Cardiac remodeling in normotensive pregnancy and in pregnancy complicated by hypertension: systematic review and meta-analysis

Transcription

1 Ultrasound Obstet Gynecol 2017; 50: Published online in Wiley Online Library (wileyonlinelibrary.com). DOI: /uog Cardiac remodeling in normotensive pregnancy and in pregnancy complicated by hypertension: systematic review and meta-analysis S. DE HAAS 1,C.GHOSSEIN-DOHA 1, L. GEERTS 1,S.M.J.VANKUIJK 2, J. VAN DRONGELEN 3 andm.e.a.spaanderman 1 1 Department of Obstetrics and Gynaecology, Maastricht University Medical Centre (MUMC+), Maastricht, The Netherlands; 2 Department of Clinical Epidemiology and Medical Technology Assessment, Maastricht University Medical Centre (MUMC+), Maastricht, The Netherlands; 3 Department of Obstetrics and Gynaecology, Radboud University Nijmegen Medical Centre, Nijmegen, The Netherlands KEYWORDS: cardiac geometry; cardiac remodeling; complicated pregnancy; physiology; pre-eclampsia; pregnancy ABSTRACT Objective The aim of this systematic review and metaanalysis was to describe comprehensively the pattern of cardiac remodeling during normotensive human singleton pregnancy and to compare it with that of pregnancy complicated by hypertension. Methods We performed a meta-analysis of the current literature on cardiac remodeling during normotensive and complicated pregnancies. Literature was retrieved from PubMed (NCBI) and EMBASE (Ovid) databases. Included studies needed to report a reference measurement (matched non-pregnant control group, prepregnancy or postpartum) and measurements made during predetermined gestational-age intervals. Mean differences between reference and pregnancy data were calculated using the random-effects model described by DerSimonian and Laird. Results Forty-eight studies were included in the meta-analysis, with publication dates ranging from 1977 to During normotensive pregnancy, most geometric indices started to increase in the second trimester. Left ventricular mass (LVM) increased by (95% CI, ) g (24%), and relative wall thickness (RWT) increased by 0.03 (95% CI, ) (10%) compared with those in the reference group. During hypertensive pregnancy, LVM and RWT increased more than during normotensive pregnancy (92 (95% CI, ) g (95%) and 0.14 (95% CI, ) (56%), respectively). Conclusions During normotensive pregnancy, most cardiac geometric indices change from the second trimester onwards. Both LVM and RWT increase, by 20% and 10%, respectively, consistent with concentric rather than eccentric remodeling. Cardiac adaptation in hypertensive pregnancy deviates from that in healthy pregnancy by a greater change in LVM (95% increase from reference) and RWT (56% increase from reference). Copyright 2017 ISUOG. Published by John Wiley & Sons Ltd. INTRODUCTION During human pregnancy, the maternal cardiovascular system undergoes major hemodynamic alterations, ensuring a normotensive course of pregnancy. The pregnancy-induced decrease in peripheral vascular resistance and compensating renin angiotensin aldosterone levels, and with it the tremendous increase in plasma volume, provoke reversible morphological alterations in the maternal heart, resembling the hypertrophic alterations often seen in athletes 1. The increase in circulating volume during pregnancy is thought to lead to an increase in cardiac left ventricular (LV) mass (LVM) and a proportional increase in ventricular wall thickness. The ratio between ventricular end-diastolic radius and wall thickness, represented by the relative wall thickness (RWT), is thought to remain stable during normal pregnancy and is referred to as eccentric remodeling 2,3. Eccentric remodeling deviates from the concentric remodeling often seen in pathologically elevated pressure load conditions such as pre-eclampsia. Concentric remodeling is characterized by an increase in wall thickness without a proportional increase in ventricular dimensions and is represented by an increased RWT. Whereas eccentric remodeling presumes healthy cardiac function during raised circulatory Correspondence to: Dr S. de Haas, Department of Obstetrics & Gynaecology, Maastricht University Medical Centre, P. Debyelaan 25, 6229 HX Maastricht, The Netherlands ( sander.dehaas@maastrichtuniversity.nl) Accepted: 3 January 2017 Copyright 2017 ISUOG. Published by John Wiley & Sons Ltd. SYSTEMATIC REVIEW

2 684 de Haas et al. needs and resolves after delivery, concentric remodeling is a prelude to the attenuation of diastolic function as a consequence of reduced ventricular compliance, and persists after gestation. Considering the changed cardiac remodeling during hypertensive pregnancy, more studies are focusing on differences in cardiac adaptation during normotensive pregnancies and pregnancies complicated by hypertension. Not all studies have consistently reported uniform results, or managed to perform echocardiography at intervals of gestational age frequently enough to study the adaptation over the whole course of pregnancy in a longitudinal setting. Moreover, only a few studies have also included non-pregnant reference groups in the analysis, which raises the question of whether the differences observed between normotensive and gestational hypertensive pregnancy were already present prior to pregnancy or originate from maladaptive cardiac remodeling. In order to explore the pattern of cardiac remodeling in normotensive and complicated pregnancies we performed a systematic review and meta-analysis of the current literature to investigate the extent and time course of changes in geometric indices during human singleton pregnancy, and differences in adaptation between both groups. METHODS Literature search We conducted an extensive literature search for articles evaluating cardiac geometry, systolic function and diastolic function during normotensive and complicated pregnancy using PubMed (NCBI) and EMBASE (Ovid). These databases provided publications from 