164 Occupational and Environmental Medicine 1995;52:164-169 Sex ratios of births, mortality, and air : can measuring the sex ratios of births help to identify health hazards from air pollution in industrial environments?

F L R Williams, S A Ogston, 0 L Lloyd

Abstract ratio in the workers' offspring, it is conceiv- Objectives-To compare the sex ratios able that air pollution from those industries of births and mortality in 12 Scottish might also cause abnormal sex ratios in the localities with residential exposure to offspring of residents of the neighbourhoods pollution from a variety of industrial exposed to those pollutants. Sex ratios have sources with those in 12 nearby and com- been studied in considerable detail2-9 but few parable localities without such exposure. studies could be found in which possible asso- Methods-24 localities were defined by ciations were investigated between sex ratios postcode sectors. SMRs for lung cancer and air pollution of the general environment. and for all causes of death and sex ratios In five retrospective studies, ''14 abnormal sex of births were calculated for each locality ratios were found in residential areas exposed for the years 1979-83. Log linear regres- to air pollution. The pollution reported in sion was used to assess the relation three of the studies had been readily seen and between exposure, sex ratios, and mor- smelt, and had also been associated with tality. significant excess mortality in the exposed Results-Mortalities from all causes were populations (possibly because of the quantity consistently and significantly higher in or quality of its toxicity).10 12 13 The pollution the residential areas exposed to air in the fourth study had also been visible pollution than in the non-exposed areas. and was often highly acrid, and had been A similar, but less consistently signifi- associated strongly with animal mortality and cant, excess of mortality from lung morbidity and with high frequencies of twin cancer in the exposed areas was also births in animals and in humans.'4-18 The fifth found. The associations between expo- study did not include any estimate of mortality sure to the general air pollution and or morbidity." abnormal sex ratios, and between The mechanism of how pollution could abnormal sex ratios and mortality, were affect the sex ratios of births is not clear, but negligible. the metabolism of the rapidly dividing cells of Conclusions-Sex ratios were not consis- the gonadal and fetal tissues is likely to be tently affected when the concentrations particularly vulnerable to the influences of or components of the air pollution were pollutants. The association between the insufficiently toxic to cause substantially nematocide DBCP (1,2-dibromo-3-chloro- increased death rates. Monitoring of propane) and low sex ratios, for example, was the does not provide a thought to be consistent with the spermatozoa reliable screening measure for detecting that bear the Y chromosome being damaged cryptic health hazards from industrial by exposure to DBCP, because both exposed air pollution in the general residential and non-exposed workers had similar environment. proportions of spermatozoa bearing the Y chromosome.'9 20 Department of (Occup Environ Med 1995;52:164-169) The question remained of whether this Epidemiology and Public Health, association between air pollution, sex ratios, Ninewells Hospital and mortality also existed in localities where and Medical School, the pollution was from a broad range of Dundee Keywords: sex ratios; mortality; residential exposure; F L R Williams industrial pollution industrial sources and probably in lower S A Ogston concentrations or of a less specific nature. If Department of Public so, the detection of such abnormalities in the Health and Abnormal sex ratios* of births (100 males sex ratio would constitute a simple screening Occupational have been clearly associated with procedure for alerting medical and environ- Medicine, The females) University ofthe specific occupational environments. ' These mental health authorities to the presence of United Arab associations have suggested that parental cryptic hazards to health from general Emirates, PO Box exposure to some occupational pollutants may industrial air pollution. This study, therefore, 17666, Al Ain, be a determinant of the sex ratio of their was designed to explore two hypotheses. o L Lloyd offspring. If exposure to pollutants within Firstly, that the sex ratio of births would Correspondence to: industries can cause an imbalance of the sex be altered, to either high or low, where the Dr F L R Williams, Department of parents have been exposed to environmental Epidemiology and Public air pollution from a range of industries. Health, Ninewells Hospital and Medical School, *(In this paper, the term "sex ratio" is used in place of the Secondly, that the mortality from lung cancer Dundee DD1 9SY. correct but longer term "secondary sex ratio" to refer to or from all causes of death would be higher in Accepted 13 October 1994 liveborn children only.) those exposed areas. Sex ratios ofbirths, mortality, and air pollution 165

