In their 1999 study of workers exposed to acrylamide, Marshet al conducted an SMR analysis, and fitted several relative risk regression models to the data.1 In each analysis, they found risk of pancreatic cancer increased by about twofold for workers in the highest cumulative exposure group, but risk of pancreatic cancer did not increase monotonically with cumulative exposure in any of their analyses. Duration of exposure was monotonically related and mean intensity showed a nearly monotonic relation with risk of pancreatic cancer.

The cut off points Marsh et al chose for the cumulative exposure groups are based on multiples of current and proposed regulated levels of exposure intensity.1 2Because these cut off points resulted in small numbers of expected deaths in the low and intermediate exposure groups, 1.08 and 2.74 respectively, we have regrouped the data to attempt to obtain more stable standardised mortality ratios (SMRs). These results are presented in table 1 and indicate a monotonic dose-response pattern with the SMRs increasing from 0.80 to 1.31 to 2.26.

In part based on the absence of a pattern of monotonically increasing risk with increased cumulative exposure, Marsh et al argue that “our findings for cancer of the pancreas should be interpreted with caution, in the context of an exploratory analysis to generate hypotheses.”1 Nevertheless, given the sufficient evidence in experimental animals for the carcinogenicity of acrylamide.3 this study plays an important part in the evaluation of safety for occupational exposures to acrylamide.

When data are sparse, it is not always clear how best to choose cut off points; the grouping we have shown results in a finding that is more compatible with the findings for duration and for intensity of exposure. It would be interesting to see if a regrouping of the exposure categories alters the results of the analyses based on internal comparisons.