Introduction
The literature on sex ratio at birth is abundant but there is still a lively discussion about the population and ecological factors affecting the sex of a child, as well as how changes in these factors explain the temporal variation in sex ratio at birth values for some populations. It is now well established that extraneous factors affect parental reproductive physiology at or near the time of conception and/or during gestation, thus favouring the specific sex of the subsequent child. These factors most probably affect the levels of parental sex hormones, which in turn cause differential conception and/or differential survival in utero.
A substantial body of evidence exists that the sex of a child is affected by parental age (James, Reference James1971; most probably father's age, see Jacobsen, Reference Jacobsen2001), birth order (Chahnazarian, Reference Chahnazarian1988), parental socioeconomic status (Teitelbaum & Mantel, Reference Teitelbaum and Mantel1971; Catalano et al., Reference Catalano, Bruckner, Gould, Eskenazi and Anderson2005b), parental occupation (Magnuson et al., Reference Magnuson, Bodin and Montgomery2007), smoking and alcohol consumption (Dickinson & Parker, Reference Dickinson and Parker1994) and nutrition (Mathews et al., Reference Mathews, Johnson and Neil2008). Also, parental exposure to environmental hazards, severe stress and other unfavourable conditions affect the sex ratio at birth. This has been demonstrated in war situations (Graffelman & Hoekstra, Reference Graffelman and Hoekstra2000; Saadat & Ansari-Lari, Reference Saadat and Ansari-Lari2004; Polasek et al., Reference Polasek, Kolcic, Kolaric and Rudan2005), and after earthquakes (Fukuda et al., Reference Fukuda, Fukuda, Shimizu and Moller1998; Saadat, Reference Saadat2008), terrorist attacks (Catalano et al., Reference Catalano, Bruckner, Anderson and Gould2005a, Reference Catalano, Bruckner, Marks and Eskenazi2006) and ecological catastrophes (Mocarelli et al., Reference Mocarelli, Gerthoux, Ferrari, Patterson and Kieszak2000). Additionally, the sex of a child is partially determined by geographical location (Masoudi & Saadat, Reference Masoudi and Saadat2007), temperature (Helle et al., Reference Helle, Helama and Jokela2008) and season of birth (Lerchl, Reference Lerchl1998; Melnikov, Reference Melnikov2003).
Although it has been shown that the yearly values of sex ratio at birth for different countries fluctuate non-randomly (James, Reference James2000), a visible secular trend emerged during the second half of the twentieth century. A decline in sex ratio at birth was observed in northern and eastern Europe, Greece, Portugal, Mexico (Parazzini et al., Reference Parazzini, La Vecchia, Levi and Franceschi1998), Canada (Allan et al., Reference Allan, Brant, Seidel and Jarrell1997), USA (Marcus et al., Reference Polasek1998; Mathews & Hamilton, Reference Mathews and Hamilton2005; Davis et al., Reference Davis, Webster, Stainthorpe, Chilton, Jones and Doi2007) and Japan (Davis et al., Reference Davis, Webster, Stainthorpe, Chilton, Jones and Doi2007), while a rise was present in southern Europe, Australia (Parazzini et al., Reference Parazzini, La Vecchia, Levi and Franceschi1998) and some Asian countries (Jha et al., Reference Jha, Kumar, Vasa, Dhingra, Thiruchelvam and Moineddin2006; Guilmoto et al., Reference Guilmoto, Hoang and Van2009). The most probable causes of lowered sex ratio at birth are increased environmental pollution (James, Reference James1987) and the use of infertility treatments (James, Reference James1985), while those of increased sex ratio are the cultural preference towards male children (Gu & Roy, Reference Gu and Roy1995) and better health care, which results in lower mortality of male fetuses (Moller, Reference Moller1996).
The sex ratio at birth in Croatia has not been thoroughly investigated. The sex ratio values in Croatia were not found to be affected by the 1991–1995 war conflict (Polasek et al., Reference Polasek, Kolcic, Kolaric and Rudan2005), although there is some evidence for the war effect in neighbouring countries (Polasek, Reference Polasek2006). A temporal trend in the yearly data was also found to be absent (Pavic, Reference Pavic2011). Because of the lack of research and relatively good and reliable sources of data, an investigation of the sex ratio in Croatia could prove fruitful. This research sought to explore more thoroughly the relationship between the population and ecological status of parents and the sex of their offspring, using data for each birth in Croatia from 1998 to 2008 provided by the Croatian Bureau of Statistics.
