Literature Review - Systematic Research 4 Pages

International Journal of Gynecology and Obstetrics 130 (2015) 116–122

Contents lists available at ScienceDirect

International Journal of Gynecology and Obstetrics

journal homepage: www.elsevier.com/locate/ijgo

REVIEW ARTICLE

A meta-analysis of risk of pregnancy loss and caffeine and coffee consumption during pregnancy

Ji Li a,1, Hong Zhao b,1, Ju-Min Song c, Jing Zhang d, Yin-Lan Tang e, Chang-Mao Xin a,⁎ a Center for Health Management, School of Health Services Management, Anhui Medical University, Hefei, China b Department of Nutrition, Jiangsu Province Traditional Chinese Medicine Hospital, Nanjing, China c Department of Social Medicine and Health Services Management, School of Public Health, Kunming Medical University, Kunming, China d Second Affiliated Hospital of Anhui Medical University, Hefei, China e Department of Rehabilitation Medicine, The Affiliated Provincial Hospital of Anhui Medical University, Hefei, China

⁎ Corresponding author at: Center for Health Ma Services Management, Anhui Medical University, H Tel./fax: +86 551 65518231.

E-mail address: [email protected] (C.-M. Xi 1 These authors contributed equally.

http://dx.doi.org/10.1016/j.ijgo.2015.03.033 0020-7292/© 2015 International Federation of Gynecology

Downloaded for Anony For personal use

a b s t r a c t

a r t i c l e i n f o

Article history:

Received 26 October 2014 Received in revised form 25 February 2015 Accepted 30 April 2015

Keywords: Caffeine Coffee Fetal death Meta-analysis Pregnancy loss Spontaneous abortion

Background: Previous reports of the relationship between pregnancy loss and caffeine/coffee consumption have been inconsistent. Objectives: To evaluate the association between pregnancy loss and caffeine and coffee consumption. Search strategy: PubMed was searched for reports published before September 2014, with the keywords “caffeine,” “coffee,” “beverage,” “miscarriage,” “spontaneous abortion,” and “fetal loss.” Selection criteria: Case-control and cohort studies were included when they had been reported in English, the exposure of interest was caffeine/coffee consumption during pregnancy, the outcome of interest was spontaneous abortion or fetal death, and multivariate-adjusted odds ratios (ORs) or risk ratios were provided or could be calculated. Data collection and analysis: Data were extracted and combined ORs calculated. Main results: Overall, 26 studies were included (20 of caffeine and eight of coffee). After adjustment for heterogeneity, caffeine consumption was associated with an increased risk of pregnancy loss (OR 1.32, 95% confidence interval [CI] 1.24–1.40), as was coffee consumption (OR 1.11, 95% CI 1.02–1.21). A dose–response analysis suggested that risk of pregnancy

loss rose by 19% for every increase in caffeine intake of 150 mg/day and by 8% for every increase in coffee intake of two cups per day. Conclusions: Consumption of caffeine and coffee during pregnancy seems to increase the risk of pregnancy loss.

© 2015 International Federation of Gynecology and Obstetrics. Published by Elsevier Ireland Ltd. All rights reserved.

1. Introduction

Pregnancy loss through spontaneous abortion (an unintended ter- mination of pregnancy before 20 weeks) or fetal death (fetal demise after 20 weeks) is not uncommon in the general population [1]. Despite decades of research, the causes of such pregnancy losses are unclear. Fernandes et al. [2] suggested that the process might be multifactorial, with the possible involvement of environmental factors such as caffeine consumption.

Caffeine increases cellular cyclic adenosine monophosphate levels by inhibiting phosphodiesterases [3], which can affect cell growth and fetal development [4]. Moreover, the structure of caffeine is similar to that of adenine and guanine, so it might be incorporated into the DNA macromolecule during mitosis, causing chromosomal anomalies [5]. A fetus can be exposed to caffeine through the amniotic fluid or umbilical cord, but has little ability to metabolize the compound [6]. Additionally,

nagement, School of Health efei, Anhui 230032, China.

n).

and Obstetrics. Published by Elsevier I

mous User (n/a) at Saudi Digital Lib only. No other uses without permiss

clearance of caffeine from the mother’s body slows down during pregnancy [7], Furuhashi et al. [8] showed high rates of chromosomal anomalies and spontaneous abortion among pregnant women who consumed many caffeinated drinks.

However, the results of epidemiological studies into the association between caffeine/coffee consumption during pregnancy and pregnancy loss are inconsistent. Although the risk estimates about the association between caffeine consumption during pregnancy and spontaneous abortion were pooled by Fernandes et al. [2], their inclusion criteria were not overly stringent. Additionally, it seems that the dose–response relationship between caffeine/coffee consumption and pregnancy loss has not yet been investigated in a meta-analysis. Thus, the aim of the present review was to elucidate potential associations between pregnancy loss and caffeine and coffee consumption.

