We appreciate the interest of Dr. Pesch and colleagues in our systematic review investigating the exposure-risk relationship for occupational hexavalent chromium (Cr(VI)) exposure and lung cancer in order to establish exposure limits (Seidler et al. 2012). Pesch et al. rightly point to uncertainties in exposure and risk assessment in existing studies of Cr(VI) exposure and lung cancer—as we also did in our original article. However, from our point of view, it is not sufficient to ask for more research regarding these uncertainties: There is no doubt that Cr(VI) exposure causes lung cancer; therefore, exposure limits are urgently required. Under the circumstances of scientific uncertainty, the best estimate based on the data available has to guide setting such exposure limits; waiting for certainty endangers workers in the meantime; and would therefore not be acceptable. The controversial discussion of definitive exposure limits is not only guided by scientific knowledge, but also by political and economic interests. Without a doubt, science will not always be successful in convincing all stakeholders of its results, especially if there is residual uncertainty. We nevertheless believe that clearly named and transparent rules of scientific decision making constitute an essential requirement for rational governmental regulation. We therefore regard the methods of evidence-based medicine—especially the systematic search for and quality assessment of evidence—as indispensable when establishing exposure limits.

These evidence-based methods were defined a priori and applied rigorously in our systematic review (Seidler et al. 2012). A comprehensive literature search was conducted, including independent screening of titles and abstracts by two reviewers. The full-texts were screened for eligibility, again by two independent reviewers. Following the “Guide for the quantification of cancer risk figures after exposure to carcinogenic hazardous substances for establishing limit values at the workplace” (AGS 2008), the included studies were required to report risk estimates for more than one level of cumulative occupational Cr(VI) exposure. An additional inclusion requirement was that the studies should adjust for or otherwise adequately consider potential confounding by smoking, the most critical confounding factor in studies on respiratory cancers. Moreover, we only included studies providing direct measurements of the exposure to Cr(IV) in the air.

According to this methodology defined a priori, the studies of Luippold et al. (2005), Birk et al. (2006), and Gerin et al. (1993)—which Pesch and colleagues explicitly refer to—were excluded from our derivation of the exposure-risk relationship, for the following reasons:

  • According to our predefined criteria, the study of Luippold et al. (2005) was not included because no exposure-risk relationship was reported there. Moreover, we would like to point to the limited power of this study, conducted among only 617 employees, leading to wide 95 % confidence intervals of 0.17–2.44 for lung cancer.

  • According to our predefined criteria, the study of Birk et al. (2006) was also not included in our derivation of an exposure-risk relationship because of the lack of direct measurements of the exposure to Cr(IV) in the air. Instead, presumed exposure to Cr(VI) was calculated from biomonitoring data of urine chromium. As we explained in our original review, measurements of urine chromium suffer from the disadvantage that it is not possible to distinguish between Cr(VI) and trivalent chromium (Cr(III)) exposure. Cr(VI) undergoes reduction after entering the human body, and urinary chromium can be detected only as Cr(III), regardless of whether it originally derived from Cr(VI) or Cr(III). Because we cannot exclude significant occupational or nonoccupational Cr(III) exposure or other confounding, an uncertainty remains as to how informative urine chromium measurements could be of Cr(VI) exposure at the chromate production plants. Furthermore, concerns exist with this study because of the skewed distributions of urine chromium measurements and the timing of urine sample collection (“before the workers went to the workplace”, rather than after the shift).

  • In a historical prospective study of lung cancer risk among a large sample of welders in nine European countries, Gerin et al. (1993) estimated cumulative Cr(VI) exposure for workers in 135 welding companies. As direct measurements of Cr(VI) were not available for the exposure assessment, the researchers developed an exposure matrix using known information regarding levels of Cr(VI) associated with different welding processes and the information provided by the welding companies with respect to specific working conditions, including the types of metals welded and the specifications of the welding process. This exposure matrix was then used to calculate the cumulative Cr(VI) exposure for each member of the cohort. Unfortunately, information regarding smoking status was known only for the study subjects in Finland and Norway, comprising 23 % of the cohort. Although the researchers state that the smoking status of the Finnish and Norwegian welders was comparable to that of the general population, the distribution of smoking status among different exposure categories was not considered. Based on their data, the researchers found no increase in risk when cumulative Cr(VI) exposure increased. However, an imbalanced distribution of smoking among the various exposure categories could have artificially distorted the dose–response relationship. Therefore, in our opinion, this study did not sufficiently consider smoking and also lacked a detailed description of exposure measurement methods. It was therefore given a low quality assessment score (“-”) and excluded from the derivation of the exposure-risk relationships in our systematic review.

Furthermore, we did in fact consider the heterogeneity between the two cohorts that were included in our analysis of the exposure-risk relationship: We presented our derived exposure limits separately for Crump et al. (2003) and Park et al. (2004); moreover, we calculated the exposure limits with different methods (i.e. conditional and life-time methods). Because direct genotoxicity was considered the predominant mechanism of carcinogenesis of Cr(VI), in accordance with the AGS (2008) recommendations, we extrapolated into the low-dose range applying linear models to fit risk data. We nevertheless discussed potential nonlinearity of the dose–response relationship in our review. However, we could find no indication for a threshold in the relationship between Cr(VI) and lung cancer; on the contrary, a possible plateau at high exposures has to be taken into consideration.

As the core result of our systematic review, we state in the above-mentioned publication (Seidler et al. 2012): “Based on an exposure level of 1 μg/m3 Cr(VI), as a ‘best estimate’, an excess risk between 2.3 and 4.9 per 1,000 was calculated (applying different reference populations, different age ranges, and different calculation methods with a β value of 1.75)”. In other words: Based on the presented best available evidence, a workplace exposure level of 1 μg/m3 Cr(VI) leads to 2.3–4.9 excess lung cancer cases per 1,000 workers. Even though—in accordance with Pesch et al.—the uncertainty in this “best estimate” should not be neglected, this uncertainty should not keep us from aiming for adequate workplace health protection by deriving concrete exposure limits from the best available evidence.