9.1 The trait.—The definition of a stillbirth employed in this study is an infant who shows no signs of life at birth, following a period of gestation of at least 20 weeks. Any infant that moves, makes respiratory attempts, or is born with a beating heart is not in the strict sense “stillborn.” This definition, which is that employed in United States vital statistics, runs counter to current Japanese usage, under which infants who showed signs of life at birth but never succeeded in establishing a pattern of regular respiration are frequently reported as stillborn. This usage was in part a matter of convenience, since a stillbirth required completing only one official form, whereas a livebirth dying an hour following delivery required completing two. The subject of the ABCC's usage of the term “stillbirth” was repeatedly taken up at conferences with the midwives.

The frequency of stillbirths encountered in our control material (both parents category 1) was in Hiroshima 1.27 per cent of all births (17,189 births) and in Nagasaki 1.31 per cent (14,450 births). In comparison, in 1950 the “official” stillbirth rate (stillbirths per 1,000 livebirths) was in Hiroshima prefecture 79.6 and in Nagasaki prefecture 87.7, or, expressed as percentage of all births, 7.37 per cent and 8.07 per cent respectively (Public Health and Welfare in Japan, 1950). Official statistics for other years are similar. There is apparent a marked discrepancy between official statistics and our own. There are at least five factors to be taken into account in an evaluation of these differences. Firstly, there is the matter of registration practices referred to above. Secondly, the official stillbirth data are for stillbirths after the third month of uterogestation, whereas by and large our figures refer to events following the fifth month of gestation. Thirdly, since the official figures are for an entire prefecture, there is the possibility of urban-rural differentials. This, however, works in the opposite direction than the first two factors, since in 1950 the stillbirth rate for all shi (cities) was 134.0 per 1,000 livebirths, whereas for all gun (villages and rural areas) the rate was 70.4. Fourthly, the official figures may include some induced terminations, whereas these were excluded from our own data. Lastly, those stillborn infants with gross malformations were excluded from our figures. The reasons for this exclusion were given in Section 6.2.

When approximate allowance is made for the exclusion from our figures of stillborn, malformed infants, the stillbirth rates observed in this study are quite similar to those currently obtaining in many parts of the United States (e.g., Michigan, 1950, 1.9 per cent). The possibility that the ABCC program was failing to obtain data on any substantial number of stillbirths seems precluded by the efficiency of pregnancy registration (Sec. 2.1) and the system of careful follow-up for all registered terminations.

9.2 The genetic argument for radiation-induced changes.—Animal experimentation on the genetic effects of irradiation indicates that the largest group of induced mutations having clear-cut effects detectable in the first generation of offspring following exposure are the autosomal lethals. As detected in the first generation of offspring these mutations would consist largely of the dominant lethals, although a few recessive lethals would be recovered due to the fortuitous combination in some individuals of an induced lethal mutant with an allelic lethal mutant of spontaneous origin. Presumably one of the manifestations of the presence of these mutant genes would be fetal death. Accordingly, we might expect, as one of the changes symptomatic of irradiation-induced genetic damage, an increase in the frequency with which infants are stillborn as parental exposure increases.

9.3 Concomitant variation known to affect the stillbirth rate.—No one of the indicators with which we have dealt or shall deal is more complex with respect to concomitant variation than the one presently under consideration. No less than seven sources of concomitant variation known to affect the stillbirth rate could, conceivably, be influencing the data which we shall analyze. These seven sources of extraneous variation are congenital syphilis, birth injury, maternal age, parity, nutrition, rate of reproduction, and paternal age.

Congenital syphilis has long been one of the major causes of fetal mortality. In Hiroshima and Nagasaki the frequency of positive serologies is from 5 to 7 per cent among post-parturient women. To assess the amount of infantile mortality attributable to congenital syphilis among the births described here would have required an expenditure of time and energy which, because of the limitations of personnel and facilities, was not possible. However, a small study was made in Nagasaki on infants born in the year 1951 in an attempt to determine the incidence of congenital syphilis among the infants being examined in connection with the genetics program (Wright, S.W. et al., 1952). Initially it was planned that this study would estimate the incidence of congenital syphilis among living infants, stillbirths, and neonatal deaths. Because of the formidable difficulties posed in estimating the latter incidences in the absence of an exhaustive post-mortem program in Nagasaki, attention had to be focused on living infants alone. Two items of interest here emerged from this study. Firstly, some seven out of every thousand liveborn infants were shown to have congenital syphilis either by clinical, serologic, or roentgenologic examination, or any combination thereof. Secondly, a gradual decline in the rate of transmission with increasing maternal age was noted. These latter data are reproduced in Table 9.1. If this same effect of maternal age is exhibited by syphilitic stillborn infants (and we have no evidence that it is not) then, because of the known differential in maternal age among the parental exposure cells, herein lies a possible source of bias. The age distribution among the exposure cells is such that this bias would tend to inflate the stillbirth rate among those parents less heavily exposed or not exposed at all. However, these potential effects would appear to be quite small, of a magnitude which could safely be ignored in this analysis.

