Pancreatic Iron Overload Is Associated With Early QTc Prolongation Before Myocardial Iron Deposition in Transfusion‐Dependent Thalassemia

ABSTRACT

Objectives

This cross‐sectional study evaluated the 12‐lead electrocardiogram (ECG) corrected QT (QTc) interval as an accessible marker for organ‐specific iron burden in adult patients with transfusion‐dependent thalassemia (TDT).

Methods

We considered 273 adult TDT patients (37.48 ± 8.75 years; 53.8% females), enrolled in the Extension‐Myocardial Iron Overload in Thalassemia Network. All patients underwent magnetic resonance imaging (MRI) for the quantification of hepatic, pancreatic, and myocardial iron ($T_2^{*}$ technique) and the assessment of biventricular size and function, alongside a standard ECG within 3 months.

Results

QTc interval showed a significant negative correlation with pancreatic $T_2^{*}$ (R = 0.220; p < 0.0001) and cardiac $T_2^{*}$ (R = −0.201; p = 0.001). A QTc > 425 ms predicted myocardial iron overload (MIO; $T_2^{*}$ < 20 ms). A significant stepwise prolongation of the QTc interval was observed across groups: no iron overload (405.45 ms), isolated pancreatic siderosis (421.76 ms), and combined pancreatic/myocardial siderosis (433.01 ms) (p = 0.001). In patients without MIO, a QTc > 416 ms predicted pancreatic siderosis ($T_2^{*}$ < 26 ms) and a multivariable regression analysis confirmed pancreatic iron, female sex, and splenectomy as independent predictors of QTc duration.

Conclusions

The QTc interval may serve as an early “metabolic sensor” of multi‐organ iron toxicity. A QTc > 416 ms is associated with pancreatic siderosis and possible cardiac risk before MRI detects cardiac iron. While QTc alone is unlikely to be sufficient for individual patient risk stratification, this inexpensive and widely available parameter may help inform decisions regarding advanced imaging.

1. Introduction

Lifelong transfusion therapy has dramatically improved survival in patients with transfusion‐dependent thalassemia (TDT) [1]. Nevertheless, chronic transfusional iron loading remains an unavoidable consequence of treatment because humans lack a physiological mechanism for active iron excretion [2]. Progressive iron accumulation within the liver, myocardium, and endocrine organs promotes reactive oxygen species (ROS) generation, triggering oxidative stress, lipid peroxidation, and cellular damage that ultimately culminate in multi‐organ dysfunction [3].

The clinical management of iron overload in TDT has been substantially improved by two key factors: the availability of iron chelating agents, which promote the removal of excess systemic iron and help maintain a negative iron balance, and the introduction of the $T_2^{*}$ magnetic resonance imaging (MRI) technique [4, 5]. As a non‐invasive “virtual biopsy”, $T_2^{*}$ MRI enables accurate and reproducible quantification of organ‐specific iron deposition, allowing early detection of tissue involvement and individualized tailoring of chelation strategies [6, 7, 8, 9]. In particular, MRI‐guided intensification of chelation therapy in response to myocardial iron overload (MIO) has markedly reduced the incidence of heart failure and cardiac mortality in TDT patients [4, 10, 11].

Despite the transformative impact of $T_2^{*}$ MRI, its global implementation remains hindered by significant practical and economic barriers. The requirement for costly infrastructure and specialized expertise makes this technology sparse in resource‐limited settings, where high costs further restrict its use for essential longitudinal monitoring [12, 13]. These limitations have stimulated interest in more accessible and cost‐effective tools for the early identification of cardiac iron toxicity. Among these, the 12‐lead electrocardiogram (ECG) represents a widely available and inexpensive method for cardiac assessment. In particular, abnormalities of ventricular repolarization, especially prolongation of the corrected QT (QTc) interval, have been associated with myocardial iron accumulation [14, 15, 16, 17]. However, currently available evidence is derived almost exclusively from pediatric cohorts and has produced inconsistent diagnostic thresholds. Zhou et al. identified a QTc cut‐off of 418.5 ms for predicting cardiac iron deposition in children, with high sensitivity but limited specificity [15], whereas Aggarwal et al. reported a substantially poorer sensitive higher threshold of 450 ms in a smaller pediatric cohort [16]. Consequently, the applicability of these findings to adult TDT patients, who experience prolonged cumulative iron exposure and a higher burden of endocrine and metabolic complications, remains uncertain.

