MRI guided iron assessment and oral chelator use improve iron status in thalassemia major patients

Abstract

Oral iron chelators and magnetic resonance imaging (MRI) assessment of heart and liver iron burden have become widely available since the mid 2000s, allowing for improved patient compliance with chelation and noninvasive monitoring of iron levels for titration of therapy. We evaluated the impact of these changes in our center for patients with thalassemia major and transfusional iron overload. This single center, retrospective observational study covered the period from 2005 through 2012. Liver iron content (LIC) was estimated both by a T2* method and by R2 (Ferriscan^®) technique. Cardiac iron was assessed as cT2*. Forty-two patients (55% male) with transfused thalassemia and at least two MRIs were included (median age at first MRI, 17.5 y). Over a mean follow-up period of 5.2 ± 1.9 y, 190 MRIs were performed (median 4.5 per patient). Comparing baseline to last MRI, 63% of patients remained within target ranges for cT2* and LIC, and 13% improved from high values to the target range. Both the median LIC and cT2* (cR2* = 1000/cT2*) status improved over time: LIC 7.3 to 4.5 mg/g dry weight, P = 0.0004; cR2* 33.4 to 28.3 Hz, P = 0.01. Individual responses varied widely. Two patients died of heart failure during the study period. Annual MRI iron assessments and availability of oral chelators both facilitate changes in chelation dose and strategies to optimize care.

Introduction

Thalassemia represents the most common monogenetic disorder that causes defects in hemoglobin synthesis worldwide [1]. By definition, patients with thalassemia major (TM) require regular red blood cell transfusions throughout life. In addition, patients with thalassemia “intermedia” phenotypes (not transfused as children) may require a switch to regular transfusions to manage symptoms or medical complications. In TM patients, the leading cause of death in most centers continues to be cardiac failure due to cardiac iron overload; in addition, iron overload in the liver, pancreas, and other organs causes progressive damage. This morbidity and mortality is preventable provided adequate chelation therapy is instituted and maintained over time (i.e., tissue iron is kept within the target range, or if it excessive, lowered to the target range by use of an effective chelation strategy which the patient is willing and able to use). Three chelators are FDA approved in the US: deferoxamine (DFO), deferasirox (DFX), and deferiprone (DFP), (approval years 1968, 2005, and 2011, respectively) [2–5]. The era of widely available oral chelators began only in 2006, with the commercial launch of DFX.

Adequate iron chelation must be guided by accurate assessment of tissue iron levels. Magnetic resonance imaging (MRI) can noninvasively measure liver iron content (LIC) and cardiac iron (cT2*), and has almost entirely supplanted liver biopsy for LIC at our center, beginning in 2005 for T2*-based methods and in 2006 for the R2 MRI method (Ferriscan®). The therapeutic goal of iron chelation is to either (a) maintain tissue iron status within a consensus target range or (b) decrease the iron burden in patients above the target range.

Large studies, including the multicenter, international Thalassemia Longitudinal Cohort (TLC) study of the NHLBI-sponsored Thalassemia Clinical Research Network, have aimed to assess the changing face of strategies to manage iron overload in the era of oral chelation [6]. Here, we present our single-center experience with MRI iron assessments and chelation changes and our ability to improve patients’ iron status over time.

Methods

Study population

Our single center, retrospective observational study covered the period from January 2005, when MRI iron assessment became standard at Boston Children’s Hospital, to December 2012. The study population included 42 patients with TM followed for both chelation and transfusion at our center, who had ≥2 MRI assessments during the study period. Permission to conduct this database review analysis was granted by the hospital IRB.

We defined baseline as January 2006 (when both MRI methods were available) or first MRI if patients joined later. We defined end of the study as December 2012 or last MRI if patients were lost to follow-up or died prior to that time. Some patients had more than one MRI per calendar year, mainly because they were part of clinical trials. In these cases, we analyze the last MRI assessment per year, except in one severely affected patient who had only two MRI assessments which occurred within the same year.

