Interpregnancy interval is the duration between the birth of an earlier born sibling and the conception of the next sibling. Research suggests that deviation from an average interpregnancy interval length of 1–3 years is associated with adverse offspring outcomes. For example, both short and long interpregnancy intervals are associated with risk for the offspring to be born preterm (less than 37 weeks of gestation), to be of low birth weight (LBW; less than 2,500 g), and to be small for gestational age (SGA; greater than 2 SDs below the mean weight for gestational age).1–6
Causal, mechanistic hypotheses linking interpregnancy interval with adverse offspring outcomes have been proposed (for review see reference 7). Although these causal hypotheses exist, small sample sizes, limited control over important covariates, and skewed measurement of interpregnancy interval length (ie, birth to birth rather than birth to conception) have hampered previous research.6,8 Researchers also have suggested that much, if not all, of the association between short interpregnancy interval and adverse birth outcomes may be the result of confounding factors,6,8–11 because there is a multitude of parental demographic, physical, and mental health factors associated with both interpregnancy interval and adverse offspring outcomes.4,12,13 Determining whether associations between interpregnancy interval and adverse birth outcomes are independent of confounding factors, and thus consistent with causal claims, has important public health implications. Interpregnancy interval is a modifiable risk factor14 and, when public health recommendations are based on studies that rigorously evaluate causal claims, change in adverse outcomes can occur.1,15,16 Therefore, the current study was designed to provide a rigorous examination of the influence of confounding on the associations between interpregnancy interval and adverse birth outcomes.
We estimated the associations between interpregnancy interval and preterm birth, LBW, and SGA applying rigorous control for potential confounding by adjusting for several measured covariates and comparing differentially exposed cousins and siblings. Cousin and sibling comparison designs rule out an influence from unmeasured environmental and genetic risks that make relatives similar.17 Previous research has used sibling comparison designs to investigate these associations.9,10 We sought to replicate those findings here; we also include a cousin comparison. Cousin comparisons address residual confounding resulting from offspring birth order and maternal age while also improving the generalizability of the findings. Furthermore, we also performed a negative control analysis using the postbirth interpregnancy interval. More specifically, we used the interval to the following (next-born) sibling to predict the outcome of the prior-born sibling. Because any association with postbirth interval cannot be the result of the pregnancy-related mechanisms through which interpregnancy interval theoretically functions, it may be taken to indicate a role of family confounding.8 To apply these methodologic advances, we used a nationwide Swedish population-based sample of families to provide the largest study on this topic to date.
MATERIALS AND METHODS
The current study used prospectively collected cohort data of individuals living in Sweden from 1973 to 2009. The institutional review board at Indiana University and the Regional Ethical Review Board in Stockholm approved this study. Data for the current study were obtained by linking information available in the several government-maintained, Swedish population-based registries. We first identified offspring and their mothers using the Swedish Medical Birth Register, which provided data on more than 96% of births in Sweden since 1973.18 After identifying fathers using the Multi-Generation Register,19 we then collected information on several parental characteristics and offspring outcomes from the following registers: 1) the National Crime Register provided information on broad violent and nonviolent criminal convictions since 1973, 2) the National Patient Register provided diagnoses for psychopathologic and substance-related inpatient hospital admissions since 1973,20 3) the Education Register provided information on the highest level of completed formal education through 2009, and 4) the Migration Register and the 5) Cause of Death Register provided information important in determining the censoring information.
The initial sample included livebirth-related information for 3,403,185 individuals with valid maternal identifiers born between 1973 and 2009 and the final cohort consisted of 1,050,271 secondborn and 368,549 thirdborn offspring (Fig. 1). These individuals were born to 1,072,081 distinct biological mothers and 1,083,329 distinct biological fathers. There were 784,640 distinct maternal-side grandmothers represented in the cohort used in cousin comparison models. Cousin comparison analyses included 331,283 differentially exposed cousin pairs. Sibling comparisons included all secondborn and thirdborn siblings, although 554,652 were differentially exposed. Postbirth interval analyses only included the 368,549 individuals who had thirdborn siblings.
