Abrupt ferritin increases as a marker of cancer in transfusion-dependent thalassemia

Current therapeutic strategies in the treatment of transfusion-dependent β-thalassemia (TDT) have resulted in significant improvement in life expectancy. Along with increased survival, cancer became a leading cause of morbidity and mortality.^1-4 Neoplasia is generally associated with increased ferritin levels, which has been attributed to the role of ferritin as a reactive protein and to the dysregulation of iron metabolism during carcinogenesis.^5,6 On the other hand, ferritin levels have been used for longitudinal evaluation of iron overload (IO) in TDT patients.⁷ The aim of this retrospective cohort study is to evaluate ferritin changes in TDT patients with IO during the occurrence of a neoplastic event.

This study has been approved by the Hospital Ethical Committee according to national guidelines for retrospective (observational) clinical studies. All TDT patients followed in our unit during the study period (1^st January 2013 to 31^st December 2024) were included. Our thalassemia unit is the largest in Greece, with the number of patients attending the unit corresponding approximately to 15% of the total TDT population nationwide. Demographic-epidemiological data, and clinical course of our TDT patients in terms of age distribution, disease complications, survival and morbidity causes, are representative of the reference population. All patients are regularly transfused and receive iron chelation therapy with different schemas which include any of the available iron chelators either in monotherapy or combination, according to the Thalassemia International Federation (TIF) guidelines.

For TDT patients who developed cancer (c-TDT), data on the type of cancer, age at diagnosis, survival status, and IO indices (serum ferritin, liver iron concentration [LIC] and cardiac iron concentration [cardiac T2*] assessed by MRI) were recorded and analyzed. The mean (Standard Deviation) ferritin levels were assessed for the following time intervals prior to cancer diagnosis: 0-6 months, 12-24 months, and 1-3 years. All 21 patients who developed cancer were regularly transfused, with no significant differences regarding transfusion burden, the year prior to cancer diagnosis and were under a stable chelation schema prior to the abrupt ferritin elevation. For TDT patients who did not develop cancer (nc-TDT), longitudinal ferritin levels for the last three years prior to last observation were captured as above. ncTDT patients for whom ferritin and LIC values were available for the last three years were used as a control group. Other causes of chronic inflammation were excluded using different diagnostic parameters (C-reactive protein, transaminases, autoimmune markers) as per standard of care.

All analyses were performed using RStudio version 3.6.2. (Integrated Development Environment for R. Posit Software, PBC, Boston, MA, USA). Comparisons were performed, using t test for independent samples and paired t test and Mann-Whitney U test for continuous variables, and χ² and Fisher’s exact test for categorical variables. Receiver Operating Characteristics (ROC) curve analysis was completed to evaluate sensitivity and specificity of ferritin increase as a screening tool for malignancy, while Area Under the Curve (AUC) values were estimated with 95% Confidence Intervals. Longitudinal trajectories of ferritin over time were evaluated with a linear mixed effects model accounting for confounders. Stratified sensitivity analyses were performed based on baseline ferritin levels and, when necessary, Holm correction for multiple comparisons was applied. Complete data were available for all variables of interest. A two-tailed P<0.05 was considered statistically significant.

A total of 376 TDT patients (192 males, 184 females) were followed in our unit during the study period, with a median age of 44 years (range 1-66) as of December 2024. During the study period, 21 patients (10 males, 11 females) were diagnosed with cancer at a median age of 45 years (Table 1), which corresponds to a cancer prevalence of 5.6%. For the reference Greek population, cancer prevalence is estimated at 0.5%, while median age at cancer diagnosis is 66 years.⁸ Eight distinct types of cancer were diagnosed, with hepatocellular carcinoma (HCC) being the most frequent. Ten patients (2.7% of our cohort) developed HCC with none of them having active hepatitis C or B. This corresponds to a significantly higher prevalence compared to the general population, which is estimated at 0.0073%.⁹ Gender distribution (4 males, 6 females) also differs from that of the general population, in which HCC is more common in males.⁹ All c-TDT patients were regularly transfused at a fixed rate, and all were under chelation treatment at a stable dose prior to ferritin elevation. When ferritin increase was noted, intensification of chelation therapy was applied in 12/21 patients without affecting serum ferritin levels.

Ferritin levels, while they were no different between one year and three years prior to cancer diagnosis (P=0.13), rose sharply for 0-6 months prior, with a mean increase of 933±1,728 (129% of baseline) to all patients except 2 (Figure 1). There was a significant difference in ferritin levels 0-6 months prior to cancer diagnosis (median 2,116 ng/mL, range 412-9,582 ng/mL) compared to levels both at 12-24 months (median 1,180 ng/mL, range 244.3-5,950 ng/mL, P=0.00014) and 1-3 years (median 707 ng/mL, range 128-4,900 ng/mL, P=0.009) prior to cancer diagnosis. In the same period, other indicators of IO (LIC and cardiac T2*) have remained stable (P>0.05) (mean change 0.39±4.09 mg Fe/g dry weight [gdw] and 0.76±6.48 msec, respectively). There was no statistical difference in ferritin for patients developing HCC compared to other malignancies (P=0.347). However, this comparison is underpowered due to the small sample size (power of two sample t test 0.144). Interestingly, a female patient presented a notable increase in ferritin levels 0-6 months prior to the diagnosis of HCC, which resolved when remission was achieved, with a similar increase in ferritin levels occurring four years later upon relapse.