1946 to June 2016 for PubMed and 1974 to June 2016 for EMBASE. The search strategy, designed for a series of meta-analyses, focused on pregnancy and pregnancy-related complications combined with echocardiography, cardiac geometry, systolic function, diastolic function and the aortic valve, as detailed in Table S1. Study selection The identified studies were assessed for eligibility in two phases, as illustrated in Figure 1. Studies were included if they were original studies, not reviews, that reported numerical values (mean with SD, SE or 95% CIs) of cardiac geometric indices during human singleton pregnancy. The mean values with SD were requested from the authors if studies reported their data differently. Data were extracted independently of the methods or formulae used to estimate these indices. Studies were only included if they also reported reference data, measured at or after 6 weeks postpartum, before conception or in non-pregnant controls. Studies examining women with a pre-existing cardiovascular history or reporting data in languages other than English or Dutch were excluded because of the research team s inability to assess the quality of these studies. Studies were not excluded based on the use of medication if no direct effect on cardiac geometry or heart function had been earlier reported for these substances. Iron, folic acid, vitamins and mineral supplements were not considered to have any effect on the indices of interest. All studies were screened independently for eligibility by two investigators (S.dH. and C.G.D.) based on the title and abstract only. Discrepancies were resolved by mutual agreement after discussion. Then, full-text articles were screened by the same investigators for eligibility based on the inclusion and exclusion criteria. Data extraction Study characteristics (study design, sample size and methods), anthropometric data (age, non-pregnant weight, height, parity and duration of amenorrhea), and effect measures with SD, SE or 95% CIs were collected from the eligible studies and entered onto predesigned data collection forms. Cardiac geometric indices of interest were left ventricular end-diastolic dimension (LVEDD), left ventricular end-systolic dimension (LVESD), interventricular septum thickness (IVST), posterior wall thickness (PWT), LVM, LVM indexed for body surface area (BSA) (LVMi), RWT and left atrial (LA) diameter. Quality assessment The included studies were assessed for quality and risk of bias using a modified list of items described in the Quality In Prognosis Studies (QUIPS) tool, adapted for the purposes of this review 4. Studies were scored with a plus or minus for risk of bias on five domains including study participation, study attrition, variable measurement, data reporting and study design. Items that were not applicable to a study were not considered for the quality of that particular study. Studies were only scored for study attrition in cases in which subjects were lost to follow-up in longitudinal study designs. Studies with a positive score of > 60% for all the items under consideration were defined as high-quality (HQ), those with a score between 60% and 30% as moderate-quality (MQ) and those with a score of < 30% as low-quality (LQ). Statistical analysis Values for geometric indices were categorized into five different gestational-age intervals (< 14, 15 21, 22 28, and weeks). These gestational-age intervals were adapted from Abudu and Sofola 5, and enabled a precise categorization of almost all indices. SD was obtained from SE or 95% CI according to the Cochrane handbook for systematic reviews of interventions 6, when necessary. Changes in the different indices were calculated separately for these predefined intervals using a random-effects model, as described by DerSimonian and Laird 7. The random-effects model allows for interstudy variation, and was chosen since observational data of different pregnant populations had been used for

3 Cardiac changes in pregnancy 685 Inclusion Eligibility Screening Identification Records identified through PubMed and EMBASE (n = ) Records remaining after duplicates removed (n = 9988) Records screened (n = 9988) Full-text articles assessed for eligibility (n = 359) Studies included in quantitative synthesis (meta-analysis) (n = 48) Records excluded (n = 9629) Full-text articles excluded (n = 311): Not original data (n = 28) No eligible reference data (n = 81) Other comorbidity (n = 6) Unsuitable study design (n = 173) Data not usable (n = 9) Duplicate reporting (n = 2) Not obtained (n = 1) Not reporting on cardiac geometry (n = 11) for each individual study and used a locally weighted scatterplot smoother (LOESS) to graph curves over the course of pregnancy. Analysis was performed in R version (R Foundation for Statistical Computing, Vienna, Austria) 10. Meta-analysis and meta-regression analysis were performed using the meta package (R Foundation) 11. RESULTS Study and data selection The electronic search produced 9988 unique studies (Figure 1). Articles published in languages other than English or Dutch were manually excluded during the first selection (n = 1552). Screening by title and abstract resulted in excluding 9629 articles in total, leaving 359 articles eligible for full-text screening. From the studies assessed by full text, 311 studies matched at least one of the exclusion criteria. Among other exclusions, we excluded 173 studies based on an unsuitable study design and 81 studies based on the lack of eligible reference data. Two articles were excluded 12,13 because of suspected double reporting 14,15. One article could not be obtained 16. Forty-eight eligible studies were included in this meta-analysis. The data from one study 3 were converted to the desired format according to the Cochrane handbook for systematic reviews of interventions 17. Eleven included studies also reported data on pregnancies complicated by hypertension. Ten