METHODS

The Cumbrian Births Database

The study area was the county currently defined as Cumbria.²¹ The Cumbrian Births Database has been described in detail elsewhere.^22,23 In summary, birth registration details of all 241 524 live births and 3234 stillbirths born to mothers usually resident in the study area, from the opening of the first crematorium in 1956 to 1993, were supplied from the Office for National Statistics and entered onto a computer database.^22,23 During this period a stillbirth was defined as a fetal death occurring after 28 weeks gestation (from 1 October 1992 fetal deaths occurring after 24 weeks gestation were included, consistent with current legal definitions).^24,25 Death registrations for the cohort, including those that occurred outside Cumbria, were supplied by the Office for National Statistics from the National Health Service Central Register (NHSCR), which was the primary source of ascertainment of deaths. NHSCR routinely records deaths of all residents of England and Wales who have ever registered with a general practitioner. However, hospital records within Cumbria and in regional referral centres outside Cumbria were searched to ascertain unregistered stillbirths and infant deaths.²⁶ All causes of stillbirth and death were coded to ICD-9. Causes of death and stillbirth were confirmed, where possible, through examination by a consultant neonatologist of details obtained from medical and/or postmortem records (the cause of about 50% of deaths were confirmed in this way). Thus, when postmortem and/or clinical records were available, causes of stillbirth and death were validated from a number of sources and derived using a more robust method than relying on death or stillbirth certificates. Neonatal death was defined as death within the first four weeks of life.

Several outcome groups were considered: stillbirth, neonatal death, stillbirth plus neonatal death, lethal congenital anomaly (overall and by cause category). Deaths from congenital anomaly (ICD740–759) were grouped by cause, using a standard classification of infant deaths,²⁷ into the following hierarchical and mutually exclusive categories: all neural tube defects (ICD740–742), congenital heart defects (ICD745–747), other congenital anomalies. Neural tube defects were subdivided into: anencephalus (ICD740), spina bifida (ICD741), other central nervous system anomalies (ICD742). All other lethal congenital anomalies were grouped because of the small numbers within each ICD-9 cause category.

The mother’s address on the child’s birth certificate was postcoded and hence grid referenced.²⁸ The father’s occupation, as recorded on the birth certificate, was assigned a social class.²⁹ Algorithms based on parents’ names were used to assign birth order and identify multiple births.²³

Geographical data (see table 1)

The grid references and dates of operation of incinerators in Cumbria were ascertained from Environment Agency records. No incinerators operated before 1977, and four operated between 1977–93. The locations of all crematoriums were ascertained from specialist digests and the dates of operation were obtained. During the period 1956–1993, three crematoriums operated. Details were captured in the geographical information system Arc/Info.³⁰ Mercury represents the main pollutant from crematoriums.^3,4 By contrast, emissions from incinerators incorporate a more complex mixture of dioxins, furans, particulates (such as chloride and sodium), heavy metals (including lead and chromium), and volatile organic compounds (such as chloroform).^1,31 Because of the differences in emissions, incinerators and crematoriums were analysed separately. Three of the four incinerators in Cumbria all dealt with materials defined as difficult by the Environment Agency,³² the other (incinerator 1) processed only inert and biodegradable material.³² Because of the historical nature of this study no detailed emissions data were available. Details of the material dealt with at each incinerator are presented in table 1. The location of all incinerators and crematoriums in Cumbria, 1956–93 is show in figure 1.

Analysis methods

A measure of exposure of each birth to incinerators and crematoriums was computed using the distance function 1/(D+0.1)² where D was the distance in km from the site and the measure was summed over all sites that were in operation at the time of birth.

Stillbirth and neonatal death rates fell substantially over the study period.³³ The cause of stillbirth was recorded on the stillbirth registration only from 1961 onwards. Hence the analysis in relation to proximity to crematoriums was stratified by time period: 1956–60, 1961–71, 1972–82, 1983–93. As incinerators in Cumbria were in operation only between 1977 and 1993, this analysis was not stratified by time period.

Multivariate logistic regression³⁴ was used to model how the risk of each outcome varied in relation to proximity to incinerators and crematoriums, adjusting for the known demographic risk factors—year of birth, social class, birth order, and multiple births—using offsets from an analysis of the effects of demographic risk factors without the exposure function. Year of birth was modelled using both quadratic and linear terms. Social class, birth order, and multiple births were treated as categorical variables (social classes I, II, IIIn, IIIm, IV, V, armed forces, and unknown, father not recorded on the births certificate; birth order 1, 2, 3, and ⩾4; multiple births, yes/no). A sensitivity analysis was carried out repeating the logistic regression, but excluding births with the greatest influence, as measured by Pregibon’s influence statistic.³⁴ For incinerators the analysis was repeated for the period before any incinerators were open, 1956–76. Because multiple births may not be considered independent events, robust estimates of variance were used and significance assessed from the corresponding p value.³⁵

RESULTS

Incinerators (see table 2)

The risk of stillbirth and neonatal death was not significantly increased closer to incinerators. However, the risk of lethal congenital anomaly was significantly higher (p<0.01). This significantly increased risk was restricted to heart defects and neural tube defects, specifically spina bifida. Sensitivity analysis demonstrated that these results remained significant when the most influential births were excluded. Replication of the analysis, using the location of incinerators for the time period before they were open, showed no increased risk for any outcome (table 2).

