Registry based perinatal epidemiology : The importance of sibling and generation data

For more than three decades, sibship organized data from the Medical Birth Registry of Norway have been the basis for studies of variation of pregnancy outcome. Also, more recently, the registry has allowed two-generation links of birth records. These population based linked materials provide research opportunities that go beyond those based on cross-sectional collections of birth records. The agenda has been pregnancy problems like preeclampsia and other maternal complications during pregnancy, and outcome of pregnancy in terms of preterm birth, low birthweight, stillbirth, malformations and infant’s survival and morbidity, and more. More recently, studies have started considering how problems in pregnancy may predict the mother’s and the father’s long-term morbidity and mortality. For example, women die at younger ages if they have delivered preterm, if they have had a child with low birth weight, or if they have developed pre-eclampsia during pregnancy. In other materials, these observations have typically been based on studies of the woman’s first pregnancy. One problem with this first pregnancy approach is that women who stop reproducing after one pregnancy have in general higher mortality than women with two or more pregnancies (1). Much of what we study is related to complications in pregnancy. The chances that a woman has a next pregnancy are contingent on the outcome of prior pregnancies in several important ways: women with a medically complicated pregnancy may be advised or choose to avoid further pregnancies; women who are mothering children with health impairments resulting from perinatal complications may be too overwhelmed to bear another child; and underlying medical factors (such as insulin resistance or vascular dysfunction) associated with pregnancy complications may also be associated with secondary infertility. These factors add up to a strong potential of selection bias when only first births are considered. Few studies have yet considered how this selection may distort the long-term associations of reproduction with women’s later health. It seems that studies of pregnancy outcome and long term health aspects both for the child and the mother by large benefit from the reproductive data of the mother: risks related to the pregnancy or the child need to be evaluated through knowledge of the mother’s previous pregnancies, while long term risks related to the mother herself calls for her complete reproductive history. In the light of the hypothesis of mother-child (genetic) conflict, a condition that improves the child’s development, may not always be beneficial to the mother’s health. One such suggested condition is preeclampsia (2). In general, the Nordic countries have unique prospects for epidemiological research given the population based health registries, with compulsory notification, in which the records are identified by the national identification number, enabling internal and external record linkage. Sibling related data, as well as generation data, are excellent examples how we in Norway have genuine research advantages. The long lasting registration of all births provides a setting for studies of variation of longevity of mothers and fathers as a function of their complete reproductive history. Such data are almost impossible to achieve outside the Nordic countries. Thus, we have, not only a unique opportunity, but an obligation to investigate the prospects of these linked data. In fact, much of the Norwegian success in the perinatal epidemiological research is due to these data qualities. We will present examples of results that cannot be achieved using conventional cross-sectional data. The Medical Birth Registry of Norway (MBRN) is the world’s first of its kind, established in 1967. Soon covering 50 years of births in Norway, analyses that deal with time trends or long term follow-up of children and parents have an advantage. Long term health effects of being born with a specific condition, for instance being born preterm (3,4), and long term health effects for women after complications in pregnancies, like preeclampsia and kidney disease are examples of excellent research that have been well received in the most prestigious journals (5-7). Also, recurrence of specific pregnancy outcomes within sibships and between generations provide new knowledge (8-11).