Methods
The Croatian Bureau of Statistics (CBS) gathers information on every childbirth in Croatia, provided by the local registrar (form DEM-1). There are twenty-seven variables on the DEM-1 form, ranging from the name of the child and place of birth to specific information about the socioeconomic status of both parents. This information is presented in an aggregate form in official publications, and the individual entries can only be acquired for research purposes, after signing a non-disclosure agreement with the CBS. This agreement has been signed by the CBS and the Center for Croatian Studies, University of Zagreb (class: 953-06/09-2/45, number: 380-1/1-09-2). The names of the children and other personal information are not available since the law forbids the disclosure of this information. The CBS provides a total of 468,783 birth records for the years 1998 to 2008, 466,656 of which were live births and 2127 were stillbirths. The variables chosen for this analysis were sex of the live-born child, birth order, county and region of birth, and age, education level and activity status of both parents. Multiple births were excluded from the analysis, as well as those entries with unknown values for any of the variables. This resulted in a total of 420,256 analysed units.
A somewhat short time period was chosen because of certain methodological constraints. Although Croatia's borders have remained approximately the same since 1946, today's internal administrative division into counties (županija) was established only in 1997. Also, up until 1998, every vital event was reported as occurring in the permanent residence of the mother instead of their usual residence, which is advised by the United Nations (United Nations, 2001). During the war period in Croatia (1991–1995) not every part of Croatia was available for registration and this situation lasted until 1998.
For the purposes of easier analysis, some of the variables were re-coded. The number of categories for the variable ‘education’ was changed from ten to just three, the first one consisting of persons with up to 3 years of vocational high school education, the second of persons with a 4-year high school diploma, and the third of those having some kind of higher education degree. Since there is no official regional division in Croatia, this investigation established four regions based on historical and geographical characteristics. The regions are Central Croatia, Eastern Croatia, Coastal Croatia and the City of Zagreb. Central Croatia consists of Bjelovarsko-bilogorska, Karlovačka, Koprivničko-križevačka, Krapinsko-zagorska, Međimurska, Sisačko-moslavačka, Varaždinska and Zagrebačka counties; Eastern Croatia consists of Brodsko-posavska, Osječko-baranjska, Požeško-slavonska, Virovitičko-podravska and Vukovarsko-srijemska counties; Coastal Croatia consists of Dubrovačko-neretvanska, Istarska, Ličko-senjska, Primorsko-goranska, Splitsko-dalmatinska, Šibensko-kninska and Zadarska counties; and the City of Zagreb is both a separate region as well as a separate county (Fig. 1). All other variables are used in the same form as received from the CBS. The parental activity variable had six values: employed, not employed (seeking first or new employment), with personal income, supported person, person working abroad and other. The ages of parents variables are continuous, while all other variables are categorical.
Fig. 1. Map of Croatia showing regions and counties.
To investigate the relationship between each of the variables and the sex of a live-born child, both the χ2 test and t-test were used. Also, the stepwise multiple logistic regression model was used to assess potential confounding effects between variables.
Results
During the period from 1998 to 2008 there were a total of 466,656 live births in Croatia, of which 239,928 were males and 226,728 were females (sex ratio 1.058). The majority of the children were born in the City of Zagreb (83,608) while the fewest were born in Ličko-senjska county (4808), which corresponds to the relative number of inhabitants in these counties. Both multiple births and births with incomplete data for any of the variables (total of 46,400 births) were excluded from the analysis. The mother's mean age at delivery was 27.78 years (SD=5.25) and the corresponding father's mean age was 31.24 years (SD=5.91).
The univariate analysis of the factors affecting the sex ratio at birth reveals that the only significant factors are the region and county of birth (Table 1). The region with the highest sex ratio is the City of Zagreb (1.073), followed by Eastern Croatia (1.068) and Coastal Croatia (1.056). The region with the lowest sex ratio at birth is Central Croatia (1.048). The sex ratio at birth at the county level ranges from 1.029 (Sisačko-moslavačka county, Central Croatia) to 1.093 (Brodsko-posavska county, Eastern Croatia).
Table 1. Univariate analysis of factors affecting the sex ratio at birth
*p<0.05.
