Abstract

Industrial and urban workers may be exposed to significant levels of air pollutants resulting from the incomplete combustion of organic matter. The authors performed a meta-analysis of 13 DNA-adduct studies (32P-DNA postlabeling technique) on occupational cohorts exposed to air pollution. The association between levels of DNA adducts and air pollution exposure was significant both in heavily exposed industrial workers and in less severely exposed urban workers. Moreover, in an analysis using the seven studies that reported measuring levels of benzo[a]pyrene (B(a)P), a typical marker of exposure, DNA adduct levels in exposed workers (versus those in referents) were significantly correlated with air levels of B(a)P. The relation between DNA adducts and B(a)P was found to be linear at low doses and sublinear at high doses, indicating that DNA adduct formation tends to reach some kind of saturation point at higher levels of exposure to the chemical mixtures present in fumes. When the authors examined the efficiency of DNA adduct production associated with increasing air pollution exposures, the production of DNA adducts per unit of exposure was significantly decreased at higher B(a)P exposure levels. These findings suggest that linear downward extrapolations based on DNA adduct levels associated with B(a)P concentrations of ≥20 ng/m3 might be affected by underestimation bias.

Humans may be heavily exposed to airborne pollutants resulting from industrial processes, residential heating, and motor vehicle exhausts (1). The incomplete combustion of organic matter (i.e., coal, petroleum fuels, and wood materials) results in the release of a complex mixture of airborne pyrolysis products, including polycyclic aromatic hydrocarbons (PAHs). PAHs constitute one of the major classes of airborne carcinogens (2) that may be metabolized into reactive derivatives capable of binding to DNA and of forming DNA adducts (1). When unrepaired, DNA adducts, such as those of benzo[a]pyrene (B(a)P) diol epoxide, can cause mutations (primarily G→T transversions) that may ultimately induce tumor formation (3).

B(a)P measurements are used as guidelines to measure PAH levels in the atmosphere (4) and as a representative exposure marker (5). PAHs are widely present in the ambient air around coke ovens, iron foundries, and aluminum production plants (6) and are present at lower levels in urban air (4). Industrial exposure to coal tar-derived products containing PAHs as major components (7) is associated with increased lung cancer risk (6). Urban exposure to air pollution has been associated with an increased smoke-adjusted risk of developing lung cancer (8). Occupational exposure to fumes generated by motor vehicles, including diesel engine exhausts, may be associated with increased lung cancer risk (8).

Over the past decade, the development of analytical methods (i.e., immunoassays, immunohistochemistry, 32P-post-labeling techniques, gas chromatography-mass spectrometry, and atomic absorbance spectrometry) has made it possible to assess DNA adduct production and use it as a biomarker that reflects carcinogen exposure and reveals individual differences in carcinogen metabolism and DNA repair rates (3, 9, 10). In the attempt to improve cancer risk evaluation, a number of studies have been conducted to measure white blood cell or lymphocyte DNA adducts in PAH-exposed workers using the 32P-DNA postlabeling technique. This assay is effective in measuring the formation of bulky aromatic DNA adducts, such as those induced by complex PAH mixtures (912).

In the present study, we performed a meta-analysis of 13 32P-DNA postlabeling studies on occupational cohorts exposed to air pollution (1325), to examine the difference in white blood cell or lymphocyte DNA adduct levels between industrial workers, urban workers, and referents and the relation between air pollution exposure and DNA adducts.

MATERIALS AND METHODS

Working procedures

Published articles on occupational cohorts exposed to air pollution were collected from recognized information sources, including MEDLINE, TOXLINE, and CANCER-CD. A careful literature search was carried out through January 2000. B(a)P was used as an indicator of air pollution exposure. B(a)P is one of the most studied carcinogenic PAHs, and it is widely used as an indicator compound of PAH mixtures in ambient air of urban and industrial areas and as a suitable marker of relative levels of PAH exposure (2, 4, 6, 7). PAH profiles are generally similar in different workplaces (4); therefore, B(a)P can be considered an acceptable marker of PAH exposure both in iron foundries and coke oven plants and in urban areas (24, 26, 27).

The criteria followed for including a study in this meta-analysis were as follows: 1) only original studies reporting white blood cell or lymphocyte DNA adduct analysis of air pollution-exposed workers using the 32P-DNA postlabeling assay were included; 2) only articles providing data on exposed workers versus referents were included (referents had to have B(a)P exposure levels less than 5 ng/m3); 3) every exposed worker–referent pair of DNA adduct values had to have been considered separately; 4) the mean level of DNA adducts and its standard deviation had to have been reported for each group, or the data needed to compute these statistics had to be present; 5) only studies considering the potentially confounding effects of smoking habits were included; and 6) duplicated studies were not included. Our final database included 13 studies (1325) which supplied 36 sets of exposed–referent pairs with DNA adduct levels (table 1). Twenty-three percent of the studies used to evaluate the qualitative association between air pollution and DNA adducts came from a Swedish laboratory (19, 20, 25), while all of the data used to analyze the dose-response relation were from different countries (1315, 18, 2325).

TABLE 1.

Characteristics of subjects included in a meta-analysis of occupational exposure to air pollution as of January 2000

