03012115-S03012115XXXXXXXX-S0301211514001067-main
Click here to load reader
-
Upload
havilah-zeki-rosa -
Category
Documents
-
view
216 -
download
0
Transcript of 03012115-S03012115XXXXXXXX-S0301211514001067-main
8/12/2019 03012115-S03012115XXXXXXXX-S0301211514001067-main
http://slidepdf.com/reader/full/03012115-s03012115xxxxxxxx-s0301211514001067-main 1/5
Gestational
age-dependent
risk
factors
for
preterm
birth:associations
with
maternal
education
and
age
early
in
gestation
Nathalie AugerQ1 a,b,c,*, Michal Abrahamowiczd, Willy Wynant d, Ernest Lo a
a Institut national de sante publique du Que bec, Montre al, Canada bResearch Centre of the University of Montre al Hospital Centre, Montre al, Canada cDepartment of Social and Preventive Medicine, University of Montre al, Montre al, Canada dDepartment of Epidemiology, Biostatistics, and Occupational Health, McGill University, Montre al, Canada
1. Introduction
Preterm birth (PTB) is an important contributor to neonatal
morbidity and mortality [1,2], especially when delivery occurs at
very early gestational ages [3,4]. Although the causes of PTB are
poorly understood [5], maternal socio-demographic characteris-
tics such as low education and older age are known risk factors
[6–10]. Few studies, however, have examined how maternal
characteristics are associated with very early (e.g., <32 gestational
weeks) as opposed to later PTB. PTB is frequently analyzed
dichotomously with early and late cases combined, an approach
which does not account for the fact that preterm delivery is the
2consequence of a dynamic process resulting in an event (birth) at a
2specific time [11], and that potentially pathologic PTBs at earlier
2gestational ages may be more strongly associated with maternal
2risk factors than PTBs at later gestational ages. A natural way to
2determine if associations vary by gestational age is to test for non-
2proportional hazards of risk factors in a time-to-event (survival)
2analysis. A particular advantage of time-to-event analysis is that
2associations between risk factors and PTB can be expressed at
2specific gestational ages if there are differences over gestation (i.e.,
3non-proportional hazards), or summarized as one overall measure
3of association if there are no differences by gestational age (i.e.,
3proportional hazards).
3Our objective was to determine if the association between PTB
3and maternal risk factors varies by gestational age at delivery using
3time-to-event analysis. We evaluated two risk factors for PTB,
3maternal education and age [12–17], both relevant to future
3research on determinants of perinatal outcomes over the range of
3gestational age, and to guiding clinical prevention [18].
European Journal of Obstetrics & Gynecology and Reproductive Biology xxx (2014) xxx–xxx
A R T I C L E I N F O
Article history:
Received 25 February 2013
Received in revised form 8 July 2013
Accepted 21 February 2014
Keywords:
Educational status
Gestational age
Maternal
age
Preterm birth
Survival analysis
A B S T R A C T
Objectives:
Pretermbirth (PTB) before 37weeks can occur
over a wide range of gestational ages,but few
studieshave assessedif associations betweenrisk factors andPTB vary over theduration of gestation.We
sought to evaluate if associations between two major risk factors (maternal education and age) and PTB
depend on gestational age at delivery.
Studydesign: Weestimatedhazard ratios ofPTBfor educationand agein a time-to-eventanalysis using a
retrospective cohort of 223,756 live singleton births from the province of Quebec, Canada for the years
2001–2005. Differences in hazards of maternal education and age with PTB were assessed over
gestational age in a Cox proportional hazardsmodel using linear and nonlinear time interaction terms,
adjusting for maternal characteristics.
Results: Associations of PTBwith lower (vs. higher) education and older (vs. younger) age strengthened
progressively at earliergestational ages, suchthat the risk of PTB formaternal education andage wasnot
constant over the course of gestation.
Conclusions: Associations of PTB with risk factors such asmaternal low education and older age may be
stronger early in gestation. Models that capture the time-dependent nature of PTB may be useful when
thegoal is to assess associations at lowgestationalages, and to avoidmaskedorbiased associationsearly
in gestation.
2014
Published by Elsevier Ireland Ltd.
* Corresponding author at: Institut national de sante publique du Quebec, 190,
boulevard Cremazie Est, Montreal, Quebec H2P 1E2, Canada.
