This is a national multicentre cross-sectional study. We collected data prospectively from a convenience sample consisting of all consecutive patients diagnosed and treated for thalassemia major or intermedia recruited from seven Brazilian hematologic centres from 2004 until 2011. These patients were treated locally and referred to our service, a private hospital in São Paulo, SP, Brazil, for MRI and clinical evaluation only.

The study was funded by the Brazilian Programme Supporting the Institutional Development (PROADI) by the Brazilian Ministry of Health and the National Health System. All the patients studied signed informed consent forms and the procedures were approved by the Hospitals Ethics Committees. This study was conducted in accordance with the Helsinki Declaration as revised in 2008.

All consecutive thalassaemic patients admitted in the participant centres were considered eligible for this study under the following conditions:

• ‐

  age ( 7 years;

• ‐

  thalassemia major or intermedia patients;

• ‐

  treated according to the Brazilian clinical guidelines1415 based on regular packed red cell transfusions according to the severity of iron overload;

• ‐

  chelation therapy based on the Brazilian protocol or with an indication for this treatment;14

• ‐

  submitted to MRI T2* exam to evaluate pancreas, heart and liver;

• ‐

  serum ferritin evaluated at admission and at least six months later (minimum follow-up of six months);

• ‐

  signed informed consent forms.

Exclusion criteria were these situations:

• • ‐

    Pacemaker carrier;

  • ‐

    Intracranial aneurysm clipping carrier;

  • ‐

    Claustrophobia;

  • ‐

    Pregnancy;

  • ‐

    History of noncompliance to chelation therapy.

The MRI equipment used to assess tissue iron content consisted of 1,5T scanner (General Electric Healthcare - Milwaukee, Wisconsin - USA) with the combination of body-matrix and spine-matrix surface coils. The myocardium sequences measurements were analysed using the ReportCard V3.3 (General Electric Healthcare - Milwaukee, Wisconsin- USA). Multi-echo segmented fast gradient echo pulse (MFGRE) sequence was used with the following parameters: FOV 44 cm, PFOV = 0,7, slice thickness 10 mm, spacing 8.0 mm, 4 slices, TR = 34 ms, FLIP 20 degrees, TEmin = 2 ms, TEmax = 16 ms, 8 echoes with 2 ms inter echo, frequency matrix = 128, phase matrix = 128, NEX = 1, receiver bandwidth = 62.5 kHz; post-processing. T2* decay fitting was performed on Report Card using the following formula: S = A*e (-TE/T2*) + C. The initial scaling factor was 25 ms, and 2 parameter fit was used (C = 0). The following equation was used to calculate hepatic iron concentration (LIC): [Fe]R2* = 0.0254 x R2* + 0.202. The same parameters were used employed for cardiac iron assessment, using short-axis view focused on the myocardial interventricular septum.

The hepatic and pancreatic relaxation times at T2* MRI were assessed at the transversal plane of the upper abdomen. Regarding the liver, an experienced radiologist selected a region from the right liver lobe in segments 6 and 7 to avoid vessels and possible artefacts.19 Paraspinal musculature signal intensity was also assessed to calculate the signal intensity ratio (SIR) between these muscles and the pancreas.1820

The severity of iron overload on each organ was classified for each organ and technique, according to Supplementary Table 1.31012202122 The radiologic report was done by two experienced radiologists.

In this study, “improvements” of iron overload correspond to the difference (Δ) between initial and final measurements in a given period. It is expressed as milliseconds (ms), with the final measure being higher than the initial one (thus a positive value), or in milligrams of iron (Fe mg/g), where it is expected that the final measure be lower than the initial (negative value).

The serum ferritin measures were done by chemiluminescence.10

All lab tests and MRI assessments were performed at our service, using the same equipment. MRI tests were analysed by two experienced radiologists who mutually agreed to the calculation of T2*. In cases where images had shown artefacts, which prevented further analysis, those patients were excluded, and results were not recorded.