Some but not all of the patients in this single-center study are included in the data analysis of the NIH-funded TLC study [6]. We took advantage in this single-site study to tailor the time frames of assessment to the dates of MRI studies over as many as 5–6 y, whereas the main analysis of iron status in TLC was from a baseline assessment and 2–3 annual follow-up studies.

To account for the potential effect of age, we performed stratified Wilcoxon signed-rank tests for the liver and cardiac iron concentration measures (stratified by age group) [7].

MRI technique

Liver iron concentration (LIC) was measured by calculating T2* and, starting April 2006, also by measuring R2 using the commercial Ferriscan® technique [8] Liver T2* was converted to LIC using a regression equation: (LIC R2* = 0.0254 × R2* + 0.202; where R2* = 1000/T2*) [9]. A three-component model was used to calculate the LIC as described by Wood et al. [9] Cardiac iron concentration was estimated as cT2*; in Fig. 2, we reported both cT2* in msec and its reciprocal cR2* (1000/cT2* in Hz) which is directly proportionally to iron.

Figure 2.

Box-and-Whisker Plots. a: Cardiac iron status from first to last MRI for each subject. Reciprocal cR2* and cT2* are on left and right Y axes, respectively. Although clinicians have become accustomed to using cT2* clinically, it is reciprocally related to iron concentration, and so gives a highly nonlinear response. b: Liver iron status estimated by R2 (Ferriscan®) and T2*. In both panels, the boxes represent IQR; the median is shown by the horizontal line. The error bars represent 95 percentile, and outliers are shown. P values are by Wilcoxon signed-rank test.

Clinical management

Throughout the study period, our center policy was to assess TM patients with MRI annually (for subjects old enough not to require sedation), and ferritin with monthly transfusions. The clinical target for LIC was ≤7 mg/g dry weight (mean of T2* and Ferriscan LIC) and cT2* ≥20 msec for cardiac iron. The approach to chelation changes was semiquantitative, namely that for values above threshold, we considered escalation of chelation. We deemed LIC values >15 mg/g dry weight, and cardiac T2* <10 msec as highly undesirable, requiring more aggressive changes in chelation strategy. This “semiquantitative” approach is used because MRI (or biopsy) iron assessments have intrinsic variability: from one study to the next, all other parameters being equal, MRI-based iron assessment may look somewhat “better” or “worse” from time to time. In this study, we define “improvement” as lower LIC or higher cT2*, of any degree. For ferritin assessments, as many values are available, we used the last nonmissing running 5-month average ferritin value from every subject. If a clear downward trend in iron burden was evident from available recent scans, we did not escalate the dose. If adherence was thought to be the main problem, we addressed this issue rather than prescribe increased doses. If transfusion burden seemed higher than expected (e.g., >0.5 cc/kg/4 weeks PRBC volume), we considered splenectomy or alternative transfusion strategies in some patients.

MRI assessment of the heart also yielded anatomical and functional data that might be taken into account in management from time to time, for example, if the left ventricular ejection fraction (LVEF) by MRI dropped below 55% (abnormal in our center for thalassemia patients), this might trigger more aggressive chelation by dosage and/or switch in chelators.

Although DFX did not become available commercially until early 2006, a fraction of our patients received DFX on research studies in 2005.

Statistical analyses

All statistical tests were two sided, and statistical significance was set at the 0.05 level. The Wilcoxon signed-rank test was used to compare the first and last MRI measurements of iron burden (liver T2* and cardiac cT2*) for each subject. The stratified Wilcoxon signed-rank test was used to adjust for age in the comparison of the first and last MRI measurements of iron burden for each subject [7]. The Bland–Altman method was used to compare the T2* and Ferriscan measures of liver iron concentration. All statistical analyses were performed using SAS software version 9.3 (SAS Institute, Cary, NC).

