**2. Results**

Speckle tracking echocardiography (STE) was performed four weeks postpartum in formerly pregnan<sup>t</sup> rats with PE-specific symptoms such as high blood pressure and albuminuria. In comparison, the same method was performed in formerly preeclamptic women 50 weeks after delivery (Figure 1). Both former preeclamptic species (PE) were compared with matched controls after healthy pregnancy (control).

**Figure 1.** Formerly preeclamptic women and rats from a transgenic animal model were characterized postpartum regarding cardiac alterations in function and structure. Early onset preeclamptic women showed a lower gestational age than controls but were matched on scanning time after delivery.

#### *2.1. Speckle Tracking Echocardiography in the Transgenic Animal Model Simulates the Human Situation*

The most important readout in STE is the global strain data. It describes the degree of deformation of the myocardium in different directions. The most sensitive parameter, the global longitudinal strain (GLS), was reduced after PE in both the animal model (control −23.5 ± 1.8% vs. PE −13.8 ± 0.6%) and the human (control −20.6 ± 0.5% vs. PE −15.5 ± 1.6%) postpartum situation (Figure 2A), and it was the same regarding the global longitudinal strain rate (Figure 2B). The global radial strain (Figure 2C) and the corresponding strain rate (Figure 2D) showed only a decreased tendency in the animal post-PE model and were unchanged in the human situation. The global circumferential strain (Figure 2E) and the corresponding strain rate (Figure 2F) were reduced in the animal model but showed no changes after human preeclamptic pregnancy. Moreover, former PE animals demonstrated a clear reduction of the ejection fraction (EF) with control 66.3 ± 2.3% vs. PE 55.5 ± 1.3%. The human data confirm the trend (Figure 2G). The stroke volume (Figure 2H) as well as the cardiac output (Figure 2I) and the end diastolic volume (Figure 2J) were not altered postpartum in any of the species. The end systolic volume only showed a slight increase in the animal model but not in the human postpartum situation (Figure 2K). In addition, echocardiography provided the first initial evidence of morphological changes. The posterior wall was clearly thickened in the animal post-PE model; in the human situation, it showed a borderline *p*-value of 0.06 (Figure 2L). If the relative wall thickness was considered, both species showed a postpartum increase (Figure 2M) with control 0.20 ± 0.01 vs. PE 0.25 ± 0.01 in the animal model and control 0.34 ± 0.01 vs. PE 0.40 ± 0.02 in the human postpartum situation. Interestingly, formerly preeclamptic animals showed an increase in LV masses (control 983.5 ± 19.0 mg vs. PE 1138.0 ± 40.6 mg); however, formerly preeclamptic women did not (Figure 2N). The LV diameter in the end diastole was not significantly changed in either species (Figure 2O). The heart rate was significantly increased in the animal model after preeclamptic pregnancy (control 306.6 ± 6.8 bpm vs. PE 356.2 ± 11.0 bpm). With a *p*-value of 0.08, this trend was also evident in the human situation (Figure 2P).

**Figure 2.** The transgenic rat model simulates cardiac alterations of a human preeclamptic pregnancy. Global longitudinal strain (**A**) and global longitudinal strain rate (**B**) were decreased after preeclampsia (PE). Global radial strain (**C**) and the corresponding strain rate (**D**) were not altered after PE. Global circumferential strain (**E**) and global circumferential strain rate (**F**) were reduced in the post-PE animals but not in the human cohort. Ejection fraction was reduced in animals and showed the same trend in the human PE data (**G**). Stroke volume (**H**), cardiac output (**I**), end-diastolic (**J**), and end-systolic volume (**K**) were not altered in either species after PE. Left ventricle (LV) posterior wall (**L**) and relative wall thickness (**M**) were increased due to PE in both species. LV mass (**N**) was only increased in the post-PE animals. LV end-diastolic diameter (**O**) was unaltered. Heart rate was higher in PE animals and showed increasing trends in humans (**P**). Mean values ± SEM, unpaired students t-test, ns. Non-significant, \* *p* < 0.05, \*\* *p* < 0.01, \*\*\* *p* < 0.001, \*\*\*\* *p* < 0.0001.

Figure 3 summarizes the described measurements in relative changes after a preeclamptic pregnancy compared to the healthy controls by a spider's web plot. Here, the examined parameters are noted in the corners. The controls were normalized to the reference value 1.0, which is shown as a grey line. Changes in the animal post-PE model are reflected by the red line, and the purple line refers to the changes in the human post-PE situation compared to healthy controls. Concludingly, similar deviations after pathological pregnancy in both species can be observed. Several functional parameters, such as EF, were reduced, and parameters representing structural remodeling, such as relative wall thickness, were increased in both species compared to controls.

