Developmental Fluoride Neurotoxicity

7/30/2019 Developmental Fluoride Neurotoxicity

1/7

1362 volume 120 |number 10 | October 2012 Environmental Health Perspectives

Review

A recent report rom the National ResearchCouncil (NRC 2006) concluded that adverse

eects o high luoride concentrations indrinking water may be o concern and thatadditional research is warranted. Fluoride maycause neurotoxicity in laboratory animals,including eects on learning and memory(Chioca et al. 2008; Mullenix et al. 1995). Arecent experimental study where the rat hip-pocampal neurons were incubated with vari-ous concentrations (20 mg/L, 40 mg/L, and80 mg/L) o sodium uoride in vitro showedthat luoride neurotoxicity may target hip-pocampal neurons (Zhang M et al. 2008).Although acute luoride poisoning may beneurotoxic to adults, most o the epidemio-

logical inormation available on associationswith childrens neurodevelopment is romChina, where luoride generally occurs indrinking water as a natural contaminant, andthe concentration depends on local geologi-cal conditions. In many rural communitiesin China, populations with high exposure touoride in local drinking-water sources mayreside in close proximity to populations with-out high exposure (NRC 2006).

Opportunities or epidemiological stud-ies depend on the existence o comparablepopulation groups exposed to dierent levels

o luoride rom drinking water. Such cir-cumstances are diicult to ind in many

industrialized countries, because uoride con-centrations in community water are usuallyno higher than 1 mg/L, even when uorideis added to water supplies as a public healthmeasure to reduce tooth decay. Multiple epi-demiological studies o developmental uo-ride neurotoxicity were conducted in Chinabecause o the high uoride concentrationsthat are substantially above 1 mg/L in wellwater in many rural communities, althoughmicrobiologically sae water has been acces-sible to many rural households as a result othe recent 5-year plan (20012005) by theChinese government. It is projected that all

rural residents will have access to sae publicdrinking water by 2020 (World Bank 2006).However, results o the published studies havenot been widely disseminated. Four studiespublished in English (Li XS et al. 1995; Luet al. 2000; Xiang et al. 2003; Zhao et al.1996) were cited in a recent report rom theNRC (2006), whereas the World HealthOrganization (2002) has considered only two(Li XS et al. 1995; Zhao et al. 1996) in itsmost recent monograph on uoride.

Fluoride readily crosses the placenta(Agency or oxic Substances and Disease

Registry 2003). Fluoride exposure to the devel-

oping brain, which is much more susceptibleto injury caused by toxicants than is the maturebrain, may possibly lead to permanent damage(Grandjean and Landrigan 2006). In responseto the recommendation o the NRC (2006),the U.S. Department o Health and HumanServices (DHHS) and the U.S. EPA recentlyannounced that DHHS is proposing to changethe recommended level o uoride in drinkingwater to 0.7 mg/L rom the currently recom-mended range o 0.71.2 mg/L, and the U.S.EPA is reviewing the maximum amount ouoride allowed in drinking water, which cur-rently is set at 4.0 mg/L (U.S. EPA 2011).

o summarize the available literature,we perormed a systematic review and meta-analysis o published studies on increaseduoride exposure in drinking water associatedwith neurodevelopmental delays. We specif-cally targeted studies carried out in ruralChina that have not been widely disseminated,thus complementing the studies that havebeen included in previous reviews and riskassessment reports.

Methods

Search strategy. We sea rched MEDLINE(National Library o Medicine, Bethesda, MD,USA; http://www.ncbi.nlm.nih.gov/pubmed),

Embase (Elsevier B.V., Amsterdam, theNetherlands; http://www.embase.com), WaterResources Abstracts (Proquest, Ann Arbor,MI, USA; http://www.csa.com/actsheets/water-resources-set-c.php), and OXNE(oxicology Data Network; National Libraryo Medicine, Bethesda, MD, USA; http://tox-net.nlm.nih.gov) databases to identiy studieso drinking-water luoride and neurodevel-opmental outcomes in children. In addition,we searched the China National KnowledgeInrastructure (CNKI; Beijing, China; http://www.cnki .ne t) database to identiy stud-ies published in Chinese journals only. Key

Address correspondence to A.L. Choi, Departmento Environmental Health, Harvard School o PublicHealth, Landmark Center 3E, 401 Park Dr., Boston,MA 02215 USA. elephone: (617) 384-8646. Fax:(617) 384-8994. E-mail: [email protected]

Supplemental Material is available online (http://dx.doi.org/10.1289/ehp.1104912).We thank V. Malik, Harvard School o Public Health,

or the helpul advice on the meta-analysis methods.Tis study was supported by internal institutional

unds.Te authors declare they have no actual or potential

competing fnancial interests.Received 30 December 2011; accepted 20 July 2012.

Developmental Fluoride Neurotoxicity: A Systematic Review and Meta-Analysis

Anna L. Choi,1 Guifan Sun,2Ying Zhang,3and Philippe Grandjean1,4

1Department of Environmental Health, Harvard School of Public Health, Boston, Massachusetts, USA; 2School of Public Health, ChinaMedical University, Shenyang, China; 3School of Stomatology, China Medical University, Shenyang, China; 4Institute of Public Health,University of Southern Denmark, Odense, Denmark

Background: Although uoride may cause neurotoxicity in animal models and acute uoridepoisoning causes neurotoxicity in adults, very little is known o its eects on childrens neuro-development.

oBjective: We perormed a systematic review and meta-analysis o published studies to investigatethe eects o increased uoride exposure and delayed neurobehavioral development.

