EBP review 2

ORIGINAL RESEARCH ARTICLE

The trend in the relationship of advanced maternal age to preterm birth and low birthweight

Kitae Sohna,b

aSchool of Economics and Finance, Curtin University, Perth, Australia; bDepartment of Economics, Konkuk University, Seoul, South Korea

ABSTRACT Purpose: As women in developed countries tend to delay childbearing, it becomes more important to understand the relationship of advanced maternal age to birth outcomes. We aimed to estimate the trend in the relationship of advanced maternal age to preterm birth and low birthweight. Materials and methods: We analysed 4,264,417 ethnically homogeneous, singleton firstborns, born in hospitals to married couples in South Korea in 1997–2014. We regressed an indicator for preterm birth or low birthweight on advanced maternal age, the baby’s sex, advanced paternal age, and a set of socioeconomic status (SES) variables by year. We then collected the coefficient on advanced maternal age and charted its trend. We repeated the same procedure for 4,153,313 second- and third births. Results: When we controlled for only the baby’s sex, the relationship between advanced maternal age and preterm birth dramatically weakened in the 2000s and slightly more thereafter: being an older mother was related to a 3.5% point increase in preterm birth in the late 1990s, but this figure decreased to less than 2% points by the early 2010s. Controlling for advanced paternal age slightly decreased the relationship and controlling for SES hardly affected the relationship. We obtained almost the same results for low birthweight. Second- and third-borns exhibited a declining, much weaker influence of advanced maternal age on the birth outcomes. Conclusions: In relative terms, mothers of advanced age were more likely to deliver preterm and low birthweight babies than younger mothers. In absolute terms, however, the risk was small in the 1990s and much smaller in the early 2010s.

ARTICLE HISTORY Received 1 May 2017 Revised 14 August 2017 Accepted 23 August 2017 Published online 26 Septem- ber 2017

KEYWORDS Maternal age; preterm birth; low birthweight; South Korea

Introduction

As women in developed countries tend to delay childbear- ing [1,2], it becomes more important to understand the relationship of advanced maternal age to birth outcomes. Among birth outcomes, preterm birth and low birthweight are important as they are related to adverse developmen- tal, health, educational and economic outcomes in child- hood and adulthood [3]. Researchers have typically tried to determine whether advanced maternal age is adversely related to these birth outcomes. Some studies reported no statistically significant relationship [4–6], but literature reviews concluded that the relationship did exist [7–9].

We contributed to the literature by charting the trend in the relationship over a period of 18 years in a homoge- neous population. The trend can offer useful information to medicine and the public. For example, if the relation- ship has weakened, one can use this information to deter- mine the mechanisms of the weakening and incidentally the relationship itself. Alternatively, a health care provider can counsel potential older mothers, paying less attention to their ages and more attention to other risk factors such as smoking and obesity. Policymakers can even haz- ard to assess the importance of advanced maternal age in the future and allocate resources accordingly. A popula- tion study is attractive because, by definition, the results are generalizable; a hospital-based study can never be free from this concern. Since the population in our study

was large, we exploited a large number of observations, enhancing estimation precision and reducing the possibil- ity of a false negative. Furthermore, our choice of a homogeneous group alleviated bias arising from omitted variables.

Researchers often neglect that babies born to older mothers tend to be born to older fathers (see studies cited in [7–9]); in our data, the correlation coefficient between maternal and paternal age was 0.64 for firstborns and 0.68 for second- and third-borns. It is, however, important to control for paternal age because advanced paternal age is also adversely related to birth outcomes [10,11]. Without controlling for paternal age, it is difficult to determine whether the putative relationship between advanced maternal age and adverse birth outcomes concerns mater- nal or paternal age. If the relationship concerns paternal age, one needs to revisit the literature. If it concerns mater- nal age, one can proceed to assess the relative strength of paternal age in mediating the relationship vis-�a-vis other potential mediators, such as socioeconomic status (SES).

South Korea provides an interesting case for several rea- sons. Almost all Korean citizens are ethnically the same [12]. In addition, Korea’s land area is relatively small, its land size being similar to Indiana’s. Hence, the population is similarly affected by macro conditions, such as culture, geography, climate, food, lifestyle, and public policy. Moreover, Korea has led a group of countries with

CONTACT Kitae Sohn [email protected] School of Economics and Finance, Curtin University, Perth, WA, Australia Supplemental data for this article can be accessed here.