Importantly, iron toxicity does not develop simultaneously across organs. Pancreatic iron accumulation has been shown to precede myocardial deposition, potentially providing an early temporal window into future cardiac risk [18, 19, 20, 21, 22]. Moreover, the pathophysiological link between the pancreas and the heart seems to be more profound, with pancreatic siderosis emerging as a powerful, distinct driver of cardiac arrhythmias [22, 23].

The present cross‐sectional study aimed to identify a QTc threshold associated with myocardial iron overload in adult TDT patients and to investigate the relationship between pancreatic iron deposition and QTc prolongation. We hypothesized that QTc prolongation may represent an early electrocardiographic marker of iron‐related metabolic and electrical dysfunction, potentially preceding overt MRI‐detectable myocardial iron accumulation.

2. Methods

2.1. Study Population

This observational, cross‐sectional study included 273 adult (age ≥ 18 years) TDT patients, retrospectively recruited from the Extension–Myocardial Iron Overload in Thalassemia (E‐MIOT) project. The E‐MIOT project is a nationwide Italian collaborative network involving 67 specialized thalassemia treatment centers and 15 MRI facilities, where all examinations are performed using highly standardized and rigorously validated protocols. The network is supported by a centralized database collecting comprehensive clinical, laboratory, and imaging data.

All patients had been receiving regular blood transfusions every 2–4 weeks since early childhood and were undergoing iron chelation therapy.

Patients with known pre‐existing cardiac arrhythmias, including atrial fibrillation, frequent ventricular ectopic beats, or advanced atrioventricular block, as well as patients receiving medications known to prolong the QT interval (e.g., class I or class III antiarrhythmic agents), were excluded from the study.

The study was conducted in accordance with the principles of the Declaration of Helsinki and was approved by the ethics committees of all participating MRI centers. Written informed consent was obtained from all participants.

2.2. Laboratory and Clinical Evaluations

Biochemical analyses were conducted at each participating thalassemia center using standard clinical chemistry analyzers and harmonized procedures for sample collection and processing. Hemoglobin and serum ferritin were assessed at least six times annually in all patients, and the corresponding mean values were computed for each parameter.

Diabetes mellitus was diagnosed in the presence of a fasting plasma glucose concentration ≥ 126 mg/dL, a 2‐h plasma glucose concentration ≥ 200 mg/dL following an oral glucose tolerance test (OGTT), or a random plasma glucose concentration ≥ 200 mg/dL accompanied by typical symptoms of hyperglycemia [24].

2.3. Electrocardiography and QTc Calculation

A standard resting 12‐lead electrocardiogram (ECG) was obtained within 3 months of the MRI examination. QTc interval values were extracted from the centralized E‐MIOT database and corresponded to the measurements routinely reported by the cardiologists of the participating thalassemia centers, based on automated analyses generated by the local ECG software systems.

In accordance with routine clinical practice and previous electrocardiographic studies in thalassemia populations [16, 25], QTc values were calculated using Bazett's formula, defined as the measured QT interval divided by the square root of the preceding R–R interval [26].

2.4. MRI

MRI examinations were conducted using clinical 1.5 T scanners (GE Healthcare, Milwaukee, WI, United States of America; Philips Healthcare, Best, The Netherlands; Siemens Healthineers, Erlangen, Germany), equipped with phased‐array surface receiver coils. Image acquisition was performed during breath‐holding at end‐expiration with ECG gating.

For iron overload assessment, a mid‐axial hepatic slice [27], a minimum of five axial slices encompassing the entire pancreas [28], and short‐axis left ventricular (LV) slices at basal, mid‐ventricular, and apical levels [29] were acquired using multi‐echo gradient‐echo $T_2^{*}$ sequences with 10 echo times and an echo spacing of 2.26 ms. Image analysis was carried out by radiologists with more than 10 years of experience using the validated custom software HIPPO MIOT. Hepatic $T_2^{*}$ values were quantified in a standard‐size circular region of interest (ROI) [27] and were converted into liver iron concentration (LIC) using Wood's calibration curve [30]. For pancreatic evaluation, $T_2^{*}$ measurements were obtained from three small ROIs placed in the head, body, and tail of the gland, while carefully avoiding major vessels, ducts, and regions affected by susceptibility artifacts from gastrointestinal gas [31]. The global pancreatic $T_2^{*}$ value was calculated as the average of the three regional measurements, with 26 ms adopted as the lower limit of normality [28]. The LV was segmented according to the standardized 16‐segment model proposed by the American Heart Association/American College of Cardiology [32]. $T_2^{*}$ values were measured for each segment [29], and a dedicated correction map was applied to minimize the effects of susceptibility artifacts [33]. The global heart $T_2^{*}$ value was calculated as the average of all segmental measurements. A threshold of 20 ms was adopted as the conservative lower limit of normal for both segmental and global myocardial $T_2^{*}$ measurements [34].

