Document baxoEa460gy956pq7rKyk4kgO

FILE NAME: Doubt Science (DBTS) DATE: 2017 DOC#: DBTS022 DOCUMENT DESCRIPTION: Journal Article - Non-Occupational Exposure to Asbestos and Risk of Pleural Mesothelioma: Review and Meta-Analysis Non-occupational exposure to asbestos and risk of pleural mesothelioma: review and meta-analysis Gary M Marsh,1,2Alexander S Riordan,3 Kara A Keeton,3 Stacey M Benson2 Additional material is published online only To view please visit the |ournal online (http //dx doi org/10 1136/ oemed-2017-104383) 'Center for Occupational Biostatistics and Epidemiology and Department of Biostatistics, Graduate School of Public Health, University of Pittsburgh, Pittsburgh, Pennsylvania, USA JCaidno ChemRisk, Pittsburgh, Pennsylvania, USA JCardno ChemRisk, Chicago, Illinois, USA Correspondence to Gary M Marsh PhD and FACE, Department of Biostatistics, Graduate School of Public Health, University of Pittsburgh, Pittsburgh, Pennsylvania, USA, Cardno ChemRisk, Pittsburgh, Pennsylvania, USA, gary m arshecardno com Received 17 February 2017 Revised 20 July 2017 Accepted 16 August 2017 Published Online First 21 September 2017 ABSTRACT O bjective To conduct an updated literature review and meta-analysis of studies of pleural malignant mesothelioma (PMM) risk among persons exposed to asbestos non-occupationally (household and neighbourhood) M etho d s We performed a literature search for articles available in the National Center for Biotechnology Information's PubMed database published between 1967 and 2016 Meta-analyses were conducted to calculate pooled PMM risk estimates, stratifying for household or neighbourhood exposure to asbestos and/ or predominant asbestos fibre type (chrysotile, amphibole or mixed) Results Eighteen studies in 12 countries comprising 665 cases met the meta-analysis inclusion criteria We identified 13 estimates of PM M risk from neighbourhood exposures, 10 from household and one from mixed exposure, and combined the estimates using randomeffects models The overall meta-relative risk (metaRR) was 5 9 (95% C M 4 to 8 7) The meta-RRs for household and neighbourhood exposures were 5 4 (95% Cl 2 6 to 11 2) and 6 9 (95% Cl 4 2 to 11 4), respectively We observed trends in risk in relation to fibre type for both household and neighbourhood studies For chrysotile, mixed and amphibole fibres, respectively, meta-RRs for neighbourhood studies were 3 8 (95% Cl 0 4 to 38 4), 8 4 (95% Cl 4 7 to 14 9) and 21 1 (95% Cl 5.3 to 84 5) and meta-RRs for household studies were 4 0 (95% Cl 0 8 to 18 8), 5 3 (95% Cl 1 9 to 15 0) and 21 1 (95% Cl 2 8 to 156 0) Conclusions PMM risks from non-occupational asbestos exposure are consistent w ith the fibre-type potency response observed in occupational settings By relating our findings to knowledge of exposureresponse relationships in occupational settings, we can better evaluate PMM risks in communities with ambient asbestos exposures from industrial or other sources It is well known that both occupational and non-occupational asbestos exposures, particularly amphibole asbestos, can increase the risk of pleural malignant mesothelioma (PMM). Earlier studies evaluating the risk of PMM due to non-occupational exposures observed elevated risks of PMM even though these exposures are generally much lower and more variable than occupational asbestos exposures. One previous meta-analysis of non-occupational asbestos exposures and PMM risk published in 2000 observed statistically significant risks that appeared dependent on fibre type. However, only eight studies were included in the original investigation published 17 years ago. Our updated review and meta-analysis included more than twice as many studies as the 2000 publication (n=18). We confirmed that non-occupational exposures to asbestos fibres are associated with an Increased risk of PMM (household=5.4 (95% Cl 2.6 to 11.2); neighbourhood=6.9 (95% Cl 4.2 to 11.4)). Our updated meta-analyses also confirmed a fibretype potency response for non-occupational exposures that is similar to the relationship observed in occupational settings. This study demonstrates more precisely a fibre-type potency response consistent with investigations of occupational asbestos exposures. By relating our fibre-type potency response findings to our knowledge of exposure-response relationships estimated in the occupational settings, we can better evaluate the risk of PMM in communities with ambient asbestos exposures from Industrial or other sources. CrossMark To cite . Ma'sh GM, Riordan AS, Keeton KA, et al Occtip Environ Med 2017,74 838-846 838 INTRODUCTION Asbestos is a tibious hvdrated magnesium silicate used extensively in building materials and insulation during the 20th century because of its indestructible nature, fire resistance and spmnabihty. Approxi mately 95% of world consumption has been chrys otile, a serpentine form of asbestos. Amphibole is the other asbestos fibre group comprising several types: amosite, crocidolite, anthophylhte, actinolite and tremolite. Exposure to asbestos can increase the risk of pleural malignant mesothelioma (PMM) and these risks appear to be considerably higher for amphiboles compared with chrysotile fibres.1 1 The increased PMM risks from occupational exposure to asbestos are well known. Some studies have shown that non-occupational (or environmental) exposures increase the risk of PMM in the general population. Two main types of non-occupational asbestos exposure are distin guished household (or domestic) and neigh bourhood (or residential) exposure. Household exposures include the installation, degradation, removal, or repan of asbestos-containing products Moreover, household exposures include paru-occupational exposures, which are those exposures taken home by asbestos workers or due to the use of asbestos-containing tools and products in the home. Neighbourhood exposures result from outside an pollution and include industrial emissions, natuial outcroppings, or erosion ot asbestos-containing Marsh GM etal Occup Environ Med 2017,74 838-846 doi 10 1136/oenned-2017-104383 BMJ building materials. Non-occupational asbestos exposures are usually much lower and more variable than exposures experi