Study selection and included studies characteristics
After an initial search through the databases, 5609 records were identified for screening. Following the removal of duplicates, titles and abstracts of the remaining studies were investigated and irrelevant or ineligible records were eliminated. Finally, 14 studies with 789 thalassemia participants (310 and 430 with and without CIO respectively and 49 with undetermined condition) met all the eligibility criteria for the systematic review and meta-analysis (15-17, 21-31) (Table-1 summarized their characteristics) (Figure-1). Eight studies with 405 patients with β-thalassemia accepted to share their data to be analyzed for IPD (12, 15, 17, 21, 22, 24, 30, 31). It is important to note that some authors have provided us with data from more number of patients than those used in their original investigation.
Global longitudinal strain
The pooled mean ± standard deviation (SD) GLS was -17.48 ± 2.69 and -19.88 ± 2.33, respectively in thalassemic patients with and without CIO. Comparison of GLS between the thalassemic groups with and without CIO showed that GLS was significantly lower in patients with CIO compared to those without CIO (WMD=1.60, 95% CI= [0.76 - 2.45], p=0.001, I2=77.1%) (Figure-2a). Heterogeneity decreased from 77.1% to 52.3% after subtraction of outliers although excluding the outliers diminished the level of association (WMD=0.90, 95% CI= [0.25 - 1.55], p=0.007). Also the mean ± SD GLS in the control group was -20.61 ± 2.26 and the difference between the control and thalassemia groups with overload was also found to be significant (WMD=2.20, 95% CI=[0.91 - 3.50], p=0.001, I2=83.90%) (Figure-2b). After excluding the outliers the level of heterogeneity decreased from 83.9% to 48% but the level of association remained high (WMD=1.58, 95% CI= [0.79 - 2.37], p=0.001).
Meta-analysis of the IPD showed a cardiac iron overload prevalence of 35.1%. Characteristic ROC curve analysis revealed that an optimal cut-off value of -19.5 for GLS could predict the cardiac iron overload in patients with thalassemia with sensitivity and specificity of 92.8% sensitivity and 34.63%, respectively (AUC=0.659, SE=0.0291, 95% CI= [0.60 - 0.72], p-value<0.0001) (Figure-2c). A GLS value ≤ -5.74% had 100% positive predictive and ≥ -24.5% had 100% negative predictive value for detection of CIO.
Left ventricular ejection fraction
Pooled mean ± SD of LVEF in the thalassemic groups with and without CIO and also control group was 57.74 ± 6.92, 60.12 ± 6.84, and 62.38 ± 4.13, respectively. No significant differences were observed regarding LVEF in patients with thalassemia with and without CIO (WMD=-2.38, 95% CI= [-4.83 - 0.06], p=0.056, I2=80.5%) (Figure-3a) Subtracting outliers reduced the level of heterogeneity from high to low (I2=0.0%) (WMD=0.22, 95% CI = [-1.12 - 1.55], p=0.752). Also between healthy subjects and thalassemia groups with CIO (WMD=-2.42, 95% CI= [-5.59 - 0.74], p=0.133, I2=78.6%) (Figure-3b). The level of heterogeneity remained high after removing outliers (I2=71.9%).
Other echocardiographic parameters
There was a significant correlation between lower levels of GCS and myocardial iron overload (WMD=2.69, 95% CI= [1.41 - 3.97], p<0.001, I2=22.3%) with possible publication bias according to Egger’s test (p=0.016) although evaluation of trim-and-fill method showed minimal evidence for publication bias (Figure-4). In contrast, no difference was found regarding the values of GRS between the two groups of thalassemia (WMD=-3.853, 95% CI=-7.796 to 0.091, p=0.055, I2=0.00%) (Supplementary material, Figure-1). E/A ratio and E/E’ ratio comparison did not reach a significant level between the two groups (E/A: WMD=0.159, 95% CI=-0.104 to 0.422, p=0.237, I2=84.4%) (Supplementary material, Figure-2) (E/E’: WMD=1.006, 95% CI=-1.574 to 3.585, p=0.445, I2=95.6%) (Supplementary material, Figure-3). No correlation was also observed regarding the deceleration time between the two thalassemia groups (WMD=-5.343, 95% CI=-38.906 to 28.221, p=0.755, I2=92.0%) (Supplementary material, Figure-4).
Risk of bias of the included studies
After evaluating the quality of the eligible studies using JBI Checklist, we decided to include all the studies in the meta-analysis due to their high-quality. Table-2 summarizes the risk of bias assessment of the eligible studies.
Evaluation of the funnel plots for the main outcomes (GLS and LVEF) demonstrated the presence of asymmetrical distribution among the studies (Supplementary material, Figure-5-8). However, p-values of Egger’s test did not reach a significant level for publication bias (p=0.144 for GLS between thalassemia patients with and without CIO, p=0.510 for GLS between thalassemia and control group, p=0.777 for LVEF between thalassemia patients with and without CIO, p= 0.263 for LVEF between the thalassemia and control group). For sensitivity analysis, we removed each single study from all the analyses to see their impact on the summary of results and no significant change was observed for all the parameters.