Study design and conduct
4. It is widely recognised that there are limitations in ecological studies of this kind. These types of study can only explore whether there is an association between the health outcomes analysed and the environmental hazard under investigation. They cannot say whether or not any reported association is causal. There is no assessment of whether the study population is actually subject to harmful exposures or if it is subject to different exposures than the reference population. It is assumed that those living near to the hazard are "exposed" and those living further away are "unexposed". Ideally, a geographical study would incorporate a "dose-response" analysis by calculating relative risks in populations living within different distance bands from the hazard under investigation. It is unfortunate that this was not possible in the present study - SAHSU informed us that further subdivisions of distance less than 2 km from the site would not be meaningful in view of the uncertainties in the geographical data both on landfill sites and on postcode locations.
5. A further limitation of ecological studies is that they can only make adjustments for variables such as socio-economic deprivation on a group level and not on an individual level. In the SAHSU study, in common with other studies, socio-economic deprivation was adjusted for using a postcode-based index, the Carstairs index, which is formed by grouping several socio-economic variables derived from census data, which may differ at the individual level. Ecological studies cannot adjust explicitly for confounders such as family history of disease, lifestyle factors such as smoking, use of medicines and occupation, which might themselves be associated with the health outcomes being studied and which are unlikely to be completely accounted for by adjusting for deprivation alone. Therefore, there is a possibility of residual confounding. In addition, this particular study may have had problems of data quality. We are informed that the landfills data, although thoroughly checked for consistency by SAHSU, may be subject to inaccuracies or omissions eg regarding the exact location of sites or the types of waste deposited. There are also limitations in some of the health statistics data sets used in the study. For example, the congenital anomalies register for England and Wales is known to be incomplete, although we note that SAHSU also used hospital admissions data to provide an independent source of data for some anomalies.
6. Nevertheless, while recognising these limitations, we consider the SAHSU study to have been well conducted and we welcome it as a useful addition to the literature. It has the advantage of being much larger than previous studies and is population based, thereby ensuring that the study and reference populations were classified on the basis of proximity to all known landfill sites, rather than a few selected sites.
7. The study analysed data on several congenital anomalies and the prevalence of stillbirths and low and very low birthweights. The adjusted rates of these health outcomes in the study population, living within 2 km of a landfill site open during the study period, were compared with standard rates, obtained from model predictions of data from the reference population. We note that the area within 2 km of a landfill site tended to be more urban and more deprived than that beyond 2 km, with 34% versus 23% of the population in the most deprived tertile of the Carstairs score. We note also that there were differences between the two populations in terms of maternal age and ethnicity. All of these differences may influence the rates of outcomes reported, and, as noted above, adjustment for these may have been incomplete.
8. The study found that the risk ratio(*) for congenital anomalies overall was 1.01 (ie the adjusted rate was 1% higher in the study population than in the reference population). When individual congenital anomalies were considered, the adjusted risk ratio for neural tube defects was 1.05 but for cardiovascular defects, in contrast, it was 0.96 (ie the adjusted rate was 4% lower in the study population than in the reference population). The congenital anomalies register data indicated a risk ratio of 1.07 for hypospadias and epispadias, but this finding was not supported by hospital admissions data, for which the adjusted risk ratio was 0.96. For abdominal wall defects, the congenital anomalies register data indicated a risk ratio of 1.08 and the hospital admissions data yielded a similar estimate. The risk ratio for hospital admissions for surgical correction of gastroschisis and exomphalos was 1.19.
* "The risk ratio is the ratio of the rate in the study population, adjusted for deprivation and other factors, to the standard rate, obtained from model predictions of data from the reference population."
9. Adjusted rates for stillbirths were the same in the study and reference populations. For low birthweight and very low birthweight, risk ratios were 1.05 and 1.04 respectively.
10. We note that the risk ratios were all close to unity, and were much smaller than the estimated relative risks reported in the EUROHAZCON study. We note also that a "before and after" sub-analysis of sites which opened during the study period indicated that the risk ratio for congenital anomalies did not increase after the landfill sites opened and, indeed, that the ratios for abdominal wall defects appeared to decrease substantially. The sub-analysis was based on smaller numbers than the primary analysis but, coupled with the difficulties in interpretation of the study which we have outlined above, indicates that it is inappropriate to draw firm conclusions on the possible health effects of landfill sites from the results of this study.
11. The "before and after" analysis found higher risk ratios for low birth weight and very low birth weight after opening. We consider that there is a possibility that residual socio-economic confounding might account, at least in part, for the differences in these two parameters between the study and reference populations.
