Arch Toxicol (2014) 88:1479–1490 DOI 10.1007/s00204-014-1296-5
Review Article
Occupational exposures to polycyclic aromatic hydrocarbons and respiratory and urinary tract cancers: an updated systematic review and a meta‑analysis to 2014 Matteo Rota · Cristina Bosetti · Stefania Boccia · Paolo Boffetta · Carlo La Vecchia
Received: 5 February 2014 / Accepted: 4 June 2014 / Published online: 17 June 2014 © Springer-Verlag Berlin Heidelberg 2014
Abstract Exposure to polycyclic aromatic hydrocarbons (PAHs) has been associated with an excess risk of respiratory tract and bladder cancers in several industries, but the issue requires further quantification. We updated a previous systematic review by reviewing in details cohort studies on workers employed in selected industries with potential PAH exposure published between 2006 and 2014, and we summarized through a meta-analytic approach the main results of all available cohort studies published between 1958 and 2014 investigating cancers of the respiratory and urinary tracts. Thirteen papers on cohort studies investigating cancer risk in workers exposed to PAHs were retrieved through the literature search. These included workers from aluminum Electronic supplementary material The online version of this article (doi:10.1007/s00204-014-1296-5) contains supplementary material, which is available to authorized users. M. Rota · C. Bosetti (*) Department of Epidemiology, IRCCS—Istituto di Ricerche Farmacologiche “Mario Negri”, Via La Masa 19, 20156 Milan, Italy e-mail:
[email protected] S. Boccia Section of Hygiene, Institute of Public Health, Università Cattolica del Sacro Cuore, Rome, Italy S. Boccia IRCCS San Raffaele Pisana, Rome, Italy P. Boffetta The Tisch Cancer Institute and Institute for Translational Epidemiology, Icahn School of Medicine at Mount Sinai, New York, NY, USA C. La Vecchia Department of Clinical Sciences and Community Health, University of Milan, Milan, Italy
production industries (seven studies), iron and steel foundries (two studies), asphalt workers (two studies), and carbon black production (two studies). In the meta-analysis, an excess risk of respiratory tract cancers (mainly lung cancer) was found in iron and steel foundries [pooled relative risk (RR) 1.31, 95 % confidence interval (CI) 1.08–1.59 from 14 studies], while a weak excess risk (pooled RR 1.08, 95 % CI 0.95–1.23 from 11 studies) emerged for aluminum production. A borderline increase risk was also observed for cancer of the bladder in the aluminum production (pooled RR 1.28, 95 % CI 0.98– 1.68 from 10 studies) and in iron and steel foundries (pooled RR 1.38, 95 % CI 1.00–1.91 from 9 studies). This updated review and meta-analysis confirm the increased risk from respiratory tract and bladder cancers in selected PAH-related occupations. It cannot be ruled out whether such excesses are due, at least in part, to possible bias or residual confounding. Keywords Polycyclic aromatic hydrocarbons · Occupational exposure · Cohort studies · Cancer risk · Systematic review · Meta-analysis Abbreviations CI Confidence interval PAH Polycyclic aromatic hydrocarbons RR Relative risk SIR Standardized incidence ratio SMR Standardized mortality ratio
Introduction Polycyclic aromatic hydrocarbons (PAHs) are lipophilic nonpolar chemical compounds characterized by the presence of two or more benzene rings. They mainly derive from the incomplete combustion of organic materials and
13
1480
are usually found in the atmosphere as a mixture of different compounds, among which the best known is benzo[a] pyrene, often used as a marker of exposure to PAHs (Straif et al. 2005). Epidemiological evidence from workers occupationally exposed to PAHs in various industries has been considered in a comprehensive review by Boffetta et al. (1997) and subsequently by Bosetti et al. (2007) in a quantitative reassessment of papers published up to 2005. They found increased risks of lung and bladder cancer in selected PAHrelated occupations. In 2009, the International Agency for Research on Cancer (IARC) reviewed such evidence and, on the basis of increased risks of cancers of the respiratory, urinary tracts and skin, classified occupational exposures during coal gasification, coke production, coal tar distillation, paving and roofing, aluminum production and chimney sweeping into Group 1 carcinogens (Baan et al. 2009; IARC Working Group 2010a, b, 2012). We aimed to update the current evidence by carrying out a systematic review of cohort studies published between 2006 and January 2014 investigating the effects of PAHs on respiratory and urinary tract cancers in workers employed in selected industries (Boffetta et al. 1997; Bosetti et al. 2007; Baan et al. 2009; IARC Working Group 2010a, b, 2012), and to update the meta-analysis of Bosetti et al. (2007) by including the additional studies retrieved from the literature.
