Bullied Children and Psychosomatic Problems a Meta-Analysis.Pdf

Bullied Children and Psychosomatic Problems: A Meta-analysis Gianluca Gini and Tiziana Pozzoli Pediatrics 2013;132;720; originally published online September 16, 2013; DOI: 10.1542/peds.2013-0614

The online version of this article, along with updated information and services, is located on the World Wide Web at: http://pediatrics.aappublications.org/content/132/4/720.full.html

PEDIATRICS is the official journal of the American Academy of Pediatrics. A monthly publication, it has been published continuously since 1948. PEDIATRICS is owned, published, and trademarked by the American Academy of Pediatrics, 141 Northwest Point Boulevard, Elk Grove Village, Illinois, 60007. Copyright © 2013 by the American Academy of Pediatrics. All rights reserved. Print ISSN: 0031-4005. Online ISSN: 1098-4275.

Downloaded from pediatrics.aappublications.org at Hospital La Paz on October 16, 2013

Bullied Children and Psychosomatic Problems: A Meta-analysis Gianluca Gini and Tiziana Pozzoli Pediatrics 2013;132;720; originally published online September 16, 2013; DOI: 10.1542/peds.2013-0614 Updated Information & including high resolution figures, can be found at: Services http://pediatrics.aappublications.org/content/132/4/720.full.ht ml References This article cites 48 articles, 12 of which can be accessed free at: http://pediatrics.aappublications.org/content/132/4/720.full.ht ml#ref-list-1

Post-Publication One P3R has been posted to this article: Peer Reviews (P3Rs) http://pediatrics.aappublications.org/cgi/eletters/132/4/720

Permissions & Licensing Information about reproducing this article in parts (figures, tables) or in its entirety can be found online at: http://pediatrics.aappublications.org/site/misc/Permissions.xh tml Reprints Information about ordering reprints can be found online: http://pediatrics.aappublications.org/site/misc/reprints.xhtml

Downloaded from pediatrics.aappublications.org at Hospital La Paz on October 16, 2013 Bullied Children and Psychosomatic Problems: A Meta-analysis

AUTHORS: Gianluca Gini, PhD, and Tiziana Pozzoli, PhD abstract Department of Developmental and Social Psychology, University BACKGROUND AND OBJECTIVE: A previous meta-analysis showed that of Padua, Padua, Italy being bullied during childhood is related to psychosomatic problems, KEY WORDS bullying, peer victimization, psychosomatic problems, health, but many other studies have been published since then, including some meta-analysis longitudinal studies. We performed a new meta-analysis to quantify the ABBREVIATIONS association between peer victimization and psychosomatic complaints in CI—confidence interval — the school-aged population. Nfs fail-safe N OR—odds ratio METHODS: We searched online databases up to April 2012, and bibliog- SES—socioeconomic status raphies of retrieved studies and of narrative reviews, for studies that Dr Gini contributed to protocol design, literature search, data examined the association between being bullied and psychosomatic extraction, statistical analysis, and writing the manuscript; Dr complaints in children and adolescents. The original search identified Pozzoli contributed to literature search, data extraction, and fi 119 nonduplicated studies, of which 30 satisfied the prestated inclusion writing the manuscript; and both authors approved the nal manuscript as submitted. criteria. www.pediatrics.org/cgi/doi/10.1542/peds.2013-0614 RESULTS: Two separate random effects meta-analyses were performed doi:10.1542/peds.2013-0614 on 6 longitudinal studies (odds ratio = 2.39, 95% confidence interval, Accepted for publication Jul 10, 2013 1.76 to 3.24) and 24 cross-sectional studies (odds ratio = 2.17, 95% Address correspondence to Gianluca Gini, PhD, Department of confidence interval, 1.91 to 2.46), respectively. Results showed that Developmental and Social Psychology, via Venezia 8, 35131, bullied children and adolescents have a significantly higher risk for Padova, Italy. E-mail: [email protected] psychosomatic problems than non-bullied agemates. In the cross- PEDIATRICS (ISSN Numbers: Print, 0031-4005; Online, 1098-4275). sectional studies, the magnitude of effect size significantly decreased Copyright © 2013 by the American Academy of Pediatrics with the increase of the proportion of female participants in the study FINANCIAL DISCLOSURE: The authors have indicated they have sample. No other moderators were statistically significant. no financial relationships relevant to this article to disclose. CONCLUSIONS: The association between being bullied and psychoso- FUNDING: No external funding. matic problems was confirmed. Given that school bullying is a wide- POTENTIAL CONFLICT OF INTEREST: The authors have indicated spread phenomenon in many countries around the world, the present they have no potential conflicts of interest to disclose. results indicate that bullying should be considered a significant international public health problem. Pediatrics 2013;132:720–729

