The effort required to identify and collect zygosity information from twins born 2001–2021 in Denmark was vastly smaller than the decades-long undertaking needed to collect similar information on the 19th century Danish twins. The present study took advantage of the Danish Civil Registration System and the high degree of digitalization of communication between public authorities and Danish citizens. Two thirds of all twin pairs responded to the questionnaire – a lower rate than in the previously collected DTR cohorts, yet still higher than most other recent Danish population studies, which have experienced declining response rates (Christensen et al. 2022). A contributing factor to the relatively high response rate in the present study was the design that provided two potential responders for each pair – either the two co-twins (for the group of twins aged 18 years and older) or the two parents (for the group of twins aged 3 to 17 years).
Like in many other (twin) studies, we observed higher response rate among females than males. However, as more than two thirds of the responses came from parents (most often the mother), the selection on the twin sex in these new cohorts in the DTR was less pronounced than typically seen when twins themselves are the responders (Lykken et al. 1987). The pairwise response rate showed very little variation between the twin birth cohorts from 2001 to 2021. For all birth cohorts, the pairwise response rate was between 60% and 70%. Encouragingly, 98% of the responding parents and 89% of the responding twins indicated their willingness to participate in future twin studies – with slightly more mothers and female twins being positive than fathers and male twins.
Zygosity classification of same-sex twin pairs was based on the same questionnaire and decision tree as used in previously collected DTR cohorts. Using genetic markers as the gold standard revealed a misclassification rate of around 5% in the new DTR cohorts with far the majority of the misclassification being MZ twin pairs classified as ssDZ twin pairs. In contrast, 99% of all twin pairs classified as MZ based on the questionnaire answers were also MZ according to the genetic markers. Similar patterns have been seen in previously collected DTR cohorts (Christiansen et al. 2003).
The marked secular variation in Danish twinning rates is only partly understood. After a relatively constant twinning rate of around 1.5% from the mid-19th century to the mid-20th century, a reduction of one third was observed from 1950 to 1970 – a reduction that could not be explained by factors known to influence the DZ twinning rate such as maternal age, parity, ethnicity, and body mass index (Rachootin et al. 1980; Westergaard et al. 1997; Basso et al. 2004a, b). The more than twofold increase in the twinning rate – from 1% to over 2% – observed from 1980 to 2000 is much easier explained. The main drivers have been the development and increased use of Medically Assisted Reproduction (MAR) and increasing maternal age at birth (Herskind et al. 2005; Pison et al. 2015). MAR included both ovarian stimulation through medication and in vitro fertilization (IVF) with typically more than one embryo transferred to the uterus at each fertilization attempt. The spontaneous twinning rate (“non-MAR twinning rate") increases approximately linearly with maternal age at birth up to age 35–39, where the rate is about double the rate for the youngest mothers. Pison et al. (2015) have estimated that in Denmark about one third of the increase in the twinning rate from 1970 to 2005 was due to increasing maternal age, whereas MAR was likely to have caused the remaining two-thirds of the increase.
In the 2000s, it was evident that twin pregnancies posed a major challenge in MAR due to the very high twin frequency in MAR pregnancies and the increased risk for complications for both mother and fetuses/children during the pregnancy and the neonatal period (Hansen et al. 2009; Monden et al. 2021). This led to a recommendation of single-embryo transfer in IVF instead of double-embryo transfer that, at the start of the new millennium, was the standard (Pinborg 2005). In the 2020s, single-embryo transfer is considered the gold standard after research has demonstrated it to be comparable in effectiveness to double-embryo transfer and superior in terms of safety and cost-effectiveness (de Neubourg et al. 2022). In accordance with these new recommendations, the Danish twinning rate has declined about one third since 2010, returning to its 19th century level of approximately 1.5%.
Beneath the secular changes in the overall twinning rates, similarly large changes in the relative rates of same-sex twins and opposite-sex twins have occurred, reflecting that secular changes in DZ twinning rates and not MZ twinning rates were the main underlying mechanism for the overall changes (Bortolus et al. 1999; Monden et al. 2021). We used Weinberg’s Differential Rule to estimate the Danish MZ twinning rate in the period 1911–2023. The validity of Weinberg’s Differential Rule has been challenged from time to time (James 1979, 1984; Kanazawa et al. 2018). However, a series of validation studies using different approaches suggest that the rule, despite its simplicity, provides a robust estimation of the MZ and ssDZ twinning rates based on the total twinning rate and the osDZ twinning rate (Husby et al. 1991; Vlietinck et al. 1988; Kanazawa et al. 2018; Eriksson et al. 1995; Fellman and Eriksson 2006).
The Weinberg estimated MZ rates in Denmark were in the range of 0.4% to 0.5% throughout the period 1911–2023 with a very modest, statistically significant increase for the period 1970–2023. This is in line with previous studies that have shown an increased frequency of MZ twinning after IVF-treatment – also when the treatment is single-embryo transfer. The mechanisms behind this increased frequency of MZ twinning in IVF-pregnancies are unknown, but specific IVF-procedures seem to play a role (Hviid et al. 2018; Kadam et al. 2023). The Weinberg estimated MZ rate and ssDZ rates were also in agreement with the relative rates of MZ and ssDZ twin pairs obtained by genetic markers in a subsample of the Danish twins born 2001–2018. The blood spots used for these genetic analyses were routinely obtained a few days after the twin births at two university hospitals in Denmark following maternal consent. Hence, the bias with an overrepresentation of MZ twins and female twins often seen in twin studies (Lykken et al. 1987) was probably not present, as the mothers’ consent to have the blood spots from the twins analyzed was unlikely to vary according to the zygosity and sex of the newborn twins. Neither did we observe an overrepresentation of MZ twin pairs in the group with questionnaire-determined zygosity in the present study. On the contrary, we found a slightly lower proportion of MZ twin pairs than expected from the Weinberg and genetic marker estimates – a finding that may be partly explained by the misclassification of MZ twins as ssDZ being the most common misclassification. Hence, MZ overrepresentation seems not to be a challenge in these new Danish 21st century twin cohorts. In agreement with the questionnaire, response rates for the twin pairs with zygosity based on genetic markers did not differ between MZ and DZ pairs.
