Environmental Pollution 225 (2017) 700e712

Contents lists available at ScienceDirect

Environmental Pollution
journal homepage: www.elsevier.com/locate/envpol

Temperature exposure during pregnancy and birth outcomes: An

updated systematic review of epidemiological evidence*
Yunquan Zhang a, Chuanhua Yu a, b, *, Lu Wang a
a
Department of Epidemiology and Biostatistics, School of Health Sciences, Wuhan University, 185 Donghu Road, Wuchang District, Wuhan 430071, China
b
Global Health Institute, Wuhan University, 8 Donghunan Road, Wuchang District, Wuhan 430072, China

a r t i c l e i n f o a b s t r a c t

Article history: Seasonal patterns of birth outcomes have been observed worldwide, and there was increasing evidence
Received 9 January 2017 that ambient temperature played as a trigger of adverse birth outcomes, such as preterm birth (PTB), low
Received in revised form birth weight (LBW), and stillbirth. To systematically review updated epidemiological evidence about the
24 February 2017
relationship between temperature exposure during pregnancy and PTB, LBW, and stillbirth, we searched
Accepted 28 February 2017
Available online 9 March 2017
for related studies published in English from electronic databases and references of identified papers. We
only included original articles that directly reported the effects of prenatal temperature exposure on
birth outcomes. The characteristics and main findings of included studies were examined. A total of 36
Keywords:
Temperature
epidemiological studies were finally included in this review. Most of these studies focused on PTB and
Pregnancy LBW, while less attention has been paid to stillbirth that was relatively rare in the occurrence. Several
Preterm birth designs including ecological (e.g., descriptive and time-series) and retrospective cohort studies (e.g.,
Low birth weight case-crossover and time-to-event) were applied to assess temperature effects on birth outcomes. Tem-
Stillbirth perature metrics and exposure windows varied greatly in these investigations. Exposure to high tem-
Review perature was generally found to be associated with PTB, LBW, and stillbirth, while several studies also
reported the adverse impact of low temperature on birth outcomes of PTB and LBW. Despite no
conclusive causality demonstrated, the current evidence for adverse effect on birth outcomes was
stronger for heat than for cold. In summary, the evidence linking birth outcomes with ambient tem-
perature was still very limited. Consequently, more related studies are needed worldwide and should be
conducted in diversified climate zones, so as to further ascertain the association between temperature
and birth outcomes. Future studies should focus on more sophisticated study designs, more accurate
estimation of temperature exposure during pregnancy, and more efficient methods to find out the
exposure windows, as well as cold-related effects on birth outcomes.
© 2017 Elsevier Ltd. All rights reserved.

1. Introduction
Maternal and child health is one of the most important global
public health problems concerning social development and human
generation, which has been listed as a notable part of United Na-
tions Millennium Development Goals (GBD 2015 SDG
Collaborators, 2016). As a widely existing health threat, adverse
birth outcomes can produce great disease burden on fetal health
status at birth and future growth and development (Saigal and
*
This paper has been recommended for acceptance by David Carpenter.
* Corresponding author. Department of Epidemiology and Biostatistics, School of
Health Sciences, Wuhan University, 185 Donghu Road, Wuchang District, Wuhan
430071, China.
E-mail address: YuCHua@whu.edu.cn (C. Yu).
Doyle, 2009). Preterm birth (PTB), for instance, defined as the
birth of an infant prior to 37 weeks’ gestation, has been linked with
a variety of adverse health outcomes, such as acute morbidity or
mortality, and long-term defects or problems in growth, behavior
and cognition (Saigal and Doyle, 2009). The estimated number of
PTB infants was approximately 15 million in 2010 worldwide,
which accounted for 11% of all live births (Blencowe et al., 2012). As
for another common adverse birth outcome, low birth weight
(LBW) refers to births less than 2500 g and also contributes a lot to
infant mortality and childhood morbidity (McCormick, 1985), as
well as individual health throughout the life course (Grace et al.,
2015). It was reported by WHO that in 2004e2005, 15.5% (more
than 20 million) of all infants were born with LBW worldwide, and
95.6% of the total LBW babies occurred in developing countries
(WHO, 2004).

http://dx.doi.org/10.1016/j.envpol.2017.02.066
0269-7491/© 2017 Elsevier Ltd. All rights reserved.
 Y. Zhang et al. / Environmental Pollution 225 (2017) 700e712 701

Adverse birth outcomes are generally considered induced by the

complex effects of multiple risk factors including genetic, behav-
ioral, socio-economic and environmental factors (Strand et al.,
2011). As one of the most important environmental stressors,
climate change has been widely regarded as the greatest global
threat for human health in the 21st century because of its broad
direct or indirect impacts on the vast majority of the world's pop-
ulation (Anthony et al., 2009; Huang et al., 2013). Numerous pre-
vious epidemiologic studies (Gasparrini et al., 2015; Gronlund et al.,
2014; Ma et al., 2015; Ye et al., 2012) including some national and
multi-country investigations have linked a series of health out-
comes (e.g., morbidity and mortality) with temperature extremes
(i.e., cold and heat). Among the potential subpopulations vulner-
able to weather changes, pregnant women have received less
attention (He et al., 2016) compared with the elderly (Bunker et al.,
2016).
In very recent years, there was increasing evidence showing that
ambient temperature exposure during pregnancy could also trigger
the occurrence of PTB, LBW, and stillbirth in some climate zones
(Basu et al., 2016; He et al., 2016; Ngo and Horton, 2016; Schifano
et al., 2016). In the context of great burden due to adverse birth
outcomes and global climate change, it is of great necessity and
Fig. 1. Flow chart of article select process.
urgency to clarify the relationship between fetal birth outcomes
and prenatal temperature exposure. In this paper, we therefore
conducted an updated systematic review of epidemiological evi- number of aspects of the research publications, which included
dence that linked ambient temperature with birth outcomes. Based clear focus on the addressed issue, acceptable recruitment strategy,
on the findings and limitations of identified studies, we also made appropriate study method, measurement bias, identification and
several recommendations about the research direction for future treatment of confounding factors, believable results, and consis-
investigations in this field. tency with other available evidence (see Box 1). Since question 6
(Q.6) related to the follow up of included subjects, and was inap-
2. Materials and methods propriate for studies linking birth outcomes with ambient