Methods: We searched the MEDLINE, EMBASE, Water Resources Abstracts, and OXNEdatabases through 2011 or eligible studies. We also searched the China National KnowledgeInrastructure (CNKI) database, because many studies on uoride neurotoxicity have been pub-lished in Chinese journals only. In total, we identifed 27 eligible epidemiological studies with highand reerence exposures, end points o IQ scores, or related cognitive unction measures with meansand variances or the two exposure groups. Using random-eects models, we estimated the stan-dardized mean dierence between exposed and reerence groups across all studies. We conductedsensitivity analyses restricted to studies using the same outcome assessment and having drinking-

water uoride as the only exposure. We perormed the Cochran test or heterogeneity between stud-ies, Beggs unnel plot, and Egger test to assess publication bias, and conducted meta-regressions to

explore sources o variation in mean dierences among the studies.results: Te standardized weighted mean dierence in IQ score between exposed and reerencepopulations was 0.45 (95% confdence interval: 0.56, 0.35) using a random-eects model.Tus, children in high-uoride areas had signifcantly lower IQ scores than those who lived in low-uoride areas. Subgroup and sensitivity analyses also indicated inverse associations, although thesubstantial heterogeneity did not appear to decrease.

conclusions: Te results support the possibility o an adverse eect o high uoride exposure onchildrens neurodevelopment. Future research should include detailed individual-level inormationon prenatal exposure, neurobehavioral perormance, and covariates or adjustment.

keywords: luoride, intelligence, neurotoxicity. Environ Health Perspect120:13621368(2012). http://dx.doi.org/10.1289/ehp.1104912[Online 20 July 2012]


2/7

Fluoride neurotoxicity

Environmental Health Perspectives volume 120 |number 10 | October 2012 1363

words included combinations o uoride ordrinking water uoride, children, neu-rodevelopment or neurologic or intelli-gence or IQ. We also used reerences citedin the articles identifed. We searched recordsor 19802011. Our literature search iden-tiied 39 studies, among which 36 (92.3%)were studies with high and reerence expo-

sure groups, and 3 (7.7%) studies were basedon individual-level measure o exposures. Telatter showed that dose-related deicits wereound, but the studies were excluded becauseour meta-analysis ocused on studies with thehigh- and low-exposure groups only. In addi-tion, two studies were published twice, andthe duplicates were excluded.

Inclusion criteria and data extraction .Te criteria or inclusion o studies includedstudies with high and reerence uoride expo-sures, end points o IQ scores or other relatedcognitive unction measures, presentation o amean outcome measure, and associated mea-sure o variance [95% conidence intervals(CIs) or SEs and numbers o participants].Interpretations o statistical signifcance arebased on an alpha level o 0.05. Inormationincluded or each study also included the frstauthor, location o the study, year o publica-tion, and numbers o participants in high-uoride and low-uoride areas. We noted andrecorded the inormation on age and sex ochildren, and parental education and incomei available.

Statistical analysis. We used SAA(version 11.0; StataCorp, College Station, X,USA) and available commands (Stern 2009)or the meta-analyses. A standardized weighted

mean dierence (SMD) was computed usingboth fxed-eects and random-eects models.he ixed-eects model uses the MantelHaenszel method assuming homogeneityamong the studies, whereas the random-eects model uses the DerSimonian and Lairdmethod, incorporating both a within-studyand an additive between-studies component ovariance when there is between-study hetero-geneity (Egger et al. 2001). Te estimate o thebetween-study variation is incorporated intoboth the SE o the estimate o the commoneect and the weight o individual studies,which was calculated as the inverse sum o

the within and between study variance. Weevaluated heterogeneity among studies usingthe I2 statistic, which represents the percentageo total variation across all studies due tobetween-study heterogeneity (Higgins andTompson 2002). We evaluated the potentialor publication bias using Begg and Eggertests and visual inspection o a Begg unnelplot (Begg and Mazumdar 1994; Egger et al.1997). We also conducted independent meta-regressions to estimate the contribution ostudy characteristics (mean age in years romthe age range and year o publication in each

study) to heterogeneity among the studies. Tescoring standard or the Combined RavensestTe Rural edition in China (CR-RC)test classiies scores o 69 and 7079 aslow and marginal intelligence, respectively(Wang D et al. 1989). We also used therandom-eects models to estimate risk ratiosor the association between uoride exposure

and a low/marginal versus normal Ravens testscore among children in studies that used theCR-RC test (Wang D et al. 1989). Scoresindicating low and marginal intelligence ( 69and 7079, respectively) were combined asa single outcome due to small numbers ochildren in each outcome subgroup.

Results

Six o the 34 studies identifed were excludedbecause o missing inormation on the num-ber o subjects or the mean and variance o theoutcome [see Figure 1 or a study selection owchart and Supplemental Material, able S1(http://dx.doi.org/10.1289/ehp.1104912) oradditional inormation on studies that wereexcluded rom the analysis]. Another study(rivedi et al. 2007) was excluded becauseSDs reported or the outcome parameter werequestionably small (1.13 or the high-uoridegroup, and 1.23 or the low-uoride group)and the SMD (10.8; 95% CI: 11.9, 9.6)was > 10 times lower than the second small-est SMD (0.95; 95% CI: 1.16, 0.75) and150 times lower than the largest SMD (0.07;95% CI: 0.083, 0.22) reported or the otherstudies, which had relatively consistent SMDestimates. Inclusion o this study in the meta-analysis resulted with a much smaller pooled

random-eects SMD estimate and a muchlarger I2 (0.63; 95% CI: 0.83, 0.44, I294.1%) compared with the estimates thatexcluded this study (0.45; 95% CI: 0.56,0.34, I2 80%) (see Supplemental Material,Figure S1). Characteristics o the 27 studiesincluded are shown in able 1 (An et al. 1992;Chen et al. 1991; Fan et al. 2007; Guo et al.1991; Hong et al. 2001; Li FH et al. 2009; LiXH et al. 2010; Li XS 1995; Li Y et al. 1994;Li Y et al. 2003; Lin et al. 1991; Lu et al. 2000;Poureslami et al. 2011; Ren et al. 1989; Serajet al. 2006; Sun et al. 1991; Wang G et al.1996; Wang SH et al. 2001; Wang SX et al.

