Maternal age at childbirth can influence offspring outcomes, especially survival, in early days after birth through biological and social mechanisms. Specifically, younger age at childbirth could be a marker for biologic vulnerabilities, including short stature, inadequate weight gain during pregnancy and potential difficulties in delivery, leading to adverse outcomes [7]. In addition, women in low- and middle-income countries who have children at younger ages are also more likely to be poor and less educated, implying a social disadvantage leading to adverse outcomes in their offspring [8]. The plausibility of biologic or/and social pathways linking maternal age to child survival provides an important basis for interpreting such epidemiologic associations. However, in the absence of rigorous mediation studies substantiating the mechanisms linking the exposure (maternal age) and child survival, caution and scepticism is warranted before attributing causality to the observed associations. Sinha and colleagues’ recommendation to prevent young motherhood as a method of improving child survival implicitly assumes a causal relationship between maternal age and offspring mortality, which given their observational study design, is impossible to ascertain. At the same time, conducting randomized controlled trials on social determinants such as maternal age at child birth are not feasible.

The key question, therefore, is whether the observed association between young motherhood (age <20 y) and an increased risk of post-perinatal mortality is causal? Without the knowledge of whether the observed association is causal, it remains unclear whether maternal age is simply a marker of unmeasured aspects that matter for child survival, or is maternal age truly an independent risk/exposure. In this editorial, we highlight one approach to improving causal inference in observational studies that consider exposures measured on mothers. Using a publicly available national data, we then apply this approach to the case of maternal age at childbirth in India. Based on other studies conducted using this approach, and insights from our own analysis, we conclude that an exclusive maternal focus for improving child survival and development in India may be problematic and misleading.

One approach, which we refer to as ‘maternal-paternal comparison’ can be considerably useful to improving causal inference in observational epidemiologic studies with an interest in maternal exposures [9,10]. The approach entails comparing the similarity in the effect size of a defined maternal and paternal exposure on the outcome. Under the framework where one anticipates a unique and substantial maternal mechanism (e.g., intrauterine mechanisms or behavioral mechanisms such as breastfeeding or providing care) – essentially aspects that fathers do not experience or undertake – one should expect that the size of effect for the maternal exposure should be significantly greater than that of the paternal effect size on the same exposure. If evidence supports a larger effect size for mothers as compared to fathers, this substantially increases our belief to attributing a causal interpretation to the maternal exposure. On the other hand, if the effect size associated with maternal and paternal exposures is highly similar, one ought to be sceptical of attributing causality to maternal exposures. Similarity in effect sizes on the same exposure for mothers and fathers are likely to imply residual confounding at the household level, including assortative mating (couples forming partnerships based on similar characteristics) [11]. While the similarity in effect size does not entirely rule out a potential causal effect of the maternal exposure, causal interpretations become highly implausible. For instance, if the effect of paternal and maternal age at childbirth is similar, one would have to believe, somewhat unrealistically, that whatever unique mechanisms that link maternal age to offspring survival are exactly similar in magnitude as the mechanisms that link paternal age to offspring survival.

The maternal-paternal comparison approach has been validated using the example of parental smoking in the Avon Longitudinal Study of Parents and their Children (ALSPAC) [10]. In this study, maternal smoking during pregnancy was inversely and strongly associated, as should be the case given the strong mechanistic linkages, with offspring birthweight while paternal smoking was not, suggesting a clear maternal specific pathway for the effect of smoking on offspring birthweight. Using this approach, studies using data on Indian populations have investigated whether maternal exposures such as Body Mass Index (BMI) [12,13], height [14], education [15], commonly considered to be causally associated with child mortality and undernutrition, are robust to this test of sensitivity. In each of the instances, the maternal effect sizes were found to be no different from the paternal effect size for the same exposure.

For perinatal and neonatal mortality, the maternal and paternal effects on offspring mortality were similar. For example, maternal age < 20 y was associated with an odds ratio of 1.31 (95% CI 0.99,1.75) for offspring mortality while paternal age <20 y at child birth had an odds ratio of 1.19 (95% CI 0.77, 1.85), with the test of difference being not statistically significant (P=0.74). Interestingly, the effect for paternal age <20 y was stronger compared to maternal age <20 y for post-perinatal mortality (P=0.006) and child mortality, although the effect for child mortality was not statistically significantly stronger compared to mothers (P=0.08). Younger mothers and fathers at childbirth seemed to show the largest effects on child mortality, with the effects of older ages being less consistent and failed to reach statistical significance.

We further examined associations between maternal and paternal age at childbirth and anthropometric failure, defined according to the 2006 WHO growth standards as stunting, wasting, and underweight (Web Fig. 1b). First, the magnitude of maternal and paternal effects was very weak, even for the youngest age group, compared to mortality. Second, a comparison of the maternal and paternal effects indicated that the associations were largely similar, with some evidence of maternal effects being stronger than paternal at ages >25 y.

In summary, given the similarity of effects of maternal and paternal age at childbirth, it seems that maternal age at childbirth may be more likely a marker than a causal exposure in its own right. Consequently, caution must be exercised while attributing causality to the maternal age at childbirth. As expected, there exists a strong correlation between maternal and paternal age at birth in India (r=0.72, P<0.0001) suggesting substantial residual confounding at the family/household level including support for a strong presence of assortative mating [11]. While it has been hypothesized that there may be mechanisms where paternal age could affect birth outcomes [17] (e.g., quality of sperm, epigenetic or DNA changes), it would mainly be observed among older males (>50 years) [18]. In the absence of any biologic mechanisms for younger fathers, we should expect the maternal effects to be much larger at younger ages.