UNIQUE DATA FOR REPRODUCTIVE EVENTS
For more than three decades, sibship organized data from the Medical Birth Registry of Norway have been the basis for studies of variation of pregnancy outcome.Also, more recently, the registry has allowed two-generation links of birth records.These population based linked materials provide research opportunities that go beyond those based on cross-sectional collections of birth records.The agenda has been pregnancy problems like preeclampsia and other maternal complications during pregnancy, and outcome of pregnancy in terms of preterm birth, low birthweight, stillbirth, malformations and infant's survival and morbidity, and more.More recently, studies have started considering how problems in pregnancy may predict the mother's and the father's long-term morbidity and mortality.For example, women die at younger ages if they have delivered preterm, if they have had a child with low birth weight, or if they have developed pre-eclampsia during pregnancy.In other materials, these observations have typically been based on studies of the woman's first pregnancy.One problem with this first pregnancy approach is that women who stop reproducing after one pregnancy have in general higher mortality than women with two or more pregnancies (1).
Much of what we study is related to complications in pregnancy.The chances that a woman has a next pregnancy are contingent on the outcome of prior pregnancies in several important ways: women with a medically complicated pregnancy may be advised or choose to avoid further pregnancies; women who are mothering children with health impairments resulting from perinatal complications may be too overwhelmed to bear another child; and underlying medical factors (such as insulin resistance or vascular dysfunction) associated with pregnancy complications may also be associated with secondary infertility.These factors add up to a strong potential of selection bias when only first births are considered.Few studies have yet considered how this selection may distort the long-term associations of reproduction with women's later health.
It seems that studies of pregnancy outcome and long term health aspects both for the child and the mother by large benefit from the reproductive data of the mother: risks related to the pregnancy or the child need to be evaluated through knowledge of the mother's previous pregnancies, while long term risks related to the mother herself calls for her complete reproductive history.In the light of the hypothesis of mother-child (genetic) conflict, a condition that improves the child's development, may not always be beneficial to the mother's health.One such suggested condition is preeclampsia (2).
In general, the Nordic countries have unique prospects for epidemiological research given the population based health registries, with compulsory notification, in which the records are identified by the national identification number, enabling internal and external record linkage.Sibling related data, as well as generation data, are excellent examples how we in Norway have genuine research advantages.The long lasting registration of all births provides a setting for studies of variation of longevity of mothers and fathers as a function of their complete reproductive history.Such data are almost impossible to achieve outside the Nordic countries.Thus, we have, not only a unique opportunity, but an obligation to investigate the prospects of these linked data.In fact, much of the Norwegian success in the perinatal epidemiological research is due to these data qualities.We will present examples of results that cannot be achieved using conventional cross-sectional data.
The Medical Birth Registry of Norway (MBRN) is the world's first of its kind, established in 1967.Soon covering 50 years of births in Norway, analyses that deal with time trends or long term follow-up of children and parents have an advantage.Long term health effects of being born with a specific condition, for instance being born preterm (3,4), and long term health effects for women after complications in pregnancies, like preeclampsia and kidney disease are examples of excellent research that have been well received in the most prestigious journals (5)(6)(7).Also, recurrence of specific pregnancy outcomes within sibships and between generations provide new knowledge (8)(9)(10)(11).

SELECTIVE FERTILITY
Especially Leiv Bakketeig, and his colleague Howard Hoffman from the US (NICHD, Bethesda), saw early the value of sibship data as the basis for research, and they published a series of unique papers in the period 1975-1986.Their focus was mostly on recurrence of perinatal events in successive sibs like small-forgestational age (SGA), preterm birth (<36 weeks) and low birthweight (<2500g) (12)(13)(14), but also weightspecific-perinatal mortality conditioned on outcome of previous births (15).
A study by Bakketeig and Hoffman that created a lot of attention, was published in BMJ in 1979 (16).
Based on sibship organize data, with an unorthodox analytical method, they challenged the traditional cross-sectional based J-shaped risk pattern of parity on perinatal mortality.Their results, focusing perinatal death in strata of parity (replicated in figure 1, based on MBRN: singleton births, first birth 1967-1986) demonstrated reduced mortality by increasing birth order within each of the fixed sibship size strata.Norman Mantel found the results interesting, but suggested reanalysis (17).Jean Golding called the results 'Perinatal epidemiology in Wonderland' (18), and suggested that the 'cross-sectional approach would have described the effects perfectly' and concluded that 'the so-called longitudinal approach to this problem leads up a blind alley' (19).In the figure, the trend based on crosssectional data is also shown.The special approach Bakketeig and Hoffman applied is valuable in many ways -one is that it triggers a discussion of choice of unit of analysis.Their analytical strategy illustrated the complexity of perinatal data.Not acknowledging the heterogeneous risk between women, the strong effects of selective fertility (a tendency to control fertility on the basis of previous pregnancy outcomes), and changing effects due to maternal age and birth order (20) may distort and bias results and interpretations.One easy solution, when parity issues complicate a study, is to confine to first pregnancy only, however, this strategy can lose the target since fertility in itself holds important clues.
Women control their fertility, the number of children they want, and when to be pregnant.These selective processes change by time and strongly influences perinatal health.Causes of variation in fertility are many, probably the most important being age of the mother.Also, the loss of a child will often lead to a 'replacement' pregnancy, while the birth of a survi-ving, but severely malformed child, may alter plans.Conditions like preterm preeclampsia may itself be associated with reduced fertility.Allen Wilcox has good descriptions of the selective fertility topic in his book 'Fertility and pregnancy' (21).
High age dramatically reduces fertility.However, if the mother has a loss, fertility is increased (22).Also, sex of the last child matters (23,24).After two children of the same sex, fertility is higher than after one of each; also twin pregnancy reduces fertility (23).Cesarean delivery reduces fertility, but more so if the infant survive than if stillborn or die in the perinatal period (25).Preeclampsia (7) and diabetes (26) both reduce the wish or ability for further reproduction.