2007; Wang ZH et al. 2006; Xiang et al. 2003;Xu et al. 1994; Yang et al. 1994; Yao et al.1996, 1997; Zhang JW et al. 1998; Zhao et al.1996). wo o the studies included in the anal-ysis were conducted in Iran (Poureslami et al.2011; Seraj et al. 2006); the other study cohortswere populations rom China. wo cohortswere exposed to luoride rom coal burning(Guo et al. 1991; Li XH et al. 2010); otherwisepopulations were exposed to uoride throughdrinking water. he CR-RC was used tomeasure the childrens intelligence in 16 stud-ies. Other intelligence measures included the

Wechsler Intelligence tests (3 studies; An et al.1992; Ren et al. 1989; Wang ZH et al. 1996),Binet IQ test (2 studies; Guo et al. 1991; Xuet al. 1994), Ravens test (2 studies; Poureslamiet al. 2011; Seraj et al. 2006), Japan IQ test(2 studies; Sun et al. 1991; Zhang JW et al.1998), Chinese comparative intelligence test(1 study; Yang et al. 1994), and the mental

work capacity index (1 study; Li Y et al. 1994).Because each o the intelligence tests used isdesigned to measure general intelligence, weused data rom all eligible studies to estimatethe possible eects o luoride exposure ongeneral intelligence.

In addition, we conducted a sensitivityanalysis restricted to studies that used similartests to measure the outcome (specifcally, theCR-RC, Wechsler Intelligence test, Binet IQtest, or Ravens test), and an analysis restrictedto studies that used the CR-RC. We alsoperormed an analysis that excluded studieswith co-exposures including iodine and arsenic,or with non-drinking-water uoride exposurerom coal burning.

Pooled SMD estimates. Am on g the27 studies, all but one study showedrandom-eect SMD estimates that indicatedan inverse association, ranging rom 0.95(95% CI: 1.16, 0.75) to 0.10 (95% CI:

Figure 1. Flow diagram o the meta-analysis.

Total abstracts identifiedfrom literature search

(n= 39)

Duplicate recordsremoved

(n= 2)

Studies excludedbecause they did not

meet inclusion criteria(n= 3)

Studies withmissing information

on outcomes(n= 6)

Studies excluded dueto questionably small

standard deviations(n= 1)

Studies included inmeta-analysis

(n= 27)

Studies for retrieval ofdetailed information

(n= 34)


3/7

Choi et al.


0.25, 0.04) (Figure 2). he study with apositive association reported an SMD esti-mate o 0.07 (95% CI: 0.8, 0.22). Similarresults were ound with the fxed-eects SMDestimates. he ixed-eects pooled SMDestimate was 0.40 (95% CI: 0.44, 0.35),with ap-value < 0.001 or the test or homo-geneity. he random-eects SMD estimate

was 0.45 (95% CI: 0.56, 0.34) with anI2 o 80% and homogeneity test p-value< 0.001 (Figure 2). Because o heterogeneity(excess variability) between study results, weused primarily the random-eects model orsubsequent sensitivity analyses, which is gen-erally considered to be the more conserva-tive method (Egger et al. 2001). Among therestricted sets o intelligence tests, the SMDor the model with only CR-RC tests anddrinking-water exposure (and to a lesserextent the model with only CR-RC tests)was lower than that or all studies combined,

although the dierence did not appear to besignifcant. Heterogeneity, however, remainedat a similar magnitude when the analyses wererestricted (able 2).

Sources of heterogeneity. We perormedmeta-regression models to assess study char-acteristics as potential predictors o eect.Inormation on the childs sex and paren-

tal education were not reported in > 80%o the studies, and only 7% o the studiesreported household income. Tese variableswere thereore not included in the models.Among the two covariates, year o publica-tion (0.02; 95% CI: 0.006, 0.03), but notmean age o the study children (0.02; 95%CI: 0.094, 0.04), was a signifcant predictorin the model with all 27 studies included. I2residual 68.7% represented the proportiono residual between-study variation due toheterogeneity. From the adjusted R2, 39.8%o between-study variance was explained by

the two covariates. he overall test o thecovariates was signifcant (p = 0.004).

When the model was res tricted to the16 studies that used the CR-RC, the childsage (but not year o publication) was a signif-cant predictor o the SMD. Te R2 o 65.6%o between-study variance was explained bythe two covariates, and only 47.3% o the

residual variation was attributable to hetero-geneity. Te overall test o both covariates inthe model remained signifcant (p = 0.0053).On urther restriction o the model to excludethe 7 studies with arsenic and iodine as co-exposures and uoride originating rom coalburning (thus including only the 9 with uo-ride exposure rom drinking water), neitherage nor year o publication was a signifcantpredictor, and the overall test o covariates wasless important (p = 0.062), in accordance withthe similarity o intelligence test outcomes andthe source o exposure in the studies included.

Table 1. Characteristics o epidemiological studies o uoride exposure and childrens cognitive outcomes.

ReerenceStudy

location

No. in high-exposure

group

No. inreerence

group

Agerange(years)

Fluoride exposure Outcome

measure ResultsAssessment Range

Ren et al.1989

Shandong,China

160 169 814 High-/low-uoridevillages

Not specifed WechslerIntelligencetesta

Children in high-uoride region had lower IQ scores

Chen et al.1991

Shanxi,China

320 320 714 Drinkingwater

4.55 mg/L (high);0.89 mg/L (reerence)

CRT-RCb The average IQ o children rom high-uoride areawere lower than that o the reerence area

Guo et al.1991

Hunan,China

60 61 713 Fluoride incoal burning

118.11361.7 mg/kg(coal burning area);Control area used wood

ChineseBinetc

Average IQ in uoride coal-burning area was lowerthan that in the reerence area

Lin et al.1991

Xinjiang,China



CRT-RCb Children in the high-uoride (low-iodine) area hadlower IQ scores compared with the children rom thereerence uoride (low-iodine) areas

Sun et al.

1991

Guiyang,

China

196 224 6.512 Rate o

uorosis

Fluorosis: 98.36%

(high); not specifed(reerence)

Japan IQ

testdMean IQ was lower in all age groups except 7 years

in the area with high uoride and aluminum (limitedto high-uoride population only)

An et al.1992

InnerMongolia,China


2.17.6 mg/L (high);0.61.0 mg/L(reerence)

WechslerIntelligencetesta

IQ scores o children in high-uoride areas weresignifcantly lower than those o children living inreerence uoride area

Li Y et al.1994

Sichuan,China

106 49 1213 Burning ohigh-uoridecoal to cookgrain in high-uoride area

4.731.6 mg/kg (high);0.5 mg/kg (reerence)

Childmental workcapacity

Early, prolonged high uoride intake causes adecrease in the childs mental work capacity

Xu et al.1994

Shandong,China



Binet-Simone

Children had lower IQ scores in high-uoride areathan those who lived in the reerence area.