2. Materials and methods

2.1. Search strategy

PubMed was searched for reports published before September 1, 2014, with the keywords “caffeine,” “coffee,” “beverage,” “miscarriage,” “spontaneous abortion,” and “fetal loss.” The references of identified

reland Ltd. All rights reserved.

rary from ClinicalKey.com by Elsevier on February 20, 2019. ion. Copyright ©2019. Elsevier Inc. All rights reserved.

117J. Li et al. / International Journal of Gynecology and Obstetrics 130 (2015) 116–122

publications (including reviews) were also searched to identify additional studies.

2.2. Selection criteria

All identified studies were independently reviewed by two investi- gators (J.L. and H.Z.). Studies were included in the present meta- analysis if they had been reported in English, were case-control or cohort investigations, the exposure of interest was caffeine or coffee consumption during pregnancy, the outcome of interest was spontane- ous abortion or fetal death, and multivariate-adjusted odds ratios (ORs) or risk ratios (RRs) with 95% confidence intervals (CIs) were provided (or data provided allowed their calculation). If data were duplicated in more than one study, the largest study was preferred. Reports were excluded before full-text assessment if they were of animal studies or were reviews, systematic reviews, or meta-analyses.

2.3. Data extraction

Data were extracted from each study by two investigators (J.L. and H.Z.) independently. The first author’s last name, year of publication, location, study period, number of cases and controls or sample size, level of maternal caffeine and/or coffee consumption during pregnancy, estimates with 95% CIs, and details of adjustment for potential confounding factors were recorded.

2.4. Statistical analyses

If ORs or RRs and 95% CIs were not provided, they were calculated using the Mantel–Haenszel method. If the incidence of disease is low, RRs are approximately equal to ORs [9]. Therefore, combined ORs were used to estimate the combined effects. The combined ORs were calculat- ed by combining logarithmic risk estimates weighted by the inverse of their variances to evaluate the strength of association between caffeine/ coffee consumption during pregnancy and risk of pregnancy loss.

Considering possible between-study heterogeneity, the I2 statistic was used to assess heterogeneity. The random effects model was used as the pooling method if an obvious between-study heterogeneity (I2 N 50%) was found [10]; otherwise, the fixed effects model was used [11]. Sensitivity analysis was performed when between-study hetero- geneity was present: studies contributing the most to the heterogeneity were sequentially removed until homogeneity was obtained [12].

A cumulative meta-analysis was performed by chronologically ordering studies by publication year to find the starting point of a risk estimate becoming statistically significant and the trend in estimated risk effect [13]. Subgroup analyses were conducted of region (Europe, North/South America, or Asia), design (cohort or case–control), adjusted (yes or no) and publication year (b2000 or ≥2000). The Egger quantitative test was used to estimate evidence for potential publication bias [14].

A dose–response meta-analysis was also conducted to explore the pooled dose–response relationship between caffeine/coffee consump- tion during pregnancy and pregnancy loss. Because caffeine consump- tion was reported in various scales, exposure data were converted into a uniform measurement (mg/day). The median of a range of caffeine or coffee consumption was considered as the corresponding exposure dose; if the median consumption was not reported, the midpoint between the upper and lower range was used. If the lowest category was open-ended, its lower boundary was set to zero. When the highest category was unrestricted for caffeine, it was assumed to be the same size as the next highest category [15]. When the highest category was unrestricted for coffee, the exposure dose was defined by the lower end value of the category multiplied by 1.5. Caffeine consumption was divided into three groups: light (b150 mg/day), moderate (150–300 mg/day), and heavy (≥301 mg/day). Coffee drinkers were also classified into three levels: light (b2 cups per day), moderate

Downloaded for Anonymous User (n/a) at Saudi Digital Library For personal use only. No other uses without permission. C

(2–3 cups per day), and heavy (≥4 cups per day) drinkers. Non- drinkers and individuals with the lowest consumption were regarded as the reference group for both caffeine and coffee.

Nonlinearity in the relationship between caffeine and coffee con- sumption and risk of pregnancy loss was assumed. Taking into account the heterogeneity among studies, a two-stage random effects dose– response meta-analysis was performed to compute the trend from the correlated log OR estimates across levels of caffeine and coffee con- sumption [16]. A restricted cubic spline model was estimated using generalized least square regression with four knots at the fifth, 35th, 65th, and 95th percentiles of the levels of caffeine and coffee consump- tion, taking into account the correlation within each set of published ORs. A fixed or random effects restricted cubic spline model according to heterogeneity combined the study-specific estimates using the meth- od proposed by Greenland and Longnecker to estimate the covariances of the OR [17]. If the P value was less than 0.05, the nonlinearity dose- response relationship was considered to exist. All statistical analyses were performed with STATA version 12.0 (Stata Corp, College Station, TX, USA).

