Although stillbirth and perinatal mortality rates for twins have been reported by many authors, infant mortality data are scarce, particularly for zygotic twins. Infant mortality rates (IMRs) for twins have decreased in Japan since 1974 [1-3]. Recognized risk factors for twin mortality include the sex and birth order of twins [1-4], zygosity [1,4], maternal age (MA) [1,3,4], gestational age (GA), birth weight (BW) [1,3,4], race [5-7], mother’s health condition, and monthly household expenditure [8].

The present study estimated IMRs for monozygotic (MZ) and dizygotic (DZ) twins in Japan between 1995 and 2008 and identified risk factors associated with these IMRs.

Data sources

Data on twin live births (LBs) between 1995 and 2008 were obtained from national statistical records. The Statistics and Information Department, Ministry of Health, Labour and Welfare (Tokyo, Japan) maintains records covering the entire Japanese population. LB certificates include details about the nationality, sex, birth date, address, GA, BW, parental birth dates and ages, single or multiple births, and birth order in multiple births. Infant death (D) certificates contain the same information as LB certificates (excluding paternal age) as well as the date and cause of death.

Describing twin data

Infant death data were obtained for twin pairs with both LBs (Ds), one LB (D), and one LB (D)–one fetal death (FD). We estimated the number of MZ and DZ twins using the Weinberg method [9]. MA and GA were not always the same for twin pairs because each twin could be born on a different date; therefore, in some cases, the number of like- or unlike-sexed twin pairs included odd numbers of twins.

To calculate IMRs for MZ and DZ twins, denominators (total live twin pairs at birth: both LBs at birth and one LB–one FD) were obtained from (Tables 1-3) of the study by Imaizumi and Hayakawa [10]. For the purpose of our study, LB twin pairs [both LBs plus (one LB–one FD)/2] were used as the denominator. The numerators used were twin pairs of both LBs–Ds plus [one LB (S)–one LB (D)]/2 plus [one LB (D)–one FD]/2. IMRs for MZ and DZ twins are presented as 1000 twin pairs of infant deaths divided by LB twin pairs.

Table 1: Infant mortality rate in zygotic twins, 1995-2008.

Annual changes in IMRs for zygotic twins

Table 1 and Figure 1 show the annual changes in IMRs for MZ and DZ twins between 1995 and 2008. IMRs for both MZ and DZ twins decreased year by year. The linear regression coefficients of IMRs for MZ and DZ twins on the year were –0.82 and –0.49, respectively. These values are significant at the 0.1% level. By 2008, IMRs had declined to approximately 1/2 of the 1995 value for MZ twins and 1/3 for DZ twins.

Figure 1: Annual change of infant mortality rate (IMR) per 1000 deliveries for MZ twins, DZ twins, and singletons, 1995-2008.

In every year except 2007, the IMR was higher in MZ than DZ twins, and was significant at the 5% level.

IMRs in zygotic twins by MA

Table 2 shows IMRs for zygotic twins by MA during the study period. IMRs were the lowest at MAs of 35–39 years for MZ twins and 30–34 years for DZ twins. IMRs for MZ twins at MAs of 35–39 years were significantly lower than those at MAs of <20 and 20–24 years. For DZ twins, IMR at MAs of 30–34 years was significantly lower than that of other MAs, with the exception at MAs of 20–24 and ≥40 years. IMRs were also significantly higher in MZ than DZ twins for MAs between 20–24 and 35–39 years.

Table 2: lnfant mortality rate in zygotic twins by maternal age, 1995-2008.

IMRs for zygotic twins by GA

Table 3 shows IMRs for zygotic twins by GA during the study period. IMR for MZ twins was 696 at GA of <24 weeks and decreased as GA increased beyond 37 weeks (3.0). IMRs for MZ twins were similar at GA of ≥ 36 weeks (3.0–6.2). For DZ twins, IMR was 589 at GA of <24 weeks and continued to decrease as GA increased beyond 39 weeks (1.9) (except for at 38 weeks). At GAs of <29 weeks and 33–34 weeks, IMRs for MZ twins were significantly higher than those for DZ twins.

Table 3: Infant mortality rate in zygotic twins by gestational age, 1995-2008.

To provide context for the present study, we used data from the study by Imaizumi et al (Table 2) [1] to recalculate IMRs for MZ and DZ twins in 1974, finding IMRs of 47.51 (116 × 1000/2441.5) for MZ twins and 45.23 (64 × 1000/1415) for DZ twins. These IMRs were similar for MZ and DZ twins [odds ratio (OR) 1.05, 95% confidence interval (CI): 0.77–1.44] and were significantly higher than those in 1995 (MZ twins: OR 2.22, 95% CI: 1.70–2.91; DZ twins: OR 3.00, 95% CI: 2.14–4.20). In the 21 years from 1974 to 1995, the corresponding IMRs for MZ and DZ twins dramatically decreased to 21.7 and 15.6, respectively. By 2008, these rates had again markedly decreased by 1/2 and 1/3 respectively. The improved medical care may explain the declining IMRs for MZ and DZ twins in the last 34 years. From (Figure 1), the ratio of IMRs for DZ twins and singletons was 4.2 in 1995 and decreased to 2.6 in 2008 where IMR was significantly higher in DZ twins than singletons in each year. Our results also indicated that risk factors such as MA and GA influence IMRs for MZ and DZ twins.

We found that IMRs were significantly higher for MZ than DZ twins in every year except 2007. During 1995–2008, the overall IMRs were 16.3 (1056 × 1000/64990) for MZ twins and 9.1 (840 × 1000/91857) for DZ twins (Table 1). The study by Imaizumi and Hayakawa [11] reported that there were 352 infant deaths due to twinto- twin transfusion syndrome (TTTS) during 1995–2008. However, overall, there were 1056 infant MZ twin pair deaths during the same period (Table 1). This indicates that 16.7% (176/1056) of IMRs for MZ twins was explained by TTTS.

There may be some errors in the estimated infant deaths for MZ and DZ twins resulting from the available data. For example, we found that the data contained cases where there were different addresses for twin pairs due to migration between birth and death. Twenty five twin pairs with one LB (D)–one FD (unknown sex) were eliminated due to unknown MZ and DZ twins. Data from Imaizumi (Table 3) [2] and Imaizumi and Hayakawa (Table 2) [3] indicate that the total number of infant deaths (males and females) was 1365 for 1995–1998 and 2670 for 1999–2008. However, we found that there were 3792 infant deaths for individual MZ and DZ twins (Table 1), giving an estimated proportion of 94.6% [3792/(4035-25)] infant deaths for zygotic twins. Another minor underestimation might be present for 2008 as some infants surviving in 2008 may have died in 2009. As far as vital statistics is used to estimate IMRs for zygotic twins, it seems not possible to avoid an error of about 5%.

We are grateful to the staff of Statistics and Information Department, Ministry of Health, Labour and Welfare in Japan.