Citrus Fruit Intake Substantially Reduces the Risk of Esophageal Cancer

Many epidemiologic studies indicate a potential association between fruit and vegetable intake and various cancers. The purpose of this meta-analysis is to investigate the association between citrus fruit intake and esophageal cancer risk. The authors conducted a comprehensive search on PubMed, EMBASE, and the Cochrane Library from inception until July 2014. Studies presenting information about citrus intake and esophageal cancer were analyzed. The authors extracted the categories of citrus intake, study-specific odds ratio or relative risk, and the P value and associated 95% confidence intervals for the highest versus lowest dietary intake of citrus fruit level. The association was quantified using meta-analysis of standard errors with a random-effects model. Thirteen case–control studies and 6 cohort studies were eligible for inclusion. Citrus intake may significantly reduce risk of esophageal cancer (summary odds ratio1⁄4 0.63; 95% confidence interval1⁄4 0.52– 0.75; P1⁄4 0), without notable publication bias (intercept1⁄4 0.79, P1⁄4 0.288) and with significant heterogeneity across studies (I1⁄4 52%). The results from epidemiologic studies suggest an inverse association between citrus fruit intake and esophageal cancer risk. The significant effect is consistent between case–control and cohort studies. Larger prospective studies with rigorous methodology should be considered to validate the association between citrus fruits and esophageal cancer. (Medicine 94(39):e1390) Abbreviations: CIs = confidence intervals, EAC = esophageal adenocarcinoma, OR = odds ratio, RR = relative risk, SCC = squamous cell carcinoma, SEs = standard errors. , MD, Xueshuai W o Yang, MD, D, Xinting Sang, MD, and Haitao Zhao, MD malignancy with a poor prognosis in the majority of cases. SCC is the predominant form of esophageal carcinoma worldwide, but a shift in epidemiology has been seen in some countries and regions like Australia, UK, USA, and western Europe, where the incidence of EAC has exceeded that of SCC. Every year, >450,000 people worldwide are diagnosed with esophageal cancer and the incidence is rapidly increasing. It is the eighth most common cancer, and the sixth most common cause of cancer-related deaths worldwide with developing nations making up >80% of total cases and deaths. The mortality from these cancers is high and the response to treatments during advanced stages is poor, so effectively reducing the chances of exposure to relative risk factors will have an important impact on the incidence of esophageal cancer. Cigarettes, red meat, alcohol, hot tea, pickled vegetables, low intake of fresh fruits and vegetables, and low socioeconomic status are associated with a higher risk of SCC. Barrett esophagus is clearly recognized as a risk factor for EAC, with other factors including gastroesophageal reflux disease, acid-suppressive medication use, obesity, tobacco use, and processed meat. Some foods can reduce the incidence of esophageal cancer. Many researchers conducted metaanalyses on diet and esophageal cancer. The study by Coleman et al suggested that dietary fiber may protect against esophageal carcinogenesis, especially esophageal adenocarcinoma. Zhu et al found that meat consumption is associated with the risk of esophageal cancer. The intake of red meat is likely to increase the esophageal SCC risk and the processed meat may increase esophageal adenocarcinoma risk; however, the consumption of fish may not be associated with esophageal cancer incidence. This phenomenon may be explained by the effects of various micronutrients such as folate, B vitamins, antioxidants, lutein, and carotenoids. Citrus