The relationship between interpregnancy interval (IPI) and perinatal health is receiving increasing attention. A recent meta-analysis concluded that the IPI, defined as time between the last delivery and conception of the current pregnancy, can be associated with adverse birth outcomes when the IPI is either too short or too long 1 . The promotion of appropriate pregnancy spacing has been recommended to achieve better birth outcomes 2 .

Although research on the IPI has been performed in many countries, relatively few studies have been conducted in developed nations characterized by low rates of adverse birth outcomes and comprehensive health insurance such as Canada 1 . Furthermore, studies on the IPI have not accounted for residential neighborhood which has been shown to independently predict birth outcomes 3 4 5 6 7 8 9 10 11 12 . Second, most studies have excluded firstborns from analyses of the IPI, thereby precluding the opportunity to explicitly compare risks across the full spectrum of birth orders. The need to consider a full spectrum of birth orders is important in evaluating effect modification by a third variable according to which the relationship between birth order and birth outcome may vary. The influence of IPI on small for gestational age (SGA) birth varies according to race 13 , but moderation by other sociodemographic variables has not been assessed. Marital status is one influence increasingly recognized as a risk factor for adverse perinatal health outcomes 14 15 16 potentially operating though social support or stress mechanisms 17 . The influence of marital status on SGA birth, a birth outcome known to be associated with psychosocial factors such as social support 18 19 , has yet to be fully understood. Being unmarried has been reported to increase the likelihood of SGA for subsequent-born relative to firstborn infants 17 .

Given the above gaps in knowledge concerning the IPI, we sought to determine the relationship between marriage, firstborn birth and subsequent birth categorized according to IPI, and the likelihood of SGA birth, accounting for residential neighborhood cluster variations. We assessed whether effect modification was present between IPI category and marital status, adjusting for neighborhood. The setting was Montréal, a large Canadian city in which SGA birth has been shown to vary according to neighborhood 20 .

Our study confirms the results of other studies that have found marital status and IPI to be associated with adverse birth outcomes 1 2 14 15 16 28 29 30 31 , and provides additional insights on factors contributing to SGA birth. First, by including neighborhood factors as explanatory variables, we were able to demonstrate that the influence of the IPI on SGA birth varies according to features of the neighborhood. This finding is consistent with the growing literature on neighborhoods and health 3 4 5 6 7 8 9 10 11 12 32 33 .

Second, through testing for effect modification between individual predictor variables, we demonstrated that the association between IPI and SGA birth depends on maternal marital status. We also showed that the association between marital status and SGA birth varied according to IPI and maternal place of origin (Canadian-born versus foreign-born). Specifically, we found that the likelihood of SGA birth associated with being unmarried was highest for subsequent births compared to firstborns, especially for short IPIs. This association was strongest for Canadian-born mothers. Foreign-born mothers might be less susceptible to health-related consequences associated with being unmarried. We are aware of two previous studies reporting that being unmarried is a greater risk factor for adverse birth outcome in subsequent births compared to firstborns; however, these studies did not address the IPI 17 28 .

Another key finding was that the odds of SGA birth conferred by being unmarried tended to be similar to that of firstborns as the IPI increased. Different mechanisms may be involved. Perhaps the presence of young siblings (i.e., short IPI) in a household contributes extra stress to unmarried mothers, thereby negatively impacting the pregnancy environment. In the case of large age gaps between siblings (i.e., long IPI), it might be that that child rearing stresses are diminished and resemble those of unmarried mothers without children. Such a mechanism suggests that a marital partner may be important for diminishing stress associated with caring for younger children. The exact nature of such stressors (e.g., fewer stressors, better coping or adaptation) remains to be elucidated, however. Also, other unmeasured socio-economic status indicators may partly explain, or confound the observed associations. Maternal age cannot explain the associations because we adjusted for this variable. An alternative interpretation for the influence of marital status is that nutritional depletion may be present in mothers with short IPIs 34 ; such mothers may be more susceptible to any effects of being unmarried. Mothers with long IPIs may have had sufficient time to restore nutritional reserves, which may in turn help buffer any adverse effects of being unmarried. Other biological mechanisms may also link the psychosocial stress of being unmarried with the likelihood of SGA birth 18 19 , and may operate through neuroendocrine or immune pathways known to be influenced by psychological stress 35 36 .