Tel.:
+1
514
864
1600x3717;
fax:
+1
514
864
1616.
E-mail address: [email protected] (N. Auger).
G Model
EURO 8453 1–5
Please cite this article in press as: Auger N, et al. Gestational age-dependent risk factors for preterm birth: associations with maternal
education and age early in gestation. Eur J Obstet Gynecol (2014), http://dx.doi.org/10.1016/j.ejogrb.2014.02.035
Contents
lists
available
at
ScienceDirect
European Journal of Obstetrics & Gynecology andReproductive Biology
jou r nal h o mepag e: w ww.elsev ier .co m / locate /e jo g rb
http://dx.doi.org/10.1016/j.ejogrb.2014.02.035
0301-2115/ 2014 Published by Elsevier Ireland Ltd.
8/12/2019 03012115-S03012115XXXXXXXX-S0301211514001067-main
http://slidepdf.com/reader/full/03012115-s03012115xxxxxxxx-s0301211514001067-main 2/5
8/12/2019 03012115-S03012115XXXXXXXX-S0301211514001067-main
http://slidepdf.com/reader/full/03012115-s03012115xxxxxxxx-s0301211514001067-main 3/5
education and age. Gestational age-specific HRs derived from the
linear interaction terms showed that associations were stronger at
lower gestational ages. With a decrease in gestational age from 36
to 22 weeks, HRs rose from 1.8 to 3.2 for education (low relative to
high), and from 1.1 to 2.0 for maternal age (older relative to
younger). Thus, the models that assumed proportional hazards
underestimated the magnitude of HRs at low gestational ages (a
difference that was statistically significant at 28 weeks), and
slightly overestimated the HR at 36 weeks.
Tests for proportional hazards in spline-based nonlinear modelsconfirmed that hazards of PTB for maternal education (P < 0.001)
and age (P < 0.01) were non-proportional (Table 2). Compared
with linear interaction terms, the nonlinear interaction terms
yielded slightly stronger associations for both maternal education
and age at lower gestational ages. The plotted spline curves
suggested a nonlinear decrease in HRs over time for both risk
factors, such that the risk of PTB for low education, and to a lesser
extent high maternal age, fell more sharply below 26 gestational
weeks (Fig. 1).
In sensitivity analyses, inclusion of births with gestational ages
<22 weeks had little influence on the shape of spline curves,
though associations were slightly attenuated (data not shown).
Exclusion of births with missing data and use of cubic (rather than
quadratic)
splines
yielded
similar
results.
Associations
includingmothers aged 20–24 years were somewhat attenuated. Last, use of
categorical gestational age-by-education (or age) interaction terms
in Cox models provided less evidence of non-proportionality
(P = 0.06).
4.
Comment
This study assessed whether associations between PTB and two
important risk factors, maternal education and age, varied over the
continuous range of gestational age. Using linear and nonlinear
gestational age interaction terms in time-to-event regression
models, we found that HRs for both lower education and older
maternal age strengthened progressively with decreasing gesta-
tional
age.
Thus,
we
demonstrated
that
the
hazards
of
PTB
for
both
1maternal education and age were not constant over gestation. The
1implication is that regression models with PTB expressed
1dichotomously (e.g., logistic regression), as well as Cox models
1that do not account for non-proportionality of hazards, yield
1average odds ratios or HRs that are heavily weighted by the much
1greater number of late PTBs [25], which mask the stronger
1associations at very low gestational ages. Researchers should be
1aware
that
results
with
PTB
measured
dichotomously
reflect1findings for late PTB primarily, and may not accurately reflect
1associations for very or extreme PTB.
1While evidence suggests that associations with maternal
1education or age may be stronger when PTB is dichotomized at
1lower gestational age cut-points than the standard 37 weeks used
1in most studies [13–17], the continuous nature of gestational age is
1not usually considered [28]. Furthermore, proportionality of
1hazards is rarely investigated in studies that do evaluate PTB as
1a time-dependent outcome. The only exception we could identify
1was an analysis of Danish birth data in which Cox regression was
1used to investigate the association between low education
1and extreme, very, or moderate PTB [12]. Although the study
1suggested that education was more strongly associated with very
1and
extreme
PTB
than
late
PTB,
the
test
for
non-proportionality
of
Table 2
Hazard of preterm birth for maternal education and age, overall and by gestational
week, singletons, Quebec, 2001–2005.