fruits include oranges, tangerines, grapefruits, lemons, and limes. They include several components, including flavonoids, folate, carotenoids, and vitamin C, which have protective effects against cancer. Previous studies have suggested that citrus intake may improve the incidence of various cancers including pancreatic, breast, and prostate cancers. Consequently, we hypothesize that citrus intake is associated with a reduced risk of esophageal cancer. Epidecohort and case–control studies on this association has not yet been summarized. Therefore, we conducted a meta-analysis to explore this hypothesis. STUDY CHARACTERISTICS earch of the English language literature sophageal cancer yielded no relevant www.md-journal.com | 1 A total of 19 articles were included in the meta-analysis, including 6 cohort studies and 13 case–control studies (Figure 1). TABLE 1. Logarithmic OR or RR (Log[OR/RR]) and Its SE for the Meta-Analysis Author Categories Log (OR/RR) SE Tuyns 1983 Total S0.4780358 0.1101486 Brown 1988 Men S0.6931472 0.2802582 Cheng et al 1992 Total S2.419119 0.6944997 Castelletto, 1994 Total 0.4700036 0.3455474 Brown et al 1995 35 Men S0.3566749 1.2199227 Zhang 1997 Total S0.1053605 0.1972423 Brown et al 1998 18 Men S0.2231435 1.97403993 Launoy 1998 Men S0.6161861 0.2530941 Levi 2000 Total S1.514128 0.4570815 Bosetti 2000 Total S0.8675005 0.2662765 Chen 2002 Total S0.7339692 0.4224383 De Stefani 2005 Total S1.272966 0.3314497 Sapkota 2008 Total S0.2744368 0.2610686 Boeing 2006 Total S0.2744368 0.2029495 Gonzalez 2006 Total S0.3147107 0.3205151 Freedman 2007 Total S0.2484614 0.2482107 Yamaji 2008 Men S0.2484614 0.2441614 Li et al 2010 Total S0.3424903 0.2531607 Steevens 2011 Men S0.2231435 0.2729166 publications from inception to July 2014. We, therefore, decided to use the key words ‘‘fruit’’ and ‘‘citrus.’’ The search terms were ([esophagus] OR [esophageal]) AND ([cancer] OR [tumor] OR [carcinoma]) AND (‘citrus’itrus OR ‘fruit’ruits). We limited the search to human adults without language restrictions. We searched the 3 major electronic databases: PubMed, EMBASE, and The Cochrane Library. Additionally, we reviewed the references from retrieved articles for additional studies. Furthermore, ethical approval was not necessary because our article is a review. Study Selection The included studies had to be epidemiologic studies such as case–control and cohort studies. The studies concerning human that addressed the association between citrus intake and incidence of esophageal cancer were collected; however, if the study provides no original data or insufficient information on the odds ratio (OR) or relative risk (RR), and their corresponding 95% confidence intervals (CIs), we excluded it. The studies not measuring the intake of citrus fruits or citrus juice at the individual level are not eligible. The instrument of assessment of citrus intake is questionnaire. Two independent reviewers read the abstracts or full-text articles to assess the eligibility of studies in a standardized manner. We resolved the disagreement by consensus. Data Abstraction We extracted important information from all eligible studies. They included study design, country of origin, years of publication, origin of control, number of cases and control, sex distribution, types of citrus fruits, types of cancer, comparison of exposure level, and potential confounding variables adjusted. The estimates of OR/RR, their associated 95% CIs, and P values were also extracted by us. If separate researches based on the same population were published, we selected the article containing more complete information for inclusion. Statistical Analyses We extracted the study specific OR/RR and 95% CIs