The aim of the study was to generate a formula to determine a synthetic Htc and ECV from the T1 blood relaxation times using MOLLI sequences obtained with an Ingenia 1.5T scanner and to determine whether linear or reciprocal regression is statistically more relevant to our data. Also, a correlation between Htc and T1 relaxation times using only native and both native and post-contrast values was aimed to be found. Formulae created were used for ECV quantification and compared with values acquired using different formulae and laboratory methods.
This retrospective study was composed of 139 subjects with a wide range of diagnoses, who underwent CMR examination in an Ingenia 1.5T scanner using both native and contrast methods. Inclusion criteria were having available both native and post-contrast T1 mapping sequences and a blood sample collected right before CMR examination (several minutes). The patients consisted of 5 groups following primary diagnoses: chronic obstructive pulmonary disease, Duchene muscular dystrophy female carriers, a group after anthracycline treatment, patients after a heart transplant, and controls. Control group included patients with a clinical indication for CMR examination, but normal CMR findings, other cardiac results and no other relevant medical history (Table 1). The appropriate size of the group to reveal even medium size effects was checked using the online available calculator for multiple regression [21], with parameters set as f2 = 0.15, power level (1-β) = 0.8, α = 0.05, and 3 predictors. The number of patients enrolled well exceeded the necessary minimum of n = 76.
Table 1
Characteristics of the patient groups included in the study.
Primary diagnosis
|
Number of patients
|
Age
|
Female
|
Mean laboratory Htc
|
Mean T1 blood relaxation time - native
|
Mean T1 blood relaxation time – post-contrast
|
Overall
|
139
|
35,42 ± 15,97
|
76 (54,7 %)
|
0,41 ± 0,04
|
1545,78 ± 84,24
|
229,76 ± 27,03
|
DMD carriers
|
40
|
37,93 ± 11,85
|
40 (100 %)
|
0,4 ± 0,03
|
1560,33 ± 79,66
|
223,92 ± 24,64
|
Patients after anthracycline treatment
|
71
|
25,99 ± 5
|
26 (36,6 %)
|
0,42 ± 0,04
|
1534 ± 80,12
|
230,9 ± 29,02
|
HT
|
8
|
53,63 ± 13,51
|
2 (25 %)
|
0,4 ± 0,08
|
1563,5 ± 104,27
|
234 ± 26,46
|
COPD
|
15
|
67,33 ± 8,5
|
5 (33,3 %)
|
0,42 ± 0,04
|
1566,13 ± 70,54
|
235,67 ± 19,72
|
Controls
|
5
|
24,4 ± 2,58
|
3 (60 %)
|
0,44 ± 0,04
|
1496,8 ± 69,15
|
235,8 ± 24,21
|
Values represent the number of patients or the median ± standard deviation. |
Htc = haematocrit, DMD carriers = genetic carriers of Duchenne muscular dystrophy, HT = heart transplant, COPD = chronic obstructive pulmonary disease. |
Patients had their Htc measured using standard laboratory methods in a centralised hospital laboratory. Afterwards, the patients underwent CMR including T1 mapping using MOLLI sequences both before the administration of a contrast agent and 15 minutes after. Similarly as described previously [22] - a balanced single-shot T1-TFE sequence with an inversion prepulse, cardiac triggering and breath-hold technique in the mid-ventricular level in the short-axis was used. With a 5s (3s) 3s MOLLI scheme for native T1 and 4s (1s) 3s (1s) 2s for enhanced T1 mapping was used with typical imaging parameters as follows: FOV 300 × 300 mm, reconstruction matrix 256, slice thickness 10 mm, acquisition voxel size 2.00 × 2.00 × 10.00 mm, time to repetition (TR) ≈ 2.2 ms, echo time (TE) ≈ 1.1 ms, flip angle 35°, SENSE factor 2. For the contrast agent, gadolinium (Gadovist, Bayer AG, Leverkusen, Germany) in the dosage of 0.2mmol/kg was used.
The regions of interest (ROIs) were contoured in both native and post-contrast images, including only a blood pool without any papillary muscles (Fig. 1, 2). First, the formula and correlation coefficient using only native values of the blood pool was calculated; afterwards, the same calculation was performed using both native and post-contrast values. Acquired Htc values were used to quantify ECV.
As for statistical analysis - Pearson correlation coefficients and related p-values were computed to assess the correlation between Htc and CMR-derived blood native and blood enhanced values. Four regression models to estimate Htc from blood native and blood post-contrast values were created in the following fashion:
1) (native linear): Htcest = b + (T1 − BN/a)
2) (native reciprocal): Htcest = b + (a/T1 − BN)
3) (combined linear): Htcest = c + (T1 − BN/a) + (T1 − BP/b)
4) (combined reciprocal): Htcest = c + (a/T1 − BN) + (b/T1 − BP)
Where Htcest is an estimated value of haematocrit, a, b, c are constants, T1 − BN is a blood native value and T1 − BP is a blood post-contrast (enhanced) value.
Further, using the estimated or measured Htc value, the extracellular volume was calculated as:
ECV = (1 – Htc) * (((1/T1 − MP) – (1/T1 − MN)) / ((1/T1 − BP) – (1/T1 − BN))
Where T1 − MN is a myocardium native value and T1 − MP is a myocardium post-contrast value. The ECV calculated using either measured or estimated Htc was compared to the values based on linear and reciprocal models by Treibel et al. using blood native values obtained by MOLLI and ShMOLLI (Treibel et al. 2015; 2016) and Bland-Altman analysis was performed to assess the systematic bias.
The statistical significance of these factors was assessed. To compare the linear and reciprocal models, a coefficient of determination (r2) was attributed to each model. A residual analysis was employed to confirm the adequacy of each model for the estimation of Htc and the residual variance of the model with the highest r2 was compared with other models using an F-test.
Finally, as the study group was heterogeneous and consisted of patients suffering from several diagnoses, the possible effect of the primary diagnosis was determined using an Analysis of Covariance (ANCOVA), with the primary diagnosis as an independent factor in the model.
Normality was tested by Kolmogorov-Smirnov test of normality and by visual inspection of histograms. To exclude substantial multicollinearity between the variables used in a model, the variance inflation factor (VIF) was computed for a model containing blood native value as an independent and blood post-contrast value as a dependent variable, as well as for the diagnoses (nominal data, one binary variable per each diagnosis, independent) and the blood native / blood post contrast value (dependent variable). Value > 2.5 was considered as a substantial multicollinearity.
Generally, results with P < 0.05 were regarded as statistically significant. The analysis was performed using STATISTICA 13.2 (TIBCO software, The United States of America).