METHODS

All registered twin births in 1994 and 1995, according to whether they were of like or unlike sex, birthweight group (< 1000 g, 1000–1499 g, 1500–1999 g, 2000–2499 g, and ≥ 2500 g), and outcome of fetal death, infant death, or survivor, were obtained from national published statistics^8,9 and from data specifically provided by the Office for National Statistics (ONS). These data comprised the denominators.

For those twins that had died, copies of the non-confidential section of the fetal and infant/child death certificates were provided by the ONS.

Information on all surviving twins was obtained from two national surveys. The first survey covered all surviving twins born in England and Wales in 1993–1995 whose co-twin suffered a fetal or infant death. The general practitioner was sent a questionnaire enquiring whether or not the child had cerebral palsy. This survey has been described previously.^5,6 The children reported here are a subset from this survey who were born in 1994 and 1995.

The second survey, the Twins Early Development Study (TEDS), covered twin pairs in which both twins survived until their first birthday. Parents of twins born in England and Wales in 1994 and 1995 were initially contacted by ONS inviting them to take part in the study. Parents who agreed to participate were sent a booklet in which they reported on the twins’ sex, birth weight, and health. Parents responded to open ended questions asking whether their twins had any medical, physical, visual, genetic, or chromosomal problems. They were asked also for general comment on their children’s health. In subsequent phases of the study, the parents continued to report on the twins’ health in a similar, open ended way. Any mention of cerebral palsy, hemiplegia, diplegia, or quadriplegia allowed the child to be included as a case of cerebral palsy.

Birthweight specific prevalence rates of cerebral palsy were determined for like-sex and unlike-sex twins in birthweight groups < 1000 g, 1000–1499 g, 1500–1999 g, 2000–2499 g, and ≥ 2500 g.

Statistical method

Crude and Mantel-Haenszel weighted relative risks with their 95% confidence intervals (CI) were determined using the EpiInfo 6 statistical package.

RESULTS

In England and Wales in 1994/1995, there were 17 200 twin maternities (34 400 registered births). Table 1 shows the number of twin births where at least one survived according to whether they were of like or unlike sex and whether the outcome of the co-twin was a fetal or infant death. Like-sex twins are at much greater risk of fetal or infant death than unlike-sex twins. Comparing like and unlike-sex twins, the relative risk for both twins to die in utero is 5.93 (95% CI 2.89 to 12.19; p < 0.0001). For one twin to die in utero and the co-twin to die in infancy, the relative risk is 2.95 (95% CI 1.33 to 6.54; p < 0.01). For both twins to be live births and both die in infancy, the relative risk is 1.80 (95% CI 1.33 to 2.45; p < 0.01).

In the first survey, there were 799 twin survivors of a co-twin fetal or infant death. (In an additional 12 cases the fetal death was registered as being of “indeterminate” sex. These were excluded from the analyses because they could not be ascribed to the like or unlike-sex twin categories). Of the 799 twin infant survivors, five from like-sex twin pairs died in childhood. For two of these, the registered cause of death was cerebral palsy. A completed questionnaire was obtained for 572 surviving children, a response rate of 71.6%.

In the second survey, there were 15 989 twin pairs where both twins survived infancy, of which 10 958 (68.5%) families returned reply cards indicating their interest in participating in the survey. These families were sent a questionnaire booklet, and 9380 completed booklets were returned, giving an overall response rate of 58.7%.

Table 2 gives the birthweight specific cerebral palsy prevalence rates of all infant survivors from twin maternities. The increasing prevalence with decreasing birth weight is clear. This confirms other reports that both low birth weight and multiple pregnancy are important risk factors in cerebral palsy.¹⁰

A breakdown of the birthweight specific cerebral palsy rates according to whether the twins were of like or unlike sex and whether or not the co-twin survived infancy shows significant differences in prevalence rates (tables 3 and 4). When the co-twin suffered a fetal or infant death, the like-sex survivor was at significantly greater risk of cerebral palsy than an unlike-sex twin (table 3; Mantel-Haenszel weighted relative risk 2.55 (95% CI 1.23 to 5.27; p = 0.01)). Among the extremely low birthweight infants, < 1000 g, cerebral palsy prevalence was marginally greater in like-sex than unlike-sex twins. In all the other birthweight groups, cerebral palsy prevalence was considerably greater in like-sex twins.

When both twins survived infancy, the birthweight specific cerebral palsy prevalence rates were significantly less than if the co-twin had died. Although like-sex twins were still at greater risk than unlike-sex twins, the difference was not significant (table 4; Mantel-Haenszel weighted relative risk 1.26 (95% CI 0.71 to 2.22).

DISCUSSION

There were important differences in the two surveys. The first survey had the aim of determining the prevalence rates of cerebral palsy in the surviving twins where the co-twin suffered a fetal or infant death. The questionnaire, which was completed by the general practitioner for more than 70% of the children, specifically inquired if the child had cerebral palsy. It is probable therefore that, in this group, there was no underascertainment of cerebral palsy among the responders.

The second survey questionnaire did not specifically inquire about cerebral palsy, was completed by the child’s parents, and the response rate was marginally under 60%. Therefore the possibility of underascertainment of cerebral palsy in this group must be entertained. The crude twin cerebral palsy prevalence among all infant survivors of 6.0 per 1000 is less than the 10.4 and 12.6 per 1000 reported from the Western Australian¹¹ and Mersey³ cerebral palsy registers respectively. This suggests that the prevalence of cerebral palsy when both twins survive infancy may have been underestimated.

A striking feature of the data reported here is the highly significant difference in birthweight specific cerebral palsy prevalence rates among surviving twins whose co-twin suffered fetal or infant death compared with twins both of whom survive (compare table 3 with table 4). This difference largely accounts for the excess prevalence of cerebral palsy in twins compared with singletons. When both twins survive, the birthweight specific cerebral palsy prevalence is slightly lower than that among singletons reported from the Mersey and Western cerebral palsy registers.^3,11 Allowing for the fact that there may have been underascertainment of cerebral palsy in the second survey, it is probable that singleton and twin cerebral palsy rates, where both twins survive, are similar.

Table 4 shows that the crude cerebral palsy prevalence is greater in like-sex twins than unlike-sex twins. However, the difference is not statistically significant. Two caveats need to be noted in interpreting this finding. The number of infants with cerebral palsy in both groups is small, and the sample size may have been insufficiently large to show a statistically significant difference, a statistical type 2 error. Also, if monozygosity is important in the development of cerebral palsy, errors in birth registration mean that twins of unlike sex are not necessarily both from a dizygotic conception. In a national study of death certificates from multiple pregnancies, there were several instances where only twins were registered although triplets or even quadruplets were recorded in early pregnancy.¹² It is still tenable that most cases of cerebral impairment manifest as cerebral palsy, even among unlike-sex twins, may be attributable to monozygosity and, specifically, to monochorionicity.