for highest versus lowest intake of citrus fruits from every study. And we calculated the standard error (SE) of the log OR/RR by using the following equation: SE1⁄4 (ln[OR/RR_upper ln OR/ RR_lower]) 3.92. Then, we summarized the overall OR and CI by using general variance-based methods of RevMan 5.0. For studies that provided OR/RR by cancer subtypes, we used a random-effects model to obtain a pooled estimate from the individual study (Table 1). We adopted the NewcastleOttawa Scale to evaluate research quality and defined them as high, middle, and low quality by score 7 to 9, 4 to 6, 1 to 3, respectively. The Grades of Recommendation, Assessment, Development, and Evaluation working group system of rating quality of evidence also were used to evaluate the research quality. The value of I was used to evaluate the extent of heterogeneity derived from study differences rather than chance. The smaller value I suggested less obvious heterogeneity. We used the random-effects model to calculate the summary OR and its 95% CI with suspecting heterogeneity. We evaluated the impact of the changes on pooled ORs by study design, cancer subtypes, geographical location, source of controls, research quality, and some adjusted confounders such as alcohol Wang et al and body mass index as prior hypotheses to explain heterogeneity through subgroup analyses and meta-regression analyses. Sensitivity analyses were conducted by removing 1 study from 2 | www.md-journal.com all studies to evaluate the impact on the pooled ORs and heterogeneity. We can, therefore, evaluate whether the results are stable. In an attempt to detect publication bias, we visually examined asymmetry in a funnel plot. We conducted Begg and Egger test to assess whether there is an obvious publication. We considered the funnel plot to be asymmetrical if the intercept of the regression line deviated from zero with P< 0.10. If the test suggests an obvious publication bias, we would conduct the trim and fill analysis to further verify. We used the Cochrane Collaboration software (Oxford, UK) to analyze the extracted data with fixed or random-effects model analysis. STATA (StataCorp, College Station, TX) was used to conduct the Egger and Begg regression asymmetry test by using the metabias command. We conducted the trim and fill analysis to observe whether the results are stable and evaluate the publication bias. RESULTS Search Results The computerized search yielded 433 references, of which 112 were included after abstract review. Citation search identified another 715 articles. Of the 827 articles that were obtained for full-text review, we excluded 808 articles based on the exclusion criteria. In particular, the result of Tuyns et al published in 1987 was replaced by Tuyns et al published in 1983, as it shared the same database. The result of De Stefani et al published in 2003 was replaced by De Stefani et al published in 2005, as the latter expanded the sample size based on the former population. Medicine Volume 94, Number 39, October 2015 OR1⁄4 odds ratio, RR1⁄4 relative risk; SE1⁄4 standard error. The estimate was obtained by fixed-effect model using the hazard ratio given by sex control. Copyright # 2015 Wolters Kluwer Health, Inc. All rights reserved. Medicine Volume 94, Number 39, October 2015 Citrus Fruit Intake Substantially Reduces the Risk of Esophageal Cancer Study Characteristics Some details of the selected studies are shown in Tables 2 and 3. All articles were published in English. Six studies were conducted among residents of the United States, 1 in Italy, 2 in Japan, 2 in France, 3 in Europe, and the remaining 5 in China, Argentina, Switzerland, Uruguay, and the Netherlands. Two of the studies recruited