Lung cancer in an urban area in Northern Italy near a coke oven plant

Lung Cancer (2005) 47, 155—164

Lung cancer in an urban area in Northern Italy near a coke oven plant Stefano Parodia,∗, Emanuele Stagnarob, Claudia Casellac, Antonella Puppoc, Enrico Daminellid, Vincenzo Fontanab, Federico Valerioe, Marina Vercellic,f a

Epidemiology and Biostatics Section, Scientific Directorate, G.Gaslini Children’s Hospital, Largo G. Gaslini, 5-16145 Genoa, Italy b Environmental Epidemiology and Biostatistics, National Cancer Research Institute, Genoa, Italy c Tumour Registry Section, National Cancer Institute, Genoa, Italy d Environmental Pollution Control, Area 08, Province of Genoa, Italy e Environmental Chemistry Section, National Cancer Research Institute, Genoa, Italy f Oncology, Biology and Genetics Department, University of Genoa, Italy Received 15 March 2004 ; received in revised form 15 June 2004; accepted 17 June 2004 KEYWORDS Lung cancer; Incidence; Environmental pollution; Deprivation; Disease mapping; Bayesian models

Summary Coke ovens are well-known sources of potentially carcinogenic air pollutants, but studies on resident populations are still poor. This study investigates the incidence of lung cancer near a coke oven in Cornigliano, a district of the Genoa municipality in Northern Italy. Genoa proper and one district similar to Cornigliano as regards socio-economic deprivation were selected as referents. Incidence data were drawn from the Ligurian Cancer Registry for 1986—1997 calendar period. Concentrations of pollutants related to the industrial activity (namely benzene, benzo[a]pyrene, PM10 , CO, NO2 and SO2 ) were collected in selected locations before and after the coke oven closing. Spatial trend around the plant was assessed by Stone’s test, while the pattern of risk across Cornigliano was evaluated via disease mapping in a Bayesian model. A gradient of air pollutants was observed around the coke oven, which disappeared after its closing. In Cornigliano, 158 lung cancer cases were observed in males and 28 in females. Only a marginal excess risk was observed versus the two selected referents, while a gradient in the areas close to the plant emerged among females. Disease mapping revealed another cluster of risk for both sexes in the Eastern part of the district, where a foundry was operative until the early 1980s.

* Corresponding author. Tel.: +39 010 5636301; fax: +39 010 3776590.

E-mail address: [email protected] (S. Parodi).

0169-5002/$ — see front matter © 2004 Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.lungcan.2004.06.010

156

S. Parodi et al. The excess risk for females is consistent with pollution measurements and with other epidemiological evidence. The geographic pattern of incidence suggests a role of industrial air pollution as a risk factor for lung cancer. © 2004 Elsevier Ireland Ltd. All rights reserved.

1. Introduction Lung cancer in Italy is the most common neoplasm in males and the third in females, accounting for 25,778 and 5398 deaths, respectively, in 1997 [1]. Even though tobacco is responsible for about 80—90% of the whole lung cancer burden [2], many other exposures have been associated to this neoplasm, comprising indoor and outdoor air pollutants [3—4]. In urban air, the more common known or suspected carcinogens are inorganic particulates (arsenic, asbestos, cadmium, chromium, nickel), radionuclides, dioxins, benzene, benzo[a]pyrene (BaP) and other polycyclic hydrocarbons [5]. In recent years also fine particulate matter (PM10 ) and SO2 have been indicated as risk factors for this neoplasm [6]. Many epidemiological studies have reported an overall relative risk (RR) of about 1.5 for people living in more versus less polluted areas, with higher RR observed in populations in extremely polluted zones [7]. Interestingly, risk estimates from population cohort studies and casecontrol investigations are similar and, in general, they are also quite consistent with results from ecological surveys, indicating that confounding effects from personal exposures and ecological fallacy have played a limited role [5,7]. For example, a recent case-control study in an area in Southern Italy (Brindisi) exposed to heavy air pollution from a petrochemical plant reported a high mortality risk for persons resident within 2 km from the plant (odds ratio = 3.2), which remained almost unchanged in multivariable analyses, including smoking habit and education as confounders [8]. However, poor communities are often located in heavy polluted areas and a low socio-economic condition tends to be associated to an elevated risk for lung cancer [9—11]. For this reason, geographical comparisons between differently exposed populations should be carried out carefully controlling deprivation factors [12]. Cornigliano is a district of Genoa municipality, in Northern Italy, where a coke oven is located few hundred meters from the residential area. In the last 20 years, pollution emitted by this plant has given rise to several complaints from the resident population and ecologist movements. Finally, in February 2002, the coke oven was closed down by an order of the Genoa Court.