We defined interpregnancy interval as the number of completed months between the birth of the preceding (earlier born) offspring and the date of conception of the index (next-born) offspring. Date of conception was obtained from information on gestational age at birth estimated from last menstrual period or ultrasonogram. In sibling comparison analyses, interpregnancy interval was calculated between the firstborn and secondborn as well as between the secondborn and thirdborn offspring. Secondborn and thirdborn offspring outcomes were compared. Postbirth intervals were calculated between the birth of the secondborn and the conception of the thirdborn sibling and used to predict secondborn outcomes. Interpregnancy intervals were categorized as 0–5 months, 6–11 months, 12–17 months, 18–23 months (referent), 24–59 months, and 60 or more months to allow for comparison across previous studies.9,10
We predicted three adverse birth outcomes in the index offspring: preterm birth (less than 37 weeks of gestation), LBW (less than 2,500 g), and SGA (greater than 2 SDs below the mean weight for gestational age). Offspring with birth weights less than 300 g were excluded from our analyses.
Depending on the model, various measured covariates were included (Table 1). These included maternal and paternal age at the index birth, highest education level, nationality, and if the earlier born offspring had a different biological father. Some adjusted models also included measured lifetime parental psychopathology. In particular, we included parental criminality as indexed by any criminal conviction under the Swedish Penal code beginning at age 15 years, the Swedish age of legal responsibility; substance use problem defined as an inpatient hospitalization involving a primary or secondary diagnosis of alcohol or any other, nonnicotine substance use disorder; suicide attempt as indicated by an attempt recorded in inpatient care records as the primary or secondary reason for care; and severe mental illness as measured by an inpatient hospitalization for bipolar disorder, broadly defined schizophrenia, or other nonorganic psychotic disorders. Except for criminality, the minimum age for all parental mental health outcomes was 12 years old. All clinical diagnoses were according to International Classification of Diseases, 8th, 9th, and 10th Revisions (codes available on request). These factors have been shown to vary with both interpregnancy interval and adverse birth outcomes.21 Some adjusted models also included binary indicators of potential adverse outcomes in the firstborn including preterm birth, LBW, and SGA.
We used logistic regression analyses when predicting the secondborn, index offspring's outcomes. We emphasize effect size interpretations, but have also used a 99% CI as a result of the number of comparisons we performed. The first model, model 1, was a baseline model that adjusted only for offspring sex and year of birth. The second model, model 2, additionally adjusted for the measured covariates of maternal and paternal age, highest education, nationality, and if the fathers were different between the firstborn and secondborn. Model 3 additionally adjusted for parental psychopathology variables, including maternal and paternal criminality, attempted suicide, substance misuse, and severe mental illness. Model 4 additionally adjusted for adverse birth outcomes of the firstborn.
Models 5 and 6 were within-family comparisons. Cousins and siblings are more similar on socioeconomic characteristics, familial culture, and genetic factors (ie, cousins share on average 12.5% and siblings on average 50% of their genes by descent) than unrelated individuals.22 As such, the increased control of environmental and genetic confounding gained by the within-family designs provides an alternative to traditional methods that compare unrelated individuals. Using both cousin and sibling comparisons allows for inferences to be validated across different model assumptions and limitations (eg, birth order does not confound the cousin comparison but necessarily limits the sibling comparison). In model 5, we limited the sample to maternal cousins and used conditional logistic regression. Model 5 adjusted for all the same measured covariates included in model 4 because these may have varied between cousins. In model 6, we compared outcomes across secondborn and thirdborn siblings who differed in interpregnancy interval category while also adjusting for measured covariates that may have varied across siblings (ie, offspring sex, birth year, parity, maternal and paternal age, and different father).