To assess biventricular size and function, steady‐state free precession (SSFP) cine images were acquired in contiguous 8‐mm short‐axis slices from the atrioventricular ring to the apex [35]. Image analysis was performed using commercially available softwares (MASS, Medis, Leiden, The Netherlands; or cvi42, Circle Cardiovascular Imaging Inc., Calgary, Canada) by manually delineating endocardial and epicardial borders at end‐diastole and end‐systole. Biventricular volumes and left ventricular mass were indexed to body surface area.

2.5. Statistical Analysis

Data analysis was performed using SPSS version 27.0 (IBM Corp., Armonk, NY, USA) and MedCalc version 19.8 (MedCalc Software Ltd., Ostend, Belgium).

Continuous data are presented as mean ± standard deviation (SD), while categorical variables are reported as frequencies and percentages.

The Kolmogorov–Smirnov test was employed to verify the normality of data distribution.

Bivariate associations were explored using Pearson's or Spearman's coefficients as dictated by data normality.

For group comparisons, the independent‐samples t‐test or one‐way ANOVA were used for normally distributed data, whereas the Mann–Whitney U or Kruskal–Wallis tests were applied for non‐parametric variables. The chi‐square test was used for categorical data comparisons. Where multiple comparisons were performed, Bonferroni post hoc correction was utilized to adjust for Type I error.

The impact of potential confounders on group differences was evaluated via analysis of covariance (ANCOVA). Covariates were included in the model if they exhibited significant inter‐group differences and were linked to the primary outcome. Logarithmic transformation was applied when necessary to satisfy the assumptions of residual normality and variance homogeneity.

Diagnostic performance was quantified through receiver operating characteristic (ROC) curve analysis. The area under the curve (AUC) was calculated with 95% confidence intervals (CI). Optimal diagnostic thresholds were determined using the Youden index, with subsequent calculation of sensitivity and specificity.

Determinants of the QTc interval were identified using univariable and stepwise multivariable linear regression analyses. Variables demonstrating a significant association in the univariable models were subsequently entered into the multivariable analysis. To ensure the stability of the final model, potential multicollinearity among independent predictors was rigorously assessed; variables were scrutinized based on the variance inflation factor (VIF) and tolerance statistics, with values exceeding 5 or falling below 0.20, respectively, indicating excessive collinearity.

Statistical significance for all two‐sided tests was defined as p < 0.05.

3. Results

3.1. Characteristics of TDT Patients

The study population had a mean age of 37.48 ± 8.75 years (range: 18–58 years) and included 53.8% females. All demographic, clinical, and instrumental characteristics are summarized in Table 1.

TABLE 1.

Demographic, clinical, and instrumental data of TDT patients.

Variable	TDT patients (N = 273)
Females, N (%)	147 (53.8)
Age (years)	37.48 ± 8.75
Transfusion starting age (months)	13.89 ± 16.93
Chelation starting age (years)	4.08 ± 4.41
Splenectomy, N (%)	177 (64.8)
Pre‐transfusion hemoglobin (g/dL)	9.59 ± 0.47
Serum ferritin (ng/mL)	891.41 ± 965.58
Diabetes, N (%)	41 (15.0)
Corrected QT interval (ms)	420.97 ± 31.08
MRI LIC (mg/g dw)	5.82 ± 8.83
Hepatic iron overload, N (%)	137 (50.2)
Global pancreas $T_2^{*}$ (ms)	11.71 ± 9.25
Pancreatic iron overload, N (%)	244 (89.4)
Global heart $T_2^{*}$ (ms)	37.34 ± 9.18
Myocardial iron overload, N (%)	23 (8.4)
Number of segments with $T_2^{*}$ < 20 ms	1.47 ± 4.17
LV end‐diastolic volume index (mL/m2)	83.96 ± 16.55
LV mass index (g/m2)	54.49 ± 13.69
LV ejection fraction (%)	62.96 ± 6.69
RV end‐diastolic volume index (mL/m2)	81.63 ± 16.74
RV ejection fraction (%)	61.14 ± 8.17

Abbreviations: LIC, liver iron concentration; LV, left ventricular; MRI, magnetic resonance imaging; N, number; RV, right ventricular; TDT, transfusion‐dependent thalassemia.