This study is aimed at evaluating the incidence risk of lung cancer in Cornigliano and at analysing its spatial distribution in relation to residential exposures to airborne pollutants emitted from the steel plant and particularly from the coke oven.

2. Materials and methods Cornigliano is located in the Western part of Genoa, in North-west Italy, along the Ligurian Sea. The urban area covers 5.2 km2 , corresponding to about 10% of the whole district, while 50% is covered by a large steel plant, which was built along the shore line in the 1950s. The facility stretches for more than 1 km along the sea coast and includes ore agglomeration, coke production, pig iron and steel production, rolling and shaping of steel ingots. The coke oven is placed in the east side of the steel plant and is made up of four oven batteries with a total of 120 ovens. Furthermore, in its extreme eastern sub-area, Cornigliano included a big foundry, operational until the early 1980s, while from East to West, a heavy trafficked thoroughfare passes through the whole district. The population of Genoa proper and of Rivarolo, the district more similar to Cornigliano as regards the socio-demographic structure, was selected as referents. More details about the areas under study have been published elsewhere [13]. Lung cancer cases (ICD-9: 162.0—162.9), diagnosed in the 1986—1997 period, were drawn from the Ligurian Cancer Registry data base. The corresponding populations at risk were estimated using the number of inhabitants by sex, age class and census tract (CT), registered at the 1991 Italian national census and provided by the Statistical Office of Genoa municipality, which also supplied data about the following socio-economic factors, referred to the same census: occupation, education and migration from Southern Italy. Prevailing wind direction was recorded in a station located in the hilly zone of the district, as described in a previous investigation [13]. Excluding northern winds, which blow the coke oven emissions far from the estates, eastern winds were prevailing. As a consequence, the central area of the district was the most heavily exposed to air pollutants emitted by the plant.

Lung cancer near a coke oven Since 1995, the Environmental Chemistry Laboratory of the National Cancer Research Institute of Genoa, in collaboration with the Genoa Province and the Regional Agency for Environmental Protection (ARPAL), has been carrying out a systematic study of air pollution in the Genoa municipality and particularly in Cornigliano. In this district, benzo[a]pyrene concentrations were registered by three monitoring stations located around the coke oven (sites A, B and C). Station A was placed at 320 m, while stations B and C were placed at about 600 m; A and B were leeward the plant. Another air sampler (D) was placed along a trafficked street in a Genoa’s district near Cornigliano, at 1200 m from the coke oven. Station A monitored also benzene concentrations (from 1998 to 2002), while other four samplers, placed very close to Station B, measured the following air pollutants concentrations: PM10 , CO, NO2 , SO2 . Briefly, the sampling and analytical method was the following: BaP adsorbed to particulate was extracted by sonication, purified by solid phase extraction and analysed by capillary gas liquid chromatography [14]. Mean PM10 fraction of airborne particles was collected for 24 h by high volume samplers, CO was daily measured by non-dispersive infrared spectrometry, while estimates of NO2 daily concentrations were obtained by ultraviolet fluorescence. Finally, daily mean SO2 concentrations were obtained by chemo luminescence analyser. Age-adjusted rates (AAR) per 100,000 inhabitants were computed, selecting the 1991 census of the Italian population as standard. Age-standardized incidence ratios (SIR) per cent between Cornigliano and the two selected reference populations were estimated via indirect standardization, using the age-specific incidence rates of Genoa population. For each census tract of Genoa municipality, a deprivation index (DI) was obtained averaging the ratio between unemployed and employed people, the proportion of low educated people and the proportion of residents immigrated from Southern Italy. After excluding Cornigliano (64 CT), the remaining 3035 CT of whole Genoa were aggregated in 30 groups on the basis of the corresponding percentiles of the above defined DI, thus obtaining a score, which allowed to estimate SIR for Cornigliano and Rivarolo, adjusted both for age and for deprivation. The corresponding 95% confidence intervals (95% CI) were obtained from the Poisson’s distribution tables [15]. The relative risk between the two districts was obtained from the ratio of the corresponding SIR, and the related 95% CI were calculated by the method reported by Breslow and Day [15].