Model 7 explored postbirth intervals as negative control analyses. Postbirth interpregnancy interval, the interval between the secondborn and thirdborn offspring, was used to predict the second born's outcomes. As a result of timing influences, if we found associations between the postbirth interval and the second born's birth outcomes, the association suggests that there are genetic or environmental family confounding effects that need to be considered because a postbirth interval cannot have a causal influence on the second born's outcomes.
We reestimated the models by predicting continuously measured gestational age and birth weight to examine whether results for binary outcomes generalize while also testing for nonlinearity. We also explored for cohort effects by stratifying the sample into three birth year ranges: 1) born 1973–1984, 2) 1985–1996, and 3) 1997–2009.
In the cohort (Table 1) and across baseline and adjusted analyses, pregnancies that followed short (less than 12 months) or long (greater than 24 months) interpregnancy intervals showed higher odds of adverse birth outcomes (Table 2). For example, in the baseline model 1 analyses, short (0–5 months) interpregnancy interval was associated with higher odds of preterm birth (odds ratio [OR] 1.88, 99% CI 1.74–2.02), LBW (OR 1.72, 99% CI 1.56–1.90), and SGA (OR 1.41, 99% CI 1.25–1.58). Odds of adverse birth outcomes were also higher if interpregnancy intervals were 24 months or greater as compared with the 18- to 23-month reference category. For example, interpregnancy intervals of 60 or more months were associated with higher odds of preterm birth (OR 1.75, 99% CI 1.67–1.84), LBW (OR 1.99, 99% CI 1.87–2.12), and SGA (OR 1.91, 99% CI 1.78–2.05). Incremental adjustment for measured covariates in models 2, 3, and 4 attenuated these associations, although across short and long intervals, the odds remained significantly higher for all, except for a short interpregnancy interval predicting SGA (Table 2).
In fixed-effects cousin comparisons (model 5), in which secondborn cousins with varying interpregnancy intervals were compared, the association between short interpregnancy interval (ie, 0–5 months) remained robust when predicting preterm birth (OR 1.72, 99% CI 1.26–2.35). The effect estimate remained elevated, but not statistically significant when predicting LBW (OR 1.42, 99% CI 0.96–2.10) but was fully attenuated when predicting SGA (OR 0.93, 99% CI 0.55–1.56). For cousin comparisons of long interpregnancy intervals (ie, 60 months or greater), the association was robust when predicting all three adverse birth outcomes: preterm birth (OR 1.42, 99% CI 1.04–1.66), LBW (OR 1.50, 99% CI 1.12–1.99), and SGA (OR 1.53, 99% CI 1.11–2.12).
Our sibling comparison analyses (Table 2, model 6) compared the odds of outcome between the secondborn and thirdborn offspring if their interpregnancy interval categories differed. The association between the shortest interpregnancy interval and preterm birth remained minimally elevated (OR 1.22, 99% CI 1.11–1.35). Associations between the shortest interpregnancy interval and LBW (OR 0.83, 99% CI 0.74–0.94) and SGA (OR 0.74, 99% CI 0.64–0.85), however, reversed direction, suggesting that siblings experiencing the shortest interpregnancy interval in the family were less likely to be born LBW or SGA as compared with their sibling. For long interpregnancy intervals, associations were attenuated, although they remained present for preterm birth (OR 1.18, 99% CI 1.08–1.28), LBW (OR 1.28, 99% CI 1.16–1.43), and SGA (OR 1.32, 99% CI 1.18–1.48).
Using the postbirth interval (ie, the interval between the secondborn and thirdborn offspring), we predicted the second born's outcomes. Results are presented in Table 2, model 7. Short postbirth intervals were associated with higher odds of preterm birth (OR 2.31, 99% CI 2.03–2.62), LBW (OR 2.99, 99% CI 2.58–3.47), and SGA (OR 2.14, 99% CI 1.79–2.56) in the second born. Thus, the length of interval after the birth of the secondborn offspring to the next sibling's conception significantly predicted the outcomes of the secondborn offspring, suggesting familial confounding. Long postbirth intervals, on the other hand, were not associated with preterm birth (OR 1.00, 99% CI 0.92–1.10), LBW (OR 0.98, 99% CI 0.88–1.11), or SGA (OR 0.92, 99% CI 0.92–1.18).