enced by workers manufacturing or using products made with asbestos.4 Compared with studies of persons exposed to asbestos occu pationally, relatively few studies have evaluated the relationship between non-occupational asbestos exposures and the risk of PMM. Numerous occupational and non-occupational studies, however, have evaluated the differential risk of PMM relative to fibre type. As early as 1972, the Occupational Safety and Health Administration (OSHA) indicated that chrysotile fibres weie less harmful than crocidolite." Hodgson and Darnton' conducted a systematic literature review that evaluated the quantitative usk of mesothelioma and lung cancer m relation to asbestos exposure. Based on occupational cohorts, the authors determined that the risk of mesothelioma due to exposure from chrysotile, amosite and crocidolite followed a 1 to 100 to 500 ratio, respectively. An updated analysis, which included additional studies, confirmed fibre potency differences associated with mesothelioma risk.4 A meta-analysis of occupational studies, conducted by Berman and Crump,' evaluated fibre-type potency with regard to mesothe lioma and found that chrysotile was between 0 and l/200th as potent as amphibole asbestos.1 Bourdes et al4 presented a literature review and meta-analysis by asbestos fibie type of the non-occupational studies of P.MM risk available around the turn of this century. Two of the eight studies identified, published between 1965 and 1998 in six different countries, specifically investigated asbestos expo sures associated with the Eternit asbestos cement factory located m Casale Monferrato, Italy6 7 The remaining studies were conducted worldwide and considered exposures to pure chryso tile, pure amphibole, mixed or unspecified fibre types. The Bourdes et al4 meta-analyses produced PMM risk effect estimates for neighbourhood and household asbestos exposures, respectively, that ranged from 1.5-4.0 for chrysotile, 6.7-8 2 for mixed or unspecified exposures, and 8.7-21.0 for amphiboles exposures. While the assessment of non-occupational asbestos exposures was generally less reliable than occupational expo sure assessments, the results of the Bourdes et al4 meta-analyses clearly supported the well known gradient in PMM risk by fibre type, or potency response, observed in the occupational studies. Since the Bourdes et al4 review' in 2000, several new and updated studies have evaluated PMM risk by fibre type among persons exposed to asbestos non-occupationally. We conducted an updated literature review and meta-analysis aimed at providing more reliable overall estimates of PMM risks in rela tion to fibre type and non-occupational asbestos exposures. We report here the results of oui updated and extended evaluation. METHODS Literature search We conducted this study using the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) and the Meta-analysis of Observational Studies in Epidemiology (MOOSE) guidelines.4 4 We performed a literature search for articles available in the National Center for Biotechnology Information's (NCBI) PubMed database published between 1967 and 2016 using the following keywords: asbestos, environ mental exposure, household exposure, neighborhood exposuie, para-occupational exposure, mesothelioma and pleural mesothe lioma. We conducted the literature search using the following operators, (((((household exposure) OR neighborhood expo sure) OR para occupational exposure) OR environmental exposure)) AND ((asbestos) AND (((pleural mesothelioma) OR mesothelioma) OR peritoneal mesothelioma)). Additionally, we systematically searched the reference lists of all studies identified in our literature review, as well as published review papers and meta-analyses, in order to identify relevant studies not captured in the primary literature search. All abstracts and arttcles weie reviewed by at least two scientists to determine if inclusion criteria were met. We selected studies based on the following criteria1. Only published studies available in English. 2. Descriptive and analytical epidemiological studies, including ecological, case-contiol and cohort studies, were considered tor inclusion. Case reports and case series were excluded 8. The medical condition of interest was defined as mesothe lioma (identified as `pleural mesothelioma' or `mesothe lioma'). Only studies that reported incidence and mortality outcomes of mesothelioma were included. Studies were excluded if the results were reported for peritoneal meso thelioma only or for pleural and peritoneal mesothelioma combined. 4 Exposures of interest were defined as non-occupational or environmental asbestos exposures (identified as `environ mental exposure'), which included household or domestic exposure (identified as `household exposure' or `para occupational exposure') and neighbourhood or residential exposure (identified as `neighborhood exposure'). House hold exposures included asbestos exposures from products contained wathin the home, or due to exposures associated with take-home asbestos from family members who poten tially brought dust containing asbestos home from their |obs on their clothing. Neighbourhood exposures were associated with living in close proximity to industrial or environmental sources that released asbestos fibres into the air. The expo sure descriptions provided by the authors of the included studies weie used to categorise reported effect estimates into neighbourhood and household exposmes. Studies weie excluded if the study subjects had definitive occupational exposure or if occupational exposure was uncertain. 5. If there were multiple studies with the same or overlapping populations, only the most recent study was considered for inclusion in the meta-analysis. 