MATERIALS AND METHODS

Characteristics of the areas under study

Cornigliano is situated in the western part of Genoa, along the Ligurian Sea. The coastline is level, with hills extending northward and all around the residential area. The Cornigliano urban area covers 5.2 km², corresponding to about 10% of the whole district, while 50% is covered by an industrial area that was reclaimed in the 1950s by filling the sea with debris from nearby hills. This new land was used to build an iron and steel factory, made up of a coke oven, blast furnaces, and plate rolling mills. The entire facility stretches for more than 1 km along the sea coast. The coke oven is located in the east of the steel plant and yields coke, which is then allotted to the blast furnace for the production of cast iron. From east to west a highly trafficked thoroughfare passes through the whole district. Finally, the extreme western subarea included a big foundry, that was operational until the early 1980s.

The populations of Genoa proper and of Rivarolo, the district more similar to Cornigliano with regard to sociodemographic characteristics, were selected as controls. Table 1 shows the distribution of demographic and socioeconomic characteristic (age, occupation, education, and immigration) of the areas under study.

Rivarolo lies to the north of Cornigliano, beyond a hilly barrier, and spread north–south for 2 km and nearly 5 km from east to west. Thirteen per cent of the district is covered by urban area. The district is crossed from north to south by a highly trafficked street and by the Genoa–Milan motorway.

Source of cases and populations

All malignant haematological cancer cases (ICD-9: 200.0–208.9) and myelodysplasias (ICD9: 238.7), diagnosed in the 1986–94 period in the areas under study, were drawn from the incidence data base of the Ligurian Cancer Registry. The corresponding populations at risk were estimated using the number of inhabitants by sex, age, class, and census tract, registered at the 1991 Italian national census and provided by the Statistical Office of Genoa municipality.

Environmental pollution control

Five conveyed benzene sources are present in the steel factory: four coke oven and one Cowper’s chimneys; emissions from the two sources reach approximately 425 g and 220 g of benzene per hour, respectively.

Benzene concentrations in air (hourly mean) were measured from April 1998 to April 1999 in a fixed monitoring station operated by the Genoa Province, placed 330 metres downwind of the coke oven’s centre. The annual mean benzene concentration measured by the fixed station was 15.1 (16.2) μg/m³. Benzene concentration was strongly dependent on wind direction. The highest benzene concentration, 28 μg/m³, was found with southeast winds, while the lowest, 3.2 μg/m³, was found with northeast winds.

During the period 15–21 April 1999, an estimate of the mean weekly benzene concentrations outside 27 Cornigliano and 28 Rivarolo houses was simultaneously performed by the Environmental Chemistry Section of the National Cancer Research Institute of Genoa (Italy),²⁰ using passive tubes placed outside residences.²¹ Unfortunately, this method did not allow an estimate of the exposure for the whole of either monitored area. In particular, with regard to Cornigliano, about 60% of its 65 census tracts were covered, mainly close to the coke oven, corresponding to 78% of the whole population. However, such measurements confirmed that benzene concentrations were higher in Cornigliano compared to Rivarolo (table 2). For 47% of the exposure time, sampling sites were downwind of the coke oven (winds from south and east–southeast). A significant difference was also observed in the mean toluene/benzene ratio: with southeast winds this value was 0.25, very similar to that found in coke oven emissions; instead, the benzene and toluene composition in samples collected with northeast winds was different, their mean ratio being 3.6, similar to that found in automobile emissions. All these results confirmed that, with southeast winds, the fixed monitoring station and the urban area of Cornigliano lie downwind of the coke oven and the diffuse benzene emissions of this plant significantly increase the benzene concentration.

The meteorological information necessary for the evaluation of these measurements was obtained from a station of the Genoa Province placed in the hilly zone of Cornigliano.

Statistical methods

Age adjusted rates (AAR) per 100 000 inhabitants were computed for all lymphohaematopoietic neoplasms (ICD-9: 200.0–208.9, 238.7) and for five selected sites: non-Hodgkin’s lymphoma (ICD-9: 201.0–201.9, 202.0–202.9), Hodgkin’s disease (ICD-9: 200.0–200.9), multiple myeloma (ICD-9: 203.0–203.9), leukaemia (ICD-9: 204.0–208.9), and myelodysplasia (ICD-9: 238.7). The 1991 census of the Italian population was selected as standard.