Models predicting continuously measured gestational age and birth weight controlling for gestational age (Appendix 1, available online at http://links.lww.com/AOG/B31) are consistent with our results predicting binary outcomes. Analyses stratified by year of birth (1] born 1973–1984, 2] 1985–1996, and 3] 1997–2009) were comparable with the main results that spanned over 35 years (Appendix 2, available online at http://links.lww.com/AOG/B31). However, adjusted analyses of short interpregnancy intervals predicting adverse birth outcomes were slightly higher for the oldest birth cohort (born 1973–1984) as compared with the younger birth cohorts.
Our study indicated that observed associations between a short interpregnancy interval and LBW and SGA were likely a result of factors that remain constant for a woman throughout her successive pregnancies (eg, prenatal care practices, genetic vulnerability). The association between short interpregnancy interval and preterm birth, on the other hand, was robust but small in magnitude.
Our short interpregnancy interval findings are in agreement with other studies that reported elevated associations with adverse birth outcomes in cross-sectional cohort data.1–6 However, with our more rigorous control of unmeasured confounding factors in sibling, cousin, and postbirth interval analyses, our interpretation is different from the causal assumptions made by previous studies. Our findings are in agreement with recent research showing that sibling comparisons attenuate the association between short interpregnancy intervals and adverse birth outcomes.9,10 Our findings are also in agreement with another paper that used postbirth interpregnancy intervals.8 In addition, one study also found that the directionality of association switches for LBW in the sibling comparison model,10 which may suggest a suppression effect.
We found a different pattern of results when examining long interpregnancy intervals. For a long interpregnancy interval, measured and unmeasured factors explain much, but not all, of the observed associations with preterm birth, LBW, and SGA, thereby suggesting an independent association of a long interpregnancy interval with an elevated risk of adverse birth outcomes. Thus, we cannot exclude a causal explanation to the observed associations, because they were robust even when controlling for genetic and environmental factors shared by cousins and siblings. It may be that unmeasured unique (ie, nonshared between siblings or cousins) confounders exist in association with long interpregnancy intervals that may still be confounding the association observed. Furthermore, the lack of an association between long a postbirth interpregnancy interval and adverse birth outcomes indicates that within-family confounding for a long interpregnancy interval is minimal, and there is no indication of an earlier born's birth outcome influencing the odds of future long interpregnancy intervals. This supports and extends previous sibling comparison studies showing elevated risk for SGA9 and LBW10 after a long interpregnancy interval by introducing results from cousin and postbirth interval analyses. A large meta-analysis1 has also previously shown increased risk for preterm birth, LBW, and SGA after a long interpregnancy interval. Although more research is needed, findings suggest that future studies and clinicians need to consider a long interpregnancy interval as an independent risk factor for adverse birth outcomes.
