Study population and design

The study population consisted of 96 healthy volunteers prospectively recruited by advertisement from the local community (76 adults, age 21–81 and 20 children, age 11–15, all caucasian). No subjects were excluded because of poor image quality. All subjects had a normal electrocardiogram (ECG) and blood pressure (systolic blood pressure (SBP) ≤ 140 mmHg and diastolic blood pressure (DBP) ≤ 90 mmHg) [10] and had no history of systolic or diastolic hypertension. Subjects with previous or current cardiovascular, systemic, metabolic disease, body mass index ≥ 30, visually overt aortic or mitral valve regurgitation in long axis MR images, or treatment with medication (except oral contraceptives [n=5], hormone replacement therapy [n=5] or oral incontinence medication [n=1]) were excluded from the study. MR imaging was performed within four weeks after inclusion with image analysis undertaken by independent observers blinded to subject characteristics. The investigation protocol and procedures were approved by the Lund University research ethics committee (reference number LU 207-00). Written informed consent was obtained from all subjects prior to inclusion.

Blood pressure

SBP and DBP were obtained in the supine or seated position by auscultation using a brachial cuff.

MR imaging

All patients were imaged in the supine position using a 1.5 T system (Magnetom Vision; Siemens, Erlangen, Germany) with a 25 mT/m gradient system and a phased-array body coil. Standard scout images were used to locate the orthogonal planes of the heart. End-expiratory ECG triggered short-axis gradient echo cine loops were then acquired throughout the left ventricle from the base (atrioventricular valve plane) to the apex. Typical imaging parameters were TR = 100 ms, and echo sharing gave an effective phase interval of 50 ms, TE = 4.8 ms, slice thickness 10 mm, field of view 350–420 mm, matrix 126 × 256, flip angle 20°. The number of cardiac phases per acquisition was determined as the integer obtained from the RR interval divided by TR. Nine to twelve slices were required to completely cover the left ventricle, depending on heart size.

MR image analysis

i) Tracing of endocardial and epicardial contours

All measurements were undertaken manually without the aid of automated image analysis software. End-diastolic and end-systolic frames were identified according to ventricular blood pool area. Measurements of the left ventricular endocardial and epicardial areas (Scion Image Beta 4.0.2, Scion Imaging Corporation) in each image frame were performed in the short-axis view. At the base of the left ventricle, the aortic outflow tract below the valve was included in volume measurements. The free papillary muscles were included for LVM assessment, and excluded for left ventricular volume assessment [11]. In the basal region of the heart where the left atrium was seen, only the portion of the slice that could be identified as left ventricle was included for measurement (Figure 1).

ii) Left ventricular mass and dimensions

The difference in area between the endocardial and epicardial contours multiplied by the slice thickness (10 mm) represented myocardial volume for a given slice. Total myocardial mass was obtained by calculating the sum of all myocardial slice volumes and multiplying by the myocardial specific gravity (1.05 g/cm³). End-diastolic volume (EDV) and end-systolic volume (ESV) were calculated as the endocardial volume at end-diastole and end-systole respectively. Stroke volume (SV) was calculated as the difference between EDV and ESV. Ejection fraction (EF) was calculated as SV divided by EDV. All parameters were divided by body surface area (BSA) to achieve values adjusted for body size (e.g. LVM/BSA for left ventricular mass adjusted for BSA). In addition, a subset of 20 subjects were analyzed by 2 readers blinded to each others results in order to examine interobserver variability. Intraobserver variability was assessed by one observer who performed repeated blinded measurements in the same 20 subjects one month later.

Statistical analysis

SPSS (version 16) was used for statistical calculations. A p value less than 0.05 was considered statistically significant. Values are expressed as mean ± SD. Unpaired Student's t test or ANOVA were used to test for significance between groups since both visual inspection and the Kolmogorov-Smirnov test showed that all measures were adequately normally distributed. Intra- and interobserver variability were assessed as the mean difference of measurements ± SD and by the intraclass correlation coeffecient (ICC) employing a two-way mixed model [12]. The coefficient of variation (CV) was calculated as the SD of the difference between two measurements expressed as percent of their mean. The coeffiecient of repeatability (CR) was calculated as two times the SD of the difference in two measurements. Pearson's correlation coefficient was used to assess the correlation between two variables, and expressed as its square (R²). BSA was calculated using a previously described technique [13]. Curve estimation and the 95% prediction intervals of LVM, dimensions, and function were defined using commercial software (Matlab curve fitting toolbox, Matlab version R12, Mathworks). The most appropriate curve fitting algorithm for LVM, dimensions and functional measures was identified as the rational polynomial of the form P(x)/Q(x) where both numerator and denominator were at most of the second degree and which had the highest adjusted R² and lowest root mean square (RMS) of the error [14]. The predicted lower, mean and upper limits for normal values of LV parameters in each decade were calculated as the average of the mean and 95% prediction interval of the predicted values for each whole year as given by the curve estimation model.