157 Spatial trend of risk was evaluated via the Stone’s method [16,17], comparing cumulative standardized incidence ratios per 100 (cSIR) estimated within four concentric circles drawn around the coke oven with a 500 m increment. The corresponding expected cases were obtained using either the whole Cornigliano or Genoa proper as reference populations, adjusting either for age only or for age and deprivation. A scaling procedure was applied to the expected cases to obtain a cumulative SIR of 100 for the last circle, which included the whole district [17]. Furthermore, on account of the prevailing direction of winds in the area under study, Stone’s test was also repeated after restricting the analyses to the census tracts located in the North-western side of the coke oven [13]. Such a restriction allowed also to exclude the CTs around the foundry. Finally, the geographic pattern of incidence within Cornigliano was evaluated by computing Full Bayes Estimates of SIR (FBE—SIR) for each census tract, by the conditional autoregressive model proposed by Besag et al. in 1991 (BYM model) [18]. FBE—SIRs were computed by the WinBugs statistical package, that makes full Bayes estimates using the Gibb’s sampler applied to Marcov Chains Monte Carlo [19,20]. FBE—SIRs were obtained after 6000 iterations, excluding the first 1000 values to assure the achievement of convergence. Moreover, convergence was checked by recomputing the FBE—SIR by means of a further running of 5000 more iterations. Stone’s test was performed using an ad hoc software program developed by Prof. Corrado Lagazio of the Department of Statistics of the University of Trieste (Italy). Standardization procedures and graphical analyses were performed using SPSS for Windows statistical package [21].

3. Results Table 1 shows the annual mean BaP concentrations measured at four different locations from the coke oven. Before its closing in 2002, a clear increasing trend of these concentrations was observed while approaching the plant. Moreover, station B, located downwind the plant, showed higher BaP concentrations than station C, located at a similar distance not leeward. Finally, after the coke oven closing, a clear drop in BaP was observed in the two stations downwind the coke oven and, to a lesser extent, also in the other two locations. Table 2 reports yearly mean concentrations of the other pollutants monitored around the coke

158

S. Parodi et al.

Table 1 Mean annual BaP concentrations (ng/m3 ) measured at four different locations from the coke oven, before and after the plant closing Year

Station A Mean

Station B S.D.

Station C

Station D

Mean

S.D.

Mean

S.D.

Mean

S.D.

Before the coke oven closing 1995 11.4 8.5 1996 7.4 9.4 1997 10.8 10.7 1998 8.8 8.0 1999 7.5 6.2 2000 8.3 5.7 2001 6.4 3.6

3.7 3.4 4.0 4.2 3.3 3.6 3.8

2.2 1.8 2.1 2.6 1.1 1.6 1.3

2.4 2.0 2.0 1.3 1.6 1.4 0.9

1.3 1.0 1.1 0.5 0.7 1.0 0.6

1.9 2.1a 1.2 1.1 1.0 0.8 0.7

0.7 0.8 0.6 0.8 0.6 0.7 0.4

After the coke oven closingb 2002 0.2

0.3

0.4

0.4

0.5

0.4

0.5

0.1

S.D.: standard deviation. Station A is located at 320 m from the coke oven, downwind; Station B at 600 m, downwind; Station C at 600 m, not downwind; Station D at 1200 m, not downwind. a Estimate based on five months: January, February, March, November and December. b Data collected from 8 February 2002 to 31 January 2003.

