The journal has published in this issue¹ and in a previous recent issue² two papers with striking evidence relating atmospheric contaminants, and mostly emissions derived from oil based combustion from engine exhausts, with increased risk for children of dying from leukaemia and other cancers. The risk was particularly high for children born in the proximities (up to 1 km) of bus stations, but also hospitals and other industrial resources with oil combustion emissions. The results from the different data and several analyses presented in both papers are firmly consistent with the author’s conclusions: childhood cancers are strongly determined by prenatal or early postnatal exposures to oil based combustion gases, especially from engine exhausts. Two key points arise from such a conclusion. Firstly, how good is evidence supporting this statement? Secondly, is there an additional argument for stricter control of atmospheric emissions?

Results from both papers, although clearly coherent, are based on data and analysis with some limitations, a common occurrence in epidemiological research, the same category of research that discovered that tobacco causes several health effects and that water born infectious diseases can be effectively prevented. In this case, Knox uses for his analysis a database of children dying from cancer between 1953 and 1980 and maps of emissions for 2001 published by the National Atmospheric Emissions Inventory. Then, most of the putative damaging exposures (prenatal or early postnatal, according to the author) occurred more than 20 years before the data on atmospheric emissions used in the study and, probably, to higher levels of emissions than levels for 2001 in the same places. But this fact does not remove any relevance to Knox’s findings: the subsequent questions are how many locations have now emissions similar, or even higher, to those happening several years ago (in UK and in other parts of the remaining world), and how safe are present levels of emissions now.

Another methodological point is that Knox’s results are based on mortality data. It will be worth examining also data on childhood cancer incidence (that is, including also child non-lethal cancers): would it be expected to find even higher risks for concerned contaminants? Also, a particularity of Knox analysis is that, as a consequence of the lack of a proper control group, the analysis uses only cases of migrant children (changing address between birth and death). This approach is based on the premise of migration equilibrium around emissions hotspots in the general child population. It would be interesting to test the validity of this approach (analysis of migrants) against confirmed effects of atmospheric emissions. Available data from well established networks of research in this area³ could be useful for this testing.

Besides limitations, Knox also finds a clear dose-response association for exposures and childhood cancers, an important criterion for evaluating causality in exposure-damage associations. Another important criterion for causality is consistency. This is to say, if different researchers, observing different populations with different approaches arrive at similar findings, it is probable that the observed association (namely, emissions from combustion from engine exhausts and childhood cancer) is causal, and not casual. This point is crucial for prevention, as only for causal associations will interventions eliminating or reducing exposure be followed by a reduction in the damage frequency or intensity. Knox² quotes studies from USA, Italy, Sweden, and France finding increased risks for leukaemia and other childhood cancer related to residence near high volume roads and petrol and vehicle repair garages. Knox also refers to some studies that have not found any relations. Apparently studies in this area are still scarce, but perhaps there is already enough information for a systematic review or meta analysis properly weighting all available evidence together.

The second point raised at the beginning of this commentary is even more crucial. Should these papers, and other available evidence, be used on behalf of stricter control of oil combustion emissions? Should government and authorities seriously consider these findings and develop any specific actions? And also: could scientific journals play a critical part in decision making processes related to issues affecting people’s health, but also strongly depending on political, economical, social, and cultural influences? Well, as much as they can. This journal has deliberately developed an editorial policy focused on publishing the best scientific evidence for socially relevant research on health determinants (“the journal of the increasingly relevant”⁴): the “policy implications” window in articles or the “Evidence based public health policy and practice” section are reassuring proofs of this editorial determination. In a recent editorial⁵ the role of public health policy was examined in its relation with public health research, particularly addressing interventions over gradually augmenting health inequalities. Also poorer people are more likely to live near more contaminating industrial resources.

The gains for health and wellbeing from searching for “appropriate engineering and social solutions”¹ to better control toxic atmospheric emissions from oil combustion surpass the expectations of any single study.^3,6–8 But also, childhood cancer is an important socially sensitive public health problem and Knox’s findings^1,2 should be seriously taken into account by decision makers concerned with environmental emissions’ limits and control.