In all three centuries, we observed higher infant mortality in twins compared to singletons with mortality before age 6 being 2–5 times higher in twins. This represented a huge absolute mortality risk for twins in the 19th century, when infant mortality was generally very high, but it also showed that there was still a substantial absolute difference as late as in the 1970s. In the 21st century, the absolute difference is more modest – in particular after the neonatal period – which reflects that the very high frequency of premature birth in twin pregnancies is the main cause of the increased infant mortality in twins compared to singletons (Roman et al. 2022).
Twins differ significantly from singletons in fetal life and in the neonatal period. They have a higher occurrence not only of prematurity but also of low birth weight and neonatal morbidity compared to singletons (Christensen and McGue 2012, 2022). This suggests that being a twin per se could be associated with certain (negative) health characteristics across the life course. A disadvantaged start to life with intrauterine growth restriction and preterm birth could leave “scars” that affect twin health throughout life. According to the hypothesis of “Fetal programming” or “Developmental Origins of Health and Disease" (DOHaD), impaired intrauterine growth increases the risk for a series of late life diseases (Barker et al.1993; Suzuki 2017), which could lead to twin-singleton differences that may impact the generalizability of results from twin studies, at least for some phenotypes. This concern has prompted a series of life-course studies of Danish twins, which took advantage of the DTR going back to the 19th century, and the possibility to link it with Danish nationwide civil and health registers (Skytthe et al. 2013).
The Danish twins born 1870 to 1900 were followed from age 6 through 1991, and their mortality rates were found to be similar to those of the background population throughout life (Christensen et al. 1995). A subsequent expansion of this study that included more than 100,000 Danish twins born 1870–1990 found no difference in mortality risks between MZ and DZ twin individuals after accounting for birth- and age-cohorts, gender differences, unknown zygosity, and the fact that twins are paired (Hjelmborg et al. 2019). Furthermore, when Nordic cohorts were included, the mortality risk of twins did not exceed that of the general population (Skytthe et al. 2019).
For cancer in general, a prospective study of more than 80,000 MZ and 120,000 ssDZ twin individuals within the Nordic population-based registers of twins followed for a median of 32 years between 1943 and 2010 found no difference in risk at any age compared to the background population (Mucci et al. 2016). For specific cancer cites, twin incidence by age, sex and zygosity has been shown not to exceed that of the general population (Skytthe et al. 2019).
Based on the DOHaD hypothesis (Suzuki 2017) it could be expected that twins had higher risk of cardiovascular diseases and diabetes. However, the 19th century Danish twins had a cardiovascular mortality risk similar to that of singletons (Christensen et al. 2001), and the prevalence of type 2 diabetes was also shown to be similar in a register study of nearly 80,000 Danish 20th century twins and a random sample of 215,000 age- and sex-matched individuals (Petersen et al. 2011). Even a study of 150 extremely birth weight-discordant young and middle-aged Danish MZ twin pairs found no indication for negative effects on glucose metabolism in the low-birth-weight twin compared to the high-birth-weight twin (Frost 2012). Danish clinical twin studies have demonstrated the same prevalence and clinical presentation in twins as reported for singletons of diseases such as Type 1 diabetes, rheumatoid arthritis, asthma, and skin diseases (Kyvik et al. 1995b; Svendsen et al. 2002; Skadhauge et al. 1999; Bryld et al. 2000).
Other studies did observe twin-singleton differences in Denmark. A decreased suicide risk among Danish twins compared to the general population has been observed, potentially because twinships provide stronger family ties that may be protective (Tomassini et al. 2003). The same mechanisms may underlie the slightly lower marriage and divorce rates observed in the mid-20th century Danish twin birth cohorts. (Petersen et al. 2011). In these cohorts, another difference was that male twins had IQ-test scores of one fifth of a standard deviation lower than those of singletons when they attended the mandatory military draft board examination at age 19 (Christensen et al. 2017). This difference seems to have vanished during the 20th century as indicated by a study of school grades in the Danish 1986–1988 birth cohorts, that showed that adolescent twins and singletons had nearly identical test scores in the ninth-grade tests (Christensen et al. 2006).
In summary, this series of studies of twin-singleton differences in health and functioning, based on data from more than a century of birth cohorts, suggests that such differences and their impact generally vanish with age and, for more recent birth cohorts, the twin-singleton differences are few and generally small after childhood. Similar observations have been made in other Scandinavian countries (Christensen and McGue 2022). It therefore seems plausible, even for traits with only minor twin-singleton differences in mean or incidence, that it is the same factors in twins and singletons that cause the observed variation. Therefore, twin-singleton differences are unlikely to cause a major challenge to the generalizability of results from twin studies.