Yang et al.1994

Shandong,China

30 30 814 Well water 2.97 mg/L (high);0.5 mg/L (reerence)

Chinesecomparativeintelligencetestf

The average IQ scores was lower in children romhigh-uoride and -iodine area than those rom thereerence area, but the results were not signifcant

Li XS et al.1995

Guizhou,China

681 226 813 Urine, DentalFluorosisIndex

1.812.69 mg/L (high);1.02 mg/L (reerence);DFI 0.83.2 (high);DFI < 0.4 (reerence)

CRT-RCb Children living in uorosis areas had lower IQ scoresthan children living in nonuorosis areas

Wang Get al. 1996

Xinjiang,China


> 1.08.6 mg/L (high);0.581.0 mg/L(reerence)

WechslerIntelligencetesta

Average IQ score was lower in children in the high-uoride group than those in the reerence group

Yao et al.1996

Liaoning,China


211mg/L (high);1 mg/L (reerence)

CRT-RCb Average IQ scores o children residing in exposed uorideareas were lower than those in the reerence area

Zhao et al.1996

Shanxi,China



CRT-RCb Children living in high-uoride and -arsenic area hadsignifcantly lower IQ scores than those living in thereerence uoride (and no arsenic) area

Yao et al.1997

Liaoning,China


2 mg/L (exposed);0.4 mg/L (reerence)

CRT-RCb IQ scores o children in the high-uoride area werelower than those o children in the reerence area

Continued


4/7



Although oicial reports o lead concentra-tions in the study villages in China were notavailable, some studies reported high percent-age (95100%) o low lead exposure (lessthan the standard o 0.01 mg/L) in drinking-water samples in villages rom several studyprovinces (Bi et al. 2010; Peng et al. 2008;Sun 2010).

Publication bias. A Beggs unnel plotwith the SE o SMD rom each study plottedagainst its corresponding SMD did not showclear evidence o asymmetry, although twostudies with a large SE also reported relativelylarge eect estimates, which may be consis-tent with publication bias or heterogeneity(Figure 3). Te plot appears symmetrical orstudies with larger SE, but with substantialvariation in SMD among the more precisestudies, consistent with the heterogeneityobserved among the studies included in theanalysis. Begg (p = 0.22) and Egger (p = 0.11)

tests did not indicate signiicant (p < 0.05)departures rom symmetry.

Pooled risk ratios. he relative risk (RR)o a low/marginal score on the CR-RC test(< 80) among children with high luorideexposure compared with those with lowexposure (16 studies total) was 1.93 (95% CI:1.46, 2.55; I2 58.5%). When the model was

restricted to 9 studies that used the CR-RCand included only drinking-water luorideexposure (Chen et al. 1991; Fan et al. 2007;Li XH et al. 2010; Li XS et al. 1995; Li Yet al. 2003; Lu et al. 2000; Wang ZH et al.2006; Yao et al. 1996, 1997), the estimate wassimilar (RR = 1.75; 95% CI: 1.16, 2.65; I270.6%). Although luoride exposure showedinverse associations with test scores, theavailable exposure inormation did not allowa ormal doseresponse analysis. However,dose-related dierences in test scores occurred ata wide range o water-uoride concentrations.

DiscussionFindings rom our meta-analyses o 27 stud-ies published over 22 years suggest an inverseassociation between high uoride exposure andchildrens intelligence. Children who lived inareas with high uoride exposure had lower IQscores than those who lived in low-exposure orcontrol areas. Our fndings are consistent with

an earlier review (ang et al. 2008), althoughours more systematically addressed study selec-tion and exclusion inormation, and was morecomprehensive in a) including 9 additionalstudies, b) perorming meta-regression to esti-mate the contribution o study characteristicsas sources o heterogeneity, and c) estimatingpooled risk ratios or the association betweenuoride exposure and a low/marginal Ravenstest score.

As noted by the NRC committee (NRC2006), assessments o luoride saety haverelied on incomplete inormation on potential

Table 1. Continued.

ReerenceStudy

location

No. in high-exposure

group

No. inreerence

group

Agerange(years)

Fluoride exposure Outcome

measure ResultsAssessment Range

Zhang JWet al. 1998

Xinjiang,China


Not specifed Japan IQTestd

Average IQ scores o children residing in high-uorideand -arsenic area were lower than those who residedin the reerence area

Lu et al.2000

Tianjin,China



CRT-RCb Children in the high-uoride area scored signifcantlylower IQ scores than those in the reerence area

Hong et al.2001

Shandong,China



CRT-RCb Average IQ scores were signifcantly lower in high-uoride group (and -iodine) than the reerence group

Wang SHet al. 2001

Shandong,China



CRT-RCb No signifcant dierence in IQ scores o children inthe high-uoride/high-iodine and reerence uoride/low-iodine areas

Li Y et al.2003

InnerMongolia,China

720 236 613 Fluorosis Endemic vs. controlregions defned by theChinese GeologicalOfce

CRT-RCb Average IQ o children in high-uorosis area waslower than that in the reerence area

Xiang et al.2003

Jiangsu,China



CRT-RCb Mean IQ score was signifcantly lower in children wholived in the high-uoride area than that o childrenin the reerence exposure area (both areas also hadarsenic exposure)

Seraj et al.2006

Tehran,Iran

41 85 Notspecifed

Drinkingwater


Raveng The mean IQ o children in the high-uoride areawas signifcantly lower than that rom the reerenceuoride area

Wang ZHet al. 2006

Shanxi,China


5.54 3.88 mg/L(high); 0.73 0.28mg/L (reerence)

CRT-RCb The IQ scores o children in the high-uoride groupwere signifcantly lower than those in the reerencegroup

Fan et al2007

Shaanxi,China



CRT-RCb The average IQ scores o children residing in thehigh-uoride area were lower than those o childrenresiding in the reerence area

Wang SX

et al. 2007

Shanxi,

China

253 196 812 Drinking

water andurine

3.811.5 mg/L (water, high);