2.1. Search strategies and article selection

Box 1
We conducted a systematic search of current epidemiologic Items of CASP guidelines used for quality assessment of
literature published in peer-reviewed English journal. We included studies in this review.
restricted the online publication time up to November 2016, and
searched electronic databases including PubMed, Medline, Scopus,
Web of Science and Google Scholar. The following keywords or
CASP guidelines Score 0/1
combinations of them were used in the search strategies: “preterm (Section A) Are the results of the study valid?
birth”, “gestational age”, “birth weight”, “stillbirth”, “birth out- Q1. Did the study address a clearly focused issue? 0/1 (No/
comes”, “climate change”, “temperature”, and “weather”. Yes)
In this review, we only included original articles that directly Q2. Was the cohort recruited in an acceptable way? 0/1 (No/
Yes)
assessed the correlation or relationship between prenatal temper- Q3. Was the exposure accurately measured to minimise 0/1 (No/
ature exposure (e.g., ambient temperature, heat waves, or cold bias? (prenatal temperature exposure) Yes)
spells) and one or more birth outcomes (i.e., preterm birth, LBW, Q4. Was the outcome accurately measured to minimise 0/1 (No/
and stillbirth). To include all of the related studies as comprehen- bias? (fetal birth outcomes) Yes)
Q5. a. Have the authors identified all important confounding 0/1 (No/
sively as possible, we also examined the references of the articles
factors? Yes)
identified. Of the 1264 articles identified in the initial database b. Have they taken account of the confounding factors in 0/1 (No/
search, 1183 articles were excluded after screening the titles, 38 the design and/or analysis? Yes)
were further excluded after reviewing the abstracts and 7 were Q6. a. Was the follow up of subjects complete enough? NA
excluded after assessing the full texts. Consequently, a total of 36 b. Was the follow up of subjects long enough? NA
(Section B) What are the results?
papers were finally included in this review (Fig. 1). Q7. What are the results of this study? (Are the results 0/1 (No/
demonstrated clearly?) Yes)
2.2. Study quality assessment Q8. How precise are the results? 0/1 (No/
Yes)
Q9. Do you believe the results? (Are the results believable?) 0/1 (No/
We conducted the quality assessment of included studies in line
Yes)
with Critical Appraisal Skills Programme (CASP) guidelines for (Section C) Will the results help locally?
evaluating quantitative studies (CASP, 2003, 2013), which were Q10. Can the results be applied to the local population? 0/1 (No/
pioneered by the National Health Service in the UK and have been Yes)
widely applied to determine the study quality of quantitative Q11. Do the results of this study fit with other available 0/1 (No/
evidence? Yes)
publications in many research fields (e.g., nursing, medical, and
Q12. What are the implications of this study for practice? 0/1 (No/
midwifery) (Carolan-Olah and Frankowska, 2014). CASP guidelines (Are there any implications for practice?) Yes)
used in this review focused on three broad issues (i.e., Section A, B,
and C) when appraising included studies, and investigated a
702 Y. Zhang et al. / Environmental Pollution 225 (2017) 700e712
temperature exposure (Carolan-Olah and Frankowska, 2014). We
then omitted Q.6 when conducting study quality assessment using
the guidelines. Each question was assigned a score of 1 for Yes/
Present and 0 for No/Absent except Q.5, each section (i.e., a & b) of
which was allocated a score of 1 (a total of 2). Consequently, study
quality scores ranged from 0 to 12 assessed by the final appraisal
tool, and a score of 12 signified the highest quality (Box 1).
3. Results
3.1. Study distribution and quality score
We identified 36 studies that examined the impact of temper-
ature on birth outcomes, most of which were conducted in devel-
oped countries/areas (e.g., North America and Europe) and
published in very recent years (after 2010). From the perspective of
study designs, 17 studies performed the ecological design
(including 12 time-series studies) and 19 studies were conducted
using retrospective cohort design (including 4 case-crossover
studies and 5 time-to-event studies). As for the study topics, 24
studies investigated the relationship between temperature and
preterm birth/gestational age, 14 studies assessed the impact of
temperature on birth weight, and 4 studies were about stillbirth.
Compared with PTB and LBW, less attention has been paid to
stillbirth that was relatively rare in the occurrence of adverse birth
outcomes.
Study quality scores assessed by CASP guidelines ranged from 7
to 12, out of a total of 12 possible points. Despite different designs
employed, aims and objectives were addressed clearly in all the
included studies. Generally, more recent epidemiological studies
showed relatively higher quality scores, and studies using case-
crossover and time-to-event designs tended to score higher
points than those using ecological (including time-series) and other
retrospective cohort designs.
3.2. Prenatal temperature and preterm birth/gestational age
Table 1 describes the characteristics and main results of the
included studies investigating the relationship between ambient
temperature and preterm birth/gestational age. 11 studies were
conducted in Europe, 6 in North America (5 from USA and 1 from
Canada), 4 in Asia (2 from Israel and 2 from China), and 2 in
Australia. Study periods persisted from within a year (Porter et al.,
1999; Yackerson et al., 2008) to more than ten years (Auger et al.,
2011; Cox et al., 2016; He et al., 2016; Ngo and Horton, 2016;
Schifano et al., 2013; Walfisch et al., 2016; Wang et al., 2013) and
sample sizes of included study cohorts varied greatly from 7585 in
Barcelona, Spain (Dadvand et al., 2011) to 838,146 births in
Guangzhou, China (He et al., 2016). Study results were also
expressed differently in each instance. For gestational age, differ-
ence in the mean gestation length (Porter et al., 1999), correlation
(Flouris et al., 2009) and regression coefficients (Dadvand et al.,
2011; Ngo and Horton, 2016) were generally reported, while the
relative risk/odds risk/hazard ratio (RR/OR/HR) or percent change
associated with per 1-unit increase in temperature (Basu et al.,
2010; Schifano et al., 2016) and RRs/ORs/HRs relative to the refer-
ence temperature (Auger et al., 2014; Cox et al., 2016; Ha et al.,
2017; Strand et al., 2012) were mostly used to present the main
results for studies that focused on PTB.
A relationship between ambient temperature and preterm birth
was firstly reported in an ecological investigation conducted in
Brooklyn, USA (Lajinian et al., 1997). The authors found a significant
linear increasing trend for the rate of preterm labour associated
with increased heatehumidity index in four study periods. This
significant evidence linking increasing preterm birth with high
temperature exposure was continuously demonstrated in subse-
quent epidemic studies worldwide conducted in different climate
zones. Most of these studies were performed within a single city,
while two or more study areas were investigated in 4 studies
(Carmichael et al., 2014; Flouris et al., 2009; Ha et al., 2017; Schifano
et al., 2016). The researchers employed a variety of study designs,
which included time-series analysis focusing on PTB counts (Arroyo
et al., 2016b; Cox et al., 2016; Schifano et al., 2013; Vicedo-Cabrera
et al., 2014, 2015; Walfisch et al., 2016), case-crossover design
developed from case-control principle (Basu et al., 2010; Ha et al.,
2017), and time-to-event method taking both birth outcomes and
gestational age into account (Schifano et al., 2016; Strand et al.,
2012; Wang et al., 2013). Furthermore, they used a number of
different temperature indicators to define high temperature
exposure. Besides direct temperature metrics such as mean tem-
perature (Flouris et al., 2009; Strand et al., 2012; Vicedo-Cabrera
et al., 2015), minimum and maximum temperature (Arroyo et al.,
2016b; Cox et al., 2016; Flouris et al., 2009; Ha et al., 2017; Lee
et al., 2008; Liang et al., 2016; Strand et al., 2012), some studies
included apparent temperatures (indices of discomfort calculated
from both air and dew-point temperatures) (Basu et al., 2010;
Schifano et al., 2016; Vicedo-Cabrera et al., 2014) instead of air
temperatures, as well as satellite-based and model-predicted
temperatures (Kloog et al., 2015). Additionally, a combination of
meteorological data (e.g., temperature and humidity indices)
(Dadvand et al., 2011; Lajinian et al., 1997; Walfisch et al., 2016) and
heat waves defined through combining cut-off temperatures and
different durations (Dadvand et al., 2011; Schifano et al., 2013;
Wang et al., 2013), were also applied to assess high temperature
exposures associated with PTB.
Despite the existing great heterogeneity in terms of design and
approach, the above-mentioned studies consistently strengthened
the evidence that high temperature exposure during pregnancy
may trigger the occurrence of preterm birth in different climate
zones. Nevertheless, compared with the relatively well-
documented relation between high prenatal temperature and
increased PTB, several studies also observed cold-related increased
risks in preterm birth (He et al., 2016; Liang et al., 2016; Yackerson
et al., 2008). Notably, He et al. (2016) revealed the adverse effects of
both extreme cold and heat on PTB, and Liang et al. (2016) and
Yackerson et al. (2008) even reported a significant protective effect
of high temperatures on PTB occurrence. Furthermore, 4 more
studies showed inconsistent results and indicated null or insignif-
icant associations between prenatal temperature exposure and
fetal preterm birth (Lee et al., 2008; Ngo and Horton, 2016; Porter
et al., 1999; Wolf and Armstrong, 2012).
3.3. Prenatal temperature and birth weight
As demonstrated in Table 2, most of the included studies
observed the impacts of exposure to high prenatal temperature on
decreased birth weight, regardless of using various designs and
assessment methods. Three ecological studies conducted in
different regions (Flouris et al., 2009; Matsuda et al., 1998; Wells
and Cole, 2002), including a worldwide investigation covering
140 populations (Wells and Cole, 2002), found significant negative
correlations between birth weight and temperature exposure, and
the correlation coefficients ranged from
0.22 (Flouris et al., 2009)
in Greece to
0.63 (Matsuda et al., 1998) in Japan (
0.59 for
worldwide). Similar results were observed for female and male
infants (Flouris et al., 2009), while there might be some potential
differences between whites and blacks in temperature effects
(Desche^nes et al., 2009). High temperature effects on birth weight
may vary by gestational stages (Arroyo et al., 2016a; Grace et al.,
2015; Kloog et al., 2015; Lawlor et al., 2005). For instance, a 1- C
Table 1
Characteristics of the included studies on ambient temperature and preterm birth/gestational age.

Study Location Study design Sample Exposure Covariates adjusted Statistical method and result Study
measurement for qualityscore
Statistical method Statistic Estimate
(0e12)

Europe
*
Lee et al. London, UK Ecological 482,568 singleton Daily max and min Long-term trend, Time-series logistic Risk change per 1  C Max temperature: 9
(2008) live births, 1988 temperature at the seasonality, day of regression increase OR ¼ 1.00 (95%CI:
e2000 time of birth the week, public 0.99e1.00, p > 0.05)
holiday Min temperature:
OR ¼ 1.00 (95%CI:
1.00e1.00, p > 0.05)
Flouris Greece Ecological 516,874 live births, Mean temperature No Correlation analysis Correlation Both sex: r ¼
0.210 8
et al. (2009) 1999e2003 during the birth coefficient between (p < 0.001)
month temperature and Males: r ¼
0.208
gestational age (p < 0.001)
Females: r ¼
0.211
(p < 0.001)
Dadvand Barcelona, Retrospective 7585 singleton births Heatehumidity index Maternal Linear regression Gestational age 5.3-day reduction 12
et al. (2011) Spain cohort spontaneous labour, demographic and model change after high (95% CI:
10.1 to