Table 1 Criteriafor selecting the industrial sources and effects of generalised industrial air pollution, a the exposed localities, and the non-exposed localities variety of sources of industrial pollution were Exposed localities: included in the study (table 2). To assess the Selection ofthefactory; Information from local Departments of Environmental consequences of exposure to a heterogeneous Health including public complaints about airborne range of sources of pollution on the sex ratio pollution. Information from the Industrial Pollution Inspectorate. was a novel part of the design of the study, Selection of the exposed locality: which we had specifically chosen to differenti- Probable effects of wind direction. Influence of local topographical features. ate it from our previously published work on Height of the chimney stack of the polluting factory. single and specific industrial sources. Sociodemographic characteristics such as ethnic profile and employment of men. Different pollutants affect the sex ratio in Non-exposed localities: different ways and may cause both abnormally Proximity to, and comparability of general environment with, the exposed locality. high and abnormally low sex ratios. To Comparable sociodemographic characteristics. accommodate this, all of the exposed localities Absence of identifiable confounding factors from other industrial sources. selected were exposed either to a single source of pollutant or, where this was not possible, to similar types of industrial pollution from a compact source of exposure. Each locality comprised a few postcode sec- Methods tors (range 1-7). For each postcode sector Twenty four localities in Scotland were identi- (the data unit of the study), data on births, fied. The basic units of the study were defined deaths, the percentage of men in employment, by the postcode sectors of the national system and the percentage of residents born in the of postcode registration. Postcodes in United Kingdom were obtained from the Scotland operate at four levels: area, district, General Register Office in Edinburgh. sector, and unit, of which there are 15, 416, Scottish rates of mortality, derived from 898, and 124634 respectively. Postcode units, contemporaneous annual reports of the although differing in size geographically, are Registrar General, were used as standards for delineated to represent roughly equal popula- the calculation of standardized mortality tion sizes. The localities consisted of 12 ratios (SMRs). exposed localities matched to 12 non-exposed Results were analysed at two levels: the 12 localities, categorised according to predeter- exposed and 12 non-exposed localities and mined criteria (table 1). No networks of air also for the individual postcode sectors that sampling stations for monitoring objectively made up those localities. the quality and quantity of air pollution in the localities were in existence, nor had any been ANALYSIS OF INDIVIDUAL POSTCODE SECTORS in the past. As is customary in these circum- There were 72 postcode sectors in the data set stances we used an indirect approach to the of which 35 were in exposed localities and 37 assessment of the probable category of expo- were in non-exposed localities. Sex ratios sure-that is, exposed or non-exposed-to were calculated for all of the sectors for the which each location should be assigned. This years 1979-83. The significance of the sex was the approach we had used successfully in a ratios in each sector was determined with the previous study,'4 and it was derived from our normal approximation to the binomial distrib- experience in earlier studies in which air ution by calculating the Z statistic: pollution samplers had been used and which Z = - NPQ had indicated the importance of local (x NP) aV topographies and the directional axes linking Where: x = number of boys; N = total num- pollution sources to populations.'3 ber of boys and ; P = the Scottish propor- Because we were evaluating the sex ratio as tion of boys (106 . 