1.611 mg/L (urine, high);0.21.1 mg/L (water,reerence);0.43.9 mg/L (urine,reerence)

CRT-RCb Mean IQ scores were signifcantly lower in the high-

uoride group than rom the reerence group in theuoride/arsenic areas

Li et al.2009

Hunan,China

60 20 812 Coal burning 1.242.34 mg/L (high);0.962 mg/L (reerence)

CRT-RCb Mean IQ was lower in children in coal-burning areascompared to those in the reerence group

Li FH et al.2010

Henan,China


2.47 0.75 mg/L (high) CRT-RCb No signifcant dierence in IQ scores betweenchildren in the exposed and reerence groups

Poureslamiet al. 2011

Iran 59 60 69 DrinkingWater


Raveng Children in the high-uoride group scored signifcantlylower than those in reerence group

aWechsler Intelligence Scale (Lin and Zhang 1986). bCRT-RC, Chinese Standardized Raven Test, rural version (Wang G et al. 1989). cChinese Binet Test (Wu 1936). dJapan test (Zhang Jet al. 1985). eBinet-Simon Test (Binet and Simon 1922). fChinese comparative intelligence test (Wu 1983). gRaven test (Raven et al. 2003).


5/7

Choi et al.


risks. In regard to developmental neuro-toxicity, much inormation has in act beenpublished, although mainly as short reportsin Chinese that have not been available tomost expert committees. We carried out anextensive review that includes epidemiologicalstudies carried out in China. Although mostreports were airly brie and complete inorma-

tion on covariates was not available, the resultstended to support the potential or uoride-mediated developmental neurotoxicity at rela-tively high levels o exposure in some studies.We did not fnd conclusive evidence o publi-cation bias, although there was substantial het-erogeneity among studies. Drinking water maycontain other neurotoxicants, such as arsenic,but exclusion o studies including arsenic andiodine as co-exposures in a sensitivity analy-sis resulted in a lower estimate, although the

dierence was not signiicant. he exposedgroups had access to drinking water with uo-ride concentrations up to 11.5 mg/L (WangSX et al. 2007); thus, in many cases concen-trations were above the levels recommended(0.71.2 mg/L; DHHS) or allowed in pub-lic drinking water (4.0 mg/L; U.S. EPA) inthe United States (U.S. EPA 2011). A recent

cross-sectional study based on individual-levelmeasure o exposures suggested that low lev-els o water uoride (range, 0.242.84 mg/L)had signifcant negative associations with chil-drens intelligence (Ding et al. 2011). hisstudy was not included in our meta-analysis,which ocused only on studies with exposedand reerence groups, thereby precluding esti-mation o dose-related eects.

Te results suggest that uoride may be adevelopmental neurotoxicant that aects brain

development at exposures much below thosethat can cause toxicity in adults (Grandjean1982). For neurotoxicants such as lead andmethylmercury, adverse eects are asso-ciated with blood concentrations as low as10 nmol/L. Serum luoride concentrationsassociated with high intakes rom drinkingwater may exceed 1 mg/L, or 50 mol/L

more than 1,000 times the levels o some otherneurotoxicants that cause neurodevelopmentaldamage. Supporting the plausibility o ourfndings, rats exposed to 1 ppm (50 mol/L)o water uoride or 1 year showed morpho-logical alterations in the brain and increasedlevels o aluminum in brain tissue comparedwith controls (Varner et al. 1998).

Te estimated decrease in average IQ asso-ciated with luoride exposure based on ouranalysis may seem small and may be within themeasurement error o IQ testing. However, asresearch on other neurotoxicants has shown, ashit to the let o IQ distributions in a popu-lation will have substantial impacts, especiallyamong those in the high and low ranges o theIQ distribution (Bellinger 2007).

Our review cannot be used to derive anexposure limit, because the actual exposureso the individual children are not known.Misclassiication o children in both high-and low-exposure groups may have occurredi the children were drinking water rom othersources (e.g., at school or in the feld).

he published reports clearly representindependent studies and are not the resulto duplicate publication o the same studies(we removed two duplicates). Several studies(Hong et al. 2001; Lin et al. 1991; Wang SH

et al. 2001; Wang SX et al. 2007; Xiang et al.2003; Zhao et al. 1996) report other expo-sures, such as iodine and arsenic, a neuro-toxicant, but our sensitivity analyses showedsimilar associations between high luorideexposure and the outcomes even ater thesestudies were excluded. Large tracts o China

Figure 2. Random-eect standardized weighted mean dierence (SMD) estimates and 95% CIs o childsintelligence score associated with high exposure to uoride. SMs or individual studies are shown as soliddiamonds (), and the pooled SMD is shown as an open diamond (). Horizontal lines represent 95% CIsor the study-specifc SMDs.

Overall (I2 = 80.0%, p= 0.000)

Xiang et al. 2003

Lu et al. 2000

Li XS et al. 1995

Wang ZH et al. 2006

Li FH et al. 2009

Hong et al. 2001

Wang SX et al. 2007

Guo et al. 1991

Wang G et al. 1996

Li Y et al. 1994

Li Y et al. 2003Wang SH et al. 2001

Li XH et al. 2010

Seraj et al. 2006

Zhang JW et al. 1998

Zhao et al. 1996

Xu et al. 1994

Yao et al. 1996

Sun et al. 1991

Yao et al. 1997

Yang et al. 1994

Fan et al. 2007

Poureslami et al. 2011

Ren et al. 1989

Study

Lin et al. 1991

An et al. 1992

Chen et al. 1991

Jiangsu

Tianjin

Guizhou

Shanxi

Hunan

Shandong

Shanxi

Hunan

Xinjiang

Sichuan

I MongoliaShandong

Henan

Tehran

Xinjiang

Shanxi

Shandong

Liaoning

Guiyang

Liaoning

Shandong

Shaanxi

Iran

Shandong

Location

Xinjiang

I Mongolia

Shanxi

0.45 (0.56, 0.34)

0.64 (0.82, 0.46)

0.62 (0.98, 0.25)

0.55 (0.70, 0.39)

0.27 (0.47, 0.06)