PREGNANCIES AND LONG TERM SURVIVAL OF WOMEN
Norwegian studies have documented that having children prolongs life expectancy.In a series of articles, Grundy and Kravdal (1,27,28) have studied the relation between reproductive history and maternal and paternal mortality.Compared to women with two or more children, the hazard of dying in the age range 45-68 years is 1.50 (95% confidence interval (CI) 1.43-1.57)for women without children, and also elevated for women with only one child, 1.31 (95% CI 1.26-1.37)(1).These excess risks are similar, but slightly lower, for men.In the follow-up study (27), they have found that these excess risks are seen in nearly all of eleven cause-of-death categories, and they have suggested that closely spaced births may negatively influence long term health of parents (28).These important results need scrutinizing by pregnancy complications and outcome of pregnancy.
In recent years there has been much focus on preeclampsia and maternal long term survival.In a frequently cited paper, Irgens et al. (5) showed that overall mortality is elevated especially in preterm preeclampsia, with a hazard ratio 2.7 (95% CI 2.0-3.7).However, they found no elevated risk for the father, so the genetic predisposition for preeclampsia that works through the father (11,29) is not reflected in long term paternal mortality.This study, as most other studies on maternal long term health and pregnancy outcome confined to first birth, without taking women's later reproduction into consideration.This is a natural strategy to avoid using the mother more than once (30).
In another Norwegian study, it was shown that the relation between preeclampsia and maternal cardiovascular disease (CVD) interacts with fertility (7).Especially after first preterm preeclampsia, the study showed a strong interaction in effect (hazard ratio for CVD death) between preeclampsia and number of lifetime pregnancies (one versus two or more), with a 9.4 (95% CI 6.5-13.7)versus 3.7 (2.7-4.8)hazard ratio, respectively -both relative women with two or more pregnancies and no preeclampsia in first pregnancy.For term preeclampsia, the study showed that women with at least one pregnancy following the first preeclamptic had no excess risk for maternal death, 1.0 (0.9-1.1) -all causes combined.
This study is encouraging, if the effect is causal, since the majority of women with preeclampsia in first pregnancy, those who go on to have more pregnancies after the first (85% of women in Norway) is shown to have no excess early mortality linked to the preeclamptic condition.They do have a slight excess CVD mortality, but reduced risk for other causes of death, for instance death due to cancer.This is a good example of how the research based on a single pregnancy leads to misleading conclusion, while a more complete reproductive picture unveil strong heterogeneities in risk.
Unaffected sisters (or brothers) may be valuable for suggesting causal relations.The Norwegian Central Population Register (NCPR) link individuals to their mother and father, and through this resource we can identify sisters and brothers that are mothers and fathers in the birth registry.This link has been utilized in preeclampsia research.
In a recent study, Vikse et al. (31) used NCPR linked to MBRN and the Cause of Death Registry to study how preeclampsia is related to kidney disease.They showed that women with a preeclampic pregnancy had a 6-fold (hazard ratio 6.0; 95% CI 4.4-8.1)risk for end stage renal disease (ESRD), but they could not find any excess risk for ESRD in unaffected sisters (1.0; 95% CI 0.6-1.6).They concluded that 'familial aggregation of risk factors does not seem to explain increased risk for kidney disease after preeclampsia.These findings support the hypothesis that preeclampsia per se may lead to kidney damage', thus there is little support for a genetical or familial predisposition for disease as suggested for CVD by Irgens et al. (5).Not surpricingly, women with pregestational diabetes who have preeclampsia in the first pregnancy and deliver preterm are at increased risk for ESRD and early death.However, effects are found only in women who have one pregnancy, demonstrating again the importance of the sib-approach (26).