3. Results

3.1. Literature search and study characteristics

A total of 26 studies were included (Fig. 1, Supplementary Material S1): 13 were case-control studies [18–30] and 13 were cohort studies [31–43]. Twenty studies [18–27,31–40] were of caffeine and eight stud- ies [28–30,36,39,41–43] were of coffee consumption. Among the studies about caffeine consumption, seven were conducted in Europe [21–24, 27,34,40], 11 in North/South America [18,19,25,31–33,35–39], and two in Asia [20,26]. Among the studies about coffee consumption, four were conducted in Europe [28,30,42,43], three in North/South America [36,39,41], and one in Asia [29].

3.2. Caffeine consumption and risk of pregnancy loss

Overall, caffeine consumption was associated with risk of pregnancy loss (OR 1.47, 95% CI 1.31–1.66; I2 = 70.9%). The association remained after two outlying studies [27,34] were excluded in the sensitivity analysis (OR 1.32, 95% CI 1.24–1.40), although the heterogeneity was no longer significant (I2 = 48.8%).

The association remained for moderate and heavy consumption in the analysis by level of consumption (Fig. 2A). In this analysis, between-study heterogeneity was detected only for heavy consump- tion. When the two outlying studies [27,34] were excluded from the analyses of moderate and heavy consumption, the associations remained significant (moderate: OR 1.28, 95% CI 1.16–1.42; heavy: 1.60, 1.46–1.76). No between-study heterogeneity was detected (moderate: I2 = 8.6%; heavy: I2 = 44.8%). No association was recorded for light consumption in the sensitivity analysis (OR 1.04, 95% CI 0.94–1.16; I2 = 0.0%).

Caffeine consumption during pregnancy was significantly associated with increased risk of pregnancy loss in both cohort and case-control studies (Table 1). The association was not recorded for either type of study when light consumption was considered; moderate and heavy caffeine consumption during pregnancy were associated with pregnan- cy loss in both study types, although the association was not significant for moderate consumption in cohort studies after adjustment (Table 1). Caffeine consumption during pregnancy was also significantly associat- ed with increased risk of pregnancy loss in other subgroups (Table 1). No association was recorded for light consumption in the other subgroups; however, moderate and heavy caffeine consumptions during pregnancy were associated with pregnancy loss in all the other subgroups (Table 1).

from ClinicalKey.com by Elsevier on February 20, 2019. opyright ©2019. Elsevier Inc. All rights reserved.

Fig. 1. Flowchart of study selection.

118 J. Li et al. / International Journal of Gynecology and Obstetrics 130 (2015) 116–122

3.3. Coffee consumption and risk of pregnancy loss

Overall, coffee consumption was associated with pregnancy loss (OR 1.31, 95% CI 1.15–1.50; I2 = 87.5%). The association remained after exclusion of two outlying studies [28,43] (OR 1.11, 95% CI 1.02–1.21), although the heterogeneity was no longer significant (I2 = 49.1%).

The association remained for heavy consumption in the analysis by level of consumption (Fig. 2B). There was significant heterogeneity for

Fig. 2. Forest plots of the association between pregnancy loss and (A) caffeine consumptio CI, confidence interval.

Downloaded for Anonymous User (n/a) at Saudi Digital Libra For personal use only. No other uses without permissio

the analyses of moderate and heavy consumption. When two outlying studies [28,43] were excluded from the analysis of heavy consumption, the association remained significant (OR 1.27, 95% CI 1.07–1.52), but the heterogeneity was no longer significant (I2 = 47.5%). No associations were recorded for light (1.00, 95% CI 0.93–1.07; I2 = 4.4%) or moderate (OR 1.05, 95% CI 0.92–1.20; I2 = 13.6%) consumption.

The subgroup analysis showed that coffee consumption during pregnancy was positively associated with risk of pregnancy loss in

n and (B) coffee consumption, by level of consumption. Abbreviations: OR, odds ratio;

ry from ClinicalKey.com by Elsevier on February 20, 2019. n. Copyright ©2019. Elsevier Inc. All rights reserved.

Table 1 Subgroup analysis.

Subgroup Light vs reference I2, % Moderate vs reference I2, % Heavy vs reference I2, % Pooled OR (95% CI) I2, %

Caffeine Study type Cohort

Crude OR 1.05 (0.91–1.22) 0 1.27 (1.02–1.58) 13.9 2.39 (1.47–3.89) 84.9 1.47 (1.20–1.79) 72.0 Amended ORa 1.05 (0.91–1.22) 0 1.16 (0.95–1.42) 0 1.54 (1.21–1.97) 32.3 1.21 (1.08–1.37) 21.9

Case-control Crude OR 1.03 (0.88–1.20) 0 1.39 (1.18–1.64) 37.4 1.88 (1.47–2.41) 75.6 1.48 (1.28–1.71) 70.4 Amended ORb 1.03 (0.87–1.22) 0 1.39 (1.22–1.57) 0 1.81 (1.59–2.06) 8.7 1.48 (1.30–1.69) 56.0