TABLE 9.1

CONGENITAL SYPHILISAMONG LIVING INFANTS BORNIN NAGASAKIIN 1951 (Incidence and relation to maternal age [after Wright, S.W. et al., 1952]).

A second major factor affecting infant survival (which it was not possible to control) is the occurrence of an injury during parturition. In the United States, it has been estimated that birth injury accounts for no less than 13.6 per cent of infantile mortality (Stander, 1941). Of the various types of injuries which may be sustained during the passage of the infant through the birth canal, by far the most common is intra-cranial injury with hemorrhage. It has been reported that 40–80 per cent of infants coming to autopsy will show evidence of this form of birth injury (Stander, 1941). In our own data among 50 autopsied infants, selected at random, 26 showed intra-cranial hemorrhages ranging from moderate to severe in degree. While the importance of birth injury in infantile death cannot be gainsaid, it is difficult to visualize circumstances which would lead to a non-random distribution of birth injuries among the parental exposure categories, and this possibility will be ignored in the analysis.

A third major factor affecting infant survival is maternal nutrition. The importance of this variable is amply attested to by an extensive literature. In the main, this literature is in agreement in indicating that rather drastic changes in maternal nutrition must occur before clearly demonstrable changes in the stillbirth rate can be shown (see, for example, Antonov, 1947, and Smith, 1947). There is no evidence that during the course of this study the limitations imposed on maternal nutrition by economic conditions prevailing in Japan reached the levels necessary to produce measurable changes in the stillbirth rate, or that (arguing again from the rough classification of economic status presented in Sec. 5.3) such differences in maternal nutrition as existed were non-randomly distributed.

Among the variables affecting the stillbirth rate previously enumerated, the effects of maternal age and parity are by far the most complex and general in scope. These effects as manifested in the present material are summarized in Tables 9.2 and 9.3 and Figures 9.1 and 9.2. The data are presented and analyzed in the same manner as age and parity effects with respect to major congenital malformations (cf. Sec. 8.4). Again, a detailed discussion of the findings will be deferred for a subsequent publication. There is a significant effect on stillbirth frequency of both age and parity, with, in both instances, an apparent increase in stillbirth frequency at the extremes of the scale. This in crease at the lower end of the scale is more definite with respect to parity than with respect to age.

TABLE 9.2

THE EFFECT OF MATERNAL AGE AT FIXED PARITY ON THE FREQUENCY OF STILLBORN INFANTS (The meaning of the symbols employed is explained in Sec. 8.4.).

TABLE 9.3

THE EFFECT OF MATERNAL PARITY AT FIXED AGE ON THE FREQUENCY OF STILLBORN INFANTS (The meaning of the symbols employed is explained in Sec. 8.4.).

FIGURE 9.1

The distribution of the frequency of stillborn infants by age of mother for specified parities.

FIGURE 9.2

The distribution of the frequency of stillborn infants by parity for specified maternal ages.

One of the more exhaustive previous studies of this relationship of maternal age and parity to infant survival is that of Yerushalmy (1945), who found that (1) “the lowest rates do not occur in one particular age group irrespective of parity. There is a consistent shifting of the minimum rate to the older ages with increasing parity.” (2) “When the stillbirth rates by age of mother are compared for the different parity groups the increase in the rate is not proportionate for the various age groups. The increase is very much higher for the younger age of mother groups than for the older.” The present data appear to agree with the first of these two conclusions, but of themselves would scarcely permit drawing the second conclusion—this may be in part a matter of the numbers involved. From the standpoint of controlling extraneous variation, these findings, taken in conjunction with our own, imply that the only really adequate control is one which takes into account not only maternal age but also parity. Among the Japanese data, to effect an adequate control would require recognizing approximately 25 parity-specific age intervals (five intervals for parity, and five for age, or 25 in all). The data, because of the already existing numerous ways of classification, would not permit so elegant an approach to the extraneous variation occasioned by mother's age and parity. Accordingly, it became a matter of judgment as to which of these two factors was to be controlled. A decision was reached to disregard age but to recognize, for analytical purposes, five parity classes, namely, parity 1, parities 2–3, 4–5, 6–7, and 8 and higher.