12. Another sub-analysis focussed on special waste landfill sites. We note that this analysis also was based on smaller numbers than the primary analysis and that a higher proportion of the population living around these sites was in the most deprived tertile of Carstairs score (36%) than in the main study population. We are informed that special waste sites are, in practice, co-disposal sites which take predominantly the same type of waste as most landfill sites, but are also licensed to take certain specified types of special waste. Without good information on emissions and controls at these and other landfill sites, it is not possible to assess whether special landfill sites may present a greater potential risk to health than other landfill sites. In this sub-analysis, the risk ratio for congenital anomalies overall was 1.07, and this may be compared to the risk ratio of 1.01 in the main analysis. The risk ratios for neural tube defects, and for low birth weight and very low birth weight, showed similar elevations to those found in the main analysis. The risk ratios for abdominal wall defects, and for hospital admissions for surgical correction of gastroschisis and exomphalos were lower than for the main analysis. For cardiovascular defects, the risk ratio was 1.11, whereas it was less than one in the main analysis. The congenital anomalies register data for hypospadias and epispadias indicated a risk ratio of 1.11 but, as in the main analysis, this finding was not supported by the hospital admissions data. The differences remain lower than those found in the EUROHAZCON study. The lack of consistency in the findings for all landfill sites and for special landfill sites further complicates any attempt to draw firm conclusions from this study about the possible health effects of landfill sites. Nevertheless, the finding of a risk ratio of 1.07 for congenital anomalies overall for populations living around special waste landfill sites, whether or not it is related to the presence of the landfill sites, merits further investigation.
13. The study analysed the incidence of childhood and adult leukaemias, hepatobiliary cancers, and cancers of bladder and brain. We understand that these endpoints were selected either to test hypotheses arising from the previous studies or on the basis of the known human carcinogenicity of certain chemicals known to be present in landfill sites. An additional uncertainty in this analysis was the use of short lag periods to allow for the latency period between initiation of cancer and the time of diagnosis/registration. SAHSU used a lag period of one year for childhood leukaemia and five years for the other cancer outcomes. We accept that this was a pragmatic approach, taken in order to increase the number of years of data available for analysis and to reduce the potential for dilution by migration. If, however, the latency period is longer, this index of potential exposure may be inappropriate, leading to dilution of any potential effect. However, taking this and other limitations of the study design into account, we consider that the finding of no excess risk for those living within 2 km of a landfill site for each of the cancer types studied provides a degree of reassurance.
14. We were informed that a programme of research and reviews is underway on congenital anomalies and landfill sites, and that this includes a project to measure emissions from landfill sites and to assess exposures of people living nearby. We welcome this as important information that was lacking from the present study and we have considered whether there is other work that could usefully be undertaken to address this issue. We consider that further exploration of the possibility of residual confounding in the SAHSU study would be a legitimate area of future research. Further, although this study found at most only small differences in adjusted rates of some birth outcomes between the study and reference populations, when considering landfill sites in general it remains possible that there are individual sites (or some subset of sites) which significantly affect the health of the local population. SAHSU has proposed that this could be investigated further by detailed mapping and statistical analysis of the existing data to provide an indication of any systematic variation in rates and to analyse any resulting variations in relation to possible explanatory variables (eg landfill characteristics, geology, other exposure sources, deprivation). We agree that this could be a useful way forward but note that the value of further analyses of the existing datasets may be limited by the known problems of some of these datasets.
15. An alternative approach to an ecological design would be to carry out case-control or cohort studies and to obtain estimates of exposure for individual study subjects, together with data on all relevant confounding factors and effect modifiers. Future studies of this type would greatly benefit from the development of accurate and specific marker(s) of exposure to chemicals released from landfill sites.
1. Dolk H, Vrijheid M, Armstrong B, Abramsky L, Bianchi F, Garne E, Nelen V, Robert W, Scott JES, Stone D and Tenconi R (1998). Risk of congenital anomalies near hazardous waste landfill sites in Europe: the EUROHAZCON study. Lancet 352 (423-427).
2. 1998 Annual Report of the Committees on Toxicity, Mutagenicity and Carcinogenicity in Food, Consumer Products and the Environment: Landfill sites and congenital anomalies, pp 12-13. Department of Health, London.
3. WHO. Methods of assessing risk to health from exposure to hazards released from waste landfills. Report from a WHO meeting, Lodz, Poland, 10-12 April 2000. Bilthoven: WHO Regional Office for Europe, European Centre for Environment and Health, 2001.
4. Birth outcomes and selected cancers in populations living near landfill sites. Report to the Department of Health. Elliott P et al., The Small Area Health Statistics Unit, Department of Epidemiology and Public Health, Imperial College. August 2001.