Y. Zhang et al. / Environmental Pollution 225 (2017) 700e712

2002e2005 clinical heat index exposure
0.05, p ¼ 0.03)
characteristics, and on the day before
infant sex delivery
*
Wolf and Two German Ecological All reported hospital Daily mean Long-term trend, Logistic time-series Temperature effect No clear evidence for 9
Armstrong States singleton births from temperature seasonality, day of regression combined (ORs) as a linear and a an association
(2012) Brandenburg (2002 the week with constrained categorical variable between temperature
e2010) and Saxony distributed lag model and PTB was found
(2005e2009) (p > 0.05)
*
Schifano et al. Rome, Italy Ecological All singleton live Maximum apparent Long-term trend, Poisson generalized Percent change During heat 10
(2013) births by natural temperature (MAT) seasonality, days of additive model during heat waves waves: þ19% increase
delivery, 2001e2010 and heat waves in the holiday, and air combined with and per 1  C increase (95% CI: 7.91e31.69)
month preceding pollution distributed lag model in MAT Per 1  C increase in
delivery MAT: 1.9% (95%CI:
0.86e2.87)
*
Vicedo-Cabrera Valencia, Ecological 20,148 singleton MAT and daily Long-term trend, Quasi-Poisson Percent change in risk 20% increase when 10
et al. (2014) Spain natural births during minimum seasonality, day of generalized additive relative to median MATthe 90th
the warm season temperature the week, public models combined temperature percentile two days
(MayeSeptember), holiday, and with distributed lag before delivery
2006e2010 relative humidity non-linear models 5% increase when
minimum
temperature90th
percentile in the last
week
*
Vicedo-Cabrera Stockholm, Ecological All singleton Daily mean Long-term trend, Quasi-Poisson Cumulative risk ratio Mean 10
et al. (2015) Sweden spontaneous births temperature during seasonality, day of generalized additive relative to median temperature ¼ 75th
collected from the the last month of the week, public models combined temperature percentile: RR ¼ 2.50
Swedish Medical gestation holiday, and with distributed lag (95% CI: 1.02e6.15)
Birth Register, 1998 relative humidity non-linear models
e2006 (gestational
age 22weeks)
*
Arroyo et al. Madrid, Spain Ecological 298,705 live Daily maximum Linear trends, Autoregressive over- Relative risks(RRs) for RR ¼ 1.055 (95%CI: 10
(2016b) singleton births, 2001 temperature seasonality, and the dispersed Poisson interquartile increase 1.018e1.092)
e2009 autoregressive regression models in temperature
nature of the series,
day of the week
*
Cox et al. Flanders, Ecological 807,835 live-born Daily minimum and Long-term trend, Quasi-Poisson Percent increase in Extreme heat (99th 11
(2016) Belgium singleton births with maximum air seasonality, day of generalized additive risk relative to vs. 50th percentile):
a gestational age temperature the week, public models combined median temperature 15.6% (95% CI: 4.8

703
(continued on next page)
Table 1 (continued )

704
Study Location Study design Sample Exposure Covariates adjusted Statistical method and result Study
measurement for qualityscore
Statistical method Statistic Estimate
(0e12)

between 22 and 42 holiday, humidity, with distributed lag e27.6) for minimum
weeks (January 1998 influenza, and air non-linear models temperature; 14.5%
eJuly 2011) pollution (95% CI: 0.5e30.6) for
maximum
temperature
Schifano et al. Rome and Retrospective Two cohorts of MAT Time trend, Cox proportional Gestational week- 22nde26th weeks: 12
(2016) Barcelona cohorty singleton live births: seasonality, and hazards model specific hazard ratios 1.071 (95% CI: 1.052
78,633 births in maternal associated with a 1- e1.091) in Rome and
Rome, Italy (2001 demographic and unit increase in MAT 1.071 (95% CI: 1.036
e2010) and 27,255 clinical e1.106) in Barcelona
births in Barcelona, characteristics 36th week: 1.032
Spain (2007e2012) (95% CI: 1.026e1.038)
in Rome and 1.033
(95% CI: 1.020e1.045)
in Barcelona
North America

Y. Zhang et al. / Environmental Pollution 225 (2017) 700e712

Lajinian et al. Brooklyn, USA Ecological Hospital birth data, Weekly average heat No Exact test for linear Adjusted preterm 1.2% in the coldest 8
(1997) 1993e1994 ehumidity index in trend labor rate for each period and 3.0% in the
the hottest and period hottest period; linear
coldest week of association test:
summer and winter p < 0.002
Porter et al. Chicago, USA Retrospective 11,792 singleton MAT at the time of Maternal ethnicity, One-tailed t-test Mean gestational age No association was 7
(1999) cohort births during warm birth educational status, of births in different found between
months (June and community MAT strata increasing MAT and
eAugust in 1995) area median shortened gestation
household income length
Basu et al. California, Retrospective 58,681 singleton Weekly average Long-term trend, Conditional logistic Percent increase in 8.6% increase (95%CI: 12
(2010) USA cohort # births spanning 16 apparent seasonality, day of regression models risk per 10  F (5.6  C) 6.0e11.3); Greater
counties (May to temperature at the the week, air increase associations were
September 1999 time of birth pollution, maternal observed for younger
e2006) ethnicity, age, mothers, blacks, and
education and Asians
infant's sex
Auger et al. Montreal, Retrospective 206,929 term and Heat episodes Birth period, Cox proportional Hazard ratio during Preterm delivery: 12
(2014) Canada cohorty 12,390 preterm (maximum month, air hazards model heat episodes relative 0.92 (95%CI: 0.74
singleton live births, temperature32  C, pollution, humidity, to other days e1.14)
JuneeSeptember duration3 days) maternal age, Early term
1981e2010 during the preceding education, marital delivery:1.17 (95%CI:
week status, 1.06e1.29)
immigration,
language spoken at
home, parity
Carmichael 957 counties Ecological Preterm delivery (20 Mean January Socioeconomic and Multivariable Percentage change in Cold climate for 10
et al. (2014) in US e36 weeks' temperature and demographic population-weighted risk of preterm blacks:0.018 (20e31
gestation) among mean July Census variables linear regression delivery per 1  C weeks,
singletons born temperature and health-related analysis increase p < 0.001);0.047 (32
during 1998e2002 variables from birth e36 weeks,
certificates p < 0.001)
Warm climate for
whites:0.010 (20e31
weeks,
p < 0.001);0.060 (32
e36 weeks,
p < 0.001)
12
Kloog et al. Massachusetts, Retrospective 450,407 live Satellite-based and Traffic density, air Linear and logistic PTB odds ratio per an OR ¼ 1.02(95% CI:
(2015) USA cohort singleton births(>22 model-predicted pollution, percent mixed models and interquartile range 1.00e1.05)
weeks of gestation), temperature of open space, accelerated failure increase (8.4  C) in
2000e2008 (1  1 km) and median income for time models temperature
ground-based every census tract,
temperature from individual-level
monitoring stations. sociodemographic
factors of the
mothers and
infants
Ha et al. 12 US sites Retrospective Medical records from Weekly average Conception month, Poisson regression Relative risk for hot Early preterm (<34 12
(2017) cohort # 223,375 singleton temperature humidity, site, sex, with generalized (>90th percentile)/ weeks): RRhot ¼ 1.11
deliveries from 12 maternal estimating equations cold (<10th (1.01e1.21);
clinical centers, 2002 demographics, and conditional percentile) exposures RRcold ¼ 1.20 (1.11
e2008 parity, insurance, logistic regression compared with mild e1.30)
pre-pregnancy BMI, temperature (10- Late preterm (34e36
pregnancy 90th percentile) weeks):
complications, and RRcold ¼ 1.09 (1.04
smoking or e1.15)
drinking during

Y. Zhang et al. / Environmental Pollution 225 (2017) 700e712

pregnancy.
Ngo and Manhattan, Retrospective Restricted birth Temperature bins Birth year, birth Fixed effect Gestational age No meaningful, 11
Horton New York, USA cohort certificate data in defined by average month, air regression model change per 1 extra significant
(2016) New York, 1985 daily temperature pollution, infant's day where association was found
e2010 sex and mother's temperature <25  F
age, education, and >85  F during
ethnicity, marital pregnancy
status, smoking,
number of previous
deliveries
Asia
Yackerson et al. Negev, Retrospective 11,979 births at the Monthly averaged Time trend, relative Multivariate Poisson PTB occurrence Regression 9
(2008) Beer-heva, cohort Soroka Medical value of daily humidity, wind regression change per 1  C coefficient:
0.09
Israel Centre during in 1999 maximum speed and direction increase (p ¼ 0.008)
temperature
He et al. (2016) Guangzhou, Retrospective 838,146 singleton Weekly mean Long-term trend Cox proportional PTB risk increase Extreme cold (7.6  C): 12
China cohorty vaginal births, 2001 temperature and seasonality, hazards model relative to the median 17.9% (95% CI: 10.2
e2011 humidity, air temperature (24.4  C) e26.2)
pollution, Extreme heat
infant'ssex and (31.9  C): 10.0% (95%
maternal CI: 2.9e17.6)
sociodemographic
variables
*
Liang et al. Shenzhen, Ecological 58,411 singleton PTB Daily mean Seasonal and long- Distributed lag non- Cumulative RR (up to 9  C (1st): 1.72 (95% 10
(2016) China (daily 22.85), 2005 temperature term trends, day of linear model 30 days) relative to CI: 1.28e2.33)
e2011 the year, day of the median 12.5  C(5th): 1.96
week, public temperature(24.5  C) (95% CI: 1.60e2.39)
holiday, humidity 29.9  C(95th): 0.69
and atmospheric (95% CI: 0.60e0.80)
pressure 30.7  C(99th): 0.62
(95% CI: 0.52e0.74)
*
Walfisch et al. Negev, Ecological 263,709 deliveries at Heat stress defined by Long-term and Time-series Poisson Incidence risk per Total PTD: RR ¼ 1.06 10
(2016) southern Israel the Soroka University the average of wet seasonal trends regression model one-unit increase in (95% CI: 1.04e1.08,
Medical Center, 1988 and dry temperature. heat-stress index p < 0.001)
e2012 Spontaneous PTD:
RR ¼ 1.07 (95% CI:
1.05e1.10, p < 0.001)
Induced PTD:
(continued on next page)

705
 706 Y. Zhang et al. / Environmental Pollution 225 (2017) 700e712

increase in temperature was linked with a decrease of 5.4 g (95% CI:

2.9e7.9) in birth weight during the first trimester (Lawlor et al.,
qualityscore

2005), while 2.0 g (95%CI: 0.4e3.5) during the last trimester

(Kloog et al., 2015). However, in a time-series ecological study
(0e12)
Study

conducted in 19 African countries, birth weight changes peaked in

11

12
the 2nd pregnancy trimester with a 0.90-g decline associated with
per hot day increase (100  F) (Grace et al., 2015).