0.43 (0.94, 0.08)

0.44 (0.85, 0.03)

0.26 (0.44, 0.07)

0.44 (0.80, 0.08)

0.38 (0.65, 0.10)

0.40 (0.74, 0.06)

0.10 (0.25, 0.04)0.50 (1.01, 0.02)

0.07 (0.08, 0.22)

0.89 (1.28, 0.50)

0.17 (0.55, 0.22)

0.54 (0.76, 0.31)

0.93 (1.35, 0.52)

0.34 (0.51, 0.17)

0.95 (1.16, 0.75)

0.43 (0.61, 0.25)

0.50 (1.01, 0.02)

0.17 (0.61, 0.27)

0.41 (0.77, 0.04)

0.75 (0.97, 0.52)

0.64 (1.01, 0.28)

0.57 (0.83, 0.31)

0.26 (0.41, 0.10)

100.00

4.52

3.20

4.68

4.34

2.38

2.94

4.46

3.26

3.88

3.39

4.712.36

4.69

3.08

3.09

4.22

2.91

4.57

4.36

4.49

2.36

2.75

3.25

4.22

% Weight

3.23

3.98

4.66

SMD (95% CI)

01.5 1 0 0.5 1

Figure 3. Beggs unnel plot showing individualstudies included in the analysis according torandom-eect standardized weighted mean dier-ence (SMD) estimates (x-axis) and the SE (se) oeach study-specifc SMD (y-axis). The solid verti-cal line indicates the pooled SMD estimate or allstudies combined and the dashed lines indicatedpseudo 95% confdence limits around the pooledSMD estimate.

0

0.05

0.10

0.15

0.20

0.25

SMD

seSMD

1 00.5

Table 2. Sensitivity analyses o pooled random-eects standardized weighted mean dierence (SMD)

estimates o childs intelligence score with high exposure o uoride.

Model

Availablestudies oranalysis SMD (95% CI) I2

p-Valuetest o

heterogeneity

1. Exclude nonstandardized testsa 23 0.44 (0.54, 0.33) 77.6% < 0.0012. Exclude nonCRT-RC Testsb 16 0.36 (0.48, 0.25) 77.8% < 0.0013. Exclude studies with other exposures (iodine, arsenic)c

or non-drinking-water uoride exposured9 0.29 (0.44, 0.14) 81.8% < 0.001

aMental work capacity (Li Y et al. 1994); Japan IQ (Sun et al. 1991; Zhang JW et al. 1998); Chinese comparative scaleo intelligence test (Yang et al. 1994). bWechsler intelligence test (An et al. 1992; Ren et al. 1989; Wang G et al. 1996);Chinese Binet IQ (Guo et al. 1991); Raven (Poureslami et al. 2011; Seraj et al. 2006); Binet-Simon (Xu et al. 1994). cIodine(Hong et al. 2001; Lin et al. 1991; Wang SH et al. 2001); arsenic [Wang SX et al. 2007; Xiang et al. 2003; Zhao et al. 1996;(Zhang JW et al. 1998 was already excluded, see note a)]. dFluoride rom coal burning [Li FH et al. 2009 (Guo et al. 1991and Li Y et al. 1994 were already excluded; see notes a and b)].


6/7



have supericial luoride-rich minerals withlittle, i any, likelihood o contamination byother neurotoxicants that would be associ-ated with uoride concentrations in drinkingwater. From the geographic distribution othe studies, it seems unlikely that luoride-attributed neurotoxicity could be attributableto other water contaminants.

Still, each o the articles reviewed haddefciencies, in some cases rather serious ones,that limit the conclusions that can be drawn.However, most deiciencies relate to thereporting o where key inormation was miss-ing. Te act that some aspects o the studywere not reported limits the extent to whichthe available reports allow a frm conclusion.Some methodological limitations were alsonoted. Most studies were cross-sectional, butthis study design would seem appropriatein a stable population where water suppliesand luoride concentrations have remainedunchanged or many years. Te current wateruoride level likely also reects past develop-mental exposures. In regard to the outcomes,the inverse association persisted between stud-ies using dierent intelligence tests, althoughmost studies did not report age adjustment othe cognitive test scores.

Fluoride has received much attention inChina, where widespread dental luorosisindicates the prevalence o high exposures.In 2008, the Ministry o Health reportedthat uorosis was ound in 28 provinces with92 million residents (China News 2008).Although microbiologically sae, water sup-plies rom small springs or mountain sourcescreated pockets o increased exposures near

or within areas o low exposures, thus repre-senting exposure settings close to the ideal,because only the uoride exposure would di-er between nearby neighborhoods. Chineseresearchers took advantage o this act andpublished their indings, though mainly inChinese journals and according to the stan-dards o science at the time. his researchdates back to the 1980s, but has not beenwidely cited at least in part because o limitedaccess to Chinese journals.

In its review o uoride, the NRC (2006)noted that the saety and the risks o uoride atconcentrations o 24 mg/L were incompletely

documented. Our comprehensive reviewsubstantially extends the scope o researchavailable or evaluation and analysis. Althoughthe studies were generally o insuicientquality, the consistency o their indingsadds support to existing evidence o uoride-associated cognitive deicits, and suggeststhat potential developmental neurotoxicityo luoride should be a high researchpriority. Although reports rom the WorldHealth Organization and national agencieshave generally ocused on beneicial eectso uoride (Centers or Disease Control and

Prevention 1999; Petersen and Lennon 2004),the NRC report examined the potentialadverse eects o luoride at 24 mg/Lin drinking water and not the beneits orpotential risks that may occur when uorideis added to public water supplies at lowerconcentrations (0.71.2 mg/L) (NRC 2006).

In conclusion, our results support the possi-

bility o adverse eects o uoride exposures onchildrens neurodevelopment. Future researchshould ormally evaluate doseresponse rela-tions based on individual-level measures oexposure over time, including more preciseprenatal exposure assessment and more exten-sive standardized measures o neurobehavioralperormance, in addition to improving assess-ment and control o potential conounders.

RefeRences

Agency or Toxic Substances and Disease Registry. 2003.

