) student who had been bullied:From the viewpoint of empowerment. Jpn J CounsSci. 2006;39:49-58.

40. Vatn AS, Bjertness E, Lien L. Bullying and healthcomplaints in children and adolescents [Mobbing oghelseplager hos barn og ungdom]. Tidsskr NorLaegeforen. 2007;127:1941-1944.

41. Zaborskis A, Vareikiene I. School bullying and itsassociation with health and lifestyle among school-children. Medicina (Kaunas). 2008;44:232-239.

42. Haavet OR, Straand J, Saugstad OD, Grunfeld B.Illness and exposure to negative life experiences inadolescence: Two sides of the same coin? A studyof 15-year-olds in Oslo, Norway. Acta Paediatr.2004;93:405-411.

43. Løhre A, Lydersen S, Paulsen B, Mæhle M,Vatten LJ. Peer victimization as reported by chil-dren, teachers, and parents in relation to children’shealth symptoms. BMC Public Health. 2011;11:278.

44. Luntamo T, Sourander A, Rihko M. Psychosocialdeterminants of headache, abdominal pain, andsleep problems in a community sample of Finnishadolescents. Eur Child Adolesc Psychiatry. 2012;21:301-313.

45. Natvig GK, Albrektsen G, Qvarnstrom U. Psycho-somatic symptoms among victims of school bullying.J Health Psychol. 2001;6:365-377.

46. Kshirsagar VY, Agarwal R, Bavdekar SB. Bullyingin schools: Prevalence and short-term impact. IndianPediatr. 2007;44:25-28.

47. Ramya SG, Kulkarni ML. Bullying among schoolchildren: Prevalence and association with commonsymptoms in childhood. Indian J Pediatr. 2011;78:307-310.

48. Fekkes M, Pijpers FIM, Verloove-Vanhorick SP.Bullying behavior and associations with psycho-somatic complaints and depression in victims.J Pediatr. 2004;144:17-22.

49. Fekkes M, Pijpers FIM, Fredriks AM, Vogels T,Verloove-Vanhorick SP. Do bullied children get ill,or do ill children get bullied? A prospective cohortstudy on the relationship between bullying andhealth-related symptoms. Pediatrics. 2006;117:1568-1574.

50. Arslan S, Hallett V, Akkas E, Akkas OA. Bullyingand victimization among Turkish children andadolescents: Examining prevalence and associatedhealth symptoms. Eur J Pediatr. 2012;171:1549-1557.

51. Karatas H, Ozturk C. Relationship between bullyingand health problems in primary school children.Asian Nurs Res. 2011;5:81-87.

52. Biebl SJ, Dilalla LF, Davis EK, Lynch KA,Shinn SO. Longitudinal associations among peervictimization and physical and mental health prob-lems. J Pediatr Psychol. 2011;36:868-877.

53. Srabstein JC, McCarter J, Shao C, Huang ZJ.Morbidities associated with bullying behaviors in

Headache 985

adolescents school based study of American adoles-cents. Int J Adolesc Med Health. 2006;18:587-596.

54. Hesketh T, Zhen Y, Lu L, Dong ZX, Jun YX, XingZW. Stress and psychosomatic symptoms in Chineseschool children: Cross-sectional survey. Arch DisChild. 2010;95:136-140.

55. Schnohr C, Niclasen BV. Bullying amongGreenlandic schoolchildren: Development since1994 and relations to health and health behaviour.Int J Circumpolar Health. 2006;65:305-312.

56. Gini G. Associations between bullying behaviour,psychosomatic complaints, emotional and behavi-oural problems. J Paediatr Child Health. 2008;44:492-497.

57. Williams K, Chambers M, Logan S, Robinson D.Association of common health symptoms withbullying in primary school children. BMJ. 1996;313:17-19.

58. Stickley A, Koyanagi A, Koposov R, McKee M,Roberts B, Ruchkin V. Peer victimisation and itsassociation with psychological and somatic healthproblems among adolescents in northern Russia.Child Adolesc Psychiatry Ment Health. 2013;7:15.

59. Pozzoli T, Gini G, Vieno A. Individual and classmoral disengagement in bullying among elementaryschool children. Aggress Behav. 2012;38:378-388.

60. Nolen-Hoeksema S, Girgus JS. The emergence ofgender differences in depression during adoles-cence. Psychol Bull. 1994;115:424-443.

61. Vaillancourt T, Hymel S, McDougall P. Why doesbeing bullied hurt so much? Insights from neuro-

science. In: Espelage D, Swearer S, eds. Bullying inNorth American Schools. New York: Taylor &Francis Group, Inc; 2010:23-33.

62. Eisenberger NI. The pain of social disconnection:Examining the shared neural underpinnings ofphysical and social pain. Nat Rev Neurosci. 2012;13:421-434.

63. Kliewer W. Violence exposure and cortisolresponses in urban youth. Int J Behav Med.2006;13:109-120.

64. Knack JM, Jensen-Campbell LA, Baum A. Worsethan sticks and stones? Bullying is linked withaltered HPA axis functioning and poorer health.Brain Cogn. 2011;77:183-190.

65. Rudolph KD, Troop-Gordon W, Granger DA. Indi-vidual differences in biological stress responsesmoderate the contribution of early peer victimiza-tion to subsequent depressive symptoms. Psycho-pharmacology (Berl). 2011;241:209-219.

66. Ladd GW, Kochenderfer-Ladd BJ. Identifyingvictims of peer aggression from early to middlechildhood: Analysis of cross-informant data forconcordance, estimation of relational adjustment,prevalence of victimization, and characteristics ofidentified victims. Psychol Assess. 2002;14:74-96.

67. Lien L, Green K, Welander-Vatn A, Bjertness E.Mental and somatic health complaints associatedwith school bullying between 10th and 12th gradestudents: Results from cross sectional studies inOslo, Norway. Clin Pract Epidemiol Ment Health.2009;23:5-6.

986 June 2014