720 GINI and POZZOLI REVIEW ARTICLE

Being bullied during childhood or ad- with poor well-being. However, Reijntjes This meta-analysis was planned, con- olescence is a risk factor for a person’s and colleagues’ review included only 2 ducted, and reported in adherence to well-being and adjustment. Studies studies that measured psychosomatic the Meta-analysis of Observational Stud- have shown that peer victimization is symptoms; unfortunately, these symp- ies in Epidemiology guidelines.13 related mainly to internalizing prob- toms were not distinguished from other lems, including low self-esteem, high types of internalizing problems (eg, de- Inclusion Criteria anxiety, and depression,1–4 and it also pression, anxiety, or loneliness), but a A study had to meet the following is linked to suicidal ideation and at- pooled correlation for each study was aprioricriteriatobeincluded.Themost tempt.5,6 Moreover, it is increasingly computed, with no comparison between basic requirement was the inclusion recognized that bullied students can bullied and nonbullied children. of measures of peer victimization at also be affected by poor physical health Since the publication of the first meta- school in childhood or adolescence and and show a variety of symptoms, such analysis in 2009, several other studies of psychosomatic symptoms. These as headache, backache, abdominal pain, that assessed the risk for psychoso- measures could include self-report skin problems, sleeping problems, bed- matic problems in bullied children have questionnaires; peer, parent, or teacher wetting, or dizziness.7–11 Given that in been added to the literature, including reports; or an interview that resulted such circumstances psychosocial pro- some studies that used a longitudinal in a quantitative rating of peer victimi- cesses seem to act as a key factor neg- design. This new meta-analysis seeks zation and health problems. Second, atively affecting children’shealth,these to update and expand both Gini and studies were required to have reported symptoms are often called psychoso- Pozzoli’s and Reijntjes and colleagues’ effect sizes and related confidence matic problems.7–10 meta-analyses3,12 by (1) including the intervals or enough information to calculate these data, for example, by re- To date, the only meta-analysis12 spe- subsequently published studies that porting comparisons between bullied cifically conducted on this issue was allowed to estimate the risk for psy- children and a control group (defined published in 2009. That meta-analysis chosomatic problems in children and aschildren from thesamepopulationof synthesized the results of 11 studies adolescents who are bullied by peers victims who were classified as not that have analyzed the association be- (ie, cases) compared with nonbullied bullied). We excluded the following tween being victimized by peers at peers (ie, controls), (2) performing types of studies: studies that did not school and the prevalence of symp- separate meta-analyses of longitudinal include a control group; studies that toms among children and adolescents and cross-sectional studies, and (3) measured psychosomatic symptoms between 7 years and 16 years of age. testing for potential moderators of vari- with items included in a larger scale, Bullied students were found to have ation in the magnitude of effect sizes. because these symptoms could not be a significantly higher risk for psycho- METHODS clearly distinguished from other prob- somatic problems than were the con- lems; studies with duplicated data; trols, that is, the agemates who were Literature Search studies that did not report analyses on not involved in bullying (pooled odds Four methods were used to identify rel- the variables of interest; and studies fi ratio [OR] = 2.00, 95% con dence in- evant studies. First, electronic searches with adults or psychiatric patients. The terval [CI], 1.70 to 2.35). Although im- in PsycINFO, PubMed, the Cochrane Li- authorsindependentlyassessedwhether portant, those results were limited by brary database, the Campbell Collabo- articlesmettheinclusioncriteria.Inthe the small number of studies included ration database, and Scopus were case of disagreement, a consensus was in the meta-analysis (which also pre- conducted in April 2012 with the following reached through discussion. cluded the possibility of testing for keywords: “bullying,”“peer victimiza- possible moderators), and the results tion” AND “somatic,”“psychosomatic,” Coding of Studies were also limited by the fact that only 2 and “physical health.” Second, the “cited Studies were coded on design (cross- of them used a longitudinal design. by” function in Scopus was used to re- sectional versus longitudinal), length Subsequently, another meta-analysis3 trieve empirical articles that have cited offollow-upforlongitudinalstudies,type has analyzed data from longitudinal the previous meta-analysis.12 Third, re- of bullying measure (self-report ques- studies that measured a variety of view articles about consequences of tionnaire versus peer or adult reports internalizing problems, including psy- bullying were reviewed for possible rel- versusinterview),typeofpsychosomatic chosomatic symptoms. Overall, this evant citations. Finally, the reference symptoms measured, type of sampling meta-analysis has confirmed that peer sections of the collected articles were procedure, sample composition and char- victimization is positively associated searched for relevant earlier references. acteristics, and geographic location of