participants in the 1980s, 5 in the 1990s, and 12 between 2000 and 2011. The factor of age was adjusted in all of the studies except Brown et al The confounding variables that were adjusted in different studies were presented in detail in Tables 2 and 3. For all of the studies, the relationship between intake of citrus fruits and esophageal cancer was not primary hypothesis and the citrus fruits were often included in a broader dietary evaluation. The ranges of adjusted ORs/RRs were from 0.089 to 1.6 and only 5 studies reached the usual threshold of P1⁄4 0.05 in the association between citrus fruits and esophageal cancer. Heterogeneity and Pooled Results There was no significant heterogeneity among the study results (I1⁄4 52%; P1⁄4 0.005). Overall summary OR using the random-effects model showed a 37%, statistically significant reduction in risk of esophageal cancer associated with citrus FIGURE 1. Flowchart of the searching and review of literatures. fruits intake (summary OR1⁄4 0.63; 95% CI1⁄4 0.52–0.75). The subgroup of case–control studies (summary OR1⁄4 0.54; 95% CI1⁄4 0.4–0.72; I1⁄4 64.2%; P1⁄4 0.001) and the subgroup of Copyright # 2015 Wolters Kluwer Health, Inc. All rights reserved. cohort studies (summary OR1⁄4 0.76; 95% CI1⁄4 0.62–0.93; I1⁄4 0%; P1⁄4 1) showed a respective 46% and 24% statistically significant reduction in risk of esophageal cancer associated with citrus fruits intake (Figure 2). In subgroup analyses defined by study type, cancer subtype, geographical location, source of controls, research quality, and adjusted confounders, citrus intake was inversely associated with risk of esophageal cancer in most subgroups, with no evidence of significant heterogeneity between subgroups with meta-regression analyses. (Table 4). Publication Bias No publication bias was observed in the selected studies. Visualization of Begg funnel plot was symmetrical (Figure 3). Formal testing using the Egger method supports the notion that there was no publication bias (intercept1⁄4 0.79, P1⁄4 0.288); however, the result of Begg test suggested an obvious publication bias (P1⁄4 0.046). And the outcome of trim and fill analysis demonstrated that there was no publication bias. DISCUSSION The overall summary OR in our study presents an inverse association between citrus fruits and esophageal cancer (summary OR1⁄4 0.63; 95% CI1⁄4 0.52–0.75; P1⁄4 0). The result is supported by the strengths of our review, which includes a systematic literature search, strict selection criteria, comprehensive data abstraction, and rigorous statistical analysis. Additionally, the results of similar reviews about the association between citrus fruits and other cancers are encouraging. www.md-journal.com | 3 T A B L E 2 . S u m m a ry o f C a se – C o n tr o l S tu d ie s In cl u d e d in th e M e ta -A n a ly si s S tu d y/ Y ea rs of P u b li ca ti on C ou n tr y N o. of C as e/ C on tr ol S ou rc es of C on tr ol s T yp es of C an ce r E xp os u re A d ju st ed F ac to rs C om p ar is on of E xp os u re L ev el A d ju st ed O R /R R (9 5% C I) P va lu e N O S S ca le T u y n s 1 9 8 3 3 2 F ra n ce 1 2 4 6 /1 9 7 6 P o p u la ti o n E C C it ru s an d ju ic es A g e, al co h o l co n su m p ti o n , to b ac co sm o k in g , an d u rb an o r ru ra l re si d en ce Y es v s n o 0 .6 2 [0 .5 0 – 0 .7 7 ] — 5 B ro w n et al 1 9 8 8 3 3 U n it ed S ta te s 2 0 7 /4 2 2 H o sp it al E C ci tr u s an d ju ic es u se o f ci g ar et te s an d al co h o l Q 3 v s Q 1 0 .5 [0 .2 9 – 0 .8 7 ] 0 .0 1 5 C h en g et al 1 9 9 2 1 0 H o n g K o n g 4 0 0 /1 5 9 8 P o p u la ti o n E C C it ru s fr u it s A g e, ed u ca ti o n al at ta in m en t, p la ce o f b ir th , h o t d ri n k s o r so u p s, g