Multivariate analysis of the Italian population

Factors affecting the sex ratio in humans: Multivariate analysis of the Italian population

Ulizzi, L


The secondary sex ratio, that is, the male/female (M/F) ratio at birth, deviates from the segregational ratio of 1:1 in almost all human populations so far examined. The magnitude of male excess differs between the main ethnic groups: The sex ratio is higher for Asian births than for white births (James 1985; Ruder 1986) and higher for white births than for black births (Beiles 1974; James 1984).

Variations in the sex ratio, however, can be found even within a population. It is well known that biological and environmental variables, such as parity, parental age, and social class, affect the sex ratio at birth (Cann and Cavalli-Sforza 1968; Rostron and James 1977; Feitosa and Krieger 1993). Nonetheless, intrapopulation variation is generally lower than that observed between ethnic groups.

The secondary sex ratio can also vary over time, and different patterns have been observed in the temporal trends of different populations. Because the length of the study period is crucial for secular trend analyses (Ulizzi and Zonta 1993), it is likely that the somewhat contradictory results reported by Gilbert and Danker (1981), Imaizumi and Murata (1981), Chahnazarian (1986), James (1987), and Feitosa and Krieger (1992) for different populations can be ascribed not only to real biological differences but also, at least in part, to the insufficient length of the study periods.

Finally, the sex ratio has been found to vary over the lifetime (Cavalli-Sforza and Bodmer 1971). Because the male mortality rate is slightly higher than the female mortality rate at most ages, the sex ratio at birth is higher than that observed later in life, for example, at 1 year of age or at reproductive age. And it is generally believed that the sex ratio at conception, that is, the primary sex ratio, is even more skewed in favor of males than is the secondary sex ratio.

It follows, therefore, that the theoretical segregational ratio of 1:1 must be affected by some modifying mechanisms, which may be of both biological and environmental origin. Such determinants may influence the sex ratio either before fertilization, by differential production or the fertilization efficiency of X- and Y-carrying spermatozoa, or after fertilization, by sex-differential mortality in utero (Astolfi and Zei 1987). Of the biological factors genetics is considered to play a primary role, especially by means of sex-linked lethal mutations, which cause intrauterine death in hemizygous males.

Although much work has been done on the determination of the sex ratio, as evidenced by the vast literature on the topic [for a review see James(1987)], the effective underlying mechanisms are still obscure.

Several researchers have studied the covariation of the secondary sex ratio and demographic variables, in particular, maternal and paternal age, parity, and parental social class. However, no general agreement has been reached on the relative roles of these variables; most of the biological and sociocultural factors studied are so strictly interrelated that it is difficult to distinguish their actual importance.

Here, we describe a partitioning of the sex ratio variability observed in the Italian population over the last two generations. This period has seen a drastic improvement in environmental conditions in the country. As a consequence, a strong reduction in perinatal and infant mortality has been observed, the extent of which is much higher in males than in females. Several biological and demographic variables, which seem to play some role in sex ratio determination, have also changed over the last 50 years, especially those related to family planning.

To isolate any specific effect, we studied the covariation over time of the sex ratio and some relevant variables, namely, stillbirth rate, maternal age, firstborn proportion, and birth order. In addition, we analyzed the relationship between the sex ratio and all the other variables simultaneously by means of a stepwise multivariate regression.

Materials and Methods

We considered all single births in Italy from 1930 to 1989. Although data on male and female births are available for a longer time interval, complete and reliable information on the other variables analyzed is retrievable for this period only.

Yearly data have been drawn from the Sommario di Statistiche Storiche dell’Italia (1861-1975) for the period 1930-1975 and from annual publications of the Istituto Centrale di Statistica (Statistiche Demografiche and Annuario Statistico Italiano) for more recent years [Istituto Centrale di Statistica 1976, 1970-1990 (various years)l.

In Italian vital statistics maternal age is given in five-year intervals.

The first group includes mothers up to 15 years of age, and the last group includes all mothers over age 50 years. The proportion of firstborns and the average birth order were computed from the annual distributions of live-borns subdivided on the basis of parity.

In accordance with current terminology, in this article we use the term sex ratio for the quantity (number of male live births)/(total number of live births). It is to be noted, however, that this proportion would be more appropriately called the sex proportion and the term sex ratio should indicate only the ratio of the number of male live births to the number of female live births (Cavalli-Sforza and Bodmer 1971).

Because the data refer to the whole Italian population, sex ratio estimates do not incur any sampling error in the common sense.

The binomial standard error, with values ranging from 9.8 x 10 sup -4 to 1.35 x 10 sup -4 , could be attributed to random sampling from the gametic pool during zygote formation.

To study the temporal trend of the sex ratio, we used three time-repeated moving averages of 3 as a smoothing approach for annual fluctuations and fitted first-, and second-, and third-degree polynomials for a description of the overall trend.