PEDIATRICS Volume 132, Number 4, October 2013 721 study. Quantitative data were extracted variance plus the between-studies remaining 64 potentially relevant stud- from text and tables; for the sake of variance t-squared (Τ2). The z statis- ies were obtained. Fourteen studies comparability with the results of the tic was calculated, and a two-tailed were excluded because they did not former meta-analysis,12 the data that P value of ,.05 was considered to in- meet the inclusion criteria (eg, they did were adjusted for important con- dicate statistical significance. Statisti- not have a control group). Fifteen stud- founders (eg, gender, age, ethnicity, or cal heterogeneity was assessed by iesdidnotreportenoughdatatocom- parental education) were preferred. using the Q statistic to evaluate pute effect sizes or confidence intervals. whether the pooled studies represent Five studies were not available in full a homogeneous distribution of effect text. The remaining 30 studies were in- Statistical Analyses sizes. Also reported is the I2 statistic, cluded for this meta-analysis. Six stud- Eleven studies reported an effect based indicating the proportion of observed ies were longitudinal studies, and 24 on a single composite score for psy- variance that reflects real differences used a cross-sectional design. chosomatic complaints, whereas the 15 in effect size. Table 1 summarizes the characteristics remaining studies reported data for Toaddressthepossiblepublicationbias of the studies included in this meta- a number of different symptoms dis- (ie, the fact that studies with non- analysis, including sample size and tinctly (eg, headache, stomachache, significant results are less likely to be response rate, age and gender com- backache, abdominal pain, dizziness, published), we computed the fail-safe N position of the sample, type of mea- sleeping problems, poor appetite, bed- (N ) according to the method Orwin16 sures, symptoms measured, study design, wetting, skin problems, vomiting; see fs proposed, which is more conservative and type of sampling. A total of 219 560 Table 1). Because the number and the than the traditional Rosenthal N .17,18 children and adolescents participated type of symptoms varied systemati- fs Orwin’s N determines the number of in the 30 studies. Across the 26 studies cally across studies, following Gini and fs additional studies in a meta-analysis that provided information about the Pozzoli’s original procedure,12 the OR yielding null effect sizes that would sample’s gender composition, 50.2% for each symptom was extracted, and be needed to yield a “trivial” OR of (range, 32.8% to 62.4%) of the partic- then a pooled OR was computed from 1.05. Researchers suggest that meta- ipants were girls. each study. (Items that referred to analysts calculate a tolerance level psychological problems, such as anxi- Five studies were from Norway, 2 of around a fail-safe N that is equal to 5 ety or depression, were not included in which were from the same publi- times the number of effects included in 28,34,35,38 this computation.) This procedure cation ;4fromtheUnited the meta-analysis plus 10 (the “5k +10” 22,23,27,43 8,11,37 allowed a direct comparison with the States ; 3 from Australia ;2 benchmark).18,19 Moreover, the associ- 44,45 results of the former meta-analysis. from the United Kingdom ; 2 from the ation between the standardized effect 7,25 31,36 The case group included victims, that Netherlands ; 2 from Finland ;2 sizes and the variances of these effects 33,40 is, children who are bullied by peers. from India ; and 1, respectively, from was analyzed by rank correlation with 26 29 30 41 The control group featured children Austria, China, France, Germany, use of the Kendall t method. If small 42 10 21 who have not been bullied. With very Greenland, Italy, Mexico, and Tur- studies with negative results were less 32 few exceptions, studies did not report key. Two articles reported data from likely to be published, the correlation 24,39 results for boys and girls separately; multiple countries. Information about between variance and effect size would therefore, we were not able to compare race or ethnicity and socioeconomic be high. Conversely, a lack of a signifi- effect sizes for these two groups of status (SES) of the participants was not cant correlation can be interpreted as children. Because most of the studies systematically reported in all studies. the absence of publication bias.20 reported the proportion of girls in the Overall, the heterogeneity of racial and fi sample, we used this information to SES classi cation within and across the test for possible moderation by gender. RESULTS studies was such that it precluded analysis by race and ethnicity or SES. Analyses were done using Compre- After the removal of duplicates, a list of hensive Meta-Analysis.14 We extracted 119 potentially eligible studies was the OR and 95% CI from each study. Data generated (Fig 1). Based on titles and Meta-analysis of Longitudinal from individual studies were pooled by abstracts, 55 articles were excluded at Studies using a random effects model. Each the first screening because they were Six studies used a longitudinal design. study was weighted by the inverse of its qualitative studies, reviews or commen- The follow-up duration ranged from 9 variance, which, under the random ef- taries, or studies that did not measure months to 11 years. Across the 6 sam- fects model, includes the within-study school bullying. Full-text copies of the ples, bullied children were found to

722 GINI and POZZOLI EITISVlm 3,Nme ,Otbr2013 October 4, Number 132, Volume PEDIATRICS TABLE 1 Characteristics of Studies Included in the Meta-analysis Source Sample Size Age Range, Bullying Measure Symptom Symptoms Reported Adjustment for Study Design Type of (response rate) y (%girls) Measure Confounders Sampling Albores-Gallo 340 (n/a) 7–11 (n/a) Peer nomination Self-report Single score None Cross-sectional Convenience et al (2011)21 questionnaire questionnaire sample Biebl et al (2011)22 65 (n/a) 12–20 at Time 1: play session; Self-report Single score Gender Longitudinal Convenience time 3 (52.9) time 3: self-report questionnaire sample questionnaire Burk et al (2011)23 344 (60%) 7–15 (52.3) Multi-informant (self-, Multi-informant Single score None Longitudinal Convenience teacher-, parent (self-, teacher-, sample report) parent report) Due et al (2005)24 123 227 (.90%) 11–15 (51) Self-report Self-report Headache, stomachache, Age, family afﬂuence, Cross-sectional Cluster random questionnaire questionnaire backache, sleeping country sampling difﬁculties, tired in the morning, dizziness, irritable, feeling nervous Fekkes et al (2004)7 2766 (100%) 9–12 (50) Self-report Self-report Headache, skin problems, Gender Cross-sectional Unknown questionnaire questionnaire abdominal pain, tense muscles, feeling tired, bad appetite Fekkes et al (2006)25 1118 (70%) 9–11 (50.3) Self-report Self-report Abdominal pain, sleeping Gender, age, having Longitudinal Unknown questionnaire questionnaire problems, headache, friends feeling tense, feeling tired, poor appetite, bedwetting Forero et al (1999)8 3918 (86%) 11–15 (54.3) Self-report Self-report Single score Clusters within Cross-sectional Cluster random questionnaire questionnaire schools sampling Gini (2008)10 565 (94%) 8–11 (52.9) Self-report Self-report Headache, abdominal pain, Gender, age Cross-sectional Simple random questionnaire questionnaire feeling tired, feeling tense, sampling sleeping problems, dizziness Gradinger 761 (95%) 14–19 (51.5) Self-report Self-report Single score None Cross-sectional Convenience et al (2009)26 questionnaire questionnaire sample Gruber and Fineran 522 (51%) 12–17 (42.9) Self-report Self-report Single score None Cross-sectional Convenience (2008)27 questionnaire questionnaire sample Haavet et al (2004)28 8316 (88%) 15 (54.4) Self-report Self-report Headache, pain from neck None Cross-sectional Population study questionnaire questionnaire or shoulder Hesketh et al 2191 (80%) 9–12 (44) Self-report Self-report Headache, abdominal pain Gender, age, residence, Cross-sectional Simple random (2010)29 questionnaire questionnaire parental education sampling Houbre et al (2006)30 291 (95%) 9–12 (n/a) Self-report Self-report Skin conditions, sleeping None Cross-sectional Convenience questionnaire questionnaire disorders, digestive sample disorders, somatic pain, vegetative symptoms,