re en le af v eg et ab le s, p ic k le d v eg et ab le s, to b ac co , al co h o l, w h et h er h ad m ea ls o r ea ti n g o u t Q 6 v s Q 1 0 .0 8 9 [0 .0 2 – 0 .3 5 ] — 7 C as te ll et to et al 1 9 9 4 3 4 A rg en ti n a 1 3 1 /2 6 2 H o sp it al S C C C it ru s fr u it s D es ig n v ar ia b le s, ag e, se x , h o sp it al , ed u ca ti o n , av er ag e n u m b er o f ci g ar et te s/ d ay , al co h o l co n su m p ti o n b ar b ec u ed m ea t, p o ta to es , ra w v eg et ab le s, co o k ed v eg et ab le s Q 3 v s Q 1 1 .6 [0 .8 – 3 .1 ] — 6 B ro w n et al 1 9 9 5 3 5 U n it ed S ta te s 1 7 4 /7 5 0 P o p u la ti o n E A C C it ru s fr u it s A g e, ar ea , sm o k in g , li q u o r u se , in co m e, ca lo ri es fr o m fo o d , an d B M I Q 4 v s Q 1 0 .7 [0 .0 6 – 7 .6 5 ] 0 .7 7 7 Z h an g et al 1 9 9 7 3 6 U n it ed S ta te s 9 5 /1 3 2 H o sp it al E A C C it ru s fr u it s A g e, se x , ra ce , ed u ca ti o n , to ta l d ie ta ry in ta k e o f ca lo ri es , p ac k -y ea rs o f sm o k in g , al co h o l u se , an d B M I Q 4 v s Q 1 0 .9 [0 .6 1 – 1 .3 2 ] 0 .5 3 2 8 B ro w n et al 1 9 9 8 1 8 U n it ed S ta te s 1 1 4 /6 8 1 P o p u la ti o n S C C C it ru s fr u it s A g e, ar ea , sm o k in g , al co h o l, an d fo o d ca lo ri es Q 4 v s Q 1 0 .8 [0 .0 2 – 3 8 .3 2 ] 0 .9 1 7 L au n o y et al 1 9 9 8 3 7 F ra n ce 2 0 8 /3 9 9 H o sp it al S C C C it ru s fr u it s A g e, in te rv ie w er , sm o k in g , b ee r, an is ee d ap er it if s, h o t C al v ad o s, w h is k y , to ta l al co h o l, an d to ta l en er g y in ta k e Q 4 v s Q 1 0 .5 4 [0 .3 3 – 0 .8 9 ] < 0 .0 5 6 B o se tt i et al 2 0 0 0 3 8 It al y 3 0 4 /7 4 3 H o sp it al S C C C it ru s fr u it s A g e, se x , ar ea o f re si d en ce , ed u ca ti o n , to b ac co sm o k in g ,a lc o h o l, d ri n k in g ,a n d n o n al co h o l en er g y Q 5 v s Q 1 0 .4 2 [0 .2 5 – 0 .7 1 ] — 7 L ev i et al 2 0 0 0 1 7 S w it ze rl an d 1 0 1 /3 2 7 H o sp it al E C C it ru s fr u it s A g e, se x , E d u ca ti o n , S m o k in g , al co h o l, an d n o n al co h o l to ta l en er g y in ta k e Q 3 v s Q 1 0 .2 2 [0 .0 9 – 0 .5 4 ] — 7 C h en et al 2 0 0 2 3 9 U n it ed S ta te s 1 2 4 /4 4 9 P o p u la ti o n E A C C it ru s fr u it s A g e, se x , en er g y in ta k e, re sp o n d en t ty p e, B M I, al co h o l u se , to b ac co u se , ed u ca ti o n Q 4 v s Q 1 0 .4 8 [0 .2 1 – 1 .1 0 ] 0 .0 3 7 D e S te fa n i et al 2 0 0 5 4 0 U ru g u ay 2 0 0 /4 0 0 H o sp it al S C C ci tr u s fr u it s A g e, se x , re si d en ce , u rb an /r u ra l st at u s, b ir th p la ce , ed u ca ti o n , sm o k in g st at u s, n u m b er o f ci g ar et te s sm o k ed p er d ay , y ea rs si n ce q u it , al co h o l d ri n k in g , m at e te m p er at u re , st ew ed m ea t, an d to ta l en er g y in ta k e Q 4 v s Q 1 0 .2 8 [0 .1 5 – 0 .5 4 ] 0 .0 0 0 1 6 S ap k o ta et al 2 0 0 8 9 Ja p an 1 1 6 / H o sp it al S C C C it ru s fr u it s A g e, ci g ar et te sm o k in g , an d al co h o l d ri n k in g Q 3 v s Q 1 0 .7 8 [0 .4 8 – 1 .2 6 ] 0 .2 1 6 B M I1⁄4 b o d y m as s in d ex , C I1⁄4 co n fi d en ce in te rv al , E A C 1⁄4 es o p ha g ea l ad en o ca rc in o m a, E C 1⁄4 es o p ha g ea l ca rc in o m a, O R 1⁄4 o d d s ra ti o , R R 1⁄4 re la ti v e ri sk , S C C 1⁄4 sq u am o us ce ll ca rc in o m a. Wang et al Medicine Volume 94, Number 39, October 2015 4 | www.md-journal.com Copyright # 2015 Wolters Kluwer Health, Inc. All rights reserved. T A B L E 3 . S u m m a ry o f C o h o rt S tu d ie s In cl u d e d in th e M e ta -A n a ly si s S tu d y/ Y ea rs of P u b li ca ti on C ou n tr y N o. of C as e/ p er so n -y ea rs S ou rc es of C on tr ol s S u b ty p e of C an ce r E xp os u re A d ju st ed F ac to rs C om