Study (ref.) and occupational category Smoking habits Mean level or range of benzo(a)pyrene exposure (ng/m3No. of subjects Mean level of DNA adducts (relative adduct labeling × 108Source of DNA 
Popp et al., 1997 (24
 Coke oven workers* Smokers and nonsmokers >2,000–3,550 0.938 (0.610), Lymphocytes 
 Coke oven workers* Smokers and nonsmokers 1,000–2,000 0.629 (0.741) Lymphocytes 
 Coke oven workers* Smokers and nonsmokers 470–<1,000 0.477 (0.614) Lymphocytes 
 Controls* Smokers and nonsmokers  23 0.445 (0.583) Lymphocytes 
Scheel et al., 1995 (21
 Coke oven workers* Smokers  17 0.45 (0.3)§ White blood cells 
 Coke oven workers Nonsmokers  17 0.39 (0.25)§ White blood cells 
 Controls* Smokers  37 0.18 (0.12) White blood cells 
 Controls Nonsmokers  21 0.19 (0.16) White blood cells 
 Coke oven workers* Smokers  14 0.75 (0.65) Lymphocytes 
 Coke oven workers Nonsmokers  0.65 (0.71) Lymphocytes 
 Controls* Smokers  14 0.39 (0.16) Lymphocytes 
 Controls Nonsmokers  0.54 (0.77) Lymphocytes 
Schoket et al., 1991 (16
 Aluminum plant I workers Smokers  14 1.82 (1.10) Lymphocytes 
 Aluminum plant I workers Nonsmokers  11 1.05 (0.48) Lymphocytes 
 Aluminum plant II workers* Smokers  18 3.35 (1.68)§ Lymphocytes 
 Aluminum plant II workers Nonsmokers  1.50 (0.62)§ Lymphocytes 
 Controls Smokers  19 1.45 (0.51) Lymphocytes 
 Controls Nonsmokers  10 1.01 (0.44) Lymphocytes 
Phillips et al., 1988 (13
 Foundry workers* Smokers and nonsmokers >200 2.75 (3.75)§ White blood cells 
 Foundry workers* Smokers and nonsmokers 50–200 1.04 (1.12)§ White blood cells 
 Foundry workers* Smokers and nonsmokers <50 16 0.21 (0.05) White blood cells 
 Controls* Smokers and nonsmokers  0.38 (0.56) White blood cells 
Savela et al., 1989 (14
 Foundry workers* Smokers and nonsmokers <50–>200 53 9.2 (23)# White blood cells 
 Controls* Smokers and nonsmokers  1.7 (0.7) White blood cells 
Reddy et al., 1991 (15
 Foundry workers* Smokers and nonsmokers >200 2.4 (0.2)§ White blood cells 
 Foundry workers* Smokers and nonsmokers 50–200 32 1.7 (0.1)§ White blood cells 
 Foundry workers* Smokers and nonsmokers <50 24 0.7 (0.1)§ White blood cells 
Controls* Smokers and nonsmokers  19 0.2 (0.1) White blood cells 
Perera et al., 1994 (18
 Foundry workers* Smokers and nonsmokers >12–60 17 2.5 (1.2) Lymphocytes 
 Foundry workers* Smokers and nonsmokers 5–12 15 2.1 (1.4) Lymphocytes 
Foundry workers* Smokers and nonsmokers <5 12 2.2 (0.8) Lymphocytes 
 Foundry workers* Smokers and nonsmokers >12–<36 2.3 (2.0) Lymphocytes 
 Foundry workers* Smokers and nonsmokers 5–12 30 1.5 (1.1) Lymphocytes 
 Foundry workers* Smokers and nonsmokers <5 1.3 (0.6) Lymphocytes 
Hemminki et al., 1997 (25
 Foundry workers* Smokers and nonsmokers >5 40 2.0 (1.3)§ White blood cells 
 Foundry workers* Smokers and nonsmokers 0–5 55 1.4 (1.0) White blood cells 
Schoket, 1993 (17
 Surface coating workers* Smokers and nonsmokers  30 12.0 (4.6)§ White blood cells 
 Controls* Smokers and nonsmokers  7.4 (2.7) White blood cells 
Nielsen et al., 1996 (22
 Mechanics and bus garage workers Nonsmokers  10 27.2 (24.6)§ Lymphocytes 
 Controls Nonsmokers  12 3.1 (2.35) Lymphocytes 
Hemminki et al., 1994 (19
 Garage workers Nonsmokers  16 3.63 (1.26)§ Lymphocytes 
 Mechanics Nonsmokers  23 2.82 (0.94) Lymphocytes 
 Others Nonsmokers  2.64 (0.48) Lymphocytes 
 Truck terminal workers Nonsmokers  11 2.65 (0.84) Lymphocytes 
 Diesel forklift drivers Nonsmokers  3.73 (1.81)§ Lymphocytes 
Electric forklift drivers Nonsmokers  2.57 (0.29) Lymphocytes 
 Repair men Nonsmokers  3.00 Lymphocytes 
 Others Nonsmokers  2.14 (0.33) Lymphocytes 
 Controls Nonsmokers  22 2.08 (0.7) Lymphocytes 
Scheel et al., 1995 (20
 Urban bus drivers Nonsmokers  26 0.9 (0.35) Lymphocytes 
 Suburban bus drivers Nonsmokers  23 1.4 (0.48)§ Lymphocytes 
 Taxicab drivers Nonsmokers  19 1.6 (0.91)§ Lymphocytes 
 Controls Nonsmokers  22 1.0 (0.32) Lymphocytes 
Merlo et al., 1997 (23
 Police officers* Smokers and nonsmokers 4.5 (3.4) 88 1.48 (1.35)§ White blood cells 
 Controls* Smokers and nonsmokers 0.15 (0.3) 48 1.01 (0.63) White blood cells 
Study (ref.) and occupational category Smoking habits Mean level or range of benzo(a)pyrene exposure (ng/m3No. of subjects Mean level of DNA adducts (relative adduct labeling × 108Source of DNA 
Popp et al., 1997 (24
 Coke oven workers* Smokers and nonsmokers >2,000–3,550 0.938 (0.610), Lymphocytes 
 Coke oven workers* Smokers and nonsmokers 1,000–2,000 0.629 (0.741) Lymphocytes 
 Coke oven workers* Smokers and nonsmokers 470–<1,000 0.477 (0.614) Lymphocytes 
 Controls* Smokers and nonsmokers  23 0.445 (0.583) Lymphocytes 
Scheel et al., 1995 (21
 Coke oven workers* Smokers  17 0.45 (0.3)§ White blood cells 
 Coke oven workers Nonsmokers  17 0.39 (0.25)§ White blood cells 
 Controls* Smokers  37 0.18 (0.12) White blood cells 
 Controls Nonsmokers  21 0.19 (0.16) White blood cells 
 Coke oven workers* Smokers  14 0.75 (0.65) Lymphocytes 
 Coke oven workers Nonsmokers  0.65 (0.71) Lymphocytes 
 Controls* Smokers  14 0.39 (0.16) Lymphocytes 
 Controls Nonsmokers  0.54 (0.77) Lymphocytes 
Schoket et al., 1991 (16
 Aluminum plant I workers Smokers  14 1.82 (1.10) Lymphocytes 
 Aluminum plant I workers Nonsmokers  11 1.05 (0.48) Lymphocytes 
 Aluminum plant II workers* Smokers  18 3.35 (1.68)§ Lymphocytes 
 Aluminum plant II workers Nonsmokers  1.50 (0.62)§ Lymphocytes 
 Controls Smokers  19 1.45 (0.51) Lymphocytes 
 Controls Nonsmokers  10 1.01 (0.44) Lymphocytes 
Phillips et al., 1988 (13
 Foundry workers* Smokers and nonsmokers >200 2.75 (3.75)§ White blood cells 
 Foundry workers* Smokers and nonsmokers 50–200 1.04 (1.12)§ White blood cells 
 Foundry workers* Smokers and nonsmokers <50 16 0.21 (0.05) White blood cells 
 Controls* Smokers and nonsmokers  0.38 (0.56) White blood cells 
Savela et al., 1989 (14
 Foundry workers* Smokers and nonsmokers <50–>200 53 9.2 (23)# White blood cells 
 Controls* Smokers and nonsmokers  1.7 (0.7) White blood cells 
Reddy et al., 1991 (15
 Foundry workers* Smokers and nonsmokers >200 2.4 (0.2)§ White blood cells 
 Foundry workers* Smokers and nonsmokers 50–200 32 1.7 (0.1)§ White blood cells 
 Foundry workers* Smokers and nonsmokers <50 24 0.7 (0.1)§ White blood cells 
Controls* Smokers and nonsmokers  19 0.2 (0.1) White blood cells 
Perera et al., 1994 (18
 Foundry workers* Smokers and nonsmokers >12–60 17 2.5 (1.2) Lymphocytes 
 Foundry workers* Smokers and nonsmokers 5–12 15 2.1 (1.4) Lymphocytes 
Foundry workers* Smokers and nonsmokers <5 12 2.2 (0.8) Lymphocytes 
 Foundry workers* Smokers and nonsmokers >12–<36 2.3 (2.0) Lymphocytes 
 Foundry workers* Smokers and nonsmokers 5–12 30 1.5 (1.1) Lymphocytes 
 Foundry workers* Smokers and nonsmokers <5 1.3 (0.6) Lymphocytes 
Hemminki et al., 1997 (25
 Foundry workers* Smokers and nonsmokers >5 40 2.0 (1.3)§ White blood cells 
 Foundry workers* Smokers and nonsmokers 0–5 55 1.4 (1.0) White blood cells 
Schoket, 1993 (17
 Surface coating workers* Smokers and nonsmokers  30 12.0 (4.6)§ White blood cells 
 Controls* Smokers and nonsmokers  7.4 (2.7) White blood cells 
Nielsen et al., 1996 (22
 Mechanics and bus garage workers Nonsmokers  10 27.2 (24.6)§ Lymphocytes 
 Controls Nonsmokers  12 3.1 (2.35) Lymphocytes 
Hemminki et al., 1994 (19
 Garage workers Nonsmokers  16 3.63 (1.26)§ Lymphocytes 
 Mechanics Nonsmokers  23 2.82 (0.94) Lymphocytes 
 Others Nonsmokers  2.64 (0.48) Lymphocytes 
 Truck terminal workers Nonsmokers  11 2.65 (0.84) Lymphocytes 
 Diesel forklift drivers Nonsmokers  3.73 (1.81)§ Lymphocytes 
Electric forklift drivers Nonsmokers  2.57 (0.29) Lymphocytes 
 Repair men Nonsmokers  3.00 Lymphocytes 
 Others Nonsmokers  2.14 (0.33) Lymphocytes 
 Controls Nonsmokers  22 2.08 (0.7) Lymphocytes 
Scheel et al., 1995 (20
 Urban bus drivers Nonsmokers  26 0.9 (0.35) Lymphocytes 
 Suburban bus drivers Nonsmokers  23 1.4 (0.48)§ Lymphocytes 
 Taxicab drivers Nonsmokers  19 1.6 (0.91)§ Lymphocytes 
 Controls Nonsmokers  22 1.0 (0.32) Lymphocytes 
Merlo et al., 1997 (23
 Police officers* Smokers and nonsmokers 4.5 (3.4) 88 1.48 (1.35)§ White blood cells 
 Controls* Smokers and nonsmokers 0.15 (0.3) 48 1.01 (0.63) White blood cells 
*

A statistically significant effect of smoking on DNA adduct levels was not shown.