THE NEW AGENDA -'EARLY ORIGIN' AND 'LIFE COURSE' EPIDEMIOLOGY
The last 25 years have largely changed the way we think of causes of disease, and especially the cause of adult diseases.'Early origin', and 'life course epidemiology' hypotheses have increased the value of perinatal epidemiological data, and medical birth registries with long observation periods.Already in 1992 Robinson (32) suggested that the 'early origin of adult diseases' represents a shift, away from the traditional 'lifestyle' paradigm.The numerous studies, starting in the early 90s, of David Barker and his colleagues at the University of Southampton (33,34), have inspired new research worldwide.Twenty years earlier, the Norwegian Anders Forsdahl reported on the relation between poor living conditions in childhood years in Finnmark and high mortality in adulthood (35), but it was Barker who revitalized and expanded this idea.Barker himself had a flexible definition of 'early origin'; he has specified the origin to adult disease with intrauterine origin (1989), fetal (1990), infant (1990), developmental (2003), social (2007), placental (2010) and obstetric origin (2013).
Inspired by the success of the 'early origin' approach to chronic disease, life course epidemiology (36) was rapidly established as a new epidemiological discipline, especially in England.Life course epidemiology challenged the traditional aetiologic model for adult disease (i.e.stress, lack of exercise, bad nutrition) by acknowledging the work of Forsdahl and Barker: '… adult chronic disease is biologically 'programmed' during critical periods of growth and development in utero or early infancy.The extent to which these effects can be modified by later experience is a key question in life course epidemiology' (36, p. 3).Thus, the basic idea is that adult chronic disease not only has its origin in adult behavior and socioeconomic circumstances, but rather accumulates throughout life from gestation, childhood, adolescence, and later adult life.'Life course epidemiology attempts to integrate biological and social risk processes rather than draw false dichotomies between them' (37).
Especially, researchers from England have been constructive in designing life course studies, both since they have pioneered the designs, but also because they have had long tradition with structured birth cohort materials for several selected years, and through these materials been able to link and update with new exposures and outcomes, for instance the Cohort of 1946 (38) and the Cohort of 1958 (39).The 1958 cohort has updated data in 1965,1969,1974,1981,1991,1999,2004,2008,2013.The original sample was based on 17,500 babies born in England, Scotland and Wales in 1958.
The Norwegian answer to these designs is the Norwegian Mother and Child Cohort Study (MoBa).MoBa has more focus on the pregnancy and the early period of the child's life than these English follow-up studies.The success of MoBa has recently been accounted for in Norsk Epidemiologi (40).
The new paradigm increases the value of well functioning population based registries.Being the oldest birth registry, the Medical Birth Registry of Norway is well suited to be a base for 'life course' studies.In a life course perspective, a birth registry can focus either the newborn or parents.Both the early origin and the life course hypotheses are built on data that are cross sectional.As a newborn, data on siblings may be of importance to the long term health prospects.However, using a medical birth registry in the health prospects of women (and men), the total reproductive experienceall their pregnancies, as well as their own birthshould be the foundation stone of the health profile.We can enlarge on these ideas through our sibling and generational data.