Relative risks (RR) and their related 95% confidence intervals (CI) between the area under study (Cornigliano) and the two selected controls were estimated via Poisson regression model.²²

Spatial trend of risk was evaluated by comparing cumulative standardised incidence ratios per 100 (cSIR) estimated within five concentric circles with a 500 m increment drawn around the coke oven. The related statistical significance was assessed by the Stone’s test.²³

The standard population utilised for the cSIR estimation was obtained by pooling together the Cornigliano and Rivarolo populations. Using the same standard population, the geographic pattern of incidence within Cornigliano was evaluated by computing full Bayes estimates of standardised incidence ratio per 100 (FBE-SIR) for each census tract, by the conditional autoregressive model proposed by Besag et al in 1991.²⁴ In each census tract, the FBE-SIR represents an estimate of the incidence risk with respect to the selected standard, computed taking into account the possible correlation between such a risk and those from the adjacent sub-areas, and may be regarded as a smoothing function applied to the more common (maximum likelihood) SIR to reduce the variability due to small population.²²

FBE-SIRs were computed by the WinBugs statistical package,²⁵ which makes full Bayes estimates using the Gibb’s sampler applied to Marcov chains Monte Carlo.²⁶

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). The remaining statistical analyses were performed using SPSS²⁷ and GLIM 4.0²⁸ statistical packages.

RESULTS

Table 3 shows incidence cases and AAR for the areas under study and the corresponding RR.

Cornigliano showed a statistically significant higher risk in males for all lymphohaematopoietic cancers, compared to both Rivarolo (RR 1.7) and Genoa (RR 1.6). Higher RR were also observed for NHL (RR 2.4 v Rivarolo and RR 1.7 v Genoa) and for leukaemia (RR 2.4 v Rivarolo and RR 1.9 v Genoa).

As far as females are concerned, statistically significant higher RR were not observed either for all lymphohaematopoietic neoplasms or for all selected sites, even though higher risks were found for NHL compared to Rivarolo (RR 1.2) and for HD compared to both control areas (RR 2.3 v Rivarolo and 1.3 v Genoa, respectively). The other selected sites showed higher or similar rates in the two control areas.

Table 4 shows the geographic trend of incidence risk for leukaemia and NHL by sex around the coke oven. In males no increased risk was observed for NHL near the coke oven, while the risk for leukaemia seemed to rise with increasing distance from the oven. Stone’s test did not reveal any statistically significant trend.

In females a rising, even if statistically non-significant (p = 0.185, Stone’s test), risk for NHL was observed approaching the plant, while a statistically significant increase in risk was observed for leukaemia, even if such a trend was due to just one case close to the coke oven, followed by zero cases in the second band (between 500 and 1000 m from the coke oven).

Figure 1 shows the pattern of incidence risk of NHL in males. In most parts of the district, FBE-SIR varied between 120 and 200 without showing any clear pattern, while in the extreme eastern part a cluster of higher risk (more than 200) was observed, due to four observed cases.

In females (fig 2), FBE-SIR of about 80–120 were observed in most of the considered area, except for the extreme eastern part where a higher risk was observed, due to one observed case in a sub-area with a small population. Higher risks, from 120 to 200, were also observed in the southeastern area, in front of the coke oven, and in some other census tracts, especially in the central-southern part close to the district’s main thoroughfare (even if in this latter case no clear pattern emerged).

Figure 3 shows the geographic pattern of leukaemia risk in males. Higher FBE-SIR were observed in the extreme west and in a small south-central sub-area. A cluster of higher incidence, corresponding to four census tracts, was observed in the extreme northern part of the district. The remaining FBE-SIR were between 160 and 200 in most central census tracts, and between 120 and 160 in southern and in eastern parts of the district.

The small number of leukaemia cases observed in females did not allow us to estimate the FBE-SIR or to evaluate the corresponding geographic pattern.