Curvilinear and stepwise multiple regression analyses were carried out on the sex ratio (as the dependent variable) against linear, quadratic, and cubic terms of stillbirth rate, mother’s age, firstborn proportion, and birth order. Arc-sine transformation had been formerly applied to sex proportion, stillbirth rate, and firstborn proportion.

Statistical analyses were performed using the software SPSS-PC+, version 4.01, and Harvard Graphics, version 2.3.

Results and Discussion

Table 1 shows the number of births and the stillbirth rate in Italy averaged over five-year intervals. (Table 1 omitted) The drastic reduction in the stillbirth rate reflects the improvements in living conditions and sanitation in Italy over the last 60 years. The effect of family planning is also evident: The total number of births has almost halved in about two generations. Although birth order has decreased almost monotonically from 3.39 to 1.76, the proportion of firstborns, which was 24% at the beginning of the period, exceeds 48% at present. Mother’s age reached its minimum of about 27 years at the beginning of the 1980s, but it seems to have been increasing in the last years.

The effects of World War II on all variables are easy to discern in both Table 1 and Figure 1. (Figure 1 omitted) As to sex ratio, Figure 1 clearly demonstrates the characteristic pattern of increase and decrease already reported by MacMahon and Pugh (1954) and Gilbert and Danker (1981) for war and postwar years in several countries.

Temporal Trend Analysis, The annual values of the sex ratio for live births are shown in Figure 1. We fitted first-, second-, and third-degree polynomials to the data. The positive linear trend, which can be inferred from the raw data, explains about 40% (p

As to the secular trend of the sex ratio, Figure 1 shows that, apart from the effects of World War II, the main feature of the pattern is a progressive approach toward and thereafter a fluctuation around the value of 0.515. This value corresponds to the M/F value of 1.06, which seems to be characteristic of whites (Cavalli-Sforza and Bodmer 1971; Beiles 1974; Gilbert and Danker 1981; James 1985, 1987; Ruder 1986; Ulizzi and Zonta 1993, 1994). It is worth noting that even the earliest data on secondary sex ratio, which refer to births in London between 1628 and 1710 [reported by Beiles (1974)], give an average value of 1.06. A previous study of the secular trend of the sex ratio in different ethnic groups (Ulizzi and Zonta 1994) revealed that no relevant change with time is generally found when the period of examination is sufficiently long. In two white populations, namely, Italians and US whites, the secondary sex ratio seems to have fluctuated around 1.06 over approximately the last 100 years.

In the same populations we compared the temporal variation of the sex ratio at birth with that at 1 year of age. We found that the two groups share a similar pattern: Differences in sex ratios calculated at birth and at 1 year of age have been reducing over recent years. In the Italians, who have been extensively studied in previous works (Ulizzi 1983; Ulizzi and Zonta 1993), we demonstrated that the sex ratio at birth tends to be maintained up to reproductive age. It can be hypothesized, therefore, that if the reproductive value is stabilized by natural selection, as is likely, new relationships between the sex ratio and selective forces may be expected to develop in human populations.

Multivariate Analysis. Before the multivariate analysis we performed curvilinear regression analyses of the third degree to identify the type of dependence of the sex ratio on stillbirth rate, mother’s age, birth order, and firstborn proportion. On the basis of the results, we chose the following model for a stepwise multiple regression to fit the data:

(formula omitted)

where SR is the sex ratio, ST is the stillbirth rate, MA is mother’s age, FP is the firstborn proportion, BO is the birth order, and YR is the year. We had previously coded individual mother’s ages as deviations from their mean to reduce the correlation between linear and quadratic terms.

The correlation coefficients of the above variables with the sex ratio are ST, -0.664; MA, -0.490; (MA) sup 2 , 0.422; FP, 0.633; BO, -0.599; and (YR) sup 3 , 0.647.

As expected, the intercorrelations of these variables, with the exception of (MA) sup 2 , are high, and their correlation coefficients vary in absolute value from 0.82 to 0.98.

Table 2 shows the results of the stepwise analysis. It can be seen that, although the stillbirth rate has the highest correlation coefficient with the sex ratio and by itself accounts for 44. 1% of the total variance, it loses its importance when the firstborn proportion enters the analysis, as shown by the t value of the regression coefficient in Eq. (a) in Table 2. (table 2 omitted) After removal of the stillbirth rate and the addition of the mother’s age linear term, 65.0% of the total variance is explained simply by the firstborn proportion and mother’s age (linear and quadratic terms), according to Eq. (b). Neither birth order nor time ever entered the analysis.

When the effect of each of the three variables on the sex ratio is singled out, the correlation coefficients of the firstborn proportion and the mother’s age quadratic term increase (from 0.633 to 0.654 and from 0.422 to 0.618, respectively), whereas that of the mother’s age linear term decreases (from -0.490 to -0.406).