206); Q = the proportion a measure of screening for the biological of girls (1 - P). WhenZ > ± 1-96,P<0-05 Table 2 Primary sources ofpollution and major pollutants The SMRs and their 95% confidence intervals for cancer and for all Area Pollutant source Primary pollutants (95% CIs) lung causes of death were calculated for each of the A Steel foundry Inorganics: arsenic, iron, lead, nickel, manganese, cadmium, zinc exposed and non-exposed postcode sectors E-P Municipal incinerator Inorganics: aluminum, antimony, arsenic, beryllium, for the years 1979-83. These years were cadmium, chromium, copper, iron, lead, nickel Organics: PCBs, and TCDDs* selected to afford stability and thus confi- E-S Chemical works Di-isocyanate, bisulphate, hydrogen sulphide dence in the population values used in E-C Hospital incinerator Inorganics: arsenic, lead, nickel Organics: PCBs, and TCDDs calculation of the SMRs. E-D Pipe coaters Tar and bitumen P Oil PAHs LOG LINEAR MODELLING WITH THE SECTOR G Petrochemicals PAHs B Chemical works Inorganics: arsenic, iron, lead, nickel, manganese, DATA and foundry cadmium, zinc was used to K Aluminium Aluminium Log linear modelling (GLIM21) M Steel foundry Inorganics: arsenic, iron, lead, nickel, manganese, analyse the mortality and sex ratio data at the cadmium, zinc sector level. This method aims to N Metal extraction Aluminium express log C Petrochemicals PAHs SMR as a linear function of explanatory variables. The variables in the model were *PCB (polychlorinated biphenyl) molecules consist of a biphenyl nucleus that has between one and 10 chlorine atoms giving 209 possible congeners that differ in the number and position of exposure (0 = no exposure, 1 = exposure), log their chlorine atoms. Many of the congeners are composed of mixtures of various isomers, each sex ratio, and percentage of men in employ- with its own independent toxic potential. 2,3,7,8-Tetrachlorodibenzo-p-dioxin (TCDD) is con- sidered to be the most toxic of the constituents of the family of polychlorinated hydrocarbons. ment. (As data on smoking habit were 166 Wiliams, Ogston, Lloyd

Table 3 Summary ofthe log linear regression ofall causes ofdeath and lung cancer side the Poisson assumption as shown by the deaths with GUM magnitude of the difference between the Death rate SEM Relative risk goodness of fit statistic (G2) and its degree of Variable Estimate (corrected) (95% CI) freedom (= 68). We report in this paper only No adjustmentfor sex ratio and employment the corrected SEM. Exposure Lung cancer 0-1708 0-0462 1-18 (1-08 to 1-30) All causes 0-1758 0-0117 1-19 (1-16 to 1-22) For comparison, we also computed the logistic regression of sex ratio on exposure Adjustmentfor sex ratio and employment Model 1: and percentage of men employed. Exposure Lung cancer 0-0558 0-0668 1-06 (0 93 to 1-21) (logSR) 0-0018 0-2895 1 00 (0-57 to 1-77) Employment -0-0181 0-0065 0-83 (0-83 to 0 85) LOCALITY ANALYSIS C.2 141-9 68 The sector data were subsumed to make Exposure All causes 0-1116 0-0290 1-12 (1-06 to 1-18) (log SR) -0-1205 0-1144 0-89 (0-71 to 111) 12 exposed matched to 12 non-exposed Employment -0 0111 0-0028 0-99 (0-98 to 0 99) localities. For the years 1979-83, the sex GP 360-4 68 ratios and SMRs for lung cancer and for all Model 2: causes of death were calculated for each of the Exposure Lung cancer 0-0537 0-0658 1-06 (0-93 to 1-20) (log SR)2 0-4932 1-3061 1-64 (0-13 to 21-18) exposed and non-exposed localities and the Employment -0-0179 0-0068 0-98 (0 97 to 1 00) results were compared. (12 141-5 68 Exposure All causes 0-1086 0-0289 1-12 (1-05 to 1-18) For each of the 12 matched localities, the (log SR)2 -0-6246 0 5579 0-54 (0-18 to 1-60) SMR and sex ratio were calculated for the Employment -0-0115 0-0028 0-99 (0-98 to 0-99) C12 356-9 68 exposed (denoted as e) and non-exposed exposed (denoted as ne) localities. The overall association was graphically represented by plotting Y against X where unobtainable, percentage of men in employ- Y = and X = ment, which reflects social class, was used SMRe-SMRne SR,-SRne instead of smoking levels.) We used two The study was planned with 12 exposed models: and 12 non-exposed localities, with 40 expected deaths in each. The significance was Model 1: log SMR = a + (exposure) + evaluated by a regression of Y = SMRe - fl,(log SR) + fi (men in employment). SMEne against X = SRe - SRne. The total sample was sufficient to detect a difference in Model 2: log SMR = a + f81 (exposure) + Y of 0-6, with a power of 85%. fl2(log SR)2 + ,6, (men in employment).

