METHODS

Subjects

Harbour workers

In a meeting organised by the employers and the investigators, all 399 seamen with potential exposure to river silt aerosols (mean age (SD) 46.7 (10.9) years, mean duration of employment 13.6 (9.2) years) were informed about the aims of the survey and invited to take part in the study. Of 161 (40%) volunteers (mean age (SD) 44.8 (11.0) years, mean duration of employment 13.0 (9.3) years) a sample of 100 was chosen for the study (mean age (SD) 45.3 (10.7) years, mean duration of employment 13.7 (9.1) years). In order to include a sufficient number of subjects with potentially higher exposure levels, all workers on suction dredgers were included (n = 30). The remaining 70 subjects were selected at random.

Controls

One hundred male office workers working in the administration office of the harbour agreed to participate in the study (mean age (SD) 46.3 (10.4) years).

Design

Harbour workers

Up to eight measurements were performed over a three year period (1994–97). Seventy eight subjects took part in all eight measurements. The reasons for drop-out were retirement (n = 5), sick leave on the day of the study (n = 5), and active refusal (n = 12).

At each survey, a medical examination and a standardised interview with special focus on nutrition and smoking habits were performed. Additionally, whole blood samples as well as a urinary samples were taken on-site at the end of the working day. The following parameters were determined:

• Blood:

  • Lead

  • Cadmium

  • Mercury

  • Frequency of micronuclei

  • Sister chromatid exchange rate (SCE rate).

• Urine:

  • Cadmium

  • Chromium

  • Nickel

  • Arsenic

  • Mercury (7th and 8th examination only).

Controls

Office workers were studied once during the last year of the survey in parallel with the seamen. Each day, six harbour workers and six age matched controls were studied alternately to take circadian rhythm as well as seasonal patterns into account (matched examinations). Calliper matching was used for age using a three year interval.

Biomonitoring

Following the Standard Operating Procedures of the German Research Council (Deutsche Forschungsgemeinschaft),⁴ atomic absorption spectrometry was used to determine metals and metalloids in blood and urine (AAS 4000 and Z3030 by Perkin Elmer, SPE 800 by Varian). Lead in blood was measured directly with palladium modifier. To determine cadmium in blood, samples were deproteinised using nitric acids. Chelatisation and extraction was used to condition urinary samples for cadmium and nickel analysis. To determine the toxicologically relevant amount of arsenic in urine (inorganic arsenic species and their mono- and dimethylated metabolites), direct hydride generation was used. Blood and urinary samples were prepared for mercury analysis using cold vapour technique/amalgamation. Precision control was carried out with commercial urine (Lyphocheck, Bio-Rad, Munich, Germany) or spiked animal blood. Relative standard deviations between series ranged from 2.6% to 13.0% (n = 17–24).

Detection limits for lead, cadmium, and mercury in blood were 10.0, 0.4, and 0.6 μg/l, respectively; the lower limit of detection for cadmium, mercury, chromium, nickel, and arsenic in urine was 0.2, 0.6, 0.6, 0.2, and 1.5 μg/l, respectively.

In order to exclude differences in the concentration of urine of harbour workers and controls, creatinine was determined in all urinary samples during the matched examinations using an enzymatic assay (Creatinine PAP, Fa. Boehringer, Mannheim; Photometer PCP 6121, Fa. Eppendorf, Hamburg, Germany).

Cytogenetic markers

A modified cytokinesis block method was used to determine the lymphocyte micronuclei frequency.^5,6 In short, lymphocytes were cultured for 72 hours, and cytochalasin B was added after 44 hours of incubation. After washing, fixing, and harvesting the cells on to slides, they were dried and stained with 4′,6-diamino-2-phenylindole (DAPI, 5 mg/l; Serva, Heidelberg, Germany). Two thousand binucleated cells per sample were scored blindly by the same reader.