diarrhea, and constipation ARTICLE REVIEW Kaltiala-Heino 17 643 (87%) 14–16 (49.3) Self-report Self-report Single score Age, gender, family Cross-sectional Convenience et al (2000)31 questionnaire questionnaire structure, parental sample education 723 724 TABLE 1 Continued Source Sample Size Age Range, Bullying Measure Symptom Symptoms Reported Adjustment for Study Design Type of IIadPOZZOLI and GINI (response rate) y (%girls) Measure Confounders Sampling Karatas and 92 (82%) 10–12 (51.1) Self-report Parent report Headache, abdominal pain, None Cross-sectional Simple random Ozturk (2011)32 questionnaire questionnaire stomachache, backache, sampling skin problems, restlessness, nervousness, sleeping problems, dizziness, respiratory problems, poor appetite Kshirsagar 500 (100%) 8–12 (62.4) Semistructured Semistructured Headache, tummy aches, Not specified Longitudinal Simple random et al (2007)33 interview interview body ache, has failed, bites sampling nails, sleep problems, vomiting, bedwetting Lien et al (2009) 3790 (88%) 15–16 (49.3) Self-report Self-report Headache; pain in neck or SES, family structure, Cross-sectional Population study (sample 1)34 questionnaire questionnaire shoulder; pain in arm, ethnicity, exposure leg, or knee; abdominal to violence, having pain; back pain close friends Lien et al (2009) 3790 (80%) 18–19 (55.9) Self-report Self-report Headache; pain in neck or SES, family structure, Cross-sectional Population study (sample 2)34 questionnaire questionnaire shoulder; pain in arm, ethnicity, exposure to leg, or knee; abdominal violence, having close pain; back pain friends Løhre et al (2011)35 419 (100%) 7–16 (n/a) Multi-informant Self-report Stomachache, headache Gender, grade Cross-sectional Convenience (self-, teacher-, questionnaire sample parent report) Luntamo et al (2012)36 2215 (91%) 13–18 (50) Self-report Self-report Headache, abdominal pain, None Cross-sectional Population study questionnaire questionnaire sleep problems McGee et al (2011)37 1806 (n/a) 14–21 (51.8) Self-report Self-report Single score Childhood aggression, Longitudinal Population study questionnaire questionnaire social or thought problem behaviors, poverty, physical punishment Natvig et al (2001)38 856 (83.7%) 13–15 (50.6) Self-report Self-report Headache, stomachache, Gender, age, school Cross-sectional Unknown questionnaire questionnaire backache, feeling dizzy, irritability, feeling nervous, sleeplessness Nordhagen 17 114 (68%) 2–17 (49.1) Parent report Parent report Single score Gender, age, country, Cross-sectional Stratified random et al (2005)39 questionnaire questionnaire living area, family sampling situation, education Ramya and Kulkarni 500 (n/a) 8–14 (32.8) Interview Interview Headache, tummy ache, None Cross-sectional Simple random (2011)40 bedwetting, fever sampling Richter et al 5650 (70%–80%) 11–15 (49.8) Self-report Self-report Single score None Cross-sectional Cluster random (2007)41 questionnaire questionnaire sampling Rigby (1999)11 78 (28.3%) Time 1: 13.8, Self-report Self-report Single score None Longitudinal Convenience Time 2: 16.7 (44.9) questionnaire questionnaire sample Schnohr and 891 (n/a) 11–15 (n/a) Self-report Self-report Headache, stomachache, Gender, age Cross-sectional Cluster random Niclasen (2006)42 questionnaire questionnaire sleeping difficulties sampling REVIEW ARTICLE Sampling sample sampling Study Design Type of Cross-sectional Convenience Cross-sectional Cluster random ed Cross-sectional Population study fi Confounders ethnic minority overweight, maternal education Not speci Gender, school year, Gender, age, race,

FIGURE 1 Flow diagram of study inclusion.