Pancreatic and cardiac iron overload were present in 89.4% (N = 244) and 8.4% (N = 23) of patients, respectively. The mean QTc interval was 420.97 ± 31.08 ms.

3.2. Demographic, Clinical, and MRI Correlates of QTc Interval

QTc interval was significantly higher in females than in males (426.22 ± 28.16 vs. 414.85 ± 33.24 ms; p = 0.005) and exhibited a weak but significant positive association with age (R = 0.137; p = 0.024).

Compared to non‐splenectomized patients, splenectomized patients showed a significantly increased QTc interval (424.14 ± 28.87 vs. 415.15 ± 34.17 ms; p = 0.036). Age significantly differed between splenectomized and non‐splenectomized patients (38.89 ± 8.48 vs. 34.88 ± 8.68 years; p < 0.0001) and was therefore used as a covariate in the ANCOVA. After the correction for age, the difference in QTc interval was at the limits of statistical significance (p = 0.056).

No significant correlations were observed between QTc interval and mean levels of hemoglobin (R = 0.048; p = 0.431) or ferritin (R = 0.022; p = 0.716).

Patients with diabetes mellitus exhibited significantly higher QTc values compared to non‐diabetic subjects (432.39 ± 33.22 vs. 418.96 ± 30.31 ms; p = 0.012). This difference remained significant after adjusting for age (p = 0.036).

QTc interval was not associated with MRI LIC (R = 0.05; p = 0.929), but showed a significant negative correlation with both pancreatic $T_2^{*}$ (R = −0.220; p < 0.0001) and cardiac $T_2^{*}$ (R = −0.201; p = 0.001). A significant correlation was found between QTc interval and number of segments with $T_2^{*}$ < 20 ms (R = 0.228; p < 0.0001).

QTc interval was not associated with LV end‐diastolic volume index (R = −0.117; p = 0.053), LV mass index (R = −0.039; p = 0.519), LV ejection fraction (R = 0.037; p = 0.544), right ventricular (RV) end‐diastolic volume index (R = −0.153; p = 0.071), and RV ejection fraction (R = −0.066; p = 0.278).

3.3. Progressive Impact of Iron Toxicity on QTc

The study population was stratified into three distinct groups based on combined cardiac and pancreatic iron status: Group A (N = 29; 10.6%) with no iron overload in either the pancreas or the heart, Group B (N = 221; 81%) with no cardiac iron overload but pancreatic iron overload, and Group C (N = 23; 8.4%) with iron overload in both organs.

The proportion of females was comparable among the three groups (48.3% vs. 53.8% vs. 60.9%; p = 0.664). Mean age was 31.32 ± 8.24 years in Group A, 38.52 ± 8.49 years in Group B, and 35.34 ± 8.59 years in Group C, with a significant difference between Groups A and B (p < 0.0001).

A significant stepwise prolongation of the QTc interval was observed across the three groups (Group A: 405.45 ± 36.51 ms; Group B: 421.76 ± 28.73 ms; Group C: 433.01 ± 38.90 ms; p = 0.001), which remained significant after adjustment for age (p = 0.010). In post hoc analysis, both Group B and Group C exhibited significantly longer QTc intervals compared with Group A (p = 0.021 and p = 0.006, respectively), whereas the difference between Group B and Group C did not reach statistical significance (p = 0.078) (Figure 1).

FIGURE 1.

QTc interval progression according to organ‐specific iron status. QTc interval is presented as mean ± standard deviation. A significant stepwise increase in QTc was observed across iron overload stages, from Group A (no iron overload) to Group B (isolated pancreatic siderosis) and Group C (combined pancreatic and myocardial siderosis). The blue horizontal bar indicates a statistically significant difference between two groups.

3.4. Dual‐Step Diagnostic QTc Thresholds

In the first step, ROC curve analysis on the entire cohort identified an optimal QTc cut‐off of > 425 ms for the prediction of MIO. This threshold yielded a sensitivity of 73.9% and a specificity of 60.0%, with an AUC of 0.65 (95% CI: 0.59–0.71; p = 0.019).

In the second step, the analysis was strictly confined to the 250 patients with a global heart $T_2^{*}$ > 20 ms. ROC analysis identified a distinct, earlier QTc cut‐off of > 416 ms for the prediction of pancreatic siderosis. This threshold yielded a sensitivity of 60.6%, a specificity of 68.9%, and an AUC of 0.66 (95% CI: 0.59–0.71; p = 0.005) (Figure 2).

FIGURE 2.