Numbers in parentheses, standard deviation.

DNA adduct levels of exposed subjects were not statistically different from those of referents.

§

DNA adduct levels of exposed subjects were statistically different from those of referents.

A significant effect of smoking on DNA adducts was reported.

#

No statistical analysis was reported.

Meta-analysis

In our meta-analysis, we compared DNA adduct levels in two groups (exposed persons and unexposed persons) when only this qualitative class definition was available. We decided to use exposed worker:referent ratios rather than absolute DNA adduct levels, since large interlaboratory differences may be present in terms of number of adducts (table 1). However, recent 32P-postlabeling studies have shown that a relatively low level of interlaboratory variability (a coefficient of variation of 29.9 percent) may be achieved using a specific 32P-postlabeling protocol (28). When exposure was also associated with B(a)P levels in air (1315, 18, 2325), we investigated the additional opportunity of evaluating dose-response relations.

To obtain a quantitative parameter of the association between PAH exposure and DNA adduct levels, we computed the frequency ratio (FRi) for the ith study (29) as follows:  

\[\mathrm{FR}_{i}\ {=}\ \frac{\mathrm{mean\ DNA\ adduct\ levels\ of\ PAH-exposed\ workers}}{\mathrm{mean\ DNA\ adduct\ levels\ of\ referents.}}\]
Use of a ratio like FRi rather than direct use of DNA adduct levels enabled us to avoid a significant fraction of interlaboratory variability and is sufficiently comparable across the studies considered (29).

To estimate an overall measure of the effect due to PAH exposure, we computed a weighted mean of the FRi's (weighted frequency ratio (wFR)) with weights related to the variance of FRi in each study (30). To stabilize the within-study variances, we performed the calculation transforming the FRi into its natural logarithm (ln(FRi)); ln(FRi) = 0 means that DNA adduct levels are the same for exposed workers and referents.

The variance of ln(FRi) is, approximately,  

\[\mathrm{Var}\left(ln\left(\mathrm{FR}_{i}\right)\right)\ {=}\ \left[\mathrm{Var}\left(me_{i}\right){/}me_{i}^{2}\right]\ {+}\ \left[\mathrm{Var}\left(mc_{i}\right){/}mc_{i}^{2}\right],\]
where mei and mci are the mean DNA adduct frequencies of the exposed and control groups, respectively, and Var(mei) and Var(mci) are the corresponding variances.

The analysis was carried out using the random effects statistical model (31, 32). The random effects model is a special form of the general linear regression model, which takes into account the variability of the exposure effect between studies (Δ2). The model is  

\[\mathrm{ln}\left(\mathrm{FR}_{i}\right)\ {=}\ {\mu}\ {+}\ {\alpha}X_{i}\ {+}\ {\beta}_{i}\ {+}\ {\tau}_{i},\]
where the parameter μ is ln(wFR); α is a vector of unknown parameters, to be estimated from the data, representing the effects of the Xi variates; βi represents the different assessments of the exposure effect from each study with variance Δ2; and τi is the error in estimating ln(FRi) in repeated samples from the same population to which the ith study belongs. The variance of τi is assumed to be known and estimated by Var(ln(FRi)).

We estimated the model parameters using the META procedure in the statistical package Epilog Plus (33). Approximate 95 percent confidence intervals for the wFR's were computed using the formula  

\[\mathrm{exp}\left[\mathrm{ln}\left(\mathrm{wFR}\right)\ {\pm}\ 1.96\ {\times}\ \sqrt{\mathrm{Var}\left(\mathrm{ln}\left(\mathrm{wFR}\right)\right)}\right].\]

RESULTS

Figure 1 shows the ln(FRi)'s and their 95 percent confidence intervals, based on the t distribution for each study. Nineteen out of 36 confidence intervals of ln(FRi)'s did not include zero, showing a significant excess of DNA adducts in exposed subjects with respect to referents. Ln(FRi) values tended to be higher in industrial workers than in urban workers (figure 1); for instance, in the third of data pairs (exposed–referents) including the highest exposures, industrial workers were overrepresented (eight pairs expected, 11 pairs found), and conversely urban workers were underrepresented (four pairs expected, one pair found). However, the Mann-Whitney test showed that the rank difference between the two groups of workers was only a trend and that it did not reach solid statistical significance (p = 0.28). Figure 1 also shows the ln(wFR) value for smokers of 0.29 (standard error 0.13), which was statistically significant. We obtained this value by applying the random effects model to nine studies meeting our inclusion criteria that analyzed white blood cell or lymphocyte DNA adduct levels in healthy non-occupationally exposed cigarette smokers (16, 21, 3440).

FIGURE 1.

Ln frequency ratios (•) according to a t distribution for 13 studies included in a meta-analysis of occupational exposure to air pollution (1325) and a weighted ln frequency ratio for smokers (○) from an analysis of nine studies (16, 21, 3440). The x axis shows the reference number (in parentheses) for each study; the @ symbol indicates a study on urban workers. FRi, frequency ratio for the ith study. Bars, 95% confidence interval based on the t distribution.

FIGURE 1.

Ln frequency ratios (•) according to a t distribution for 13 studies included in a meta-analysis of occupational exposure to air pollution (1325) and a weighted ln frequency ratio for smokers (○) from an analysis of nine studies (16, 21, 3440). The x axis shows the reference number (in parentheses) for each study; the @ symbol indicates a study on urban workers. FRi, frequency ratio for the ith study. Bars, 95% confidence interval based on the t distribution.

Figure 1 also shows that results in which DNA adduct levels of exposed groups were similar to those of corresponding referents have frequently been published (41). We investigated whether papers with results located at the left end of the graph in figure 1 had been published in journals that differed significantly in terms of impact factor (42) from the journals that published results at the right end of the graph. The impact factors were similar in the two cases, which suggests that the type of results obtained did not directly affect their publishability. Moreover, in figure 2, we plotted ln(FRi)'s versus their standard errors to evaluate the spread of every set of data. Results with a ln(FRi) value around zero showed a spread similar to results with higher ln(FRi)'s. The distribution of standard errors was skewed to the right but was reasonably narrow, with three fourths of standard errors falling between zero and 0.3 units (see inset of figure 2).

FIGURE 2.

Plot (•) of ln frequency ratios as a function of their standard errors in a meta-analysis of occupational exposure to air pollution (1325). The inset shows that the distribution of standard errors is skewed to the right but reasonably narrow, with three fourths of the standard errors falling between zero and 0.3 units. FRi, frequency ratio for the ith study.

FIGURE 2.

Plot (•) of ln frequency ratios as a function of their standard errors in a meta-analysis of occupational exposure to air pollution (1325). The inset shows that the distribution of standard errors is skewed to the right but reasonably narrow, with three fourths of the standard errors falling between zero and 0.3 units. FRi, frequency ratio for the ith study.

Table 2 reports the estimates of the weighted means of the FRi's (wFR's) and their confidence intervals resulting from the random effects model after exponential transformation. The wFR's were statistically significant in both industrial workers (p < 0.0001) and urban workers (p < 0.0204) and were equal to 2.012 and 1.512, respectively. This means that the frequencies of DNA adducts in industrial workers and urban workers are approximately 100 percent and 50 percent higher than those in corresponding referents, respectively. Moreover, we compared the wFR of industrial workers with that of urban workers and found that the difference between the two wFR values was only a trend and was not statistically significant (p ≅ 0.22, table 2).

TABLE 2.

