Fetal cardiac contractility can be objectively assessed by measurement of the cardiac output using spectral Doppler analysis and shortening fraction of ventricles. Each of these techniques has its own limitations. Measurement of cardiac output by spectral Doppler is rarely used in clinical practice because it is technically difficult, time-consuming, and needs high expertise. Assessment of shortening fraction of ventricles is much simpler either using two-dimensional ultrasonography or M-mode for dimension shortening. Many authors^{1–7} propose an assessment of shortening fraction of ventricular dimension using M-mode (transverse diameter at end-diastole−diameter at end-systole/diameter in end-diastole) and construct nomograms for ventricular shortening fraction. However, measurement using M-mode is associated with some unique technical difficulties in exact placement of the M-mode cursor perpendicular to the interventricular septum and variation of ventricular diameters in different views through the ventricles. The difficulty in obtaining a proper plane before M-mode display and measurement may explain the different normative values among various studies.^{1–7} For example, the left ventricular shortening fraction is 0.26, 0.33, and 0.48 during similar gestational age in the studies reported by Koyanagi et al,^{4} De Vore et al,^{2} and Hsieh et al,^{3} respectively. In addition, some studies found no change of ventricular shortening fraction throughout pregnancy, whereas some showed a decrease^{3} or an increase^{8} with gestational age. Moreover, the nomograms of such studies were based on a small sample size, including only 13 to 104 fetuses. It seems that such difference is likely the result of nonperfect planes of measurements because it is difficult to get proper planes with two-dimensional ultrasonography attributable to the fact that the exact reliable plane needs to be controlled in all three dimensions of the heart. It is therefore very difficult to be sure that the interventricular septum is exactly horizontal in all axes. To be more accurate, therefore, we created our own reference ranges for shortening fraction using innovative cardiospatiotemporal image correlation with M-mode technology with a large sample size to use as a standard in evaluating shortening fraction of fetuses with hydrops fetalis secondary to fetal anemia and congenital heart defects or congestive heart failure.^{9} Shortening fraction may be a useful tool in differentiating causes of hydrops fetalis. The purpose of this study was to estimate fetal ventricular shortening fraction representing cardiac contractility derived from cardiospatiotemporal image correlation with M-mode in fetuses with hydrops fetalis secondary to high-output (fetal anemia) and low-output causes (congenital heart defect).

#### MATERIALS AND METHODS

This cross-sectional study was conducted in the Department of Obstetrics and Gynecology, Faculty of Medicine, Chiang Mai University, Thailand, between September 1, 2007, and March 31, 2010. This research was approved by the research ethics committee of our institute. The study consisted of two parts simultaneously performed. Part I was a construction of normal reference ranges of shortening fraction as a standard tool for subsequent assessment of shortening fraction in pathologic fetuses in part II. In part I, normal pregnant women attending the antenatal care clinic were invited to the study only once for each pregnancy. The inclusion criteria were: 1) gestational age between 14 and 40 weeks based on regular menstruation with accurate date of the last menstrual period and ultrasonography in the first half of pregnancies; and 2) low-risk pregnancies without known obstetric and medical complications. Exclusion criteria were: 1) multifetal pregnancy; 2) fetal anomalies or chromosome abnormalities; 3) abnormal intrauterine growth (estimated fetal weight of less than 10th percentile or more than 90th percentile); and 4) cardiospatiotemporal image correlation with M-mode display volume data sets could not be satisfactorily obtained.

In part II, all fetuses with hemoglobin Bart's disease, usually diagnosed at 18–22 weeks by fetal hemoglobin typing either with or without hydropic changes and hydropic fetuses caused by cardiac defects were recruited. The prenatal diagnosis of hydrops fetalis was defined by demonstration of fluid collections in at least two fetal body compartments such as ascites, pleural effusion, pericardial effusion, or skin edema. These fetuses underwent detailed ultrasonography to identify associated abnormalities. Hydrops fetalis secondary to any causes other than hemoglobin Bart's disease and congenital heart defects were excluded. Unsatisfactory or poor-quality cardiospatiotemporal image correlation with M-mode display volume data sets were also excluded.

All ultrasonographic examinations were performed by the four authors, who are experts in fetal echocardiography. The examination technique was standardized through hands-on training and a conference before initiation of the study.