� 2017 The European Society of Contraception and Reproductive Health

THE EUROPEAN JOURNAL OF CONTRACEPTION & REPRODUCTIVE HEALTH CARE, 2017 VOL. 22, NO. 5, 363–368 https://doi.org/10.1080/13625187.2017.1372569

extremely low fertility rates.1 Since advanced maternal age is closely related to low fertility, our results are of great interest to this group of countries. As total fertility rates have decreased even in many developing countries (not- ably, China), our results are also relevant to some develop- ing countries.

Methods

We analysed publicly available birth information on all births at South Korea in 1997–2014. A person enjoys legal rights only if his or her birth was reported to the govern- ment. Hence, the data probably permitted observation of all births in Korea during the period. We included babies who were as homogeneous as possible, while maintaining representativeness: ethnically Korean singletons born in hospitals to married couples. Marriage in Korea is socially a prerequisite to birth: of all births during the period, the highest proportion of babies outside marriage (including births with no information on the status) was only 2.35% in 2009. Babies born to married couples could differ from those born to unmarried couples, and the former were rep- resentative in our data. We thus considered only babies born to married couples. During the period, about 98% of births took place in hospitals, and these hospitals provided birth information, such as birthweight and gestational length. Birth information supplied by hospitals is presum- ably more accurate than that provided by other locations,2

and babies born in hospitals were representative and could differ from those born in other locations. We thus consid- ered only babies born in hospitals.

Despite Korea’s extremely low total fertility rate [2], accompanied by a higher standard of living [13,14], almost all married couples have at least one child: of married women aged 20–44 in a nationally representative survey, 90% reported having at least one child in 2012 [15, p. 130]. Since fertility was higher before this year, the rate must have been higher in earlier years. In contrast, having a second child is not as prevalent as having a first child, so the family background of firstborns could differ from that of babies of higher parity. We thus focused on firstborns, but for completeness, we briefly reported results for second- and third-births combined. Twins are significantly lighter than singletons for reasons other than advanced maternal age, and twins accounted for less than 4% of all births in 2014. We thus considered only singletons. The Korean population has long consisted of one ethnic group [12], and Statistics Korea disclosed information on parental citizenship (roughly identical to ethnicity) for 2010–2013 (but not for 2014). Of 436,455 births recorded in 2013, 94.05% were born to Korean mothers and Korean fathers. To consider babies of the same ethnicity, we included only babies born to Korean mothers and Korean fathers.

To reduce bias and enhance estimation precision, we excluded observations with extreme values. We restricted birthweights to 500 g or greater, gestational lengths to 22 weeks or longer. Since the number of the excluded obser- vations was minuscule, the exclusion was inconsequential (not shown). Since teenage births are extremely rare in Korea (e.g., 0.59% in 2014), we restricted maternal ages to 20–49; ages younger than 20 exhibited very imprecisely measured estimates (not shown). Smoking is an important

determinant of birth outcomes [16], but our data did not contain this information. This lack of information is, how- ever, inconsequential because almost no pregnant women in Korea were smokers. For example, when Jhun et al. [17] tested 1057 urine samples from pregnant women who visited 30 randomly sampled obstetric clinics and hospitals for prenatal care in 2006, only 3.0% turned out to be cotinine-validated smokers. We detailed the process of excluding observations in Supplementary Appendix Table A-2.

In this study, preterm birth refers to birth at a gesta- tional length <37 weeks, and low birthweight refers to a birthweight <2500 g. We used a cut-off age of 35 years to define advanced maternal age and justified this age below. Considering the age difference in a couple (about three years, see Table 1), we defined the paternal age 40 years and older as advanced paternal age – older mothers and fathers for brevity.3 In addition to the baby’s sex and pater- nal age, we controlled for the following covariates: parental education (at least a college education vs. otherwise), par- ental occupation (nine categories) and municipality fixed effects (the number varied depending on the birth year, but fluctuated around 250 categories). By controlling for municipality fixed effects, we compared babies born in the same municipality. This is an effective way of comparing babies who were as homogeneous as possible because a municipality in Korea is small in size (e.g., the greatest number of births took place in Songpa-gu during the period, and its size is only slightly greater than half of Manhattan’s). Education and occupation are important components of SES, and residential area reflects the resi- dent’s SES. We thus attempted to control for parental SES by controlling for this set of covariates. The number of observations were 4,264,417 for firstborns and 4,153,313 for second- and third-births.