Toxicological Profle or Fluorides, Hydrogen Fluoride, and

Fluorine (Update). Available: http://www.atsdr.cdc.gov/

toxprofles/tp11.pd [accessed 5 April 2010].

An JA, Mei SZ, Liu AP, Fu Y, Wang CF. 1992. Eect o high level

o uoride on childrens intelligence [in Chinese]. Chin J

Control Endem Dis 7(2):9394.

Begg CB, Mazumdar M. 1994. Operating characteristics o

a rank correlation test or publication bias. Biometrics

50:10881101.

Bellinger DC. 2007. Interpretation o small eect sizes in occu-

pational and environmental neurotoxicity: individual ver-

sus population risk. Neurotoxicology 28:245251.

Bi WJ, Zheng X, Lan TX. 2010. Analysis on test results o drink-

ing waters quality in Janan Railway Bureau rom 2005

2009 [in Chinese]. Prev Med Trib 16(6):483485.

Binet A, Simon T. 1922. The Measurement o the Mental

Development o the Child (translated into Chinese by Jie

FP). Shanghai:Commercial Press.

Centers or Disease Control and Prevention. 1999. Achievements

in public health, 19901999: uoridation o drinking water to

prevent dental caries. MMWR 48(41):933940.

Chen YX, Han F, Zhou Z, Zhang H, Jiao X, Zhang S, et al. 1991.Research on the intellectual development o children in high

uoride areas. Chin J Control Endem Dis 6(suppl):99100.

Available: http://www.uoridealert.org/chinese/ [accessed

20 August 2012].

China News. 2008. Twenty-eight provinces were aected by

luorosis in China [in Chinese]. Available: http://news.

qq.com/a/20081216/001707.htm [accessed 3 July 2012].

Chioca LR, Raupp IM, Da Cunha C, Losso EM, Andreatini R.

2008. Subchronic luoride intake induces impairment

in habituation and active avoidance tasks in rats. Eur J

Pharmacol 579:196201.

Ding Y, Gao Y, Sun H, Han H, Wang W, Ji X, et al. 2011. The

relationships between low levels o urine uoride on chil-

drens intelligence, dental uorosis in endemic uorosis

area in Hulunbuir, Inner Mongolia, China. J Hazard Mater

186:19421946.

Egger M, Davey Smith G, Altman DG. 2001. S ystematic Reviews

in Health Care: Meta-Analysis in Context. London:BMJ

Publishing.Egger M, Davey Smith G, Schneider M, Minder C. 1997. Bias in

meta-analysis detected by a simple, graphical test. BMJ

315:629634.

Fan ZX, Dai HY, Bai AM, Li PO, Li T, LI GD, et al. 2007. Eect

o high luoride exposure in childrens intelligence [in

Chinese]. J Environ Health 24(10):802803.

Grandjean P. 1982. Occupational luorosis through 50 years:

clinical and epidemiological experiences. Am J Ind Med

3(2):227336.

Grandjean P, Landrigan P. 2006. Developmental neurotoxicity o

industrial chemicals. Lancet 368(9553):21672178.

Guo XC, Wang R, Cheng C, Wei W, Tang L, Wang Q, et al. 1991.

A preliminary exploration o IQ o 713 year old pupils in a

uorosis area with contamination rom burning coal. Chin

J Endemiol 10:98100. Available: http://www.uoridealert.

org/chinese/ [accessed 20 August 2012].

Higgins JP, Thompson SG. 2002. Quantiying heterogeneity in a

meta-analysis. Stat Med 21:15391558.

Hong F, Cao Y, Yang D, Wang H. 2001. A study o luorine

eects on childrens intelligence development under

dierent environments. Chin Prim Health Care 15:5657.


20 August 2012].

Li FH, Chen X, Huang RJ, Xie YP. 2009. Intelligence impact o

children with endemic uorosis caused by uoride rom

coal burning [in Chinese]. J Environ Health 26(4):338340.

Li XH, Hou GQ, Yu B, Yuan CS, Liu Y, Zhang L, et al. 2010.

Investigation and analysis o childrens intelligence and

dental uorosis in high uoride area [in Chinese]. J Med

Pest Control 26(3):230231.

Li XS, Zhi JL, Gao RO. 1995. Eect o uoride exposure on intelli-

gence in children. Fluoride 28(4):189192.

Li Y, Jing X, Chen D, Lin L, Wang Z. 2003. The eects o endemic

uoride poisoning on the intellectual development o children

in Baotou. Chin J Public Health Manag 19(4):337338.


20 August 2012].

Li Y, Li X, Wei S. 1994. Eect o excessive luoride intake on

mental work capacity o children and a preliminary study o

its mechanism. J West China Univ Med Sci 25(2):188191.


20 August 2012].

Lin C, Zhang H. 1986. Wechsler Children Intelligence Scale. Revised

Edition in China. Beijing:Beijing Normal University P ress.

Lin FF, Ai HT, Zhao HX, Lin J, Jhiang JY, Maimaiti, et al. 1991.

High uoride and low iodine environment and subclinical

cretinism in Xinjiang [in Chinese]. Endem Dis Bull 6(2):6267.

Lu Y, Sun ZR, Wu LN, Wang X, Lu W, Liu SS. et al. 2000. Eect o

high-uoride water on intelligence in children [in Chinese].

Fluoride 33(2):7478.

Mullenix PJ, Denbesten PK, Schunior A, Kernan WJ. 1995.

Neurotoxicity o sodium luoride in rats. Neurotoxicol

Teratol 17:169177.

NRC (National Research Council). 2006. Fluoride in Drinking

Water: A Scientifc Review o EPAs Standards. Washington,

DC:National Academies Press.

Peng YP, Zou J, Yang DF, Li XH, Wu K. 2008. Analysis o water

quality rom homemade wells in Leshan downtown

during 20042006 [in Chinese]. J Occup Health Damage.

23(4):219221.

Petersen PE, Lennon MA. 2004. Eective use o uorides or the

prevention o dental caries in the 21st century: the WHO

approach. Community Dent Oral Epidem 32(5):319321.

Poureslami HR, Horri A, Atash R. 2011. High uoride exposure

in drinking water: eect on childrens IQ, one new report.Int J Pediatr Dent 21(suppl 1):47.