Estimated weighted frequency ratios for DNA adducts resulting from occupational exposure to benzo(a)pyrene, according to the random effects model, in a meta-analysis

Random effects model Weighted frequency ratio 95% confidence interval Z statistic (p value) 
Industrial workers 2.012 1.533, 2.641 <0.0001* 
Urban workers 1.512 1.066, 2.144 <0.0204* 
Industrial workers vs. urban workers   0.2232 
Random effects model Weighted frequency ratio 95% confidence interval Z statistic (p value) 
Industrial workers 2.012 1.533, 2.641 <0.0001* 
Urban workers 1.512 1.066, 2.144 <0.0204* 
Industrial workers vs. urban workers   0.2232 
*

Against their own referents.

In order to analyze the dose-response relation between B(a)P exposure and ln(FRi)'s, we applied the random effects model to the seven studies reporting measured levels of B(a)P exposure (1315, 18, 2325). We used as exposure values the average point of the B(a)P ranges reported by each study (table 1), as reported by Greenland (32). The highest B(a)P values were presumed to be 3,000 ng/m3 in the studies by Phillips et al. (13), Savela et al. (14), and Reddy et al. (15), based on the industrial B(a)P measurements observed in Finnish foundries in 1979–1980 (26), and 60 ng/m3 in the study by Hemminki et al. (25), considering B(a)P levels in 1990 (18).

A slope significantly different from zero between the ln(FRi) and ln(B(a)P) values was found (p ≅ 0.026), indicating that DNA adduct levels of exposed workers, with respect to referents, were associated with B(a)P concentrations. The relation we found is illustrated by equation 5, and a graphic representation is given in figure 3.  

\[\mathrm{ln}\left(\mathrm{FR}_{i}\right)\ {=}\ {-}0.178\ {+}\ 0.215\ {\times}\ \mathrm{ln}\left[\mathrm{B}\left(a\right)\mathrm{P}_{i}\right].\]
If the FRi values were directly proportional to those of B(a)P, the slope would be 1. Since the estimate of the slope was 0.215 and its 95 percent confidence interval lay between 0.043 and 0.387, without including 1, the relation of the dose-response curve was seemingly not linear. After exponential transformation, equation 5 becomes  
\[\mathrm{FR}_{i}\ {=}\ \mathrm{exp}\left({-}0.178\right)\ {\times}\ \mathrm{B}\left(a\right)\mathrm{P}_{i}^{0.215}\]
(see figure 4). We repeated the same calculations that brought us to equation 6 and figure 4, using not only average B(a)P data but also the lower term of the exposure range, and obtained qualitatively similar results (data not reported).

FIGURE 3.

Relation (•) between ln frequency ratios and ln benzo[a]pyrene (ln[B(a)P]) concentrations in a meta-analysis of occupational exposure to air pollution (1325). The solid line shows predicted values of ln(FRi) based on the random effects model (equation 5), using the seven studies that reported external B(a)P levels (1315, 18, 2325). FRi, frequency ratio for the ith study. Dotted lines, 95% confidence interval.

FIGURE 3.

Relation (•) between ln frequency ratios and ln benzo[a]pyrene (ln[B(a)P]) concentrations in a meta-analysis of occupational exposure to air pollution (1325). The solid line shows predicted values of ln(FRi) based on the random effects model (equation 5), using the seven studies that reported external B(a)P levels (1315, 18, 2325). FRi, frequency ratio for the ith study. Dotted lines, 95% confidence interval.

FIGURE 4.

Dose-response relation between frequency ratios and external benzo[a]pyrene (B(a)P) concentrations in work environments in a meta-analysis of occupational exposure to air pollution (1325), as predicted from equation 6. The inset shows an extrapolated dose-response curve at low exposure doses, assuming a linear dose-response relation, for B(a)P levels between 0 and 4.5 ng/m3, the lowest value in the database. FRi, frequency ratio for the ith study.

FIGURE 4.

Dose-response relation between frequency ratios and external benzo[a]pyrene (B(a)P) concentrations in work environments in a meta-analysis of occupational exposure to air pollution (1325), as predicted from equation 6. The inset shows an extrapolated dose-response curve at low exposure doses, assuming a linear dose-response relation, for B(a)P levels between 0 and 4.5 ng/m3, the lowest value in the database. FRi, frequency ratio for the ith study.

The inset of figure 4 shows the relation between DNA adducts and B(a)P levels in the range of B(a)P values between 0 and 4.5 ng/m3, the lowest value in our database, assuming that the dose-response curve was linear at low exposure doses (43). The straight dashed line in the inset of figure 4, whose equation is  

\[\mathrm{FR}_{i}\ {=}\ 1\ {+}\ 0.039\ {\times}\ \mathrm{B}\left(a\right)\mathrm{P}_{i},\]
is the tangent at curve 6 (solid line in inset). We obtained it by setting its intercept to 1 and by considering the fact that the first derivative of a curve in a given point is the slope of the tangent to the curve in that point, and that a curve and its tangent have one common point. The coordinates of the intercept of the straight line were set to y = 1, x = 0 (figure 4), since the average values of DNA adducts for exposed workers and referents were expected to be the same in the absence of B(a)P exposure, and consequently FRi = 1 when B(a)P = 0. This appears to be a realistic extrapolation of the relation between DNA adducts and B(a)P at low doses. The value of FRi would tend to be zero at a B(a)P level of 0 ng/m3 without the assumption of dose-response linearity at low doses. We also examined the relative efficiency of DNA adduct formation for increasing air pollution exposures. The values of DNA adducts per unit of exposure decreased in a statistically significant manner at higher B(a)P levels (p = 0.009), indicating less efficient production of DNA adducts per unit of exposure in heavily exposed workers (figure 5).

FIGURE 5.

Levels of DNA adducts (relative adduct labeling × 10−8) per unit of exposure versus exposure expressed as ln benzo[a]pyrene (ln[B(a)P]) levels in a meta-analysis of occupational exposure to air pollution (1325). RAL, relative adduct labeling. The subscript i represents the ith study.

FIGURE 5.

Levels of DNA adducts (relative adduct labeling × 10−8) per unit of exposure versus exposure expressed as ln benzo[a]pyrene (ln[B(a)P]) levels in a meta-analysis of occupational exposure to air pollution (1325). RAL, relative adduct labeling. The subscript i represents the ith study.

DISCUSSION

In this meta-analysis, levels of DNA adducts in industrial workers were significantly higher than those in corresponding referents (p < 0.0001). Similarly, urban workers had significantly higher DNA adduct levels than did controls (p < 0.02, table 2). Different average half-lives have been reported for different types of white blood cells (8 hours for monocytes, 7–24 hours for granulocytes, and months for lymphocytes (44)). We analyzed whether a significant increase in DNA adduct frequency was present in exposed workers with respect to referents in studies that used whole white blood cell counts. Even if lymphocytes represent only about one fourth of the white blood cell population, because of their much longer life, their DNA adduct contribution is probably the most important. Our database included six studies of industrial workers in which white blood cell counts had been used (1315, 17, 21, 25). The association between DNA adducts and industrial air pollution exposure was statistically significant in the studies that used white blood cells (p < 0.05).

Air pollution exposure resulting from the incomplete combustion of organic matter is suspected of contributing to cancer risk (6, 8). Our results support the presence of harmful effects, in terms of DNA adducts, as a consequence of air pollution exposure, not only in heavily exposed industrial workers but also in urban workers.

All of the studies included in our work used the nuclease P1 method for analysis (table 1). This method is mainly effective at detecting bulky hydrophobic DNA adducts, such as those formed by PAHs (45). Occupational and urban air pollution exposures have also been analyzed using the butanol extraction procedure (46, 47), a method adopted primarily to detect aromatic amines bound to the C-8 position of guanine and nitro-PAH DNA adducts (45). However, neither of those two studies (46, 47) satisfied the necessary criteria for inclusion in our analysis.