studies reported on pregnancies complicated by pre-eclampsia, while one study reported data on LVM, RWT, LVEDD and LVESD in pregnancies with pregnancy-induced hypertension (PIH) 18. All studies included reference data; one used prepregnancy data as reference, 28 used postpartum data as reference and 19 used a non-pregnant control group. Figure 1 Flowchart summarizing study selection process. anthropometric and clinical characteristics. Thus, we would not expect a common effect size. Egger s regression test for funnel-plot asymmetry was performed to test for the presence of publication bias 8. The primary outcome for each study was the mean difference in the geometric indices between pregnancy and reference, and is reported with 95% CIs. The relative increase from reference (percent) is reported in parentheses. The ratio between total heterogeneity and total variability (I 2 statistic) is presented as a measure of heterogeneity. I 2 can distinguish true heterogeneity from sampling variance and is expressed as a percentage 9. Sources of heterogeneity (type of reference, quality of study and formula used to calculate LVM, LVMi and RWT) and differences between normotensive and complicated pregnancies were investigated by meta-regression analysis using a mixed-effects model. Reference curves were constructed assuming that changes were normally distributed over the course of the pregnancy. We computed 5 th,50 th and 95 th percentiles Characteristics Study characteristics and anthropometric information of the study populations of the included articles on normotensive pregnancies and pregnancies complicated by hypertension are listed in Tables 1 and 2, respectively. Most studies reported overall characteristics, but did not present these characteristics per investigated subgroup, thus overall characteristics are reported here. Quality assessment The quality assessment of the included studies is depicted in Table S2. Of the 48 included studies, six (13%) matched the criteria for low quality, 31 (65%) for moderate quality and 11 (23%) for high quality. The items that scored lowest among all studies were the use of prepregnant data as reference values (n = 47) and details of ethnicity of the studied populations (n = 43). Left ventricular mass In the first trimester, LVM did not change in normotensive pregnancy compared with the non-pregnant reference

4 686 de Haas et al. Table 1 Characteristics of included studies in reference (Ref) and normotensive pregnancy (Preg) groups Subjects (n) Mean age (years) Mean non-pregnant weight (kg) Mean Parity/gravidity (n) height (cm) Nulli- Primi- Multi- Study Ref Preg Ref Preg Ref Preg Ref Preg Ref Preg Ref Preg Ref Preg Reference GA (weeks)* Melchiorre (2016) NP controls Ando (2015) NP controls 21 & 33 Cong (2015) PP (7.8 months) Cong (2015) NP controls Song (2015) PP (6 months) 35.8 Papadopoulou (2014) NP controls 8 36 Ducas (2014) PP (16 weeks) 34 Tso (2014) NP controls 22.1 & 34 Zanati Bazan (2013) NP controls Estensen (2013) PP (6 months) Savu (2012) PP (4 months) Dennis (2012) NP controls 36 Yosefy (2012) NP controls Rossi (2011) NP controls 20 & 32 Yoon (2011) NP controls 28.4 Pandey (2010) PP (8 12 weeks) Ogueh (2009) Prepregnant 6 36 Hamad (2009) PP (3 6 months) 33 Valensise (2008) PP (1 year) 24 Vasapollo (2008) PP (6 8 months) 24 Tzemos (2008) NP controls 2 nd trim. Fok (2006) PP (6 8 weeks) 8 36 Valensise (2006) PP (6 8 weeks) Freire (2006) PP (19 weeks) 33.7 Desai (2004) PP (6 12 weeks) Schannwell (2003) PP (8 weeks) 9 33 Schannwell (2002) PP (8 weeks) 9 33 Simmons (2002) PP (13 weeks) Kametas (2001) NP controls Borghi (2000) NP controls 30.9 Mesa (1999) PP (6.8 weeks) Wolfe (1999) PP (14 weeks) 2 nd &3 rd trim. Geva (1997) PP (12 14 weeks) Gilson (1997) PP (6 weeks) Poppas (1997) PP (6 months) 1 st to 3 rd trim. Mone (1996) PP (9 weeks) Mabie (1994) PP (12 weeks) 8 39 Nisell (1993) PP (3 months) 35 Sadaniantz (1992) PP (7.8 weeks) 34.4 Vered (1991) PP (8 weeks) 12 & 38 Escudero (1988) NP controls Airaksinen (1986) PP (14.6 weeks) Sánchez (1986) NP controls 32 Krajewski (1983) NP controls Kuźniar (1983) NP controls 36 Larkin (1980) NP controls 34.4 Katz (1978) PP (6 12 weeks) Rubler (1977) NP controls 13 to term Only first author is given for each study. Parity and gravidity data are in normal and italic font, respectively. *Unless stated otherwise. Mean value. GA, gestational age at measurement; NP, non-pregnant; PP, postpartum; trim., trimester. (mean difference, 3.01 (95% CI, 4.98 to 11.00) g (% increase relative to reference, 2.49 (95% CI, 4.12 to 9.09)%)) (Table 3, Figure 2). Thereafter, LVM increased progressively until weeks, at which point the mean difference was (95% CI, ) g (% increase, (95% CI, )%). Between 36 and 41 weeks, LVM increased by a mean of (95% CI, ) g (% increase, (95% CI, )%). There was statistically significant funnel-plot asymmetry in the week interval (P = 0.04). The corrected mean difference was (95% CI, ) g. We found that heterogeneity was significantly affected by MQ studies compared with LQ and HQ studies (P = 0.048). Additionally, the mean difference was statistically significantly higher using prepregnant data