oven. After the coke oven closing, benzene concentration dropped from about 18 ␮g/m3 in 2000—2001 to 2.7 ␮g/m3 in 2002. A clear reduction was also observed for SO2 and, to a lesser extent, for PM10 , while no noteworthy variation was evident for CO and for NO2 . Fig. 1 shows the association between lung cancer incidence in Genoa, expressed as SIR (reference population: Genoa proper), and the DI score. In males, a linear trend in risk by DI levels was evident (R2 = 0.87), while in females, no clear pattern emerged. A similar result was obtained after stratifying data by age group (i.e. 40—69 years and 70 years or more, data not shown). Fig. 2 shows the relationship between DI and the distance from the coke oven in Cornigliano for each

CT. The fitted lowess line (smoothing obtained by moving median with 50% of width) indicates poor socio-economic conditions for the sub-areas close to the plant and a tendency to reach a plateau at about 1000 m of distance. Table 3 presents incidence cases and AAR for the areas under study and the corresponding SIR. In males, lung cancer incidence was similar in Cornigliano and in Rivarolo (AAR = 137.8 and 131.6, respectively) and slightly higher than in Genoa proper (AAR = 118.7). SIR was significantly higher in both the districts adjusting for age only, and they became similar to the referent (Genoa) after controlling for DI. In females, AAR in Cornigliano was similar to Genoa proper (AAR = 21.3 and 21.9, respectively), while in Rivarolo, a lower risk was ob-

Table 2

Benzene, PM10 , SO2 , NO2 and CO concentrations before and after the plant closing

Year

Benzene (␮g/m3 )

PM10 (␮g/m3 )

CO (mg/m3 )

NO2 (␮g/m3 )

SO2 (␮g/m3 )

Mean

Mean

S.D.

Mean

S.D.

Mean

S.D.

Mean

S.D.

Before the coke oven closing 1995 — — 1996 — — 1997 — — 1998 20.1a 43.6 33.3 1999 10.8b 2000 18.5 31.4 2001 18.4 36.3

79 59 65 61 59 59 57

42.0 24.7 29.2 22.3 22.5 22.5 18.4

4.0 3.6 4.5 3.1 3.1 2.3 1.8

2.3 2.3 3.2 2.4 2.2 1.5 11.4

52 79 48 59 — 69 62

31.3 37.2 38.4 31.9 — 27.4 28.5

29 29 52 79 72 78 67

16.5 12.2 41.1 59.0 60.7 57.2 42.8

After the coke oven closing 2002 2.7 2.7

43

15.6

1.6

1.2

69

27.1

14

15.3

S.D.

S.D.: standard deviation. a Data recorded from May to August. b Data recorded from January to April.

Lung cancer near a coke oven

159

Fig. 1. Lung cancer risk in Genoa municipality (1986— 1997) by deprivation index.

served (AAR = 15.1). SIR adjusted only for age confirmed a similar risk between the Cornigliano and Genoa populations and a lower one for Rivarolo residents, with a corresponding RR between the two districts of 1.4 (95%CI: 0.83, 2.27). After adjusting for DI, the SIR in Cornigliano was slightly lowered, while the risk in Rivarolo remained almost unchanged. Table 4 shows the geographic trend of incidence risk by sex around the coke oven. In males, the highest risk was observed in the first circle drawn around the plant (SIR = 117.2 versus whole Cornigliano and SIR = 116.0 versus Genoa proper). However, the corresponding spatial trends were not statistically significant (P = 0.232 and P = 0.236, respectively). After adjusting for DI, this pattern tended to disappear. Similar results were obtained restricting the analyses to the Western part of the district (data not shown). Among females, the highest risk was also observed around 500 m from the coke oven Table 3

Fig. 2. Deprivation index by distance from the coke oven in Cornigliano area.

