This systematic review complies with PRISMA statement.5

The search was limited by publication date (1 August 2016 and 1 August 2021) and study language. We included original articles (clinical trials, cohort studies, registry-based cohort studies, case–control studies) with longitudinal echocardiographic analysis of CR and RR during pregnancy and post-partum (quantification of left atrium volume, LV mass, LV end-diastolic or end-systolic volumes) in humans. We excluded case reports or case series, abstract proceedings and original studies with cardiac assessments other than transthoracic echocardiography.

A systematic search was performed on the MEDLINE and Web of Sciences Core Collection databases. A manual search was done on reference lists of the included articles. The search strategy included six search components: pregnancy, echocardiography, publication date and article language, excluding offspring terms and specific publication types. Each component combined MeSH terms (controlled language) and free-text keywords selected by the study authors. The complete search strategy is defined in Table S1.

One author (AFF) screened the titles and abstracts of all citations identified by the searches using an online reviewing software (Rayyan Systems Inc., USA). Potentially eligible studies were assessed for inclusion criteria. Reference lists from included articles were also reviewed.

Data extracted from the selected papers included: author(s) name; year of publication; country; study design; study groups, assessment time points (both during pregnancy or post-partum); recruitment period; total number of participants; number of participants in each study group (if applicable); indexation method; cardiac chambers indexed (LV mass and volumes, and left atrium volume); statistical methods for between-group comparison and for within-group/longitudinal comparison; discussion or comment about indexation methods in the article.

Results are presented using descriptive statistics.

A convenience sample of pregnant women who had their first medical appointment at the Obstetrics Department of Centro Hospitalar Universitário de São João, Porto, from March 2019 to April 2021 were invited to participate. Pregnant women aged <18 years, with pre-existing cardiomyopathy, renal disease, chronic obstructive airway disease, active systemic infection, genetic syndromes, type-1 diabetes mellitus, obesity, arterial hypertension or gestational hypertension were excluded.

Women were evaluated at the following time points: (1) first trimester (10–15 weeks, baseline conditions, the initial phase of cardiac remodeling); (2) third trimester (30–35 weeks, peak of cardiac remodeling, when cardiac adaptations are expected to be most prominent); (3) 6–8 weeks after delivery; and (4) 6–7 months after delivery (cardiovascular RR stages to assess cardiac recovery). The evaluation included clinical characterization through questionnaires, clinical examination and cardiac assessment by transthoracic echocardiography.

All study participants provided written informed consent. The Ethics Committee of Centro Hospitalar Universitário São João approved this study in November 2018 (ID 201/18). Confidentiality and data anonymity were complied with the guidelines emanating from the Declaration of Helsinki of 1964, revised in Fortaleza, in 2013.

Clinical evaluation included measurement of blood pressure (after 5-minutes rest, sitting position), height and weight. Body mass index (weight (kg)/height (m)2) and body surface area [Du Bois formula: 0.007184×height (cm)0.725×weight (kg)0.425] were calculated.6

Maternal cardiovascular health, obstetric and perinatal outcomes, maternal health-related habits, mother's smoking habits, drinking and medical history were systematically collected. For the present pregnancy, obstetric outcomes (especially, gestational diabetes, type of delivery and gestational age at delivery) were also registered.

At each evaluation moment, a single operator performed conventional transthoracic echocardiography (TTE) with a 3 MHz phased-array probe (ACUSON SC2000 PRIMETM). All images, measurements and variables were obtained from standard views according to the recommendations of the European Society of Cardiology for chamber quantification and diastolic function evaluation and at least three cardiac cycle images were collected for data analysis.7,8 TTE was analyzed and validated by an independent cardiologist and discrepancies were settled by discussion and consensus agreement. Structural parameters in mm (interventricular septum in diastole – IVSd and posterior wall thickness in diastole – PWd, left ventricular internal diameter in diastole – LVEDd and systole – LVESd) and their derived parameters (LV mass and relative wall thickness) were obtained from a 2-dimensional parasternal long-axis view. LV mass was estimated using the following equation 0.8×(1.04×[(LVEDd+PWd+IVSd)3−(LVEDd)3]+0.6).

Left ventricular mass (LVM) was presented according to each previously stated methods: (1) absolute values, without indexing; (2) indexing to the basal (previous pregnancy) BSA; (3) using allometric indexes; or 4) indexing to BSA measured at the same day of cardiac assessment.

Continuous variables were reported as median, minimum and maximum, and categorical as absolute values and relative frequencies.

To evaluate cardiac structural changes throughout pregnancy and after delivery, we used the Friedman test (F-statistics, for within-subject assessment) and Kruskal–Wallis test (H-statistics, for between-subject assessment), computed for LV mass, using the four indexation methods. Non parametric tests were used because the LV mass values in each time-point period did not reveal a normal distribution (assessed through visual analysis of histograms in tandem with Q–Q plots’ examination). The Friedman test effect size (Kendall's W value) and the Kruskal–Wallis effect size (η2) will be estimated as a standardized metric, regardless of the scale or method used to measure the outcome variable – cardiac chambers dimensions – in (1) absolute value or (2) basal BSA indexed (LVM/basal BSA, using Du Bois formula); or (3) using allometric indexes; or (4) BSA indexed according to TTE moment.