The first model tested for whether a high Results sex ratio indicated a high mortality and the SECTOR ANALYSIS second model tested for whether an extreme According to the data in the 1981 , sex ratio (either high or low) indicated high the numbers and percentages of people born mortality. The model that included (log SR)2 in the United Kingdom exposed and non- was probably more appropriate than that exposed sectors did not differ appreciably (P which included (log SR), as exposure was > 005). Overall, the mean percentages born coded 1 for exposed and 0 for non-exposed, in the exposed and non-exposed sectors fi represented the difference in the log SMR respectively were 96 6% and 95 7%. between the exposed and non-exposed locali- ties. The antilog of fi gave the relative risk for LOG LINEAR MODELLING exposure adjusted for the other variables in Regression of log SMR on exposure alone the model. Thus, the equation described the showed that mortality for lung cancer and for effect of exposure on the SMR. The GLIM total mortality was significantly higher in the system reports the parameter estimates for a exposed than in the non-exposed localities andfl, with their SEMs and tests of signifi- (table 3). cance. Two SEMs were calculated. The first Table 3 shows the regression coefficients SEM was based on a Poisson assumption and SEMs for the regression of log SMR for obtained from GuM, the second SEM was lung cancer and for all causes of mortality corrected and allowed for the variability out- on (log SR), (log SR)2, % employment, and

Table 4 Summaries ofSMRs and sex ratios for the 12 localities: 1979-83 Exposed Non-exposed Exposed Non-exposed Exposed Non-exposed Lung cancer Lung cancer AU causes All causes Locality SMR (95% CI) SMR (95% CI) SMR (95% CI) SMR (95% CI) Sex ratio Z score Sex ratio Z score A 68 (47 to 87) 81 (64 to 98) 96 (90 to 102)* 85 (81 to 89)* 101 0 30 98 -0 27 E-P 122 (100 to 144) 95 (76 to 114) 95 (90 to 100)* 81 (77 to 85)* 114 0-06 105 -0-66 E-S 186 (153 to 219)* 92 (65 to 119)* 118 (111 to 125)* 88 (81 to 95)* 100 -0 07 121 1-53 E-C 133 (105 to 156) 91 (64 to 119) 107 (102 to 112)* 93 (86 to 100)* 127 1-41 97 -1-43 E-D 121 (94 to 148) 103 (84 to 126) 101 (95 to 107)* 84 (79 to 89)* 102 -0-83 118 0-93 P 78 (58 to 98) 55 (35 to 75) 103 (97 to 109)* 86 (80 to 92)* 108 -0 35 112 0 44 G 93 (73 to 113) 109 (86 to 133) 88 (83 to 93)* 101 (95 to 107)* 104 0-27 100 -0 30 B 85 (50 to 120) 69 (40 to 99) 102 (92 to 112) 106 (97 to 115) 109 0-32 103 -0 34 K 182 (88 to 335) 96 (39 to 180) 102 (81 to 123) 97 (79 to 115) 92 -0-06 94 0-05 M 99 (85 to 113) 91 (76 to 106) 108 (104 to 112)* 87 (83 to 91)* 110 1 00 103 -0-93 N 87 (73 to 101) 95 (69 to 121) 112 (108 to 116)* 94 (88 to 101)* 114 0-63 106 -0-86 C 102 (85 to 119) 81 (66 to 96) 109 (104 to 114)* 79 (75 to 83)* 100 -0-44 104 0-44 *P < 0 05 exposed v non-exposed localities. Sex ratios ofbirths, mortality, and airpollution 167

Table 5 The overall means ofthe sex ratios and of the localities; however, none of these differences SMRsfor lung cancer and all causes in the localities and non-exposed to industrial pollution was significant (table 4). For lung cancer, exposed nine of the 12 exposed localities had SMRs Mean (SD) t Statistic P value higher than the SMRs in their equivalent Sex ratio: non-exposed localities, but for only one of Exposed 107 (9) 0-436 0-672 Non-exposed 105 (8) these nine was the difference significant (table SMRs lung cancer: 4). For all causes of death, by contrast, 10 of Exposed 113 (38) 2-461 0-032 Non-exposed 88 (15) the exposed localities had higher SMRs than SMRs all causes: the non-exposed localities, nine of these Exposed 103 (8) 3-68 0 004 differences were significant, and in only one Non-exposed 90 (8) area was the SMR for the exposed locality sig- nificantly lower than its counterpart (table 4). The scatter plot (fig 1) of the relation between the difference in the SMR for lung exposure. The SMRs for mortality from cancer for the exposed and non-exposed all causes in the exposed localities were localities (SMRe-SMRne) on the y axis and significantly higher than those in the non- the difference between the sex ratios for exposed localities (relative risk 1-12, 95% CI the exposed and non-exposed localities 1 06-1 18). The difference between the (SRe-SRne) on the x axis, showed only a SMRs for lung cancer in the exposed and weak association (rho-0 224). Only three non-exposed localities was reduced to a non- localities showed extreme differences. significant relative risk (1 *06, 95% CI Between all causes of death and the sex ratio 0-93-1-21). For all causes of death, the addi- (fig 2), the relation was slightly stronger tion of (log SR)2 in the model maintained the (rho - 0306). significant association between total mortality Because the distributions of the SMRs and and exposure (relative risk 1 12, 95% CI sex ratios among the localities were consistent 1-05-1-18). Lung cancer mortality showed a with normal distributions, the overall means positive association with (log SR)2 and expo- for the sex ratios and SMRs were compared sure, but all causes mortality showed a nega- with the Student's t test. The means of the tive association with (log SR). sex ratios did not differ significantly between The regression of the sex ratio on exposure the exposed and non-exposed localities and employment was not significant (odds (table 5). The mean SMRs for the exposed ratio 1-03, 95% CI 0-98-1-07). This mirrored localities were significantly higher both for the results of the locality analysis. lung cancer and for all causes of death than the corresponding SMRs for the non-exposed LOCALITY ANALYSIS localities (table 5). The sex ratios were higher in the exposed localities in only seven of the 12 matched Discussion The United Kingdom's Clean Air Acts of

Figure 1 Scatter plot 100 - the 1 950s and 1 960s have successfully showing the differences in reduced the gross health hazards caused the SMRfor lung cancer by matter and sulphurous com- and the sex ratio between 80 _ * particulate 0~ pounds. The environmental air pollution, of exposed and non-exposed x localities. 0) 60 increasing importance to public health in the 1970s and beyond, includes the less visible ~0 40 pollutants that contain the heavy metals and C'a organic compounds (table 2). Perinatal and 0. 20 * were once crucial variables by x a) * which the quality of the public's health was 0 - measured; with improved general health and CI) access to medical care, however, these vari- -201 - o I I ables have become less discriminating. It has -30 -20 -10 0 10 20 30 40 been suggested that as the sex ratio of births Sex ratio (exposed-not exposed) might be considered to reflect deaths at much earlier stages, this variable might be used to monitor the quality of public health. Obstetric epidemiology would have the Figure 2 Scatter plot 40 - advantage over other variables of chronic showing the differences in a) disease of a relatively short latency between 0 the SMRfor all causes of 0L L30 -exposure and measurable outcome. death and the sex ratio x a1) studies showed that the sex ratio between exposed and non- 20 - Previous exposed localities. ~0 of births could be abnormal where pollution 10 - concentrations from specific industrial 0) (0 o processes were sufficiently toxic to cause clear 0 0

x * cut increases in the contemporaneous rates -10 - for all causes of death and in particular for (I) -20 I I lung cancer. As one of the aspirations of -30 -20 -10 0 10 20 30 40 environmental epidemiology is to identify Sex ratio (exposed-not exposed) hazardous environments before they cause 168 Williams, Ogston, Lloyd

gross morbidity or mortality, a study was was pursued exhaustively, there was no con- needed in which communities exposed to founding of the pattern through occupational less obvious pollution from a wider range experience. of sources were examined. In our study, How best to estimate the flow of air pollu- however, no convincing evidence was found tion in the nearby localities and hence their that exposure to generalised airborne pollu- assignation into exposed and non-exposed tion was associated with abnormal sex ratios, categories is the second methodological and a non-significant negative relation was difficulty. It is often met in environmental present between abnormal sex ratios and total epidemiological studies where (usually) no mortality. comprehensive network of sites exists for The SMRs in the exposed localities were monitoring the spread of air pollution from significantly higher than those in the non- defined industrial sources. In studies of exposed localities for total mortality and similar relations, conventionally only the (less consistently) for lung cancer. The proximity between the addresses of cases and non-exposed localities had been selected the sources of pollution has been used.23 because of their geographical proximity to, In addition to this measure, we also took and environmental comparability with, the into account the important influences of exposed localities. Hence, an allowance for wind direction, topography, and height of socioeconomic variables in the study units the sources of pollution on the probable was intrinsic in the comparisons. After the distribution of the pollution. Nevertheless, SMRs had been adjusted further by taking this method ultimately and