(SIR = 188.4 versus whole Cornigliano and SIR = 187.6 versus Genoa proper), even if in the last circle (more than 1500 m), an elevated incidence also emerged (SIR = 150.3 versus whole Cornigliano and SIR = 148.5 versus Genoa proper). Statistical significance of spatial trend was borderline (P = 0.07 for both analyses). Adjusting for DI resulted in a small reduction of risk close to the coke oven (SIR = 155.6). The exclusion of the Eastern part of the district caused a drop of SIR in the last circle in all analyses (from about 150 to 81—103, data not shown). Fig. 3 displays the geographic pattern of lung cancer risk in males. The highest FBE—SIRs (between 110 and 130) were found in the Eastern part of the district (where 12 cases were observed), in a census tract close to the coke oven (which included five cases) and in few other tracts, without a clear geographic pattern, while the lowest risks were

Lung cancer incidence in Cornigliano and in the two referent areas (1986—1997)

Area Males Cornigliano Rivarolo Genoa Females Cornigliano Rivarolo Genoa

N

AAR

SIR1

95% C.I.

158 343 5833

137.8 131.6 118.7

118.6 113.5 100

100.8 101.8 —

28 44 1258

21.3 15.1 21.9

97.5 70.8 100

64.8 51.4 —

SIR2

95% C.I.

137.8 125.8 —

98.9 102.8 100

83.5 92.2 —

114.3 114.0 —

140.9 95.0 —

89.2 71.9 100

59.3 52.3 —

128.9 96.5 —

N: number of incidence cases; AAR: age-adjusted rates per 100,000 inhabitants; SIR1: standardized incidence ratio adjusted for age only; SIR2: standardized incidence ratio adjusted for age and deprivation index.

160

Table 4

S. Parodi et al.

Lung cancer incidence spatial trend: SIR at selected distance from the coke oven Males Observed

Females SIR

cSIR

Observed

SIR

cSIR

3 15 6 4

188.4 92.7 79.3 150.3

188.4 101.2 94.7 100.0

3 15 6 4

187.6 93.5 78.4 148.5

187.6 102.0 94.8 100.0

Referent: Genoa population stratified by age and deprivation index Bands (m) 0.730 ≤500 10 101.1 101.1 3 500—1000 81 92.5 88.9 15 1000—1500 53 129.0 100.4 6 >1500 14 96.3 100.0 4

155.6 93.9 81.0 149.0

155.6 100.5 94.8 100.0

Referent: Cornigliano population stratified by age Bands (m) ≤500 10 117.3 117.3 500—1000 81 88.2 90.7 1000—1500 53 124.5 100.8 >1500 14 92.7 100.0 Referent: Genoa population stratified by age Bands (m) ≤500 10 116.0 116.0 500—1000 81 88.5 90.9 1000—1500 53 124.2 100.9 >1500 14 91.9 100.0

P 0.230

0.242

P 0.072

0.073

0.124

cSIR: cumulative standardized incidence ratio per 100.

observed in a South-western sub-area (FBE—SIR ≤ 90). In females (Fig. 4), clusters of risk were observed close to the coke oven (FBE—SIR: 110—130, due to three observed cases), in a South-western sub-area (FBE—SIR: 110—130, due to five cases), in the hilly zone in the Northern part of the district (FBE—SIR: 130—150, two cases) and in some census tracts at the extreme Eastern part (FBE—SIR: 130—150, four cases).

4. Discussion Coke ovens are well-known sources of potential carcinogenic and toxic compounds, comprising benzene, benzo[a]pyrene, other polycyclic aromatic hydrocarbon, SO2 and fine particulate matter [22,23]. Many occupational studies highlighted an excess risk for lung cancer in people working close to coke oven plants [24,25], whereas few evidences have been so far collected about the risk for resident populations [12,26]. Coke ovens are in general located within other industrial plants, in particular steel mills and mining activities, then emissions from other parts of the plant may confound the relationship between cancer risk and carcinogenic exposure from coke oven [22]. However, spatial trend of BaP concentration in Cornigliano and the drop of BaP, benzene, SO2 and particulate matter, after the coke oven closing, indicate that this plant