The potentially confounding effect of smoking habits on DNA adduct levels was considered in all of the studies included in this meta-analysis (table 1). Conflicting results (with white blood cells or lymphocytes used as a source of DNA) have been reported for smokers compared with nonsmokers (48). However, a synergistic effect between smoking habits and other inhalation exposures and a significant correlation between DNA adducts in lymphocytes and bronchi have been reported (49, 50). Our meta-analysis on smoking habits showed an increased frequency of DNA adducts in smokers with respect to referents. Notice that some heavy occupational exposures to air pollution induce DNA adduct levels greater than the ones observed in average smokers (figure 1). This agrees with previous observations (51).

Admittedly, we realize that our study could ideally have been conducted at the level of bronchi, which are more exposed to airborne carcinogens and more competent in terms of metabolic activation (52). Bronchial biopsies are not practical for routine analysis, however, and to our knowledge no occupational studies have been performed using bronchial mucosa. Conversely, peripheral blood cells are an accessible source of DNA (1012), perhaps linked to cancer risk (39, 53), that have been reported to be an acceptable surrogate target for lung tissue (49).

The difference between the wFR for industrial-professional exposures and the wFR for urban-environmental exposures did not reach the level of statistical significance (table 2). This could reflect the fact that some of the studies included in our meta-analysis were also performed in industries where external PAH levels were strictly regulated. The comparison of urban workers with industrial workers might oversimplify B(a)P exposure classifications, since it does not take into account the exposure of some industrial and urban workers to similar air pollution levels.

A large interindividual variation in DNA adducts for apparently similar occupational exposure levels has been reported in some of the studies included in our meta-analysis—for instance, a 20- to 30-fold range in DNA adducts in aluminum plant workers (16), a 15-fold range in iron foundry workers (18), and a 10-fold range in garage workers, mechanics, and controls (22)—a variation that is at least partly attributable to the genetic polymorphism of metabolizing enzymes and of DNA repair enzymes (3). In the case of occupational exposures, misclassification is also possible. The average B(a)P levels for a given broad classification of occupational activity can be accompanied by important variations in real exposure. These finer details are usually not available. We recognize the fact that, at the individual level, concomitant “nature” and “nurture” individual susceptibilities can cause very important deviations of an individual risk from average risk. However, average risk means number of expected cancers in a given population, and this information, at the level of public health concerns, remains of the utmost importance.

We further analyzed the relation between B(a)P exposure and DNA adducts using the seven studies that reported measuring B(a)P concentrations (1315, 18, 2325). Our results showed that levels of DNA adducts in exposed workers, with respect to referents, were associated with B(a)P levels (p ≅ 0.026, figure 3), and that the relation of the dose-response curve was sublinear in heavily exposed industrial workers (figure 4). Moreover, saturation of DNA adducts was observed at higher B(a)P exposure levels (p ≅ 0.009, figure 5).

We noticed, as figure 4 also shows, that equation 6 gives origin to a curve that has a form similar to the well known Michaelis-Menten equation (54):  

\[V\ {=}\ V_{\mathrm{max}}\ {\times}\ \left[S\right]{/}\left(\left[S\right]\ {+}\ K_{\mathrm{m}}\right),\]
where V is the catalytic velocity of a given enzyme, [S] is the substrate concentration, and Km is the Michaelis-Menten constant. The large experimental variability of our data is probably compatible not only with equation 6 but also with some appropriate combination of equations 6 and 7 (figure 4). It is reasonable to think that some rate-limiting enzyme involved in the activation of PAH-type procarcinogens and perhaps other chemical families of procarcinogens will reach saturation at increasing concentrations of our xenobiotics. An equation of the type of equation 8 is therefore biologically plausible and to some degree expected.

Figure 4 indicates that the association between B(a)P exposure and DNA adducts in peripheral blood cells is approximately linear at low doses and sublinear at high doses. Since the levels of cytochrome P450 enzymes necessary to metabolize the carcinogens into reactive metabolites capable of forming DNA adducts are higher in bronchial cells than in lymphocytes (52, 55), higher concentrations of carcinogens are probably necessary to saturate DNA adducts in bronchi with respect to peripheral blood cells (a consideration that implies a higher risk). Increased DNA adducts in bronchi also depend on the fact that the local exposure to air pollution is higher in bronchial mucosa, with respect to peripheral blood cells. These two considerations are partially supported by the observation that DNA adducts are 2–3 times higher in bronchial epithelium than in lymphocytes of smokers (37, 49). When looking at figure 4, for bronchial mucosa we would hypothesize a curve that in principle would be similar to the one observed for blood cells but reaching saturation at higher FRi's. We do not have parallel DNA adduct/external exposure data for bronchial mucosa, but the saturation levels reached in the bronchial mucosa could indeed be several times higher than those in blood cells. It has also been reported that the enzymatic patterns of Phase I enzymes can indeed be qualitatively different in the two targets (52, 55).

Our findings are consistent with two previous studies (not included in our meta-analysis because they did not provide data on exposed workers versus referents) reporting DNA adduct saturation in the lymphocytes of 105 smoker and nonsmoker aluminum workers exposed to average B(a)P levels of 20–1,500 ng/m3 (56) and in the white blood cells of 71 smoker and nonsmoker coke oven workers exposed to average PAH levels of 1,200–17,000 ng/m3 (46). Lutz and Gaylor (43) have reported that deviations from linearity of dose-response relations for DNA adducts are expected when the mechanisms of DNA repair, activation, and detoxification of carcinogens are saturated. In chronic exposures to environmental carcinogens, it is reasonable that DNA adducts will reach a steady state when the rate of DNA adduct formation is balanced by the rate of DNA adduct removal (cell replication and DNA repair).

We have considered the possibility that a flattening of the relation between B(a)P exposures and DNA adducts, for high B(a)P concentrations, may depend on technical aspects of the 32P-postlabeling methodology. Spot intensity could be sublinear with respect to the real adduct frequency. However, a linear dose-response relation between increased doses of the reactive B(a)P diol epoxide and DNA adducts has been reported for human fibroblasts (57). Treatments with B(a)P diol epoxide at concentrations up to 10 mm for 30 minutes at 37°C have yielded linear relative adduct labeling values up to approximately three DNA adducts per 105 normal nucleotides.

Exposure misclassification has been reported to influence the exposure-response relation (58, 59). The attenuation of the linearity of DNA adduct formation at high exposure levels might be due to the fact that a consistent portion of subjects were incorrectly assigned to higher-than-real exposure levels, especially for the highest categories of exposure. To avoid the possible influence of exposure misclassification, we analyzed the B(a)P exposure-response relation using as external exposure values not only the average point of the B(a)P ranges reported by each study but also the lowest value of the B(a)P ranges (table 1); similar sublinear DNA adduct kinetics were obtained (data not reported). Epidemiologic studies have also shown that sublinearity of exposure-response relations might be explained by the tendency of highly exposed workers with ill health to leave the workplace (i.e., health-related job selection) (6, 60). A possible consequence of the depletion of the pool of more susceptible individuals could be that increased exposures will have less-than-proportional effects on DNA adduct formation for the remaining less susceptible highly exposed workers. However, the similitude of the frequency distributions of two genotypes involved in PAH metabolism—cytochrome P450 1A1 (CYP1A1) and glutathione transferase M1 (GSTM1) in Finnish iron foundry workers (25) and Polish urban and rural populations (61), respectively—does not support the above hypothesis, at least for CYP1A1 and GSTM1 genotypes.

Several parameters, such as age, gender, consumption of charcoal-broiled food, consumption of fruits and vegetables, and seasonal variations, have been reported to influence levels of DNA adducts (6264). Even an earlier age of onset of smoking has recently been shown to increase lung DNA adduct persistence in former smokers (long-lasting perturbations of mucosa cells (49)). Dietary PAHs from charcoal-broiled food may induce white blood cell or lymphocyte DNA adducts (63), while a high intake of fruits and vegetables may have a protective effect against white blood cell DNA adducts (64). Seasonal variations affecting white blood cell or lymphocyte DNA adducts have mainly been associated with increased levels of air pollution for residential heating in winter, with a seasonal variation of aryl hydrocarbon hydroxylase inducibility, or with more intense photochemical modifications of PAHs in the atmosphere during summer (63). Only a few of these confounding factors, such as age, gender, and consumption of charcoal-broiled food, have been evaluated in the studies included in our meta-analysis; no effects or borderline effects have been reported (1325). The modulating effects of genetic polymorphism of enzymes involved in carcinogen metabolism (e.g., GSTM1 and CYP1A1 genotypes) on DNA adduct levels have been analyzed only in a group of Finnish iron foundry workers (25); only borderline effects on DNA adducts have been observed in individuals with CYP1A1 MspI genotypes or in subjects lacking the GSTM1 gene. It is probably impossible to completely exclude residual confounding effects dependent on unreported factors.