HRs ranged from 1.13

(95% CI: 1.03e1.24) to
0.97e1.01, p ¼ 0.699)

was associated with

Higher temperature

different heatwave
HR ¼ 0.96 at 15  C
RR ¼ 0.99 (95% CI:

Similar to temperature-PTB associations, inconsistent evidence

2.00 (95% CI: 1.37

and 1.02 at 25  C.
shorter gestation.

e2.91) by using
was also observed for temperature effects on birth weight in
several studies. Low temperatures rather than high temperatures

definitions
Estimate

were linked with decreased birth weight in two previous retro-

spective cohort studies, with sample sizes of 3333 and 447,499
singleton live births (Elter et al., 2004; Murray et al., 2000). More
recently, despite no significant temperature-PTB associations
Hazard ratio relative

observed, prenatal exposures to both cold and heat were regarded

compared to non-
Adjusted hazard

as the potential independent risk factor to trigger low birth weight

heatwave days

heatwave days
ratios during

in New York City (Ngo and Horton, 2016). In addition, there were
another three studies showing no effect of ambient temperature on
to 21  C
Statistic

birth weight (Diaz et al., 2016; Tustin et al., 2004; Wolf and
Statistical method and result

Armstrong, 2012), one of which demonstrated no clear evidence

for an relation between temperature and PTB as well (Wolf and
Armstrong, 2012).
hazards regression
Statistical method

Cox proportional

Cox proportional
hazards model

3.4. Prenatal temperature and stillbirth

Table 3 outlines the characteristics and findings of the 4

Notes: *: Time-series study; #: case-crossover study; y: time-to-event study; PTB: preterm birth; MAT: maximum apparent temperature.
model

included studies assessing the impact of temperature exposure

during pregnancy on fetal stillbirth. These studies were all con-
Covariates adjusted

ducted in developed areas and published very recently. They

baby's gender, and
sociodemographic

Indigenous status,
pollution, infant's
sex and maternal
Long-term trend,

Long-term trend,

economic index,
pollution, area-

consistently reported increased RRs of stillbirth related to prenatal

marital status,
maternal age,
humidity, air

humidity, air

heat exposure, while there was no standardized definition of still-

seasonality,

seasonality,

level social
variables

birth (Arroyo et al., 2016a; Basu et al., 2016; Strand et al., 2012).
parity

Strand et al. (2012) conceived a time-to-event design taking

for

advantage of Cox proportional hazards models and firstly provided

strong evidence of an association between high ambient temper-
durations (at least 2, 3
Heatwave defined by

temperatures (90th,

ature and increased risk of stillbirth in Brisbane, Australia. Further

and 4 consecutive
a series of cut-off

time-stratified case-crossover studies conducted in North America

daily maximum
calculated from

percentile) and
95th and 98th
Weekly mean
measurement

(Auger et al., 2016; Basu et al., 2016), observed greater risks of

temperature

hourly data

stillbirth for male fetuses, younger mothers, and less educated

Exposure

mothers (Basu et al., 2016), and linked elevated outdoor tempera-

days)

ture with term stillbirth (gestational weeks  37) due to undeter-

mined and maternal causes, but not other causes (Auger et al.,
2016).
50,848 spontaneous
singleton live births
between November
births registered in

and March in 2000

101,870 singleton

Collection, 2005
the Queensland
Perinatal Data

4. Discussion
Sample

In the present study, we reviewed the current epidemiologic

e2009

e2010

evidence examining the relationship between prenatal tempera-

ture and birth outcomes. Under the background of global climate
change, these findings may have great implications in public
Retrospective

Retrospective
Study design

decision-making regarding maternal and child health to reduce

cohorty

cohorty

temperature-related adverse birth outcomes. Up to now, however,

the evidence was still very limited in this research area. More well-
designed studies were thus needed worldwide and several points
as follows should be taken into account in the future.
Brisbane,

Brisbane,

4.1. Study design and statistical approach

Australia

Australia
Location

Several design types have been used to study the effect of

Table 1 (continued )

temperature exposure on birth outcomes. Generally, earlier in-

vestigations were descriptive and conducted relatively simple sta-
Strand et al.

Wang et al.

tistical analyses (e.g., t-test, ANOVA, and correlation analysis)

(2012)

(2013)
Australia

(Matsuda et al., 1998; Porter et al., 1999; Tustin et al., 2004). Recent
Study

epidemiological studies were often time-series or retrospective

cohort designs and a number of multivariate regression models
Table 2
Characteristics of the included studies on ambient temperature and birth weight.

Study Location Study design Sample Exposure measurement Covariates adjusted for Statistical method and result Study
qualityscore
Statistical method Statistic Estimate
(0e12)

Matsuda Japan Ecological Hospital births in 1982 Mean temperature in No Simple correlation Correlation coefficient r ¼
0.63 (p < 0.01) 7
et al. (1998) each prefecture (47 analysis between temperature
included) and birth weight
Murray Northern Retrospective 447,499 singleton live Mean daily max Year and month of birth, Linear regression Birth weight change per Females:3.50g (95%CI: 9
et al. (2000) Ireland cohort births after 36 weeks of temperatures in the duration of gestation, model 1  C increase 1.78e5.22)
pregnancy, 1971e1986 second trimester maternal age, number of Males:1.02g (95%
previous pregnancies, CI:
0.06e2.75)
sex, and social class of
infants at birth
Wells and Cole Worldwide Ecological Births in 140 Heat stress/index latitude, altitude, gross Stepwise linear Percent change in birth
2.7% (95%CI:
3.8 to 10
(2002) populations (5558, domestic product, regression weight per 1-unit
1.6) (r ¼
0.59,
median sample size) average daily energy increase in heat index p < 0.01)
intake, maternal height
and infant mortality
Elter et al. Marmara, Retrospective 3333 singleton live Daily mean Maternal age and parity, Stepwise multiple Birth weight change in Regression 9

Y. Zhang et al. / Environmental Pollution 225 (2017) 700e712

(2004) Turkey cohort births after 36 temperature during the mode of delivery, infant's regression analysis birth weight per 1  C coefficient ¼ 0.001
completed weeks of second pregnancy sex, humidity, rainfall and increase (p ¼ 0.018)
pregnancy, 1992e2003 trimester daylight
Tustin et al. Dunedin, Retrospective 7039 full-term births, Mean temperature in No One-way ANOVA Mean birth weight F ¼ 0.03, p > 0.05. There 7
(2004) New Zealand cohort 1999e2003 the second trimester between peak and was no effect of
trough temperature ambient temperature
periods on birth weight.
Lawlor et al. Aberdeen, Retrospective 12,150 live births, 1950 Mean temperature in Maternal age, infant's sex, Linear regression Birth weight change per 1st trimester: 5.4g (95% 11
(2005) Scotland cohort e1956 the mid 10-day period birth year, birth order and model 1  C increase CI:
7.9 to
2.9)
of each trimester social class 2nd trimester:1.8g (95%
CI: 0.7e4.3)
3rd trimester:1.3g (95%
CI: 0.50e2.1)
Desche^nes Colombia, USA Retrospective 37.1 million singleton Mean daily maximum Maternal age, fertility Fixed effect regression Percent change in birth Whites:
0.0022% 10
et al. (2009) cohort births, 1972e1988 and minimum history, marital status, model weight per hot day (p < 0.05)

temperatures in each education attainment, increase (85 F) in the Blacks:
0.0036%
pregnancy trimester long-term trend and whole pregnancy (p < 0.05)
seasonality trimester
Flouris et al. Greek Ecological 516,874 live births, Mean temperature No Correlation analysis Correlation coefficient Both sex: r ¼
0.218 8
(2009) 1999e2003 during the birth month between temperature (p < 0.001)
and birth weight Males: r ¼
0.217
(p < 0.001)
Females: r ¼
0.218
(p < 0.001)
*
Wolf and Two German Ecological All reported hospital Daily mean Long-term trend, Logistic time-series Temperature effect No clear evidence for an 9
Armstrong States singleton births from temperature seasonality, day of the regression combined (ORs) as a linear and a association between
(2012) Brandenburg (2002 week with constrained categorical variable temperature and LBW
e2010) and Saxony distributed lag model was found (p > 0.05)
(2005e2009)
*
Grace et al. (2015) 19 African Ecological 67,769 recorded birth Gridded monthly Mothers' characteristics Generalized linear Birth weight change per 1st trimester: 0.63g 10
countries weights from temperature derived (age, height, marital model and standard hot day increase (p < 0.01)
demographic and from satellite and status, education), child's linear regression (100  F) in each 2nd trimester: 0.90g
Health Surveys, 1986 ground-based weather characteristics (sex, birth pregnancy trimester (p < 0.001)
e2010 stations order, birth month), 3rd trimester: 0.44g
urban residence, rainfall, (p < 0.10)
etc.
Kloog et al. (2015) Massachusetts, Retrospective About 500 thousand Satellite-based and Traffic density, air Linear and logistic Birth weight change During the last 12
USA cohort live singleton model-predicted pollution, percent of open mixed models and and LBW odds ratio per trimester: 16.7 g lower