5 Cardiac changes in pregnancy 687 Table 2 Characteristics of included studies in reference (Ref) group and group of pregnancies complicated by hypertension (CP) Subjects (n) Mean age (years) Mean non-pregnant weight (kg) Mean Parity/gravidity (n) height (cm) Nulli- Primi- Multi- Study Ref CP Ref CP Ref CP Ref CP Ref CP Ref CP Ref CP Reference GA (weeks) Cong (2015) NP controls 29 & 36.4 Dennis (2012) NP controls 36 Hamad (2009) PP (18 weeks) 35 Valensise (2008) PP (1 year) 24 Vasapollo (2008) PP (28 weeks) 24 Valensise (2006) PP (7 weeks) Simmons (2002) PP (15 weeks) 35 Borghi (2000) * NP controls 28.4 Escudero (1988) NP controls > 26 Sánchez (1986) NP controls 32 Kuźniar (1983) NP controls 35 Only first author is given for each study. Parity and gravidity data are in normal and italic font, respectively. *Mean value. GA, gestational age at measurement; NP, non-pregnant; PP, postpartum. Table 3 Pooled changes in geometric indices during normotensive pregnancy Geometric Gestational-age interval (weeks) index < IVST (cm) MD 0.03 ( 0.07 to 0.00) 0.01 ( 0.02 to 0.04) 0.03 ( 0.01 to 0.07) 0.09 (0.06 to 0.13) 0.03 ( 0.01 to 0.07) % 3.82 ( 8.92 to 0.00) 1.33 ( 2.66 to 5.33) 3.89 ( 1.30 to 9.09) (7.47 to 16.19) 3.93 ( 1.31 to 9.17) PWT (cm) MD 0.00 ( 0.03 to 0.03) 0.02 ( 0.02 to 0.07) 0.03 (0.00 to 0.07) 0.08 (0.06 to 0.11) 0.07 (0.02 to 0.11) % 0.00 ( 3.80 to 3.80) 2.44 ( 2.44 to 8.54) 3.96 (0.00 to 9.23) 9.71 (7.28 to 13.35) 9.01 (2.58 to 14.16) RWT MD 0.01 (0.00 to 0.02) 0.01 ( 0.03 to 0.05) 0.03 (0.02 to 0.03) 0.03 (0.02 to 0.05) 0.03 (0.02 to 0.05) % 3.62 (0.00 to 7.25) 2.94 ( 8.82 to 14.71) (6.91 to 10.36) 9.13 (6.09 to 15.22) (6.83 to 17.08) LVEDD (cm) MD 0.06 ( 0.11 to 0.01) 0.09 ( 0.01 to 0.20) 0.13 (0.09 to 0.17) 0.15 (0.09 to 0.22) 0.21 (0.13 to 0.28) % 1.30 ( 2.39 to 0.22) 1.96 ( 0.22 to 4.35) 2.83 (1.96 to 3.70) 3.30 (1.98 to 4.84) 4.55 (2.82 to 6.06) LVESD (cm) MD 0.11 ( 0.17 to 0.04) 0.01 ( 0.16 to 0.18) 0.06 (0.01 to 0.11) 0.11 (0.06 to 0.16) 0.13 (0.06 to 0.20) % 3.72 ( 5.74 to 1.35) 0.34 ( 5.40 to 6.08) 2.10 (0.35 to 3.85) 3.83 (2.09 to 5.57) 4.38 (2.02 to 6.73) LA diameter (mm) MD 0.06 ( 0.67 to 0.56) 1.41 ( 1.28 to 4.10) 2.33 (1.05 to 3.62) 2.31 (1.44 to 3.19) 4.24 (3.76 to 4.71) % 0.19 ( 2.21 to 1.85) 4.39 ( 3.99 to 12.77) 7.54 (3.40 to 11.71) 7.40 (4.61 to 10.21) (12.08 to 15.13) LVM (g) MD 3.01 ( 4.98 to 11.00) 4.30 ( 4.63 to 13.24) (7.72 to 21.27) (19.73 to 37.00) (17.00 to 32.87) % 2.49 ( 4.12 to 9.09) 3.46 ( 3.72 to 10.65) (6.50 to 17.91) (16.42 to 30.79) (13.71 to 26.51) LVMi (g/m 2 ) MD 0.16 ( 4.58 to 4.90) 3.18 ( 5.60 to 11.96) 6.15 ( 0.42 to 12.71) (5.53 to 21.83) 7.57 (2.59 to 12.54) % 0.24 ( 6.87 to 7.35) 4.52 ( 7.95 to 16.99) 8.84 ( 0.60 to 18.27) (7.83 to 30.91) (3.79 to 18.36) Values are reported as absolute mean difference (MD) and relative change (%) compared with reference, with 95% CI. IVST, interventricular septum thickness; LA, left atrial; LVEDD, left ventricular end-diastolic dimension; LVESD, left ventricular end-systolic dimension; LVM, left ventricular mass; LVMi, left ventricular mass indexed for body surface area; PWT, posterior wall thickness; RWT, relative wall thickness. as reference compared with using postpartum or non-pregnant reference groups (P = 0.04). The formula used to calculate LVM did not affect the heterogeneity (0.37 P 0.95). In pregnancies complicated by hypertension, LVM increased compared with the reference, and by weeks the mean difference was (95% CI, ) g (% increase, (95% CI, )%). It increased further as pregnancy progressed, until at weeks the mean difference was (95% CI, ) g (% change, (95% CI, )%) compared with reference data (Table 4, Figure 3). Only one included study reported LVM at term (36 41 weeks) in pre-eclamptic women, with an increase of 92 (95% CI, ) g (% increase, (95% CI, )%) 19.The increase was significantly higher in pre-eclamptic and pregnancies complicated by PIH compared with normotensive pregnancies (P < ). Left ventricular mass index We identified studies indexing LVM for BSA (g/m 2 ), height (g/m or g/m 2.7 ) and volume (g/cm 3 ), although the latter two were uncommon. Therefore, the meta-analysis was conducted for LVM indexed for BSA only (LVMi). We found no change in LVMi in the first trimester of pregnancy (mean difference, 0.16 (95% CI, 4.58 to 4.90) g/m 2 (% change, 0.24 (95% CI, 6.87 to 7.35)%)) compared with the non-pregnant reference (Table 3, Figure S1). Thereafter, LVMi increased progressively up to weeks, but the first statistically significant increase occurred in the period of weeks, with a mean difference of (95% CI, ) g/m 2 (% increase, (95% CI, )%). Publication bias was present in the data for the first trimester (P = 0.04), the week interval (P = ) and the week interval (P = 0.003). Corrected mean differences were 4.80 (95% CI, to 0.57) g/m 2,

6 688 de Haas et al. Study Sample size Preg Ref MD (95% CI) < 14 weeks Melchiorre (2016) (4.92 to 25.08) Papadopoulou (2014) ( to 21.40) Savu (2012) ( 8.53 to 12.53) Ogueh (2009) ( 2.90 to 23.90) Ogueh (2009) ( 3.77 to 27.63) Fok (2006) ( to 3.83) Schannwell (2003) ( to 5.11) Schannwell (2002) ( to 5.85) Simmons (2002) ( to 7.06) Mesa (1999) ( to 3.38) Geva (1997) (6.40 to 28.60) Mone (1996) ( to 15.44) Mabie (1994) ( to 10.04) Mabie (1994) ( to 10.66) Vered (1991) ( 4.43 to 46.43) Katz (1978) (19.38 to 28.62) Summary ( 4.98 to 11.00) I 2 = 91% weeks Estensen (2013) ( to 4.84) Ogueh (2009) (8.98 to 38.62) Fok (2006) ( to 10.79) Desai (2004) ( to 0.66) Kametas (2001) (14.36 to 34.84) Wolfe (1999) ( to 12.96) Geva (1997) (12.72 to 40.08) Gilson (1997) ( 7.07 to 7.07) Mone (1996) ( to 18.44) Mabie (1994) ( to 17.34) Mabie (1994) ( to 21.69) Summary ( 4.63 to 13.24) I 2 = 81% weeks Melchiorre (2016) (8.58 to 27.42) Papadopoulou (2014) ( 5.96 to 39.34) Tso (2014) ( to 15.33) Estensen (2013) ( to 4.08) Savu (2012) (4.76 to 27.24) Yoon (2011) (6.52 to 33.48) Pandey (2010) ( to 2.84) Ogueh (2009) (12.90 to 40.50) Valensise (2008) (25.30 to 28.70) Vasapollo (2008) (25.25 to 30.75) Desai (2004) ( to 18.38) Desai (2004) ( 6.81 to 20.81) Schannwell (2003) (47.17 to 58.83) Schannwell (2002) (41.86 to 54.14) Simmons (2002) ( 2.62 to 16.62) Mesa (1999) ( to 10.67) Wolfe (1999) ( 8.04 to 18.44) Gilson (1997) ( 6.33 to 10.33) Mabie (1994) (2.44 to 25.56) Katz (1978) (2.27 to 11.73) Summary (7.72 to 21.27) I 2 = 95% Change in LVM (g) Figure 2 Continued over. Forest plot of changes in left ventricular mass (LVM) in normotensive pregnancies. Only first author of each study given. Some studies appear more than once within a gestational-age interval because they provided data for different gestational weeks within that interval. Preg, pregnant group; Ref, reference group.

