SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Acknowledgments
  7. Disclosure Statement
  8. References

Iron can be a potent pro-oxidant and, on this basis, elevated body iron may increase the risk of cancer. Although epidemiological evidence is mixed, there is overall support for this possibility. In addition, because of this same oxidative capacity, body iron levels may alter radiation sensitivity. In the present study, a nested case-control study of breast cancer was conducted in Japanese atomic bomb survivors. Stored serum samples from the Adult Health Study cohort were assayed for ferritin levels and joint statistical analyses were conducted of ferritin and radiation dose on the risk of breast cancer. Serum ferritin is the best feasible indicator of body iron levels in otherwise healthy people. A total of 107 cases and 212 controls were available for analysis. The relative risk (RR) of breast cancer for a 1 log unit increase in ferritin was 1.4 (95% confidence interval 1.1–1.8). This translates to an RR of 1.64 comparing high and low values of the interquartile range among controls (58 and 13.2 ng/mL, respectively). The results support the hypothesis that elevated body iron stores increase the risk of breast cancer. However, the study was inconclusive regarding the question of whether body iron alters radiation-induced breast cancer risk. (Cancer Sci 2011; 102: 2236–2240)

The carcinogenic impact of radiation exposure following the atomic bombing of Hiroshima and Nagasaki may be influenced by the individual sensitivity of the person exposed:(1) a given dose may be a greater risk for one person than another. This sensitivity could be influenced by diet and lifestyle factors, as well as by genetic variants in proteins that either protect a cell from oxidative damage or repair damage that has occurred. One factor that may alter radiosensitivity is body iron stores.(2,3)

There are a limited number of studies of body iron and cancer that tend to support an association between the two,(4,5) although other studies do not support this relationship.(6) The rationale for these studies rests primarily on the pro-oxidant capacity of iron,(7–9) which may also help cycle catecholestrogen metabolites that themselves can cause radical damage to DNA and other biomolecules.(10) Recently, studies have appeared that have focused on breast cancer in particular and offer support for elevated dietary iron intake and/or body iron stores as a risk factor.(11–14) Other studies have examined mutations in the hemochromatosis gene HFE and breast cancer risk(15,16) based on the idea that a diet high in iron would be particularly disadvantageous to those people carrying these mutations because increased iron absorption would lead to oxidative stress in mammary tissue from excess tissue iron.(17,18) Even though HFE mutations are very rare in Japan,(19) other genetic variants involved in iron absorption from the diet and/or physiological processing undoubtedly exist, although they have not yet been discovered.

In the present study, we conducted a nested case-control study of the association between body iron levels (inferred from serum ferritin measurements) and the subsequent risk of breast cancer based on serum samples from atomic bomb survivors taken as part of the Adult Health Study. We also hoped to determine whether elevated body iron increased the effectiveness of radiation exposure to cause breast cancer.

Materials and Methods

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Acknowledgments
  7. Disclosure Statement
  8. References

Subjects.  The Adult Health Study (AHS) comprises approximately 20 000 people selected from the Life Span Study (LSS) of atomic bomb survivors in Hiroshima and Nagasaki who have undergone physical examinations and interviews every 2 years since 1958; serum samples have been collected and stored since 1969. The cases in the present study were women from the AHS cohort who had been diagnosed with breast cancer between 1971 and 2001. Sera from two women who were at-risk on the date of each case diagnosis and who had been exposed to a known radiation dose were chosen as controls. Each case was matched with potential controls for age at time of blood collection (±2 years from index case), sample collection year (same decade), and city (Hiroshima versus Nagasaki), and counter-matched with cancer cases on breast radiation dose group (low, medium, and high with cut-off points at 5 mGy [for low] and 790 mGy [for high]).(20) In addition, height and weight were assessed at the time of blood donation and body mass index (BMI) was then calculated in kg/m2.

Parity, age at first full-term pregnancy, age at menarche, and age at menopause for some subjects were obtained from periodic questionnaires, but were not known for all subjects. Menopausal status at the time of blood collection was assigned by age (<43 and >55 years for pre- and post-menopause, respectively). If age at collection was between 43 and 55 years, menopausal status was determined after serum collection on the basis of follicle-stimulating hormone (FSH) concentrations based on the manufacturer’s test guidelines (FSH, Tokyo Special Reference Labs, http://www.srl.info/srlinfo/news/2003-14.htm, 2003). Matched controls whose menopausal status at the time of collection did not match that of the corresponding case were excluded. After further excluding cases with missing data for radiation dose or ferritin, a total of 107 cases and 212 matched controls remained: 15 premenopausal cases (29 controls) and 92 post-menopausal cases (183 controls). Exclusions and the numbers of samples are shown in Figure 1.

image

Figure 1.  Exclusion criteria and numbers of subjects included in the present study.

