The associations of birth intervals with small-for-gestational-age, preterm, and neonatal and infant mortality: a meta-analysis

Providing access to family planning in low- and middle-income countries (LMIC) has social and economic benefits, but is also a critical public health intervention that may increase survival and improve health of mothers and newborns. Short and long birth intervals, or the time between previous and index live births, have been linked to adverse neonatal outcomes, including child mortality, low birthweight, preterm birth, and intrauterine growth restriction (IUGR) [1–3]. Studies have reported J-shaped risk associations, with the highest risk occurring for children born after the shortest birth intervals, then dropping to a plateau roughly around 36 months, then a gradual increase beginning around 60 months [1, 3]. Better quantifying the magnitude of these associations may provide invaluable information to estimate the possible impact of family planning interventions in reducing adverse birth outcomes, and the potential for saving newborn lives and reducing stillbirths.

Several mechanisms have been proposed linking short birth intervals with adverse pregnancy outcomes, which were recently systematically reviewed by Conde-Agudelo et al [4]. The maternal depletion syndrome (MDS) postulates that a mother may not be physiologically recovered from the previous birth if she conceives the next child shortly thereafter, leading to adverse outcomes [5, 6]. Alternately, short birth intervals may simply be an indicator of non-biological mechanisms. The sibling competition theory hypothesizes that too many children shortly spaced may place resource pressures on families. Short intervals may also be a result of the mother suboptimally breastfeeding the previous child, as proper breastfeeding delays the mother regaining fecundity. Finally, a mother may have a history of preterm births, making the short interval its product rather than preterm birth a product of the short interval.

Long birth intervals, whether intended or unintended, may also have negative outcomes, and thus important to understand the associations. A woman’s physiologic and anatomic capacity to accommodate fetal growth may revert to a nulliparous state if she has undergone a long period since her last birth, and that the infant subsequent to a long interval may experience the same risks as a first birth [7]. The long interval may also be correlated with negative outcomes if it is not a result of conscious family planning; for instance, mothers may be struggling with secondary infertility.

There are limitations to the current literature on this topic. Much of the literature on this subject utilizes cross-sectional studies such as Demographic and Health Surveys (DHS), especially with long recall periods [1–3, 8]. Causal inference is difficult to draw from cross-sectional studies, and the quality of both exposure and outcome measures may be poor in datasets that heavily depend on maternal recall. Another major drawback to synthesizing the current evidence is the substantial heterogeneity in definitions of exposures and outcomes across studies.

Thus, the aim of this work is to address some of these limitations by examining the association between birth intervals and poor neonatal outcomes (small-for-gestational-age (SGA), preterm, neonatal and infant mortality), using original data from prospective birth cohort studies conducted in LMIC, and conducting analyses using standardized categorizations and definitions of risk exposure and outcome variables. We controlled for available socioeconomic, nutritional, and reproductive health confounders in each dataset. The ultimate objective is to generate estimates to feed into the Lives Saved Tool (LiST). LiST is a computer-based tool that estimates the impact of scaling up various health interventions, such as family planning, on maternal and child mortality [9]. This work was conducted to make recommendations regarding the link connecting contraceptive use to adverse neonatal outcomes.

In our meta-analysis, short birth interval (<18 months) was significantly associated with SGA (aOR 1.51), preterm (aOR 1.58), and infant mortality (aOR 1.83). We observed a dose response relationship, with the magnitude of risk increasing as the birth intervals got shorter from the reference 36-<60 month category. Birth interval <18 months carried a substantially higher (three-fold) risk of delivering an infant who is both preterm and SGA compared to those who had a reference birth interval; preterm-SGA babies carry substantially higher risk of mortality than those born term-AGA [10].