All fetuses underwent ultrasonographic examination, including fetal biometry and anatomic survey and cardiospatiotemporal image correlation using real-time machines Voluson E8 equipped with three-dimensional transabdominal, 2- to 4-MHz curvilinear transducers. The acquired volume data sets were stored for subsequent offline analysis with cardio-spatio-temporal image correlation with M-mode display (4D Views 9). To measure ventricular shortening fraction, the cardio-spatio-temporal image correlation with M-mode display examination was performed as follows: 1) volume acquisition: volume data sets were obtained as a standard technique acquired with the cardiac axis oriented horizontally, ultrasonographic beam perpendicular to the interventricular septum. The virtual cardiac volumes were stored in a hard disk for subsequent offline analysis; 2) offline volume data set processing: the stored volume data sets were adjusted for brightness, contrast, and speed to optimize the images for proper analysis. On multiplanar view, the best image of the four-chamber view was identified and displayed on panel A and then was maneuvered by moving the reference dot and rotating along the three orthogonal axes (x, y, and z) to get the interventricular septum to be in exactly the horizontal plane in the three panels. The proper multiplanar views for analysis are as follows: the interventricular septum in panel A must be in the same line with that in panel B (exactly horizontal) and total en face view of interventricular septum was displayed in panel C (Fig. 1, upper panel); 3) cardiospatiotemporal image correlation with M-mode display: when panel A was selected, the cardiospatiotemporal image correlation with M-mode line was placed perpendicular to the interventricular septum at the level of maximal ventricular dimension, usually just below level of the atrioventricular valves and the cardio-spatio-temporal image correlation with M-mode tracing was displayed (Fig. 1, lower panel); and 4) ventricular shortening fraction measurement: on cardiospatiotemporal image correlation with M-mode panels, the measurements were made at the end-diastole and the end-systole by placing the M-mode cursor at the level of the greatest dimension. The dimensions were measured included: left ventricular internal dimension, which was measured from the endocardium of the left ventricular wall to the endocardium of the left side of the interventricular septum; and right ventricular internal dimension, which was measured from the endocardium of the right ventricular wall to the endocardium of the right side of the interventricular septum. These parameters were measured in the same tracings at both end-diastole and end-systole. The best measurements were selected and recorded for analysis. Left ventricle shortening fraction and right ventricle shortening fraction were calculated using equation: shortening fraction=dimension at end-diastole−dimension at end-systole/dimension at end-diastole.

In part I, regression analysis of right shortening fraction and left shortening fraction was performed to determine the best-fit equations as follows: gestational week, as an independent variable, and ventricular shortening fraction as dependent variables were used to generate the best-fit regression equations. Various built-in models, including linear, quadratic, cubic, exponential, and logarithmic models, were tested and compared. The simpler model was chosen if the more complex model did not result in significant improvement. The residuals were tested for normality of distribution using Kolmogorov-Smirnov and Shapiro-Wilk tests. If the distribution was not normal, the data were transformed and the new residuals were checked to determine whether they conformed to a normal distribution. Data were also examined to determine whether the standard deviation of the residuals varied across the range of values for the independent variable using Levine's test. If a significant degree of heteroscedasticity was found, weighted regression of absolute residuals was used to adjust the standard deviation. To construct normal reference range including +1, 2, and 3 standard deviation curve with restricting the standard error of the limits of reference range to 10% of the standard deviation, a sample size of approximately 550 fetuses (14–40 weeks)^{10} was needed. The results in part II were related to the reference range data of left ventricle shortening fraction and right ventricle shortening fraction. The data were also compared between the group of hydrops fetalis resulting from hemoglobin Bart's disease and the group of that caused by congenital heart defects. Statistical analysis was performed using SPSS 17 for Windows and GraphPad Prism 5. Analysis of variance and Kruskal-Wallis nonparametric test were used to compare the mean and median of shortening fraction as well as Z-score as appropriate. Dunn's test was used for nonparametric multiple pairwise comparisons. *P* value of <.05 was considered statistically significant.

#### RESULTS

In part I, in construction of normal reference ranges of shortening fraction, a total of 606 measurements from 606 fetuses (group 1) were performed for left ventricle shortening fraction and right ventricle shortening fraction (Table 1). The mean number of measurements per week was 22.4 ranging from 15 to 35 measurements. The mean left ventricle shortening fraction and right ventricle shortening fraction was 0.32±0.52 and 0.31±0.49, respectively. Both parameters slightly but significantly decreased with increasing gestational age. The weighted regression analysis, using weighted least square technique, reveals a fitted linear equation as follows:

Predicted mean left ventricle shortening fraction=0.397−0.003GA(week) (r^{2}=0.141, *P*<.001) and

Predicted mean right ventricle shortening fraction=0.355−0.002GA(week) (r^{2}=0.055, *P*<.001).