Our study confirmed that firstborns are at greater risk of being SGA than their siblings 37 . Furthermore, our data indicate that the protective effects of being a subsequent birth are greater for infants born to married compared to unmarried women. Being married appears to augment the protective effects of a multiparous uterine environment. This finding is difficult to explain, and we suspect that marriage may serve as a proxy for other determinants of SGA birth. It is well known that unmarried mothers are more likely to have unfavorable lifestyles (e.g., smoking) associated with lower socioeconomic status. These and other unmeasured risk factors linked to being unmarried may account for or partly mediate the lesser protective effects of IPI among unmarried women.

Lastly, our study confirms the association between IPI and SGA birth 1 38 39 . Our novel finding is that this association varies depending on marital status. More specifically, intermediate IPIs were significantly more protective than long IPIs for married mothers only. Thus our results support the recommendation that mothers should avoid prolonged IPIs, but this applies, for unknown reasons, primarily to married mothers. Our data do not support the finding that short IPIs are associated with a greater risk of SGA, and this applies for both married as well as unmarried mothers. We did not evaluate extremely short IPIs in this study.

Beyond the influence of IPI on SGA birth, marital status is an especially strong predictor of this outcome. While we suggest neither a causal association nor a strict interpretation of "attributable risk", the estimated attributable fractions indicate that being unmarried (population attributable fraction = 5.3%) is a more important contributor to SGA birth than short or long IPIs (population attributable fraction = 3.2%). One study reported an attributable risk of 9.4% for short or long IPIs, but because the study was restricted to subsequent-born infants (i.e. excluded firstborns) and did not consider marriage, this estimated attributable risk cannot be directly compared with ours 34 .

Our study may be subject to several limitations. First, we used broad categorizations of marital status and IPI which may inadvertently mask underlying associations. For example, because our data did not permit finer categorization, we defined "unmarried" as not having a legal marital arrangement; however sub-groups of unmarried women such as those in stable cohabitation may be subject to different associations. Similarly, we categorized foreign-born mothers as one group when in fact differences may exist based on nationality or length of residence in Canada, but this was unavoidable because data on duration of residence is not available in the birth registry. Second, we used an administrative definition of neighborhood that may not correspond to residents' perception of neighborhood; effect estimates might differ for other neighborhood boundaries. Third, we do not have data on potential confounders such as infertility treatment which may partly account for the observed associations, although we excluded multiple births 40 . We do not know how factors such as income or alternate classifications of socio-economic status could influence our results. We also could not correct reduced precision resulting from correlation between siblings as our data do not allow the identification of siblings, although we do not suspect this effect could be substantial. Last, the extent to which our results might be generalizable to other populations is unknown. Nevertheless, these limitations are countered by a large sample size, representing all births over five years in a large Canadian city.

Our results have a number of implications for current infant health promotion practices. Present obstetric guidelines focus on promoting an appropriate IPI (i.e., intermediate IPI) to mothers contemplating subsequent pregnancies. Our results suggest that married mothers may be more likely to benefit from such recommendations than unmarried mothers. Thus, prevention strategies for unmarried mothers may well need to differ from those for married mothers.

Differential benefits to married mothers may be compounded when we consider that the IPI is associated with the social characteristics of the neighborhood (Table 3). Although no other study has yet addressed neighborhood influences on the association between IPI and birth outcomes, many studies have found neighborhoods to be important for perinatal health outcomes 3 4 5 6 7 8 9 10 11 12 . Thus, prevention strategies may need to take neighborhood factors into account.

Our results bring into question current public health recommendations in obstetrics that appropriate IPIs should be emphasized as an important intervention for newborn health for all women. Focusing on the IPI as an intervention may be differentially successful depending on the social group a mother belongs to. In fact, unmarried mothers who are most at risk of SGA birth may be the least likely to benefit from such an intervention. Marital status in particular might need to be accounted for in prevention strategies for improving birth outcomes.

CI: Confidence interval; IPI: Interpregnancy interval; OR: Odds ratio; SGA: Small-for-gestational-age.

The authors declare that they have no competing interests.

NA developed the research design, guided the data analysis, interpreted the results, and wrote the manuscript. MD, RWP, and ZCL contributed to the research design, analysis and review of the manuscript. YW performed the statistical analysis. RC contributed to conception pf the study and development of the research design. All authors have seen and approved the final version of the manuscript.