Materials and methods Study identification and data collection We carried out a literature search in MEDLINE, ISI Web of Science, SCOPUS and EMBASE of all cohort studies published as original articles, without any language restriction, from January 1, 2006, to January 31, 2014, on workers from selected industries characterized by exposure to PAH (Boffetta et al. 1997; Bosetti et al. 2007; Baan et al. 2009; IARC Working Group 2010a, b, 2012)—including those of the aluminum production, coal gasification, coke production, iron and steel foundries, coal tar and related products, carbon black and carbon electrodes production—and providing standardized mortality ratios (SMRs) and/or standardized incidence ratios (SIRs) of respiratory and urinary tract neoplasms. The following search string has been used: (polycyclic aromatic hydrocarbons OR PAH) AND (cohort OR prospective OR follow* OR longitudinal OR retrospective) AND ((occupation* cancer) OR (occupation* tumor) OR (occupation* cancer)). We hand-checked the reference list of the retrieved articles to identify additional relevant studies. When multiple studies were published on the same cohort, we included in the meta-analysis only the most informative one.
13
Arch Toxicol (2014) 88:1479–1490
This systematic review and meta-analysis were conducted according to the MOOSE guidelines (Stroup et al. 2000). Statistical analysis The overall SMR for each neoplasm of interest was calculated through an unweighted ratio of the sum of observed and expected death/cases (Breslow and Day 1987). Pooled relative risks (RRs) and the corresponding 95 % confidence intervals (CIs) were computed by using random-effects models in order to take into account the heterogeneity of risk estimates. Each study log(SMR) or log(SIR) was weighted by the inverse of its variance, which is mainly determined by the number of observed cases, plus the between-studies variance component τ2 estimated through the moment estimator (DerSimonian and Laird 1986). We assessed between-studies heterogeneity using chi-square test and quantified the inconsistency using the I2 statistic, which is the proportion of total variation contributed by between-studies variance (Higgins and Thompson 2002). The results of the meta-analysis were presented through cumulative meta-analysis over time in which studies were added one a time according to publication year, and results summarized as each new study was added. Forest plots of study-specific and overall RRs were also reported as supplementary materials, too. In some cases, the study-specific 95 % CIs might slightly differ from those published in the original publications because of rounding. Presence of publication bias was assessed by visual examination of funnel plot (Peters et al. 2008) and by applying the tests proposed by Begg and Mazumdar (1994), and by Egger et al. (1997).
Results The process for article selection is described in Supplementary Figure 1. Sixty-two out of 474 non-unique papers identified through the literature search were fully assessed for eligibility, and after evaluation of full-texts, 49 of these 62 articles were excluded due to investigation of other occupations or other cancer sites, case–control design, missing quantitative estimates of mortality (SMRs) or incidence (SIRs) for the investigated cancer sites, or multiple publications on the same cohort. A total of 13 papers (Dell et al. 2006; Hoshuyama et al. 2006; Spinelli et al. 2006; Wellmann et al. 2006; Gibbs et al. 2007; Gibbs and Sevigny 2007a, b; Bjor et al. 2008; Armstrong and Gibbs 2009; Behrens et al. 2009; Sim et al. 2009; Westberg et al. 2013; Zanardi et al. 2013) were considered eligible and included in the systematic review. From each of these studies, we extracted information related to the production type, period