6. Only studies that reported effect estimates or provided data that enabled the calculation of an effect estimate were considered for inclusion m this evaluation. For all studies that met the above inclusion criteria, the following data were extracted, first author, publication year, study design, study location, outcome classification, asbestos exposure type and sources of exposure, the predominant asbestos fibre type, sex of the exposed subjects, number of cases and effect estimates (relative risk (RR), odds ratio (OR), stan dardised mortality ratio (SMR), standardised incidence ratio (SIR) and variance or confidence interval (Cl)) for each relevant exposure group (table 1) Meta-analysis Meta-analyses were performed to calculate pooled risk esti mates, specifically stratifying for- (1) household or neighbour hood exposure to asbestos, (2) predominant asbestos fibre type of the exposure (amphibole, chrysotile or mixed) and (3) asbestos fibre type by household and neighbourhood exposure The unspecified fibre type studies were only used in the anal ysis of all studies combined. In five of the publications that met the inclusion criteria, we manually calculated crude (unadjusted) Marsh GM, eta/ Occup Environ Med 2017,74 838-846 doi 10 1136/oemed-2017-10438B 839 OO o Marsh GM, et a l Occup Environ M ed 2017,74 838-846 doi 10 1136/oemed-2017-104383 Table 1 Studies on non-occupational exposure and pleural mesothelioma A u th o rs Year C o u n try O utc o m e Exposure type Source of exposure Fibre typ e Ecological studies Camus et a/18 Fazzo et aF' Mensi et a/22 1998 2014 2015 Canada Italy Italy Mortality Incidence Incidence Neighbourhood N e ig h b o u rh o o d Household N e ig h b o u rh o o d Asbestos mines or mills Industrial-- asbestos processing plant Industrial-- asbestos processing plant Chrysotile Unspecified Mixed M etm tas e fa /17 2002 Turkey Incidence N e ig h b o u rh o o d Geological Mixed Case-control studies McDonald and McDonald^ Madkour e fa /11 1980 2009 Canada/USA Egypt Incidence Incidence Household Neighbourhood N e ig h b o u rh o o d Industrial-- asbestos processing plant Asbestos mines or mills Industrial-- asbestos processing plant Chrysotile Chrysotile Lacourt etal23 Maule e fa /29 2014 2007 France Italy Incidence Incidence Mixed non-occupationa! Household N e ig h b o u rh o o d Unspecified/various Industrial-- asbestos processing plant Chrysotile Mixed Ferrante e ra /24 Luce e fa /25 Yaztcioglu et a/12 Newhouse and Thompson13 Howe/ e fa /19 2016 2000 1980 1965 1997 Magnani et a/26 2000 Italy N ew Caledonia Turkey UK UK Various (Italy, Spam, S w it z e r la n d ) Incidence Incidence Incidence Incidence Mortality Incidence Household N e ig h b o u rh o o d Household Household Neighbourhood Household Neighbourhood Household N e ig h b o u rh o o d Industrial-- asbestos processing plant Geological Geological Unspecified/various Industrial-- asbestos processing plant Industnal--asbestos processing plant Unspecified/various Mixed Tremolite Tremolite Mixed Unspecified Unspecified Cohort studies Hansen et a/14 Magnani eta/6 Ferrante e fa/27 1998 1993 2007 Australia Italy Italy Incidence Mortality Incidence N e ig h b o u rh o o d Household Household Asbestos mines or mills Industnal--asbestosprocessing plant Industrial-- asbestos processing plant Crocidolite Mixed Mixed M o ra lity Kurumatam and K u m a g a i28 2008 Japan Mortality N e ig h b o u rh o o d Industrial-- asbestos processing plant Mixed ` Disease was specified as pleural cancer or as a pleural tum our tEffect estimates were combined together using DerStmoman and Laird random effects m odels1516 Crude (unadjusted) effect estimates and/or 95% CIs were calculated from the raw data provided in the study Italicised studies were included in Bourdes et a t meta-analysts Technical details provided in online supplementary appendix Effect estim ate Sex C as es (n) (9 5 % Cl) F 7 M+F 30 M 5 M+F 72 M+F 24 M+F 8 M+F 1 M+F 83 M+F 22 M+F 75 M+F 60 M+F 33 Unspecified 14 M+F 22 M+F 7 M+F 11 M+F 17 M+F 5 M+F 30 M+F 25 SMR-7 6 (3 1 to 15 7)* SfR-1 7(1.1 to 2.4) SIR=2 0 (0.7 to 4 7) SIR=6 6 (5.2 to 8 3) SIR-87.0 (32 5 to 233.0)4 OR=4.0 (0 8 to 18 8)4 OR=0 2 (0.02 to 2 2)t OR=26 7 (5 7 to 581 0 )t OR-3.8 (1.6 to 8 9)+ OR=1 8(1.1 to 2 8) OR-6 1 (3 6 to 10 5 )t OR=2.4 (1.3 to 4 4) OR = 40 9 (5 2 to 325 0) RR-21 1 (2 8 to 156 0)7 OR=2S 7 (3 Oto 220 0)* OR=5 5 (1 7 to 173)7 O R -58{1.7 to 192) OR-6.6 (0.9 to 50 0) OR=3 7 (1 6 to 8 7)t OR=8 6 (1 7 to 42 2)t M+F 27 SIR=12 3 (1 9 to 79 8 )t* F 3 SM R-7 7 (1 6 to 21.9) F 11 SIR=25 2 (12 6 to 45.1 ) M+F 73 SM R=4.3 (3.4 to 5.4) Figure 1 Flow diagram for included studies effect estimates and corresponding Cls using data provided m the corresponding manuscript10-14The Appendix provides tech nical details of the data extraction and calculations performed to generate these estimates manually. In instances where more than one effect estimate was provided per stratification of interest (eg, separate estimates for men and women), but the predom inant fibre type exposure was the same, effect estimates were combined together using Der Simoman and Laird random effects models" 16 to determine an overall effect estimate. Upon the extraction or derivation of all efiect estimates and Cls from included publications, we performed fixed effects meta-analyses and the corresponding I2 tests to assess the homo geneity of effect estimates from the studies. DerSimoman and Laird random effects models were subsequently performed on all iterations or stratifications described above." 16 The meta-analyses results are reported as meta-RR with corresponding 95% Cl. Funnel plots were also created to evaluate the potential of publication bias. All statistical analyses were perfoimed using StataMP V.14 (StataCorp, College Station, Texas, USA). RESULTS Literature search Our PubMed literature search returned 1795 results (figure 1). Numerous seal ch results were excluded because they did not meet the initial inclusion criteria: 314 research articles were written in languages other than English, 83 studies were conducted on experimental animals, 408 articles were commentaries, which included opinion letters and letters to the editor, and case series or case reports. When limiting the disease outcome to pleural mesothelioma, 797 studies were excluded. Limiting to non-occupational exposure excluded an additional 151 studies. Of the resulting 42 studies, 18 descriptive epidemiology studies were excluded. Cohorts were then limited to the most recent study, yielding 16 studies. Furthermore, 99 review ai tides were iden tified, specific to non-occupational exposures to asbestos. Two additional