Results

Forty-two patients (55% male, mean age at end of study 21.6 y) met the inclusion criteria, and the mean age at first MRI was 17.5 y (range 1.9–43). The mean follow-up period was 5.13 ± 1.8 y. A total of 190 MRIs were performed, with a median (interquartile range) of 4.5 (3–6) MRIs per patient. Abnormal LVEF was noted in 4 of 23 adults and none of the young patients at time of last MRI. Summaries of patient demographics and baseline characteristics are presented overall and by age group in Table I.

TABLE I.

Demographics and Baseline Characteristics for all the Patients and Patients Divided by Age Group

Characteristics of the study population			Characteristics of the study population by age at last MRI
Parameter	All Subjects (n = 42)		Age Group
			<10 years old (n = 9)	10–18 years old (n = 10)	>18 years old (n = 23)
Male, n (%)	23 (54.7)		5 (55.6)	6 (60)	12 (52.1)
Ethnic Group, n (%)			Ethnic Group, n (%)
Mediterranean	23 (54.8)		2 (22.2)	5 (50)	16 (69.6)
Asian	15 (35.7)		5 (55.5)	3 (30)	7 (30.4)
Other	4 (9.5)		2 (22.2)	2 (20)	0
Categories of Genotype, n (%)			Categories of Genotype by age group, n (%)
β-Thalassemia Syndromea,b	39 (90.7)		9 (100)	8 (80)b	22 (95.6)a,b
α-Thalassemia Syndromec,d	3 (7)		0	2 (20)c	1 (4.4)d
MRI, Median (IQR)		P-valuee	MRI, Median (IQR)			P-valuef
Baseline LICh	7.3 (3.3–11.4)		8.28 (6.7–10.7)	6 (4.3–9.1)	8.2 (2.3–16)
Last LIC (mg/g)	4.5 (2.9–6.9)		4.06 (3.34–5.2)	3.3 (2–4.5)	5.8 (3.4–7.8)
Change in LIC	−2.6 (−5.8, −0.05)	0.0003	−4.6 (−6.1, −1.4)	−3.2 (−4.6, −1.5)	−1.7 (−6.5–1.4)	0.003
Baseline cT2*	29.6 (12.7–37.6)		40.8 (32–42.5)	30.2 (25–37.6)	16.1 (8.6–36.2)
Last cT2*(msec)	35.1 (20.8–38.3)		38 (30.7–39.3)	35.7 (35–37.3)	28.3 (13.4–38.3)
Change in cT2*	1.25 (−1.5–8.1)	0.03	−4.5 (−5.1, −1.4)	0.7 (−0.6–7.7)	1.9 (−0.5–9.4)	0.01
Baseline cR2*	33.8 (26.6–78.7)		24.5 (23.5–31.4)	33 (26.5–40)	62.1 (27.6–116.3)
Last cR2* (Hz)	28.5 (26.1–48)		26.3 (25.4–32.6)	28 (26.8–28.5)	35.33 (26.1–74.2)
Change in cR2*	−1.47 (−11.5, −2.7)	0.02	3 (0.8–4)	−0.7 (−6.9, −0.3)	−4.7 (−51.5–0.5)	0.01
LVEF, n (%)h			LVEF, n (%)h
<55%	4 (9.5)		0	0	4 (17.4)
≥55%	37 (88)		8 (88.8)	10 (100)	19 (82.6)
Ferritin (5 month running average), Median (IQR)			Ferritin “running average” over 5 months, Median (IQR)i
1238.2 (909.26–2021)			1569.2 (1144.9–2021)	1140.2 (857.4–1349)	1175.2 (953.3–2383.2)

^aTwo β-Thalassemia patients were treated as intermittently transfused β-Thalassemia intermedia in the past.

^bThree patients with Hb E/Beta0 thalassemia (2 >18 years old and 1 at the age group 10–18 years old).

^cIncludes two patients with alphathalassemia/Hb Dartmouth compound heterozygotes.

^done patient with 4α-gene deletion.

^eP-value for the change in iron concentration by the Wilcoxon signed-rank test.