7 Cardiac changes in pregnancy 689 Study weeks Melchiorre (2016) (11.97 to 32.03) Ducas (2014) (47.00 to 57.00) Papadopoulou (2014) ( 7.77 to 37.89) Tso (2014) (7.40 to 37.00) Savu (2012) (20.45 to 43.55) Pandey (2010) (16.24 to 35.86) Hamad (2009) (15.69 to 24.31) Fok (2006) (1.18 to 30.02) Freire (2006) ( to 28.32) Valensise (2006) (29.53 to 44.47) Desai (2004) (2.87 to 35.13) Desai (2004) ( 3.28 to 25.28) Schannwell (2003) (67.96 to 80.04) Schannwell (2002) (54.64 to 67.36) Simmons (2002) (10.28 to 33.72) Borghi (2000) (7.17 to 50.43) Mesa (1999) (0.02 to 23.98) Geva (1997) (24.86 to 48.74) Mone (1996) (7.56 to 38.44) Mabie (1994) ( 8.69 to 26.69) Mabie (1994) (12.60 to 49.40) Escudero (1988) (12.72 to 55.28) Sánchez (1986) (19.90 to 54.10) Larkin (1980) (4.70 to 33.90) Summary (19.73 to 37.00) I 2 = 94% weeks Melchiorre (2016) (25.11 to 44.89) Estensen (2013) ( 8.85 to 16.85) Dennis (2012) (21.26 to 46.74) Ogueh (2009) (19.73 to 45.47) Desai (2004) (9.25 to 52.75) Wolfe (1999) ( 3.15 to 21.79) Geva (1997) (32.65 to 55.95) Gilson (1997) ( 1.07 to 13.07) Mone (1996) (5.82 to 38.18) Mabie (1994) (19.37 to 60.63) Vered (1991) (8.82 to 61.18) Katz (1978) (14.34 to 23.66) Summary (17.00 to 32.87) I 2 = 83% Sample size Preg Ref Change in LVM (g) MD (95% CI) Figure 2 (Continued) Forest plot of change in left ventricular mass (LVM) in normotensive pregnancies. Only first author of each study is given. Some studies appear more than once within a gestational-age interval because they provided data for different gestational weeks within that interval. MD, mean difference; Preg, pregnant group; Ref, reference group (95% CI, ) g/m 2 and (95% CI, ) g/m 2, respectively. Study quality contributed significantly to heterogeneity (MQ vs LQ and HQ, P = ). LVMi, calculated with the formula LVMi = 1.04 ((LVEDD + PWT + IVST) 3 (LVEDD) 3 ) 13.6g, were, on average, statistically significantly higher than the values given by the other formulae (P = 0.02). Four LVMi measurements in pregnancies complicated by hypertension were identified, ranging from 22 weeks to term (Table 4, Figure S2). With the latter data we were unable to observe a statistically significant difference in LVMi increase between normal pregnancies and pregnancies complicated by pre-eclampsia (P = 0.11). Relative wall thickness In the first trimester, RWT did not change (mean difference, 0.01 (95% CI, ) (% change, 3.62 (95% CI, )%)) (Table 3, Figure 4). Only one included study provided a value for RWT suitable for inclusion in the week gestational-age interval (0.01 (95% CI, 0.03 to 0.05) (% change, 2.94 (95% CI, 8.82 to 14.71)%)) 20. The first statistically significant increase in RWT was seen between 22 and 28 weeks, with a value

8 690 de Haas et al. Table 4 Pooled changes in geometric indices during pregnancy complicated by hypertension, and comparison with normotensive pregnancy Gestational-age interval (weeks) Geometric index P* IVST (cm) MD 0.11 (0.02 to 0.20) 0.16 (0.12 to 0.21) < % (2.47 to 24.68) (14.81 to 25.91) PWT (cm) MD 0.09 (0.06 to 0.12) 0.15 (0.12 to 0.18) % (7.90 to 15.80) (14.79 to 22.19) RWT MD 0.05 (0.02 to 0.08) 0.06 (0.04 to 0.09) 0.14 (0.09 to 0.19) < % (6.01 to 24.04) (12.07 to 27.16) (36.00 to 76.00) LVEDD (cm) MD 0.07 ( 0.11 to 0.25) 0.18 (0.10 to 0.26) 0.33 (0.14 to 0.52) 0.96 % 1.49 ( 2.34 to 5.31) 3.85 (2.14 to 5.56) 7.47 (3.17 to 11.77) LVESD (cm) MD 0.33 (0.07 to 0.59) 0.14 (0.01 to 0.27) 0.32 (0.16 to 0.48) 0.15 % (2.54 to 21.38) 4.77 (0.34 to 9.21) (5.70 to 17.09) LA diameter (mm) MD 5.00 (3.45 to 6.55) 3.17 (1.23 to 5.11) 7.85 (6.56 to 9.15) 0.02 % (11.20 to 21.27) (4.03 to 16.74) (22.48 to 31.21) LVM (g) MD (14.83 to 42.18) (46.77 to 74.26) 92 (75.46 to ) < % (11.89 to 33.83) (36.93 to 58.63) (77.79 to ) LVMi (g/m 2 ) MD (11.51 to 32.49) (13.32 to 30.24) (8.73 to 23.27) 0.11 % (13.08 to 36.92) (20.69 to 46.98) (14.81 to 39.44) Values are reported as absolute mean difference (MD) and relative change (%) compared with reference, with 95% CI. *Comparison between hypertensive and normotensive pregnancies. IVST, interventricular septum thickness; LA, left atrial; LVEDD, left ventricular end-diastolic dimension; LVESD, left ventricular end-systolic dimension; LVM, left ventricular mass; LVMi, left ventricular mass indexed for body surface area; PWT, posterior wall thickness; RWT, relative wall thickness. Study Sample size Preg Ref weeks Valensise (2008) Valensise (2008) Vasapollo (2008) Borghi (2000) Summary I 2 = 79% weeks Hamad (2009) Valensise (2006) Valensise (2006) Simmons (2002) Escudero (1988) Sánchez (1986) Summary I 2 = 93% weeks Dennis (2012) Summary Not applicable Change in LVM (g) MD (95% CI) (5.80 to 22.20) (7.15 to 46.85) (20.16 to 35.84) (33.79 to 80.21) (14.83 to 42.18) (48.25 to 55.75) (49.45 to 58.55) (29.29 to 44.71) (9.88 to 62.12) (21.67 to 94.33) ( to ) (46.77 to 74.26) (75.46 to ) (75.46 to ) Figure 3 Forest plot of change in left ventricular mass (LVM) in pregnancies complicated by hypertension. Only first author of each study is given. Some studies appear more than once within a gestational-age interval because they provided data for different gestational weeks within that interval. MD, mean difference; Preg, pregnant group; Ref, reference group. of 0.03 (95% CI, ) (% increase, (95% CI, )%). Funnel-plot asymmetry was found in this interval (P = 0.04) but did not affect the mean difference. RWT remained constant during the first half of the third trimester (mean difference, 0.03 (95% CI, ) (% change, 9.13 (95% CI, )%)) and the end of the third trimester (mean difference, 0.03 (95% CI, ) (% change, (95% CI, )%)). Statistically significant funnel-plot asymmetry was found in the week interval (P = 0.03), with a corrected mean difference of 0.01 (95% CI, ). Heterogeneity was significantly affected by using postpartum reference data compared with non-pregnant control data (P = 0.002) and MQ vs LQ and HQ studies (P < ). No significant difference was found in RWT increase between studies that used different formulae (2 (PWT/LVEDD) vs (IVST + PWT)/LVEDD) to calculate RWT (P = 0.25).