articles that specifically discussed PMM and met the rest of the inclusion criteria were referenced in the review articles. Ultimately, 18 studies in 12 countries comprising 665 cases met the criteria for inclusion m the meta-analysis (table 1). Ot these 18 studies, 11 reported efiect estimates for PPM with either household or neighbourhood exposure. There were four ecolog ical studies, 10 case-control studies and four cohort studies. Six papers reported effect estimates for both household and neighbouihood exposures, while one paper reported an effect esti mate for a mixed non-occupanonal exposme. Overall, 24 effects estimates were reported, 13 due to neighbourhood exposures, 10 due to household exposures and one for mixed non-occupational exposures. Overall effect estimates ranged from an OR of 0.2 (95% Cl 0.02 to 2.2) tor neighbourhood exposures to chrvsotile10 to an SIR of 87.0 (95% Cl 32.5 to 233.0) for neighbour hood exposures to geological sources of mixed asbestos fibres 17 Four of the reported effect estimates were not statistically signif icant. Only seven studies provided effect estimates for the risk of mesothelioma due to pure fibre exposures, with the majority being studies of mixed or unspecified fibre type. Comparison to Bourdes ef a/.4 Literature Search Results The previous meta-analysis conducted by Bourdes ct al4 included eight papers/' ' 111121118-20 Since the initial meta-analysis, 10 addi tional studies11 " 21-28 and one update of a previously evaluated cohort24 have been published that examined the risk ot PMM following environmental exposure to asbestos. The overlapping studies included in Bourdes et alAand tile current analysis are Marsh GM, ef a/ Occup Environ Med 2017,74 838-846 doi 10 1136/oemed-2017-104383 841 Table 2 Random effects meta-analyses by asbestos exposure type and by fibre type Neighbourhood exposure Household exposure N Meta-RR 95% Cl N Meta-RR 95% Cl A ll studies* 13 6.9 4,2 to 11.4 10 5.4 2 6 to 11.2 Fibre type Chrysotile 3 3.8 0 4 to 38.4 1t 40 0.8 to 18.8 Mixed 6 84 4 7 to 14 9 6 53 1 9 to 15 0 Amphibole 2 21.1 5 3 to 84 5 n 21.1 2.8 to 156.0 *ln five of the publications, data reported in the manuscript were used to calculate crude (unadjusted) effect estimates and CIs See online supplementary appendix for technical details tA s reported by the sole study per category, Yaziooglu et alu and McDonald and McDonald Note Of the total 24 effect estimates included in this manuscript, 23 are presented in this table as one estimate pertained to a mix of both neighbourhood and household exposures23 lcaltciscd in tabic 1. Botha et al,10 which was included in the earlier meta-analysis, was excluded from our analysis because the reported number of deaths from PMM and asbestosts could nor be distinguished. Additionally, one study14 was included m our meta-analysts, as it evaluated PMM due to neighbourhood exposure, but was not included in the Bourdes et a t meta-analysis. When accounting for the five studies1'1' 14 where reported data were used to estimate risks and 95% Cl, our calculations wete consistent with those reported by Bourdes et a lt except for those calculated from the data reported m Newhouse and Thompson.15 Here, our calculated OR was similar, but the 95% Cl was larger than that reported m Bourdes et a lt Meta-analysis Table 2 presents Jesuits of the random effects meta-analyses models and corresponding 95% CIs, by neighbourhood or household exposure type, fibre type, and neighbourhood or household exposure and fibre type concomitantly. When fixed effects models were run on all subgroups, only two subgroups failed to reject the null hypothesis of homogeneous effect esti mates used from the studies: amphibole total and neighbour hood amphibole, both of which demonstrated an I2 of 0 and p-vulues of 0.70 and 0.40, respectively Of note, rhe results of the random effects analyses on these two subgroups were not markedly different from those found under fixed-effects models. In all othei instances, the findings of fixed-effects models demon strated p-values <0.0001 for an I2 test, indicative of heteroge neity Therefore, to account for heterogeneity demonstrated across the strata, we present results from our random effects models exclusively. Forest plots were generated to graphically represent the meta-RR, variability and heterogeneity of those studies by type of exposure (figure 2). When we considered effect estimates from all studies (n=24), the meta-RR was 5.9 and statistically significant (95%Cl 4 1 to 8 7). However, because household exposures are generally not equivalent to neighbourhood exposures, the remaining analysis considered these two exposure scenarios separately (excluding one study that had mixed neighbourhood and environmental exposures combined).2' When stratified by asbestos exposure type, the meta-RR for household exposure was 5.4 and statisti cally significant (95% Cl 2.6 to 11.2). The meta-RR for neigh bourhood exposuie was slightly higher (6.9) and statistically significant (95% Cl 4.2 to 11.4) (figure 2). We observed similar trends m asbestos fibre-type potency in relation to PMM risk for both household and neighbourhood studies (table 2). For household exposure studies with mixed fibre type, the random effects meta-RR was 5.3 and statistically significant (95%CI 1 9 to 15.0). Only one effect estimate was available for household studies with chrysotifc or amphibole exposure, precluding the calculation of a meta-RR for these subgroups. For compara tive purposes, we reported results from each of these studies m table 2 (Amphibole: as reported in Yazicioglu et a l'1: OR=21.1 (95% Cl 2.8 to 156.0); Chrysotde- as reported in McDonald and McDonald10. O R=4.0 (95% Cl 0.8 to 18.8)). For neigh bourhood studies, the meta-RR for mixed fibre exposure was 8.4 (95% Cl 4.7 to 14 9); the meta-RR for chrysotile exposure was 3.8 (95% Cl 0.4 to 38.4); and the meta-RR for amphibole exposure was 21 1 (95% CL 5.3 to 84.5). DISCUSSION Overall, the results of our updated meta-analysts demonstrated that non-occupational exposure to asbestos fibres is associated with increased PMM risk. The risk of PMM between household exposures and neighbourhood exposures was comparable with