Our findings produced a similar magnitude of associations as previous literature for short intervals with SGA and preterm outcomes, although the results cannot be directly compared due to different birth interval cut-offs and definitions. Conde-Agudelo et al.’s meta-analysis found an adjusted odds ratio of 1.26 (95% CI: 1.18-1.33) for SGA and 1.40 (95% CI 1.24, 1.58) for preterm, examining an interpregnancy interval (IPI) (period between birth and conception) of <6 months, against a reference of 18 to 23 months [1]. However, the definition of SGA accepted for inclusion in the meta-analysis was not clearly defined. In Wendt et al.’s meta-analysis [20], preterm associations had similar magnitudes also looking at an IPI of six months with a range of reference intervals, but they did not examine SGA or IUGR because of the inconsistencies in definitions across studies. Using IPI <6 month exposure and 18-<24 month reference, a study examining 173,205 children from Utah birth records (1989-1996) saw an SGA aOR of 1.3 (95% CI: 1.2-1.4) and a preterm aOR of 1.4 (95% CI: 1.3-1.5) [7]. A separate study used Michigan birth records and linked births by mother to create longitudinal cohorts; that study noted statically significant aORs with low birthweight, ranging from 1.2 to 1.5 depending on birth order of the children [21]. However, the low birthweight outcome is not directly comparable to SGA or preterm.

In our meta-analysis, short birth interval was significantly associated with increased infant mortality risk, however had no significant association with neonatal mortality risk. This finding may be driven by the smaller number of neonatal deaths, compared to infant deaths; we noticed increased risk in all datasets, but confidence intervals were wide and crossed unity in the pooled association. Incomplete neonatal mortality information in the Zimbabwe dataset may also have affected the association. Another possible explanation may be the confounding effect of breastfeeding. Those who fail to breastfeed will regain their fecundity sooner than those who do, leading to shorter birth intervals. We also expect mothers to repeat negative breastfeeding patterns for the subsequent child [22], which impacts the child’s survival in the infant period. Therefore it may not be the physiological effect of short birth intervals, but breastfeeding practices correlated with short intervals that lead to adverse infant outcomes. We did not have relevant information available to explore this hypothesis. A meta-analysis using 17 DHS datasets found neonatal and infant mortality associations with birth interval <18 months stronger than what we found in our data (neonatal: aOR 2.72, 95% CI 2.3-3.2, infant: aOR 2.84, 95% CI 2.5-3.2); however the study used cross-sectional data and a reference category of 36-47 months [2].

Long intervals do not appear to have a strong adverse association with neonatal outcomes; we only observed a statistically significant 12% increase in odds in SGA but no association with any other adverse outcomes. A meta-analysis of DHS data showed no adverse association for long birth intervals as well [2]. In contrast, Conde-Agudelo et al’s meta-analysis found 36% increased odds of SGA (95% CI: 1.20, 1.54) and 27% increased odds of preterm (95% CI: 1.17, 1.39) for IPI over 60 months (birth interval of approximately 69 months), but had a different reference category [1].

Some researchers have hypothesized that maternal depletion drives the association between short birth intervals and adverse neonatal outcomes; a mother may not have nutritionally and physiologically recovered enough before conceiving the next child. Upon controlling for available maternal nutritional variables, we witnessed no significant change in the magnitude of the associations. This may imply that nutritional depletion either plays no or a small role. A systematic review also found weak evidence to support the maternal depletion hypothesis [4], examining 15 studies that used anthropometric outcomes, maternal anemia or iron deficiency, and micronutrient deficiency as indicators of depletion. However, a separate study noted that only short birth interval children of higher birth order had a high risk of death [23]. It may be that nutritional depletion only plays a role following a cumulative effect of having multiple children or multiple short interval children. The same study also revealed fundamental background differences between mothers who completed their reproductive period with high fertility versus low fertility, and that low birth orders of high fertility mothers are worse off after a short interval than low birth orders of low fertility mothers after a short interval. Mothers who have low completed fertility may have background characteristics (i.e. better socioeconomic status) that allow them to tolerate nutritional and economic demands of short interval births, while high completed fertility mothers, who start worse off than low completed fertility mothers, may not have the capacity to handle those demands. Parity may be modifying the effect of birth intervals only when certain socioeconomic and/or nutritional conditions are present. The prospective cohort studies presented in our meta-analyses do not have information like mother’s final fertility that may serve as a proxy for effect modifiers or residual confounders that are not captured by available variables; the findings in the aforementioned study [23] suggests that we may have failed to address either some effect modification and/or confounding in our analysis.