Occupational exposure to carcinogens is of great public health concern. Increased relative risks for lung cancer have been reported in several occupational categories involving exposure to air pollution (table 3 (6, 6570)). The relative risks reported in table 3 for professional workers seem to be in acceptable agreement with the wFRs for DNA adducts resulting from our meta-analysis. Coke oven and foundry workers, who show a higher lung cancer risk, belong to the category of industrial workers presenting the highest levels of DNA adducts. Comparing DNA adduct data in figure 1 and cancer risk data in table 3, cigarette smokers seem to have a much higher cancer risk associated with relatively low levels of DNA adducts in peripheral blood cells, in comparison with professional exposures. The major reasons for this apparent discrepancy are probably the following: 1) length of exposure tends to be longer and is usually accompanied by an earlier onset in smokers, with respect to professional workers (67, 68); 2) the composition of fumes related to professional exposures is different from that of cigarette smoke (6, 71); and 3) levels of some PAHs, such as B(a)P, benz[a]anthracene, benzo[b]fluoranthene, benzo[j]fluoranthene, and benzo[k]fluoranthene, may be hundreds of times higher in industrial atmospheres than in cigarette smoke, at least in some coke oven plants (6, 71). The type of 32P-postlabeling method considered in our meta-analysis is known to be especially suitable for detecting bulky, hydrophobic adducts like PAH-generated adducts (45). Since the relative risks vary within a relatively small range in various exposed workers, our observations are relevant in terms of public health concerns for different cohorts of people (expected number of cancers in a population or a group). At the individual level, the combined effect of multiple individual susceptibilities/predispositions could easily be more important (larger) than average relative risks.

TABLE 3.

Range of relative risk estimations for lung cancer among industrial workers, vehicle drivers, and cigarette smokers in various studies

Category Relative risk Source(s) 
Aluminum plant workers 1.0–3.7 International Agency for Research on Cancer (6) and Samet (66
Iron and steel foundry workers 1.2–7.1 International Agency for Research on Cancer (6), Samet (66), and Moulin et al. (67
Coke oven workers 1.6–10 International Agency for Research on Cancer (6), Costantino et al. (68), and Schottenfeld and Fraumeni (69
Professional drivers 1.5–2.6 Jakobsson et al. (70) and Soll-Johanning et al. (71
Cigarette smokers* 3.7–25.4 Schottenfeld and Fraumeni (69
Category Relative risk Source(s) 
Aluminum plant workers 1.0–3.7 International Agency for Research on Cancer (6) and Samet (66
Iron and steel foundry workers 1.2–7.1 International Agency for Research on Cancer (6), Samet (66), and Moulin et al. (67
Coke oven workers 1.6–10 International Agency for Research on Cancer (6), Costantino et al. (68), and Schottenfeld and Fraumeni (69
Professional drivers 1.5–2.6 Jakobsson et al. (70) and Soll-Johanning et al. (71
Cigarette smokers* 3.7–25.4 Schottenfeld and Fraumeni (69
*

For an interval of 1–>40 cigarettes/day.

Epidemiologic studies have been used, mainly by the US Environmental Protection Agency, to calculate a lung cancer unit risk factor of 1 × 10−1 (10 percent of people) following a lifetime exposure to B(a)P of 1,000 ng/m3 (8). This unit risk factor was used by Hemminki et al. (8) to estimate the annual number of cancers (mainly lung cancers) attributable to urban air pollution, assuming a linear dose-response relation between higher and lower B(a)P exposure levels. A lifelong exposure to B(a)P at a concentration of 0.7 ng/m3 was estimated by Hemminki et al. to cause 1.0 case of lung cancer annually per 106 inhabitants. From figure 4, we can deduce that B(a)P levels of 1,000 ng/m3 and 0.7 ng/m3 correspond to frequency ratios of approximately 3.69 (2.69 over background) and 1.03 (0.03 over background), respectively, assuming that the relation between DNA adducts and B(a)P is linear at low doses, when the enzymes involved in the activation and detoxification of carcinogens are not saturated (43). It has been reported that at low exposure levels an assumption of linearity of the relation between cancer risk and exposure to a carcinogenic agent (or DNA adducts in our case) may represent a reasonable choice only according to some multihit and multistage models but not others (72). As a first approximation, multihit models that at low levels of carcinogen exposure depict cancer risks that are the product of multiple factorials of the type  

\[\left(a_{1}\ {+}\ {\varepsilon}_{1}\right)\left(a_{2}\ {+}\ {\varepsilon}_{2}\right)\left(a_{3}\ {+}\ {\varepsilon}_{3}\right)\ {\ldots}\ \left(a_{n}\ {+}\ {\varepsilon}_{n}\right).\]
where εiai is the increment over background of individual mutation risks for a discrete number of altered oncogenes (typically 4–7 (73)), tend to predict a quasilinear relation between mutation risks (which are directly dependent on DNA adduct frequency) and cancer risks. Our dose-response curve for DNA adducts has been estimated in blood cells (figure 4) rather than in bronchial epithelium. For more precise assessment of the relation between DNA adducts and lung cancer risk, a similar dose-response relation related to the target tissue would be of the utmost importance.

The two graphs presented in figure 4 tend to suggest that adduct levels associated with B(a)P levels of 20 ng/m3 and above can generate underestimation of the real levels of adducts formed at lower B(a)P concentrations. There is probably no need to refer to very high B(a)P levels (e.g., ≥1,000 ng/m3) in order to be confronted with this downward underestimation problem.

Molecular epidemiologic studies may improve our knowledge of human cancer risk in addition to studies of human carcinogen exposure (10), thus reducing the uncertainties associated with extrapolating from high environmental doses to low environmental doses. Usually, classical epidemiologic studies require relatively high levels of exposure and relatively large samples in order to observe lung cancer relative risks greater than 1 at a statistically significant level. The use of molecular epidemiologic parameters, such as DNA adducts in our case, may allow for some kind of comparison between an epidemiologically significant situation (for relatively heavy exposures and sufficiently large samples) and an epidemiologically silent situation (for lighter exposures and/or smaller samples).

In summary, air pollution exposure may induce measurable levels of DNA adducts not only in heavily exposed industrial workers but also in urban workers experiencing milder exposures. The relation between B(a)P exposure and DNA adducts is linear at low doses and sublinear at high doses in peripheral blood cells.

Reprint requests to Prof. Silvio Parodi, Unit of Experimental Oncology, National Cancer Institute, Largo Benzi 10, 16132 Genoa, Italy (e-mail: parodis@hp380.ist.unige.it).

This work was partially supported by grants from the Italian Association for Research on Cancer and from the European Community (contract QLRT-1999-00927) to M. Peluso, and by grants to S. Parodi from the Ministry of Health and the Ministry of University and Scientific and Technological Research.

The authors thank Drs. Paolo Boffetta, Federico Valerio, and Paolo Vineis for their comments and Donatella Camposeragno for data management. The authors are grateful to Drs. Ulderica Parodi and Massimo Caccia for their mathematical help in the elaboration of equation 7 and the inset of figure 4.

