Over all variability of mitral annular plane peak systolic velocity and peak global longitudinal strain rate in relation to age, body size, and sex: The HUNT Study

Left ventricular (LV) systolic global function can be assessed by peak annular systolic velocity S′. Global longitudinal strain rate (GLSR) is relative LV shortening rate, equivalent to normalizing S′ for LV length (S′n). It has previously been shown that mitral annular plane systolic excursion (MAPSE) and global longitudinal strain (GLS) have similar biological variability, but GLS normalizes for one dimension only, inducing a systematic error, increasing body size dependence. The objective of this study was to compare S′ with GLSR in the same way, comparing biological variability and body size dependence.

EF 7 and has prognostic predictive value in hypertension. 8 By tissue Doppler, both systolic and diastolic function can be measured by the same method, showing the interdependence between systolic and diastolic function. 2,3,9 Our group has previously published age and sex-specific normal values for S′ from the HUNT Study, 9 showing higher values in men, and lower velocities with increasing age. S′ shows differences between walls in healthy individuals, so for a global measure, measurements of S′ from different walls have to be averaged [9][10][11] as shown in Figure 1A.
Systolic strain rate is velocity difference per length unit 12 and has been validated as a measure of segmental systolic dysfunction [13][14][15] (Figure 1B). Peak global longitudinal strain rate (GLSR), being the mean of all segments is a global measure of LV function, the rate of LV shortening normalized for LV length. Normalizing annular velocity (S′ n ) for wall length (WL) is also a measure of global strain rate as shown in Figure 2. Thus, both GLSR and S′ n are measures of shortening rate, normalized for LV length. Normalized measures were supposed to compensate for differences due to heart, and thus, body size, reducing biological variability. However, recent data from the Nord-Trøndelag Health Study (HUNT) have shown that for MAPSE and global longitudinal strain (GLS) normalizing for LV length did not reduce biological variability, in fact normalizing for LV length induced a systematic error, with subsequently increased dependence on body size and sex. 16 The basic measures of S′, wall lengths and GLSR in this material, have all been published previously, 9,17,18 but the normalized values of S′ n , and the comparisons of variability and relation to body size are new. The aim of the present study was to ascertain the biological variability of S′ vs the normalized values S′ n and GLSR in terms of the relations to age, body size and sex, and to see if the relations were similar to those of MAPSE and GLS.

| Study subjects
The study population was recruited from 50 839 participants of the third wave of the Nord-Trøndelag Health Study (HUNT3) and has been extensively described in previous papers. 9,10,16,17,19 Briefly, a random sample restricted to two communities was invited to the echocardiographic study, excluding subjects with a history of heart disease, hypertension, or diabetes. A total of 1296 were included.
After echocardiography, another 30 individuals with significant pathology on the echocardiogram were excluded. Thus, the study consisted of 1266 subjects aged 19-89 years. All subjects in the HUNT Study gave their written consent to participate in both main and substudies. The study was approved by the regional ethical committee (REK 4.2009.397 and 2018/929). Basic characteristics are given in Table 1. Echocardiographic measures were all normally distributed.
Blood pressure was measured during the visit, as the mean of the two last of three automated measures. Despite excluding subjects with known hypertension, it is evident from Table 1 that some untreated hypertension may be present in the material, although this was spot blood pressures on a single day. In the three age groups, ˂40, 40-60, and >60 years, respectively, 7%, 18%, and 44% had Systolic blood pressure (SBP) >140 mm Hg, and 3%, 9%, and 10% had diastolic blood pressure (DBP) >90 mm Hg. Both (SBP) and (DBP) correlated with age: R = .40 and .24, respectively (both P < .001).

| Echocardiography
One experienced echocardiographer (HD) conducted all the examinations. Subjects were examined in left lateral supine position with a Vivid 7 scanner (version BT06, GE Vingmed Ultrasound).
The examination included apical four-and two-chamber and apical long-axis views. Mean B-mode frame rate was 44 FPS. Pulsed-wave tissue Doppler recordings of mitral annular velocities were acquired from the base of the septum and lateral wall in four-chamber, and anterior and inferior walls in the two-chamber views, S′ was measured by spectral tissue Doppler and averaged per patient for global measure in accordance with general usage. [1][2][3]6,9 Wall lengths (WL) were approximated by the straight line from the apex to the mitral ring at end-diastole in B-mode, as shown in Figure 1B. S′ n was calculated as S′/WL and averaged for all four walls per patient for global value.
Color tissue Doppler images from the three standard apical planes were acquired separately with mean Doppler frame rate of 100 FPS, and B-mode images in the background. Peak systolic strain rate was measured by the combined segmental tissue Doppler and speckle tracking method described earlier 17,20 as shown in Figure 1B. Care was taken to avoid visible clutter areas. Global average (GLSR) was calculated of all six walls (16 segment model) from the three standard planes. 21 We also calculated the mean from the four walls of the four-chamber and two-chamber views, for comparison with S′ and S′ n . Longitudinal systolic strain rate is shortening, being negative, but the main objective was to compare with S′, so strain rate values are here referred as numerical values.

