Abstract
Dear Editor
In their metaanalysis CondeAgudelo et al (2016) conclude that short interpregnancy intervals (IPIs) are associated with a significantly increased risk of autism spectrum disorder (ASD). We wish to point out that none of the studies in the metaanalysis controlled for the birth order of the index child. In a recent paper (Beenstock, Levine and Raz 2015) we showed that recurrence risk of ASD varies inversely with birth order, as well as with birthspacing. Birth order matters because the more parents succeed in having typically developed children, the less likely it is that they have a genetic predisposition to having children with ASD. Therefore, given everything else, if the index child is a firstborn the incidence of ASD is expected to be larger than for index children who are secondborns, and so on for higher birth orders.
If birthspacing is unrelated to birth order, the omission of birth order effects would not matter for estimates of birth gap effects from linear regressions. Matters are different with nonlinear oddsratio methods used by CondeAgudelo et al where the estimated effect of birthspacing is biased in an unknown direction. If birthspacing and birth order are positively correlated, the estimated birthspacing effect is positively biased in linear regression, and it is almost surely positively biased in oddsratio methods. If birthspacing and birth order are negatively correlated, the bias is negative, which means that the estimated negative effect of birthspacing on ASD risk in the metaanalysis might be a spurious result induced by the failure to specify the birth orders of the index children.
If parents have targets for family size, birthspacing is expected to be negatively correlated with birth order (Heckman, Hotz and Walker 1985). Given everything else, if child 2 was slower to arrive, parents who desire three or more children may try to make up for lost time in producing their third child. Heckman et al used data for Swedish women to establish this negative correlation empirically. Target family size theory inevitably casts doubt on the robustness of the main result in CondeAgudelo et al (2016).
Using population cohort data for 4976 index children in Israel with ASD, Beenstock et al (2015) found that recurrence risk is 40% larger among younger siblings born within at least two years of their index sibling. Also, if the index is a firstborn and the second child is typically developed, recurrence risk for the third child falls by 0.5% (9.1% in terms of relative risk).
In our study index children have ASD whereas in CondeAgudelo et al they do not. Therefore, the fact that we found birth spacing is statistically significant after controlling for birth order does not necessarily mean that the same result would be found for the general population. Future research on risk factors of neurodevelopmental disorders should aspire to specify birthspacing and birth order among the covariates.
Yours sincerely
References:
Beenstock M, Levine H, Raz R (2015) Birth Gap and the Recurrence Risk of Autism Spectrum Disorders: A Populationbased Cohort Study. Research in Autism Developmental Disorders, 17: 8694.
CondeAgudelo A., RosasBermudez A., Norton MH (2016) Birth spacing and risk of autism and other neurodevelopmental disabilities.
In their metaanalysis CondeAgudelo et al (2016) conclude that short interpregnancy intervals (IPIs) are associated with a significantly increased risk of autism spectrum disorder (ASD). We wish to point out that none of the studies in the metaanalysis controlled for the birth order of the index child. In a recent paper (Beenstock, Levine and Raz 2015) we showed that recurrence risk of ASD varies inversely with birth order, as well as with birthspacing. Birth order matters because the more parents succeed in having typically developed children, the less likely it is that they have a genetic predisposition to having children with ASD. Therefore, given everything else, if the index child is a firstborn the incidence of ASD is expected to be larger than for index children who are secondborns, and so on for higher birth orders.
If birthspacing is unrelated to birth order, the omission of birth order effects would not matter for estimates of birth gap effects from linear regressions. Matters are different with nonlinear oddsratio methods used by CondeAgudelo et al where the estimated effect of birthspacing is biased in an unknown direction. If birthspacing and birth order are positively correlated, the estimated birthspacing effect is positively biased in linear regression, and it is almost surely positively biased in oddsratio methods. If birthspacing and birth order are negatively correlated, the bias is negative, which means that the estimated negative effect of birthspacing on ASD risk in the metaanalysis might be a spurious result induced by the failure to specify the birth orders of the index children.
If parents have targets for family size, birthspacing is expected to be negatively correlated with birth order (Heckman, Hotz and Walker 1985). Given everything else, if child 2 was slower to arrive, parents who desire three or more children may try to make up for lost time in producing their third child. Heckman et al used data for Swedish women to establish this negative correlation empirically. Target family size theory inevitably casts doubt on the robustness of the main result in CondeAgudelo et al (2016).
Using population cohort data for 4976 index children in Israel with ASD, Beenstock et al (2015) found that recurrence risk is 40% larger among younger siblings born within at least two years of their index sibling. Also, if the index is a firstborn and the second child is typically developed, recurrence risk for the third child falls by 0.5% (9.1% in terms of relative risk).
In our study index children have ASD whereas in CondeAgudelo et al they do not. Therefore, the fact that we found birth spacing is statistically significant after controlling for birth order does not necessarily mean that the same result would be found for the general population. Future research on risk factors of neurodevelopmental disorders should aspire to specify birthspacing and birth order among the covariates.
Yours sincerely
References:
Beenstock M, Levine H, Raz R (2015) Birth Gap and the Recurrence Risk of Autism Spectrum Disorders: A Populationbased Cohort Study. Research in Autism Developmental Disorders, 17: 8694.
CondeAgudelo A., RosasBermudez A., Norton MH (2016) Birth spacing and risk of autism and other neurodevelopmental disabilities.
Original language  American English 

Article number  e20153482 
Number of pages  1 
Journal  Pediatrics 
Volume  137 
Issue number  5 
DOIs 

State  Published  Aug 2016 