have a signiﬁcantly higher risk for effect size to a trivial effect (5k +10

Symptoms Reported Adjustment for psychosomatic problems than non- benchmark = 40). wetting, headache, tummy ache breathing problems, nausea, poor appetite dizziness, backache, irritability, feeling nervous, sleeping problems Sleeping problems, bed Sore throat, cold or coughs, Headache, stomachache, bullied agemates were (OR = 2.39, 95% CI, 1.76 to 3.24, Z = 5.62, P , .0001). Meta-analysis of Cross-Sectional Figure 2 shows the forest plot for this Studies meta-analysis. Studies were highly ho- Across the 24 samples that were in-

Measure 2 interview questionnaire questionnaire mogeneous (Q = 4.94, P = .42, I = 0%). cluded in the cross-sectional studies, Semistructured Self-report Furthermore, no evidence of publica- bullied children were found to have tion bias was present. Kendall’s t was a signiﬁcantly higher risk for psycho- .53 with two-tailed P = .13. An additional somatic problems than were non- 102 studies with null effect sizes would bullied peers (OR = 2.17, 95% CI, 1.91 to be needed to attenuate this omnibus 2.46, Z = 12.09, P , .0001). Figure 3 interview questionnaire Bullying Measure Symptom 9 (49.6) Interview Parent report 10 (n/a) Semistructured 15 (53.5) Self-report – – – y (%girls) Age Range, 2962 (93.1%)1639 (82%) 7 6 (response rate) 15 305 (83%) 11 45

Continued FIGURE 2 43 44 Source Sample Size Forest plot for random effects meta-analysis of the association between being bullied and psychosomatic problems: longitudinal studies. Note: Effect sizes are expressed as odds ratios. Studies are (2006) (1996) ’ TABLE 1 n/a, not available. Srabstein et al Wolke et al (2001) Williams et al represented by symbols whose area is proportional to the studys weight in the analysis.

PEDIATRICS Volume 132, Number 4, October 2013 725 shows the forest plot for this meta- 95% CI, 1.61 to 2.90, respectively; Q = sampling design or a whole population analysis. Effect sizes within this group 0.004, P = .95). of students and a good response rate of studies were not homogeneous (Q = Moreover,thepotentialmoderatingrole (.80%). Twelve studies satisfied both 2 103.06, P , .001, I =77.7%).Again,no of a methodological feature, namely the criteria. We then performed a separate evidence of publication bias was pres- type of informant, was tested. Twenty meta-analysis of this subgroup of ent. Kendall’s t was .05 with two-tailed studies used the participant as an in- studies, and the results were OR = 2.10, P = .75. An additional 325 studies with null formant for involvement in bullying (ie, 95% CI, 1.87 to 2.46. effect sizes would be needed to attenuate used self-report questionnaires or this omnibus effect size to a negligible interviews with the child), and only 4 DISCUSSION value (5k +10benchmark=130). studies collected data through other Our meta-analysis showed that bullied Moderator analyses with gender com- informants (ie, peers or parents). Effect pupils are at least two times more likely position of the sample, geographic lo- sizes did not vary as a function of the than nonbullied agemates to have cation, and type of informant were type of informant associated with bul- psychosomatic problems. Thus, this performed to explore possible explan- lyingexperiences(OR=2.17,95%CI,1.86 updated meta-analysis confirmed the ations for heterogeneity in the effect to2.53forself-reports,OR=2.18,95%CI, findings of the former meta-analytic sizes across cross-sectional studies. 1.55 to 3.06 for other informants; Q = synthesis12 with a much larger sample The proportion of girls in the sample 0.00, P = .98). Similarly, 19 studies col- of studies. Importantly, the same result lected information about symptoms was available for 20 of the 24 cross- was found not only with cross-sectional from the participants themselves (OR = sectional studies, and it was used as studies but also in a meta-analysis of 6 2.21, 95% CI, 1.90 to 2.58), whereas 5 a continuous predictor in a weighted studies that used a longitudinal design. studies asked other informants (OR = mixed-effects metaregression. The Finally, the meta-regression analysis fi 2.00, 95% CI, 1.47 to 2.72). Also, the ef- magnitudeoftheeffectsizesigni cantly showed that the strength of the re- fect of this moderator was not statis- decreased with an increase in the lationship between being bullied and tically significant: Q = 0.37, P = .57. number of female participants in the having health problems is higher when 2 2 12 study sample (B = 0.04, 95% CI, 0.07 Finally,asintheformermeta-analysis, samples contain proportionally more 2 , ’ to 0.02, P .002). The studys geo- a sensitivity analysis was performed boys. Given the explorative nature of graphic location (coded as Europe based on the quality of the studies. this analysis, a significant finding is not versus other countries) was not a sig- Quality was assessed through 2 crite- to be considered definitive, but it does fi ni cant moderator (k = 15, OR = 2.19, ria (beyond those required as inclusion suggest a direction for additional re- 95% CI, 1.82 to 2.62, and k = 8, OR = 2.16, criteria): the use of a randomized search. A possible explanation might deal with the fact that a school or classroom environment with a higher proportion of male students is a con- text in which bullying behavior is more likely to happen and where supportive and helping behaviors in favor of the bullied pupils are less frequent.46 This could increase the negative impact of being bullied on children’s health. The influence of the school environment’s gender composition on peer victimization and its consequences for children’s well-being is a topic that warrants additional research. Since the former meta-analysis, the number of studies testing the association between bullying experiences and FIGURE 3 psychosomatic problems has tripled. Forest plot for random effects meta-analysis of the association between being bullied and psychosomatic problems: cross-sectional studies. Note: Effect sizes are expressed as odds ratios. Studies are We can reasonably conclude that this represented by symbols whose area is proportional to the study’s weight in the analysis. association is established, and we call

