Characterization of Transfusion-Derived Iron Deposition in Childhood Cancer Survivors

With advances in anticancer therapy and supportive care, the combined 5-year survival rate for individuals diagnosed with cancer less than 20 years of age in the United States exceeds 80% (1). Unfortunately, a variety of treatment-related complications may cause or contribute to clinically significant chronic health problems resulting in diminished quality of life and premature death in many childhood cancer survivors (CCS).

One recently described late effect is iron overload (IO) resulting from multiple packed red blood cell (PRBC) transfusions administered to correct anemia, a common acute complication of cancer therapy. Multiple PRBC transfusions can result in accumulation of excess iron due to disruption of the homeostasis of iron storage and distribution when exogenous iron is loaded (2). Transfusional IO has been well documented among children with nononcologic conditions such as chronic hemolytic anemia (3). Although preliminary evidence for transfusional IO in pediatric oncology patients has been documented (4–18), its prevalence, organ distribution, and severity remain incompletely characterized. Historically, direct measurement of tissue iron has been limited almost exclusively to liver biopsy. More recently, magnetic resonance imaging (MRI) has emerged as an accurate, noninvasive means for measuring iron in multiple organ systems (19).

We undertook this study of an institutional CCS cohort across a spectrum of malignancies and treatment intensities to determine the prevalence, anatomical pattern, and severity of iron deposition in liver, pancreas, and heart by MRI. Our hypothesis was that increased organ iron deposition would be associated with higher cumulative PRBC transfusion volumes received during prior cancer treatment.

Participants were recruited to this cross-sectional study if they had been diagnosed with cancer before age 21 between June 1, 2004, and December 31, 2009, received all treatment and transfusions at Children's Hospital Los Angeles (CHLA), were able to read and write in English or Spanish, and were able to tolerate an MRI procedure lasting 45 to 60 minutes. Patients less than 7 years of age were excluded to avoid the need for sedation for MRI scans. Study procedures were approved by the CHLA human subject protection committee (Institutional Review Board). The CONSORT flow diagram for the study is shown in Fig. 1. The study population included 4 patients who underwent hematopoietic stem cell transplant (HSCT; Table 1).

MRI studies were performed on a Philips 1.5 Tesla Achieva magnet with 80 mT/m gradients and a four element torso array coil. Liver/pancreas R2* and heart T2*, indicators of tissue iron content, were assessed using multiecho gradient echo technique with echo times from 1 to 16 ms. Liver R2, a second metric of liver iron, was assessed using single-spin echo imaging with echo times from 3 to 30 ms. Custom Matlab routines (The MathWorks) were used to calculate R2 and R2* coefficients and convert them into iron values (20). In addition to the study radiologist's blinded interpretation of MRIs relative to study aims, a separate, blinded CHLA radiologist reviewed MRIs for evidence of noncardiac abnormalities. Blood samples were obtained on the same day as the MRI evaluation. Disease/treatment and demographic information was abstracted from medical records. Treatment intensity was graded by two experienced pediatric oncology clinicians (K.S. Ruccione and D.R. Freyer), using the validated Intensity of Treatment Rating Scale (ITR 3.0) to classify treatment as level 1 (least intensive), level 2 (moderately intensive), level 3 (very intensive), or level 4 (most intensive; ref. 21).

The primary outcome variable was quantification of liver iron content (LIC), and the primary exposure of interest was total and weight-adjusted cumulative PRBC volume. Medical and demographic variables selected as potential predictors or effect modifiers included sex, race/ethnicity, diagnosis, age at diagnosis and at study evaluation, history of tumor resection, and treatment intensity. Biomarkers related to iron status included serum ferritin, iron, and iron-binding capacity with the percentage of transferrin saturation. Two mutations of the hereditary hemochromatosis gene (HFE) were included as potential confounders based on their association with high iron absorption and progressive body IO (22).

MRI and laboratory measurements were treated as continuous variables in univariate and multivariate analyses. Log transformation of ferritin values was performed because of nonnormal distribution. Spearman rank correlation was used to assess the relationship between organ iron concentrations and iron biomarkers. Univariate and multivariate analyses were conducted to examine the effects of patient and clinical characteristics on LIC. Unique multivariate models were built starting with predictors with an F test P value of <0.15 on the univariate analysis. For the multivariate analysis, cumulative PRBC volume was retained whereas other candidate predictor variables with P < 0.15 via the likelihood ratio test (LRT) on univariate analysis were removed following a step-wise iterative process until the multivariate LRT P value fell below 0.05 for each of the remaining covariates. This approach was validated by assessing changes in significance through reintroduction of eliminated predictors while retaining adjusted cumulative PRBC volume in the model. Treatment intensity rating was not included in multivariate regression modeling because it was highly correlated with cumulative PRBC volume. All analyses were performed as two-sided tests with a significance level of 0.05. Statistical computations were performed using the statistical software package Stata (StataCorp. 2009; Stata Statistical Software: Release 11: StataCorp LP).