707
(continued on next page)
 708 Y. Zhang et al. / Environmental Pollution 225 (2017) 700e712

have been applied in these studies. Thus, these studies were able to
qualityscore

quantitatively estimate the effects of studied birth outcomes

associated with per 1-unit change in temperature. For time-series
(0e12)
Study

studies, researchers usually fitted autoregressive Poisson regres-

10

11
9
sion models (Arroyo et al., 2016a; Diaz et al., 2016; Walfisch et al.,

LBW (results not given)

(95% CI:
29.7 to e3.7)

shown to be related to
2016) or generalized additive models (Cox et al., 2016; Liang
pregnancy: OR ¼ 1.04

and cold waves is not

Average temperature

Average temperature
Temperature in heat
OR ¼ 1.003 (95% CI:
(95% CI: 0.96e1.13)

et al., 2016; Vicedo-Cabrera et al., 2014). To fully capture both the

nonlinear and delayed effects of temperature, the regression anal-
Over the entire

<25  F:
1.8g

>85  F:
1.7g
1.002e1.004)

ysis in some studies also incorporated with distributed lag

nonlinear model, which was recently developed by Gasparrini
Estimate

(2011) and widely applied in time-series analysis assessing

temperature-mortality relationship. Differentiating from time-

Birth weight change per

temperature <25  F and

maximum temperature

series studies based on population characteristics, retrospective

LBW odds ratio per an
accelerated failure time an interquartile range

Relative risks of LBW

pregnancy trimester

cohort studies focused on individual-level exposure and outcome.

increase (8.4  C) in

1 extra day where

per 1  C increase
10  C increase in

Generally, these studies selected Cox proportional hazards model

>85  F during
in the second

(Auger et al., 2014; He et al., 2016; Strand et al., 2012) from the
temperature

pregnancy
perspective of survival analysis and conditional logistic regression
Statistic

(Auger et al., 2016; Basu et al., 2010; Basu et al., 2016) on the basis of
Statistical method and result

case-crossover principle. In addition, generalized estimating

equations (Ha et al., 2017) and mixed models (linear or logistic)
Fixed effect regression
Autoregressive, over-

Autoregressive, over-

(Kloog et al., 2015) could also be the alternative statistical models in

Statistical method

dispersed Poisson

dispersed Poisson
regression model

regression model

some cases.
The great variety of study designs and statistical models may
result in the difficulty for comparing the results between these
models

studies (Strand et al., 2011). Nevertheless, none of the current

model

studies can establish definitive causal links between temperature

exposure and birth outcomes due to the limitations of epidemio-
minimum temperature the autoregressive nature
of the series, weeks of the

Birth year, month, infant's

space, median income for

sociodemographic factors

Linear trend, seasonality,

Linear trend, seasonality

marital status, smoking,

logic designs themselves. Ecological studies linked population-level

Exposure measurement Covariates adjusted for

and the autoregressive

sex, and mother's age,

education, ethnicity,

temperature exposure with birth outcomes (e.g., counts or rates)

number of previous
month, and pollinic

maximum (34  C) and nature of the series

of the mothers and
every census tract,

and were unable to assess the association at the individual level. It

individual-level

was thus not so appropriate and persuasive to detect an association

deliveries

by applying simple correlation analysis or ANOVA. There might

pollution
infants

exist some inevitable ecologic bias resulting from the study design
itself and a number of unadjusted potential confounders (e.g.,
seasonality). In time-series studies, seasonality and temporal
temperature (1  1 km)

temperatures (-2  C)

trends (long-term trend and short-term fluctuation) played a very

Heat and cold waves

<25  F, 25e45  F, 45
Daily maximum and
monitoring stations

defined by average
daily temperature:
defined by weekly
and ground-based
temperature from

e65 F, 65e85 F,
Temperature bins

important confounding role, and could be well eliminated by per-



forming a white noise test for model residual series in ARIMA

analysis (Arroyo et al., 2016a; Diaz et al., 2016) and conducting
minimum

model fit assessment or sensitivity analyses (Gasparrini and

<85 F



Armstrong, 2010) (e.g., varying degrees of freedom of natural cu-

bic spline function for time) to check the robustness of the results
gestation), 2000e2008

certificate data in New

298,705 live singleton

(gestation weeks37)

obtained from generalized additive models (Liang et al., 2016;

1492 non-premature
births(>22 weeks of

infants, 2001e2009
births, 2001e2009

Schifano et al., 2013; Vicedo-Cabrera et al., 2014, 2015). These

York, 1985e2010
Retrospective Restricted birth

time-series analyses provided flexible methods to investigate the

complex nonlinear association between temperature and health
such as birth outcomes (Gasparrini and Armstrong, 2010). How-
Sample

ever, time-series evidence could still have been limited by the

underpowered ecologic design due to inability of accounting for
Study design

bias arising from aggregation of individual-level data (Auger et al.,

*

Notes: *: Time-series study; LBW: low birth weight.

Ecological

Ecological

2014).
cohort

Case-crossover design is a useful emerging epidemiological

method in assessing the acute effects of temperature exposures on
birth outcomes (Auger et al., 2014; Basu et al., 2010) and could
New York, USA
Madrid, Spain

Madrid, Spain

adjust for both known and unknown confounders (e.g., time-

Manhattan,

invariant individual characteristics) by matching each individual

Location

as his or her own control (Maclure, 1991; Maclure and Mittleman,

2000). In addition, long-term trends as well as day of the week
could be further controlled for in time-stratified case-crossover
Table 2 (continued )

studies, since control periods are usually selected every a week

Diaz et al. (2016)

Ngo and Horton

during the same calendar month (Auger et al., 2016; Basu et al.,
Arroyo et al.

2016). However, case-crossover analyses do not account for

(2016a)

(2016)

fetuses-at-risk, and results could be biased if days after delivery are

Study

selected as control periods because cases are no longer at risk

(Auger et al., 2014). Time-to-event design (i.e., survival analysis)
 Y. Zhang et al. / Environmental Pollution 225 (2017) 700e712 709

employed in some recent studies included gestational age as the

qualityscore
time variable (i.e., survival time) and birth outcomes (e.g., pre-
term/term birth, live birth/stillbirth) as the binary survival
(0e12)
Study

Higher ambient temperatures 11 outcome variable. This approach could simultaneously account

Maximum temperature: 1.012 10

12

12
for the impacts of both time-dependent exposures (e.g., air pol-

younger mothers, less educated

pregnancy increased the risk of

28  C relative to undetermined causes and 1.46

lutants, and humidity) and individual risk factors (e.g., behavioral,

Minimum temperature: 1.037

stillbirth. The hazard ratio for

in risk for every e16.8); Risk were greater for

(95% CI: 1.010e1.013) in the

(95% CI: 1.035e1.039) in the

Odds ratios at 1.19 (95% CI: 1.02e1.40) for

and metabolic), and was thus expected to have combined the

Percent change 10.4% increase (95% CI: 4.4

mothers, and male fetuses

stillbirth was 0.3 at 12  C
relative to 21  C in the last 4 weeks of the

advantages of time-series analysis as well as cohort study (Wang

(95% CI: 1.03e2.07) for

et al., 2013). Also, time-to-event study design had relatively high
degree of statistical power because all fetuses at risk rather than

maternal causes
2nd trimester;

3rd trimester
births only were accounted for in the survival analysis (Auger
et al., 2014; Wang et al., 2013). Further, by including gestational
Estimate

age as the time scale in the survival analysis, time-to-event

design accounted for the fact that the likelihood of giving birth
increases with increasing gestational weeks and allowed for

10  F (5.6  C)
comparing the risks of birth among fetuses with the same
Hazard ratio

Autoregressive, Relative risk

Statistical method and result

gestational age. This could thus avoid the potential confounding

over-dispersed per 10  C
increase

increase
Statistic

by different gestational ages, which other study designs and

20  C

statistical models have paid less attention to (Strand et al., 2012).

Therefore, survival analysis methods are recommended for future
regression and
hazards model

meta-analysis

studies when evaluating the association between time-varying

proportional

Conditional

Conditional
regression

regression
Statistical

exposures and birth outcomes (Auger et al., 2014; Strand et al.,

method

Poisson

logistic

logistic
model

2012).
Cox

4.2. Exposure assessment and exposure window

Long-term trend, seasonality, humidity,

the week, maternal characteristics (e.g.,

California, USA Retrospective 8510 fetal deaths (gestation20 Mean apparent Long-term time trends, day of the week,

factors (e.g., age, ethnicity, education),

maternal age, smoking, marital status,

Year and season of conception, day of

variables including sex of the infant,

age, ethnicity, socioeconomic status,

air pollution, maternal demographic

air pollution, and sociodemographic

autoregressive nature of the series

Linear trends, seasonality, and the
indigenous status, and pregnancy

The majority of identified studies used the averages of

monthly, weekly, or daily temperature collected from ground
weather stations as the exposure metrics. As demonstrated in
Covariates adjusted for

smoking, comorbidity)

previous investigations, atmosphere environmental factors

itself, day of the week

including air pollutants and ambient temperature could also vary

greatly both spatially and temporally (Meng et al., 2016; Wang
complications

et al., 2016). The use of station-based temperature exposure can

introduce inevitable measurement bias and exposure misclassi-
fetal sex

fication, which may have limited the researchers’ ability to

accurately estimate temperature-related effects on health out-
comes (Tian et al., 2012). In recent years, exposure assessments
maximum and
Weekly mean
measurement

temperature

temperature

temperature

weeks) during the warm season temperature

based on satellite remote sensing and GIS technologies (e.g., land

maximum
minimum
Exposure

use regression) were applied in more and more studies focusing

Daily

Daily

on health impacts (including birth outcomes) of exposure to

ambient pollution (Estarlich et al., 2016; Hyder et al., 2014;
Lakshmanan et al., 2015; Wang et al., 2016; Zhang et al., 2016).
Perinatal Data Collection, 2005

298,705 live singleton births,

registered in the Queensland

(MayeOctober), 1999e2009

However, very limited studies have taken advantage of satellite-

Characteristics of the included studies on ambient temperature and stillbirth.