Predicted standard deviation (SD) was constant throughout gestation; predicted SD for left ventricle shortening fraction and right ventricle shortening fraction were 0.01989 and 0.011444, respectively.

Therefore, Z-scores for left ventricle shortening fraction or right ventricle shortening fraction for each measurement can be simply calculated using the equation: Z score=(measurement−predicted mean)/predicted SD.

In part II, 63 fetuses with hemoglobin Bart's disease without hydrops fetalis (group 2), 48 fetuses with hemoglobin Bart's disease with hydrops fetalis (group 3; high-output hydrops fetalis), and 21 hydropic fetuses with congenital heart diseases (group 4; low-output hydrops fetalis) were successfully recorded for cardiospatiotemporal image correlation. The groups and the various causes of congenital heart disease in group 4 are presented Table 2. The mean gestational ages for groups 2, 3, and 4 were 20.3±5.1 weeks, 24.7±5.3 weeks, and 25.7±3.7 weeks, respectively.

All fetuses with hemoglobin Bart's disease either with or without hydrops fetalis had cardiomegaly but nearly all of them had normal shortening fraction. Because shortening fraction values significantly, although slightly, changed with gestational age, to compare left ventricle shortening fraction and right ventricle shortening fraction among the four groups, statistical control of gestational age was performed. Among the three groups (2, 3, and 4), analysis of covariance tests showed no significant effect of gestational age on left shortening fraction and right shortening fraction (*P*>.05). Post hoc test revealed that group 2 and group 3 were not significantly different (*P*>.05). On the contrary, left ventricle shortening fraction and right ventricle shortening fraction in hydropic fetuses with congenital heart defects were markedly decreased when compared with all other groups (*P*<.001) (Table 3; Fig. 2).

Table 3 Image Tools |
Fig. 2 Image Tools |

Likewise, Z-scores tended to be decreased in group 2 and in group 3 and even much more decreased in group 4, when compared with normal fetuses as shown in Table 4. Left ventricle shortening fraction and right ventricle shortening fraction of fetuses with hemoglobin Bart's disease were approximately one SD below the mean, whereas those of fetuses with hydrops fetalis secondary to congenital heart defects were 5 SDs and 8 SDs below the mean, respectively. Medians and range of Z-scores for left shortening fraction and right shortening fraction were −0.4703 (−9.28 to 6.41) and −0.4817 (−16.22 to 11.72), respectively. Kruskal-Wallis test showed statistical significance between groups (*P*<.001). Dunn's posttests comparing the differences between each pair group showed significant differences between group 4 compared with group 1 (*P*<.001), group 4 compared with group 2 (*P*<.001), and group 4 compared with group 3 (*P*<.001), whereas there was no significance between the others (*P*>.05).

#### DISCUSSION

Assessment of fetal cardiac contractility with M-mode was introduced by Wladimiroff et al^{6} as a simple objective quantification obtained by measuring the shortening fraction. However, standardization of measurement of the end-diastolic and the end-systolic diameter is difficult as a result of problems with two-dimensional M-mode alignment. As a result, interobserver limits of agreement range approximately ±15%.^{11,12} Nevertheless, the shortening fraction remains a widely used parameter of cardiac contractility. This may be helpful in evaluating contractility in abnormalities that may affect wall motion such as cardiomyopathies or congestive heart failure. With innovative three-dimensional cardiospatiotemporal image correlation M-mode, the problems of alignment and improper measurement have been resolved. This technique is simple, less time-consuming, and highly reliable in obtaining the proper orientation. Unlike measurement with two-dimensional ultrasonography in which we have to orientate a transducer to find the proper plane, with three-dimensional spatiotemporal image correlation offline analysis we can control and orientate the virtual heart to get the proper plane, not orientate the transducer. The interventricular septum can simply be maneuvered in the exact horizontal line in panels A and B and total en face view of the interventricular septum can be displayed in panel C, resulting in more reliable measurements.