Study sample demographic and clinical characteristics are shown in Table 1. Most patients were Latino/Hispanic (n = 45, 60%), reflecting the racial/ethnic composition of the CHLA patient population. The median age at the time of study evaluation was 14 years (range, 8–25.6). Median follow-up times to study evaluation from treatment completion and most recent PRBC transfusion were 4.4 years (range, 0.2–7.6 years) and 4.9 years (range, 1.4–7.9 years), respectively. Cancer diagnoses included acute lymphoblastic leukemia (ALL) or acute myelogenous leukemia (AML; n = 33) and six types of solid tumors (n = 42). Of the cohort members with solid tumors, 33 (78.6%) had undergone tumor resection. Sixty-seven (89%) patients had received at least one PRBC transfusion while undergoing cancer treatment, with a median of 11 (range = 1–47) received per patient (data not shown). Treatment history included HSCT in 4 (5.3%) patients. Treatment intensity ratings, reflecting the purposeful construction of the sample to encompass a spectrum of treatment intensities, included levels 1 (least intensive) and 2 (moderately intensive; n = 39, 52%), level 3 (very intensive; n = 22, 29.3%), and level 4 (most intensive; n = 14, 18.7%). No study subjects had homozygous HFE C282Y mutations or C282Y/H63D compound mutations.

Median values for serum ferritin, iron, and iron-binding capacity with the percentage of transferrin saturation fell within the reference range, but the ranges were wide and exceeded the upper or lower ranges of normal in 36 subjects. These included elevated levels of ferritin in 21 of 74 (28.4%), iron in 2 of 71 (2.8%), iron-binding capacity in 3 of 75 (4%), and the percentage of transferrin saturation in 5 of 71 (7%). Results of MRI iron quantitation and calculated tissue iron concentration for liver, pancreas, and heart are shown in Table 2. The median liver R2* value in the study sample (n = 73) was 40.8 Hertz (range, 27.6–925.7). LIC was calculated using the equation LIC = R2* × 0.0254 + 0.2 (20). The study sample's median and range values for LIC were 1.2 and 0.9 to 23.7 mg/g, respectively. LIC was elevated above the reference range of 0.8 to 1.2 mg/g in 36 (49.3%) patients. Because no pancreatic iron calibration exists to convert R2* values into pancreatic iron concentration (23), pancreatic iron loading was estimated by R2* results alone, with a normal R2* being <30 Hertz. Pancreas R2* results (median, range in Hertz), for the study sample (n = 72) were 26 and 18 to 128. Although 19 of 72 (26.4%) study subjects had abnormal pancreatic R2* results, the degree of pancreatic loading was small in the majority of patients. One outlier had a pancreatic R2* value of 128; when this value was verified in the medical record, it was noted that this result was interpreted by the radiologist as moderate pancreatic siderosis (values of 30–100 Hertz constitute mild pancreatic siderosis, 100–400 Hertz moderate siderosis, and >400 Hertz severe siderosis; ref. 24). No cardiac T2* values were abnormal in the study sample (n = 74); the median cardiac T2* value was 32.8 ms (range, 24.3–40.5).

Statistically significant correlations for liver and pancreatic R2* were found (data not shown). Both hepatic and pancreatic R2* were positively correlated with serum ferritin, serum iron, the percentage of transferrin saturation, and weight-adjusted cumulative PRBC volume. Given our previous report involving this cohort (15), which described projected LIC based on transfusion volume received, projected LIC was compared with the LIC calculated from actual MRI liver R2* results and found to be significantly correlated (Spearman rank correlation r = 0.472, P < 0.0001; data not shown).