Two patients (Patients 1 and 16) did not experience these changes in ferritin kinetics. Patient 1 had thyroid carcinoma with low IO (ferritin levels 230-454 ng/mL, LIC 1.1mg Fe/gdw) and presented a 10% increase in chelation treatment dose. Patient 16 was diagnosed with HCC with mild IO (ferritin levels 2,000-2,750 ng/mL, LIC 4.4 mg Fe/gdw) and had hepatic lesions under a 2-year close evaluation prior HCC diagnosis. Different explanations for the different kinetics in ferritin levels in these 2 patients are hypothesized; ferritin elevation may not be observed in cases of low-grade neoplasia with slow evolution, while the over 6-month extent of the period of malignant development prior to cancer diagnosis may create problems in establishing the baseline ferritin values. None of the patients had active hepatitis, although 8 patients had previously received treatment for HCV infection. There was no significant association between history of HCV exposure and development of HCC or other malignancies. Mean change in serum ferritin over time did not differ significantly between patients with positive versus negative HCV exposure (P=0.942).

The sensitivity and specificity for changes of serum ferritin as a prognostic marker for malignancy were calculated by evaluating data on serum ferritin, LIC and heart T2* for all 376 patients followed in our unit. Thirty-two out of these patients, none of whom had cancer, showed significant increase in iron load, as demonstrated by increased LIC >50%, and were excluded from the ROC analysis. The nc-TDT patients, that were included in the analysis, had stable iron load, as reflected by minimal variations in their ferritin levels and MRI-assessed cardiac and hepatic iron load.

Table 1.Characteristics of transfusion-dependent thalassemic patients with malignancy.

c-TDT patients were on a stable chelation schema prior to abrupt ferritin elevation. ROC analysis showed that the sensitivity and specificity of the changes in ferritin levels as a marker of malignancy were estimated at 90.48% and 100%, respectively, for an increase > 70.55% from baseline, and 96.9% and 85.7%, respectively, for an absolute increase > 442.4 ng/mL from baseline (Figure 2). Positive prognostic values were estimated at 100% for a relative ferritin increase > 70.55%, and at 64.29% for an absolute increase > 442.4 ng/ mL, while negative prognostic value is 99.38% and 99.05%, respectively.

Sensitivity stratified ROC analysis was based on baseline ferritin levels in 2 groups: 1) <1,000 ng/mL, and 2) >1,000 ng/mL; this analysis revealed similar results. Specifically, sensitivity and specificity of changes in ferritin levels as a marker of malignancy for the aforementioned 2 groups were estimated as follows: (i) 86% and 81%, respectively (cut off: 33.28%) for the <1,000 ng/mL baseline ferritin group, and (ii) 57.1% and 91.0%, respectively (cut off: 31.20%) for the >1,000 ng/mL baseline ferritin group, indicating that this marker performs best for the lower baseline ferritin group.

Figure 1.Trends in serum ferritin levels during the three years prior to the diagnosis of cancer in 21 transfusion-dependent thalassemic patients who developed cancer. (A) Individual trends. (B) Combined trend of ferritin level trajectories, assessed by a mixed effects model.

The limitations of this study include its retrospective nature and being from a single institution. On the other hand, as a single-institution cohort study, patients were treated uniformly by the same medical team, even though there is heterogeneity in transfusion and iron chelation schemas which are individualized according to the needs of the patients. The study population is limited; therefore, perspective investigation in larger cohorts is needed to confirm our findings. In conclusion, this study investigated the clinical significance of serum ferritin as a prognostic factor for cancer development in TDT patients. Sharp increases in ferritin levels were shown to be indicative of malignant processes, with a high specificity and sensitivity, even in the light of secondary

Figure 2.Receiver Operating Characteristics curves for the changes in ferritin levels as a marker of cancer in transfusion-dependent thalassemic patients. (A) Model of relative (%) changes compared to previous levels observed 1-3 year(s) prior to cancer diagnosis. (B) Model of absolute changes (in ng/mL) compared to previous levels observed 1-3 year(s) prior to cancer diagnosis. Sensitivity and specificity of the changes in ferritin levels as a marker of malignancy were estimated at 90.5% and 100%, respectively, for increase > 71% from baseline, and 96.9% and 85.7%, respectively, for absolute increase > 442 ng/mL from baseline.

IO. Thus, sustained unexplained ferritin increases in TDT patients should prompt timely evaluation for an underlying malignancy.

Footnotes

• Received June 10, 2025
• Accepted January 7, 2026

Correspondence

*EC and PD contributed equally as first authors

A. KATTAMIS - ankatt@med.uoa.gr

Disclosures

AK reports payment or honoraria, not related to this work, from Agios Pharmaceuticals, Bristol Myers Squibb, Novo Nordisk, Pfizer, and Vertex. All the other authors have no conflicts of interest to disclose.

Contributions