The regional factor retains its significance in the logistic regression model, where the interaction effect of a mother's age and a father's age emerges as significant as well (Table 2). As in the univariate analysis, all other factors remain insignificant.
Table 2. Results of the stepwise logistic regression of factors affecting the sex ratio at birth by sex of the live-born child
*p<0.05.
The odds ratio values suggest that Central Croatia has a significantly lower sex ratio at birth than Eastern Croatia and the City of Zagreb. The interaction of parental ages lowers the sex ratio at birth but this effect is clearly discernable only at high ages of both mother and father (Fig. 2). At the lower and middle ages of a father, the rise in a mother's age actually slightly increases the sex ratio at birth.
Fig. 2. Probability of male birth (Prob) by age of mother and father.
Discussion
The results of this study are somewhat different from previous studies in the sense that the most replicated factors affecting the sex ratio at birth, namely birth order and age of the father, were found to be not significant. This could be due to the size of the investigated sample. It seems that the effect of father's age becomes visible when the sample size is more than half a million births, while birth order needs samples over 1 million births (James & Rostron, Reference James and Rostron1985; Jacobsen et al., Reference Jacobsen, Moller and Mouritsen1999). As already mentioned, the sample used in this study was restricted for methodological and material reasons.
However, an older age of both parents seems to lower the sex ratio at birth, i.e. a higher paternal age (40 and older) produces a lower sex ratio at birth in older mothers. It seems that a higher mother's age plays a major role and the present results can be partly explained by the ‘over-ripeness ovopathy hypothesis’, and more specifically by the ‘dose–response fallacy’. It seems that although older women conceive more male offspring there is a threshold of danger for the weaker male fetuses that, when surpassed, results in more male losses in utero, which in turn leads to a lower sex ratio at birth (Jongbloet, Reference Jongbloet2004).
The significant regional variation in sex ratio at birth can be very perplexing given the size of the Croatian territory and its number of inhabitants, and the geographical, climatic and economic differences between Croatian regions. One of the most viable explanations for the negative impact on human reproductive physiology and consequently the sex ratio at birth is environmental pollution (James, Reference James2008). While serving as a good explanation for the declining trend in sex ratio at birth at the population level, it is not clear whether environmental pollution can account for the regional differences in sex ratio at birth. It is true that the region with the lowest sex ratio at birth (Central region) has a high concentration of industrial plants and corresponding air pollution, yet there are other regions in Croatia with similar industrial activity and pollution levels that do not exhibit a lower sex ratio at birth.
Some elements of lifestyle, namely nutrition, have been shown to be associated with a biased sex ratio at birth (Mathews et al., Reference Mathews, Johnson and Neil2008). The nutritional and physical inactivity patterns fit the regional sex ratio at birth picture well. Central Croatia has the highest percentage of obese and overweight women and women with an increased waist circumference (Fišter et al., Reference Fišter, Kolčić, Milanović Musić and Kern2009). Also the highest prevalence of unhealthy nutrition, defined by the intake of animal fats, the use of smoked meat products and salt, and the rare consumption of fruits, was found in Central and Eastern Croatia (Jelinic et al., Reference Jelinic, Pucarin-Cvetkovic, Nola, Senta, Milosevic and Kern2009). This might seem unusual since Eastern Croatia has a relatively high sex ratio at birth. Despite their unhealthy diet, women in Eastern Croatia have the lowest level of physical inactivity in Croatia (Milosevic et al., Reference Milosevic, Golubic, Mustajbegovic, Jelinic, Holcer and Kern2009), which makes them less obese and overweight than their counterparts from Central Croatia. It seems that the regional variation in sex ratio at birth in Croatia is best explained by the complex interaction between nutrition and physical activity patterns. The inhabitants of the continental part of Croatia (Zagreb, Central and Eastern Croatia) have traditionally a diet rich in animal fats. This could, along with reduced physical activity, lead to an increased number of obese and overweight women, and consequently to a lower sex ratio at birth.
This study has several limitations, some of which have already been mentioned. The data on gestational ages and the children born by assisted reproduction were not available, but because of the small number of such births in Croatia, this could not have affected the results of the study. Because of the constraints (especially the sample size), socioeconomic and environmental factors cannot be ruled out as potential explanations for the observed sex ratio at birth in Croatia. This calls for more detailed monitoring of these factors as well as the refinement in the methods of their reporting.