Crematoriums (see table 3)

During 1956–93 there was a significantly increased risk (p<0.01) of stillbirth closer to crematoriums, reflecting a consistently increased risk from 1961 onwards. The risk of anencephalus was also significantly increased during this period (p<0.05), due to a significantly increased risk in 1961–71. Although most (92%) cases of anencephalus were stillborn, the significantly increased risk of stillbirth remained after exclusion of anencephalus cases from the analysis. From 1972 onwards there was an increased risk of all other congenital anomalies, excluding neural tube defects and heart defects, with increasing proximity to crematoriums, which was significant (p<0.01) for the period 1983–1993. These findings remained significant after exclusion of the most influential births.

DISCUSSION

Summary

We found a significantly increased risk of lethal congenital anomaly (specifically spina bifida and heart defects) in relation to proximity to incinerators, but not of stillbirth or neonatal death. In contrast with Elliott et al,¹² who found an increased risk for certain congenital anomalies in areas where landfill sites were later opened, we found no increased risk for any outcome in areas where incinerators were subsequently opened. Hence, there was no evidence that these increased risks might be attributable to features of the environment where incinerators were located.

Around crematoriums, there was a consistently increased risk of stillbirth from 1961 onwards. There was also a significantly increased risk of anencephalus during 1961–1971, when case ascertainment was highest because this time period largely pre-dated antenatal screening for this outcome. In the later two time periods there were very few cases of anencephalus in term pregnancies and hence statistical power to detect an effect was greatly reduced. There was a significantly increased risk of all other lethal congenital anomalies around crematoriums from 1983 onwards. This increased risk was not observed in earlier time periods despite a greater number of cases, suggesting either that a small association was obscured in earlier time periods by cases due to causes that were eliminated or reduced during 1983–93, or that the significant association in 1983–93 was a chance finding.

The significant statistical associations are different for incinerators and crematoriums. While we cannot infer a causal effect from these statistical associations, the inconsistency may be attributable to the different pollutants emitted by crematoriums and incinerators,^1,4,5,31 or it may reflect confounding with other unmeasured risk factors, or it may be a chance finding. In addition, the time periods of operation of incinerators and crematoriums were different (1977–93 and 1956–93 respectively). Hence, while we observed a significantly increased risk of anencephalus with proximity to crematoriums during 1961–1971, we did not in the later time periods for either crematoriums or incinerators. It is unlikely that any association between proximity to incinerators or crematoriums and the risk of anencephalus would be detectable in later time periods when the number of cases was low because of prenatal screening and therapeutic termination.

Strengths and weaknesses of the study

Our study covered 38 years, allowing us to investigate a potential environmental hazard with a large cohort of 244 758 births. Changes in medical practices over time may have affected the results. Medical advances, such as improved antenatal care, which allows more fetuses to be carried to at least 28 weeks, and improved gestational dating, may have increased the number of deaths classified as stillbirths. However, other advances, such as better fetal monitoring and improved resuscitation, may have decreased the number of stillbirths either by shifting potential stillbirths into the category of neonatal deaths or by preventing infant death. The introduction of antenatal screening and elective termination reduced the number of stillbirths and deaths attributable to congenital anomalies in recent years.³⁶ Thus the clinical characteristics of the cases in the 1950s and 1960s may be intrinsically different from those in the 1990s. However, all analyses were adjusted for year of birth, such that the risk of stillbirth, lethal congenital anomaly or neonatal death to mothers living close to incinerators or crematoriums was, in effect, compared with that of other mothers giving birth around the same time. Hence, the objectives of our study were not affected by changes in the nature of cases over time.

Because the Cumbrian Births Database recorded all birth registrations in Cumbria during the study period by date of birth and postcode of mother’s residence, we had precise data on the population at risk and the location of each birth. Consequently we were able to estimate exposure and risk within a continuous model unconstrained by the availability of population statistics from other sources and we did not have to restrict our analysis to traditional geographical areal units. However, a limitation of our study was the unavailability of data on pregnancies less than 28 weeks gestation (24 weeks since 1 October 1992), which will affect the population at risk because some serious congenital anomalies might not continue to this stage of maturity, either through spontaneous abortion or termination. The inability to include such cases in our study is likely to have resulted in a conservative estimate of the effects of proximity to sources of pollution. A further limitation of our study was the exclusion of non-lethal congenital anomalies, although we were rigorous in our ascertainment of deaths, stillbirths, and lethal congenital anomalies.²⁶ Because no data were available for non-lethal congenital anomalies these cases could not be excluded from the live birth control group. However, as the control group comprised all live births that survived over 28 days non-lethal congenital anomalies would have comprised a very low proportion of the comparison group.