^fP-value for the change in iron concentration by the stratified Wilcoxon signed-rank test.

^gLIC is expressed as the mean of T2* and Ferriscan® methods, that is ([LIC ferriscan + LIC T2*]/2).

^hLVEF assessed by MRI.

ⁱIQR = Interquartile range.

Chelation status of the study population

The launch of DFX in 2006 resulted in dramatic changes in the chelator usage across our population, so that DFO usage decreased from 70% of subjects to 10% in 2009. DFX usage increased from 26% at baseline (mostly research use) to 73% by 2009. See Table II.

TABLE II.

Chelator Regiments in our Population From the Baseline (2005–2006), Mid Study (2009) and Last Assessment (2012)

Chelators regimens	Baseline 2005–2006; n = 30a	Mid study 2009; n = 40	Last assessment 2012b; n = 41
DFO n (%)	3 (10)	4 (10)	5 (12)
Exjade n (%)	27 (90)	29 (73)	27 (66)
L1 n (%)	–	–	3 (7)
DFO/L1 n (%)	–	3 (8)	2 (5)
DFO/Exjade n (%)	–	4 (10)	2 (5)
Research chelators n (%)	–	–	2 (5)

^aThree new patients in 2006. Sixteen patients switched DFO for Exjade between 2005 and 2006. One patient switched DFO/L1 for Exjade between 2005 and 2006.

^bLast assessment in 3 patients was in 2010 and in 2 patients in 2011.

Figure 1a shows dosing changes: most of the changes were due to iron burden, but 3% were dose decreases for side effects. Figure 1b shows changes in chelator regimen (exclusive of dose changes): the majority of the changes were predominantly related to either the commercial launch of DFX or to clinical trials. The median number of chelation changes (dose change or chelator change) was 1.4 per patient per year (IQR 0.9–1.9).

Figure 1.

a: Changes in chelator dosage regimen (without change in chelator). “Other Reason” included research studies which defined a dose and one subject who modified his own doses (without change in prescription) when ferritin levels were very close to 500 ng/mL. b: Changes in chelator regimen only. “Changes” could include i) changes in monotherapy, ii) change to multiple chelators, and iii) drug holidays for toxicity or for low ferritin (DFX).

Iron status of the study population

Thirteen of 42 patients (31%) remained within the target range for cardiac T2* (cT2*) and LIC throughout the study period, while 69% of the overall population had at least one cT2* or LIC out of the target range (comprising 97 scans overall of the 190) during the follow-up period. Thirty-eight of these 97 (40%) out of range values prompted a change in chelation strategy (see earlier).

At the initial MRI, 16 of 41 (40%) patients were in the target range for both LIC and cT2*, 4 of 41 (10%) were in the highest (undesirable) range of LIC >15 mg/g dry weight and/or cardiac T2* <10 msec. Cardiac and liver iron burdens significantly decreased over the follow-up period for patients in our center (P = 0.003 and P = 0.0003, respectively; Fig. 2).

From first to last cT2* assessment (n = 38), 63% of patients started and ended within the target range, 13% improved from abnormal to target range, 24% remained out of the target range. For LIC (n = 42), 45% remained in the target range throughout, 33% started out of target range and ended within, 12% improved but not to the target, 7% worsened, and one outlier remained severe.

The observed results were similar to the nonstratified tests (see Table I) with adjusted P-values of 0.0036, 0.0103, and 0.0131 for the change in LIC and the change in cT2*, respectively. (See supplemental Fig. 2) As shown by others, we found an association between elevated myocardial iron concentration (T2* <20) and low LVEF. A retrospective study showed a significant correlation between LVEF and cardiac T2* in patients with cardiac siderosis [10,11].