**Figure 3.** Relative values of post-preeclamptic changes in the animal model and in the human situation. Controls of each species were normalized to one. Grey line = controls, purple line = PE human, red line = PE animal; PE preeclampsia.

#### *2.2. Intraobserver Variability Shows an Excellent Correlation within One Observer*

Analyses of the intraobserver variability in the animal data showed an excellent correlation between two repeated evaluations of EF, r = 0.97, *p* < 0.0001 (Figure 4A), and GLS, r = 0.98, *p* < 0.0001 (Figure 4B) within one observer. The very strong agreemen<sup>t</sup> between repeated evaluations was substantiated in the corresponding Bland–Altman plots, which showed only a marginal bias mean difference (95% CI for limits of agreement): 0.75 (4.50 to −3.00) for EF (Figure 4C) and −0.50 (1.30 to −2.31) for GLS (Figure 4D). In the human data analyses, there was likewise an excellent intraobserver correlation regarding measurements of EF, r = 0.94, *p* < 0.0001 (Figure 4E), and GLS, r = 0.98, *p* < 0.0001 (Figure 4F). Only a minor bias was seen in the corresponding Bland–Altman plots: 0.44 (4.48 to −3.61) for repeated evaluation of human EF (Figure 4G) and −0.11 (1.36 to −1.58) for human GLS (Figure 4H). The intraclass correlation coefficients (ICCs) for reliability of the evaluations were excellent both in animal and in human analyses (95% CI): 0.96 (0.89 to 0.99) for animal EF, 0.97 (0.92 to 0.99) for animal GLS, 0.93 (0.83 to 0.98) for human EF, 0.98 (0.95 to 0.99) for human GLS evaluations.

**Figure 4.** Intraobserver comparison. In analyses of animal data, there was an excellent correlation between the two repeated evaluations of ejection fraction, r = 0.97, *p* < 0.0001 (**A**), and global longitudinal strain, r = 0.98, *p* < 0.0001 (**B**) within one observer. The excellent agreement between the two evaluations was substantiated in the corresponding Bland–Altman plots, which showed only a marginal bias, mean difference (95% CI for limits of agreement) 0.75 (4.50 to −3.00) for ejection fraction (**C**) and −0.50 (1.30 to −2.31) for global longitudinal strain (**D**). In human data analyses, there was likewise an excellent intraobserver correlation regarding measurements of ejection fraction, r = 0.94, *p* < 0.0001 (**E**), and global longitudinal strain, r = 0.98, *p* < 0.0001 (**F**). Only minor bias was seen in the corresponding Bland–Altman plots: 0.44 (4.48 to −3.61) for repeated evaluation of human ejection fraction (**G**) and −0.11 (1.36 to −1.58) for global longitudinal strain (**H**). EF = ejection fraction, GLS = global longitudinal strain.

#### *2.3. Interobserver Comparison Displays Strong Correlation between Two Di*ff*erent Observers*

The interobserver variability comparison showed moderate to strong correlation between the assessments of the two observers for animal EF (Figure 5A) and GLS (Figure 5B) with a moderate bias, as shown in the Bland–Altman plots, mean difference (95% CI for limits of agreement) of 1.56 (−4.96 to 8.09) for animal EF (Figure 5C) and −1.01 (−9.30 to 7.28) for animal GLS (Figure 5D). A moderate to strong correlation was also shown between the assessments of the two observers for human EF (Figure 5E) and GLS (Figure 5F) with a moderate bias, as shown in the Bland–Altman plots, mean difference of −4.50 (−12.27 to 3.27) for human EF (Figure 5G) and −1.08 (−5.79 to 3.64) for human GLS (Figure 5H). The repeatability of the interobserver assessments fluctuated between good ICC (95% CI) for interobserver animal EF assessments: 0.89 (0.69 to 0.96) and 0.63 (−0.04 to 0.88) for human EF, 0.68 (0.31 to 0.88) for animal GLS, and 0.73 (0.39 to 0.90) for human GLS interobserver assessments.

**Figure 5.** Interobserver comparison. In analysis by two experts, variability comparison showed moderate to strong correlation between the assessments for animal EF (**A**) and GLS (**B**) with a moderate bias, as shown in the Bland–Altman plots, mean difference (95% CI for limits of agreement) of 1.56 (−4.96 to 8.09) for animal EF (**C**) and −1.01 (−9.30 to 7.28) for animal GLS (**D**). A moderate to strong correlation was also shown between the assessments of the two observers for human EF (**E**) and GLS (**F**) with a moderate bias, as shown in the Bland–Altman plots, mean difference of −4.50 (−12.27 to 3.27) for human EF (**G**) and −1.08 (−5.79 to 3.64) for human GLS (**H**).