DISCUSSION

Differences in benzene concentrations measured in Cornigliano versus Rivarolo and the lower toluene/benzene ratio observed in the former area, particularly in the high floors, confirmed that the coke oven was the main benzene source in the district. Coke ovens are a well known industrial source of benzene and other environmental pollutants, including polycyclic aromatic hydrocarbons (PAHs).⁵ In particular, a previous investigation in the Genoa municipality, covering the 1994–95 calendar period, indicated that the coke oven in Cornigliano was also the main responsible source of the high concentrations of benzo(a)pyrene in the area.²⁹ However, since benzene related PAHs are thought to be more associated to lung cancers and some other neoplasms,^5,14 the present investigation focused mainly on the possible association between environmental benzene pollution and lymphohaematopoietic cancers.

In the study period, 89 incidence lymphohaematopoietic cancer cases were observed in the district of Cornigliano, of which 62 were either leukaemia or NHL, accounting for nearly 70% of the total number of cases. The small number of cases observed for the other sites considered makes comparison of the results from the two control areas difficult to evaluate. For instance, the risk for Hodgkin’s disease for females living in Cornigliano was 1.3 times higher than the risk of the corresponding population of Genoa proper and more than twice that of Rivarolo. This excess incidence was caused by only three cases observed in Cornigliano, and it is impossible, especially in a descriptive context, to state whether the lack of statistical significance was a result of low statistical power or the absence of any real effect. For these reasons, this study endeavoured to evaluate the risk pattern for the two more common lymphohaematopoietic sites.

In the district of Cornigliano, a statistically significant higher risk for leukaemia was observed in males, showing a tendency to be higher versus the population of Rivarolo than versus the population of Genoa proper. This difference is in agreement with the hypothesis of a slight protective effect of deprivation index variables on the risk of leukaemia, suggested for some European areas, including Turin in Italy,³⁰ even if the related 95% confidence intervals are largely overlapped.

Although leukaemia is recognised to constitute a group of several diseases with possibly different aetiological factors,^15,31 we opted to analyse the risk for the entire category because of the small number of observed cases. A common objection to such an approach is that leukaemia types other than AML seem to be less associated with benzene exposure.^6,31 However, in the Cornigliano population two cases of AML were observed during the period analysed; any analysis of such a subgroup would therefore yield unreliable results.

A higher risk of leukaemia with respect to Rivarolo was also observed in male children (under 15); the estimated RR were 4.6 versus Rivarolo and 3.6 versus Genoa (data not shown). These estimates, however, based only on two cases in males, do not enable us to draw reliable conclusions. Furthermore, there seems to be no relation between the residence of these cases and the coke oven (data not shown).

A higher incidence was also observed for NHL in males. The lack of an excess risk in females, for both leukaemia and NHL, could suggest an effect due to occupational exposure, which is more likely to occur in males. The lower risk for leukaemia in females, even if not statistically significant, remains difficult to interpret.

To our knowledge, there is little evidence in the scientific literature of a different effect by gender of chronic exposure to benzene. Women are under-represented in occupational studies, but their higher body fat ratio and their tendency to metabolise benzene more effectively than men should indicate a higher risk for such a group.^14,32 It is not clear, however, whether these factors also play a negative role under chronic low exposure.

As far as children are considered, the higher incidence of leukaemia in males, observed in most countries around the world,³³ was suggested to be due more to higher exposure from different activity patterns, such as time spent in garages and workshops or an intensive use of consumer products containing benzene, rather than to a higher susceptibility to the pollutant.³⁴ Nevertheless, a different behaviour of male children living in Cornigliano with respect to those living in the control area, seems to be unlikely. In addition, a recent study conducted in Rome found a higher risk for leukaemia in children, particularly males, living within 150 metres of heavily trafficked thoroughfares.³⁵

Benzene concentrations outside Cornigliano residences suggest that people living along the area downwind of the coke oven are likely to be more heavily exposed to environmental benzene than people living in other Cornigliano areas or in Rivarolo. However, in most of the studies aimed to estimate the risk of leukaemia associated with highly trafficked roads, a distance under 200 m is usually employed.^10,35 Therefore, the impact of benzene from the coke oven, which is located roughly 300 metres from the residential area, might concern only a small part of the population; this would make the results from the spatial trend analysis difficult to interpret, especially without accurate measurements of air benzene concentrations. In particular, one case of leukaemia observed in a female residing in the first circle around the coke oven (0–500 m), even if accounting for a nearly doubled risk, is hard to ascribe to environmental pollution. Furthermore, the instability of a risk estimated on only one case does not allow any clear cut conclusions to be drawn.