From patient to patient, changes in iron status over time were plotted as time-directional arrows in the Supplemental Fig. 1. Individuals have wide variations in responses. Two patients died of heart failure at our center due to iron overload during the study period; both had taken DFP before their deaths but for different durations (3 days and 5 years). The two patients who died were among the persistent abnormal cardiac T2* group and had long histories of poor adherence to prescribed chelation therapy. One patient had the last MRI 12 days before death: LIC was 18.1 and cT2* was 4.6; the other patient had a MRI 10 months before death: LIC 3.1 and cT2* 6.4.

Our center measures ferritin with monthly transfusion visits. Whereas single-ferritin values are extremely variable, running 5-month averages smooth monthly variation. We use these running averages to monitor changes in iron status between annual MRIs. The change in ferritin values for patients with abnormal MRI was not significantly different between patients, who did and those who did not have a change in chelation treatment within 3 months of the abnormal MRI (P = 0.2639).

We performed a Bland–Altman analysis to compare the two MRI liver iron assessment methods. The Bland–Altman plot indicates that the T2* method yields a higher apparent LIC than Ferriscan (difference >0) at low overall iron content, but the T2* method is relatively insensitive to very high iron levels, and therefore, yields a lower LIC iron estimate in the face of marked iron overload (>20 mg/g dry weight), see Fig. 3. These data allow us to assess the potential clinical impact of using one or the other method as opposed to the mean of the two: during the study period, there were 176 observations for which both T2* and Ferriscan measurements were obtained. Treating LIC >7 mg/g dry weight as the cut point for a clinical recommendation to increase chelator treatment, there were 25 observations (14.2%) for which chelator treatment would be recommended by T2* and not Ferriscan. There was only one observation (0.6%) for which chelator treatment would be recommended by Ferriscan and not T2*. Thus, treatment decisions based solely on T2* would be more conservative (in favor of more aggressive chelation) than those based on Ferriscan.

Figure 3.

Bland–Altman Plot comparing LIC assessment by T2* and Ferriscan methods: as described in the Results section, Ferriscan is assigned as the “standard” method. Horizontal lines represent the mean bias and 95% confidence interval for the difference between ferriscan and T2*. Note the positive bias of T2* at low LIC and negative bias of T2* at very high LIC.

Discussion

In patients with TM, the last several years have brought major changes to both assessment and treatment of iron overload. In this single center, retrospective analysis, we demonstrated that the introduction of routine MRI assessments of liver and cardiac iron concentration, together with the introduction of oral chelators, has increased the proportion of thalassemia patients with liver and cardiac iron concentration within the target range. This observation is consistent with previous reports, where the introduction of MRI assessment of cardiac iron levels helped to identify patients at highest risk and intensified iron chelation treatment [12,13].

In our experience, younger patients are more likely to be compliant with chelator use as they are given their medication by a parent/caregiver. This may lead to differing rates of decline in liver and cardiac iron burden over time (Supplemental Fig. 1).

In this study, two patients died of heart failure due to iron overload; both had taken DFP before their deaths, one briefly and one for years. These patients had poor compliance with their chelators which can explain their severe cardiac siderosis and death [14]. Their markedly abnormal cT2* values prior to death are captured in our data (as two of the three outliers in Fig. 2a and as patients #5 and #40 in Supplemental Fig. 2c).

We note that in addition to these two subjects, 12% of patients in our cohort were within the target range for LIC (<7 mg/g dry weight) but with abnormal cT2* (<20 msec). We noted this discrepancy previously in a cross-sectional analysis of our population [15], and we attribute the discordance to the complicated pharmacodynamics of iron chelation therapy; it is much easier to rapidly unload iron from the liver than the heart, so that for severely overloaded patients treated with aggressive chelation, the liver values may come into range while the heart remains dangerously overloaded [16–18]. The treatment strategies for patients with the clinical scenario of low LIC (sometimes below target range) and cardiac iron overload are evolving with the availability of DFP as a commercial choice in the United States since December 2011. This data do not have enough longitudinal follow-up after this launch date to comment on efficacy.

Adherence to prescribed chelation regimens is one of the key determinants of success of chelation, and ultimately of survival in transfusion-dependent anemias. Other key determinants include transfusional iron burden and biological responsiveness to a given chelation regimen [19].