Retrospective 5047 stillbirths during the

eSeptember), 1981e2011

Notes: *: Time-series study; #: case-crossover study; y: time-to-event study.

Retrospective 101,870 singleton births

based predicting models to conduct individual-level exposure

assignment. Kloog et al. (2015) estimated satellite-based spatio-
warm months (April

temporally resolved and monitor-based temperature exposure,

and evaluated the association between ambient temperature and
2001e2009

birth outcomes using these two exposure assessment methods.

Sample

They observed some inconsistent and even contrary effects of

e2009

satellite-based and monitor-based temperatures on birth weight

and gestational age. Therefore, future studies should aim at more
Study design

accurate assessment of temperature exposure, so as to further

*
Ecological

and better understand how temperature metrics exactly affect

cohort #

#
cohorty

cohort

birth outcomes.
Various exposure windows have been used to investigate the
relationship between temperature and birth outcomes in previ-
Madrid, Spain

Continental

ous studies, because there was no consensus about which period

Brisbane,
et al. (2012) Australia
Location

et al. (2016) Quebec,

during the pregnancy was more vulnerable. Some studies limited

Canada

the exposure to the last week preceding delivery or several days

before birth (Auger et al., 2016; Auger et al., 2014; Lee et al., 2008;
et al. (2016)

Vicedo-Cabrera et al., 2014) and some focused on exposure dur-

ing the month preceding delivery (Flouris et al., 2009; Schifano
(2016a)
et al.

et al., 2013; Strand et al., 2012; Vicedo-Cabrera et al., 2015).