Arch Toxicol (2014) 88:1479–1490
of workers employment, duration of follow-up, outcome, selected cancers investigated, observed and expected number of death or cases, and standardized mortality ratio (SMR) or incidence ratio (SIR) (Table 1). Since three papers (Gibbs and Sevigny 2007a; Armstrong and Gibbs 2009; Behrens et al. 2009) partly included the data of other papers, the quantitative meta-analysis considered 10 out of 13 papers (Dell et al. 2006; Hoshuyama et al. 2006; Spinelli et al. 2006; Wellmann et al. 2006; Gibbs et al. 2007; Gibbs and Sevigny 2007b; Bjor et al. 2008; Sim et al. 2009; Westberg et al. 2013; Zanardi et al. 2013). Most papers (n = 7) published after the Bosetti et al. 2007 review were on aluminum production industries (Spinelli et al. 2006; Gibbs et al. 2007; Gibbs and Sevigny 2007a, b; Bjor et al. 2008; Armstrong and Gibbs 2009; Sim et al. 2009), two were on iron and steel foundries (Hoshuyama et al. 2006; Westberg et al. 2013), two on asphalt workers (Behrens et al. 2009; Zanardi et al. 2013) and two on carbon black production (Dell et al. 2006; Wellmann et al. 2006). Table 1 gives summary information and major results of these papers. No further papers were published from January 1, 2006, to January 31, 2014, on coal gasification workers, coke production workers and carbon electrode manufacture. Supplementary Table 1 gives summary information of papers included in the meta-analysis published before January 2006. Aluminum production Seven papers were published between 2006 and January 2014 on cohorts of aluminum production workers (Spinelli et al. 2006; Gibbs et al. 2007; Gibbs and Sevigny 2007a, b; Bjor et al. 2008; Armstrong and Gibbs 2009; Sim et al. 2009) (Table 1). Five of them were from Canada (Spinelli et al. 2006; Gibbs et al. 2007; Gibbs and Sevigny 2007a, b; Armstrong and Gibbs 2009), one from Australia (Sim et al. 2009) and one from Sweden (Bjor et al. 2008). The Canadian study by Spinelli et al. (2006) was a 14-year update of a previously published historical cohort study of aluminum reduction plant workers (Spinelli et al. 1991). It included 6,423 workers employed between 1954 and 1997 at a Söderberg plant. No excess was found for cancer of the lung both for incidence (147 observed cases vs. 133.36 expected, SMR = 1.10, 95 % CI 0.93–1.30) or mortality (120 observed deaths vs. 112.15 expected, SMR = 1.07, 95 % CI 0.89–1.28). Similarly, no excess risk was found for laryngeal cancer both for incidence (10 observed cases vs. 12.67 expected, SMR = 0.79, 95 % CI 0.38–1.45) or mortality (4 observed deaths vs. 4.12 expected, SMR = 0.97, 95 % CI 0.27–2.49). For what regards bladder cancer, an excess incidence (SIR = 1.80, 95 % CI 1.45–2.21) was observed, while no excess mortality emerged (12 observed deaths vs. 8.63 expected,
1481
SMR = 1.39, 95 % CI 0.72–2.43). No excess was also found for kidney cancer both for incidence (SIR = 1.00, 95 % CI 0.62–1.52) or for mortality (SMR = 0.74, 95 % CI 0.30–1.52). A series of Canadian papers (Gibbs et al. 2007; Gibbs and Sevigny 2007a, b; Armstrong and Gibbs 2009) were based on workers employed in different time periods at the same aluminum reduction plant, i.e., on payroll before 1951 (Gibbs et al. 2007), update of data from Gibbs (1985) or first employed between 1950 and 1999 (Gibbs and Sevigny 2007b). Male workers on payroll before 1951 (Gibbs et al. 2007) had an increased mortality from lung (SMR = 1.36, 95 % CI 1.25–1.48, based on 538 deaths) and bladder (SMR = 2.24, 95 % CI 1.77–2.79, based on 78 deaths) cancer, but not from laryngeal (SMR = 0.91, 95 % CI 0.56–1.39, based on 21 deaths) nor from kidney cancer (SMR = 1.06, 95 % CI 0.70–1.55). A cohort of 10,454 first employees since 1951 followed up until the