Numerous other hypotheses on mechanisms linking birth intervals to adverse health outcomes have been identified [4]. Mechanisms with more evidence base include folate deficiency. While there is substantial evidence reporting folate deficiency following pregnancy, very few report on likelihood of folate deficiency among short birth interval mothers, and on the association between short birth interval and birth outcome among mothers who were not supplemented with folic acid. Regarding sibling competition theory, evidence implied that competition was not a major factor linking short birth intervals to neonatal mortality, but possibly for post-neonatal mortality. Other hypotheses include cervical inefficiency and vertical transmission of infections, but there is no clear evidence that supports these hypotheses.

The strength of our analysis is the use of high quality exposure, outcome, and confounder data from prospective birth cohorts. Unlike some cross-sectional survey data, the outcome information is collected soon before or around the time of birth. Also, by standardizing the categorization of birth intervals and outcomes, we were able to meta-analyze five studies with the same exposure and outcome definitions. One of the largest methodological issues with other meta-analyses is the heterogeneity of birth interval categorization, definition of SGA, use of birth-to-birth intervals versus birth-to-conception intervals (or IPI), and other exposure and/or outcome definitions.

The main weakness of this study and of almost all other studies reporting on birth interval is the inability to examine the associations taking into account the length of each component of a birth interval (birth to fecundity, fecundity to conception, and conception to birth.) Non-live birth outcomes (abortion, miscarriage, stillbirth) in between two live births may attenuate the association between short intervals and adverse outcomes by attributing more adverse outcomes to the reference or long birth interval. A study conducted in Bangladesh [24] noted differences in associations between an IPI <6 months with induced abortion, miscarriage, and stillbirth, depending on what outcome the IPI began with. The magnitude of association was highest among IPIs starting with live births for outcomes of induced abortion and miscarriage, but with stillbirths for outcome of stillbirth, although not all associations were statistically significantly different from each other. Furthermore, the first live births that we excluded from our analysis may contribute more information, as these children may have followed a pregnancy that ended in a non-live birth. The length of gestation would also affect the conception-to-birth period, and preterm outcomes could have a variety of etiologies that may not be captured through available confounding variables. We did not have complete pregnancy histories that would help us better explore these issues. Future research on birth intervals would benefit greatly from collecting appropriate data to distinguish these birth interval components and their predictors. Finally, there may have been residual confounders that were not fully captured in the available data, such as the mother’s history of preterm births and breastfeeding practices, as we only controlled for available nutritional and socioeconomic variables.

The associations we see between short birth intervals and adverse outcomes emphasize the importance of family planning interventions and the timing of the interventions. As Conde-Agudelo et al. have also stated [1], unmet need for family planning is not only a socioeconomic issue, but a public health issue for both the mother and the child. Assuming a 10% prevalence of short birth interval (<18 months) and infant mortality aOR of 1.83, lengthening the birth intervals of those individuals to ≥18 months could reduce infant mortality by 7.7%, a magnitude that is of public health significance. Furthermore, if a differential impact of short birth intervals exists by mothers’ background characteristics, unmet need poses a major health equity problem. While modern contraceptive use among women of reproductive age is ~70% in North America and Western Europe, it is only ~15% in Sub-Saharan Africa, with many countries reporting single digit figures [25]. Within countries also, there are huge equity gaps; taking Burkina Faso and Mozambique as examples, they have a close to a 30 percentage point difference in modern contraceptive use between the lowest wealth quintile (6.3% in Burkina Faso [26], 3.9% in Mozambique [27]) and the highest wealth quintile (35.5%, 34.8%). Equitable access to family planning interventions need to consciously target the most vulnerable women, as they may carry the highest health risks associated with short intervals and are also the least likely to have access to health education, contraceptives, and medical care.

Our results suggest that birth intervals under 18 months are associated with adverse neonatal outcomes, with as high as an 83% increase in odds of infant mortality. Policymakers have the responsibility to secure family planning access to all women, which would benefit the health and the economy of the population. Programs and policies also need to focus on vulnerable mothers, as they may have higher risks associated with short intervals and are least likely to have interventions reach them. Finally, more operations research needs to be conducted to determine the most effective ways to delay subsequent births in LMICs.

In LiST, we recommend the inclusion of the associations between short birth intervals and adverse outcomes, with the understanding that birth intervals may have differential impact on neonatal and infant outcomes, depending on the baseline condition of the mother.