has been the major source of air pollution in the district. These observations are consistent with a previous investigation estimating benzene concentrations in the same district, using passive tubes placed outside residences [13]. These results also pointed out a role of wind direction, indicating a strong impact of air pollutants on the central part of the district and, in particular, in the area located within 500 m of the plant, without any fallout at a higher distance. Moreover, air pollutants from other part of the steel mill and other sources, comprising car traffic, seemed to have played a marginal role. The absence of other major sources of pollution in the area under study, except for the above cited foundry in the extreme Eastern part of the district, suggests that a similar pattern of spatial distribution is likely to be present since 1954, when the coke oven began its activity. Deprivation index used in this study does not include some variables commonly used in other investigations, like car ownership or family income [27,28], because they were not available in the Italian national census database. On the contrary, we decided to include the immigration from Southern Italy because neighbourhoods with a higher proportion of immigrants are likely to be poorer. An objection to such an approach might be that people living in Southern Italy have a reduced risk of many cancer sites, including lung cancer. However, the percentage of immigrants was positively correlated with

Lung cancer near a coke oven

161

Fig. 3. Pattern of incidence risk of lung cancer in Cornigliano area—–males.

the other two selected socio-economic variables, while no protective effect of immigration was observed on lung cancer risk (data not shown). The DI computed in this investigation showed a very high correlation with lung cancer in males and, as expected, it was also associated to the distance from the industrial plant. Moreover, removing the immigration rate from DI computation caused a reduction of such an association (data not shown), suggesting that the inclusion of this variable as socioeconomic indicator might have reduced the risk of residual confounding. Finally, the strong association between DI and lung cancer in males and a lack of a similar pattern in females are also consistent with results from a mortality investigation in Rome [29]. For these reasons, the DI estimated in this analysis seems to be an adequate proxy for the

socio-economic condition at least in Italian Northern cities. Results from standardization on DI score indicate that the excess risk observed among males in Cornigliano and in Rivarolo is mainly attributable to differences in socio-economic characteristics. As regards females, Cornigliano shows a risk similar to Genoa proper, but more elevated than Rivarolo and such excess remains after adjusting for deprivation. However, this finding seems to be mainly due to a low risk for Rivarolo. On the contrary, the high risk observed in females living close to the coke oven and the spatial trend of incidence around the plant are in good agreement both with our hypothesis and with the environmental measurements. Furthermore, they are also consistent with results by Pless-Mulloli et al. [12],

162

S. Parodi et al.

Fig. 4. Pattern of incidence risk of lung cancer in Cornigliano area—–females.

who reported a similar risk for females living in a region in Northern England exposed to many industrial sources of air pollution, including a coke oven (SMR = 393 in the more polluted zone versus SMR = 242 and SMR = 185 in the two less exposed sub-areas, corresponding to RR = 1.6 and RR = 2.1, respectively), even if such finding was limited to the 0—64 age group. Moreover, a recent investigation on the mortality risk for lung cancer around a big coke oven plant in Sidney, Nova Scotia (Canada), found a statistically significant higher risk in both sexes, more evident among females (SMR = 141 and 176, respectively, referent: Canadian population) [26] (results from this study are also available at the Internet website: http://www.safecleanup.com/health/band.htm). Interestingly, such a risk was also in a very good agreement with that reported in the present study

(Table 4). Finally, results from our investigation are also in accordance with estimates generally observed in ecological surveys around other industrial sources of air pollutants [4], including a recent study in La Spezia Province (Italy), which revealed an RR between 1.5 and 2 for females living in two areas exposed to emissions from many industrial plants [30]. The lack of a similar pattern among males in our investigation might be attributable to a lower rate of residential exposure and to confounding from occupational activities and lifestyle factors, in particular smoking habits. Measurements of tobacco consumption for the areas under study are not available, but an estimate for the whole Italy, referred to 1965, reported a very low proportion of smokers in females (7.7%) compared to males (60.0%), even if only 10 years later, the corresponding smoking prevalence had