an evaluation of al! studies demonstrating a statistically signifi cant increased risk of PMM (meta-RR=5.9; 95% Cl 4.1 to 8.7). However, there was a pronounced effect of fibre type on the meta-risk estimates. For neighbourhood and household expo sures, respectively, exposure to amphibole fibtes alone had a meta-RR of PMM 2.5-3.2 times greater than those exposed to mixed fibres, and 5.3-5.6 times greater than those exposed to chrysotile fibres alone While our meta-RRs for PMM aie generally consistent with those observed in the earlier literature review and meta-analysis of Bourdes et a l t the variability associated with our estimates was similar or even greater for some subgroups considered, despite having included more than twice as many studies overall. These instances of mci eased variability are due to the heterogeneity of the effect estimates associated with the studies included m our update. These studies were conducted on populations located worldwide, and several different potential sources of asbestos were included: asbestos cement plants, asbestos textile facilities, asbestos mines and mills, and naturally occurring geological sources. The subpopulations evaluated also varied, with some studies including only men or women and some including both Depending on fibre type, the range in the reported risks of PMM in the individual studies was quite large For chrysotile-speufic exposures the risk ranged from O R=0.2 (95% Cl 0.02 to 2.2)10 to OR=26.7 (95% Cl 5.7 to 581.0).11 Mixed fibre exposures reported PMM risks that ranged from O R = l 8 (95% Cl 1 1 to 2.8)29 to SIR=87.0 (95% Cl 32 5 to 233 0).17 There was roughly a threefold increased risk of PMM for amphibole expo sures between the lowest (SIR=12.3; 95%C1 1.9 to 79.8)14 and highest reported estimates (OR=40.9; 95% Cl 5.2 to 325.0).21 Moreover, a potential limitation is the inclusion of seven crude (unadjusted) effect estimates, which could have led to the large 842 Marsh GM, eta/ Occup Environ Med 2017,74 838-846 doi 10 1136/oemed-2017-104383 !C S 35% Cl) W&igh? Usf&Coi*&t a! n MeDc'SkfJair,c18, <'? .. .......*...... }13.J3siSi ;2$'?*!*}(st si 22, al 17, aa!rt}{**} Fa.2212etasli12i Fsc-Ae! et si t@( <2t) Si 6H(*'3000/ -- * 2670(670 58060} 191 020<002 220} ?60(310, 1570} 51 1816 430(3<0,540} 663 w, 666100((352600,S10305}0} 659602 - 700(3250 23320) 460 ....1 1*.... 6125300(1(17900,.1779308)0) 245140 4-.---:>...... -- 4090(520,32500) 223 170H10,240} 624 ... yi...... 660(090,5000) 231 60(170,4220} 303 <p> 6SO(418 1140} 5288 UiCO'-'alBsf'-s 'i-2; ida! 'T'\ et sj 24 tier's'leltealta:!28c, *f si '221 H'S-'VCiOviSP-S7P<5rn$,'iZP yvOtyfelit at S-fetc'sl et-isaiva{fe2t$?'*S3c%t *0003 LBs-a-3>5^"-si al iZ2` ^ t ~t Overall vl'TSiy^'fiO*38 f uo t*-CT s-'Srfrerrirs-tdems<3ss"ai/5-i5 00??$ ** ~"i#. 1, "" --i "' t o 15 <o O 1 426SO28(0$18206108,4S6O)10} 309 662 240(130.440} 570 710700((116100,221809}0) 030771 250(070 470} 469 22157100$$23000,2125060000}) 223112 50(170 1920) 396 357400((216500 8119023}} 442S314 380(160 8O0; 407 380(161,S96} 437 694(40? 867) 10000 ! 031 Figure 2 Forest plots of random effects models of household and neighbourhood exposures range of reported effect estimates of PMM observed between studies. Additional sources of the heterogeneity m mcta-RRs include the overall mixture of asbestos fibre exposures and exposure levels to those fibres. The mixed fibre studies included asbestos cement plants,6 11 14 17"1'1 geological deposits17 and proximity to asbestos factories.1' All of these sources emitted different amounts and different ratios of asbestos fibre types. Further more, only three studies14111~4 estimated actual exposure levels. Other studies consideied distance from an asbestos emitting fauhtv or mine, but the majority of the studies reviewed based the designation of exposure on geneial definitions of residential location or living with an individual who worked in an asbestos related industry. Another source of potential exposure misclassification could come m the form of recall bias. Only two of the ease-control studies11 directly interviewed the cases. The lemaming case-control studies used family members to provide information on exposures or used proxy measures of exposure, such as residential location in relation to a potential asbestos source. The aforementioned exposure designations may have led to increased potential for exposure misclassification, which in turn increased the variability m the study findings and directly corresponded to the externa] inconsistencies observed between the evaluated studies. Notably, six studies were conducted in Italy in locations adja cent to various asbestos cement product manufacturing facilities. The effect estimates for these studies ranged from a SIR of 1.8 (95% Cl 1.1 to 2.8) foi a hospital survey of new PMM cases diagnosed between 1 January 1987 and 30 June 1993, among residents living in Casale Monferiato24 to a SIR of 21.2 (95% Cl 12.6 to 45.1) for incident PMM cases for wives of workers employed at the same facility.27 Thus, considerable internal inconsistency exists among the Italian observations as well as within the observations associated with the same asbestos emit ting facility. Only two studies we evaluated provided PMM effect estimates by estimated level of non-occupational exposure to asbestos. 424 Hansen et a lu evaluated the risk of PMM due to environmental crocidohte exposures among residents of Wittenoom, Australia As we noted in a lecent letter to the editoi,10 the highest expo sure groups considered by Hansen et al had cumulative expo sures of at least 20 f/mL-years (fibers/mL-years), with RRs ranging from 3.6 to 6.3 depending on the type of censoring.14 In contrast, Ferrante et al2* found an OR of 23.3 (95% Cl 2.9 to 186.9) for their highest non-occupational cumulative expo sure category of 10.0 f/ml.