Sister chromatid exchange (SCE) rates were assessed as described previously.^7,8 Whole blood cultures were established in chromosome medium containing 5-bromo-2-deoxy-uridine for 72 hours. Colcemid (0.1 μg/ml) was added to each culture two hours before the end of incubation. Lymphocytes were washed, harvested on to slides, and air dried chromosome preparations were made. Slides were stained with Giemsa and scored blindly by the same reader. Fifteen well differentiated diploid second division metaphases with at least 40 chromosomes were scored in each subject.

Statistical analyses

Because of the non-normal distribution of the exposure and effect levels, descriptive statistics are presented as quartiles with ranges. Values below the detection limits were entered as half of the detection limit.

Comparisons between exposed subjects and controls were based on the matched measurements. Bivariate associations between potential risk factors and levels of exposure and cytogenetic markers were analysed using the Mann-Whitney U test. Seasonal trends over the course of the survey were tested using MANOVA. Multivariate analyses were done using multiple linear regression models. In these models, all variables bivariately associated with exposure or effect levels were included. The model assumptions for linear regression models have been tested graphically using residual plots.

For statistical analyses the Statistica (Tulsa, USA) and SPSS (USA) software packages have been used.

RESULTS

Descriptives

Exposed subjects were more likely to smoke than controls (41% v 29% active smokers, respectively). Fifty per cent of both groups reported that they eat fish at least once a week.

The results of the physical examinations among the exposed subjects did not indicate any job specific diseases.

Biomonitoring results

Harbour workers

Table 1 shows the biomonitoring results of the harbour workers over the course of the survey. Eighty two per cent of the urinary chromium concentrations were below detection limit; therefore, these results were not further analysed. Apart from cadmium in blood (26% of the measurements below the detection limit) less than 11% of the biomonitoring results were below the detection limit.

One level of mercury in blood was higher than the current biological exposure indices (BEI, 15 μg/l). All other concentrations of heavy metals in blood and urinary samples were below these limits.

In contrast, 70 urinary samples (9.4%) had arsenic levels above the current BEI (35 μg/l). The maximum level of arsenic in urine (356 μg/l) was found during the first examination. However, no trend in arsenic levels was seen over the course of the survey. The same was true for all exposure parameters under study besides lead concentrations in blood. These levels decreased over the course of the survey (p_Friedman < 0.001). No seasonal trends were seen.

Potential risk factors and levels of exposure

Table 2 shows the results of bivariate analyses. Within the group of harbour workers, only cadmium levels in urine were slightly higher in workers on suction dredgers compared to other harbour workers (p < 0.05). Duration of employment was not associated with exposure levels during the matched surveys.

Comparing harbour workers and controls, harbour workers had significantly higher levels of cadmium in blood and urine. In contrast, lead levels in blood and nickel concentrations in urinary samples were significantly lower among harbour workers. The latter associations have been confirmed in the multivariate models. However, only 8% of the variance could be explained by these models (table 3).

Age was associated with higher biomonitoring results for almost all parameters under study. These associations were significant for mercury concentrations in blood as well as cadmium and arsenic levels in urinary sample and were robust for adjustment in the multivariate models.

Current smoking was a strong predictor for cadmium levels in blood and urine in the bivariate as well as in the multivariate models. In contrast, fish consumption in the five days prior to the examination was strongly associated with mercury levels in blood and arsenic levels in urine. These associations were confirmed in the multivariate models. Including ex-smoking as predictor in the multivariate models did not change the results (data not shown). Additionally, log transformation of the dependent parameters did not change the conclusions.

Cytogenetic markers

Micronuclei frequencies as well as SCE rates over the course of the survey are given in table 1. No seasonal trend was seen nor did the results change systematically over the three years under study.

Comparing the matched examinations of harbour workers and controls (table 4), controls had significantly higher micronuclei frequencies than harbour workers. The major predictor of micronuclei frequencies was age. The latter result was confirmed in the multivariate regression analysis (table 5). In this model the association between exposure status and micronuclei frequency was no longer significant.

SCE rates were bivariately associated with cadmium levels in blood and urine. Additionally, smokers had non-significantly higher SCE rates than non-smokers. In the multiple linear regression model, current smoking and cadmium levels in blood were independent predictors of SCE rates.