We were able to incorporate exposures of each birth to putative pollution from several sites. In addition, we had demographic information for each birth and hence were able to take account of individual risk factors, such as social class, which we have shown previously to be a better predictor of stillbirth rates than community based deprivation measures such as the Townsend score.²³ Grid references for incinerators and crematoriums were supplied to an accuracy of 100 metres.

We assumed that the mother’s residence during pregnancy was the same as that recorded on the birth registration. Hence migration of mothers during pregnancy may have resulted in misclassification of exposure, which would have tended to obscure any association between risk of adverse pregnancy outcome and proximity to crematoriums or incinerators.

A further limitation was that, as actual pollution levels around each site were unknown and would be impossible to ascertain retrospectively over such a long time period, we relied on a function of distance as a surrogate for potential exposure. The form of the exposure function, 1/(D+0.1)², assumed that exposure increased rapidly with proximity to the sites. A potential mechanism for absorption of toxic pollutants from incinerators or crematoriums by pregnant women might involve direct inhalation of pollutants or contact through food, soil, or water contamination. We assume higher pollution levels closer to the point source and thus the distance function is a reasonable surrogate indicator that has been used in many similar studies investigating health risks around pollution sources.^11–13,15 Although we could not consider any changes in pollution levels over time all analyses were adjusted for year of birth, so the risk of adverse pregnancy outcomes for mothers living close to crematoriums/incinerators was compared with that of mothers giving birth in the same year.

The facilities in Barrow in Furness and Dalton in Furness are located near to industrial sites defined as hazardous by the Environment Agency. Hence, there is some potential for confounding between proximity to incinerators/crematoriums and proximity to hazardous industrial sites.

There is potential for confounding between distance from incinerators and crematoriums and unmeasured risk factors, such as diet, lifestyle, or occupational exposures. However, we adjusted for individual social class, which is likely to be related to such lifestyle factors. Hence, this study can only identify a potential statistical association between exposure to incinerators or crematoriums (modelled by a function of distance) and adverse pregnancy outcomes. We cannot establish the biological plausibility of these findings given the lack of detailed emissions data. Further studies are now required using actual pollution levels around crematoriums and incinerators to investigate the biological plausibility of our findings.

We undertook a large number of comparisons and hence it is possible that some of the significant results may be chance findings, arising through multiple significance testing. However, our results of raised risk of stillbirth, congenital heart defects, and neural tube defects were generally consistent between time periods and sensitivity analysis showed they were robust, which lessens the probability of them being chance findings. As with all geographically based studies, there was potential for confounding with lifestyle and sociodemographic risk factors that were not included in the analysis, although we were able to adjust for individual level socioeconomic status, which has not been possible in many other studies.

Comparison with other studies

Although several studies have considered pregnancy outcomes for mothers living close to hazardous waste and municipal landfill sites,^{11–15,17,37,38} there is a paucity of epidemiological data concerning pregnancy outcomes around incinerators and crematoriums with which to compare our study. Our finding of an increased risk of lethal congenital anomalies, in particular neural tube defects and congenital heart defects, in babies born close to incinerators is consistent with the results of some studies of congenital anomalies around hazardous waste and municipal landfill sites^11–15 but not with others.^37,38 Nevertheless, our findings need to be interpreted cautiously, as both the pollutants and exposure pathways associated with these sources differ. While incinerators are sources of a range of chemicals, including some also emitted by hazardous waste and municipal landfill sites, they also emit dioxins, heavy metals, and particulates.³¹ Furthermore, the exposure pathways from incinerators and landfill sites are different^12,13,31: exposure of humans to landfill pollution results from water supply contamination, groundwater run off, and atmospheric contamination from landfill gases,^12,13 whereas pollutants from incinerators are primarily dispersed atmospherically.

Although incinerators and crematoriums in Cumbria were located in urban areas, there were so few in operation that only 10% of the Cumbrian birth cohort were born within 2 km of an incinerator or crematorium, in contrast with the finding by Elliott et al¹² that 80% of the population in England and Wales live within 2 km of a landfill site.

Conclusions

We found an increased risk of lethal congenital anomaly (specifically spina bifida and heart defects) in relation to proximity to incinerators and an increased risk of stillbirth and anencephalus in relation to proximity to crematoriums. In view of the scarcity of published data and our use of a distance function to represent potential exposure it is difficult to assess whether these statistical associations reflect a causal effect. Further investigations using actual pollution levels and high quality data, including lethal and non-lethal outcomes in term pregnancies and elective terminations, are required. Sufficient investment must be made in national registration systems to ensure these issues can be investigated adequately. The UK system for registration of congenital anomalies is known to be incomplete and this severely restricts its credibility.³⁹