-years to 24.2 f/mL-years. Thus, the higher exposure levels estimated by Hansen et a /14yielded PMM risks 3.0-fold to 6.5-fold lower than those reported by Ferrante et al.2* This inconsistency was exacerbated by the fact that subjects in Ferrante et al24 were exposed to mixed chrysotile, amphibole fibres, whereas the residents of Wittenoom were exposed to pure crocidohte fibres. This finding conflicts with the established view that PMM risks are higher for exposure to amphibole fibres than other fibre types.1"1 The Ferrante et al2* study also contained internal inconsis tencies in PMM effect estimates relative to asbestos exposure. For example, Boffetta11 recently noted that the ORs for envi ronmental asbestos exposure appeared to be higher for the same MarshGM, efa/ Occup [nv>for, Med 2017,74 838-846 doi 10 1136/oemed-2017-104383 843 level of exposure than the corresponding ORs for occupational asbestos exposure. This was particularly evident in the category of highest exposure (>10f/mL-years) despite exposure levels being lowei on aveiage in that category (15 f/ml.-years vs 260 f/ mL-years) It is difficult to estimate how these inconsistencies may have influenced this analysis as overall effect estimates from each stud) were calculated and used in the reported meta-anal ysis, rather than risks reported by exposure levels. The numerous inconsistencies and considerable variability v\e observed for asbestos related PMM effect estimates across and within the individual studies are due in part to limitations associated with these studies and the corresponding meta-analysis. For one, the overall quality of the studies is mixed. Seven effect estimates from the ISi studies were based on 10 or fewer PMM cases, giving rise to less precise individual and combined elfect estimates. Five studies did not confirm the diagnosis of mesothelioma histologically.6 I! 14 1621 Misdiagnosis of meso thelioma could lead to inflated risk estimates and ultimately result in meta-RRs that are higher than expected Relatively few studies evaluated non-occupational exposures to pure asbestos fibre types Also, as stated above, the potential for exposure irusclassification and related study bias is high for many studies because of the lack of quantitative exposure estimates. Further more, as indicated by Bourdes et al,4 the pooled risk estimates from this most recent meta-analyses most likely overestimated PMM risk due to lion-occupational exposures as the evaluated groups experienced higher levels of exposures than individuals living in countries without specific sources of exposure. The quality of the included studies and the control or compar ison groups used by the individual studies could also influence the results of tins meta-analysis. Several studies calculated effect estimates by comparing the exposed population to national28 or world14 ' standardised populations. Due to the varying rates of PMM across an individual country or the world, reported effect estimates could be biased. For example, Delgermaa et al32 compaied mesothelioma rates across the world and showed that the UK had age-adjusted moitality rates 5.6-fold higher than Japan. A review article of incident mesothelioma rates showed that Australia had the highest incident rate, followed by Great Britain and Belgium while South Korea, Morocco and Tunisia had the lowest, approximately a 30-fold difference.1' For coun tries with high mortality and incidence rates, comparison to the global population would overestimate the risk Conversely, a country with lower incidence and mortality rates that compares to the world population may underestimate their risk estimates. Moreover, rates of mesothelioma have also been shown to vary considerably across regions within individual countries. An evaluation of mortality rates in Italy showed that standardised rates can vary by more than sevenfold for the total population, depending on the location of interest.34 Historical evaluations of pleural mesothelioma deaths in USA have shown that rates are higher m north-eastern USA, Colorado, Wyoming, Florida, Illinois and the Pacific Coast than in other regions of USA.1' Similar to the use of global standardised rates, comparison to national rates could also lead to biased risk estimates. Studies that were able to draw comparison groups from populations in the surrounding region may have been able to minimise this potential source of bias. Another limitation of the presented analysis is the inclusion of ecological studies. We understand that these are not usually included in meta-analyses due to the limitations associated with exposure assessment and the lack of adjustment for confounding factors. In this instance, we felt it was appropnate to include the ecological studies because all individuals were selected based on their potential exposure to asbestos rather than then case status. Furthermore, the exposures for some of the ecological studies were equivalent, and in some instances, better than the proxy measures of exposure reported in some of the case-con trol studies. For instance, Camus et a l]s used historical records of asbestos use, ambient asbestos air concentrations recorded by the government, and asbestos fibre concentrations in the air reported by the asbestos industiy. This historical information, along with information on the duration of residency, allowed foi the estimate of cumulative exposure levels. Metintas et al'' collected soil samples and indoor and outdoor air samples to confirm exposure to naturally occurring asbestos. Furthermore, it is unlikely that confounding factors contributed to the devel opment of mesothelioma because the primary risk factor for the development of mesothelioma is exposure to asbestos. In order to evaluate how some of the limitations stated above may have impacted our overall findings, additional analvses were conducted. The potential for publication bias (figure 3) was also evaluated. We conducted sensitivity analyses to determine how the inclusion of the crude (unadjusted) estimates calculated by the authors and the inclusion ol the ecological studies may have impacted our overall findings Neither the exclusion of the crude (unadjusted) estimates nor the ecological studies changed our overall findings, with effect estimates remaining above 5 and heterogeneity unchanged As shown in figure 3, the poten tial for publication bias appears limited for the neighbourhood studies. The household studies do not appear as equally distrib uted, however, any studies possibly omitted most likely would have had lower effect estimates and been weighted less than the reported studies. In this hypothetical scenario, inclusion of such 4 > 5 2 f 1 0 2 Neighborhood ---------; ----- * -- -- f 10 Weigh 3.5 3 c' 1. 