DISCUSSION

Repeated measurements in a large number of harbour workers did not indicate increased biomonitoring levels of heavy metals. In contrast, urinary arsenic concentrations were above recent exposure limits in a large number of subjects. Besides age, the major predictor of these enhanced exposure levels was fish consumption. Likewise, the major predictors for heavy metal concentrations in biomonitoring and cytogenetic markers were non-occupational factors such as smoking habits.

Selection bias cannot be ruled out in our survey as only 40% of the harbour workers volunteered to participate in the survey. However, it seems unlikely that this bias was differential for the exposure under study. Eighty two of the 100 harbour workers stayed under study until the end of the survey. Therefore, selection bias over the course of the study could be minimised. While drop-outs were mainly due to retirement or loss of interest in the study, some participants were lost to follow up due to long term health reasons. It cannot be ruled out that these health complaints were due to occupation. Using matched surveys with alternating examinations of six harbour workers and six controls on the same day climatic factors, seasonal variations, and circadian rhythm were taken into account. This study was done in order to generate hypotheses rather than to test specific hypotheses. Therefore, multiple statistical testing has been done and should be considered in interpretation of borderline significant results.

Our results indicate high levels of arsenic in urine of some exposed workers. These levels were not different from controls and were mainly associated with recent fish consumption. Fish and mussels are known to contain arsenic.⁹ In our study, many of the participants reported regular consumption of salted young herring. We have recently shown that the consumption of salted young herring is associated with levels of arsenic in urine >100 μg/l.¹⁰ In general, subjects daily consuming fish have a mean daily intake of arsenic of 1.38 μg/kg body weight while those who never consume fish have an average daily intake of arsenic of 0.94 μg/kg body weight.¹¹ The German National Environmental Health Survey has found mean levels of arsenic in urine of subjects with regular fish consumption of 18 μg/l.¹² The levels found in harbour workers under study were within this range.

As for arsenic the association between fish consumption and mercury concentrations in blood is well established.^13–15 Only mercury concentrations in urine were weakly associated with amalgam fillings. This might be due to the fact that mercury levels in blood reflect inorganic and organic mercury compounds, while urinary mercury levels mainly indicate the inorganic portion of mercury.^16,17

The decrease in lead levels in blood over the course of the study are consistent with the decrease found in the general population.¹² The well known associations between active smoking and cadmium levels in blood and urine were confirmed in our study.^18,19

The cytogenetic markers were used to assess potential carcinogenic effects of the exposures under study.^20–22 Occupational effects on cytogenetic markers have not been observed in our survey. Age was the only significant predictor of micronuclei frequency. This is in agreement with current literature.^23–27 As in other studies no association was seen between smoking and micronuclei frequency.^23,25,27

SCE rates were significantly associated with active smoking and cadmium levels in blood. The relation between smoking and SCE rate is well established.^26,28 Results on the association between cadmium levels in blood and SCE rate are conflicting.^28,29 Most studies that were not able to show such a correlation were limited by small numbers.^29,30 In vitro, a genotoxic effect of cadmium has been described.^31,32 Using human lymphocytes this effect could not be confirmed.^33,34 In vivo, Fahmy and Aly³⁵ found a dose dependent effect of cadmium chloride on SCE rate. The reason for this inconsistency might be that cadmium may serve as a marker of environmental tobacco smoke (ETS) exposure.^18,19 Therefore, the association seen between cadmium and SCE rate may indicate the carcinogenic effect of passive smoke exposure.^36,37 Unfortunately, ETS exposure was not assessed in our study; therefore the association between ETS exposure and cadmium levels could not be tested.

In conclusion, we could not show an effect of long term occupational exposure to river silt aerosols on heavy metal concentrations in blood and urine of exposed workers. Additionally, no effects of occupational exposure on cytogenetic markers have been observed. Regular fish consumption might increase urinary arsenic concentrations well above current BEI levels. After adjustment for smoking habits, cadmium levels in blood were still a significant predictor of SCE rates. Therefore, cadmium levels in blood may reflect current exposure to ETS.