REFERENCES

1.
Pershagen G. Air pollution and cancer. In: Vainio H, Sorsa M, McMichael AJ, eds. Complex mixtures and cancer risk. (IARC Scientific Publication no. 104). Lyon, France: International Agency for Research on Cancer, 1990:240–51.
2.
International Agency for Research on Cancer. Polynuclear aromatic compounds. Part I. Chemical, environmental, and experimental data. (IARC Monographs on the Evaluation of Carcinogenic Risks to Humans, vol 32). Lyon, France: International Agency for Research on Cancer, 1984.
3.
Perera FP. Molecular epidemiology: insights into cancer susceptibility, risk assessment, and prevention.
J Natl Cancer Inst
 
1996
;
88
:
496
–509.
4.
World Health Organization. Air quality guidelines for Europe. (World Health Organization regional publications, European series, no. 23). Copenhagen, Denmark: World Health Organi-zation, 1987.
5.
Perera FP, Tang DL, O'Neill JP, et al. HPRT and glycophorin A mutations in foundry workers: relationship to PAH exposure and to PAH-DNA adducts.
Carcinogenesis
 
1993
;
14
:
969
–73.
6.
International Agency for Research on Cancer. Polynuclear aromatic compounds. Part III. Industrial exposure in aluminium production, coal gasification, coke production, and iron and steel founding. (IARC Monographs on the Evaluation of Carcinogenic Risks to Humans, vol 34). Lyon, France: International Agency for Research on Cancer, 1984.
7.
International Agency for Research on Cancer. Polynuclear aromatic compounds. Part IV. Bitumens, coal-tars and derived products, shale-oils and soots. (IARC Monographs on the Evaluation of Carcinogenic Risks to Humans, vol 35). Lyon, France: International Agency for Research on Cancer, 1985.
8.
Hemminki K, Pershagen G. Cancer risk of air pollution: epidemiological evidence.
Environ Health Perspect
 
1994
;
102
:
187
–92.
9.
Poirier MC. DNA adducts as exposure biomarkers and indicators of cancer risk.
Environ Health Perspect
 
1997
;
105(suppl 4)
:
907
–12.
10.
Poirier MC, Weston A. Human DNA adduct measurements: state of the art.
Environ Health Perspect
 
1996
;
104
:
883
–94.
11.
Beach A, Gupta RC. Human biomonitoring and the 32P-postlabeling assay.
Carcinogenesis
 
1992
;
13
:
1053
–74.
12.
Taningher M, Malacarne D, Mancuso T, et al. Methods for predicting carcinogenic hazards: new opportunities coming from recent developments in molecular oncology and SAR studies.
Mutat Res
 
1997
;
391
:
3
–32.
13.
Phillips DH, Hemminki K, Alhonen A, et al. Monitoring occupational exposure to carcinogens: detection by 32P-postlabelling of aromatic DNA adducts in white blood cells from iron foundry workers.
Mutat Res
 
1988
;
204
:
531
–41.
14.
Savela K, Hemminki K, Hewer A, et al. Interlaboratory comparison of the 32P-postlabelling assay for aromatic DNA adducts in white blood cells of iron foundry workers.
Mutat Res
 
1989
;
224
:
485
–92.
15.
Reddy MV, Hemminki K, Randerath K. Postlabeling analysis of polycyclic aromatic hydrocarbon-DNA adducts in white blood cells of foundry workers.
J Toxicol Environ Health
 
1991
;
34
:
177
–85.
16.
Schoket B, Phillips DH, Hewer A, et al. 32P-Postlabelling detection of aromatic DNA adducts in peripheral blood lymphocytes from aluminium production plant workers.
Mutat Res
 
1991
;
260
:
89
–98.
17.
Schoket B. Monitoring occupational exposure to carcinogens. In: Phillips DH, Castegnaro M, Bartsch H, eds. Postlabelling methods for detection of DNA adducts. (IARC Scientific Publication no. 124). Lyon, France: International Agency for Research on Cancer, 1993:341–7.
18.
Perera FP, Dickey C, Santella R, et al. Carcinogen-DNA adducts and gene mutation in foundry workers with low-level exposure to polycyclic aromatic hydrocarbons.
Carcinogenesis
 
1994
;
15
:
2905
–10.
19.
Hemminki K, Soderling J, Ericson P, et al. DNA adducts among personnel servicing and loading diesel vehicles.
Carcinogenesis
 
1994
;
15
:
767
–9.
20.
Hemminki K, Zhang LF, Kruger J, et al. Exposure of bus and taxi drivers to urban air pollutants as measured by DNA and protein adducts.
Toxicol Lett
 
1994
;
72
:
171
–4.
21.
Scheel C, Popp W, Kraus C, et al. 32P-Postlabeling analysis of DNA adducts in different populations.
Toxicol Lett
 
1995
;
77
:
299
–307.
22.
Nielsen PS, Andreassen A, Farmer PB, et al. Biomonitoring of diesel exhaust-exposed workers: DNA and hemoglobin adducts and urinary 1-hydroxypyrene as markers of exposure.
Toxicol Lett
 
1996
;
86
:
27
–37.
23.
Merlo F, Bolognesi C, Peluso M, et al. Airborne levels of polycyclic aromatic hydrocarbons: 32P-postlabeling DNA adducts and micronuclei in white blood cells from traffic police workers and urban residents.
J Environ Pathol Toxicol Oncol
 
1997
;
16
:
157
–62.
24.
Popp W, Vahrenholz C, Schell C, et al. DNA single strand breakage, DNA adducts, and sister chromatid exchange in lymphocytes and phenanthrene and pyrene metabolites in urine of coke oven workers.
Occup Environ Med
 
1997
;
54
:
176
–83.
25.
Hemminki K, Dickey C, Karlsson S, et al. Aromatic DNA adducts in foundry workers in relation to exposure, life style and CYP1A1 and glutathione transferase M1 genotype.
Carcinogenesis
 
1997
;
18
:
345
–50.
26.
Perera FP, Hemminki K, Young TL, et al. Detection of polycyclic aromatic hydrocarbon-DNA adducts in white blood cells of foundry workers.
Cancer Res
 
1988
;
48
:
2288
–91.
27.
Menichini E, Monfredini F, Merli F. The temporal variability of the profile of carcinogenic polycyclic aromatic hydrocarbons in urban air: a study in a medium traffic area in Rome, 1993–1998.
Atmospheric Environ
 
1999
;
33
:
3739
–50.
28.
Phillips DH, Castegnaro M. Standardization and validation of DNA adduct postlabelling methods: report of interlaboratory trials and production of recommended protocols.
Mutagenesis
 
1999
;
14
:
301
–15.
29.
Bonassi S, Montanaro F, Ceppi M, et al. Is human exposure to styrene a cause of cytogenetic damage? A re-analysis of the available evidence.
Biomarkers
 
1996
;
1
:
217
–25.
30.
Kleinbaum DG, Kupper LL, Morgenstern H. Epidemiologic research: principles and quantitative methods. New York, NY: Van Nostrand Reinhold Company, 1982.
31.
DerSimonian R, Laird N. Meta-analysis in clinical trials.
Controlled Clin Trials
 
1986
;
7
:
177
–88.
32.
Greenland S. Quantitative methods in the review of epidemiologic literature.
Epidemiol Rev
 
1987
;
9
:
1
–30.
33.
Epicenter Software. Epilog Plus, version 3. Pasadena, CA: Epicenter Software, 1993.
34.
Vulimiri SV, Wu X, Baer-Dubowska W, et al. Analysis of aromatic DNA adducts and 7,8-dihidro-8-oxo-2′-deoxyguanosine in lymphocyte DNA from a case-control study of lung cancer involving minority populations.
Mol Carcinog
 
2000
;
27
:
34
–46.
35.
Savela K, Hemminki K. DNA adducts in lymphocytes and granulocytes of smokers and nonsmokers detected by the 32P-postlabelling assay.
Carcinogenesis
 
1991
;
12
:
503
–8.
36.
Jahnke GD, Thompson CL, Walker MP, et al. Multiple DNA adducts in lymphocytes of smokers and nonsmokers determined by 32P-postlabeling analysis.
Carcinogenesis
 
1990
;
11
:
205
–11.
37.
Phillips DH, Schoket B, Hewer A, et al. Influence of cigarette smoking on the levels of DNA adducts in human bronchial epithelium and white blood cells.
Int J Cancer
 