The classical dilemma -nature or nurture
A problem with the early origin of adult diseases is that social and behavioral factors are inherited from one generation to the next.Also, few of Barker's initial articles had good adjustments for crucial social factors, like education and smoking during pregnancy.In a series of papers based on data from the Norwegian Mother and Child study (MoBa), utilizing one essential question 'Did your mother smoke when pregnant with you?', we have found that in utero exposure to cigarette smoke increases risks of stillbirth, fertility problems, obesity, hypertension, gestational diabetes, unfavorable plasma lipids and more (41)(42)(43)(44)(45).Smoking in pregnancy is socially inherited from mother to daughter, and may easily explain at least some of these results.A recent study has shown that a woman born with low birthweight is more often a smoker as an adult, even after adjusting for maternal education (46).Low birthweight in mothers may be due to grandmothers' smoking habits.In a fascinating study, Naess and colleagues (47) have found strong relations between CVD mortality in grandparents and grandchild birthweight, suggesting an important role of smoking during pregnancy.Thus, the social inheritance of smoking from mother to daughter explains part of the link between low birthweight and adult cardiovascular death -the increased risk of smoking in pregnancy is four fold in women with a mother who smoked during pregnancy relative other women.These Norwegian studies all show the complexity in causal inference, disentangling biological (genetic) and environmental factors.Part of the effect of early origin to adult diseases reported by Barker may be due to 'inherited' social factors.

Birth order and performance in adulthood
In another generation study we have focused on the mother's own birth order.It is well known that first births are more often growth retarded, premature or born after a preeclamptic pregnancy.Also, they are at higher risk for perinatal death than later born children.Since birthweight between generations are strongly correlated (48), it would be natural to assume that first born women would have children with lower birthweight than second or higher birth order women.We have found the opposite (49).First born women, compared to later born, have offspring with higher birthweight.Our interpretation is that the negative birth experience is overruled by the social setting of being first born.Birth order also has an effect on measured IQ, again with first born scoring highest.In a very interesting study, again based on sib data (50), Kristensen and Bjerkedal showed that if the first born dies, the next child will have higher IQ, even higher than the first born.The dynamic of being born first, second or last in a sequence of sibs holds fascinating variation.In the book 'Born to rebel', Frank Sulloway (51) has described the importance of birth order and performance in adulthood.The overall conclusion is that first born children typically are conservative and respectful of their parents, while later born are more open to experience and change, are rebellious and even more creative.

Fertility, reproduction and recurrence of events
Recurrence of pregnancy outcomes between generations is an obvious topic when generational data are available, and it holds a potential for understanding causes of 'early origin'.With data covering the years 1967-2014, 48 years, we now have almost 10 years (1967-76) with complete two-generations data, i.e. for women and men born in these early years we have records covering their complete reproduction within the follow-up years.We have, among other conditions, studied malformations, preeclampsia, preterm and postterm birth (10,11,(52)(53)(54) but also relations between preterm birth and low birthweight in the first generation and perinatal mortality in the next.
An important issue of these generational data is lack of reproduction in the next generation -women and men born with low birtweight, born preterm, or born with a malformation, have reduced fertility as adults (3,4,10,52) and this reduction seems to be higher for males than for females.

Birthweight and gestational age in 1 st generation, perinatal survival in the next
Likewise, preterm born mothers have excess risk for perinatal loss of own children, while we observe no such effect for preterm born fathers (55,56) and this effect is especially high when the mothers deliver twins (57).
In studies of variation of birthweight or gestational age for the mother's own pregnancy, and consequences for offspring perinatal survival we find strong 'doseresponse' relations, with high mortality for preterm and low birthweight born mothers.We will suggest this effect as evidence for an 'early origin' mechanism, the preterm born mother have reduced capacity to deliver a healthy baby (55,56).This is particularly observed when preterm born mothers are pregnant with twins, with 3.8 (95% CI 1.6-9.4),2.4 (1.3-4.5) and 1.9 (1.2-3.0)excess risk for extreme preterm (<32 weeks), moderate preterm (32-34 weeks) and slight preterm birth (35-36 weeks), respectively, all compared to 37-42 weeks gestation.The effect was especially high after assisted reproductive therapy (ART) pregnancy (57).
A paradoxical results is the relative importance of maternal birthweight on weight specific offspring mortality: if the child is born small, the survival is better when the mother herself also was small at birth (55).This effect is equivalent to what we find for a similar evaluation between first and second sib in maternal successive sibs (55,58,59).