All patients had been assisted by their own treating physicians at each centre, who referred them to our service for the tests (MRI and laboratory) and medical evaluation at least every six months to record all relevant clinical information (clinical symptoms, medications/chelation, transfusion requirement etc.). Therefore, our centre did not interfere with treatments. Each centre assistant physicians were allowed to assess the MRI and lab tests results, so the treatment choice was not blinded. These physicians received general guidance to follow the Brazilian chelation protocol, but each physician, according to his/her discretion, could adjust the patient's treatments according to specific individual characteristics and tolerability. They also were required to inform our Hospital about any modification on the chelation therapy or patient treatment during the follow-up. In consequence, we could not ascertain strict compliance (assistant physician-patient) to the national guidelines for treatment, which limited a reliable comparison between each type of iron chelation and explains why some patients, although eligible, were not undergoing combined chelation.

In general, the Brazilian protocol1415 recommends chelation therapy with desferrioxamine or deferasirox to patients with hepatic iron overload and combined therapy (desferrioxamine plus deferiprone) to those with concomitant cardiac iron overload.

The variables under investigation were:

• •

  Serum ferritin levels (ng/ml)

• •

  T2* MRI (results expressed in milliseconds, ms)

• •

  R2* (hertz) to assess pancreatic iron; formula R2* = 1000/T2*

• •

  Fe mg/g dry weight liver to assess LIC; [Fe]R2* = 0.0254xR2* + 0.202

• •

  Fe mg/g dry weight to assess myocardium iron, using the Carpenter equation [Fe]= 45.0•(T2*)−111

• •

  Severity and proportions of iron overload according to the thresholds established in the literature (Supplementary Table 1) for MRI.

• •

  Δ ferritin/year (ng/ml/year) and ΔR2*/year (hertz/year): were defined as the difference between the final and the initial ferritin or R2* organ measurement rationalized by the time (year)

• •

  Δ Fe mg/g/year: was defined as the difference between the final and the initial Fe mg/g dry weight organ measurement rationalized by the time (year)

The continuous variables (serum ferritin:, LIC, myocardium T2*, and pancreatic R2*) were described by medians, means, range and standard deviation. The Shapiro-Wilk method was applied to test for normality distribution. The correlations between those variables were analysed using the Spearman's rank test. The median variations (Δ) on iron concentrations were adjusted by the time (year) elapsed from initial to final evaluation. The Kruskal-Wallis independent test was used to compare iron variations (Δ) during follow-up between more than two chelation treatment groups.

The Wilcoxon paired test was used to compare these variables at baseline and follow-up in the same group. While, the Mann-Whitney unpaired test was used to compare those variables between different patient groups.

Compliance was assessed indirectly. Firstly, we separated patients with complete data available in two groups according to the median follow-up: short-term (1.2 years) versus long-term (5.2 years), then we compared the iron parameters as described above. Also, we analysed the estimated incidence ratios of new cases of cardiac or pancreatic iron overload.

The non-parametric receiver operating characteristic (ROC) curve was applied to test the accuracy of serum ferritin and pancreatic R2* thresholds to predict cardiac iron overload. The STATA software version 11 and SPSS version 24 were used for the statistical analysis.

In the study period, from a total of 211 patients enrolled in the study, we included 157 that we assessed at baseline and at least once more during follow up (54 were excluded due to missing final MRI results). However, from the 157 elected patients, we excluded further 21 because medical records were incomplete (lacking data from the second evaluation mostly).

The baseline clinical features of the 136 included patients were: major thalassemia (126/92,5%) and intermedia thalassemia (10/7,5%); male (47/34%) and female (89/66%); median age of 18 years old (range 7-55); median follow up time of 1.2 years (range 0.6–5.2 years); median serum ferritin of 2033 ng/ml (range 59–14.123). The proportion of patients observed on each chelation treatment was: desferrioxamine (51/37.5%), combined therapy (38/28%), deferasirox (35/26%) and deferiprone alone (12/8.5%). Only 16/57 (28%) of patients with myocardium iron overload were being treated with combined iron chelation at the baseline evaluation. During long-term follow-up, this proportion increased to 16/33 (50%).

At baseline evaluation, the organ most commonly affected by iron overload was the liver, affecting 156 (99,3%) of the patients, followed by pancreas 117 (74,3%) and heart 57 (36%). Supplementary Table 2 summarizes the distribution and severity of iron concentration in each organ.

