Editor—In their ecological study of 27 countries Grech et al reported a higher male to female ratio at birth for southern Europe than for central and Nordic Europe.^1-1 A reversed latitudinal relation emerged for North America, with Mexico and the United States yielding lower sex ratios than any European country; Canada's sex ratio (although not stated) was somewhat higher. The authors were unable to explain this cross continental pattern reversal and concluded that a temperature related (or latitudinal) effect on sex ratios is unsupported. Here we show that cross continental differences are likely to be artefacts and that there is a latitudinal effect on sex ratios.

Firstly, the study periods differ noticeably: whereas European data are for 1950-99, North American data are for 1958-97. Sex ratios peaked internationally after the second world war,^1-2 and different onsets of study periods may thus account for observed cross continental differences in sex ratios.

Secondly, Canada, the United States, and Mexico are too large for ecological analysis. Their combined area equals 4.8 times the combined area of the 24 European countries investigated. Noticeable differences in sex ratios within countries have been reported for Canada^1-3 and may well extend to the United States and Mexico.

Thirdly, the authors obscured the relation between latitude and sex ratio by aggregating the 24 European countries' sex ratios into three latitude bands. We reanalysed only the European data presented, since the reported North American sex ratios were not suitable (see above). Regressing the latitude of the countries' capital city on their sex ratio yielded 22% variance in sex ratios explained by latitude (P=0.021). Subsequent entry of a squared latitude term into the regression model explained an increase of a further 27.7% variance in sex ratios (P=0.003). The regression equation is sex ratio=0.542 – 0.001064 * latitude + 0.000009982 * latitude squared.

Within Europe the lowest sex ratios are found in moderate latitudes, where the amplitudes of seasonal change in climate are most pronounced. In contrast, higher sex ratios are found in both Nordic and southern Europe, where less ambient seasonality is experienced. Our finding of a curvilinear relation of latitude and sex ratio is consistent with an effect related to photoperiod, as has been put forward for numerous other observations, such as seasonality in the incidence of suicide^1-4 and the ratio of the length of the index finger to that of the ring finger.^1-5

Editor—Grech et al report that significantly more male than female infants were born in southern latitudes in Europe, with the reverse finding in North America.^2-1 The reasons for this are unexplained. We have made an incidental finding that maternal cytomegalovirus seropositivity is associated with an increased proportion of female infants.

We pooled data from two studies from which the prevalence of cytomegalovirus infection in women of childbearing age in Northern Ireland could be calculated. The first study related to the prevalence of viral infections in infancy.^2-2 Cord blood was collected at birth for 236 consecutive mothers and analysed for cytomegalovirus IgG by standard immunofluorescence. The cord samples (analysed as described) from the second study were taken as part of a study of fetal loss associated with parvovirus B19 infection.^2-3

In the first study 96 (41%) samples of cord serum were positive for cytomegalovirus. Altogether, 123 male infants and 113 female infants were delivered to this group of women. The sex ratio (male:female) was 1.08:1. A relation between cytomegalovirus seropositivity in cord blood and female sex was noted (odds ratio 1.9, confidence interval 1.12 to 3.21, P=0.017). In the second study 625 consecutive samples were tested; 268 (43%) were positive for cytomegalovirus. There were 311 male infants and 314 female infants (sex ratio (male: female) 0.99:1). Cytomegalovirus seropositivity was again associated with female sex (odds ratio 1.46, 1.06 to 2.00, P=0.02). Pooling the data from both studies gave a cohort of 861 with a male:female ratio of 1.02:1; 364 (42.3%) cord blood samples were positive for cytomegalovirus.

A highly significant relation between cytomegalovirus seropositivity in cord blood and female sex was obtained, with an odds ratio of 1.6 (1.19 to 2.06, P=0.001). Maternal cytomegalovirus seropositivity most probably occurs before pregnancy, as seroconversion rates during pregnancy are low. There is an association between cytomegalovirus seropositivity and unmarried status and social deprivation.

The cervix is thought to harbour latent cytomegalovirus, and cytomegalovirus shedding has been shown in vaginal secretions.^2-4 Cytomegalovirus may influence sex determination by effects on the quantity and quality of cervical mucus (for example, facilitating penetration of sperm carrying the X chromosome or blocking penetration of sperm carrying the Y chromosome), sperm motility, success of implantation, and selective male fetal loss.

Any attempt to explain a change in the sex ratio and variation among different racial groups should consider the possible influence of the prevalence of cytomegalovirus infection.^2-5