Population description

Normal variation of left ventricular measures with age

i) Left ventricular mass

Information on the age dependence of LV mass and wall stress in the majority of this population have been published and discussed previously [14]. The current study, however, provides tabular values in order to ease the use of this information as reference values. Figure 2 and Figure 3 describe the changes in left ventricular mass with age.

ii) End-diastolic and End-systolic volume

Ventricular volumes varied markedly between males and females (Figure 2 and Figure 3). Notably, there was a rise in EDV and ESV during adolescence and early adulthood in males, with a decline thereafter, possibly reflecting the concurrent change of LVM with age. This was absent in females with a trend towards a decrease of EDV and EDV/BSA with age and increase of ESV and ESV/BSA with age. Ventricular volumes in the adolescent group were similar between males and females (EDV: 120 ± 29 ml vs. 114 ± 23 ml, p = 0.66; EDV/BSA: 76 ± 11 ml/m² vs. 77 ± 11 ml/m², p = 0.95; ESV: 35 ± 8 ml vs. 32 ± 11 ml, p = 0.48; ESV/BSA: 22 ± 4 ml/m² vs. 21 ± 7 ml/m², p = 0.65). However, differences between genders became apparent for all ventricular volumes in the adult group with higher values in the male group (EDV: 153 ± 30 ml vs. 118 ± 19 ml, p < 0.001; EDV/BSA: 76 ± 13 ml/m² vs. 69 ± 9 ml/m², p = 0.006; ESV: 63 ± 17 ml vs. 43 ± 11 ml, p < 0.001; ESV/BSA: 31 ± 8 ml/m² vs. 25 ± 6 ml/m², p < 0.001).

iii) Left ventricular stroke volume and ejection fraction

In general, both SV and EF demonstrated a decline with age (Figure 3). Although differences in SV and EF between males and females were present in the adult group (SV: 91 ± 19 ml vs. 75 ± 15 ml, p < 0.001; SV/BSA: 45 ± 9 ml/m² vs. 44 ± 9 ml/m², p = 0.51; EF: 0.59 ± 0.07 vs. 0.64 ± 0.08, p = 0.014), SV and EF were similar between genders in the adolescent group (SV: 85 ± 22 ml vs. 83 ± 16 ml, p = 0.82; SV/BSA: 54 ± 9 ml/m² vs. 55 ± 7 ml/m², p = 0.70; EF: 0.71 ± 0.04 vs. 0.73 ± 0.06, p = 0.36). Notably, left ventricular EF in the adolescent group was higher (Figure 3) compared to adult subjects (0.72 ± 0.05 vs. 0.61 ± 0.07, p < 0.001) with a decline in both males and females from ~70% to ~60% with age.

Normal ranges for left ventricular mass, volumes and function

All variables varied according to age group (ANOVA, p < 0.05) except for EDV/BSA (p = 0.13) and ESV/BSA (p = 0.06) in males, and EDV (p = 0.39), EDV/BSA (p = 0.15), ESV/BSA (p = 0.28), SV (p = 0.14) and LVM/BSA (p = 0.24) in females.

Intra- and interobserver variability

A) Left ventricular mass

The findings regarding left ventricular mass in the present study and its relation to earlier studies has been discussed earlier [14]. Other non-invasive techniques including three dimensional echocardiography [15, 16] and computerized tomography [17] have overcome many of the shortcomings of two dimensional echocardiography although no reference values for normal patients have been established in a large population.

B) Left ventricular volumes, stroke volume and ejection fractions

Six CMR studies [3–5, 7–9] have described overall values of EDV, ESV, SV, and EF, which are consistent with findings from other imaging modalities and are, in the corresponding age ranges, broadly consistent with the findings in this study, with one exception. Sandstede et al used a gradient echo sequence and reported as much as 20–30% lower values for LV mass and volumes. Those authors chose to only measure LV volumes and mass in short axis slices which contained more than 50% of the circumference of the LV wall in both end-diastole and end-systole. This approach does not take into account the long axis movement of the basal parts of the left ventricle [11], which in part may explain the lower values obtained in that study. Hudsmith et al, showed differences between subjects above and below the age of 35 [8], but continuous age-specific values of these parameters, however, have not been described using gradient echo CMR in populations that extend over the breadth of age as in the present study. We show that adult females had a slight progressive decrease in ESV with age, with a resulting slight progressive increase in EF, whereas Maceira et al [9], showed the opposite. BMI and blood pressure were similar for both populations, hence it is likely some other factor which contributes to the difference between the studies. The discrepancy is difficult to interpret, and one can only speculate as to what difference between the populations may explain this discrepancy.

The inotropy of the left ventricle has been shown to be related to growth hormone/IGF levels [18]. Although our study has not measured growth hormone levels, one might speculate that the higher EDV, SV and EF in younger subjects may possibly, in part, reflect the higher growth hormone levels within this age group. Alternatively, these higher values may be due to physical activity rather than hormones. However, we have previously published that self-reported physical activity increased with age in our population [14]. Thus, it is unclear exactly why these measures are larger in younger subjects.

Reproducibility of measurement

The reproducibility of CMR LV measures obtained in our study was consistent with previously reported studies of intra- and interobserver variability [19–21].

Limitations

Regurgitation in the aortic or mitral valve was only assessed visually in long axis CMR images, and this is a limitation. However, it is likely that this was sufficient to serve the purposes of this study in this population with otherwise unremarkable ECG, medical history and physical examination. The number of included subjects in each gender and decade is less than the minimally suggested n = 10 [21] for the size of a group needed to determine reference values. However, the suggestion of n = 10 is based on the use of the mean +/- 2SD of those subjects' measurements as a basis for calculating reference values. In contrast, we used a curve estimation model and its 95% prediction interval to determine our reference values. Using this method, the robustness of the reference values is based on data from the entire population and less susceptible to small numbers of subjects in individual decade groups. The normal values provided in this study are appropriate for studies undertaken using similar gradient echo sequences at 1.5 T. It has been shown that steady-state free precession based sequences may result in slightly lower LVM and greater LV volumes [7, 22, 23]. These differences are likely to be systematic in nature and do not alter the physiological significance of the age trends reported in the current study. Furthermore, the use of steady-state free precession based sequences at a field strength of 3 T is currently hampered by artifacts [24]. The presented normal values for gradient echo sequences at 1.5 T may thereby also be of value for assessing results from gradient echo cardiac imaging at 3 T.