p ar is on of E xp os u re L ev el A d ju st ed O R /R R (9 5% C I) P va lu e N O S S ca le B o ei n g et al 2 0 0 6 4 1 E u ro p ea n 3 5 2 /2 ,1 8 2 ,5 6 0 P o p u la ti o n S C C C it ru s fr u it s A g e, se x , ce n te r, B M I, en er g y fr o m fa t so u rc es , en er g y fr o m n o n fa t so u rc es , ed u ca ti o n , sm o k in g st at u s ca te g o ri es Q 5 v s Q 1 0 .7 6 [0 .5 1 – 1 .1 3 ] 0 .1 2 9 8 G o n za le z et al 2 0 0 6 4 2 E u ro p ea n 6 5 /3 ,1 1 0 ,0 3 4 P o p u la ti o n E A C C it ru s fr u it s S ex , h ei g h t, w ei g h t, ed u ca ti o n le v el , to b ac co sm o k in g , ci g ar et te sm o k in g in te n si ty , w o rk an d le is u re , p h y si ca l ac ti v it y , al co h o l in ta k e, en er g y in ta k e, re d m ea t in ta k e, an d p ro ce ss ed m ea t in ta k e Q 3 v s Q 1 0 .7 3 [0 .3 9 – 1 .3 7 ] 0 .2 2 8 F re ed m an et al 2 0 0 7 3 1 U n it ed S ta te s 1 0 3 /2 ,1 9 3 ,7 5 1 P o p u la ti o n S C C C it ru s fr u it s S ex , ag e at en tr y in to co h o rt , B M I, ed u ca ti o n , al co h o l in ta k e, ci g ar et te sm o k ed o se , v ig o ro u s p h y si ca l ac ti v it y , u su al ac ti v it y th ro u g h o u t th e d ay , an d to ta l en er g y Q 3 v s Q 1 0 .5 8 [0 .3 4 – 0 .9 9 ] 0 .0 4 6 7 F re ed m an et al 2 0 0 7 3 1 U n it ed S ta te s 2 3 3 /2 ,1 9 3 ,7 5 1 P o p u la ti o n E A C C it ru s fr u it s S ex , ag e at en tr y in to co h o rt , B M I, ed u ca ti o n , al co h o l in ta k e, ci g ar et te sm o k ed o se , v ig o ro u s p h y si ca l ac ti v it y , u su al ac ti v it y th ro u g h o u t th e d ay , an d to ta l en er g y Q 3 v s Q 1 0 .9 6 [0 .6 9 – 1 .3 5 ] > 0 .0 5 7 Y am aj i et al 2 0 0 8 1 6 E u ro p ea n 1 1 6 /2 9 7 ,6 5 1 P o p u la ti o n S C C C it ru s fr u it s A g e, al co h o l, sm o k in g Q 3 v s Q 1 0 .7 8 [0 .4 8 – 1 .2 5 ] 0 .2 1 9 L i et al 2 0 1 0 4 3 Ja p an 1 5 1 /3 2 9 ,9 8 5 p o p u la ti o n E C C it ru s ag e (c o n ti n u o u s v ar ia b le ), se x (f o r to ta l p ar ti ci p an ts ), jo b st at u s, y ea r o f ed u ca ti o n , B M I, p h y si ca l ac ti v it y , en er g y in ta k e, ci g ar et te , al co h o l Q 3 v s Q 1 0 .7 1 [0 .4 3 – 1 .1 6 ] 0 .1 7 8 9 S te ev en s et al 2 0 1 1 1 5 N et h er la n d s 1 0 1 /5 0 ,7 8 5 P o p u la ti o n S C C C it ru s fr u it s A g e (y ea rs ), se x , ci g ar et te sm o k in g (c u rr en t sm o k in g [y es /n o ]) , fr eq u en cy (n u m b er o f ci g ar et te s p er d ay ), d u ra ti o n (n u m b er o f y ea rs )] , al co h o l co n su m p ti o n (g et h an o l/ d ay ), to ta l v eg et ab le in ta k e an d al l o th er fr u it s, co n su m p ti o n o f re d m ea t (g /d ay ), co n su m p ti o n o f fi sh (g /d ay ) Q 5 v s Q 1 0 .5 4 [0 .2 7 – 1 .0 7 ] 0 .3 8 9 S te ev en s et al 2 0 1 1 1 5 N et h er la n d s 1 4 4 /5 0 ,7 8 5 P o p u la ti o n E A C C it ru s fr u it s A g e, se x , ci g ar et te sm o k in g , fr eq u en cy , d u ra ti o n , al co h o l co n su m p ti o n , to ta l v eg et ab le in ta k e an d al l o th er fr u it s, co n su m p ti o n o f re d m ea t, co n su m p ti o n o f fi sh Q 5 v s Q 1 0 .9 7 [0 .9 0 – 1 .0 4 ] 0 .3 7 9 B M I1⁄4 b o d y m as s in d ex , C I1⁄4 co n fi d en ce in te rv al , E A C 1⁄4 es o p ha g ea l ad en o ca rc in o m a, E C 1⁄4 es o p h ag ea l ca rc in o m a, O R 1⁄4 o d d s ra ti o , R R 1⁄4 re la ti v e ri sk , S C C 1⁄4 sq u am o u s ce ll ca rc in o m a o f es o p ha g u s. Medicine Volume 94, Number 39, October 2015 Citrus Fruit Intake Substantially Reduces the Risk of Esophageal Cancer Copyright # 2015 Wolters Kluwer Health, Inc. All rights reserved. www.md-journal.com | 5 NOTE: Weights are from random effects analysis . . Overall (I−squared = 52.0%, p = 0.005) Tuyns, A. J.1983