Since our main aim was to chart the trend in the rela- tionship between advanced maternal age and the birth outcomes, we first estimated the following specification by a linear probability model (LPM) for each year:4

Yim ¼ b1oldmomim þ b2sexim þ b3olddadim þ Ximb3 þ mm þ uim (1)

where Yim refers to an indicator of preterm birth or low birthweight of baby i born in municipality m, oldmom to whether the mother was aged 35 þ or not, sex to the baby’s sex, olddad to whether the father was aged 40þ or not, X to a set of covariates, m to municipality fixed effects, u to the error term and the b’s to coefficients to estimate. We took advantage of the large number of observations by flexibly controlling for all covariates in dummy form. We then plotted the trend of b1 over the 18 years studied.

Our secondary aim was to estimate the relative power of advanced paternal age (versus SES) in mediating the relationship between advanced maternal age and the birth outcomes. We thus incrementally added the independent variables. The first set consisted of oldmom and sex. We then added olddad. We next replaced olddad with SES (namely, X and m) since we competed olddad with SES. For completeness, we finally controlled for both olddad and SES. In preliminary analyses, we checked whether estima- tion results varied by sex and confirmed that they did not. We thus pooled boys and girls.

364 K. SOHN

Results

Background results

Table 1 presents descriptive statistics. To save space, we only described the variables of interest (for more, see [18,19]). The percentage of older mothers was 6.3% for first- borns and 15.2% for second- and third-borns. The percent- age of preterm births was 3.6% for firstborns and 3.9% of second- and third-borns. The corresponding figures for low birthweights was 3.5% and 2.9%.

Researchers typically use a cut-off point for advanced maternal age at 35, 40, or even 45 years [4,20,21]. Since we had a large number of observations, we could precisely estimate the relationship between each year of maternal age and the birth outcomes by pooling all birth years and regressing each of the birth outcomes on a series of mater- nal age dummies (and the baby’s sex); the age 20 was the reference age. Supplementary Figure A-1 shows the coeffi- cient on each year of age with capped spikes indicating 95% confidence intervals. The wider confidence intervals for 40þ indicate a small number of babies born to mothers in this age range. Relative to 20, the probability of having a preterm birth was slightly lower in the 20s, but switched to be higher between 30 and 35. We obtained the same

results for low birthweight (Figure A-2 in the Supplementary appendix). At 35þ, the probability was dis- cernibly higher. Considering the wide confidence intervals for 40þ and the discernible difference in 35þ, we defined advanced maternal age as 35þ. Another reason for this cut-off point is that it is common in the literature [22]; choosing it thus facilitates the comparison of our results with others.

Trends in the relationship of advanced maternal age to birth outcomes

Figure 1 shows that when we controlled for only the baby’s sex (thick solid line), the relationship between advanced maternal age and preterm birth dramatically weakened in the 2000s and slightly more thereafter: being an older mother was related to about a 3.5% point increase in pre- term birth in the late 1990s, but this figure decreased to under 2% points by the early 2010s. When we added olddad (thick dashed line), the trend in the relationship slightly shifted down. When we replaced olddad with SES (thin solid line), the trend almost reverted to the original one, indicating that SES hardly mediated the relationship. Comparisons of the thick dashed line and thin solid line

Table 1. Descriptive statistics.