9 Cardiac changes in pregnancy 691 Study Sample size Preg Ref MD (95% CI) < 14 weeks Melchiorre (2016) ( 0.01 to 0.03) Cong (2015) ( 0.02 to 0.06) Cong (2015) ( 0.01 to 0.03) Savu (2012) ( 0.01 to 0.03) Simmons (2002) (0.00 to 0.02) Summary (0.00 to 0.02) I 2 = 0% weeks Ando (2015) ( 0.03 to 0.05) Summary ( 0.03 to 0.05) Not applicable weeks Melchiorre (2016) ( 0.01 to 0.03) Cong (2015) ( 0.01 to 0.07) Cong (2015) (0.00 to 0.04) Savu (2012) (0.00 to 0.04) Yoon (2011) ( 0.01 to 0.03) Valensise (2008) (0.03 to 0.03) Vasapollo (2008) (0.02 to 0.04) Simmons (2002) (0.01 to 0.03) Summary (0.02 to 0.03) I 2 = 29% weeks Melchiorre (2016) (0.01 to 0.07) Ando (2015) (0.00 to 0.06) Savu (2012) (0.01 to 0.05) Hamad (2009) (0.00 to 0.02) Valensise (2006) (0.02 to 0.06) Simmons (2002) (0.02 to 0.04) Escudero (1988) (0.04 to 0.06) Summary (0.02 to 0.05) I 2 = 87% weeks Melchiorre (2016) (0.03 to 0.07) Cong (2015) (0.00 to 0.08) Cong (2015) (0.01 to 0.05) Zanati Bazan (2013) (0.00 to 0.04) Summary (0.02 to 0.05) I 2 = 38% Change in RWT Figure 4 Forest plot of change in relative wall thickness (RWT) in normotensive pregnancies. Only first author of each study given. MD, mean difference; Preg, pregnant group; Ref, reference group. In pregnancies complicated by hypertension, RWT increased by 0.05 (95% CI, ) (% change, (95% CI, )%) between 22 and 28 weeks compared with reference and continued to increase until weeks, at which point the mean difference was 0.06 (95% CI, ) (% change, (95% CI, )%). In addition to pre-eclamptic pregnancies, the latter interval included pregnancies complicated by PIH from one study 18. At term, only one included study investigated RWT in pregnancies complicated by late-onset pre-eclampsia (n = 34) 21. Compared with non-pregnant controls, an increase of 0.14 (95% CI, ) (% increase, (95% CI, )%) was observed (Table 4, Figure 5). The maximum increase was significantly higher, at 45.6%, than the increase seen in normotensive pregnancies (P < ). Interventricular septum thickness During normotensive pregnancies, IVST did not change up to the early third trimester compared with reference (Table 3, Figure S3). IVST statistically significantly increased between 29 and 35 weeks, by 0.09 (95% CI, ) cm (% increase, (95% CI, )%) compared with reference. At term, IVST

10 692 de Haas et al. Study Sample size Preg Ref MD (95% CI) weeks Valensise (2008) (0.04 to 0.08) Valensise (2008) ( 0.02 to 0.04) Vasapollo (2008) (0.06 to 0.08) Summary (0.02 to 0.08) I 2 = 86% weeks Cong (2015) (0.09 to 0.19) Hamad (2009) (0.04 to 0.04) Valensise (2006) (0.06 to 0.08) Valensise (2006) (0.09 to 0.11) Simmons (2002) (0.00 to 0.06) Escudero (1988) (0.02 to 0.04) Summary (0.04 to 0.09) I 2 = 96% weeks Cong (2015) (0.09 to 0.19) Summary (0.09 to 0.19) Not applicable Change in RWT Figure 5 Forest plot of change in relative wall thickness (RWT) in pregnancies complicated by hypertension. Only first author of each study is given. Some studies appear more than once within a gestational-age interval because they provided data for different gestational weeks within that interval. MD, mean difference; Preg, pregnant group; Ref, reference group. had not changed compared with reference. Funnel-plot asymmetry was present at term (P = 0.03). The corrected mean difference was 0.01 (95% CI, 0.05 to 0.03) cm compared with reference. No significant contributor to heterogeneity was found, either for type of reference group (postpartum vs non-pregnant controls, P = 0.82) or for quality of the included studies (P = 0.53 and P = 0.62). During the late second trimester in pregnancies complicated by pre-eclampsia, IVST increased by 0.11 (95% CI, ) cm (% increase, (95% CI, )%) compared with reference. In the early third trimester, IVST further increased, by 0.16 (95% CI, ) cm (% change, (95% CI, )%) compared with the non-pregnant condition (Table 4, Figure S4). This increase in pregnancy complicated by hypertension was significantly higher than the changes observed in normotensive pregnancy (P < ). Posterior wall thickness One study was omitted from the meta-analysis and meta-regression analysis owing to a SD of zero in the reference data 22. During the first trimester of normal pregnancies, no alterations in thickness of the LV posterior wall were observed compared with reference (mean difference, 0.0 (95% CI, 0.03 to 0.03) cm). The first statistically significant increase in PWT, of 0.08 (95% CI, ) cm (% change, 9.71 (95% CI, )%) compared with the reference, occurred between 29 and 35 weeks. The posterior wall was thicker by 0.07 (95% CI, ) cm (% change, 9.01 (95% CI, )%) at term compared with reference (Table 3, Figure S5). No statistically significant funnel-plot asymmetry was found in the data (0.052 P 0.66). Meta-regression analysis showed no contribution of study quality (0.14 P 0.68) nor for type of reference measurement (postpartum vs non-pregnant controls, P = 0.67) to the heterogeneity. In the pre-eclampsia group, the posterior wall thickened by 0.09 (95% CI, ) cm (% change, (95% CI, )%) between 22 and 28 weeks compared with reference data (Table 4, Figure S6). A further increase was observed between 29 and 35 weeks (0.15 (95% CI, ) cm (% change, (95% CI, )%)). The increase observed in the pregnancy group complicated by hypertension was significantly higher than the increase observed in the normotensive pregnancy group (P = ). Left ventricular end-diastolic dimension LVEDD decreased in the first trimester by 0.06 (95% CI, 0.11 to 0.01) cm (% change, 1.30 (95% CI, 2.39 to 0.22)%), after which it progressively increased, by 0.21 (95% CI, ) cm (% change, 4.55 (95% CI, )%) at term compared with reference (Table 3, Figure S7). Funnel-plot asymmetry was observed in the LVEDD values between 36 and 41 weeks (P = 0.046). After correction for funnel-plot asymmetry, the increase in LVEDD compared with reference was 0.32 (95% CI, ) cm.