726 GINI and POZZOLI REVIEW ARTICLE for new research efforts aimed at that the major limitations of the liter- their overlapping, research has dem- elucidating the mechanisms through ature that were highlighted in the first onstrated the importance of dis- which bullying affects children’s health. meta-analysis are still present is star- tinguishing the 2 forms of victimization We also call for research that inves- tling to see. For example, much variability because they may be differentially re- tigates how other environmental fac- exists in the methods and instruments lated to personal adjustment.48 Future tors interact with peer victimization used to assess the prevalence of symp- studies should analyze the negative experiences to determine health risk. toms and peer victimization experiences. health consequences of physical and However, not all children are at the The majority of studies used a variety of relational victimization experiences. same risk for developing health prob- self-report questionnaires, both for peer Finally, our meta-analysis shares the lems: Some children may be more re- victimization and for children’s health same limitations of all meta-analyses of silient than others against a high-risk complaints. In some cases, these mea- observational studies. Because indi- environment. To explain this adaptive sures were reduced to a single-item vidualscannotbe randomlyallocatedto success, protective factors must be questionnaire. Self-report measures groups, the influence of confounding considered. For example, supportive are very common in bullying research variables cannot be fully evaluated. parent–child relationships, character- and are usually considered to be valid Although many studies controlled for ized by parental warmth, supervision, and reliable.47 However, possible pro- important confounding variables, such support, and involvement, may protect blems with these instruments are that as parental education and socioeco- children from adverse life experiences they require a good level of respond- nomic status or exposure to violence at school, such as being bullied by ents’self-consciousness and that some outside of school, other unknown con- peers, and thus reduce negative con- bullied children may deny their condi- founders could be partially responsible sequences. Similarly, attachment to tion. Finally, associations between data for the effect observed. school, sense of belonging, and school derived from the same source (ie, support may be related to better stu- when children self-report both bullying Implications for Practitioners dent health. Longitudinal studies that experiences and health problems) Thestudiesreviewedsupportedthefact address the mediating role of these might be inflated by the common and other environmental factors on the that bullied children have more fre- method variance. For these reasons, peer victimization–health problems quent psychosomatic problems than we stress the need for future studies to link are much needed. nonbullied pupils. Moreover, this meta- collect information through multiple analysis significantly complements the independent informants, such as chil- growing body of research that docu- Strengths and Limitations dren themselves, their peers within the ments the poor personal adjustment of The strengths of this meta-analysis in- class, and their teachers or parents. bullied children and adolescents, in clude the much larger number of Moreover, it is important that research- terms of both internalizing and exter- studies that were available this time ers choose validated and widely used nalizing problems, which other recent compared with the former meta- instruments rather than ad hoc or newly meta-analyses on the psychosocial analysis. Another strength is the wide developed scales with no evidence of consequences of peer victimization3,4,49 geographic distribution of the samples, reliability or validity. Also, the assess- summarize. Altogether, these results which were derived from several dif- ment of children’s physical health must have significant implications for pedia- ferent countries around the world. be improved. For example, none of the tricians, child psychologists, and other Furthermore, we were able to perform available studies included independent health care professionals. It is very im- separate meta-analyses of longitudinal objective information, such as children’s portant that these professionals be and cross-sectional studies, which school absenteeism extracted from ready to identify children who are at yielded the same results. Finally, we did school attendance records or their visits risk for being bullied because the po- not find evidence of publication bias to the school nurse office. tential negative health, psychological, that may have led to overestimating the Furthermore,thestudiesincludedinthis and educational consequences of bul- association between bullying and psy- meta-analysis, and in the former meta- lying experiences are far-reaching. chosomatic problems. analysis, did not measure different Pediatricians can play an important role Meta-analysis is an invaluable tool for forms of victimization separately (ie, in detecting potential victims of bullying integrating priorresearch, illuminating physical and relational victimization) or if they consider bullying as a possible research gaps, and defining priorities did not report separate analyses for risk factor in any patient with recurrent for future research. However, the fact different forms of victimization. Despite headaches, breathing problems, poor

PEDIATRICS Volume 132, Number 4, October 2013 727 appetite, sleeping problems, and so on. Asking whether the child feels safe at development of social skills and as- Any recurrent and unexplained somatic school can also allow the pediatrician to sertiveness in their children. Pedia- symptom can be a warning sign of bul- gain insight into the level of concern the tricians’ suggestions are likely to be lying victimization. Because children do child is experiencing. particularly effective given the high not easily talk about their bullying conﬁdence that parents usually put in Moreover, pediatricians could rou- experiences, pediatricians could ap- these professionals. Furthermore, par- tinely review the warning signs of proach the issue of bullying through ents should be encouraged to ask for bullying with parents to help them general questions, for example, by in- school support when a case of bullying quiring about the child’s experience and identify problems with bullying their emerges. Breaking the cycle of victimi- friends in school. If the child seems to be child may be experiencing. Preventive zation through early identiﬁcation and withdrawn from peers, the pediatrician measures can also include counseling prompt intervention may prevent per- should ask for the reason and de- parents about bullying experiences as sistent physical and mental health termine whether teasing, name calling, a risk factor for children’s well-being problems in children who experience or deliberate exclusion may be involved. and the importance of promoting bullying.