Firstborn 2nd and 3rd-born

Continuous variable Mean (SD) Mean (SD) Girl’s birthweight (g) 3202 (426) 3226 (422) Boy’s birthweight (g) 3296 (441) 3330 (435) Maternal age (year) 28.4 (3.8) 30.7 (3.8) Paternal age (year) 31.1 (4.1) 33.5 (4.0)

Discrete variable % % Maternal age <35 years 93.7 84.8 Maternal age � 35 years 6.3 15.2 Paternal age <40 years 96.7 92.9 Paternal age � 40 years 3.3 7.2 Gestational length

�36 3.6 3.9 37 4.5 6.9 38 14.5 25.0 39 24.6 25.8 40 40.2 32.6 �41 12.6 5.7

Birthweight �2500 g 96.5 97.2 Birthweight <2500 g 3.5 2.9 Girls 48.6 47.6 Boys 51.4 52.4 Maternal education < College or above 41.6 52.0 Maternal education � College or above 58.4 48.1 Maternal Occupation

Manager 0.6 0.5 Professional 7.8 5.0 Clerical and kindred 14.2 8.3 Sales and service worker 3.7 2.8 Farmer, fisherman, forest worker 0.3 0.6 Craftsman 0.4 0.2 Operative 0.1 0.1 Labourer 0.2 0.1 Student, housewife, unemployed, missing, or soldier 72.7 82.3

Paternal education < College or above 37.5 43.6 Paternal education � College or above 62.5 56.4 Paternal occupation

Manager 3.1 3.6 Professional 12.2 10.8 Clerical and kindred 44.3 41.7 Sales and service worker 18.9 21.0 Farmer, fisherman, forest worker 1.6 2.5 Craftsman 7.8 8.6 Operative 3.7 4.2 Labourer 1.7 2.2 Student, housewife, unemployed, missing, or soldier 6.6 5.5

N 4,264,417 4,153,313

SD: standard deviation.

THE EUROPEAN JOURNAL OF CONTRACEPTION & REPRODUCTIVE HEALTH CARE 365

indicate that advanced paternal age exerted more influence on the mediation than SES. As expected, when we con- trolled for both olddad and SES (thin dotted line), the trend was almost identical to that with only olddad.

We repeated the same exercise with low birthweight and obtained almost the same results (Figure A-3 in the Supplementary appendix). Note that the proportion of pre- term births increased from 2.6% in 1997 to 4.5% in 2014; the corresponding figures for low birthweights were 2.9% and 4.2% (Figure A-4 in the Supplementary appendix), but the relationship between advanced maternal age and the birth outcomes weakened in absolute terms. Hence, the effect of advanced maternal age on the birth outcomes fur- ther decreased in relative terms. We measured the relative size of the coefficient on advanced maternal age by divid- ing the coefficient by the mean of the birth outcome for each birth year. To estimate the total relationship between advanced maternal age and the birth outcomes, we con- trolled for only the baby’s sex. Figure 2 shows a fast decrease up to 2007 and thereafter a slow decrease for both outcomes. In the late 1990s, being an older mother was related to an increase in the birth outcomes by more than 100% of the mean of the outcomes, but by 2014, the figure decreased to about 40%. We repeated the exercise

for second- and third-borns and found the same downward trends (Figure 3). The only difference was that the levels for second- and third-borns were lower than those for first- borns by about 40% points. Hence, advanced maternal age poses a substantially lower risk for preterm births and low birthweights in the later years of the study. It is unclear what drove the upward trends in preterm birth and low birthweight, but it is clear that advanced maternal age was not the main driver.

Discussion

Differences in results and conclusions

We charted the trends in the relationship of advanced maternal age to preterm birth and low birthweight. This exercise improved on the literature in that it went beyond asking whether the relationship existed. We found not only did the relationship exist during the entire period but also that it weakened over time. Some researchers argued that they did not find a statistically significant relationship between advanced maternal age and birth outcomes, including preterm birth and low birthweight. We believe that they were false negatives because they considered a small number of older mothers. For example, Grimes and Gross [6] dismissed the difference between 30% and 24% in the rate of low birthweight between 149 hypertensive mothers aged 35þ vs. 462 hypertensive mothers aged 34�, saying that it was not statistically significant. In contrast, Kirz et al. [5] found a statistically significant difference, but their finding is counter-intuitive: 54 mothers aged 35þ experienced a lower rate of low birthweights than 251 mothers aged 20–25 (6.5% vs. 9.9%). Even if the two find- ings were true, the concern of selection remains because they analysed not a national population of babies but a selected group, babies who were delivered at a single hos- pital. Another example worsens this concern. Berkowitz et al. [4] estimated an adjusted odds ratio of 1.0 for pre- term birth between mothers aged 20–29 and mothers aged 35þ and concluded that there was no difference between the two groups. This ratio does not present a grave con- cern for a false negative or a counter-intuitive outcome. However, most of the older women in their study were white (87.2% were whites), married (90.6% were married),