Adjusted Yes

Crude OR 1.07 (0.94–1.21) 0 1.26 (1.04–1.52) 36.1 2.17 (1.56–3.03) 80.9 1.51 (1.29–1.78) 73.0 Amended ORc 1.08 (0.93–1.24) 6.1 1.25 (1.05–1.48) 13.4 1.86 (1.48–2.34) 39.0 1.39 (1.21–1.60) 52.0

No Crude OR 0.98 (0.81–1.19) 0 1.42 (1.23–1.64) 0 1.94 (1.39–2.69) 80.0 1.44 (1.22–1.71) 66.8 Amended OR d 0.98 (0.81–1.19) 0 1.40 (1.21–1.61) 0 1.65 (1.43–1.90) 15.7 1.36 (1.19–1.55) 42.2

Publication year b2000

Crude OR 1.02 (0.89–1.17) 0 1.42 (1.14–1.75) 44.2 2.50 (1.33–4.70) 87.7 1.46 (1.19–1.78) 75.7 Amended ORa 1.02 (0.89–1.17) 0 1.36 (1.09–1.68) 40.0 1.44 (1.11–1.88) 18.6 1.22 (1.08–1.38) 33.9

≥2000 Crude OR 1.08 (0.91–1.27) 0 1.28 (1.11–1.48) 3.0 1.87 (1.49–2.34) 73.3 1.48 (1.29–1.71) 65.9 Amended ORb 1.09 (0.91–1.31) 0 1.32 (1.14–1.52) 0 1.80 (1.57–2.07) 21.5 1.49 (1.31–1.69) 45.0

Country Europe

Crude OR 1.22 (0.66–2.28) 33.1 1.35 (1.08–1.68) 40.5 2.66 (1.84–3.84) 88.0 1.97 (1.56–2.50) 83.4 Amended ORe 2.20 (0.69–7.01) – 1.30 (1.11–1.52) 0 1.84 (1.62–2.09) 25.6 1.61 (1.46–1.78) 46.6

North/South America Crude OR 1.03 (0.92–1.15) 0 1.30 (1.09–1.55) 19.3 1.53 (1.28–1.82) 18.1 1.22 (1.11–1.35) 31.7

Coffee Study type Cohort

Crude OR 1.01 (0.91–1.13) 17.9 1.02 (0.94–1.11) 0 1.41 (1.16–1.42) 88.9 1.23 (1.07–1.42) 91.5 Amended ORf 1.01 (0.91–1.13) 17.9 1.02 (0.94–1.11) 0 1.21 (1.09–1.35) 8.9 1.10 (1.01–1.19) 50.6

Case-control Crude OR 1.08 (0.83–1.39) 29.7 1.49 (0.96–2.30) 45.7 4.17 (2.74–6.36) 0 1.60 (1.12–2.28) 76.2

Adjusted Yes

Crude OR 1.03 (0.92–1.15) 28.2 1.23 (0.89–1.70) 81.4 1.50 (1.04–2.16) 89.2 1.23 (1.07–1.41) 84.2 Amended ORg 0.98 (0.93–1.04) 0 1.03 (0.94–1.12) 0 1.17 (1.06–1.30) 0 1.06 (0.99–1.13) 37.5

No Crude OR 1.08 (0.80–1.45) 30.7 1.20 (0.80–1.80) 38.6 1.84 (1.30–2.62) 87.5 1.39 (1.12–1.74) 77.5 Amended OR h 1.08 (0.80–1.45) 30.7 1.39 (0.74–2.64) 49.4 6.28 (1.69–23.35) 0 1.30 (0.93–1.82) 48.8

Publication year b2000

Crude OR 1.03 (0.92–1.15) 28.2 1.23 (0.89–1.70) 81.4 1.50 (1.04–2.16) 89.2 1.23 (1.07–1.41) 84.2 Amended ORg 0.98 (0.93–1.04) 0 1.03 (0.94–1.12) 0 1.17 (1.06–1.30) 0 1.06 (0.99–1.13) 37.5

≥2000 Crude OR 1.08 (0.80–1.45) 30.7 1.20 (0.80–1.80) 38.6 1.84 (1.30–2.62) 87.5 1.39 (1.12–1.74) 77.5 Amended ORh 1.08 (0.80–1.45) 30.7 1.39 (0.74–2.64) 49.4 6.28 (1.69–23.35) 0 1.30 (0.93–1.82) 48.8

Country Europe

Crude OR 1.16 (0.75–1.80) 29.5 1.34 (0.93–1.93) 57.1 2.18 (1.51–3.15) 89.9 1.61 (1.29–2.01) 79.9 Amended ORh 1.16 (0.75–1.80) 29.5 1.49 (0.96–2.30) 45.7 4.17 (2.74–6.36) 0 1.73 (1.17–2.54) 72.1