inevitably lacks into account the percentage of men in precision because the exposed and the not employment (used instead of social class and exposed localities (if nearby) cannot be hence for smoking) the difference between the accurately demarcated in the absence of SMRs for lung cancer of the exposed and measurements of chemical concentrations in the unexposed localities was no longer signifi- environmental samples. The cause-effect cant, although that difference did remain relation is likely to be artefactually weakened significant for all causes of death. Because in the exposed localities and strengthened in of the known relation between lower social the non-exposed localities. classes and both higher consumption of In our study, the assessment did not tobacco and higher mortality, we decided include information about the total exposure to introduce this variable and treat it as a of the parents to the pollution, and so some potential confounding factor. births might have been misclassified with These observations showed two method- regard to antenatal exposure. Uncertainties ological difficulties frequently encountered on this point exist for both the at risk exposed in environmental epidemiology. The first areas and the non-exposed comparison areas, difficulty was how to control adequately for and there is no reason to suppose that one differences of general socioeconomic factors differed appreciably from the other. Lastly, without risking overcontrol. Our first method other factors that can affect the sex ratio (such of matching the two types of locality for as parental age, parity, and exposure to general environmental circumstances and disease) were unavailable and therefore could social class (and hence smoking), by geo- not be taken into account. In view of all these graphical proximity, ethnicity, and general circumstances, therefore, any impact of the perception of the localities, had credibility in pollution upon the sex ratios (and mortality) being derived from a direct assessment of the that was found in this study must have been study areas themselves, but was partly subjec- artefactually low. Nevertheless, despite the tive. With this method, the close comparability inherent imprecision of the exposure cate- of the values of ethnicity between exposed and gories, differences in mortality between the non-exposed populations supported the view exposed and non-exposed types of locality that the socioeconomic structures of the remained, with the mortality from all causes two types of locality might be similar. Our remaining significantly high with all the additional method, incorporating the percent- analytical tests. age of men in employment instead of social The tests for the two hypotheses of this class, was based totally on an objective study led to two conclusions with their related variable and was valid on a national basis. We implications. Firstly, the sex ratios of births had no evidence, however, that this indirect were not consistently affected in areas where measure of smoking accurately reflected any the toxicity of the pollution was either of the differences of social class (and smoking wrong chemical composition to produce that habits) between the neighbouring localities in effect or was insufficient to cause large and this particular study. unequivocally significant increases in the It is always conceivable that occupational SMRs from both total mortality and lung exposures (for which data are not available) cancer. Therefore, the routine monitoring of could play a part in environmental patterns sex ratios of births is unreliable as a screening of disease and sex ratios of births. The dis- measure for the detection of cryptic health tances from the pollution sources of most hazards in general industrial environments. of the non-exposed localities were compar- Hence, the search for further specific types able to those from the exposed localities. of pollution that affect the sex ratios of Indeed the comparison localities had been births should be pursued. Secondly, the chosen because they satisfied that condition. areas exposed to pollution from a range of In a previous study in which this question industrial sources had higher SMRs for lung Sex ratios of births, mortality, and air pollution 169