.2

.4

.6

.8

A d va

n ce

d M

a te

rn a l A

g e C

o e ff

ic ie

n t

/ O

u tc

o m

e M

e a n

1997 2002 2007 2012 Birth Year

Preterm Birth Low Birthweight

Figure 3. Trends in the relative strength of the relationship of advanced maternal age to preterm birth and low birthweight: Second- and third-borns. To estimate the coefficient on the dummy for advanced maternal age, we controlled for only the baby’s sex.

.015

.02

.025

.03

.035

.04 M

a te

rn a l A

g e C

o e ff

ic ie

n t

1997 2002 2007 2012

Birth Year

Sex Sex & Paternal Age Sex & SES Sex, Paternal Age, & SES

Figure 1. Trends in the strength of the relationship between advanced maternal age and preterm birth: firstborns. Sex refers to the baby’s sex, par- ental age to a dummy indicating a paternal age of 35þ, and SES to parental education and occupation and municipality fixed effects.

.4

.6

.8

1

1.2

1.4

A d va

n ce

d M

a te

rn a l A

g e C

o e ff ic

ie n t / O

u tc

o m

e M

e a n

1997 2002 2007 2012 Birth Year

Preterm Birth Low Birthweight

Figure 2. Trends in the relative strength of the relationship of advanced maternal age to preterm birth and low birthweight: firstborns. To estimate the coefficient on the dummy for advanced maternal age, we controlled for only the baby’s sex.

366 K. SOHN

college educated (94.5% attended college), non-smoking (96.5% did not smoke during pregnancy), and healthy (88.2% reported no chronic medical conditions). Therefore, external validity is doubtful.

We left the possibility open that we failed to control for important mediators or confounders, and we are sympa- thetic to the conclusion of Newburn-Cook and Onyskiw [8] in their literature review: maternal age exerts an effect on both gestational age and birthweight but it is unclear whether maternal age exerts an independent and direct impact on the birth outcomes or acts indirectly through its association with age-dependent confounders. This lack of information can be a critical limitation if one attempts to determine the existence of the relationship between advanced maternal age and the birth outcomes and to identify the mechanisms behind the relationship. However, this is not what we pursued. Since we aimed to chart the trend in the relationship and to run a race between advanced paternal age and SES, the lack of information should not detract from this study.

Furthermore, given the homogeneity of babies in our data and the extensive array of covariates in our specifica- tion, it appears that advanced maternal age exerts inde- pendent and direct effects on the birth outcomes. If this interpretation is unconvincing, at least, we can say that whatever the mechanisms were, their influence weakened over time. What then can explain the weakening? Risk fac- tors for preterm birth and low birthweight include socio- demographic risk factors, medical risks before pregnancy, risks of the current pregnancy, a lack of prenatal care, and environmental and behaviour risks [23,24]. Considering the short period, we could eliminate genetic and constitutional influences. We instead argue that the fast improvement in the monetary and nonmonetary standard of living can explain our results. For example, the increase in Korean GDP per capita in 1953–2000 (14.9 times) was the greatest among 142 countries [25]. A better growth environment increases height and decreases age at menarche [26–35], and it is revealing that the speed of increasing height and of decreasing age at menarche in Korea was fastest in the world [14]. Furthermore, the close connection between the early growth environment and health in adulthood, known as the fetal origins hypothesis [36], implies that Korean mothers of the same age became healthier; mothers in the late 30s in 2014 might be as healthy as or even healthier than those in the early 30s in 1997. The growing purchas- ing power allowed Korean mothers to fully utilize the development of medical technology and consequently to be healthier and to mitigate the said risk factors. One measure of health succinctly summarises the improvement in health. Life expectancy at birth for Koreans increased fastest among OECD countries [37] from 52.4 years in 1960 to 78.5 years in 2005, or 5.8 years per decade. If our explan- ations are plausible, future research will replicate our results for countries that experienced or are experiencing fast eco- nomic growth.