Several fetal conditions such as congenital heart defects, anemia, or infection may lead to myocardial dysfunction and eventually cardiac decompensation resulting in hydrops fetalis. Therefore, evaluation of shortening fraction may be helpful in determining the degree of cardiac decompensation or differentiating causes of fetal hydrops. The extent of shortening fraction changes with gestational age may be small but significant. Therefore, to compare shortening fraction among fetuses with cardiac pathology, gestational age must be taken into account. Accordingly, we controlled for gestational age and used Z-scores to compare shortening fraction among fetuses at different gestational ages. Our results showed the different patterns of shortening fraction compromise between the group of hydrops fetalis associated with low cardiac output (cardiac defects) and the group with high cardiac output (fetal anemia). Low-output hydropic fetuses demonstrated impairment in contractility, whereas high-output hydropic fetuses had very minimally impaired cardiac contractility despite hyperdynamic circulation or hypervolemia, indicating that hydrops fetalis had developed before cardiac decompensation. This finding is contrary to a long-believed concept that hydrops fetalis secondary to anemia is a consequence of congestive heart failure.

Surprisingly, our study suggests that fetal anemia or hypervolemia may not be associated with significantly low shortening fraction. Most of these fetuses had normal shortening fraction, although there was a trend for low shortening fraction suggesting that hydrops fetalis is not primarily caused by cardiac decompensation. Although long-lasting hypervolemia can lead to marked cardiomegaly, and finally to cardiac failure,^{13} cardiomegaly reflects rather high competency of cardiac compensation than cardiac failure. Despite hypervolemia, normal shortening fraction implies effective ventricular contractility or good adaptive response. Hypervolemia may be the result of a low blood viscosity resulting from anemia and effective cardiac contractility. On the contrary, in fetuses with cardiac defects and hydrops fetalis, low shortening fraction indicates cardiac decompensation resulting from the cardiac pathology, leading to hydrops fetalis, whereas hydrops fetalis resulting from anemia is not caused by cardiac decompensation. We postulate that during the early stage of anemia, hydropic change can occur; the fetus effectively compensates by increasing the cardiac dimension and effective contractility to maintain peripheral perfusion. Congestive heart failure does not develop in most fetuses with high-output hydrops fetalis if they are not in the end stage. This study implies that the normal fetal heart is capable of compensation to cope with anemic hypoxia. This is consistent with our observation that several fetuses with hemoglobin Bart's disease can show reactive nonstress test despite marked cardiomegaly without other signs of contractility compromise. Our findings indicate that in fetal anemia, local and systemic compensatory mechanisms are highly effective and may help the fetus to survive. On the other hand, some of these mechanisms at the same time increase the imbalance of Starling forces, resulting in enhanced interstitial fluid accumulation before the compensatory mechanisms become exhausted.

Although anemia in this study was uniformly lethal and the shortening fraction measurements are no of value, it could be a model for nonlethal causes of anemia such as Rh erythroblastosis or parvovirus B19. Although fetuses with hydrops fetalis resulting from anemia may have poor prognoses overall, a better prognosis can still be expected if the shortening fraction is normal. In such cases, shortening fraction assessment may be used to differentiate hydropic fetuses with congestive heart failure (poor contractility) from ones without heart failure. Thus, shortening fraction assessment may be helpful in predicting the prognosis and in making a choice of management. However, this study is preliminary and further study of the clinical application is needed.

The limitation of this study is that resolution of the volume data set in several fetuses may be somewhat lower when compared with conventional two-dimensional ultrasonographic resolution. Moreover, the problems of unsatisfactory acquisition can be encountered, especially as a result of fetal movement or maternal breathing. It is impossible to obtain good acquisition in many fetuses with very active movement for a long period or in fetuses in poor position. Additionally, even with a high-resolution machine, the small cardiac structures in the early second trimester could not be clearly demonstrated in several cases, leading to exclusion. Finally, because the examiners were aware of the underlying causes of the fetal hydrops, this could possibly lead to a potential source of bias in examination.

The strengths of this study include 1) the data were derived from a reliable technique, three-dimensional cardiospatiotemporal image correlation with M-mode display with offline analysis without time limit, enabling us to obtain the best image for measurement in all high-quality volume data sets; and 2) the large sample size per week of gestation to construct the reference ranges, providing a reliable tool for evaluation of shortening fraction in the pathologic groups.

In conclusion, fetuses with hydrops fetalis secondary to cardiac causes and fetal anemia have a different pattern of shortening fraction. This study provides data that myocardial dysfunction and cardiac decompensation are a primary cause of hydrops fetalis in fetuses with congenital heart defects, whereas hypervolemia may be a primary cause in fetuses with anemia with cardiac failure or decompensation occurring when the cardiac compensatory mechanism is exhausted.