For univariate and multivariate analyses, LIC was used as the outcome variable inasmuch as pancreatic iron levels were minimally abnormal and cardiac iron levels were not elevated in this sample. In univariate analysis, variables significantly associated with increased LIC included older age at diagnosis (P = 0.01) and at MRI evaluation (P = 0.05), higher treatment intensity rating (P < 0.01), and having received higher cumulative PRBC volume (P < 0.0001). Having undergone HSCT was associated with increased LIC (P < 0.0001), but because there were only 4 patients in this group, this variable was not used in the multivariate analyses. The final multivariate regression model comprised weight-adjusted cumulative PRBC volume (P < 0.0001) and older age at diagnosis (P < 0.0001). Weight-adjusted cumulative PRBC volume was associated with a 0.03 mg/g increase in LIC for each mL/kg transfused. Age at diagnosis was associated with a 0.29 mg/g increase in LIC for each 1-year increase in age. The final model explained 52% of the variance in LIC among the sample of 73 CCSs who underwent liver MRI examination. Length of follow-up was added to the model to examine for change in model statistics. This resulted in a 0.02 increase in R2 and no change in the P value; therefore, length of follow-up was not retained in the final model.

To our knowledge, these study findings represent the first time the prevalence, anatomical distribution, and severity of organ iron deposition have been characterized in a relatively large single institution cohort of CCSs treated for a variety of malignancies over a range of therapeutic intensity and cumulative PRBC volume transfused. The finding of elevated calculated LIC in 49.3% of 73 study participants who had evaluable liver MRI examinations has potential clinical significance that merits further study due to the risks posed by excessive iron accumulation in hepatocytes, including hepatocellular injury that may eventually lead to the development of fibrosis and cirrhosis (25), and hepatocellular carcinoma (26, 27). Although the degree of pancreatic iron loading was relatively small in the subjects who had abnormal pancreatic R2* measurements, this finding also has potential clinical implications in that pancreatic beta cell damage due to IO may lead to glucose intolerance and diabetes mellitus, and hepatic disturbance of glucose utilization has been shown to accelerate pancreatic beta cell depletion due to hyperinsulinemia (2). Further research is needed to better understand the prospective risk of subsequent glucose dysregulation in individuals with increased pancreatic R2* values, what threshold levels of pancreatic iron loading might be critical in patients who also have excessive LIC, as well as the additive risk of liver/pancreatic damage that may arise from specific chemotherapeutic or other toxic exposures.

In this sample, we found no MRI evidence of cardiac iron loading. Although the absence of measurable cardiac iron is a pertinent negative finding that seems reassuring, caution is warranted as isolated case reports of cardiac iron loading have been published (10, 28) and the problem may have been under-represented in our cohort. Furthermore, much remains unknown about the fate of transfusion-derived iron during pediatric cancer treatment. Research is needed to elucidate how normal iron homeostasis might be modified in this population through suppression of erythropoiesis by chemotherapy, the inflammatory effects of infections and tissue breakdown, altered nutrition, and circulating factors emanating from the neoplasms themselves that could affect organ function (10, 28).

Cumulative PRBC transfusion volume was positively associated with LIC as measured by MRI in both univariate and multivariate analyses. A statistically significant correlation between projected (based on cumulative PRBC volume) and measured (based on MRI R2*) LIC values also was demonstrated, suggesting the potential value of tracking cumulative PRBC volume as a risk factor for IO, something that currently is not a standard practice. Statistically significant correlations were found between liver R2*, pancreatic R2*, and several serum iron biomarkers, including ferritin. The correlation between liver iron (as reflected in liver R2*) and serum ferritin, in particular, is confirmatory of other studies and of the longstanding clinical practice of using trends in serum ferritin to indirectly monitor iron stores in transfused patients (29). Ferritin assays are easily available, relatively inexpensive, and well standardized (30), but because ferritin levels are known to vary with clinical conditions other than IO (such as chronic inflammation, chronic liver damage, and malignancies), as well as with intensity of transfusion therapy, its value as an independent marker of iron balance has been questioned (2, 31). Taken together, however, study findings suggest that systematic monitoring and tracking of cumulative PRBC volume would be prudent in clinical practice to facilitate the most judicious use of PRBC transfusions and to identify patients who should be monitored with serum ferritin levels that could trigger selective evaluation with MRI for early detection of iron loading. Staging algorithms for MRI examinations could be developed that classify pediatric oncology patients into low- and high-risk categories, based on cumulative PRBC volume, ferritin levels, and clinical characteristics as has been done for patients with thalassemia, with the advantage that such triaging could be cost-effective, conserve magnet time, and facilitate timely detection of IO (24).