Arroyo
Strand

Auger
Study
Table 3

Basu

Another approach was to report the results for certain or each

trimester of the pregnancy (Arroyo et al., 2016a; Desche ^nes et al.,
710 Y. Zhang et al. / Environmental Pollution 225 (2017) 700e712
2009; Elter et al., 2004; Murray et al., 2000; Ngo and Horton, 2016).
Additionally, some investigators conducted several window-
specific stratified analyses separately and explored the potential
exposure windows in developed areas (Ha et al., 2017; Kloog et al.,
2015; Schifano et al., 2016). However, a great variation in exposure
windows could be expected given the diversified meteorological
characteristics in different climate zones. It could be difficult but of
great significance to identify exposure windows period by period
from the whole pregnancy. More efficient methods or models
should be developed to simplify the process of detecting exposure
windows.
4.3. Effect modifiers
Some previous ecological studies assessing the relationship
between prenatal temperature and birth outcomes seldom
adjusted for any potential confounders, thus may result in low-
grade or invalid evidence. As previous studies demonstrated,
possible effect modifiers could possibly include a number of factors,
such as maternal age, smoking or drinking status, previous preg-
nancies, pregnancy complications, air pollution, humidity,
ethnicity, pre-pregnancy BMI, and socio-economic status (Ha et al.,
2017; Kloog et al., 2015; Strand et al., 2011). These confounding
factors should be fully controlled in future studies so as to obtain
valid effect estimates close to reality.
Among the above-mentioned confounders, several factors
including maternal age, fetal gender, ethnicity, and socio-economic
status (e.g., education level) have been discussed by stratified an-
alyses, while the evidence was still of lack. Generally, younger
mothers were found to be more susceptible to temperature-related
PTB effect (Basu et al., 2010; Schifano et al., 2013), because they
were usually of lower socioeconomic status, less education attain-
ment, and more risky behavioral patterns (Basu et al., 2010). As
another marker of socioeconomic status, maternal race/ethnicity
could also have played an important confounding role. In America,
black and Asian women exhibited higher risks of preterm delivery
compared with white and Hispanic women (Basu et al., 2010;
Carmichael et al., 2014). However, less consistent results were re-
ported in Flanders, Belgium, and no obvious differences in effect
estimates were identified among subpopulations when investi-
gating the relationship between temperature and preterm delivery
(Cox et al., 2016). As for birth outcomes of stillbirth and birth
weight/LBW, very few studies explored the potential effect modi-
fiers (Basu et al., 2016). In 16 counties of California, stronger
temperature-stillbirth associations were found among male fe-
tuses, mothers with Hispanic ethnicity, younger ages, and lower
education level (Basu et al., 2016), while a greater effect associated
with birth weight was reported for female fetuses rather than male
fetuses (Murray et al., 2000). The insufficient and less consistent
evidence shown in these studies may limit the ability for the gov-
ernment to conduct more targeted and effective decision-making.
Notably, air pollution has been identified as one of the envi-
ronmental risk factors for adverse birth outcomes (Ha et al., 2014;
Hyder et al., 2014; Qian et al., 2016). Air pollution levels generally
vary substantially between seasons and ambient temperatures
change as seasons alternate. In previous studies, season and tem-
perature were widely found to modify the associations between air
pollution and morbidity/mortality (Chen et al., 2013; Hsu et al.,
2017; Li et al., 2017). Also, air pollution may play a confounding
role in temperature-related mortality effects (Qin et al., 2016).
Based on the existing epidemiological evidence, there is a reason to
infer that ambient pollution could be a potential effect modifier in
the relationship between birth outcomes and temperature expo-
sure. However, the current studies often only included air pollution
as an independent adjusted factor. Hence, future studies should be
further conducted to clarify this potential effect modification.
4.4. Cold-related effects
Despite the protective effect of high temperature observed for
PTB in Shenzhen, China (Liang et al., 2016), most of the identified
studies showed the adverse effects of high temperature exposures
on birth outcomes. To study the heat-related effect, some in-
vestigators limited the study periods to warm season (Auger et al.,
2016; Auger et al., 2014; Basu et al., 2016). Ambient temperatures
have shown nonlinear effects on morbidity/mortality-related
health outcomes (Bunker et al., 2016; Gasparrini et al., 2015), and
increasing evidence was also observed in recent years that cold
temperatures may increase the risks of PTB and LBW (Carmichael
et al., 2014; Ha et al., 2017; He et al., 2016; Liang et al., 2016; Ngo
and Horton, 2016). Therefore, further studies should also focus on
cold-related effects on birth outcomes when investigating the
relationship between temperature and birth outcomes.
Several earlier studies applied linear relationship in the effect of
whole-year temperature exposure (Elter et al., 2004; Murray et al.,
2000), which would likely result in inability to detect or underes-
timate the true effect of temperature, as the slope would be biased
towards the null by the opposing risk of low or high temperatures
(Strand et al., 2011). Statistical models allowing for nonlinear
relationship, such as generalized additive model and distributed lag
nonlinear model, are recommended to be used in the analytical
process so as to detect the potential heat or cold-related effect.
Alternatively, temperature-birth outcome association can also be
assessed separately in cold season and warm season (Schifano et al.,
2013) or using different temperature-stratified exposures (Ngo and
Horton, 2016). In addition, similar to heat waves applied for heat-
related exposure (Wang et al., 2013), cold spells defined by the
several-day's duration of temperature decreasing below the given
low temperature threshold, can also be used as the exposure metric
of cold ambient environment.
5. Conclusions
In the past few decades, an increasing attention has been paid to
investigating adverse birth outcomes associated with prenatal
exposure to temperature. In addition to various study designs and
statistical approaches, the identified studies in the current review
used a number of different temperature indicators as the exposure
assessment. Also, time windows of exposure differentiated
considerably from study to study. These great inconsistencies of
included studies limited our ability to perform a meta-analysis for
quantitative consolidation of the results. In most of the published
epidemiological studies, high temperatures have been identified as
the risk factor for preterm birth, low birth weight, and stillbirth.
Meanwhile, relatively less evidence was also reported for cold-
related effects. Generally, current epidemiologic evidence was
still very limited and mainly came from Europe and North America.
Consequently, more related studies are needed worldwide and
should be conducted in more diversified climate zones, so as to
further ascertain the association between temperature and birth
outcomes.
Future studies should focus on more sophisticated study designs
with large samples, to produce more high-grade evidence based on
scientific effect evaluation of extreme temperatures on birth out-
comes. More accurate temperature exposure during pregnancy
should be estimated and assigned to individual women using the
satellite remote sensing and GIS technologies (e.g., land use
regression). Efforts should be made to find out the exposure win-
dows if there exist vulnerable periods, which could make the
estimated effects comparable between studies using the same
 Y. Zhang et al. / Environmental Pollution 225 (2017) 700e712
exposure periods. Also, the nonlinear temperature impact and cold-
related effect on birth outcomes should be taken into account.
Additionally, more investigations should be conducted aiming at
exploring the potential individual-level modifiers in the effects of
temperature exposure on birth outcomes. These continuous efforts
and further findings would have important implications for
decision-making of public intervention strategies to reduce the
burden of adverse birth outcomes due to prenatal temperature
exposure.
Author contributions
Chuanhua Yu and Yunquan Zhang conceived and designed the
experiments; Yunquan Zhang and Lu Wang reviewed the articles;
Yunquan Zhang wrote the paper; Chuanhua Yu revised the
manuscript.
Conflicts of interest
The authors declare they have no competing financial interests.
Acknowledgments
The study was supported by National Natural Science Founda-
tion of China (81273179). We appreciated the anonymous re-
viewers very much, whose comments and suggestions contributed
a lot to improving the quality of the manuscript.
References
Anthony, C., Mustafa, A., Adriana, A., Sarah, B., Sarah, B., Richard, B., Sharon, F.,
Nora, G., Anne, J., Maria, K., Lee, M., Levy, C., Maslin, M., McCoy, D., McGuire, B.,
Montgomery, H., Napier, D., Pagel, C., Patel, J., Antonio, J., Oliveira, P.d.,
Redclift, N., Rees, H., Rogger, D., Scott, J., Stephenson, J., Twigg, J.W., Patterson, C.,
2009. Managing the health effects of climate change: lancet and university
college london institute for global health commission. Lancet 373, 1693e1733.
Arroyo, V., Diaz, J., Carmona, R., Ortiz, C., Linares, C., 2016a. Impact of air pollution
and temperature on adverse birth outcomes: madrid, 2001-2009. Environ.
Pollut. 218, 1154e1161.
Arroyo, V., Diaz, J., Ortiz, C., Carmona, R., Saez, M., Linares, C., 2016b. Short term
effect of air pollution, noise and heat waves on preterm births in Madrid
(Spain). Environ. Res. 145, 162e168.
Auger, N., Kuehne, E., Goneau, M., Daniel, M., 2011. Preterm birth during an extreme
weather event in Quebec, Canada: a "natural experiment". Matern. Child.
Health J. 15, 1088e1096.
Auger, N., Naimi, A.I., Smargiassi, A., Lo, E., Kosatsky, T., 2014. Extreme heat and risk
of early delivery among preterm and term pregnancies. Epidemiology 25,
344e350.
Auger, N., Fraser, W.D., Smargiassi, A., Bilodeau-Bertrand, M., Kosatsky, T., 2016 May
9. Elevated outdoor temperatures and risk of stillbirth. Int. J. Epidemiol. pii:
dyw077.
Basu, R., Malig, B., Ostro, B., 2010. High ambient temperature and the risk of preterm
delivery. Am. J. Epidemiol. 172, 1108e1117.
Basu, R., Sarovar, V., Malig, B.J., 2016. Association between high ambient tempera-
ture and risk of stillbirth in California. Am. J. Epidemiol. 183, 894e901.
Blencowe, H., Cousens, S., Oestergaard, M.Z., Chou, D., Moller, A.B., Narwal, R.,
Adler, A., Vera, G.C., Rohde, S., Say, L., Lawn, J.E., 2012. National, regional, and
worldwide estimates of preterm birth rates in the year 2010 with time trends
since 1990 for selected countries: a systematic analysis and implications. Lancet
379, 2162e2172.
Bunker, A., Wildenhain, J., Vandenbergh, A., Henschke, N., Rocklov, J., Hajat, S.,
Sauerborn, R., 2016. Effects of air temperature on climate-sensitive mortality
and morbidity outcomes in the elderly; a systematic review and meta-analysis
of epidemiological evidence. EBioMedicine 6, 258e268.
Carmichael, S.L., Cullen, M.R., Mayo, J.A., Gould, J.B., Loftus, P., Stevenson, D.K.,
Wise, P.H., Shaw, G.M., 2014. Population-level correlates of preterm delivery
among black and white women in the U.S. PLoS One 9, e94153.
Carolan-Olah, M., Frankowska, D., 2014. High environmental temperature and
preterm birth: a review of the evidence. Midwifery 30, 50e59.
CASP, 2003. Critical Appraisal Skills Programme (CASP): Appraisal Tools. Public