Differentiating maternal and fetal causes for disease
Two generational materials on reproduction are especially valuable when we have the opportunity to compare recurrence from mother to offspring with recurrence from father to offspring.Genetic influence from the father can only be expressed through the fetus.It is natural to assume a similar fetal influence also coming from the mother.In addition, the mother's own genes and constitution have independent and strong impact on the fetus.In studies where we observe the same recurrence from the father as from the mother, it is natural to assume that only fetal genes are at work, and when there is no recurrence effect from the father we can assume that only the mothers own genes or mitochondrial inheritance are at work.

Malformations through generations
Our first generational recurrence focus was on malformations from parents to offspring (10,52), inspired by our previous sib study on recurrence from first to second pregnancy (9).Overall, we found recurrence to be stronger from father (OR=2.4;95% CI 1.7-2.8)than from mother (OR=1.6;1.2-2.0).We found strong recurrence for same type of defect, both from mother (OR=6.8;4.5-10.0)and father (OR=6.5;4.0-10.4).For other defects we found no excess risk from the mother (OR=1.0;0.7-1.4),while a significant effect (OR=1.8; 1.3-2.5)from the father.Surely, the results suggest fetal causes since the father's influence is stronger than the mother's (60).

Breech presentation and generations
Recurrence of breech presentation is surprising (61).In the literature it was difficult to find any mentioning of the rule of the father, however we found that the recurrence of breech presentation from one generations to the next to be as strong from men who themselves were delivered in breech, as from women who were delivered in breech (OR were both 2.2, with 95% confidence intervals 1.8-2.7 and 1.9-2.5, respectively).These results were for term birth only, no recurrence could be demonstrated for men or women born preterm.The interpretation is again that there is a strong familial predisposition for term breech delivery that is due to genetic inheritance, predominantly through the fetus.

Gestational age and generations
Gestational age between generations show interestingly complex relations.Population based generational data for preterm birth is a good example of 'macro-genetic' relations that can guide or target the search for specific genes.We find a clear maternal effect for preterm birth between generations, but there is no fetal effect, i.e. a preterm born father does not elevate the risk for preterm birth in the next generation (53).There is however both a maternal and fetal effect for postterm birth (54), and this is clearly linked to a similar effect within the term (37)(38)(39)(40)(41)(42) weeks (62).We especially find an effect of paternal birthweight where high birthweight reduces the gestational age in the next generation (62).The conclusion we have for gestational age confined to the term weeks, is that the fetus and the mother play equally important roles in determining time of delivery.

Preeclampsia and generations
Another example of generational effects is for preeclampsia.Fathers who were born after a pregnancy complicated by preeclampsia increased the risk of preeclampsia in the pregnancy of their partner with 50% (OR=1.5;95% CI 1.3-1.7),while the risk was increased 120% (OR=2.2;2.0-2.4) for women who themselves were born after a preeclamptic pregnancy (11).In this study we also evaluated the risk for unaffected sisters and brothers, and we found a 100% increase in risk for the sister (OR=2.0;1.7-2.3),while no effect for the brother.Thus, the interpretation is that preeclampsia is mainly caused by maternal factors, but there is also an effect that works through the fetus itself.This design was also used for recurrence of hyperemesis across generations (8).We found no evidence for fetal effects, but a relatively strong maternal effect from the affected woman to daughters as well as from older unaffected sisters to their daughters -a 3-fold increase.