In both univariate and multivariate analyses, older age was associated with increased LIC. It has been speculated that because CCSs usually terminate their transfusions upon completion of cancer treatment, they might substantially draw down their iron load over time through continued growth and development (7, 18). Given there is no mechanism for active excretion of excess iron, our results suggest that older children and adolescents exposed to multiple transfusions during cancer therapy may be less able to expend iron stores and sustain lifelong exposure to toxic iron, which could result in organ dysfunction many years after cure of their malignancy. However, age is a complex construct in pediatric populations because the timing and velocity of growth and development vary among individuals and by sex, and are influenced by nutrition, pubertal status, and other factors. There are variations in the types of malignancies that occur at different ages within the pediatric age range as well, and these malignancies vary in the intensity of treatment and transfusion support required for their management. Although further study is needed to better understand the interplay among multiple variables to explain why older age was a risk factor for increased LIC in this study, our data suggest that IO, along with other known treatment-related toxicities (32–36) may represent another adverse outcome with particular relevance for the adolescent and young adult population.

A limitation of this study is its cross-sectional design, which represents a single time point after therapy and cannot portray trends in organ iron disposition over time, nor could the potential clinical significance of long-term exposure to relatively lower levels of excessive liver iron concentration be determined. Nine (12.2%) study subjects received treatment with dexrazoxane, a chelating agent used as a cardioprotectant during the administration of cardiotoxic chemotherapy. Although its mechanism of action is believed to involve iron binding (37), it seems unlikely to have influenced organ iron concentrations inasmuch as this agent is not used therapeutically as an iron chelator and is administered only during anthracycline chemotherapy. Potential threats to generalizability of our findings may include volunteer bias due to undetected differences between participants and nonparticipants; selection bias due to exclusion of participants' age <7 years and those who could not tolerate an MRI scan; and survivor bias. In addition, though our cohort size is reasonably large, it is not known to what extent the study sample is representative of all CCSs with the same diagnoses and a similar length of follow-up.

In this cross-sectional study, we have established the prevalence, anatomic distribution, and severity of transfusion-related iron deposition in an institutional cohort of 75 CCSs who underwent quantitative MRI scanning to measure iron content in the liver, pancreas, and heart. In addition, we validated the expectation that tissue iron is correlated with cumulative transfusion volume and discovered that older patients (i.e., adolescents and young adults) seem to be at greatest risk. Our results point to the need for practice guidelines addressing the monitoring and management of transfusion-related IO in cancer survivors, exploration of the additive risk of liver/pancreatic damage from specific chemotherapeutic/other exposures, as well as for the design of patient/survivor education interventions aimed at minimizing other forms of liver/pancreatic damage during and after cancer treatment. Results of this study provide the groundwork for further research aimed at better understanding the biologic mechanisms of IO in pediatric oncology patients and translation of that knowledge into clinical practice. Importantly, further research—particularly of a prospective, longitudinal nature, and in a larger cohort—that links assessment of organ iron content to organ function as well as to genetic, genomic, and epigenetic parameters that may mediate iron organ iron uptake is needed to improve understanding of the eventual fate of iron-loaded organs in CCSs.

J.C. Wood is a consultant/advisory board member for Shire, BioMed Informatics, and ApoPharma. No potential conflicts of interest were disclosed by the other authors.

Conception and design: K.S. Ruccione, J.C. Wood, D.R. Freyer

Development of methodology: K.S. Ruccione, J.C. Wood, D.R. Freyer

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): K.S. Ruccione, J.C. Wood, C. Chen, D.R. Freyer

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K.S. Ruccione, J.C. Wood, R. Sposto, J. Malvar, D.R. Freyer

Writing, review, and/or revision of the manuscript: K.S. Ruccione, J.C. Wood, R. Sposto, J. Malvar, C. Chen, D.R. Freyer

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): K.S. Ruccione, J.C. Wood, R. Sposto

Study supervision: K.S. Ruccione, D.R. Freyer

The authors acknowledge the assistance of Lakshmi Damerla, Libertad Garcia, Marielena Sandoval, Ada Santa Cruz, and Octavio Zavala, as well as the Children's Hospital Los Angeles Clinical Trials Unit staff, and members of the Department of Radiology.

This work was supported in part by grant number UL1TR000130, Children's Hospital Los Angeles from the National Center for Advancing Translational Sciences (NCATS) at the NIH. K.S. Ruccione, principal investigator: St. Baldrick's Foundation, Concern Foundation, ThinkCure Foundation, Oncology Nursing Society/Sigma Theta Tau Nursing Honor Society, DAISY Foundation, and the Children's Center for Cancer and Blood Diseases at Children's Hospital Los Angeles.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.