In Figure 2 the expected values computed with Eq. (b) in Table 2 are graphically compared with the sex ratios observed for the 60-year period. (Figure 2 omitted)

The results of multivariate analysis show that a quadratic function of the firstborn proportion and mother’s age can be a fairly good predictor of sex ratio values. It is likely, therefore, that the two variables play a main role in the determination of the sex ratio. This finding is in agreement with the results of Cann and Cavalli-Sforza (1968), Rostron and James (1977), Astolfi and Zei (1987), and Feitosa and Krieger (1993). Although some published observations have failed to show a significant association with either maternal age or birth order, this lack of association is often attributed to different analytical techniques, sample size, or other aspects of the sampling procedure (James 1987).

It is to be stressed, however, that the interpretation of the still unexplained variation observed in our population requires that further sources of variability, both biological and environmental, be taken into account.

1. Dipartimento di Genetica e Biologia Molecolare, Universita “La Sapienza,” Piazzale A. Moro 5, 00195 Rome, Italy.

2. Dipartimento di Genetica e Microbiologia, Universita di Pavia, V. Abbiategrasso 207, 27100 Pavia, Italy.

Acknowledgments We wish to thank S. Zanoli for her technical help. This research was supported by the Ministero dell’Universita e della Ricerca Scientifica (MURST).

Received 22 April 1994; revision received 7 July 1994.

Literature Cited

Astolfi, P., and G. Zei. 1987. Variation of the human secondary sex ratio and lethal recessive X-linked mutations. J. Biosoc. Sci. 19:283-294.

Beiles, A. 1974. A buffered interaction between sex ratio, age difference at marriage, and population growth in humans and their significance for sex ratio evolution. Heredity 33:265-278.

Cann, H.M., and L.L. Cavalli-Sforza. 1968. Effects of grandparental age, birth order, and geographic variation on sex ratio of live-born and stillborn infants. Am. J. Hum. Genet. 20:381-91.

Cavalli-Sforza, L.L., and W.F. Bodmer. 1971. The Genetics of Human Populations. San Francisco. CA: Freeman.

Chahnazarian, A. 1986. Determinants of the sex ratio at birth. Ann Arbor, MI: University Microfilms International.

Feitosa, M.F., and H. Krieger. 1992. Demography of the human sex ratio in some Latin American countries, 1967-1986. Hum. Biol. 64:523-530.

Feitosa, M.F., and H. Krieger. 1993. Some factors affecting the secondary sex ratio in a Latin American sample. Hum. Biol. 65:273-278.

Gilbert, K., and H. Danker. 1981. Investigations on the changes of sex ratio in Germany from 1826 up to 1978. Acta Anthropogenet. 5:89-110.

Imaizumi, Y., and M. Murata. 1981. The changing sex ratio in Japan. Jpn. J. Hum. Genet. 26:71-81.

Istituto Centrale di Statistica. 1976. Sommario di Statistiche Storiche dell’Italia (1861-1975). Rome, Italy: Istituto Centrale di Statistica.

Istituto Centrale di Statistica. Various years, 1970-1990. Annuario Statistico Italiano. Rome, Italy: Istituto Centrale di Statistica.

Istituto Centrale di Statistica. Various years, 1970-1990. Statistiche Demografiche. Rome, Italy: Istituto Centrale di Statistica.

James, W.H. 1984. The sex ratio of black births. Ann. Hum. Biol. 11:9-44.

James, W.H. 1985. The sex ratio of Oriental births. Ann. Hum. Biol. 12:485-487.

James, W.H. 1987. The human sex ratio. I. A review of the literature. Hum. Biol. 59:721-752.

MacMahon, B., and T.F. Pugh. 1954. Sex ratio of white births in the United States during the Second World War. Am. J. Hum. Genet. 6:284-292.

Rostron, J., and W.H. James. 1977. Maternal age, parity, social class, and sex ratio. Ann. Hum. Genet. 41:205-217.

Ruder, A. 1986. Paternal factors affect the human secondary sex ratio in interracial births. Hum. Biol. 58:357-366.

Ulizzi, L. 1983. Relationship between mortality and sex ratio: Changes over the last century in Italy. Ann. Hum. Genet. 47:321-327.

Ulizzi, L., and L.A. Zonta. 1994. Sex ratio and natural selection in humans: A comparative analysis of two Caucasian populations. Ann. Hum. Genet. 57:211-219.

Ulizzi, L., and L.A. Zonta. 1994. Sex ratio and selection by early mortality in humans: Fifty-year analysis in different ethnic groups. Hum. Biol. 66(6):1037-1048.

Copyright Wayne State University Press Feb 1995

Provided by ProQuest Information and Learning Company. All rights Reserved