The limited independent association between a short interpregnancy interval and elevated odds of preterm birth may be the result of maternal nutritional depletion23 or suboptimal implantation of the placenta.24 An additional possible contributing mechanism to the independent association between a short interpregnancy interval and preterm birth could be a failure of contraction-related proteins to return to prepregnancy levels.25 Although more research is also needed on potential mechanisms linking a long interpregnancy interval and adverse birth outcomes, it has been suggested that there is a gradual decline in reproductive capacity after a birth.3 The gradual physiologic regression contributes to the parous woman presenting a similar birth outcome profile as a primigravid woman.3 Alternatively, or perhaps in conjunction, infections may contribute to fertility issues, thereby lengthening the interpregnancy interval as well as increasing adverse pregnancy outcomes for the exposed offspring.3,26 Sibling comparisons do not have the ability to control for factors that differentially influence each pregnancy and also influence the outcome such as infection. However, a sibling comparison is able to control for underlying medical conditions such as hypertension that remain consistently present in the mother in each successive pregnancy. Future research would benefit from exploring the role of breastfeeding in this complex association, because breastfeeding has been shown to elongate the interpregnancy interval and deplete maternal nutrient stores.27
From a research perspective, our study advances the field for several reasons. First, we include a sibling comparison model, a cousin comparison model, and a postbirth interval analyses, thereby making methodologic advances from previous work.9,10 These methodologic advances improve the clinical relevance of our study because several different approaches are supporting similar conclusions. With the inclusion of a cousin comparison we were able to eliminate the potential residual influence of birth order and maternal age and strengthen the external validity of the findings. Although a sibling comparison accounts for a larger extent of genetic and environmental factors than a cousin comparison, using sibling comparisons alone to examine the interpregnancy interval is problematic because of potential carryover effects among siblings, possible differences in families with more than two siblings, and potential confounding by birth order and maternal age.22 Second, we were also able to include several important individual, parent, and family structure confounds including if the father was biologically the same between pregnancies. Third, sensitivity analyses indicated that our results were not the result of bias introduced in creating binary outcomes and also not influenced by cohort effects. Although we did identify a trend appearing to show stronger associations between a short interpregnancy interval and adverse birth outcome in older cohorts, this may have been driven by a general trend of reduction in adverse birth outcomes and improved preconception care. Finally, this is the largest study to date examining the association between interpregnancy intervals and adverse birth outcomes providing more statistical power to examine these rare outcomes.
Our findings can be applied to both family planning and disease prevention efforts. For example, health care providers may use these findings to support families in making informed decisions about birth spacing. Given that we are only showing a minimal risk between a short interpregnancy interval and preterm birth, families desiring closer birth spacing with low risk for preterm birth may opt for birth spacing shorter than the traditionally recommended 2 years. On the other hand, families should be aware of the increased risk for adverse birth outcomes if birth spacing is longer than 5 years. Our findings also suggest that rather than intervening to modify the interpregnancy interval, families at risk for adverse birth outcomes may benefit from interventions aimed at changing other modifiable family-level risk factors.
Limitations, however, must also be considered. For example, various factors may influence the generalizability of our findings. As a result of the relative ethnic homogeneity of the Swedish population, future research should perform within-family analyses across ethnic and racial groups.4,13 Similarly, prenatal care is advanced and comprehensive in Sweden. This may have influenced both interpregnancy interval length and birth outcomes,16 and replication in different populations is needed. The selection of exposure-discordant pairs in cousin and sibling comparisons may also have increased bias as a result of measurement error and nonshared confounding factors.22 Unfortunately, we were unable to control for miscarriage and spontaneous or induced abortions. These factors may be especially important to consider in the long interpregnancy interval relations. Additionally, cousin (and sibling) comparisons are not randomized controlled studies; therefore, the design cannot rule out all possible confounding factors.22
In conclusion, our findings suggest that modification to increase the interpregnancy interval (ie, reducing a short interpregnancy interval) would only have a minimal independent effect on reducing the likelihood of preterm birth and no independent effect on reducing LBW or SGA. Rather than intervening to lengthen short interpregnancy intervals, at-risk families may benefit more from interventions that alter other modifiable risk factors for adverse birth outcomes. On the other hand, our findings also suggest that an unusually long interpregnancy interval is associated with a moderate independent effect of elevated odds of preterm birth, LBW, and SGA. More research into the mechanisms driving these associations is necessary to direct intervention and prevention efforts for this risk factor after replication of these findings.
1. Conde-Agudelo A, Rosas-Bermúdez A, Kafury-Goeta AC. Birth spacing and risk of adverse perinatal outcomes: a meta-analysis. JAMA 2006;295:1809–23.
2. Fuentes-Afflick E, Hessol NA. Interpregnancy interval and the risk of premature infants. Obstet Gynecol 2000;95:383–90.
3. Zhu BP, Rolfs RT, Nangle BE, Horan JM. Effect of the interval between pregnancies on perinatal outcomes. N Engl J Med 1999;340:589–94.