Download figure to PowerPoint

Radiation doses to the breast, based on Dosimetry System 2002,(21) were adjusted to correct for regression bias resulting from random errors (uncertainties) in dose estimates.(22) Eligibility was further restricted to subjects who were free of certain medical conditions (oophorectomy, current exogenous hormone treatment, prior malignancies [or later diagnosis within 2 years], or pregnancy) and whose menopausal status at the time of serum collection could be classified. Perimenopausal samples were not included because there were too few of them to allow for separate estimation of effects. Seventeen samples were removed because they were drawn during the perimenopausal period. Five cases and five controls had both a pre- and post-menopausal serum sample; when ferritin was analyzed without regard to menopausal status at time of serum collection, the pre- and post-menopausal values of log ferritin were averaged for these subjects.

The present study (RERF Research Protocol 6–02) was reviewed and approved by the Research Protocol Review Committee and the Human Investigation Committee of the Radiation Effects Research Foundation (Hiroshima, Japan).

Sample storage and serum assays.  Serum samples were collected between 1969 and 1991 in the Hiroshima and Nagasaki laboratories and stored in vials at −80°C. Assays were not performed in duplicate owing to the limited amounts of sera available. All assays were performed blinded to case-control status, age, city of collection, and radiation exposure. When measured values exceeded the upper limit of detection, samples were diluted and remeasured. Serum levels of ferritin were measured at a single laboratory (SRL Laboratory, Hachioji, Japan) using the conventional chemiluminescent enzyme immunoassay (CLEIA).

Sera were collected at least 2 years prior to diagnosis; the mean time between serum collection age and age at diagnosis was 13 years, with the maximum being 26 years.

Statistical methods and analysis.  The counter-matched nested case-control design is analyzed using a partial likelihood with sampling weights to overcome the biased control selection,(23,24) which can be implemented using conditional binary regression software. The generalized regression program for matched sets with binary outcomes (GMBO) of Epicure (Hirosoft, Seattle, WA, USA) was used in the present study. Analyses of radiation risk in the full LSS cohort(25,26) have used an excess relative risk (ERR) model(27) for the effect of radiation, so the same model was used in the present analyses. The effect of ferritin was assessed using the natural logarithm, either grouped into tertiles or as a continuous factor. Possible confounding of the marginal effect of ferritin and its joint effect with radiation was tested by including BMI and parity in a log-linear baseline model. Data on age at menarche and (in post-menopausal cases and controls) age at menopause were too sparse to examine possible confounding effects.

The joint effects of radiation and ferritin were assessed using both additive and multiplicative models,(28) and departure from purely additive or purely multiplicative joint effects was assessed using interaction terms(29) and a mixed model,(28,30) where 0 signifies additivity, 0–1 signifies superadditive but submultiplicative, 1 signifies multiplicativity, and >1 signifies supermultiplicative. The additive and multiplicative models can be compared in terms of their deviances, which are equivalent to the Akaike Information Criterion (AIC) and Bayes Information Criterion (BIC),(31) because both models have the same number of parameters.

The design of the study provided approximately 80% of the power of the full cohort to detect interaction,(20) but it is well known that observational studies generally lack power to detect low levels of interaction.

Results

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Acknowledgments
  7. Disclosure Statement
  8. References

Table 1 shows subject characteristics of cases and controls used in the analyses. Cases had a higher mean radiation dose and lower parity than controls, but similar BMI.

Table 1.   Summary of study sample characteristics
 CasesControls
  1. Data are the mean ± SD with the number of individuals given in parentheses. †Five cases and five controls had samples of both pre- and post-menopausal sera. ‡The control mean ± SD dose are weighted by the counter-matched sampling weights, otherwise mean control dose would be biased by the oversampling of high-dose controls used to increase the precision of analyses of effect modification.(20) Cases were not selected according to their dose, so the mean dose among cases is not weighted.

Premenopausal breast cancer (15 cases, 29 controls)
 Log serum ferritin (units)2.78 ± 0.95 (15)2.89 ± 1.09 (29)
 Radiation dose (Gy)0.72 ± 1.00 (15)0.65 ± 0.93‡ (29)
 Body mass index (kg/m2)22.9 ± 3.0 (14)22.4 ± 3.0 (29)
 Parity1.9 ± 1.0 (14)2.2 ± 1.2 (22)
Post-menopausal breast cancer (92 cases, 183 controls†)
 Log serum ferritin (units)
  Premenopausal sera3.10 ± 1.05 (41)2.85 ± 1.16 (79)
  Post-menopausal sera4.06 ± 0.88 (56)3.92 ± 0.78 (109)
 Radiation dose (Gy)0.61 ± 0.90 (92)0.35 ± 0.62‡ (183)
 Body mass index (kg/m2)23.2 ± 3.7 (87)22.6 ± 3.2 (175)
 Parity2.4 ± 1.8 (84)3.1 ± 1.9 (178)

Associations of ferritin and radiation with risk are given in Tables 2 and 3. Table 2 shows the results for serum ferritin and breast cancer risk stratified in several different ways. The primary analysis is presented at the top for “All cancer, All serum samples”. The crude RR associated with a 1 log unit change in ferritin was 1.4. The geometric mean ferritin level in controls was 26 ng/mL; the interquartile range was 13–58 ng/mL. Thus, the RR of 1.4 corresponds to a 1 unit increase above the log of the baseline value, a level of 71 ng/mL; an RR of 2.0 would be expected if levels increased to 194 ng/mL. The RR of 58 ng/mL compared with 13 ng/mL (the interquartile range) is 1.64. These values of serum ferritin are still well within the normal range of serum ferritin; clinically, <12 ng/mL is taken as evidence of iron deficiency, whereas >1000 ng/mL is evidence of severe iron overload.(32) The radiation-adjusted RR associated with a 1 log unit change in ferritin was 1.3.