0 s 0 Household 10 15 20 Weight (%) Figure 3 Plots to evaluate publication bias for neighborhood (A) and household (B) studies that reflect the natural log of the reported effect estimates versus the weights used in the random effects meta-analysis for each group of studies separately 844 Marsh GM, era/ Oaup Environ Med 2017,74.838-846 doi 10 1136/oemed-2017-104383 studies would not be expected to influence the overall results of the meta-analysis. The observed funnel plots are consistent with the neighbourhood exposure evaluation previously conducted by Bourdes et al.4 Another critical component of these studies is to ensure adequate latency for disease development. Six studies reported latency \alues of at least 20years for a majority of, if not all, individuals diagnosed with PMM 6 22 24 2' 28 22 Mensi et al22 observed a latency period of 54 years, on average, for those with familial exposure and a median of SOyears since first exposure foi those with envnonmental exposuie. Maule et a r accounted tor latency in domestic or environmentally exposed subjects by excluding information about residential location for the last 20 years leading up to diagnosis. We used the reported effect estimate for women married to asbestos cement plant workers that had latency values between 20 years and 3 8 years'' An evaluation of individuals who lived near an industrial source of asbestos in Japan had an average latency of 43 years.28 However, the remaining studies were not dear on the duration of time between exposure and disease diagnosis. A strength of our evaluation was the systematic approach to locating and summarising relevant studies. In addition, our update included more than twice the number of studies included in the previous Bourdes et al4 evaluation which included the most recent updates for cohorts that have been evaluated at numerous time points. The increased sample size of studies allowed for the evaluation of confirmed mixed fibres without the inclusion of unspecified fibre-type studies. We also calculated PMM effect estimates from the raw data provided in the publica tions when the requisite estimates for our meta-analysis were not provided by the authors Finally, we were able to account for the observed heterogeneity m study design and population by using random effects analytical techniques. Though limitations inherent to the studies included in this meta-analysis are evident (specifically with inadequate compar ison gioups, exposure misclassification, misdiagnosed mesothe lioma and recall bias), these studies are the current literature a\ ailable evaluating non-occupational exposures. Addressing this important potential source of bus in a formal way would require access to medical and pathology records, and to pathology spec imens It is unlikely that these sources are available. However, a discussion of this source of bias, including simulation-based sensitivity analyses using results of independent validation studies, would be useful. Future studies could benefit from implementing simulation-based sensitivity analyses, as well as more formal confirmation of medical diagnoses and non-occupational exposures. CONCLUSIONS We observed elevated risk of PMM for both neighbourhood and household exposures. The risk of PMM was directly related to asbestos fibre type, with chrysotile fibres not showing statistically elevated risk of PMM However, PMM effect esti mates observed across and within the 18 individual studies revealed considerable heterogeneity, and the sample sizes for the meta-analyses conducted by fibre type were small for pure fibre studies. Nevertheless, PMM risks from non-occupational asbestos exposure are consistent with the fibre-type potency response obseived m occupational settings By relating our findings to knowledge of exposure-response relationships in occupational settings, we can better evaluate PMM risks in communities with ambient asbestos exposures from industrial or other sources Contributors G M M had the idea for the manuscript and was responsible for the overall quality of the work He also drafted the text and assisted w ith the biostatistical approach to the analysis ASR, KAK and SMB reviewed the literature and confirmed the inclusion and exclusion of various studies KAK and SMB were responsible for drafting the text and creating the figures ASR was responsible for conducting the analyses and drafting the text Funding Schellenberg Wittmer, a Swiss business law firm tha t provides comprehensive legal services to domestic and international clients Competing interests All authors are employed by Cardno ChemRisk, a consulting firm that provides scientific advice to the government, corporations, law firms and various scientific/professional organisations Cardno ChemRisk has been engaged by Schellenberg Wittmer, a law firm in Switzerland, to provide general consulting and expert advice on scientific matters, as well as litigation support This paper was prepared and written exclusively by the authors, w ithout review or comment by Schellenberg W ittm er counsel One of the authors (GM M ) has previously testified on behalf of Schellenberg W ittm e r in asbestos litigation The study, the preparation of the paper, including the synthesis of the findings, the conclusions drawn, and recommendations made are the exclusive professional work product of the authors, and may not necessanly be those of their employer Provenance and peer review Not commissioned, externally peer reviewed Article author(s) (or their employer(s) unless otherwise stated in the text of the article) 2017 A ll rights reserved No commercial use is permitted unless otherwise expressly granted REFERENCES 1 Berman DW, Crump KS A meta-analysis of asbestos-related cancer risk that addresses fiber size and mineral type Cnt Rev Toxicol 2008,38 (Suppl 1) 49-73 2 OSHA 29 CFR Part 1910 Standard exposure to asbestos dust Notice of proposed rulemaking Fed Reg 1972,37 466-8 3 Hodgson JT, Darnton A The quantitative risks of mesothelioma and lung cancer in relation