America Crude OR 1.01 (0.91–1.13) 17.9 1.03 (0.94–1.12) 0 1.17 (1.06–1.30) 0 1.06 (0.99–1.14) 36.6

a Domínguez-Rojas et al. [34] removed. b Maconochie et al. [24] and Stefanidou et al. [27] removed. c Domínguez-Rojas et al. [34] and Maconochie et al. [24] removed. d Stefanidou et al. [27] removed. e Domínguez-Rojas et al. [34], Maconochie et al. [24], and Stefanidou et al. [27] removed, f Andersen et al. [43] removed. g Parazzini et al. [28] removed. h Wisborg et al. [42] and Andersen et al. [43] removed.

119J. Li et al. / International Journal of Gynecology and Obstetrics 130 (2015) 116–122

both cohort and case-control studies (Table 1). The association was not recorded for either study type for light or moderate consumption; heavy consumption was associated with pregnancy loss in both cohort and case-control studies (Table 1). Coffee consumption during pregnan- cy was also significantly associated with increased risk of pregnancy loss in some other subgroups (Table 1). No associations were recorded for light or moderate consumption in the other subgroups; however,

Downloaded for Anonymous User (n/a) at Saudi Digital Library For personal use only. No other uses without permission. C

heavy caffeine consumption during pregnancy was associated with pregnancy loss in all the other subgroups (Table 1).

3.4. Cumulative meta-analysis

The association between caffeine and risk of pregnancy loss was sta- ble from approximately the year 2000 (Fig. 3A). There was a tendency of

from ClinicalKey.com by Elsevier on February 20, 2019. opyright ©2019. Elsevier Inc. All rights reserved.

Fig. 3. Cumulative meta-analysis results of the association between risk of pregnancy loss and (A) caffeine or (B) coffee consumption during pregnancy. Abbreviations: OR, odds ratio; CI, confidence interval.

120 J. Li et al. / International Journal of Gynecology and Obstetrics 130 (2015) 116–122

a more obvious association between coffee and risk of pregnancy loss as studies were added by publication year (Fig. 3B).

3.5. Dose–response meta-analysis

Overall, 23 studies were included in final dose–response meta- analysis: two identified articles [20,29] had only two levels of consump- tion including the reference and one [27] presented no data for cases and controls. Among the 23 included articles, 18 articles were used to explore the dose–response relationship between caffeine consumption during pregnancy and risk of pregnancy loss [18,19,21–26,31–40], and six were used to explore the dose–response relationship between coffee consumption during pregnancy and risk of pregnancy loss [28,30,36,41–43].

For caffeine, the nonlinear relationships of caffeine consumption during pregnancy with pregnancy loss were not observed (P for nonlinearity = 0.76; Fig. 4A), and some evidence of heterogeneity was found. The risk of pregnancy loss increased by 19% for every 150 mg/day increase in caffeine consumption in the random effect dose–response model, assuming linearity (OR 1.19, 95% CI 1.16–1.23). Compared with never/lowest drinkers, the pooled ORs were 1.04 (95% CI 0.97–1.11), 1.09 (95% CI 0.99–1.20), 1.16 (95% CI 1.05–1.28), 1.24 (95% CI 1.14–1.36), 1.34 (95% CI 1.22–1.46), 1.43 (95% CI 1.30–1.56), 1.52 (95% CI 1.38–1.66), 1.61 (95% CI 1.47–1.76), and 1.99 (95% CI 1.73-2.29) for caffeine consumptions of 50, 100, 149.5, 199.5, 249.5, 300, 349.5, 400, and 599.5 mg/day, respectively. When two studies were removed [24,34], no evidence of heterogeneity was found. The risk of pregnancy loss increased by 21% for every 150 mg/day increment in caffeine consumption in a fixed effects dose–response model, assuming linearity (OR 1.21, 95% CI 1.17–1.25).

For coffee, the nonlinear relationships of coffee consumption during pregnancy with pregnancy loss were noted (P for nonlinearity = 0.01; Fig. 4B). The risk of pregnancy loss increased by 8% for every increase

Downloaded for Anonymous User (n/a) at Saudi Digital Libra For personal use only. No other uses without permissio

in coffee consumption of two cups per day in the random effect dose– response model, assuming linearity (OR 1.08, 95% CI 1.07–1.10). Com- pared with never/lowest drinkers, the pooled ORs were 1.00 (95% CI 0.96–1.04), 1.06 (95% CI 1.01–1.11), 1.15 (95% CI 1.10–1.20), 1.24 (95% CI 1.18–1.30), 1.28 (95% CI 1.21–1.35), 1.38 (95% CI 1.29–1.48), 1.44 (95% CI 1.34–1.54), 1.58 (95% CI 1.43–1.75), and 1.64 (95% CI 1.41-1.91) for coffee consumptions of 1, 2, 3, 4, 4.5, 6, 7, 12, and 15 cups per day, respectively. The heterogeneity was no longer observed when three studies were removed [28,42,43]. A linear relationship was found. The risk of pregnancy loss increased by 3% for every increase in coffee consumption of two cups per day in a fixed effects dose– response model, assuming linearity (OR 1.03, 95% CI 1.01–1.05).