cancer for all causes of 7 Hytten FE. Commentary: boys and girls. Br Obstet and particularly Gynaecol 1982;89:97-9. mortality than the unexposed areas, even 8 Erickson JD. The secondary sex ratio in the United States definition of exposure was based 1969-71: association with race, parental ages, birth when the order, paternal education, and legitimacy. Ann Hum on estimates rather than on data from a Genet 1976;40:205-12. sampling network for air pollution. This find- 9 Chandra HS. Is human X chromosome inactivation a sex- determining device? Proc Nad Acad Sci USA 1985; ing indicated the need for investigations to 82:6947-9. exclude the possible effects of other 10 Lyster WR, Bishop MWH. An association between rainfall and sex ratio in man. Jf Reprod Fertil 1965;10:35-47. confounding variables and to determine the 11 Lyster WR. Sex ratio of human births in a contaminated natures and sources of the toxic air pollutants area. Med jAust 1977;1:829-30. 12 Lloyd 0, Lloyd MM, Holland Y, Lyster WR. An unusual in those environments. It also suggested that sex ratio of births in an industrial town with mortality authorities could monitor the problems. BrJ Obstet Gynaecol 1984;91:910-07. public health 13 Lloyd OL, Smith G, Lloyd MM, Gailey F. Raised mortal- mortality in their populations exposed to ity from lung cancer and high sex ratios of births associ- even in the absence of ated with industrial pollution. Br Ind Med 1985; industrial pollution 42:475-80. detailed information from pollution monitor- 14 Williams FLR, Lawson AB, Lloyd OL. Low sex ratios ing networks. If the cause-effect relation sug- of births in areas at risk from air pollution from incinerators, as shown by geographical analysis and 3- ested in this paper is confirmed by such dimensional mapping. IntJ7 Epidemiol 1992;21:311 -9. would also indicate that 15 Williams FLR. The epidemiological surveillance of mortal- studies, the findings ity and other health indices in traditional communities the legislation on air quality control should be in Scotland. Dundee: University of Dundee, 1989. (PhD to minimise exposures to these thesis.) strengthened 16 Lloyd OL, Lloyd MM, Williams FLR, Lawson AB. environmental hazards. Twinning in human populations and in cattle exposed to air pollution from incinerators. Br Ind Med 1988;45:56-60. We thank Mrs Wendy Mitchell for help with data processing 17 Smith GH, Lloyd OL. Soil pollution from a chemical and Professor C du V Florey for his very helpful suggestions waste dump. Chemistry in Britain 1986;22:139-41. for improving this paper. This work was supported by an 18 Lloyd OL, Lloyd MM, Williams FLR, McKenzie A, Hay MRC grant: MRC A 601/108. A. Ragwort poisoning, fat cow syndrome, or industrial chemical toxicity? The value of an epidemiological analysis. Sci Total Environ 199 1;106:83-96. 1 McDowall ME. Occupational reproductive epidemiology: the 19 Whorton D, Milby TH, Krauss RM, Strubbs HA. use of routinely collected statistics in England and Wales Testicular function in DBCP exposed pesticide workers. 1980-82. Studies on medical and population subjects. Occup Med 1979;21:161-6. London: HMSO, 1982:50. 20 Potashnik G, Goldsmith J, Insler V. Dibromochloro- 2 Goldsmith JR, Potashnik G, Israeli R. Reproductive out- propanes-induced reduction of the sex ratio in man. comes in families of DBCP-exposed men. Arch Environ Andrologia 1984;16:213-8. Health 1984;39:85-9. 21 Baker RG, Nelder JA. The GLIM system. In: Release 3 ed. 3 James WH. The human sex ratio. Part 2: a hypothesis and Oxford: Numerical Algorithms Group, 1978. a programme of research. Hum Biol 1987;59:873-900. 22 Lloyd OLL, Ireland E, Tyrrel H, Williams FLR. 4 James WH. The human sex ratio. Part 1: a review of the Respiratory cancer in a Scottish industrial community: a literature. Hum Biol 1987;59:721-52. retrospective case-control study. Soc Occup Med 1986; 5 Rehan NE. Sex ratio of live-born Hausa infants. Br 36:2-8. Obstet Gynaecol 1982;89:136-4. 23 Elliott P, Hills M, Beresford J, Kleinschmidt I, Jolley D, 6 Jakobovits AA. Sex ratio of spontaneously aborted fetuses Pattenden S, et al. Incidence of cancers of the larynx and and delivered neonates in second trimester. EurJ Obstet lung near incinerators of waste solvents and oils in Great Gynaecol Reprod Biol 1991;40:211-3. Britain. Lancet 1992;339:854-8.

Correspondence and editorials Occupational and Environmental Medicine wel- minimum. Letters are accepted on the comes correspondence relating to any of the understanding that they may be subject to material appearing in the journal. Results editorial revision and shortening. from preliminary or small scale studies may The journal also publishes editorials which also be published in the correspondence are normally specially commissioned. The column if this seems appropriate. Letters Editor welcomes suggestions regarding should be not more than 500 words in length suitable topics; those wishing to submit an and contain a minimum ofreferences. Tables editorial, however, should do so only after and figures should be kept to an absolute discussion with the Editor.