Although the relationship between advanced maternal age and the birth outcomes weakened over time, the rela- tionship was not small in relative terms even when it was the weakest. When we controlled for only the baby’s sex, the relationship was still about 40% of the mean of the birth outcome (Figure 2). That said, we assure worried future mothers of advanced age that in absolute terms, the

risk of preterm birth and low birthweight is small, and if the trend continues, will be smaller. When we controlled for only the baby’s sex, thereby estimating the maximum relationship, the probability of having preterm birth was 3.7% points greater for older mothers than the rest in 1997–1999 but 1.9% points in 2012–2014. The figure for low birthweight decreased from 3.4% points to 1.9% points. We thus agree with the conclusion of Carolan and Frankowska [7] and Usta and Nassar [9] in their literature reviews that although the likelihood of adverse birth out- comes increases with maternal age, the outcomes are gen- erally favourable for older mothers in the absence of pre- existing medical conditions. We can say the same to future fathers of advanced age.

Conclusions

We analyzed more than eight million babies born in South Korea in 1997–2014 and obtained encouraging results for mothers of advanced age. In relative terms, mothers of advanced age were more likely to deliver preterm and low birthweight babies than younger mothers. In absolute terms, however, the risk was small in the late twentieth century and became much smaller by the second decade of this century.

Acknowledgements

I am grateful to the anonymous reviewer and Irving Sivin for helpful comments and suggestions.

Disclosure statement

There is no conflict of interest.

Notes

1. Detailed data are available at: https://data.oecd.org/pop/fertility- rates.htm.

2. Gestational length in the hospital report was determined by either the last menstrual period or an ultrasound exam, the latter being quite accurate. The data did not indicate which method was used, but indirect evidence suggests that the length was accurate. We analysed publicly available data on all infant deaths between 2009 and 2014, which contained information as to whether the mother received prenatal care. Prenatal care in Korea involves ultrasound exams, and if it does not adversely affect birth outcomes (the opposite is the intention), this information should provide lower bounds of percentages of mothers who received prenatal care for all births. Table A-1 in the Supplementary appendix shows that most received prenatal care, implying that hospitals provide accurate gestational lengths for most births.

3. Slightly varying the cut-off points for advanced maternal and paternal ages did not change the substance of our results (not shown).

4. Applying a probit or logit regression did not change the substance of our results (not shown).

References

[1] Burkimsher M. Europe-wide fertility trends since the 1990s: turning the corner from declining first birth rates. Demogr Res. 2015;32:621–656.

THE EUROPEAN JOURNAL OF CONTRACEPTION & REPRODUCTIVE HEALTH CARE 367

[2] Sohn K. Parents are rapidly getting older in South Korea. Human Fertil. 2017;20:212–216.

[3] Currie J, Almond D, Human capital development before age five. In: Card D, Ashenfelter O, editors. Handbook of labor economics, volume 4B. 1st ed. Amsterdam: Elsevier; 2011. p. 1315–1486.

[4] Berkowitz GS, Skovron ML, Lapinski RH, et al. Delayed child- bearing and the outcome of pregnancy. N Engl J Med. 1990;322:659–664.

[5] Kirz DS, Dorchester W, Freeman RK. Advanced maternal age: the mature gravida. Am J Obstet Gynecol. 1985;152:7–12.

[6] Grimes DA, Gross GK. Pregnancy outcomes in black women aged 35 and older. Obstet Gynecol. 1981;58:614–620.

[7] Carolan M, Frankowska D. Advanced maternal age and adverse perinatal outcome: a review of the evidence. Midwifery 2011;27:793–801.

[8] Newburn-Cook CV, Onyskiw JE. Is older maternal age a risk fac- tor for preterm birth and fetal growth restriction? A systematic review. Health Care Women Int. 2005;26:852–875.

[9] Usta IM, Nassar AH. Advanced maternal age. Part I: obstetric complications. Am J Perinatol. 2008;25:521–534.