end of 1999 (Gibbs and Sevigny 2007b) showed no excess mortality from lung (140 observed deaths vs. 120.69 expected, SMR = 1.16, 95 % CI 0.97–1.36), bladder (5 observed deaths vs. 5.88 expected, SMR = 0.85, 95 % CI 0.28–1.98) and kidney (4 observed deaths vs. 8.16 expected, SMR = 0.49, 95 % CI 0.13–1.25) cancer, while a modest, but not significant, excess mortality from laryngeal cancer was observed (11 observed deaths vs. 7.01, SMR = 1.57, 95 % CI 0.78–2.81). Moreover, cancer incidence for 16,431 workers employed at the same plant and followed up from 1980 to 1999 was also investigated (Gibbs and Sevigny 2007a). An excess incidence was found for laryngeal (SIR = 1.32, 95 % CI 1.01–1.70, based on 60 cases), lung (SIR = 1.20, 95 % CI 1.10–1.31, based on 519 cases) and bladder (SIR = 1.82, 95 % CI 1.59–2.07, based on 230 cases) cancers, while no excess emerged for kidney cancer (SIR = 0.96, 95 % CI 0.72–1.26, based on 51 cases). In a subsequent re-evaluation of the combined cohort of 16,431 workers followed up from 1980 to 1999, Armstrong and Gibbs (2009) found an excess lung cancer mortality (677 observed deaths vs. 512.88, SMR = 1.32, 95 % CI 1.22–1.42). The Swedish study by Bjor et al. (2008) included 2,264 men employed for at least one year in the foundry between 1942 and 2000. An excess incidence was observed for cancer of the lung (40 observed cases vs. 27.3 expected, SIR = 1.48, 95 % CI 1.06–2.02), but no excess risk was found for bladder cancer (22 observed cases vs. 21.57 expected, SIR = 1.02, 95 % CI 0.64–1.54) nor for kidney cancer (12 observed cases vs. 11.54 expected, SIR = 1.04, 95 % CI 0.54–1.81). The study by Sim et al. (2009) included 4,396 males who had worked for at least three months between 1983 and 2002 in two Australian prebake aluminum smelters. There was no excess lung cancer mortality (SMR = 1.08,
13
13 1954–1997
Period of employment
Canada 10,454 workers (9726 male and 728 female) at 3 Prebake and Söderberg plants
Canada 16,431 workers (15,703 male and 728 female) at 3 Prebake and Söderberg plants
Sweden 2,264 males at 1 prebake and Söderberg smelter
Canada 16,431 workers (15,703 male and 728 female) at 3 Prebake and Söderberg plants
Gibbs and Sevigny (2007a)c
Bjor et al. (2008)
Armstrong and Gibbs (2009)c
1950–1999 Mortality
1958–2005 Incidence
1942–2000
1950–1999 (cohorts A, B, C, D)
1980–1999 Incidence
1950–1999 Mortality
1950–1999 Mortality
1957–1999 Mortality, incidence
Period of follow-up, outcome
1950–1999 (cohorts A, B, C, D)
First employed 1950–1999 (cohorts A, B, C, D)
On payroll before 1951 Canada 5,977 males at 3 Prebake and Söder- (cohorts A, B, C) berg plants
Canada 6,423 workers at 1 Söderberg plant
Country and study population
Gibbs and Sevigny (2007b)
Gibbs et al. (2007)b
Aluminum production Spinelli et al. (2006)a
Authors and publication year
4 120 12 7 10 147 90 21 21 538 78 27 11 140 5 4 60 519 230 51 40 22 12 677
Incidence Larynx Lung Bladder Kidney Larynx Lung Bladder Kidney Larynx Lung Bladder Kidney Larynx Lung Bladder Kidney Lung Bladder Kidney Lung
Observed deaths/ cases
Mortality Larynx Lung Bladder Kidney
Cancer
1.20 (1.10–1.31)d 1.82 (1.59–2.07)d 0.96 (0.72–1.26)d 1.48 (1.06–2.02) 1.02 (0.64–1.54) 1.04 (0.54–1.81) 1.32 (1.22–1.42)
0.79 (0.38–1.45) 1.10 (0.93–1.30) 1.80 (1.45–2.21) 1.00 (0.62–1.52) 0.91 (0.56–1.39) 1.36 (1.25–1.48) 2.24 (1.77–2.79) 1.06 (0.70–1.55) 1.57 (0.78–2.81) 1.16 (0.97–1.36) 0.85 (0.28–1.98) 0.49 (0.13–1.25) 1.32 (1.01–1.70)d
0.97 (0.27–2.49) 1.07 (0.89–1.28) 1.39 (0.72–2.43) 0.74 (0.30–1.52)
SMR/SIR (95 % CI)
Table 1 Summary information of cohort studies of workers exposed to polycyclic aromatic hydrocarbons in various industries and occupations, published between January 1, 2006, and January 31, 2014
1482 Arch Toxicol (2014) 88:1479–1490