While improving the conditions of the mother has intrinsic significance (including perhaps delaying age of marriage or childbirth), their instrumental role in improving child outcomes may be exaggerated. Scrutiny and scepticism is warranted on existing observational studies with an interest in maternal exposures. Perhaps most critically, there is an urgent need to move away from an exclusive maternal lens to a more household perspective to addressing child survival and development [19-21], since more often than not, in countries such as India, the vulnerabilities at household level often are considerably greater in magnitude than vulnerabilities between individuals within a household.

Funding: None; Competing interest: None stated.

1. Corsi DJ, Subramanian SV. Revisiting the discourse on accomplishing MDG-4. Int J Epidemiol. 2013;42:648-53.

2. You D, Hug L, Ejdemyr S, Idele P, Hogan D, Mathers C,
et al. Global, regional, and national levels and trends in under-5 mortality between 1990 and 2015, with scenario-based projections to 2030: a systematic analysis by the UN Inter-agency Group for Child Mortality Estimation. Lancet. 2015;386:2275-86.

3. Bhutta ZA, Ahmed T, Black RE, Cousens S, Dewey K, Giugliani E, et al. What works? Interventions for maternal and child undernutrition and survival. Lancet. 2008;371:417-40.

4. Gakidou E, Cowling K, Lozano R, Murray CJ. Increased educational attainment and its effect on child mortality in 175 countries between 1970 and 2009: a systematic analysis. Lancet. 2010;376:959-74.

5. Bhutta ZA, Das JK, Rizvi A, Gaffey MF, Walker N, Horton S, et al. Evidence-based interventions for improvement of maternal and child nutrition: what can be done and at what cost? Lancet. 2013;382:452-77.

6. Sinha S, Aggarwal AR, Osmond C, Fall CHD, Bhargava SK, Sachdev HS. Association between maternal age at childbirth and perinatal and under-five mortality in a prospective birth cohort from Delhi. Indian Pediatr. 2016;53:871-7.

7. Alam N. Teenage motherhood and infant mortality in Bangladesh: maternal age-dependent effect of parity one. J Biosoc Sci. 2000;32:229-36.

8. Finlay JE, Ozaltin E, Canning D. The association of maternal age with infant mortality, child anthropometric failure, diarrhoea and anaemia for first births: evidence from 55 low-and middle-income countries. BMJ Open. 2011;1:e000226.

9. Richmond RC, Al-Amin A, Davey Smith G, Relton CL. Approaches for drawing causal inferences from epidemiological birth cohorts: a review. Early Human Dev. 2014;90:769-80.

10. Davey Smith G. Assessing intrauterine influences on offspring health outcomes: can epidemiological findings yield robust results? Basic Clin Pharmacol Toxicol. 2008; 102:245-56.

11. Davey Smith G. Epidemiology, epigenetics and the ‘Gloomy Prospect’: embracing randomness in population health research and practice. Int J Epidemiol. 2011;40:537-62.

12. Subramanian SV, Ackerson LK, Smith GD. Parental BMI and childhood undernutrition in India: an assessment of intrauterine influence. Pediatrics. 2010;126: e663-71.

13. Corsi DJ, Subramanian SV, Ackerson LK, Davey Smith G. Is there a greater maternal than paternal influence on offspring adiposity in India? Arch Dis Child. 2015;100:973-9.

14. Subramanian SV, Ackerson LK, Davey Smith G, John NA. Association of maternal height with child mortality, anthropometric failure, and anemia in India. JAMA. 2009;301:1691-701.

15. Vollmer S, Bommer C, Krishna A, Harttgen K, Subramanian SV. The association of parental education with childhood undernutrition in low- and middle-income countries: comparing the role of paternal and maternal education. Int J Epidemiol. 2016 Aug 8. pii: dyw 133.[Epub ahead of print]

16. International Institute for Population Sciences (IIPS) and Macro International. National Family Health Survey (NFHS-3), 2005–06: India: Volume I. Mumbai: IIPS; 2007.

17. Shah PS, Knowledge Synthesis Group on determinants of preterm/low birthweight b. Paternal factors and low birthweight, preterm, and small for gestational age births: a systematic review. Am J Obstet Gynecol. 2010;202: 103-23.

18. Sharma R, Agarwal A, Rohra VK, Assidi M, Abu-Elmagd M, Turki RF. Effects of increased paternal age on sperm quality, reproductive outcome and associated epigenetic risks to offspring. Reprod Biol Endocrinol. 2015;13:35.

19. Subramanian SV, Mejia-Guevara I, Krishna A. Rethinking policy perspectives on childhood stunting: time to formulate a structural and multifactorial strategy. Mat Child Nutr. 2016;12(Suppl 1):219-36.

20. Corsi DJ, Mejia-Guevara I, Subramanian SV. Improving household-level nutrition-specific and nutrition-sensitive conditions key to reducing child undernutrition in India. Soc Sci Med. 2016;157:189-92.