Receiver operating characteristics (ROC) curve illustrating the diagnostic performance of the QTc interval for identifying pancreatic iron overload ($T_2^{*}$ < 26 ms) in patients without myocardial iron overload.

3.5. Independent Predictor of QTc Prolongation

To confirm the independent role of pancreatic toxicity, a linear regression model was built focusing solely on patients without myocardial iron (Table 2). In the univariable models, female sex, age, splenectomy, diabetes, global pancreas $T_2^{*}$ values, and biventricular end‐diastolic volume indexes were significantly associated with QTc interval. In multivariable analysis, the independent predictors of increased QTc were female sex, splenectomy, and increased pancreatic iron levels (F = 10.45; p < 0.0001). No variable was excluded from the multivariable models due to excessive collinearity.

TABLE 2.

Determinants of QTc interval identified by univariable and multivariable linear regression analyses in patients with normal global heart $T_2^{*}$ values.

	Univariable regression		Multivariable regression
	β	p	β	p
Female sex	0.173	0.006	0.195	0.002
Age	0.181	0.004
Splenectomy	0.163	0.010	0.167	0.008
Pre‐transfusion hemoglobin	0.068	0.289
Serum ferritin	−0.055	0.387
Diabetes	0.155	0.014
MRI LIC	−0.025	0.699
Global pancreas $T_2^{*}$	−0.247	< 0.0001	−0.210	0.001
LV end‐diastolic volume index	−0.146	0.021
LV mass index	−0.070	0.273
LV ejection fraction	0.127	0.055
RV end‐diastolic volume index	−0.153	0.015
RV ejection fraction	−0.057	0.366

Abbreviations: LIC, liver iron concentration; LV, left ventricular; MRI, magnetic resonance imaging; RV, right ventricular.

4. Discussion

This multicenter study is framed within the ongoing effort to identify simple and widely accessible markers capable of detecting early cardiac iron toxicity in adult TDT patients. Our findings suggest that the standard 12‐lead ECG, through the analysis of the QTc interval, may serve as an early indicator associated with systemic iron toxicity and organ‐specific iron burden.

We found an association between QTc prolongation and myocardial iron burden in adults with TDT, extending previous observations in pediatric and very young cohorts [14, 15, 16, 17, 36] to an older population. The prolongation of the QTc interval in the setting of MIO is a complex, multifactorial process driven by a cascade of direct toxic effects. Free non‐transferrin‐bound iron (NTBI), the principal circulating labile iron species driving extra‐hepatic toxicity in transfusional iron overload, enters cardiomyocytes via L‐type calcium channels (LTCC) [37], where it may interfere with ventricular repolarization through modulation of key ion currents, particularly delayed rectifier potassium and calcium channels [36]. Concurrently, intracellular iron, through the promotion of oxidative stress and lipid peroxidation, may activate profibrotic signaling pathways, including transforming growth factor‐beta signaling [38]. The resulting structural remodeling and myocardial fibrosis contribute to electrical heterogeneity, slowed conduction, and further QTc prolongation [39]. We identified a QTc cut‐off > 425 ms for predicting MIO, with moderate diagnostic performance. However, by the time this threshold is reached, significant iron accumulation is often already present, limiting its role in early detection.

The most innovative aspect of our research is the identification of a significant association between reduced pancreatic $T_2^{*}$ and QTc prolongation, even in the presence of normal cardiac iron levels. This finding supports the role of the pancreas as a sentinel organ of systemic iron exposure, identifying a potential “temporal window” during which early cardiac electrical alterations precede the detection of myocardial iron on MRI. Clinically, this is evidenced by the significant stepwise prolongation of the QTc interval observed across our study groups, where patients with isolated pancreatic siderosis already exhibited significantly longer intervals compared to those without any iron overload. This relationship is largely driven by shared exposure to circulating NTBI, although tissue‐specific uptake mechanisms and kinetics modulate organ susceptibility, resulting in different patterns of iron handling in the pancreas and the heart. Pancreatic iron accumulation appears particularly sensitive to sustained NTBI exposure, largely driven by DMT1 and ZIP14‐mediated transport. In contrast, myocardial iron uptake is more closely associated with L‐type calcium channel activity, which is repeatedly engaged during each cardiac contraction [19], although other transport pathways may also contribute. Because the pancreas accumulates iron earlier and more rapidly than the myocardium [19, 22], a reduced pancreatic $T_2^{*}$ serves as an early indicator of an expanded and toxic labile iron pool. Furthermore, pancreatic siderosis drives progressive beta‐cell dysfunction and the development of diabetes mellitus [22, 40], which is associated with significantly longer QTc intervals in our cohort. Diabetes may contribute to this electrical instability through autonomic neuropathy, altered sympathovagal balance, and increased oxidative stress, all of which can impair cardiac ion channel function, particularly delayed rectifier potassium currents [41]. In this context, we identified a lower QTc threshold of > 416 ms as a sensitive marker of pancreatic siderosis in patients with preserved myocardial iron levels. From a clinical perspective, the identified threshold may represent an early warning marker of pancreatic involvement and impending cardiac risk, potentially enabling earlier risk stratification, closer surveillance, and timely intensification of chelation therapy before the development of overt MIO.