3.6. Publication bias

The Egger test showed no evidence of significant publication bias for the analysis between risk of pregnancy loss and caffeine consumption (P = 0.20) or coffee consumption (P = 0.24).

4. Discussion

The present meta-analysis has evaluated the association between pregnancy loss and caffeine and coffee consumption during pregnancy. Overall, it has shown that caffeine and coffee consumption during preg- nancy are significantly associated with pregnancy loss. Furthermore, a dose–response relationship was observed between risk of pregnancy loss and caffeine and coffee consumption during pregnancy.

The strengths of the present meta-analysis were that a systematic and quantitative assessment was used to detect the potential associa- tion between pregnancy loss and caffeine and coffee consumption during pregnancy, and that a dose–response relationship was also explored. Although heterogeneity was evident in these findings, the ev- idence of heterogeneity disappeared when the most heterogeneous

ry from ClinicalKey.com by Elsevier on February 20, 2019. n. Copyright ©2019. Elsevier Inc. All rights reserved.

Fig. 4. The dose–response analysis between risk of pregnancy loss and (A) caffeine or (B) coffee consumption during pregnancy with restricted cubic splines in a random effects dose– response model. The solid red line and the dashed blue line represent the estimated relative risk and its 95% confidence interval. The dashed black line represents the linear relationship.

121J. Li et al. / International Journal of Gynecology and Obstetrics 130 (2015) 116–122

studies were removed. Additionally, the results were confirmed in a subgroup analysis, suggesting that they were robust.

However, several potential limitations should also be considered. First, caffeinated food items—important sources of caffeine—could not be included because of limitations of the data available. Second, con- founding was inherent in all observational studies, and some case- control studies could not avoid recall bias, which cannot be solved at a meta-analysis level. Although most studies had adjusted for important confounders, such as age and smoking status, it is not possible to rule out potential confounding factors such as pregnancy history. Third, be- cause the consumption of coffee in all studies was determined according to the number of cups and the size of a cup might vary, misclassification of coffee consumption could be inevitable. Finally, because one database was searched, publication bias could be of concern. However, no evidence of publication bias was found.

Compared with the lowest level (the reference), moderate and heavy levels of caffeine consumption during pregnancy were associated with an increased risk of pregnancy loss. All the results in subgroup analyses including case-control studies and cohort studies also showed statistical significance. However, some evidence of heterogeneity was found across studies. When the most heterogeneous studies were removed, the heterogeneity was no longer significant. Heterogeneity was mainly from three studies [24,27,34], but the reasons were unclear. It might have been caused by varying characteristics—e.g. age, ethnic origin, sample size, adjusted estimates. Additionally, heterogeneity could have been caused by recall bias in case-control studies. Selection bias could also be another major source of heterogeneity, but was reduced by removing studies in sensitivity analyses.

Although some evidence of heterogeneity was found across studies, which might produce an overestimate of the true association, studies that gave rise to heterogeneity did not significantly change the overall estimates. Therefore, the present estimates seem to be reliable. Indeed, a previous meta-analysis [8] also suggested that caffeine consumption during pregnancy might be associated with an increased risk of spontaneous abortion.

Compared with the lowest level (reference), heavy consumption of coffee during pregnancy was associated with an increased risk of pregnancy loss. This result was robust across cohort and case-control studies. However, some evidence of heterogeneity was found across studies. When remote studies were removed, the heterogeneity was no longer significant. Three studies were the major sources of heteroge- neity [28,42,43], and some of the possible reasons are the same as for the studies of caffeine consumption. Additionally, in subgroup analyses, few studies of the relationship between coffee consumption during pregnancy and pregnancy loss were included.

Heterogeneity in the findings of the dose–response analyses was also evident. For caffeine, the evidence of heterogeneity disappeared when two studies were removed [24,34]. For coffee, the heterogeneity was no longer observed when three studies were removed [28,42,43]. The results of several previous studies [19,27,40] have also shown a

Downloaded for Anonymous User (n/a) at Saudi Digital Library For personal use only. No other uses without permission. C

linear relationship between caffeine consumption during pregnancy and pregnancy loss.

In conclusion, the present meta-analysis has indicated that caffeine and coffee consumption during pregnancy could be associated with an increased risk of pregnancy loss. A dose–response relationship has also been observed. Further studies with larger sample sizes and greater statistical power are needed to confirm the findings.

Conflict of interest

The authors have no conflicts of interest.