1990
;
46
:
569
–75.
38.
Gallagher J, Mumford J, Shank T, et al. DNA adduct profiles and levels in placenta, blood and lung in relation to cigarette smoking and smoky coal emissions: postlabelling methods for detection of DNA adducts (IARC Scientific Publication no. 123). Lyon, France: International Agency for Research on Cancer, 1993:283–92.
39.
Peluso M, Airoldi L, Munnia A, et al. WBC-DNA adducts, smoking, diet, and NAT2 and GSTM1 genotypes in bladder cancer: a case-control study.
Cancer Epidemiol Biomarkers Prev
 
1998
;
7
:
341
–6.
40.
Wang Y, Ichiba M, Iyadomi M, et al. Effects of genetic polymorphism of metabolic enzymes, nutrition, and lifestyle factors on DNA adduct formation in lymphocytes.
Ind Health
 
1998
;
36
:
337
–46.
41.
Vandenbroucke JP. Passive smoking and lung cancer: a publication bias?
Br Med J
 
1988
;
296
:
391
–2.
42.
Institute for Scientific Information. Science Citation Index journal citation reports. Philadelphia, PA: Institute for Scientific Information, 1996.
43.
Lutz WK, Gaylor D. Significance of DNA adducts at low dose: shortening the time to spontaneous tumor occurrence.
Regul Toxicol Pharmacol
 
1996
;
23
:
29
–34.
44.
Santella RM, Grinberg-Funes RA, Young TL, et al. Cigarette smoking related polycyclic aromatic hydrocarbon-DNA adducts in peripheral mononuclear cells.
Carcinogenesis
 
1992
;
13
:
2041
–5.
45.
Gupta RC, Earley K. 32P-adduct assay: comparative recoveries of structurally diverse DNA adducts in the various enhancement procedures.
Carcinogenesis
 
1988
;
9
:
1687
–93.
46.
Lewtas J, Walsh D, Williams R, et al. Air pollution exposure-DNA adduct dosimetry in humans and rodents: evidence for non-linearity at high doses.
Mutat Res
 
1997
;
378
:
51
–63.
47.
Binkova B, Lewtas J, Miskova I, et al. DNA adducts and personal air monitoring of carcinogenic polycyclic aromatic hydrocarbons in an environmentally exposed population.
Carcinogenesis
 
1995
;
16
:
1037
–46.
48.
Phillips DH. DNA adducts in human tissues: biomarkers of exposure to carcinogens in tobacco smoke.
Environ Health Perspect
 
1996
;
104
:
453
–8.
49.
Wienke JK, Thurston, SW, Kelsey KT, et al. Early age at smoking initiation and tobacco carcinogen DNA damage in the lung.
J Natl Cancer Inst
 
1999
;
91
:
614
–19.
50.
Rojas M, Alexandrov K, Auburtin G, et al. Anti-benzo[a]pyrene diolepoxide–DNA adduct levels in peripheral mononuclear cells from coke oven workers and the enhancing effect of smoking.
Carcinogenesis
 
1995
;
16
:
1373
–6.
51.
Lewtas J, Mumford J, Everson RB, et al. Comparison of DNA adducts from exposure to complex mixtures in various human tissues and experimental systems.
Environ Health Perspect
 
1993
;
99
:
89
–97.
52.
Lang DS, Becker S, Devlin RB, et al. Cell-specific differences in the susceptibility of potential cellular targets of human origin derived from blood and lung following treatment with 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD).
Cell Biol Toxicol
 
1998
;
14
:
23
–38.
53.
Tang D, Santella RM, Blackwood AM, et al. A molecular epidemiological case-control study of lung cancer.
Cancer Epidemiol Biomarkers Prev
 
1995
;
4
:
341
–6.
54.
Lehninger AL, Nelson DL, Cox MM, eds. Principles of biochemistry. 2nd ed. New York, NY: Worth Publishers, Inc, 1993.
55.
Hukkanen J, Hakkola J, Anttila S, et al. Detection of mRNA encoding xenobiotic-metabolizing cytochrome P450s in human bronchoalveolar macrophages and peripheral blood lymphocytes.
Mol Carcinog
 
1997
;
20
:
224
–30.
56.
van Schooten F-J, Godschalk RW, Breedijk A, et al. 32P-Postlabelling of aromatic DNA adducts in white blood cells and alveolar macrophages of smokers: saturation at high exposures.
Mutat Res
 
1997
;
378
:
65
–75.
57.
Baan RA, van den Berg PT, Steenwinkel M-J, et al. Detection of benzo[a]pyrene-DNA adducts in cultured cells treated with benzo[a]pyrene diol-epoxide by quantitative immunofluorescence microscopy and 32P-postlabelling: immunofluorescence analysis of benzo[a]pyrene-DNA adducts in bronchial cells from smoking individuals. In: Methods for detecting DNA damaging agents in humans: application in cancer epidemiology and prevention. (IARC Scientific Publication no. 89). Lyon, France: International Agency for Research on Cancer, 1988:146–54.
58.
Dosemeci M, Wacholder S, Lubin JH. Does nondifferential misclassification of exposure always bias a true effect toward the null value?
Am J Epidemiol
 
1990
;
132
:
746
–8.
59.
Correa-Villasenor A, Stewart WF, Franco-Marina F, et al. Bias from nondifferential misclassification in case-control studies with three exposure levels.
Epidemiology
 
1995
;
6
:
276
–81.
60.
Petsonk EL, Daniloff EM, Mannino DM, et al. Airway responsiveness and job selection: a study in coal miners and non-mining controls.
Occup Environ Med
 
1995
;
52
:
745
–9.
61.
Butkiewicz D, Grzybowska E, Hemminki K, et al. Modulation of DNA adduct levels in human mononuclear white blood cells and granulocytes by CPY1A1, CP2D6 and GSTM1 genetic polymorphisms.
Mutat Res
 
1998
;
415
:
97
–108.
62.
Perera FP. Environment and cancer: who are susceptible?
Science
 
1997
;
278
:
1068
–73.
63.
Schoket B. DNA damage in humans exposed to environmental and dietary polycyclic aromatic hydrocarbons.
Mutat Res
 
1999
;
424
:
143
–53.
64.
Peluso M, Airoldi L, Magagnotti C, et al. White blood cell DNA adducts and fruit and vegetable consumption in bladder cancer.
Carcinogenesis
 
2000
;
21
:
183
–7.
65.
Coultas DB. Other occupational carcinogens. In: Samet JM, ed. Epidemiology of lung cancer. New York, NY: Marcel Dekker, Inc, 1994:299–333.
66.
Moulin JJ, Lafontaine M, Mantout B, et al. Mortality due to bronchopulmonary cancers in workers of 2 foundries.
Rev Epidemiol Sante Publique
 
1995
;
43
:
107
–21.
67.
Costantino JP, Redmond CK, Bearden A. Occupationally related cancer risk among coke oven workers: 30 years of follow-up.
J Occup Environ Med
 
1995
;
37
:
597
–604.
68.
Blot WJ, Fraumeni JF Jr. Cancers of the lung and pleura. In: Schottenfeld D, Fraumeni JF Jr, eds. Cancer epidemiology and prevention. New York, NY: Oxford University Press, 1996:637–65.
69.
Jakobsson R, Gustavsson P, Lunderberg I. Increased risk of lung cancer among male professional drivers in urban but not rural areas of Sweden.
Occup Environ Med
 
1997
;
54
:
189
–93.
70.
Soll-Johanning H, Bach E, Olsen JH, et al. Cancer incidence in urban bus drivers and tramway employees: a retrospective cohort study.
Occup Environ Med
 
1998
;
55
:
594
–8.
71.
International Agency for Research on Cancer. Tobacco smoking. (IARC Monographs on the Evaluation of Carcinogenic Risks to Humans, vol 38). Lyon, France: International Agency for Research on Cancer, 1985.
72.
Krewski D, Murdoch D, Dewanji A. Statistical modeling and extrapolation of Carcinogenesis data. In: Moolgavkar SH, Prentice RL, eds. Modern statistical methods in chronic disease epidemiology. New York, NY: John Wiley and Sons, Inc, 1986:259–82.
73.
Hahn WC, Counter CM, Lundberg AS, et al. Creation of human tumour cells with defined genetic elements.
Nature
 
1999
;
400
:
464
–8.