4. Khoshnood B, Lee KS, Wall S, Hsieh HL, Mittendorf R. Short interpregnancy intervals and the risk of adverse birth outcomes among five racial/ethnic groups in the United States. Am J Epidemiol 1998;148:798–805.
5. Klerman LV, Cliver SP, Glodenberg RL. The impact of short interpregnancy intervals on pregnancy outcomes in a low-income population. Am J Public Health 1998;88:1182–5.
6. Smith GC, Pell JP, Dobbie R. Interpregnancy interval and risk of preterm birth and neonatal death: retrospective cohort study. BMJ 2003;327:313.
7. Conde-Agudelo A, Rosas-Bermudez A, Castaño F, Norton MH. Effects of birth spacing on maternal, perinatal, infant, and child health: a systematic review of causal mechanisms. Stud Fam Plann 2012;43:93–114.
8. Erickson JD, Bjerkedal T. Interpregnancy interval. Association with birth weight, stillbirth, and neonatal death. J Epidemiol Community Health 1978;32:124–30.
9. Ball SJ, Pereira G, Jacoby P, de Klerk N, Stanley FJ. Re-evaluation of the link between interpregnancy interval and adverse birth outcomes: retrospective cohort study matching two intervals per mother. BMJ 2014;349:g4333.
10. Hanley GE, Hutcheon JA, Kinniburgh B, Lee L. Interpregnancy interval and adverse pregnancy outcomes: an analysis of successive pregnancies. Obstet Gynecol 2017;129:408–15.
11. Klebanoff MA. Interpregnancy interval and pregnancy outcomes: causal or not? Obstet Gynecol 2017;129:405–7.
12. Crittenden CP, Boris NW, Rice JC, Taylor CA, Olds DL. The role of maternal health factors, behavioral factors, and past experiences in the prediction of rapid repeat pregnancy in adolescence. J Adolesc Health 2009;44:25–32.
13. Schelar E, Franzetta K, Manlove J. Repeat teen childbearing: differences across states and by race and ethnicity. Child trends research brief. Washington, DC: Child Trends; 2007.
14. Schachar BZ, Lyell DJ. Interpregnancy interval and obstetric complications. Obstet Gynecol Surv 2012;67:584–96.
16. Teitler JO, Das D, Kruse L, Reichman NE. Prenatal care and subsequent birth intervals. Perspect Sex Health Reprod Health 2012;44:13–21.
17. Rutter M. Proceeding from observed correlation to causal inference: the use of natural experiments. Perspect Psychol Sci 2007;2:377–95.
18. Centre for Epidemiology. The Swedish medical birth register—a summary of content and quality. Stockholm (Sweden): Socialstyrelsen; 2003.
19. Statistics Sweden. Multi-generation register 2005—a description of contents and quality. Orebro (Sweden): Statistics Sweden; 2006.
21. Goldenberg RL, Culhane JF, Iams JD, Romero R. Epidemiology and causes of preterm birth. Lancet 2008;371:75–84.
22. D'Onofrio BM, Lahey B, Turkheimer E, Lichtenstein P. Critical need for family-based, quasi-experimental designs in integrating genetic and social science research. Am J Public Health 2013;103(suppl 1):S46–55.
23. Winkvist A, Rasmussen KM, Habicht JP. A new definition of maternal depletion syndrome. Am J Public Health 1992;82:691–4.
24. Getahun D, Oyelese Y, Salihu HM, Anath CV. Previous cesarean delivery and risks of placenta previa and placental abruption. Obstet Gynecol 2006;107:771–8.
25. Norwitz ER, Robinson JN, Challis JR. The control of labor. N Engl J Med 1999;341:660–6.
26. Jacob JA. Another frontier in microbiome research: preterm birth. JAMA 2015;314:1550–1.
27. Smits LJ, Essed GG. Short interpregnancy intervals and unfavorable pregnancy outcome: role of folate depletion. Lancet 2001;358:2074–7.