Education HR (95% CI)
(low vs. high)
Age HR (95% CI)
(high vs. low)
Averaged over all
preterm
birthsa
Unadjusted 1.76 (1.64, 1.88) 0.97 (0.91, 1.03)
Adjusted 1.92 (1.78, 2.06) 1.21 (1.13, 1.28)
By
gestational
weekLinear interactionb
22 weeks 3.17 (2.19, 4.57) 1.96 (1.60, 2.39)
28 weeks 2.48 (2.08, 2.97) 1.55 (1.44, 1.66)
32 weeks 2.11 (1.91, 2.34) 1.32 (1.29, 1.36)
36 weeks 1.80 (1.66, 1.94) 1.13 (1.12, 1.14)
P value interaction
<0.001
<0.001
Nonlinear
interactionc
22 weeks 4.57 (2.62, 7.98) 2.48 (1.46, 4.24)
28 weeks 2.32 (1.93, 2.80) 1.55 (1.29, 1.85)
32
weeks
2.01
(1.76,
2.30)
1.26
(1.11,
1.44)
36 weeks 1.74 (1.59, 1.90) 1.17 (1.07, 1.28)
P value interaction
<0.001
<0.01
a Cox models assuming proportional hazards, adjusted for education, age, marital
status, language, and parity.b Cox models accounting for non-proportional hazards with linear education (or
age)-by-gestational
age
interaction
terms,
adjusted
for
education,
age,
marital
status,
language,
and
parity.c Cox models accounting for non-proportional hazards with nonlinear education
(or age)-by-gestational age spline interaction terms, adjusted for education, age,
marital status, language, and parity.
Fig. 1. Association between maternal education/age and preterm birth by
gestational age. Hazard ratio (bold) and 95% confidence interval from Cox
models accounting for non-proportional hazards with nonlinear education (or
age)-by-gestational
age
spline
interaction
terms,
adjusted
for
education,
age,
marital status, language, and parity (first imputation; results for other imputations
were similar).
N. Auger et al. / European Journal of Obstetrics & Gynecology and Reproductive Biology xxx (2014) xxx–xxx 3
G Model
EURO 8453 1–5
Please cite this article in press as: Auger N, et al. Gestational age-dependent risk factors for preterm birth: associations with maternal
education and age early in gestation. Eur J Obstet Gynecol (2014), http://dx.doi.org/10.1016/j.ejogrb.2014.02.035
8/12/2019 03012115-S03012115XXXXXXXX-S0301211514001067-main
http://slidepdf.com/reader/full/03012115-s03012115xxxxxxxx-s0301211514001067-main 4/5
8/12/2019 03012115-S03012115XXXXXXXX-S0301211514001067-main
http://slidepdf.com/reader/full/03012115-s03012115xxxxxxxx-s0301211514001067-main 5/5
[17] Martius JA, Steck T, Oehler MK, Wulf KH. Risk factors associated with preterm(<37+0 weeks) and early preterm birth (<32+0 weeks): univariate andmultivariate analysis of 106 345 singleton births from the 1994 statewideperinatal survey of Bavaria. Eur J Obstet Gynecol Reprod Biol 1998;80:183–9.
[18] Ashton DM, Lawrence HC, Adams NL, Fleischman AR. Surgeon general’sconference on the prevention of preterm birth. Obstet Gynecol 2009;113:925–30.
[19] Auger N, Abrahamowicz M, Park AL, Wynant W. Extreme maternal educationand preterm birth: time-to-event analysis of age and nativity-dependent risks.Ann Epidemiol 2013;23:1–6.
[20] Acevedo-Garcia D, Soobader M, Berkman L. Low birthweight among US
Hispanic/Latino
subgroups:
the
effect
of
maternal
foreign-born
status
andeducation. Soc Sci Med 2007;65:2503–16.[21] Auger N, Luo ZC, Platt RW,Daniel M.Do mother’s education and foreign-born
status interact to influence birth outcomes? Clarifying the epidemiologicparadox and the healthy migrant effect. J Epidemiol Community Health2008;62:402–9.