Supplementary data to this article can be found online at http:// dx.doi.org/10.1016/j.ijgo.2015.03.033.

References

[1] Robinson GE. Pregnancy loss. Best Pract Res Clin Obstet Gynaecol 2014;28(1): 169–78.

[2] Fernandes O, Sabharwal M, Smiley T, Pastuszak A, Koren G, Einarson T. Moderate to heavy caffeine consumption during pregnancy and relationship to spontaneous abortion and abnormal fetal growth: a meta-analysis. Reprod Toxicol 1998;12(4): 435–44.

[3] Francis SH, Sekhar KR, Ke H, Corbin JD. Inhibition of cyclic nucleotide phosphodies- terases by methylxanthines and related compounds. Handb Exp Pharmacol 2011; 200:93–133.

[4] Grosso LM, Bracken MB. Caffeine metabolism, genetics, and perinatal outcomes: a review of exposure assessment considerations during pregnancy. Ann Epidemiol 2005;15(6):460–6.

[5] Ebrahimi A, Habibi-Khorassani M, Akher FB, Farrokhzadeh A, Karimi P. Caffeine as base analogue of adenine or guanine: a theoretical study. J Mol Graph Model 2013;42:81–91.

[6] Aldridge A, Aranda JV, Neims AH. Caffeine metabolism in the newborn. Clin Pharmacol Ther 1979;25(4):447–53.

[7] Knutti R, Rothweiler H, Schlatter C. Effect of pregnancy on the pharmacokinetics of caffeine. Eur J Clin Pharmacol 1981;21(2):121–6.

[8] Furuhashi N, Sato S, Suzuki M, Hiruta M, Tanaka M, Takahashi T. Effects of caffeine ingestion during pregnancy. Gynecol Obstet Invest 1985;19(4):187–91.

[9] Zhao NQ. Assessment for Relative Risk RR = 1, RR N 1 and RR b 1 by Using the Re- sults of Odds Ratio Test. (in Chinese) J Shanghai Med Univ 1999;26(6):425–7.

[10] DerSimonian R, Laird N. Meta-analysis in clinical trials. Control Clin Trials 1986;7(3): 177–88.

[11] Mantel N, Haenszel W. Statistical aspects of the analysis of data from retrospective studies of disease. J Natl Cancer Inst 1959;22(4):719–48.

[12] Lohmueller KE, Pearce CL, Pike M, Lander ES, Hirschhorn JN. Meta-analysis of genetic association studies supports a contribution of common variants to susceptibility to common disease. Nat Genet 2003;33(2):177–82.

[13] Lau J, Antman EM, Jimenez-Silva J, Kupelnick B, Mosteller F, Chalmers TC. Cumula- tive meta-analysis of therapeutic trials for myocardial infarction. N Engl J Med 1992;327(4):248–54.

[14] Egger M, Davey Smith G, Schneider M, Minder C. Bias in meta-analysis detected by a simple, graphical test. BMJ 1997;315(7109):629–34.

[15] Discacciati A, Orsini N, Wolk A. Coffee consumption and risk of nonaggressive, ag- gressive and fatal prostate cancer—a dose-response meta-analysis. Ann Oncol 2014;25(3):584–91.

[16] Orsini N, Bellocco R, Greenland S. Generalized least squares for trend estimation from summarized dose–response data. Stata J 2006;6(1):40–57.

[17] Greenland S, Longnecker MP. Methods for trend estimation from summarized dose- response data, with applications to meta-analysis. Am J Epidemiol 1992;135(11): 1301–9.

from ClinicalKey.com by Elsevier on February 20, 2019. opyright ©2019. Elsevier Inc. All rights reserved.

122 J. Li et al. / International Journal of Gynecology and Obstetrics 130 (2015) 116–122

[18] Fenster L, Eskenazi B, Windham GC, Swan SH. Caffeine consumption during preg- nancy and spontaneous abortion. Epidemiology 1991;2(3):168–74.

[19] Infante-Rivard C, Fernández A, Gauthier R, David M, Rivard GE. Fetal loss associated with caffeine intake before and during pregnancy. JAMA 1993;270(24):2940–3.

[20] al-Ansary LA, Babay ZA. Risk factors for spontaneous abortion: a preliminary study on Saudi women. J R Soc Health 1994;114(4):188–93.

[21] Cnattingius S, Signorello LB, Annerén G, Clausson B, Ekbom A, Ljunger E, et al. Caf- feine intake and the risk of first-trimester spontaneous abortion. N Engl J Med 2000;343(25):1839–45.

[22] Rasch V. Cigarette, alcohol, and caffeine consumption: risk factors for spontaneous abortion. Acta Obstet Gynecol Scand 2003;82(2):182–8.