[10] K€uhnert B, Nieschlag E. Reproductive functions of the ageing male. Hum Reprod Update. 2004;10:327–339.

[11] Sartorius GA, Nieschlag E. Paternal age and reproduction. Hum Reprod Update. 2010;16:65–79.

[12] Sohn K. Men’s revealed preference for their mates’ ages. Evolution & Human Behavior. 2017;38:58–62.

[13] Sohn K. Sexual stature dimorphism as an indicator of living standards? Ann Hum Biol. 2016;43:537–541.

[14] Sohn K. Improvement in the biological standard of living in 20th century Korea: evidence from age at menarche. Am J Hum Biol. 2017;29:e22882.

[15] Kim S, Park J, Kim Y, et al. 2012 national survey of marriages and births. Seoul: Korea Institute for Health and Social Affairs; 2012.

[16] Hackshaw A, Rodeck C, Boniface S. Maternal smoking in preg- nancy and birth defects: a systematic review based on 173 687 malformed cases and 11.7 million controls. Hum Reprod Update. 2011;17:589–604.

[17] Jhun H-J, Seo H-G, Lee D-H, et al. Self-reported smoking and urinary cotinine levels among pregnant women in Korea and factors associated with smoking during pregnancy. J Korean Med Sci. 2010;25:752–757.

[18] Sohn K. Birthweight as a measure of holiday stress. Women Health. 2017. DOI: 10.1080/03630242.2017.1306605

[19] Sohn K. Disappearing seasonality in birthweight. Am J Hum Biol. 2016;28:767–773.

[20] Chan BC-P, Lao TT-H. Effect of parity and advanced maternal age on obstetric outcome. Int J Gynaecol Obstet. 2008;102:237–241.

[21] Dildy GA, Jackson GM, Fowers GK, et al. Very advanced mater- nal age: pregnancy after age 45. Am J Obstet Gynecol. 1996;175:668–674.

[22] Heffner LJ. Advanced maternal age-how old is too old? N Engl J Med. 2004;351:1927–1929.

[23] Goldenberg RL, Culhane JF, Iams JD, et al. Epidemiology and causes of preterm birth. Lancet. 2008;371:75–84.

[24] de Bernab�e JV, Soriano T, Albaladejo R, et al. Risk factors for low birth weight: a review. Eur J Obstetr Gynecol Reprod Biol. 2004;116:3–15.

[25] Rhee YH. Economic stagnation and crisis in Korea during the eighteenth and nineteenth centuries. Aust Econ Hist Rev. 2014;54:1–13.

[26] Sohn K. Biological standards of living: age at menarche vs height. Ann Hum Biol. 2017;44:21–27.

[27] Sohn K. The influence of climate on age at menarche: augmented with the influence of ancestry. HOMO. 2016;67:328–336.

[28] Sohn K. The influence of birth season on height: evidence from Indonesia. Am J Phys Anthropol. 2015;157:659–665.

[29] Sohn K. Is leg length a biomarker of early life conditions? Evidence from a historically short population. Am J Hum Biol. 2015;27:538–545.

[30] Sohn K. The trend in age at menarche in Indonesia: birth cohorts 1944–1988. J Biosoc Sci. 2015;47:407–412.

[31] Sohn K. Age and size at maturity in Indonesian women: a norm of reaction? Am J Hum Biol. 2014;26:713–715.

[32] Sohn K. The height premium in Indonesia. Econ Hum Biol. 2015;16:1–15.

[33] Sohn K. The value of male height in the marriage market. Econ Hum Biol. 2015;18:110–124.

[34] Sohn K. Now, the taller die earlier: the curse of cancer. J Gerontol Ser A. 2015;71:713–719.

[35] Sohn K. The influence of birth season on mortality in the United States. Am J Hum Biol. 2016;28:662–670.

[36] Barker DJP, Mothers, babies and health in later life. Amsterdam (The Netherlands): Elsevier Health Sciences; 1998.

[37] Yang S, Khang Y-H, Harper S, et al. Understanding the rapid increase in life expectancy in South Korea. Am J Public Health. 2010;100:896–903.

368 K. SOHN

Copyright of European Journal of Contraception & Reproductive Health Care is the property of Taylor & Francis Ltd and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.