11 Cardiac changes in pregnancy 693 Only postpartum reference data compared with non-pregnant controls reference data contributed significantly to the heterogeneity (P = 0.001). In pregnancies complicated by hypertension, LVEDD did not change compared with reference between 22 and 28 weeks (mean increase, 0.07 (95% CI, 0.11 to 0.25) cm (% change, 1.49 (95% CI, 2.34 to 5.31)%)). LVEDD increased by 0.18 (95% CI, ) cm (% change, 3.85 (95% CI, )%) between 29 and 35 weeks compared with reference (Table 4, Figure S8). Only one included study reported LVEDD between 36 and 41 weeks, with a value of 0.33 (95% CI, ) cm (% change, 7.47 (95% CI, )%) 21. We did not find a statistically significant difference in LVEDD increase between pregnancies complicated by hypertension and normotensive pregnancies (P = 0.96). Left ventricular end-systolic dimension LVESD decreased during the first trimester by 0.11 (95% CI, 0.17 to 0.04) cm (% change, 3.72 (95% CI, 5.74 to 1.35)%) and thereafter progressively increased with advancing pregnancy, with a maximum increase of 0.13 (95% CI, ) cm (% change, 4.38 (95% CI, )%) at term compared with reference (Table 3, Figure S9). Funnel-plot asymmetry was observed in the week gestational-age interval (P = 0.03), and after correcting for funnel-plot asymmetry the increase was 0.20 (95% CI, ) cm compared with reference. Meta-regression analysis revealed a significant contribution of using postpartum reference data and LQ studies to the observed heterogeneity (P = and P = 0.016, respectively). In pregnancies complicated by hypertension, the increase in LVESD was 0.14 (95% CI, ) cm (% change, 4.77 (95% CI, )%) in the early third trimester (Table 4, Figure S10). The increase was not significantly different from that in uncomplicated pregnancies (P = 0.15). Only two studies were identified that investigated LVESD in pregnancies complicated by pre-eclampsia over other time intervals. One included study reported an increase of 0.33 (95% CI, ) cm (% change, (95% CI, )%) in 40 pre-eclamptic pregnancies at a mean gestational age of 28 weeks 23. Another study reported an LVESD increase of 0.32 (95% CI, ) cm (% change, (95% CI, )%) in 34 pre-eclamptic women between 36 and 41 weeks compared with non-pregnant controls 21. Left atrial diameter LA diameter did not change in the first trimester compared with reference values from normotensive pregnancies. LA diameter started to increase between 22 and 28 weeks, with a maximum increase at term of 4.24 (95% CI, ) mm (% change, (95% CI, )%) compared with reference (Table 3, Figure S11). No funnel-plot asymmetry was observed (0.07 P 0.80). Meta-regression analysis showed no statistically significant contribution of postpartum vs non-pregnant controls as the reference group (P = 0.10) nor for study quality (LQ vs MQ and HQ, P = 0.30 and MQ vs LQ and HQ, P = 0.40) to the heterogeneity. In pregnancies complicated by hypertension, LA diameter increased by 3.17 (95% CI, ) mm (% change, (95% CI, )%) compared with reference data in the first half of the third trimester and continued to increase, such that in the last weeks of these pregnancies the increase was 7.85 (95% CI, ) mm (% change, (95% CI, )%) (Table 4, Figure S12). Only one study reported LA diameter data between 22 and 28 weeks, with an increase of 5.00 (95% CI, ) mm 23. The increase in LA diameter was statistically significantly higher than the increase seen in normotensive pregnancies (P = 0.02). Reference curves The reference curves calculated from intrastudy variance are depicted in Figures S13 S19. The mean assessment period was estimated from studies reporting gestational intervals by selecting the center of the interval. Each point estimate is indicated as a value from a low-, moderate- or high-quality study as determined by the quality assessment. Only LVM values calculated using the Devereux formula were used for the reference curve of LVM. DISCUSSION This systematic review with meta-analysis comprehensively describes and expands our understanding of cardiac remodeling during normotensive pregnancy and deviated remodeling during hypertensive pregnancy. Mixed remodeling during normotensive pregnancies Pregnancy is characterized by a substantially increased blood volume During pregnancy, an increment in cardiac output is achieved by a rise in ventricular radius paralleled with a rise in preload, resulting initially in increased end-diastolic wall stress. Following LaPlace s law, increased wall stress can be offset by either ventricular wall thickening or by decreased ventricular radius. We have shown that end-diastolic and end-systolic LV dimensions started to increase significantly at 22 weeks gestation, reaching a maximum increase of 4.6% and 4.4%, respectively, towards term. The increase in LV dimensions is paralleled by a relatively much larger increase in LA diameter, most probably to accommodate the increased volume load in a high-cardiac output state 21,27. The interventricular septum and LV wall thicken significantly, starting at 29 weeks gestation. Simultaneously,