[23] Giannelli M, Doyle P, Roman E, Pelerin M, Hermon C. The effect of caffeine consump- tion and nausea on the risk of miscarriage. Paediatr Perinat Epidemiol 2003;17(4): 316–23.

[24] Maconochie N, DoyleP, Prior S, Simmons R. Risk factors for first trimester miscarriage– results from a UK-population-based case-control study. BJOG 2007;114(2):170–86.

[25] Matijasevich A, Barros FC, Santos IS, Yemini A. Maternal caffeine consumption and fetal death: a case-control study in Uruguay. Paediatr Perinat Epidemiol 2006; 20(2):100–9.

[26] Zhang BY, Wei YS, Niu JM, Li Y, Miao ZL, Wang ZN. Risk factors for unexplained re- current spontaneous abortion in a population from southern China. Int J Gynecol Obstet 2010;108(2):135–8.

[27] Stefanidou EM, Caramellino L, Patriarca A, Menato G. Maternal caffeine consumption and sine causa recurrent miscarriage. Eur J Obstet Gynecol Reprod Biol 2011;158(2): 220–4.

[28] Parazzini F, Chatenoud L, Di Cintio E, Mezzopane R, Surace M, Zanconato G, et al. Cof- fee consumption and risk of hospitalized miscarriage before 12 weeks of gestation. Hum Reprod 1998;13(8):2286–91.

[29] Nonaka T, Takakuwa K, Tanaka K. Analysis of the polymorphisms of genes coding biotransformation enzymes in recurrent miscarriage in the Japanese population. J Obstet Gynaecol Res 2011;37(10):1352–8.

[30] Agnesi R, Valentini F, Fedeli U, Rylander R, Meneghetti M, Fadda E, et al. Maternal ex- posures and risk of spontaneous abortion before and after a community oriented health education campaign. Eur J Public Health 2011;21(3):282–5.

Downloaded for Anonymous User (n/a) at Saudi Digital Libra For personal use only. No other uses without permissio

[31] Srisuphan W, Bracken MB. Caffeine consumption during pregnancy and association with late spontaneous abortion. Am J Obstet Gynecol 1986;154(1):14–20.

[32] Wilcox AJ, Weinberg CR, Baird DD. Risk factors for early pregnancy loss. Epidemiol- ogy 1990;1(5):382–5.

[33] Mills JL, Holmes LB, Aarons JH, Simpson JL, Brown ZA, Jovanovic-Peterson LG, et al. Moderate caffeine use and the risk of spontaneous abortion and intrauterine growth retardation. JAMA 1993;269(5):593–7.

[34] Domínguez-Rojas V, de Juanes-Pardo JR, Astasio-Arbiza P, Ortega-Molina P, Gordillo- Florencio E. Spontaneous abortion in a hospital population: are tobacco and coffee intake risk factors? Eur J Epidemiol 1994;10(6):665–8.

[35] Dlugosz L, Belanger K, Hellenbrand K, Holford TR, Leaderer B, Bracken MB. Maternal caffeine consumption and spontaneous abortion: a prospective cohort study. Epide- miology 1996;7(3):250–5.

[36] Fenster L, Hubbard AE, Swan SH, Windham GC, Waller K, Hiatt RA, et al. Caffeinated beverages, decaffeinated coffee, and spontaneous abortion. Epidemiology 1997; 8(5):515–23.

[37] Wen W, Shu XO, Jacobs Jr DR, Brown JE. The associations of maternal caffeine con- sumption and nausea with spontaneous abortion. Epidemiology 2001;12(1):38–42.

[38] Weng X, Odouli R, Li DK. Maternal caffeine consumption during pregnancy and the risk of miscarriage: a prospective cohort study. Am J Obstet Gynecol 2008;198(3): 279.e1–8.

[39] Savitz DA, Chan RL, Herring AH, Howards PP, Hartmann KE. Caffeine and miscarriage risk. Epidemiology 2008;19(1):55–62.

[40] Greenwood DC, Alwan N, Boylan S, Cade JE, Charvill J, Chipps KC, et al. Caffeine in- take during pregnancy, late miscarriage and stillbirth. Eur J Epidemiol 2010;25(4): 275–80.

[41] Armstrong BG, McDonald AD, Sloan M. Cigarette, alcohol, and coffee consumption and spontaneous abortion. Am J Public Health 1992;82(1):85–7.

[42] Wisborg K, Kesmodel U, Henriksen TB, Hedegaard M, Secher NJ. A prospective study of maternal smoking and spontaneous abortion. Acta Obstet Gynecol Scand 2003; 82(10):936–41.

[43] Andersen AM, Andersen PK, Olsen J, Grønbæk M, Strandberg-Larsen K. Moderate al- cohol intake during pregnancy and risk of fetal death. Int J Epidemiol 2012;41(2): 405–13.

ry from ClinicalKey.com by Elsevier on February 20, 2019. n. Copyright ©2019. Elsevier Inc. All rights reserved.