Letter to the Editor

Scand J Work Environ Health 1995;21(4):301-309    pdf

doi:10.5271/sjweh.1365

Pleural mesothelioma in oil refinery workers.

by Wong O

In a recent publication, Gennaro et al (1) described mortality from pleural mesothelioma in a cohort of 1661 blue-collar workers at two oil refineries in northern Italy. Ten men were reported to have died from pleural mesothelioma, and the corresponding standardized mortality ratio (SMR) of 266, based on provincial rates, was statistically significant at the level of =0.05 [95% confidence interval (95% CI) 128--489]. I would like to raise some questions and offer some comments on the paper. In the introduction of their paper, Gennaro et al cited a 1980 paper by Lilis et al, who suggested that the number of pleural mesotheliomas and lung cancers might be elevated among maintenance workers in oil refineries. As opposed to mesothelioma, lung cancer is a target of almost all cohort studies of oil refinery workers. Using meta-analyses, my co-workers and I have previously summarized mortality by cancer site for oil refinery workers (2). Based on all relevant studies available at the time, altogether 1903 lung cancer deaths were observed among oil refinery workers, compared with 2469.60 expected. The corresponding summary SMR for lung cancer was 77 (P<0.00001) and the 95% CI was 74--81. Furthermore, there was no difference in lung cancer mortality between operating and maintenance workers, or any upward trend by length of employment in refineries (3, 4). Thus, if lung cancer is used as a measure for asbestos-related cancers in refinery workers, there was no such increase. As for the study by Gennaro et al, results on lung cancer were not reported. Was there an increase in lung cancer in the two refineries in Genoa and La Spezia? Considering available data on exposure to asbestos in refineries (primarily from the United States), one would not expect any significant increase in asbestos-related diseases. For example, in its evaluation of cancer risk associated with exposures in petroleum refining, the International Agency for Research on Cancer (IARC) (5) reported that asbestos exposures for employees working with pipe insulation ranged from 0.1 to 0.9 fibers per centimeter (f·cc-1) and that the 8-h time-weighted average concentrations during turn-around activities involving the removal of lagging, gaskets, and insulation ranged from 0.01 to 0.15 f·cc-1. For insulation stripping and valve and joint repacking, measurements ranged from <0.01 to 0.02 f·cc-1. Removing insulation from compressors generated 0.06 f·cc-1 (6). Gennaro et al did not report any asbestos measurements for the two refineries in Genoa and La Spezia. It would be interesting to know whether asbestos exposures in these two Italian refineries were substantially higher than those recorded in refineries in the United States. In numerous epidemiologic studies, shipyard workers (regardless of specific trades) have had a significantly elevated risk of mesothelioma (7). In particular, those who worked in shipyards during World War II were at even higher risk than other shipyard workers. The primary reason for the increased risk was the extremely high asbestos exposure in the enclosed environment on board vessels (8--10). For example, fiber concentrations associated with the removal of blue sprayed asbestos ranged from 112 to 1906 f·cc-1, and with bagging asbestos debris from 106 to 3815 f·cc-1 (10). In northern Italy, several studies have reported increased risks for both lung cancer and mesothelioma among shipyard workers in Genoa (11), Trieste (12), and Monfalcone (13). Gennaro et al stated that there was widespread asbestos exposure in both Genoa and La Spezia because of the shipbuilding industry in the two provinces, and they were correct in pointing out that employment at the shipyards could have been a major confounding factor in their study. Because of the rarity of mesothelioma and the extremely high risk associated with shipyard employment, even a few cases with previous shipyard exposure would have been sufficient to produce a spurious association. It would be important to obtain the complete (life-long) employment histories of all 10 cases and to determine if any had been employed in the shipbuilding industry. The median latency (time since first employment) among the 10 cases was 27.5 years. For pleural mesothelioma, the average latency was approximately 40 years for persons heavily exposed. For example, Giarelli et al reported an average latency of 47 years and a median of 48 years for persons with pleural mesothelioma in Trieste, most of whom were shipyard workers (12). For exposures at much lower levels (such as those encountered in refineries), the latency would have been much longer, even if it is assumed that such exposures could increase the risk of mesothelioma. The observation of a median latency of 27.5 years for refinery workers in the Gennaro et al study is, therefore, not consistent with the interpretation that the mesotheliomas were caused by refinery employment. More likely, exposures from previous employments contributed significantly, which would bring the median latency more in line with that of other studies. In the San Francisco Bay Area in northern California, many of the maintenance trades in the shipbuilding industry in the 1940s or 1950s moved to other industries such as oil refining as the demand for shipbuilding declined beginning in the 1960s. It appeared that the rise and fall of shipbuilding in northern Italy might be similar to that in northern California (11, 13). With a large population of workers with mixed exposures, it is important to assign appropriate weights to exposures from specific occupations, particularly in the presence of the extremely high mesothelioma risk associated with shipyard employment. McDonald & McDonald (14) have developed a hierarchy for ordering the relative importance of asbestos exposure by occupation or industry. Apart from confounding exposures, the question of diagnostic comparability should also be examined. Gennaro et al stated that five cases had mesothelioma specified on the death certificates, while the diagnosis was less precise in the remaining five cases. It is not clear what "less precise" means and what rationale the authors used in arriving at the decision of assigning mesothelioma to the remaining five cases. The relevant question is "Would these five cases have been coded as pleural mesothelioma in the provincial or national mortality data?" According to both the 1991 abstract (15) and the 1994 report (1), it appeared that all cases were "documented" by a general practitioner. The fact that the findings of the entire study relied on one practitioner's judgment also raises some concern. The issue is not the accuracy of diagnosis but rather the comparability of diagnosis. Doll & Peto (16) have previously commented on the difficulties in interpreting comparisons based on noncompatible diagnostic criteria in studying asbestos-related neoplasms. Perhaps it would have been more balanced to present two SMR estimates, one assuming the remaining five cases were indeed mesothelioma and one assuming they were not. The two SMR estimates would have been 266 (95% CI 128--489) and 133 (95% CI 43--311), respectively. Thus the result of the study was completely dependent on the diagnostic decision of the five cases whose death certificates contained insufficient information. As a minimum, the actual causes of death listed on the death certificates should be reported in the paper so that the readers can make their own assessment. With respect to the statistical analysis, there are also several questions. First, as stated in the introduction of the article, it appeared that the study was initiated by two clusters of mesothelioma previously discovered in the refineries. Strictly speaking, the P-value in hypothesis testing, such as that in a cohort mortality study, is only applicable to a priori hypotheses fully and explicitly formed before the data are examined (17). Otherwise the P-value has a distinctly different meaning and carries a much reduced weight. The usual methods and formulas for testing hypotheses may be applied to post hoc hypotheses, but the usual interpretation may be profoundly weakened. If the population of a study is chosen in an a posteriori manner (such as choosing the two refineries with known clusters of mesothelioma cases), an ascertainment bias may be introduced which most likely increases the likelihood of demonstrating an excess (18). Second, because of the limited availability of the reference rates of mesothelioma, the statistical adjustments in the study were inadequate. In cohort studies, the usual practice is to use five-year age intervals. Only three age groups (25--54, 55--74, 75) were used by Gennaro et al. The rates within such wide age groups may not be homogeneous enough. For example, between 1980 and 1984 in the United States, the annual mesothelioma incidence rate for white males aged 70--74 years was 115 per 100 000, which was substantially higher than the rate of 21 per 100 000 for white males aged 55--59 years, a factor of more than fivefold (19). Yet these two age groups were combined into a single category in the Gennaro et al study. Furthermore, the reference rates were available for only three years (1975--1977), whereas the observation period of the study spanned approximately four decades. Instability of the reference rates presented yet another problem. How many deaths were the age-specific rates based on in each age group? Third, the authors put much emphasis on the results for workers with more than 10 years of employment and 20 years of latency. How meaningful this result was depended on what other exposures the workers might have had outside the refineries. Certainly the amount of asbestos fibers inhaled even for a year or two at the shipyards during World War II would far exceed 10 years of exposure at oil refineries. For mesothelioma, a latency of 40 years rather than 20 years would have been more appropriate. Furthermore, the authors indicated that no consistent pattern was noted when the two refineries were analyzed separately. Thus, although the analyses in this study might resemble those in a traditional cohort mortality study, the interpretation of the findings might have been substantially weakened as a result of the choice of refineries based on previously identified clusters, inadequate adjustments in SMR calculation, unstable reference rates, and incompatible diagnosis. Although the authors made a cautious statement that the role of asbestos exposure inside the refineries could not be completely resolved on the basis of the reported data and that a case-referent study would be needed, the authors, nevertheless, concluded that pleural mesothelioma should be considered an occupational disease among oil refinery workers. On the basis of the questions and criticisms I have raised, such a conclusion appeared both premature and unjustified.

This article refers to the following text of the Journal: 1994;20(3):213-215
The following article refers to this text: 1996;22(1):67