The geographic pattern of risk emerging from the FBE-SIR maps is consistent with the results from the spatial trend analyses and did not highlight any cluster of cases around the source of benzene. In males the risk of leukaemia seems to be higher in the extreme west and in northern subareas, very far from the coke oven and in the hilly part where no other major sources of benzene seem to be present, and might at least partially reflect the distribution of dwellers who are occupationally exposed to benzene or other toxic chemicals.

With regard to NHL, a cluster of risk was observed in the extreme eastern part of the area under study, both for males and for females. Interestingly, the higher FBE-SIR was observed in a census tract, where a foundry, another possible source of industrial pollution, was operational until the early 1980s. The lack of risk for leukaemia in this area might be due either to exposure to pollutants other than benzene, or to cohort effects resulting from different lag times between exposure and disease occurrence.

It seems unlikely that the prevailing winds, reported on the figures, influenced the observed patterns of risk. In fact, excluding northern winds, which are expected to blow the coke oven emissions far from the residential area, easterly winds prevailed; therefore, the central, rather than the western subarea would have been expected to show the highest risks for the considered diseases. However, we evaluated the trend of risk around the coke oven using Stone’s method²³ without including wind direction in the analyses, since the collection of such information in the studied area was begun only in 1994 at the end of the period under examination. Moreover, the presence of the sea coast may create local breezes, thus making it hard to estimate the influence of the wind on the distribution of pollutants.³⁶ Furthermore, the difficulty in obtaining reliable measures of exposure is a problem commonly encountered in ecological investigations.³⁷ For example, in a study cited above,¹⁹ carried out in an area of South Wales close to a petrochemical facility and with elevated leukaemia rates, monitoring data found no increased levels of benzene near the plant.

Finally, inspection of the geographic risk distribution, as shown in figs 1, 2, and 3, did not reveal any particular pattern close to the main Cornigliano thoroughfare, thus excluding a major confounding effect of motor vehicular exhaust.

CONCLUSIONS

The main finding of the present study is the detection of statistically significant higher risks for leukaemia and NHL for males living in an industrialised area. However, the lack of higher risks in females and the pattern of geographic distribution of NHL and leukaemia in both sexes do not suggest any association between the location of the coke oven and the residence of the cases.

Ecological studies, as is this investigation, aim to identify areas at higher risk for selected disease. Because they make use of routinely collected data, they are not costly and may be carried out over a short time. Moreover, ecological comparisons have been recognised as the first natural approach to study the long term effects of air pollution, due to the nature of the exposure and to measurement related problems.³⁷ Their main limit is the inability to assess personal exposure; as a result, causal inference may not be drawn. In fact, the attribution of the exposure levels in an aggregated way may cause the well known “ecological fallacy”—that is, either a systematic under- or overestimation of the association under study.^22,38 Furthermore, risks from environmental concentrations of benzene are likely to barely exceed the risk of the general population, and thus the low statistical power makes it hard to detect differences in risk distribution.⁴

Finally, benzene is a very common air pollutant given off by several sources.^4,34,39 The Rivarolo control area was selected because of the similarity of its sociodemographic characteristics with the Cornigliano area. A different pattern of exposure due to personal habits may thus be considered unlikely. However, a different distribution of occupational exposures, due to a tendency of workers to live near their workplace, cannot be ruled out. In light of such considerations, further investigation, in particular via epidemiological-aetiological studies, are needed in order to trace the exposures that may have contributed to the observed risks and to their geographic distribution.