Although the two MRI liver iron assessment methods, that is, Ferriscan and T2*/R2*, are generally concordant, neither has emerged as a true “gold standard,” and the prior standard, liver biopsy, has faded almost entirely from use. Since April 2006, our center’s clinicians have used the mean of LIC by the two methods ([LIC ferriscan + LIC T2*]/2) for clinical decision making in thalassemia. As seen from the Bland–Altman analysis, across the population there is a modest bias toward higher values of estimated liver iron with T2* when the degree of iron overload is low, and a reverse bias, of poor sensitivity of T2* to very high liver iron burdens. But the degree of bias varies from individual to individual, and the mean values are a reasonable approach until additional analysis of larger datasets can allow conclusions about whether either method could be dispensable. While neither method is a “gold standard” in terms of accuracy, Ferriscan is valid across a wider range of LIC, and includes an internal control and standardization across a wide range of centers, countries, and MRI instruments. T2* methods for LIC determination require less time in the MRI scanner, and do not have a specific fee for proprietary analysis, in comparison to Ferriscan.

In our population, changes in dose were much more common than changes in chelator, as one might expect (221 vs. 76). Considering only the latter, changes from one chelator to another monotherapy or to combination therapy due to iron status alone accounted for 1/3 of all chelator changes (25/76). Of the remaining two thirds of changes in chelation regimen, the majority were changed when oral chelators became available, though a small fraction (10.5%) were due to side effects.

The reasons that an MRI out of range might not directly trigger a change in chelation strategy [dose or chelators] (defined for the purposes of this study as a change within 3 months after an MRI) could be several: (a) if a patient was on a trend from undesirable values toward the target, but had not achieved the target, no change would be made; (b) a chelator increase might have been initiated based on ferritin in the interval between the penultimate MRI and the most recent (so that by definition, we could not attribute the change to the last MRI); (c) a patient might have been nonadherent (as judged by self report, low prescription refills or worsening iron status on a stable dose), and in this case, dosage increase was not indicated so much as attention to adherence; (d) some patients had achieved maximum tolerated doses of DFX, and other measures were attempted to improve iron status (including splenectomy and/or changes in transfusion targets); and (e) ferritins are measured monthly in our program compared to annual MRI, so that trends may be revealed between annual MRIs (though ferritin can be much less accurate in patients with Hepatitis C).

Noetzli et al. studied chelation over time by MRI of heart and liver and noted distinct patterns of unloading or loading liver faster than heart, expressed as graphical “areas” [16]. We did not observe a similar phenomenon in our present cohort. The reason for this difference is not certain. (See Supplemental Fig. 2).

In our patients, we demonstrated that T2* estimates of LIC have significant negative bias over values of roughly 25 mg/g dry weight (Fig. 2). Andrade et al. mention that T2*-weighted sequences are susceptible to signal loss caused by the accumulation of iron in the tissues, which may explain the findings in our study [20]. The relative insensitivity of T2* hepatic iron is only noted at liver iron values that are clinically “much too high,” and trigger more aggressive therapy regardless of the numerical value.

Potential limitations of the study include (a) that it was retrospective and (b) included a single center with (c) a relatively small sample population. However, although the study was retrospective and purely observational, data ascertainment was relatively complete. Also, a potential benefit of this single-center study was the uniformity of the clinical approach over the follow-up period. This can be an advantage in particular compared to larger national or international multicenter datasets [6], in which variable approaches to measurement or patient care across centers and countries may obscure some of the beneficial changes over time. Thus, despite the small population, we demonstrated that the introduction of MRI methods to assess liver and cardiac iron concentration together with the introduction of oral chelators improved overall iron status, and increased the fraction of patients in the target iron range at our center.

Larger, prospective, multicenter studies, including the NHLBI-sponsored TLC, will be best suited to determining optimum chelation strategies in transfusional iron overload.