Table 1 shows the MRI measures of iron concentration in each organ and serum ferritin at baseline and during follow up (median 1.2 years; range 0.6 to 5.2) on patients being treated with iron chelation. We observed a significant decrease in serum ferritin (median 2033 to 1735; p = 0.001) and myocardium Fe (median 0.86 to 0.77; p = 0.024), while we did not detect any significant improvement in LIC and pancreas R2*.

As illustrated in Table 2, we found statistically significant moderate (0.527) correlations between LIC and ferritin parameters and also between R2* pancreas and Fe mg/g myocardium (0.476).

As shown in Table 3, the subgroup of patients (n = 50) with cardiac iron overload had a significant improvement in the myocardium Fe mg/g, reaching a median reduction of 0.27 Fe mg/g per year (p = 0.003). The majority of patients improved the myocardium iron concentration, corresponding to 38/50 (75%).

We also analysed the heart using T2* MRI (ms), which resulted in a significant median improvement of 1 ms/year (range 0.4–2.1ms/year). For the R2* pancreas and LIC, we did not observe a significant iron load improvement during follow up.

Analysing each iron-chelating therapy in the short-term follow up (median 1.2 years), we did not find any differences between combined, desferrioxamine, deferasirox and deferiprone therapies in neither iron parameter, as shown in Table 4.

Because we observed a moderate correlation between R2* pancreas and myocardium Fe mg/g and considering that ferritin could be a very sensitive parameter of iron overload, we performed a non-parametric ROC analysis using these variables. We observed that R2* pancreas or ferritin alone would not be enough to guide us for myocardium iron. The optimal R2* pancreas cut-off was 148 Hertz, achieving 78% sensitivity and 73% specificity (Figure 1). However, when combining the R2* pancreas cut off ≤ 50 Hertz and a ferritin ≤ 1222 ng/ml, we could reach a negative predictive value (NPV) of 98% for cardiac siderosis.

Figure 1.

Diagnostic accuracy of R2* pancreas (dashed line) and serum ferritin (solid line) assessments to predict myocardium iron overload.

(0.1MB).

As compliance is a critical issue affecting the effectiveness of iron chelation, we indirectly assessed the adherence to treatment verifying if patients with a longer follow-up could achieve better iron parameters after chelation than those with a short-term. Our follow-up compliance seems to be much lower than the optimal target (> 80%). We had complete data from 136 patients at a median of 1.2 years (short term), but from only 33 patients at 5.2 years of follow-up (long term).

We compared the Δ Fe mg/g in the myocardium in the subgroup of 33 patients that had been compliant to MRI assessments and persisted with the same chelation therapy (i.e., a “long-term” subgroup) with the patients in the “short-term” subgroup. We observed that the median improvement rates were not significant different: -0.035 Δ Fe mg/g/year (equivalent T2* rate of +0.9 ms/year) in the long-term versus -0.023 Δ Fe mg/g/year in the short-term (p = 0.611). Half of those patients followed on long-term were using combined chelation, 40% desferrioxamine and 10% deferiprone.

Furthermore, to indirectly assess compliance, we also analysed the estimated annual incidence rates of new cases diagnosed with pancreatic or myocardium siderosis at follow-up, as one would not expect new cases of iron overload following adequate chelation. We found incidence rates of new cases of 6,1% (95% IC: 3,0-12,3%) in the myocardium and 16,4% (95%IC: 6,8-39%) in the pancreas.

We also compared if patients doing combined iron chelation were performing better than those with other chelators, and we found a significant difference in the median rate of myocardium iron overload improvement (Δ Fe mg/g/year): -0.085 in combined versus -0.011 in other chelating therapies (p 0.023), as illustrated in Figure 2.

Figure 2.

Long-term (median 5.2 years) follow-up of myocardium iron concentration according to chelation therapy: combined versus other types.

(0.16MB).

Comparing each organ iron improvement between the subgroups with longer versus shorter follow-up, as shown in Supplementary Table 3, we could not observe a significant difference in iron overload measures.