Lung cancer near a coke oven changed to 16.3% and 53.2%, respectively, continuing to decrease in males and tending to a plateau in females (45.6% and 17.7% in 1983, respectively) [31]. These figures could also explain the very high correlation between socio-economic index and lung cancer incidence in males and the lack of a clear association between these variables in females, also reported by two other Italian studies [29,30]. Bayesian disease mapping highlighted a small cluster of risk for females living near the coke oven plant, in agreement with the results from the Stone’s test. Furthermore, an elevated risk in both sexes was observed in an extreme Eastern sub-area, where a big foundry was operational until the early 1980s. Interestingly, such finding is consistent with the pattern of incidence of non-Hodgkin lymphoma, observed in a recent study on the same area [13]. The major limit of this investigation is probably the lack of information about individual exposures, which prevents from drawing causal inference. Residual confounding, in particular from smoking habits, cannot be completely ruled out even after adjusting for socio-economic factors [32]. Moreover, uncontrolled migration rates could have also influenced the pattern of risk estimated in this investigation. However, higher emigration rates are likely for people living close to the industrial site, so probably reducing the risk estimates around the plant. In spite of the unavoidable limits of this study, the observed spatial pattern of incidence is consistent both with air pollution measurements and with estimates from other epidemiological surveys [8,12,13,26], suggesting a role of industrial air pollution as a risk factor for lung cancer. However, further investigations are needed to trace back the exposures responsible for the excess risks observed.

Acknowledgements We thank Dr. Arvati, Dr. Romani and Mr. D’Orazi (Genoa Municipality) that provided us with the data about the population and geographic maps; Dr. Ceppi (National Cancer Research Institute of Genoa) for precious methodological advice; Prof. Lagazio (University of Trieste), who implemented the statistical routine for Stone’s test; and Mr. Vittorio Meirano for language revision.

References [1] Negri E, La Vecchia C, Decarli A. Cancer mortality in Italy, 1997: quantify the fall in rates in women and men. Tumori 2001;87:290—8.

163 [2] Doll R, Peto R, Wheatley K, Gray R, Sutherland I. Mortality in relation to smoking: 40 years’ observations on male British doctors. BMJ 1994;309:901—11. [3] Williams MD, Sandler AB. The epidemiology of lung cancer. Cancer Treat Res 2001;105:31—52. [4] Bilello KS, Murin S, Matthay RA. Epidemiology, etiology, and prevention of lung cancer. Clin Chest Med 2002;23:1—25. [5] Cohen AJ. Outdoor air pollution and lung cancer. Environ Health Perspect 2000;108(S4):743—50. [6] Pope 3rd CA, Burnett RT, Thun MJ, Calle EE, Krewski D, Ito K, et al. Lung cancer, cardiopulmonary mortality, and longterm exposure to fine particulate air pollution. J Am Med Assoc 2002;287:1132—41. [7] Katsouyanni K, Pershagen G. Ambient air pollution exposure and cancer. Cancer Causes Control 1997;8:384—91. [8] Belli S, Benedetti M, Comba P, Lagravinese D, Martucci V, Martuzzi M, et al. Case-control study on cancer risk associated to residence in the neighborhood of a petrochemical plant. Eur J Epidemiol 2004;19:49—54. [9] Kogevinas M, Pearce N, Susser M, Boffetta P (Eds.), Social Inequalities and Cancer, Lyon, IARC Scientifical Publications, 1997. [10] Hart CL, Hole DJ, Gillis CR, Smith GD, Watt GC, Hawthorne VM. Social class differences in lung cancer mortality: risk factor explanations using two Scottish cohort studies. Int J Epidemiol 2001;30:268—74. [11] Mao Y, Hu J, Ugnat AM, Semenciw R, Fincham S, the Canadian Cancer Registries Epidemiology Research Group. Socioeconomic status and lung cancer risk in Canada. Int J Epidemiol 2001;30:809—17. [12] Pless-Mulloli T, Phillimore P, Moffatt S, Bhopal R, Foy C, Dunn C, et al. Lung cancer, proximity to industry, and poverty in northeast England. Environ Health Perspect 1998;106:189—96. [13] Parodi S, Vercelli M, Stella A, Stagnaro E, Valerio F. Lymphohaematopoietic system cancer incidence in an urban area near a coke oven plant: an ecological investigation. Occup Environ Med 2003;60:187—93. [14] Valerio F, Pala M, Lazzarotto A, Stella A, Ciccarelli F, Balducci D. Air quality standard for benzo(a)pyrene (BaP) in Genoa (1994—1995). Polycyclic Aromat Compd 1996;9:61—6. [15] Breslow NE, Day NE. Statistical Methods in Cancer Research—–Volume II—–The Design and Analysis of Cohort Studies. Lyon, International Agency for Research on Cancer; 1987. [16] Stone RA. Investigations of excess environmental risks around putative sources: statistical problems and a proposed test. Stat Med 1988;7:649—60. [17] Shaddick G, Elliott P. Use of Stone’s method in studies of disease risk around point sources of environmental pollution. Stat Med 1996;15:1927—34. [18] Besag J, York J, Molli´ e A. A Bayesian image restoration, with two application in spatial statistics. Ann Inst Stat Math 1991;43:1—59. [19] Spiegelhalter D, Thomas A, Best N, WinBUGS Version 1.3 User Manual, 2000. Available at rnet website: http://www.mrc-bsu.cam.ac.uk/bugs. [20] Molli´ e A. Bayesian mapping of disease. In: Gilks WR, Richardson S, Speigelhalter DJ, editors. Markov Chain Monte Carlo in Practice. Boca Raton/London/New York/Washington DC: Chapman & Hall; 1996. p. 359—79. [21] SPSS Inc., SPSS 9.0 User’s Guide, Chicago, SPSS Inc., 1999. [22] Dolk H, Thakrar B, Walls P, Landon M, Grundy C, Saez Lloret I, et al. Mortality among residents near cokeworks in Great Britain. Occup Environ Med 1999;56:34—40.