12 694 de Haas et al. we observed a significant increase in LVM from 22 weeks onwards, reaching a maximum change of 23.6% in the third trimester, which seems to be in line with the results of the study of Meah et al. 28, who presented a reference curve for weighted means of LVM clustered for five different gestational-age intervals. Our reference graph was constructed by plotting the means of LVM against the mean gestational age reported by each individual study. A LOESS curve was plotted with the 5 th and 95 th percentiles weighted by the inverse variance, which may partly explain the slight differences between our curve and theirs. We observed a significant increase in RWT, of 10%, by the end of normotensive gestation, consistent with concentric remodeling, in contrast to what was previously thought to be eccentric remodeling 1.Ithas been suggested that cardiac remodeling during pregnancy is a physiological adaptation in response to increased volume load and increased cardiac demands 3,29. However, a recent study linked the persistent remodeling of the maternal heart at term pregnancy with impaired myocardial relaxation and diastolic dysfunction in a significant proportion of a healthy pregnant study population, indicating that at a certain point, the exaggeratedly increased volume load may contribute to a shift in the pathological behavior of LV diastolic function 3. Whether and at what point during pregnancy this shift occurs is open to speculation, but should involve volume- and pressure-loading conditions as well as circulating biochemical factors that affect myocyte hypertrophy, such as renin angiotensin aldosterone system hormones, lipids, glucose and insulin factors, and sex hormones such as estrogen and progesterone 30. Unfortunately, none of the studies combined cardiac adaptation with changes in volume and pressure load and biochemical factors, thus not helping our understanding of the complete physiological mechanisms. We showed in an earlier meta-analysis that the increase in circulating plasma volume reaches a plateau in the late second trimester, while cardiac remodeling does not plateau earlier than the third trimester 31. We investigated whether the type of reference would affect the observed geometric changes. The mean differences in LV dimensions in the postpartum group were systematically lower throughout gestation. This is in line with our current understanding that cardiac-chamber dimensions do not return to baseline values within weeks postpartum. This may explain our finding of decreased LV dimensions in the first trimester, as most studies (58%) used postpartum values as reference. It may be that our chosen interval is still too short to allow full recovery towards prepregnancy values, especially in pregnancies complicated by hypertension, in which changes seem to persist for several months to a year postpartum 1,32,33. Trend towards concentric remodeling in hypertensive pregnancy LVM and RWT increased in hypertensive pregnancies. Although RWT increased by approximately five times that in normotensive pregnancies, the change in LVEDD did not differ statistically significantly from that of normotensive pregnancies, suggesting that the excessive increase in RWT relates mainly to an excessive increase in wall thickness. Extensive thickening of LV walls may relate to the increased pressure load rather than lower volume load that are known to accompany hypertensive pregnancy 21. Volume load is known to be an important determinant of LV dimensions and may not per se affect adaptation of these dimensions, but may still affect the baseline dimensions. The LA was more enlarged in hypertensive pregnancies, which might be secondary to an increased LA afterload due to decreased LV compliance caused by hypertrophy 21,27. Nevertheless, it is evident that cardiac morphological alterations differ significantly between normotensive pregnancies and pregnancies complicated by hypertension. The number of eligible studies on pregnancies complicated by hypertension was low and they did not provide data in early pregnancy before the clinical onset of hypertension, thus preventing us from describing the early pattern of change seen in such pregnancies. Limitations of the study Several shortcomings of this study need to be addressed. First, some longitudinal studies reported reference and pregnant values from the same women. These studies often did not present sufficient information to determine the covariance between measurements. Instead of estimating covariances, we ignored their dependence. Not taking the covariance into account can only have resulted in slightly conservative estimates of precision, since the variance of the mean difference decreases as the covariance between the two measurements increases. Point estimates are unaffected by these dependent observations. Second, we are aware that there was not always enough power to test for funnel-plot asymmetry. Therefore, the corrected effect sizes should be interpreted with caution. Finally, reference curves are preferably constructed using individual patient data and specifically designed statistical methods that normalize the data to account for skewness, if present. We combined estimates from previously published studies, making it impossible to test the assumptions we made. As a result, the percentile curves may be less accurate than are curves estimated using individual patient data. Moreover, the pooled estimates of the geometric indices cannot account for individual differences throughout pregnancy. Conclusions We have shown in this meta-analysis that both LVM and RWT increase during normotensive pregnancy, consistent with concentric rather than eccentric remodeling. Furthermore, cardiac adaptation in hypertensive pregnancies deviates from that in healthy pregnancies by a greater change in LVM (95% increase from reference) and RWT (56% increase from reference).