For the third method, allometric indexes were estimated using a constant denominator: height1.7 or height2.7, that have been proposed as an advantage method over indexing to BSA in both normal subjects and obese people, by avoiding within variability.9,10 We sought to discover whether pregnant population could also benefit from this method and used the two most agreed upon methods in the literature. Mathematically, indexation methods (1), (2) and (3) were expected to produce a similar effect size since there is a common denominator for each one: (1) equals to one (e.g., absolute values); (2) equals to basal BSA (constant across evaluation follow-up); (3) equals to height1.7 or height2.7 (constant across evaluation follow-up).

The level of significance was 0.05.

Twenty-seven eligible studies were included in this systematic review (Figure 1), whose characteristics are shown in Tables 1 and 2. All studies had an observational design, with three case control studies,11–13 six retrospective longitudinal cohort studies14–19 and 18 prospective longitudinal cohort studies.20–36 Four studies did not include echocardiographic assessment during the post-partum period.11,15,27,36 However, in these articles, the investigators performed at least two echocardiographic evaluations during pregnancy. Although all longitudinal studies repeated cardiac assessments during the follow-up period, four articles did not analyse the progression of cardiac remodeling and/or reverse remodeling during pregnancy and after delivery, respectively.12,14–16 Regarding echocardiographic variables, namely LA volume, LV volumes or mass, six studies used absolute values without indexing them to BSA.13,17,18,21,24,35 LV mass was the most indexed variable to BSA, followed by LA volume. Golinska-Grzybala et al. was the only study to highlight the importance of BSA indexation in analyzing the echocardiographic changes in subjects with distinct body shapes and sizes.

Figure 1.

Flow diagram of the search for eligible studies cardiac remodeling and reverse remodeling during pregnancy and postpartum.

(0.53MB).

Four studies used a non-pregnant control group.11,20,22,32 Ten articles did not include study group comparisons (eight used consecutive pregnant women recruitment,21,23,26,28–30,33,37 one focused on pregnant women with pulmonary hypertension19 and one included only women with hypertensive pregnancy disorders).35 Related to the statistical analysis used for longitudinal evaluation of CR and RR, the selection of tests (such as repeated measures ANOVA, linear mixed models, Wilcoxon and paired t-test) from within-subjects design was reported in 17 articles13,18,21–29,31–33,35–37 and not specified in three manuscripts (reporting only used of Student's t-test17,19,32). In addition, between-subject design was reported in only three studies to assess the cardiac adaptations during the follow-up period.11,20,30

We enrolled 45 pregnant women with a median age of 33 [25,41] years, of whom 40% were multiparous. Only three participants developed gestational diabetes. Regarding smoking habits, five pregnant women maintained smoking during gestation, three stopped smoking at the beginning of pregnancy, and ten were former smokers. The variation of body mass index and BSA during pregnancy and postpartum are represented in Table 3. Data were analyzed considering four indexation methodologies and using a within and a between design.

We focused on the longitudinal variation of LVM to analyse the CR and RR during pregnancy and postpartum, respectively. Figures 2 and 3 show the variation of LVM, using Friedman and Kruskal–Wallis tests, respectively. In these plots, each point represents one pregnant woman, and a dashed connective line depicts the change between evaluation times.

Figure 2.

Longitudinal analysis of LVM variation of LVM using Friedman test.

(0.91MB).

Figure 3.

Longitudinal analysis of LVM variation of LVM using Kruskal–Wallis test.

(0.61MB).

Related to LVM progression, a significant increase was found from 1st to 3rd trimesters followed by a significant decrease from the 3rd trimester to 6 months postpartum regardless of the indexation method used (Figures 2 and 3). However, when LVM was indexed to BSA at each time-point, the LVM regression from the 3rd trimester to 1st month after delivery lost statistical significance, contrasting with results from other indexation methodologies (Figures 2 and 3).

The effect size can be used as a standardized metric to compare the magnitude of reported effects for LVM indexation (LVMi), using four different normalization methods (1) absolute values, without indexing; (2) indexing to pre-pregnancy BSA; (3) allometric indexing or (4) indexing to BSA measured at the same day of cardiac assessment).

Regarding the within-subjects design, the effect size of the variation of LVM analysis was similar when indexed to pre-pregnancy BSA, without indexation or using allometric indexation (Table 4). However, the effect size was smaller when indexed to BSA computed at each time-point (Table 4). The comparisons made by the within-subjects design were all statistically significant (p=<0.05).

Concerning the between-subjects design, the effect size of the LVM comparison among four time points was similar using pre-pregnancy BSA indexation, allometric indexes or without any indexation method (Table 4). The effect size reported in LVM indexed to BSA calculated for each time-point of cardiac assessment was lower (Table 4). The comparisons made by the between-subjects design were all statistically significant (p=<0.05).

Focusing on comparisons between within and between-subjects designs, the first one displayed higher effect size values in all indexation methods (Table 4). Additionally, we found that the effect size calculated from LVM progression indexed to BSA in each evaluation through within-subject design was similar to the effect size quantified in the between-subjects design when indexed to pre-pregnancy BSA, without indexation or using allometric indexes.