Of note, the younger age of the patients free of both pancreatic and cardiac iron loading likely reflects a profound cohort effect. These individuals have benefited from early, lifelong exposure to optimized and flexible chelation strategies from the very onset of their transfusion history, coupled with timely, longitudinal $T_2^{*}$ MRI tracking. Conversely, older patients in our cohort likely spent their early childhood under less systematic monitoring and older therapeutic protocols, resulting in a higher cumulative iron burden over time.

Our multivariable analysis further identified female sex and splenectomy as independent predictors of QTc prolongation, alongside pancreatic iron burden. Although sex‐related differences in QTc duration are well established in the general population [42, 43], this is, to our knowledge, the first study demonstrating their persistence in adult TDT patients. The association with splenectomy likely stems from the loss of the spleen as a protective reservoir for non‐toxic iron storage, exposing patients to increased extra‐hepatic deposition [22, 44]. Furthermore, experimental studies have suggested that splenectomy may alter autonomic neural control, contributing to electrophysiological abnormalities [45].

Although anemia may influence myocardial depolarization and repolarization through sympathetic activation, chronic volume overload, neurohormonal activation, and oxidative stress [46], we found no association between QTc interval and hemoglobin levels, likely due to the tightly controlled pre‐transfusion hemoglobin values and low inter‐individual variability in our well‐treated TDT cohort.

Surprisingly, in patients without MIO, QTc duration was inversely associated with biventricular end‐diastolic volumes in the univariable analysis, suggesting that QTc prolongation reflects early iron‐related injury. In the early stages of MIO, increased myocardial and vascular stiffness can reduce ventricular volumes [47], whereas progressive iron accumulation can later lead to ventricular dilation and systolic dysfunction [48, 49].

As a potential avenue for future research, it would be of interest to evaluate whether QTc intervals change following acute chelation therapy. Paired ECG recordings obtained before and approximately 2–3 h after administration of a chelating agent, together with concurrent measurements of transferrin saturation as a surrogate of circulating NTBI, may help clarify whether circulating labile iron has direct and reversible effects on cardiac electrophysiology independent of changes in tissue iron burden.

4.1. Limitations

The cross‐sectional design precludes the establishment of a definitive longitudinal causal relationship.

Although Bazett's formula may overestimate QTc at high heart rates and underestimate it at low heart rates, it remains the most widely used correction method in clinical practice [50] and thalassemia studies [16], facilitating comparison with previous literature. While the use of routine multicenter ECG data without centralized reanalysis may represent another limitation, standardized automated QTc measurements likely reduced inter‐observer variability and improved reproducibility across this large cohort.

Other metabolic parameters influencing repolarization, such as serum calcium and magnesium levels, were not included in the multivariable analysis.

5. Conclusions

A QTc interval > 416 ms may represent a marker of systemic metabolic distress and possible pancreatic involvement. In settings where MRI access is limited, this simple parameter may help identify patients who may benefit from further evaluation with $T_2^{*}$ MRI and closer longitudinal monitoring.

Author Contributions

A.M. designed the study, analysed the data, and drafted the initial manuscript. L.P. was responsible for data collection. G.B.R., F.L., R.R., V.R., and G.M. collected the clinical data. G.P., M.Z., N.V., P.F., and A.B. analysed the images. V.P. developed the software for $T_2^{*}$ image analysis. A.B. supervised the study and is the guarantor of this work. All authors assisted with interpretation, commented on drafts of the manuscript, and approved the final version.

Funding

The E‐MIOT project received “no‐profit support” from industrial sponsorships (Chiesi Farmaceutici S.p.A. and Bayer). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Conflicts of Interest

The authors declare no conflicts of interest.

Data Availability Statement

The data underlying this article cannot be shared publicly due to privacy reasons. The data will be shared on reasonable request to the corresponding author.