164 [23] Lambert TW, Lane S. Lead, arsenic and polycyclic aromatic hydrocarbons in soil and house dust in the communities surrounding the Sydney, Nova Scotia, Tar Ponds. Environ Health Perspect 2004;112:35—41. [24] Moolgavkar SH, Luebeck EG, Anderson EL. Estimation of unit risk for coke oven emissions. Risk Anal 1998;18:813— 25. [25] Costantino JP, Redmond CK, Bearden A. Occupationally related cancer risk among coke oven workers: 30 years pf follow-up. Occup Environ Med 1995;37:597— 604. [26] Band P, Camus M, Henry J, Zielinski J, Jiang H, Semenciw R, et al. Mortality rates within Sidney Nova Scotia. Ottawa, Ontario, Canada, Environmental Protection Service, Environment Canada, 2003. [27] Townsend P, Simpson D, Tibbs N. Inequalities in health in the city of Bristol: a preliminary review of statistical evidence. Int J Health Serv 1985;15:637—63.

S. Parodi et al. [28] Carstairs V. Deprivation indices: their interpretation and use in relation to health. J Epidemiol Community Health 1995;49(S2):3—8. [29] Michelozzi P, Perucci CA, Forastiere F, Fusco D, Ancona C, Dell’Orco V. Inequality in health: socioeconomic differentials in mortality in Rome, 1990—95. J Epidemiol Community Health 1999;53:687—93. [30] Parodi S, Baldi R, Benco C, Franchini M, Garrone E, Vercelli M, et al. Lung cancer mortality in a district of La Spezia (Italy) exposed to air pollution from industrial plants. Tumori 2004;90:181—5. [31] La Vecchia C. Smoking in Italy, 1949—1983. Prev Med 1986;15:274—81. [32] Dolk H, Mertens B, Kleinschmidt I, Walls P, Shaddick G, Elliott P. A standardisation approach to the control of socioeconomic confounding in small area studies of environment and health. J Epidemiol Community Health 1995;49(S2):9—14.