Raven J, Raven JC, Court JH. 2003. Manual or Ravens

Progressive Matrices and Vocabulary Scales. San Antonio,

TX:Harcourt Assessment.

Ren DL, Li K, Lin D. 1989. An investigation o intelligence

development o children aged 814 years in high-uoride

and low-iodine areas. Chin J Control Endem Dis 4:251.


20 August 2012].

Seraj B, Shahrabi M, Falahzade M, Falahzade FP, Akhondi N.

2006. Eect o high luoride concentration in drinking

water on childrens intelligence. J Dental Med 19(2):8086.

[abstract in English]. Available: http://journals.tums.ac.ir/

upload_fles/pd/_/2530.pd [accessed 24 August 2012].

Stern JAC. 2009. Meta-analysis in Stata: An Updated Collection

rom the Stata Journal. College Station, TX:Stata Press.

Sun LY. 2010. Survey o drinking water quality in Jintang County

[in Chinese]. J Occup Health Damage 25(5):277280.

Sun MM, Li SK, Wang YF, Li FS. 1991. Measurement o intelli-gence by drawing test among the children in the endemic

area o Al-F combined toxicosis [in Chinese]. J Guiyang

Med College 16(3):204206.

Tang QQ, Du J, Ma HH, Jiang SJ, Zhou XJ. 2008. Fluoride and

childrens intelligence: a meta-analysis. Bio Trace Elem

Res 126:115120.

Trivedi MH, Verma RJ, Chinoy NJ, Patel RS, Sathawara NG.

2007. Eect o high uoride water on intelligence o school

children in India. Fluoride 40(3):178183.

U.S. EPA. 2011. EPA and HHS Announce New Scientiic

Assessments and Actions on Fluoride: Agencies Working

Together to Maintain Benefts o Preventing Tooth Decay

while Preventing Excessive Exposure. Available: http://

yosemite.epa.gov/opa/admpress.ns/bd4379a92ceceeac85

25735900400c27/86964a577c37ab285257811005a8417!Open

Document [accessed 7 January 2011].


7/7

Choi et al.


Varner JA, Jensen KF, Horvath W, Isaacson RL. 1998. Chronic

administration o aluminum-uoride or sodium-uoride to

rats in drinking water: alterations in neuronal and cerebro-

vascular integrity. Brain Res 784:284298.

Wang D, Di M, Qian M. 1989. Chinese Standardized Raven Test,

Rural Version. Tianjin, China:Tianjin Medical University.

Wang G, Yang D, Jia F, Wang H. 1996. Research on intelligence

quotient o 4-7 year-old children in a district with a high

level o uoride. Endem Dis Bull 11:6062. Available: http://

www.uoridealert.org/chinese/ [accessed 20 August 2012].

Wang SH, Wang LF, Hu PY, Guo SW, Law SH. 2001. Eects o

high iodine and high uorine on childrens intelligence and

thyroid unction [in Chinese]. Chin J Endemiol 20(4):288290.

Wang SX, Wang ZH, Cheng XT, Li J, Sang ZP, Zhang XD, et al.

2007. Arsenic and uoride exposure in drinking water: chil-

drens IQ and growth in Shanyin County, Shanxi Province,

China. Environ Health Perspect 115:643647.

Wang ZH, Wang SX, Zhang XD, Li J, Zheng XT, Hu CM, et al.

2006. Investigation o childrens growth and development

under long-term uoride exposure [in Chinese; abstract in

English]. Chin J Control Endem Dis 21(4):239241.

World Bank. 2006. Water Quality Management: Policy and

Institutional Considerations. Available: http://siteresources.

worldbank.org/INTEAPREGTOPENVIRONMENT/Resources/

China_WPM_fnal_lo_res.pd [accessed 13 June 2012].

World Health Organization. 2002. Fluorides. Geneva:World

Health Organization. Available: http://whqlibdoc.who.int/

ehc/WHO_EHC_227.pd [accessed 5 September 2012].

Wu TM. 1936. Second revision o Chinese-Binet Intelligence

Test. Shanghai:Commercial Press (in Chinese).

Wu T. 1983. The Chinese Comparative Intelligence Test

Guidebook. 3rd ed. Beijing:Beijing University Press.

Xiang Q, Liang Y, Chen L, Wang C, Chen B, Chen X, et al. 2003.

Eect o luoride in drinking water on childrens intel-

ligence. Fluoride 36(2):8494.

Xu YL, Lu CS, Zhang XN. 1994. Eect o uoride on childrens

intelligence [in Chinese]. Endem Dis Bull 2:8384.

Yang Y, Wang X, Guo X, Hu P. 1994. Eects o high iodine and

high uorine on childrens intelligence and the metabolism

o iodine and luorine. Chin J Pathol 15(5):296298.


20 August 2012].

Yao LM, Deng Y, Yang SY, Zhou JL, Wang SL, Cui JW. 1997.

Comparison o childrens health and intelligence between

the lu oro sis are as wit h and wit hout alt eri ng wate r

sources [in Chinese]. Lit In Prev Med 3(1):4243.

Yao LM, Zhou JL, Wang SL, Cui KS, Lin FY. 1996. Analysis o

TSH levels and intelligence o children residing in high

uorosis areas [in Chinese]. Lit In Prev Med 2(1):2627.

Zhang J, Gung Y, Guo J. 1985. Children Intelligence Scale

Handbook. Beijing:Captial Institute o Pediatrics Heatlh

Research Ofce.

Zhang JW, Yao H, Chen Y. 1998. Eect o high level o uoride

and arsenic on childrens intelligence [in Chinese]. Chin J

Public Health 17(2):57.

Zhang M, Wang A, Xia T, He P. 2008. Eects o uoride on DNA

damage, S-phase cell-cycle arrest and the expression

o NF-B in primary cultured rat hippocampal neurons.Toxicol Lett 179:15.

Zhao LB, Liang GH, Zhang DN, Wu XR. 1996. Eect o a high

uoride water supply on childrens intelligence. Fluoride

29(4):190192.

Developmental Fluoride Neurotoxicity

Documents

Transcript of Developmental Fluoride Neurotoxicity