| Calculation and statistics
Calculations and statistics were done in SPSS (IBM, corp).
Echocardiographic indices are presented as means and standard deviations (SD). Strain is given in numeric values. Differences between genders were tested by independent samples Student's t-test, differences between age groups by one-way ANOVA, with Bonferroni post hoc comparisons. As strain rate and velocity have different units, the relative SD (SD/mean) was used as variability measurement. Correlations were assessed by Pearson's correlation coefficient. Linear regression was used for assessing the interaction of BSA, age, and gender.
Repeatability of the different measures has been extensively studied previously. 9,10,16-18 Shortly inter-observer variations in repeated acquisitions had a coefficient of repetition (CoR) of 1.7 cm/s and mean error (ME) of 8% for S′ averaged from four walls which increased to 11% when using mean of only two walls. The corresponding CoR and ME was 0.2 s −1 and 9% for GLSR and 1.6 mm and 4% for WL, respectively.
However, as segmental strain is susceptible to clutter, almost 40% of segments were rejected in order to obtain representative normal segmental values for SR and strain. 16 Feasibility per wall was 96%-97% for all four walls for S′, 93% for S′ n and 58%-90% for SR, lowest in anterior wall.
Distribution of measures are shown in Figure 3, panel A. GLSR was normally distributed, while S′ and S′ n showed significant, but modest skewness (near normal) of −0.24 and 0.35, respectively. As seen by Table 2, the relative SD was not very different between S′ and the two normalized measures S′ n and GLSR.
Sex-and age-related values are given in Table 2. For direct comparison, numerical values are given and discussed, even though the correct usage for GLSR and S n should be negative systolic values as explained in Figure 2.
As pulsed-wave tissue Doppler conventionally is taken in four points from four-and two-chamber views, values for the mean of all four walls as well as for only septal and lateral walls (being a time-saving practice in everyday clinic) are given as secondary measures. Differences between two and four walls for S′ and S′ n were both significant, although differences are negligible in practice. GLSR was measured in all six walls, but for comparison, mean of the same four and two walls are also given. Again, differences were statistically significant, but in practice negligible. GLSR, however, was significantly higher than S′ n .
S′ correlated with MAPSE (R′ = 0.55). There was a weak positive univariate correlation of BSA with S′ of 0.13, while S n and GLSR showed numerically slightly higher, but negative correlations with BSA as seen in Table 3. The relation between the three measures and BSA is shown in Figure 3, panel B. BSA was also significantly different between sexes (Table 1, P < .001). In linear regression with sex, age, and BSA, BSA was independently associated with S′ n , and sex was independently associated with S′ and GLSR as seen from Table 3.

F I G U R E 1
Measurements of peak mitral annular velocities, left ventricular wall lengths, and systolic longitudinal strain rate. A, Peak systolic mitral annular plane velocity (S′) by pulsed-wave spectral tissue Doppler curves from the septal and lateral points. Diastolic measures are obtained by the same method; e′ is peak early diastolic velocity (relaxation), a′ is peak late diastolic velocity (atrial contraction). It can be easily and robustly measured in the mitral annular points. It is evident that the lateral S′ is higher than the septal, so for global S′, at least two points have to be averaged. B: Wall lengths (WL) were approximated by the straight line from the apex to the mitral ring at end-diastole (green straight lines). Normalized S′ n was measured by S′/WL per wall. WL overestimates the LV diastolic length (Ld-white line), while underestimating the true wall length (green broken line). Segmental strain rate was measured with a combined tissue Doppler-speckle tracking method tracking kernels at the segmental borders (magenta squares), in the longitudinal direction by tissue Doppler, and in the lateral direction by speckle tracking. Strain rate was calculated as the temporal derivative of strain, converted from Lagrangian to Eulerian strain rate Differences between age groups were highly significant (post hoc P < .001 overall and for difference between all age groups) as seen in Table 2. Relations with age are shown in Figure 3, panel C. There was almost no correlation between age and BSA (R = −.06, P = .04), not significant if multiple correlations were taken into account. In line with this, there was little difference between univariate and multivariate correlations of the echocardiographic measures and age as seen from Table 3.
Age correlated with BP, in linear regression against age and blood pressure, age had the strongest association with S′, S′ n , and GLSR with β of −0.39, −0.245, and −0.25, respectively (all P < .001), SBP was not significant, while DBP was also associated with all three (β of −0.1, −0.19, and −0.12, all P < .005).