Germany 1,535 male workers at one production plant
USA 5,011 male workers at 18 facilities
Italy 415 male asphalt workers
Germany 7,919 male asphalt workers
Mid 1930s–2003 Mortality 1976–1998 Mortality
1960–1998
Update of data from Gibbs (1985)
95 % CI calculated by using Byar’s formula (Breslow and Day 1987)
Data of a German sub-cohort of asphalt workers already included in Boffetta et al. (2003)
e
d
Data partly included in Gibbs et al. (2007) and Gibbs and Sevigny (2007b)
c
b
Update of data from Spinelli et al. (1991)
a
1964–2001 Mortality
Mid 1930s–2000
1964–1997
1965–2004 Mortality
1958–2004 Incidence
1958–2004
1975–1997
1980–1993 Mortality
1983–2002 Mortality, incidence
1983–2002
Before 1980–1993
Period of follow-up, outcome
Period of employment
CI confidence interval, SIR standardized incidence ratio, SMR standardized mortality ratio
Wellmann et al. (2006)
Carbon black production Dell et al. (2006)
Zanardi et al. (2013)
Asphalt workers Behrens et al. (2009)e
Westberg et al. (2013)
China 121,846 male workers in one company Sweden 3,045 male workers at ten iron and steel foundries
Australia 4,396 male workers in 2 prebake smelters
Sim et al. (2009)
Iron and steel foundry Hoshuyama et al. (2006)
Country and study population
Authors and publication year
Table 1 continued
3 39 13 14
Incidence Larynx Lung Bladder Kidney
Kidney
Lung Urinary organs Larynx Lung Bladder
Larynx Lung Bladder Lung
Larynx Lung Bladder Kidney
3
138 13 0 50 1
14 101 14 5
6 53 29 9
1,522
2 2
Lung
28
Bladder Kidney
Observed deaths/ cases
Mortality Lung
Cancer
1.04 (0.22–3.05)
0.97 (0.82–1.15) 0.72 (0.42–1.24) 0 (0–2.82) 2.18 (1.61–2.87) 0.38 (0.01–2.13)
3.74 (2.21–6.31) 1.77 (1.46–2.16) 3.29 (1.95–5.55) 1.00 (0.40–2.40)
1.77 (0.65–3.85) 1.58 (1.18–2.06) 1.27 (0.85–1.83) 0.71 (0.33–1.36)
0.85 (0.81–0.89)
0.76 (0.25–2.37) 1.23 (0.90–1.72) 1.26 (0.73–2.16) 1.99 (1.12–3.35)
1.01 (0.25–4.02) 0.79 (0.20–3.16)
1.08 (0.75–1.57)
SMR/SIR (95 % CI)
Arch Toxicol (2014) 88:1479–1490 1483
13
1484
95 % CI 0.75–1.57, based on 28 deaths). With reference to bladder cancer mortality, the SMR was 1.01 (95 % CI 0.25–4.02), on the basis of two observed deaths. Incidence data showed an excess risk for kidney cancer (14 observed cases vs. 7.04 expected, SIR = 1.99, 95 % CI 1.12–3.35), but not for laryngeal (SIR = 0.76, 95 % CI 0.25–2.37), lung (SIR = 1.23, 95 % CI 0.90–1.72) or bladder cancer (SIR = 1.26, 95 % CI 0.73–2.16). Overall, 1,314 lung cancer cases/deaths versus 1,154.7 expected were observed in ten studies (Giovanazzi and D’Andrea 1981; Rockette and Arena 1983; Moulin et al. 2000; Romundstad et al. 2000; Carta et al. 2004; Spinelli et al. 2006; Gibbs et al. 2007; Gibbs and Sevigny 2007b; Bjor et al. 2008; Sim et al. 2009), with a pooled RR of 1.07 (95 % CI 0.93–1.23, Table 2 and Supplementary Figure 2 for the cumulative meta-analysis over time). With reference to laryngeal cancer, there were 71 cases/deaths observed in seven studies (Giovanazzi and D’Andrea 1981; Moulin et al. 2000; Romundstad et al. 2000; Spinelli et al. 2006; Gibbs et al. 2007; Gibbs and Sevigny 2007b; Sim et al. 2009) versus 63.4 expected, with a pooled RR of 1.15 (95 % CI 0.91–1.45, Table 2 and Supplementary Figure 2 for the cumulative meta-analysis over time). Considering all cancers of the respiratory tract—including mainly lung cancer—a total of 1,349 cases/deaths were observed in 11 studies (Milham 1979; Giovanazzi and D’Andrea 1981; Rockette and Arena 1983; Moulin et al. 2000; Romundstad et al. 2000; Carta et al. 2004; Spinelli et al. 2006; Gibbs et al. 2007; Gibbs and Sevigny 2007b; Bjor et al. 2008; Sim et al. 2009) compared to 1,183.9 expected, with a pooled RR of 1.08 (95 % CI 0.95–1.23, Table 2, Supplementary Figure 3). A significant heterogeneity between studies was observed (p