to asbestos exposure Ann Occup Hyg 2000,44 565-601 4 Bourdes V, Boffetta P, Pisam P Environmental exposure to asbestos and risk of pleural mesothelioma review and meta-analysis Fui J [pidcm iol 2000,16 41' -1 5 Hodgson JT, Darnton A Mesothelioma risk from chrysotile Occup Environ Med 2010,67 432 6 Magnam C, Terracint B, Ivaldi C, et al A cohort study on mortality among wives ol workers in the asbestos cement industry in Casale Monferrato, Italy Br 1 Ind Med 1993,50 779-84 7 Magnam C, Terracim B, Ivaldi C, et al Pleural malignant mesothelioma and nonoccupational exposure to asbestos m Casale Monferrato, Italy Occup inviron Med 1995,52 362-7 8 Moher D, Liberati A, Tetzlaff J, et al Preferred reporting items for systematic reviews and meta-analyses the PRISMA statement Ann Intern Med 2009,151 264-9 9 Stroup DF Meta-analysis of observational studies in epidemiology a proposal for reporting JAMA 2008,2000 283 10 McDonald AD, McDonald JC Malignant mesothelioma in North America Cancer 1980,46 1650-6 11 Madkour MT, El Bokhary MS, Awad Allah HI, et a! Environmental exposure to asbestos and the exposure-response relationship with mesothelioma EastMedtterr Elealth J 2009,15 25-38 12 Yazicioglu S, llgayto R, Balci K, et al Pleura! calcification, pleural mesotheliomas, and bronchial cancers caused by tremolite dust Thorax 1980,35 564-9 13 Newhouse ML, Thompson H Mesothelioma of pleura and peritoneum following exposure to asbestos m the London area B r J Ind Med 1965,22 261-9 14 Hansen J, de Klerk IiH, Musk AW, et al Environmental exposure to crocidolite and mesothelioma exposure-response relationships Am J Respir Cnt Care Med 1998,157 69-75 15 DerSimoman R, Laird N Meta-analysis in clinical trials Control Clin Trials 1986,7 177-88 16 ElwoodJM Critical appraisal o f epidemiological studies and clinical trials NewYork Oxford University Press, 2007 17 Metmtas S, Metintas M, Ucgun I, et al Malignant mesothelioma due to environmental exposure to asbestos follow-up of a Turkish cohort living in a rural area Chest 2002,122 2224-9 18 Camus M, Siemiatycki J, Meek B Nonoccupational exposure to chrysotile asbestos and the risk of lung cancer N Engl I Med 1998,338 1565-71 19 Howel D, Arblaster L, Swinburne L, et al Routes of asbestos exposure and the development of mesothelioma in an English region Occup Environ Med 1997,54 403-9 20 Botha 1L, Irwig LM, Strebel PM Excess mortality from stomach cancer, lung cancer and asbestosis and/or mesothelioma in crocidolite mining districts in South Africa Am J Epidemiol 1986,123 30-40 21 Fazzo L, Menegozzo S, Soggiu ME, et a l Mesothelioma incidence in the neighbourhood of an asbestos-cement plant located in a national priority contaminated site Ann 1st Super Sanita 2014,50 322-7 Marsh GM, eta/ Occup Environ Med 2017,74 838-846 doi 10 1136/oemed-2017-104383 845 27 Merisi C, Riboldi L, De Matteis S, eta! Impact of an asbestos cement factory on mesothelioma incidence global assessment of effects of occupational, familial, and environmental exposure Environlnt 2015,74 191-9 23 Lacourt A, Gramond C, Rolland P, et al Occupational and non-occupational attributable risk of asbestos exposure for malignant pleural mesothelioma Thorax 2014,69 532-9 24 Ferrante 0, Mtrabeili D, Tunesi S, e ta i Pleural mesothelioma and occupational and non-occupational asbestos exposure a case-control study with quantitative risk assessment Occup Environ Med 2016,73 147-53 25 Luce D, Bugel I, Goldberg P, et al Envnonmental exposure to tremolite and respiratory cartcenn New Caledoma a case-control study Am J Epidemiol 2000,151 259-65 26 Magnam C, Agudo A, Gonzalez CA, et al Multicentric study on malignant pleural mesothelioma and non-occupational exposure to asbestos Br J Cancer 2000,83 104-11 27 Ferrante D, Bertolotti M, Todesco A, et al Cancer mortality and incidence of mesothelioma in a cohort of wives of asbestos woikers in Casale Monferiato, Italy Environ Health Perspeci 2007,115 1401-5. 28 Kuiumatam N, Kumagai 5 Mapping the risk of mesothelioma due to neighborhood asbestos exposure Am J Respir Cnt Care Med 2008,178 624-9 29 Maule MM, Magnam C, Dalmasso P, et a l Modeling mesothelioma risk associated with environmental asbestos exposure Environ Health Perspect 2007,115 1066-71 30 Marsh GM, Benson SM Response to 'Pleural mesothelioma and occupational and non-occupational asbestos exposuie a case-control study with quantitative risk assessment' by Ferrante et al OccupEnviron Med 2017,74 156-7 31 Boffetta P Response to Pleural mesothelioma, and occupational and nonoccupational asbestos exposure a case-control study with quantitative risk assessment Occup Environ Med 2016,73 712 32 Delgermaa V, Takahashi K, Park K, et al Global mesothelioma deaths reported to the World Health Organization between 1994 and 2008 Bull World Health Organ 2011,89 716-24 33 Bianchi C, Bianchi T Malignant mesothelioma global incidence and relationship with asbestos Ind Health 2007,45 379-87 34 Fazzo L, Minelli G, De Santis M, et al Mesothelioma mortality surveillance and asbestos exposure tracking in Italy Ann 1st Supei Sanaa 2012,48 300-10 35 Enterline PE, Henderson VL Geographic patterns for pleural mesothelioma deaths in the United States, 1968-81 J Natl Cancel Inst 1987,79 31-7 846 Marsh GM, eta! Occup Environ Med 2017,74 838-846 doi 10 1136/oemed-2017-104383 Non-occupational exposure to asbestos and risk of pleural mesothelioma: review and meta-analysis Gary M Marsh, PhD, FACE, Alexander S Riordan, MPH, Kara A Keeton, MPH and Stacey M Benson, PhD Occup Environ Med 2017 74' 838-846 originally published online September 21, 2017 doi: 10.1136/oemed-2017-104383 Updated information and services can be found at: http://oem.bmj.com/content/74/11/838 References Email alerting service These include: This article cites 34 articles, 8 of which you can access for free at. http://oem.bmj.eom/content/74/11/838#BIBL Receive free email alerts when new articles cite this article. Sign up in the box at the top right corner of the online article. Notes To request permissions go to. http://group.bmj.com/group/rights-licensing/permissions To order reprints go to: http://journals.bmj.com/cgi/reprintform To subscribe to BMJ go to: http://group.bmj.com/subscribe/