New partner and interval between pregnancies
In a study on malformations and siblings, we evaluated the effect of 'new partner' and 'new municipality' (9) and found that the recurrence of same defect was strongest for women with same partner and same municipality (OR=11.6;95% CI 9.3-14.0),and weakest for women with new partner and who also had moved (OR=4.9;1.7-8.2).These differences suggest an effect of the environment.
The occurrence of preeclampsia is strongly related to parity, and cross-sectional analyses show that the occurrence is stronger in first (5-6%) than in later (1.5%) pregnancies.However, in the early 1990s several research groups observed that a new partner in second pregnancy almost eliminated this reduction, the mother was back to being 'nullipara'; and immunological mechanisms was suggested (64)(65)(66)(67).A series of studies from the Nordic countries suggested that distance between pregnancies was a strong confounder (68)(69)(70).It was shown that a new partner was associated with prolonged interval, and that interval increased the risk of preeclampsia.Adjusting for distance, the increased risk for preeclampsia in second pregnancy due to a new partner disappeared, in fact a new partner slightly reduced the risk of preeclampsia.This reduction may be due to social differences, with more smoking in women with a new partner, and indeed in a recent study from Sweden (71), further adjusting for maternal smoking demonstrated no difference in effect on preeclampsia by a new partner.
On the other hand, change of partner is in general associated with excess risk of adverse pregnancy outcome.In a study where 2 nd pregnancy outcomes for women with new partners were compared to that of women with same partner for both 1 st and 2 nd pregnancy, infant mortality, preterm birth (< 37 weeks) and low birth weight (<2500 gram) were all elevatedwith adjusted ORs corresponding to 1.8 (95% CI 1.6-2.1),2.0 (1.9-2.1) and 2.5 (2.3-2.6), respectively (72).Women with low education tend to change partner more than those with higher education.The relative number of women with new partners is increasing, both between first and second, and between second and third pregnancy.Less than 5% of women had a new partner in the second pregnancy in the period 1967-1976.This increased to almost 15% in the year 2000.A similar increase for a new partner is observed also following the second pregnancy.
Recently we evaluated the association between preeclampsia and assisted reproductive technology (ART), again using sibship data (73).A new partner indirectly increased the risk for preeclampsia since a new partner increase the distance, and this prolonged distance was associated with ART.Thus, a new partner leads to ART pregnancy and preeclampsia, both caused by a prolonged time between pregnancies.In fact, ART pregnancies, more than other pregnancies, leads to preeclampsia since these women are less likely to smoke.
Grundy and Kravdal used data from NCPR (1,27,28) to study long term survival of women and men.Lie et al. (29) also used data from NCPR, linked to MBRN, to identify brothers and sisters and their pregnancies.They showed that if a woman becomes pregnant by a man who earlier fathered a preeeclamptic pregnancy to another woman, her risk for preeclampsia is increased (OR=1.8;95% CI 1.2-2.6).They also showed that if two sisters have the same father but different mothers, the risk for recurrence for preeclampsia is again 1.8 (1.01-2.9).

Adverse pregnancy events, evaluation of the other pregnancies
We have evaluated recurrence risks for adverse outcome in successive sibling.In the study of malformations within sibships, we found that the recurrence was much higher for a similar malformation (OR=7.6;95% CI 6.5-8.8)than for dissimilar malformations (1.5; 1.3-1.7)(9).The recurrences of same defects varied largely between defects (from 5 to 45), but the recurrences were all low for dissimilar defects.Recurrence was reduced with a new partner, but even more with a new address.Much the same results are seen for other conditions like sudden infant death (SIDS) and stillbirths.Early we evaluated recurrence risk of SIDS in sibships (74), and even 12 years later we found a similar risk estimate (OR=5.8;2.1-13.2) (75).Evaluating 7 different causes of death in the 1 st year of life, no other cause was related to SIDS.For stillbirths, the recurrence risk was especially high for another stillbirth in the same gestational period.Extremely high recurrence risk was observed for early stillbirths, 20-27 weeks (OR=26; 20-33) (75,76), but also recurrence loss in weeks 28-32, and 33-36 was high, while the recurrence risk was low and not significant for a loss in the term weeks followed by a loss in early preterm.In general, the recurrence is stronger for the same than for different types of losses.
In a series of articles, we have evaluated the variation in birthweight of siblings (and other characteristics of the birth or the pregnancy) in families with one or more SIDS cases, one or more children with malformations, one or more perinatal deaths.The patterns of birthweight in the affected and the unaffected are very different depending of the condition.In sibships with malformations, the birthweights were 250-400 grams reduced relative to the same parity births in sibships without malformations, while the birthweights for the non-malformed were very close (and nonsignificantly different) to that of families without malformations for the same parity (77).This pattern was so even in families with three pregnancies where two of the fetuses had malformations.
In families with SIDS, however, there is a negative effect on birthweight even for unaffected siblings (78,79) Also in families with one or more perinatal losses, the birthweight was reduced for non-affected siblings (80).
The value of sibship data was particularly obvious in birth characteristics of non-ART pregnancies for women with one or more ART pregnancy.The conclusion of the study was that the observed variation in adverse outcome due to assisted fertilization could be attributed to factors leading to infertility rather than to the reproductive technology in itself (81).