| D ISCUSS I ON
The main findings in this study were as follows: • Normalization of mitral annular velocities for LV length, or the use of GLSR, do not reduce overall variability compared with S′.
where v(x) and v(x+∆x) are velocities in two different points, and ∆x is the distance between the points. As the apex is nearly stationary, v in apex is near zero. Peak systolic annular velocity (S′) is the summed velocity of a whole wall. Wall length (WL) will then be ∆x, and the strain rate formula will then give peak SR ≈ 0 − S � WL = −S � WL . Panels to the right show top: pulsed-wave spectral Doppler curve from the base of the septum, and bottom, a strain rate curve from most of the length of the septum of the same subject  (Tables 2 and 3 global strain rate, S′; peak mitral annular systolic longitudinal velocity, S′ n ; peak mitral annular systolic longitudinal velocity normalized for wall length, e′; peak mitral annular early diastolic longitudinal velocity, e′; peak mitral annular late diastolic longitudinal velocity, WL; wall length between them remains constant across the BSA range. 18 As the main contribution to the stroke volume is the AV-plane motion, 22 this means that with larger BSA and larger hearts, the main SV increase is related to the cross-sectional LV area, and the square of the radius.
Thus, a larger heart generates a larger SV even without the effect of increased AV-plane motion, as illustrated in Figure 4, meaning that even with a higher SV, there is very little increase in MAPSE with increasing BSA. This is thus also the case for S′, which correlates strongly with MAPSE, and thus changes very little with heart size, despite change in SV. The weak correlation between S′ and BSA, however, is sufficient to give a statistically significant sex difference, although for practical purposes the difference is negligible as seen in Table 2, the mean sex differences are small compared with the prediction interval.
As previously shown for global strain, 16 GLSR and S′ n show a slightly stronger, negative correlation with BSA, due to the systematic error that they are normalized for LV length only as explained in Figure 4. As S′ is nearly unchanged with larger heart size, GLSR and S′ n will decrease by the larger denominator (length), as seen by the numerically higher, but negative correlations with BSA.
The diagnostic discriminatory capability of a parameter is related to both the variability of the parameter and the separation of means between normalcy and disease. As normalization for length does not reduce overall biological variability compared to S′, there seems to be no advantage of using global strain rate as a global LV contractility measure compared to S′. Basically, strain rate is a method to assess regional dysfunction and differences in timing, while regional S′ is not. 4 This is important in diseases with regional uneven function, as coronary artery disease and conduction abnormalities. Here, GLSR may still be important in assessing global function. The diagnostic accuracy in specific patient populations must be assessed in direct comparative studies.
Global longitudinal strain rate and S′, on the other hand, have been shown to have about the same reproducibility, where image quality is good. 10 However, spectral tissue Doppler is much more robust in the presence of clutter. 23

| Limitations
The HUNT study is among the largest normal studies but is limited by the lack of ethnic and geographical differences. This limits the generalizability of the normal values. However, for the main issue of comparing GLSR and S′ with relation to BSA and age, this is less important. As the conventional LV dimensions and FS in this material are in line with other M-mode studies, 25,26 the population seems to be fairly representative.
As this is a cross-sectional study, the age differences are between cohorts and not true aging. However, it reflects the age relations as seen today.
The present study only discusses the variability of the indices within normal ranges. In conditions with reduced regional function, the diagnostic performance of global strain rate may still be better than S′, although this needs to be established in studies by direct comparison.

| CON CLUS ION
Systolic annular velocity shows similar biological variation as global strain rate in normal adults, despite global strain rate are normalized for LV length. Further, the normalization for LV length actually increases the body size dependence. Age is the main determinant of LV shortening. Thus, it is dubious that strain rate adds diagnostic information about global function, compared to mitral annular peak velocity, although strain rate being important for assessing regional function.

ACK N OWLED G M ENTS
The study was fully sponsored by the Norwegian University of Science and Technology, as a PhD grant, as well as the HUNT Study providing the infrastructure for the Echocardiography in HUNT substudy. There was no relation to industry.

CO N FLI C T O F I NTE R E S T
The authors have no conflicts of interest.

Asbjørn Støylen
https://orcid.org/0000-0002-2245-7066 F I G U R E 4 Relation between stroke volume and mitral annular plane systolic excursion (MAPSE) in ventricles of different size. It has been shown that while heart size increases with body size, the ratio between length (LVEDL) and external diameter (LVEDD) does not. As the stroke volume is mainly determined by the systolic shortening (MAPSE), a larger ventricle has a larger radius, and thus, a larger stroke volume (increasing proportional to the square of the radius) even without any differences in MAPSE, as shown by the very low correlations between MAPSE and BSA. Thus, in the ventricle to the right, for the same MAPSE, the SV is far higher. As length increases proportional to the diameter, GLS being MAPSE/ LVEDL, GLS actually decreases with increasing heart size. This is a systematic error that occurs due to the one-dimensional normalization. As there is a strong correlation between S′ and MAPSE, this is also the case for S′ vs GLSR, even if those measures are more closely related to contractility than stroke volume