REFERENCES

1. Callaghan S, Joseph S. Self-concept and 11. Rigby K. Peer victimisation at school and en México [Bullying and its association peer victimization among schoolchildren. the health of secondary school students. Br with psychiatric disorders in a Mexican Pers Individ Dif. 1995;18(1):161–163 J Educ Psychol. 1999;69(pt 1):95–104 students sample]. Salud Publica Mex. 2011; 2. Hodges EVE, Perry DG. Personal and in- 12. Gini G, Pozzoli T. Association between bul- 53(3):220–227 terpersonal antecedents and consequences lying and psychosomatic problems: a meta- 22. Biebl SJ, Dilalla LF, Davis EK, Lynch KA, Shinn of victimization by peers. J Pers Soc Psy- analysis. Pediatrics. 2009;123(3):1059–1065 SO. Longitudinal associations among peer chol. 1999;76(4):677–685 13.StroupDF,BerlinJA,MortonSC,etal.Meta- victimization and physical and mental 3. Reijntjes AHA, Kamphuis JH, Prinzie P, Telch analysis of observational studies in epidemi- health problems. J Pediatr Psychol. 2011;36 MJ. Peer victimization and internalizing ology: a proposal for reporting. Meta-analysis (8):868–877 problems in children: a meta-analysis of of Observational Studies in Epidemiology 23. Burk LR, Armstrong JM, Park JH, Zahn- longitudinal studies. Child Abuse Negl. (MOOSE) group. JAMA. 2000;283(15):2008– Waxler C, Klein MH, Essex MJ. Stability of – 2010;34(4):244 252 2012 early identified aggressive victim status in fi 4. Tto MM, Farrington DP, Losel F, Loeber R. 14. Comprehensive Meta-Analysis [computer elementary school and associations with Do the victims of school bullies tend to program]. Version 2.2. Englewood, NJ: Bio- later mental health problems and func- become depressed later in life? A system- stat; 2006. tional impairments. J Abnorm Child Psy- atic review and meta-analysis of longitu- chol. 2011;39(2):225–238 15. Higgins JP, Thompson SG, Deeks JJ, Altman dinal studies. JACPR. 2011;3(2):63–73 DG. Measuring inconsistency in meta- 24. Due P, Holstein BE, Lynch J, et al; Health Behaviour in School-Aged Children Bullying 5. Cleary SD. Adolescent victimization and analyses. BMJ. 2003;327(7414):557–560 associated suicidal and violent behaviors. Working Group. Bullying and symptoms 16. Orwin RG. A fail-safe N for effect size in Adolescence. 2000;35(140):671–682 among school-aged children: international meta-analysis. J Educ Stat. 1983;8(2):157– 6. Kim YS, Koh YJ, Leventhal B. School bullying comparative cross sectional study in 28 159 and suicidal risk in Korean middle school countries. Eur J Public Health. 2005;15(2): students. Pediatrics. 2005;115(2):357–363 17. Rosenthal R. Combining results of in- 128–132 dependent studies. Psychol Bull. 1978;85(1): 7. Fekkes M, Pijpers FIM, Verloove-Vanhorick 25. Fekkes M, Pijpers FIM, Fredriks AM, Vogels 185–193 SP. Bullying behavior and associations with T, Verloove-Vanhorick SP. Do bullied chil- fi psychosomatic complaints and depression 18. Rosenthal R. The le drawer problem and dren get ill, or do ill children get bullied? A in victims. J Pediatr. 2004;144(1):17–22 tolerance for null results. Psychol Bull. prospective cohort study on the relation- 1979;86(3):638–641 8. Forero R, McLellan L, Rissel C, Bauman A. ship between bullying and health-related fi Bullying behaviour and psychosocial health 19. Muellerleile P, Mullen B. Suf ciency and symptoms. Pediatrics. 2006;117(5):1568– among school students in New South stability of evidence for public health inter- 1574 Wales, Australia: cross sectional survey. ventions using cumulative meta-analysis. 26. Gradinger P, Strohmeier D, Spiel C. Tradi- BMJ. 1999;319(7206):344–348 Am J Public Health. 2006;96(3):515–522 tional bullying and cyberbullying: identifi- 9. Berntsson LT, Köhler L, Gustafsson JE. Psy- 20. Begg CB, Mazumdar M. Operating charac- cation of risk groups for adjustment chosomatic complaints in schoolchildren: teristics of a rank correlation test for problems. Z Psychol. 2009;217(4):205–213 a Nordic comparison. Scand J Public publication bias. Biometrics. 1994;50(4): 27. Gruber JE, Fineran S. Comparing the impact Health. 2001;29(1):44–54 1088–1101 of bullying and sexual harassment victimi- 10. Gini G. Associations between bullying be- 21. Albores-Gallo L, Sauceda-García JM, Ruiz- zation on the mental and physical health of haviour, psychosomatic complaints, emo- Velasco S, Roque-Santiago E. El acoso escolar adolescents. Sex Roles.2008;59:1–13 tional and behavioural problems. J (bullying) y su asociación con trastornos 28. Haavet OR, Straand J, Saugstad OD, Grünfeld Paediatr Child Health. 2008;44(9):492–497 psiquiátricos en una muestra de escolares B. Illness and exposure to negative life