[22] Moutquin JM. Classification and heterogeneity of preterm birth. Br J ObstetGynaecol
2003;110(Suppl.
20):30–3.[23] Greenland S. Avoiding power loss associated with categorization and ordinal
scores in dose–response and trend analysis. Epidemiology 1995;6:450–4.[24] Sterne JAC, White IR, Carlin JB, et al. Multiple imputation for missing data in
epidemiological and clinical research: potential and pitfalls. BMJ 2009;338:b2393.
[25] Allison P. Survival analysis using SAS: a practical guide. Cary NC: SAS Institute;2010.
[26] Quantin C, Abrahamowicz M, Moreau T, et al. Variation over time of the effectsof prognostic factors in a population-based study of colon cancer: comparisonof statistical models. Am J Epidemiol 1999;150:1188–200.
[27] Abrahamowicz M,MacKenzie T. Joint estimation of time-dependent and non-
linear
effects
of
continuous
covariates
on
survival.
Stat
Med
2007;26:392–408.[28] Greenland S. Dose–response trend analysis in epidemiology: alternatives to
categorical analysis. Epidemiology 1995;6:356–65.
3[29] Kyrklund-Blomberg NB, Cnattingius S. Preterm birth and maternal smoking:3risks related to gestational age and onset of delivery. Am J Obstet Gynecol31998;179:1051–5.3[30] Tegethoff M, Greene N, Olsen J, Meyer A, Meinlschmidt G. Maternal psycho-3social adversity during pregnancy is associated with length of gestation and3offspring size at birth: evidence from a population-based cohort study.3Psychosom Med 2010;72:419–26.3[31] Peacock JL, Bland JM, Anderson HR. Preterm delivery: effects of socioeconomic3factors, psychological stress, smoking,alcohol, and caffeine. BMJ 1995;311:531–5.3[32] Goldenberg RL, Hauth JC, Andrews WW. Intrauterine infection and preterm3delivery. N Engl J Med 2000;342:1500–7.
3[33]
Delbaere
I,
Verstraelen
H,
Goetgeluk
S,
Martens
G,
De
Backer
G,
Temmerman 3M. Pregnancy outcome in primiparae of advanced maternal age. Eur J Obstet3Gynecol Reprod Biol 2007;135:41–6.3[34] Basso O, Wilcox AJ. Might rare factors account for most of the mortality of 3preterm babies. Epidemiology 2011;22:320–7.3[35] Fortes Filho J, Costa M, Eckert G, Santos P, Silveira R, Procianoy R. Maternal3preeclampsia protects preterm infants against severe retinopathy of prema-3turity.
J
Pediatr
2011;158(37):2–6.3[36] Love ER, Crum J, Bhattacharya S. Independent effects of pregnancy induced3hypertension on childhood development: a retrospective cohort study. Eur J3Obstet Gynecol Reprod Biol 2012;165:219–24.3[37] Mann JR, McDermott S, Griffith MI, Hardin J, Gregg A. Uncovering the complex3relationship
between
pre-eclampsia,
preterm
birth
and
cerebral
palsy.
Pae-3diatr Perinat Epidemiol 2011;25:100–10.3[38] VanderWeele TJ, Mumford SL, Schisterman EF. Conditioning on intermediates3in perinatal epidemiology. Epidemiology 2012;23:1–9.3[39] Auger N, Delezire P, Harper S, Platt R. Maternal education and stillbirth:3estimating gestational-age-specific and cause-specific associations. Epidemi-3ology 2012;23:247–54.
4[40]
Schempf
AH,
Branum
AM,
Lukacs
SL,
Schoendorf
KC.
Maternal
age
and
parity-4associated risks of preterm birth: differences by race/ethnicity. Paediatr4Perinat Epidemiol 2007;21:34–43.
N. Auger et al. / European Journal of Obstetrics & Gynecology and Reproductive Biology xxx (2014) xxx–xxx 5
G Model
EURO 8453 1–5
Please cite this article in press as: Auger N, et al. Gestational age-dependent risk factors for preterm birth: associations with maternal
education and age early in gestation. Eur J Obstet Gynecol (2014), http://dx.doi.org/10.1016/j.ejogrb.2014.02.035