Health Resource Unit. NHS, Oxford.
CASP, 2013. 12 Questions to Help You Make Sense of a Cohort Study, Critical
Appraisal Skills Programme. Public Health Resource Unit, Oxford.
Chen, R., Peng, R.D., Meng, X., Zhou, Z., Chen, B., Kan, H., 2013. Seasonal variation in
the acute effect of particulate air pollution on mortality in the China Air
711
Pollution and Health Effects Study (CAPES). Sci. Total Environ. 450e451,
259e265.
Cox, B., Vicedo-Cabrera, A.M., Gasparrini, A., Roels, H.A., Martens, E.,
Vangronsveld, J., Forsberg, B., Nawrot, T.S., 2016. Ambient temperature as a
trigger of preterm delivery in a temperate climate. J. Epidemiol. Community
Health 70, 1191e1199.
Dadvand, P., Basagana, X., Sartini, C., Figueras, F., Vrijheid, M., de Nazelle, A.,
Sunyer, J., Nieuwenhuijsen, M.J., 2011. Climate extremes and the length of
gestation. Environ. Health Perspect. 119, 1449e1453.
Desche ^nes, O., Greenstone, M., Guryan, J., 2009. Climate change and birth weight.
Am. Econ. Rev. 99, 211e217.
Diaz, J., Arroyo, V., Ortiz, C., Carmona, R., Linares, C., 2016. Effect of environmental
factors on low weight in non-premature births: a time series analysis. PLoS One
11, e0164741.
Elter, K., Emine, A., Uyar, E., Kavak, Z.N., 2004. Exposure to low outdoor temperature
in the midtrimester is associated with low birth weight. Aust. N. Z. J. Obstet.
Gynaecol. 44, 553e557.
Estarlich, M., Ballester, F., Davdand, P., Llop, S., Esplugues, A., Fernandez-
Somoano, A., Lertxundi, A., Guxens, M., Basterrechea, M., Tardon, A., Sunyer, J.,
Iniguez, C., 2016. Exposure to ambient air pollution during pregnancy and
preterm birth: a Spanish multicenter birth cohort study. Environ. Res. 147,
50e58.
Flouris, A.D., Spiropoulos, Y., Sakellariou, G.J., Koutedakis, Y., 2009. Effect of seasonal
programming on fetal development and longevity: links with environmental
temperature. Am. J. Hum. Biol. 21, 214e216.
Gasparrini, A., 2011. Distributed lag linear and non-linear models in R: the package
dlnm. J. Stat. Softw. 43, 1e20.
Gasparrini, A., Armstrong, B., 2010. Time series analysis on the health effects of
temperature: advancements and limitations. Environ. Res. 110, 633e638.
Gasparrini, A., Guo, Y., Hashizume, M., Lavigne, E., Zanobetti, A., Schwartz, J.,
Tobias, A., Tong, S., Rocklov, J., Forsberg, B., Leone, M., De Sario, M., Bell, M.L.,
Guo, Y.L., Wu, C.F., Kan, H., Yi, S.M., de Sousa Zanotti Stagliorio Coelho, M.,
Saldiva, P.H., Honda, Y., Kim, H., Armstrong, B., 2015. Mortality risk attributable
to high and low ambient temperature: a multicountry observational study.
Lancet 386, 369e375.
GBD 2015 SDG Collaborators, 2016. Measuring the health-related sustainable
development Goals in 188 countries: a baseline analysis from the global burden
of disease study 2015. Lancet 388, 1813e1850.
Grace, K., Davenport, F., Hanson, H., Funk, C., Shukla, S., 2015. Linking climate
change and health outcomes: examining the relationship between tempera-
ture, precipitation and birth weight in Africa. Glob. Environ. Change 35,
125e137.
Gronlund, C.J., Zanobetti, A., Schwartz, J.D., Wellenius, G.A., O'Neill, M.S., 2014. Heat,
heat waves, and hospital admissions among the elderly in the United States,
1992-2006. Environ. Health Perspect. 122, 1187e1192.
Ha, S., Hu, H., Roussos-Ross, D., Haidong, K., Roth, J., Xu, X., 2014. The effects of air
pollution on adverse birth outcomes. Environ. Res. 134, 198e204.
Ha, S., Liu, D., Zhu, Y., Kim, S.S., Sherman, S., Mendola, P., 2017. Ambient temperature
and early delivery of singleton pregnancies. Environ. Health Perspect. 125,
453e459. http://dx.doi.org/10.1289/EHP97.
He, J.R., Liu, Y., Xia, X.Y., Ma, W.J., Lin, H.L., Kan, H.D., Lu, J.H., Feng, Q., Mo, W.J.,
Wang, P., Xia, H.M., Qiu, X., Muglia, L.J., 2016. Ambient temperature and the risk
of preterm birth in Guangzhou, China (2001-2011). Environ. Health Perspect.
124, 1100e1106.
Hsu, W.H., Hwang, S.A., Kinney, P.L., Lin, S., 2017. Seasonal and temperature modi-
fications of the association between fine particulate air pollution and cardio-
vascular hospitalization in New York state. Sci. Total Environ. 578, 626e632.
Huang, C., Barnett, A.G., Xu, Z., Chu, C., Wang, X., Turner, L.R., Tong, S., 2013. Man-
aging the health effects of temperature in response to climate change: chal-
lenges ahead. Environ. Health Perspect. 121, 415e419.
Hyder, A., Lee, H.J., Ebisu, K., Koutrakis, P., Belanger, K., Bell, M.L., 2014. PM2.5
exposure and birth outcomes: use of satellite- and monitor-based data.
Epidemiology 25, 58e67.
Kloog, I., Melly, S.J., Coull, B.A., Nordio, F., Schwartz, J.D., 2015. Using satellite-based
spatiotemporal resolved air temperature exposure to study the association
between ambient air temperature and birth outcomes in Massachusetts. En-
viron. Health Perspect. 123, 1053e1058.
Lajinian, S., Hudson, S., Applewhite, L., Feldman, J., Minkoff, H.L., 1997. An associa-
tion between the heat-humidity index and preterm labor and delivery: a pre-
liminary analysis. Am. J. Public Health 87, 1205e1207.
Lakshmanan, A., Chiu, Y.H., Coull, B.A., Just, A.C., Maxwell, S.L., Schwartz, J.,
Gryparis, A., Kloog, I., Wright, R.J., Wright, R.O., 2015. Associations between
prenatal traffic-related air pollution exposure and birth weight: modification by
sex and maternal pre-pregnancy body mass index. Environ. Res. 137, 268e277.
Lawlor, D.A., Leon, D.A., Davey Smith, G., 2005. The association of ambient outdoor
temperature throughout pregnancy and offspring birthweight: findings from
the Aberdeen Children of the 1950s cohort. BJOG 112, 647e657.
Lee, S.J., Hajat, S., Steer, P.J., Filippi, V., 2008. A time-series analysis of any short-term
effects of meteorological and air pollution factors on preterm births in London,
UK. Environ. Res. 106, 185e194.
Li, J., Woodward, A., Hou, X.Y., Zhu, T., Zhang, J., Brown, H., Yang, J., Qin, R., Gao, J.,
Gu, S., Li, J., Xu, L., Liu, X., Liu, Q., 2017. Modification of the effects of air pol-
lutants on mortality by temperature: a systematic review and meta-analysis.
Sci. Total Environ. 575, 1556e1570.
Liang, Z., Yan, L., Ma, Y., Lei, Z., Xue, Z., Li, L., Zhang, S., Cheng, Y., Zhou, X., Lin, H.,
712 Y. Zhang et al. / Environmental Pollution 225 (2017) 700e712
2016. The association between ambient temperature and preterm birth in
Shenzhen, China: a distributed lag non-linear time series analysis. Environ.
Health 15, 84.
Ma, W., Wang, L., Lin, H., Liu, T., Zhang, Y., Rutherford, S., Luo, Y., Zeng, W., Zhang, Y.,
Wang, X., Gu, X., Chu, C., Xiao, J., Zhou, M., 2015. The temperature-mortality
relationship in China: an analysis from 66 Chinese communities. Environ. Res.
137, 72e77.
Maclure, M., 1991. The case-crossover design: a method for studying transient ef-
fects on the risk of acute events. Am. J. Epidemiol. 133, 144e153.
Maclure, M., Mittleman, M.A., 2000. Should we use a case-crossover design? Annu.
Rev. Public Health 21, 193e221.
Matsuda, S., Furuta, M., Kahyo, H., 1998. An ecologic study of the relationship be-
tween mean birth weight, temperature and calorie consumption level in Japan.
J. Biosoc. Sci. 30, 85e93.
McCormick, M.C., 1985. The contribution of low birth weight to infant mortality and
childhood morbidity. N. Engl. J. Med. 312, 82e90.
Meng, X., Fu, Q., Ma, Z., Chen, L., Zou, B., Zhang, Y., Xue, W., Wang, J., Wang, D.,
Kan, H., Liu, Y., 2016. Estimating ground-level PM(10) in a Chinese city by
combining satellite data, meteorological information and a land use regression
model. Environ. Pollut. 208, 177e184.
Murray, L.J., O'Reilly, D.P., Betts, N., Patterson, C.C., Davey Smith, G., Evans, A.E.,
2000. Season and outdoor ambient temperature: effects on birth weight.
Obstet. Gynecol. 96, 689e695.
Ngo, N.S., Horton, R.M., 2016. Climate change and fetal health: the impacts of
exposure to extreme temperatures in New York City. Environ. Res. 144,
158e164.
Porter, K.R., Thomas, S.D., Whitman, S., 1999. The relation of gestation length to
short-term heat stress. Am. J. Public Health 89, 1090e1092.
Qian, Z., Liang, S., Yang, S., Trevathan, E., Huang, Z., Yang, R., Wang, J., Hu, K.,
Zhang, Y., Vaughn, M., Shen, L., Liu, W., Li, P., Ward, P., Yang, L., Zhang, W.,
Chen, W., Dong, G., Zheng, T., Xu, S., Zhang, B., 2016. Ambient air pollution and
preterm birth: a prospective birth cohort study in Wuhan, China. Int. J. Hyg.
Environ. Health 219, 195e203.
Qin, R.X., Xiao, C., Zhu, Y., Li, J., Yang, J., Gu, S., Xia, J., Su, B., Liu, Q., Woodward, A.,
2016. The interactive effects between high temperature and air pollution on
mortality: a time-series analysis in Hefei, China. Sci. Total Environ. 575,
1530e1537.
Saigal, S., Doyle, L.W., 2009. An overview of mortality and sequelae of preterm birth
from infancy to adulthood. Lancet 371, 261e269.
Schifano, P., Lallo, A., Asta, F., De Sario, M., Davoli, M., Michelozzi, P., 2013. Effect of
ambient temperature and air pollutants on the risk of preterm birth, Rome
2001-2010. Environ. Int. 61, 77e87.
Schifano, P., Asta, F., Dadvand, P., Davoli, M., Basagana, X., Michelozzi, P., 2016. Heat
and air pollution exposure as triggers of delivery: a survival analysis of
population-based pregnancy cohorts in Rome and Barcelona. Environ. Int. 88,
153e159.
Strand, L.B., Barnett, A.G., Tong, S., 2011. The influence of season and ambient
temperature on birth outcomes: a review of the epidemiological literature.
Environ. Res. 111, 451e462.
Strand, L.B., Barnett, A.G., Tong, S., 2012. Maternal exposure to ambient temperature
and the risks of preterm birth and stillbirth in Brisbane, Australia. Am. J. Epi-
demiol. 175, 99e107.
Tian, Z., Li, S., Zhang, J., Jaakkola, J.J., Guo, Y., 2012. Ambient temperature and cor-
onary heart disease mortality in Beijing, China: a time series study. Environ.
Health 11, 56.
Tustin, K., Gross, J., Hayne, H., 2004. Maternal exposure to first-trimester sunshine is
associated with increased birth weight in human infants. Dev. Psychobiol. 45,
221e230.
United Nations Children’s Fund and World Health Organization, 2004. Low Birth
Weight, New York. http://whqlibdoc.who.int/publications/2004/9280638327.
pdf (accessed 23 November 2016).
Vicedo-Cabrera, A.M., Iniguez, C., Barona, C., Ballester, F., 2014. Exposure to elevated
temperatures and risk of preterm birth in Valencia, Spain. Environ. Res. 134,
210e217.
Vicedo-Cabrera, A.M., Olsson, D., Forsberg, B., 2015. Exposure to seasonal temper-
atures during the last month of gestation and the risk of preterm birth in
Stockholm. Int. J. Environ. Res. Public Health 12, 3962e3978.
Walfisch, A., Kabakov, E., Friger, M., Sheiner, E., 2016. Trends, seasonality and effect
of ambient temperature on preterm delivery. J. Matern. Fetal Neonatal Med.
1e5.
Wang, J., Williams, G., Guo, Y., Pan, X., Tong, S., 2013. Maternal exposure to heat-
wave and preterm birth in Brisbane, Australia. BJOG 120, 1631e1641.
Wang, Y., Kloog, I., Coull, B.A., Kosheleva, A., Zanobetti, A., Schwartz, J.D., 2016.
Estimating causal effects of long-term PM2.5 exposure on mortality in New
Jersey. Environ. Health Perspect. 124, 1182e1188.
Wells, J.C., Cole, T.J., 2002. Birth weight and environmental heat load: a between-
population analysis. Am. J. Phys. Anthropol. 119, 276e282.
Wolf, J., Armstrong, B., 2012. The association of season and temperature with
adverse pregnancy outcome in two German states, a time-series analysis. PLoS
One 7, e40228.
Yackerson, N., Piura, B., Sheiner, E., 2008. The influence of meteorological factors on
the emergence of preterm delivery and preterm premature rupture of mem-
brane. J. Perinatol. 28, 707e711.
Ye, X., Wolff, R., Yu, W., Vaneckova, P., Pan, X., Tong, S., 2012. Ambient temperature
and morbidity: a review of epidemiological evidence. Environ. Health Perspect.
120, 19e28.
Zhang, Z., Laden, F., Forman, J.P., Hart, J.E., 2016. Long-term exposure to particulate
matter and self-reported hypertension: a prospective analysis in the nurses'
health study. Environ. Health Perspect. 124, 1414e1420.