728 GINI and POZZOLI REVIEW ARTICLE

experiences in adolescence: two sides of the 36. Luntamo T, Sourander A, Rihko M, et al. Psy- behaviors in adolescents. School based same coin? A study of 15-year-olds in Oslo, chosocial determinants of headache, abdom- study of American adolescents. Int J Ado- Norway. Acta Paediatr. 2004;93(3):405–411 inal pain, and sleep problems in a community lesc Med Health. 2006;18(4):587–596 29. Hesketh T, Zhen Y, Lu L, Dong ZX, Jun YX, sample of Finnish adolescents. Eur Child 44. Williams K, Chambers M, Logan S, Robinson – Xing ZW. Stress and psychosomatic symp- Adolesc Psychiatry. 2012;21(6):301 313 D. Association of common health symptoms toms in Chinese school children: cross- 37. McGee TR, Scott JG, McGrath JJ, et al. Young with bullying in primary school children. sectional survey. Arch Dis Child. 2010;95 adult problem behaviour outcomes of ado- BMJ. 1996;313(7048):17–19 – – (2):136 140 lescent bullying. JACPR. 2011;3(2):110 114 45. Wolke D, Woods S, Bloomﬁeld L, Karstadt L. 30. Houbre B, Tarquinio C, Thuillier I, Hergott E. 38. Natvig GK, Albrektsen G, Qvarnstrøm U. Bullying involvement in primary school and Bullying among students and its con- Psychosomatic symptoms among victims of common health problems. Arch Dis Child. sequences on health. Eur J Psychol Educ. school bullying. J Health Psychol. 2001;6(4): 2001;85(3):197–201 2006;21(2):183–208 365–377 46. Pozzoli T, Gini G, Vieno A. Individual and 31. Kaltiala-Heino R, Rimpelä M, Rantanen P, 39. Nordhagen R, Nielsen A, Stigum H, Köhler L. class moral disengagement in bullying Rimpelä A. Bullying at school: an indicator Parental reported bullying among Nordic among elementary school children. Ag- of adolescents at risk for mental disorders. children: a population-based study. Child gress Behav. 2012;38(5):378–388 J Adolesc. 2000;23(6):661–674 Care Health Dev. 2005;31(6):693–701 47. Ladd GW, Kochenderfer-Ladd BJ. Identifying 32. Karatas H, Ozturk C. Relationship between bul- 40. Ramya SG, Kulkarni ML. Bullying among victims of peer aggression from early to lying and health problems in primary school school children: prevalence and associa- middle childhood: analysis of cross- children. Asian Nurs Res. 2011;5(2):81–87 tion with common symptoms in childhood. informant data for concordance, estimation 33. Kshirsagar VY, Agarwal R, Bavdekar SB. Bul- Indian J Pediatr. 2011;78(3):307–310 of relational adjustment, prevalence of vic- lying in schools: prevalence and short-term 41. Richter M, Bowles D, Melzer W, Hurrelmann timization, and characteristics of identiﬁed impact. Indian Pediatr. 2007;44(1):25–28 K. Bullying, psychosoziale Gesundheit und victims. Psychol Assess. 2002;14(1):74–96 34. Lien L, Green K, Welander-Vatn A, Bjertness Risikoverhalten im Jugendalter [Bullying, E. Mental and somatic health complaints psychosocial health and risk behaviour in 48. Crick NR, Bigbee MA. Relational and overt associated with school bullying between adolescence]. Gesundheitswesen. 2007;69 forms of peer victimization: a multiinfor- 10th and 12th grade students: results from (8–9):475–482 mant approach. J Consult Clin Psychol. cross sectional studies in Oslo, Norway. 42. Schnohr C, Niclasen BV. Bullying among 1998;66(2):337–347 Clin Pract Epidemiol Ment Health. 2009;5:6 Greenlandic schoolchildren: development 49. Reijntjes AHA, Kamphuis JH, Prinzie P, Boe- 35. Løhre A, Lydersen S, Paulsen B, Mæhle M, since 1994 and relations to health and len PA, van der Schoot M, Telch MJ. Pro- Vatten LJ. Peer victimization as reported by health behaviour. Int J Circumpolar Health. spective linkages between peer victimization children, teachers, and parents in relation 2006;65(4):305–312 and externalizing problems in children: to children’s health symptoms. BMC Public 43. Srabstein JC, McCarter RJ, Shao C, Huang a meta-analysis. Aggress Behav. 2011;37(3): Health. 2011;11:278 ZJ. Morbidities associated with bullying 215–222

A Message From the Editor of Pediatrics: In its January 2013 issue, Pediatrics published a case report entitled “Lethal Effect of a Single Dose of Rasburicase in a Preterm Newborn Infant.” The authors included two physicians (Patrizia Zar- amelia, MD, and Alessandra DeSalvia, PhD, MD), who disclosed that they served as paid expert witnesses in the case reported in this article. Although the authors of the case report refer to the infant as “our patient,” Pediatrics has since learned that none of the authors of the case report treated the infant who is the subject of the case report. Pediatrics has also learned that this case is the subject of pending criminal and civil proceedings in Italy, that Drs. Zaramelia and DeSalvia were appointed as expert witnesses for the prosecution in the criminal proceedings and that the medical conclusions in the article are being contested by the opposing parties and their experts.

PEDIATRICS Volume 132, Number 4, October 2013 729