Brown and Rushton¹ have conducted a retrospective cohort mortality study of 2700 workers in the industrial sand industry. Work in the industrial sand industry results in exposure to crystalline silica, and the focus of the study was whether exposure to silica causes lung cancer. Retrospective exposure assessment, based on air measurements since 1978, and some assumptions about exposure before then, was used to estimate exposure levels for different jobs in the industry over time. The resulting job-exposure matrix was used to assign estimated exposure levels to each worker and to estimate cumulative silica exposure, which is commonly the exposure measure of interest for chronic diseases such as lung cancer.

Brown and Rushton did not find an excess of lung cancer in this cohort compared to the general population (lung cancer SMR 0.99, 77 deaths), nor did they find any excess silicosis (only two silicosis deaths were observed). Furthermore, they did not find a positive exposure-response trend for lung cancer by cumulative exposure category (rate ratios of 1.0, 1.24, 1.42, and 0.88 by increasing exposure).

Should this negative result be considered surprising? After all, the International Agency for Research on Cancer (IARC) declared in 1997 that crystalline silica was a group I (definite) human carcinogen, based on lung cancer findings across a large number of existing occupational studies and positive animal studies.² The National Toxicology Program (NTP) (www.ntp.niehs.nih.gov/ntp/roc/toc11.html) in the USA followed this by declaring silica a known human carcinogen in 2000. Our own subsequent pooled analysis of 10 large silica exposed cohorts (65 000 workers, 1000 lung cancer deaths) found a significant positive exposure-response trend with cumulative silica exposure,³ supporting the IARC decision.

I would argue that the negative results of Brown and Rushton are not that surprising, for two reasons. First, IARC noted in 1997 that positive findings for lung cancer were not consistent across studies, and that specific mineralogic properties of the silica might vary and result in different levels of toxicity. Therefore it is quite possible that a given study might be negative.

Second, the exposure levels in the Brown and Rushton study were rather low. Our pooled exposure-response analysis showed that silica is not a strong lung carcinogen compared to the classic occupational carcinogens such as nickel, arsenic, and asbestos. This is why it took over 20 years of studies to convince IARC and other agencies that silica should be considered a known carcinogen. In our pooled analysis, the lung cancer rate ratios did not increase in categorical analyses until after a cumulative exposure of 2 mg/m³-years (RRs of 1.0, 1. 0, 1.3, 1.5, 1.6 for <0.4, 0.4–2.0, 2.0–5.4, 5.4–12.8, and 12.8+ mg/m³-years). The geometric mean cumulative exposure in Brown and Rushton was 0.31 mg/m³-years, a level which a priori might not be expected to result in detectable excess of lung cancer. The high exposure group in the Brown and Rushton categorical analysis (which showed no lung cancer excess) was >1.0 mg/m³-year, still a relatively low level (for example, 10 years of working at 0.1 mg/m³, which is the occupational limit in many countries). We do not know the distribution of exposures within the high exposure group, but it may be that there were relatively few highly exposed workers there. The relatively low level of exposure in the Brown and Rushton cohort may be the reason for the fact that only two cases of death from silicosis were observed. It is interesting to note that another cohort of 4600 industrial sand workers in the USA had a similar low overall level of cumulative exposure, but there were 17 deaths from silicosis.⁴ A lung cancer excess was also observed in this cohort, concentrated among those with the highest exposures. Furthermore, a second study of US industrial sand workers,⁵ with some but not complete overlap with Steenland and colleagues,⁴ found an SMR for lung cancer of 1.39 and a significant positive exposure-response trend for cumulative and average exposure.

Figure 1 shows the exposure-response curve for the Brown and Rushton study (UK industrial sand) and the IARC multicentre study (note: I have used the exposure-response data from Brown and Rushton which did not control in the model for quarry, as quarry may be a surrogate for exposure). The figure makes clear that the UK study is restricted to relatively low exposures and that the odds ratios have relatively wide confidence intervals—overall the findings from the UK study are compatible with the IARC study.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

 Exposure-response trends for silica and lung cancer in two studies.

There are other aspects of the Brown and Rushton cohort which might tend to result in negative findings for lung cancer. There were only eight lung cancers among those with 20 or more years latency, where an effect might be most anticipated. The overall deficit of lung cancer primarily resulted from a marked deficit at one plant, which also had a significant deficit for all cause mortality and for all cancer mortality, possibly due to that plant having less follow up time and a greater healthy worker effect.

Does this negative finding for the Brown and Rushton cohort mean that there is a threshold below which silica exposure is not dangerous? There are several reasons why this may not be the case. Biologically, it is probably not a good idea to postulate a threshold for carcinogens which act via initiating a mutation in the DNA in a stochastic process. Statistically, it is often difficult to determine the shape of the exposure-response curve in the low dose reason, even with a large sample size. We looked at this question in our own pooled analysis.⁶ A threshold model provided only modest improvement in the log likelihood over a non-threshold model, and the optimal threshold was low, about 0.33 mg/m³-years. Regulators and policy makers often consider exposure levels for a hypothetical 45 year working lifetime. To stay below a threshold of 0.33 mg/m³-years would require a allowable level of less than 0.01 mg/m³, which is less than a tenth of the current standard in the USA. This very low level is unlikely to be technically feasible in most realistic occupational settings.

My own view is that the current standard in the USA, used in many other countries as well, is clearly too high. Several exposure-response studies in recent years have shown that the current US standard is not sufficiently protective to prevent silicosis.^7,8,9,10 The evidence is strong that silica can cause lung cancer,¹¹ and evidence is mounting that silica causes not only silicosis and lung cancer, but also renal disease and autoimmune diseases like arthritis and scleroderma.^12,13 A reduction of occupational limits to 0.05 mg/m³ (the NIOSH recommended standard) or to some lower but technically feasible level below 0.05 mg/m³ would go a long way to reducing disease due to occupational exposure. This point has been made repeatedly by other authors.^7,11

A broader question raised by the findings of Brown and Rushton is how we evaluate consistency in epidemiological studies, and when one might consider that a controversy about a putative carcinogen might be laid to rest. Clearly for silica one should expect that some new studies will be negative for lung cancer, especially when exposure levels are low. Nonetheless, policy makers must judge the weight of the evidence, as has been done by IARC and the NTP, agencies which evaluated more than 30 epidemiological studies of silica exposed workers. Given this level of past research, it might be considered high time for policy makers to act without the perennial call that “more research is needed”.