Ecological bias? Gunther Craun, Gunther F. Craun & Associates 5 June 2012 We find the ecological study of Linos et al. of interest but disagree that the finding ¿supports the hypothesis of hexavalent chromium (Cr+6) carcinogenicity via the oral ingestion pathway of exposure.¿ Potential confounding by personal or occupational exposures coupled with poorly defined Cr+6 ingestion exposures raise serious concerns about the validity of the hypothesized association. In ecological studies such as this, where the population group is the unit of observation for exposure and outcome, it is also important to consider ecological bias--the failure of the reported association to reflect an association at the individual level. Whether Cr+6 poses a cancer risk from ingestion exposures in humans is an important public health question (Stout et al. 2008). Occupational epidemiological studies have associated inhalation exposures with increased lung cancer. However, an increased risk of other cancers has not been consistently reported from occupational exposures (Gatto et al. 2010). There is scant epidemiological information about cancer risks associated with the ingestion of environmental levels of Cr+6. Since the study of Linos et al. does not provide information about a possible casual association, we urge caution in the use of these data as the basis for making regulatory decisions about waterborne levels of Cr+6. Municipalities in the study area were transformed into industrial areas beginning in the early 1970s with little or no restrictions on disposal of liquid industrial wastes into the Asopos River, which runs through Oinofita, the municipality with elevated cancer mortality. In 2009 there were 700 industries operating in the area with 500 generating undisclosed liquid industrial waste. Occupational, airborne, waterborne, or other environmental exposures to a wide range of undisclosed chemicals could account for the reported elevated risks. If liquid wastes were discharged into the surface and/or ground water sources during the period of industrialization, various chemicals would undoubtedly be present in Oinofita municipal water, possibly contributing to confounding or effect modification. Modest concentrations of arsenic were found in the water, and it is possible that other chemicals may have been present but not measured. Linos et al. note that the available water quality measurements for Oinofita water did not reveal high concentrations of other substances but do not provide a list of chemicals that were measured. The authors also note that the highest concentration (156 ppb) of Cr+6 was measured in a well close to the village of Agios Thomas (population 1090) and that three of the six liver cancer deaths were among residents of this village. Agios Thomas is one of the four villages in Oinofita municipality. This information seems to have been presented as evidence of a strong association between waterborne Cr+6 and liver cancer mortality, but readers are not informed whether the contaminated well is part of the Oinofita municipal system and used by residents of Agios Thomas. Rather, the authors offer contradictory statements about this specific exposure: ¿...exposure is expressed as residing in the area assuming that all residents consumed water provided by the municipality (Oinofita)¿ and ¿Well water is very rarely used for drinking in Greece and the level of contamination is very similar.¿ Confusion about the use of this contaminated well also raises concerns that the assumed ecological exposures may not necessarily reflect individual exposures. Exposure misclassification bias is also possible as ¿Initial concerns were raised after Oinofita area citizens complained about the discoloration and turbidity of their drinking water....¿ with regular protests from 1990s onward. Beaumont et al. (2008) noted that waterborne exposure to Cr+6 in an ecological study in China may have been self-limiting due to the lack of palatability of drinking water at high concentrations of Cr+6. Armienta-Hernandez and Rodriguez-Castillo (1995) also noted that participants in a study in Mexico did not consume water because of the yellowish color associated with Cr+6 concentrations above 500 ppb. Oinofita residents may also have limited their consumption of municipal water and used alternative sources of drinking water (e.g., bottled water). Bottled water use would lower actual exposures, and if bottled water use had been more prevalent among Oinofita residents, nondifferential misclassification bias would result in an over estimate of risk in the current analysis. All death certificates were scrutinized to exclude all metastatic liver cancers, and this reduces the likelihood for the misclassification of liver cancer mortality. While this strengthens the finding of elevated risks of liver cancer in Oinofita, it does not address the possibility that hepatitis B infection or alcohol consumption may be contributing risk factors. We agree with and strongly support the authors¿ recommendation that ¿Further studies are needed to determine whether this association is causal, and to establish preventative guidelines and public health recommendations.¿ However, to better understand possible risks associated with the ingestion of environmental levels of Cr+6, additional epidemiological studies should not use an ecological study design. Competing interests We have no competing interests. We accepted an honorarium from the American Chemistry Council to review this article. These comments are based on our review.