The population average birthweight does not apply equally well to all women
Based on sibship data it was evident that the population average birthweight (around 3500 grams) is not the norm for all women.Admittedly a mother that smokes heavily in all her pregnancies will have a reduction in birthweight (200-300 grams) for all her births.However, given a relatively small child in the first pregnancy (<2500 grams), compared to a large first child (≥4500 grams), the difference in birthweight for second births for these two groups of mothers is close to 1000 grams (58), and this strong relation persists after accounting for gestational age (59).A very similar result is seen when evaluating variation in gestational age in itself (82).These differences in expected size or gestational age for a second child have consequences for risk evaluations; for instance perinatal mortality is twice for a preterm born child with an older sibling born to term (40-41 weeks), compared to an older sibling born early preterm (35-36 weeks) (82).These results led us to suggest 'customized' birthweight by gestational age standards, accounting for birthweight of previous births to the same mother (and for first births similarly utilized the mothers own birth weight) (59).Certainly, it is better to account for the birthweight of previous pregnancies than to adjust for parity.Also, it is our suggestion that previous pregnancies are better predictors for birthweight in the next pregnancy than height or weight of the mother, as used in current 'customized' standards (83,84).MBRN has since 2007 registered height and weight, so we can now evaluate the value of maternal height and weight against birthweight and gestational age of previous births as resources for customizing growth standards.

THE IMPORTANCE OF DEPENDENCIES IN SUCCESSIVE PREGNANCIES
The complicating issue of lack of independence in successive pregnancy outcomes has led researchers to study only the first pregnancy to the mother, especially when studying long term survival of the parents (5,30).This selection approach is not ideal, since it excludes around 60% of all pregnancies.Also, we and others have shown that the presence of further pregnancies may have an impact on maternal health.
For studies of outcome of pregnancy, or survival of the child, it is customary to adjust for parity.In a commentary on when to use pregnancy history in studies of reproductive failure (85), Jørn Olsen concludes that 'proper adjustment for past pregnancy history goes much further than just adjustment for parity'.Adjusting for parity cannot be a substitute for lack of sibship organized data.Many problems in perinatal epidemiology need the mother as the unit of analysis, not the single child.Even the practice to account for dependencies (for instance by use of procedures in STATA by using the mother as a 'cluster' in the analysis) is not optimal, since this strategy will camouflage the strong effects in the sib relations, also the sequence of events is lost.
Studying the next pregnancy conditional on outcomes of previous pregnancies will provide clinical useful results.As a final example we have graphed the recurrence of preterm pregnancies in successive pregnancies (1 st to 4 th ) in two independent periods (figure 3).We have previously presented similar graphs for perinatal death and preeclampsia (86,87).For preterm birth the results are remarkably consistent by period, and if anything, the recurrence risk is higher in the last period.
We have with this overview tried to convince the reader that the correlated structure in perinatal data provide huge research opportunities.The dependencies in successive pregnancies hold important clues for risk heterogeneity between women.Also, intergenerational data reveal strong influence of socioeconomic and behavioral factors, not only genetic inheritance (46,49).

Figure 2 .
Figure 2. Generational recurrence of preeclampsia: First panel: Risk of preeclampsia in own pregnancy for women born after a preeclamptic pregnancy.Second panel: Risk of preeclampsia in partner's pregnancy for men born after a preeclamptic pregnancy.Third panel: Risk of preeclampsia in pregnancy for unaffected sisters.Fourth panel: Risk of preeclampsia in partner's pregnancy for unaffected brothers.

Figure 2 (
from ref. 11) illustrates the four sets of risk scenarios.

Figure 3 .
Figure 3. Risk for preterm birth, 1st to 4th singleton birth, by outcome of previous births.Solid lines: previous preterm.Dashed lines: previous term.