9.4 The data.—Tables 9.4 and 9.5 present the distribution and analysis of stillborn infants by sex, city, and parental exposure without further subdivision by parity. Inspection of Table 9.4 suggests a difference in the stillbirth rate when the mother was not exposed as opposed to when she was. From the analytical table (9.5a) we note a significant effect of mother's exposure but not of sex or father's exposure. Attention is called to the fact that the sex-mother's exposure interaction is not significant, in contrast to expectation in view of the possible effect of exposure on the sex ratio described in Chapter VII. With regard to the reality of the effect of mother's exposure on the stillbirth rate, it should be borne in mind that these data are uncorrected for parity and the more heavily exposed cells are biased in a way which would increase the stillbirth rate in these cells. If in the analysis of Table 9.4 we limit our attention to those exposure cells in which both parents were exposed, which would reduce the differences between cells with regard to parity, the apparent effect of maternal exposure disappears (see Table 9.5b). This observation is of considerable importance in the interpretation of these data since (a) maternal exposure confounds the effects of genetic and somatic damage and hence a significant effect of maternal exposure not reflected in paternal exposure might well be a somatic rather than a genetic effect, and (b) there is evidence which has been presented in Chapter V and elsewhere in this report that the comparison of pregnancy terminations occurring to exposure category 1 parents with terminations occurring to parents in any other exposure category is biased.

TABLE 9.4

FREQUENCY OF STILLBIRTHS BY SEX, PARENTAL EXPOSUREAND CITY (Unrelated parents).

TABLE 9.5

CHI-SQUARE ANALYSIS OF THE FREQUENCY OF STILLBIRTHS BY SEX, PARENTAL EXPOSURE, AND CITY (Unrelated parents).

Let us now turn to a consideration of what these data reveal when the sex of the infant is ignored but terminations are classified with respect to parity. Tables 9.6 and 9.7 present the distribution and analysis of stillborn infants by city and parental exposure with further subdivision by parity. It should be noted that, as expected, there are significant differences in the stillbirth rate among different parities. From Table 9.7 we further note that when all exposure cells are considered with reference to parity, there is no effect of city or of maternal exposure. However, an effect of paternal exposure significant at the 5 per cent level now appears. Inexplicably this latter effect seems to be largely, if not entirely, attributable to those fathers whose infants were firstborn. The frequency of stillbirths among firstborn infants by city and paternal exposure is presented in Table 9.8. The data suggest a regular increase in stillbirth frequencies with increasing paternal exposure. To the end of determining whether this apparent effect was explicable in terms of the residual variation in maternal age not removed by the parity classification, Table 9.9 was prepared. An examination of this table reveals a number of points of interest. Of the five age groups investigated, the stillbirth rate increases with paternal exposure in three and decreases in two. In one of the three age groups (21–25) the increase with paternal exposure primarily reflects a “1” vs. “non-1” difference. In substance, then, when maternal age as well as parity is taken into account, there emerges no clearly consistent relationship between paternal exposure and the stillbirth rate.

TABLE 9.6

THE FREQUENCY OF STILLBIRTHS BY PARENTAL EXPOSURE, CITY, AND PARITY (Unrelated parents).

TABLE 9.7

CHI-SQUARE ANALYSIS OF THE FREQUENCY OF STILLBIRTHS BY PARENTAL EXPOSURE, CITY, AND PARITY (Unrelated parents).

TABLE 9.8

THE FREQUENCY OF STILLBIRTHSAMONG FIRSTBORN INFANTS BY CITY AND PATERNAL EXPOSURE (Unrelated parents).

TABLE 9.9

THE FREQUENCY OF STILLBIRTHS AMONG FIRSTBORN INFANTSBY PATERNAL EXPOSURE AND MATERNAL AGE (Unrelated parents).

9.5 Summary.—Analysis of the stillbirth data, taking into account differences between exposure cells in the parity distributions, fails to reveal significant differences between cities or consistent significant effects of parental exposure. The deviations from expectation observed are so small and inconsistent that they can scarcely be labeled as favoring or not favoring genetic hypothesis.