Table 2.   Crude and mutually adjusted relative risks for ferritin and radiation according to menopausal status at diagnosis and at the time of serum collection
 Crude RR† (95% CI)P value‡Adjusted RR (95% CI)P value‡
  1. †The relative risk (RR) for ferritin is calculated from a log-linear model as: log (RR) = β × log(ferritin). The RR for radiation is calculated from an excess relative risk (ERR) model as: ERR = γ × dose. ‡P values for the likelihood ratio test; confidence bounds are likelihood-based bounds. CI, confidence interval.

All breast cancer
 All serum samples (pre- and post-menopausal)
  Ferritin1.4 (1.1–1.8)0.0201.3 (1.0–1.7)0.033
  Radiation1.9 (1.3–2.8)<0.0011.9 (1.3–2.9)<0.001
 Premenopausal serum samples only
  Ferritin1.3 (1.0–1.9)0.0851.2 (0.9–1.7)0.24
  Radiation2.3 (2.3–4.2)0.0012.1 (1.4–3.9)0.003
Postmenopausal breast cancer
 All serum samples
  Ferritin1.5 (1.1–2.0)0.0081.4 (1.1–1.9)0.020
  Radiation2.1 (1.4–3.3)<0.0012.0 (1.3–3.1)<0.001
 Premenopausal serum samples only
  Ferritin1.5 (1.0–2.2)0.0321.3 (0.9–1.9)0.25
  Radiation3.0 (1.5–6.2)<0.0012.6 (1.3–5.6)0.003
 Premenopausal serum samples only
  Ferritin1.3 (0.9–2.1)0.171.4 (0.9–2.1)0.17
  Radiation1.7 (1.1–3.1)0.0261.8 (1.1–3.2)0.025
Premenopausal breast cancer
 Premenopausal serum samples
  Ferritin0.9 (0.4–1.8)>0.51.0 (0.5–1.9)>0.5
  Radiation1.4 (0.8–4.3)0.411.4 (0.7–4.3)0.43
Table 3.   Further aspects of the joint effects of ferritin and radiation on post-menopausal breast cancer risk
ModelRelative risk (95% CI)P value for the interaction
Ferritin tertile† (vs lowest)Radiation
MiddleHigh
  1. †Premenopausal serum log ferritin was split at cut-off points of 2.4 and 3.5, whereas post-menopausal log serum ferritin was split at cut-off points of 3.8 and 4.4. CI, confidence interval.

Premenopausal sera
 Multiplicative2.2 (0.8–5.9)1.1 (0.4–3.5)3.1 (1.5–6.8)0.23
 Additive3.6 (1.0–15)0.8 (0.1–5.2)4.3 (1.9–15)0.48
Post-menopausal sera
 Multiplicative2.3 (0.9–5.6)2.5 (1.1–5.7)1.9 (1.1–3.5)>0.5
 Additive2.5 (0.9–8.7)2.8 (1.0–9.2)2.5 (1.1–7.6)>0.5

Because high BMI may have affected our results, we removed all subjects with BMI >25 kg/m2. This reduced the number of post-menopausal cases (with known radiation dose) from 92 to 60. The crude risk for 1 log unit ferritin after this exclusion was 1.67 (95% confidence interval [CI] 1.18–2.38; = 0.004), and the radiation-adjusted risk for ferritin was 1.44 (95% CI 1.03–2.03; = 0.032). The median ferritin levels in this subset of subjects were 41.3 ng/mL among cases and 31.8 ng/mL among controls.

Similarly, because anemia may have affected our analysis, when post-menopausal women who had hemoglobin levels ≤11 g/dL were excluded from analysis, the estimated radiation-adjusted risk for a 1 log unit of ferritin was 1.47 (95% CI 0.99–2.18) based on 76 cases and 137 controls.

When stratified according to menopausal status at the time of cancer diagnosis, the radiation-adjusted RR for ferritin in post-menopausal cases was 1.4 based on all serum samples. When restricted to only those postmenopausal women for whom a premenopausal sample was collected, the radiation-adjusted RR was 1.3 and, when restricted to postmenopausal serum samples only, the RR was 1.4. Figure 2 shows the distributions of log serum ferritin among post-menopausal cases and their corresponding controls. The distribution in post-menopausal sera among cases was shifted towards larger values compared with controls. In premenopausal sera, cases had a paucity of lower values, but otherwise similar distribution as controls. The mean serum ferritin among controls before menopause was more than threefold lower than that after menopause, yet the estimates of RR associated with differences between cases and controls using pre- and post-menopausal serum samples was nearly equivalent among the post-menopausal cases of breast cancer. Because the phase of the menstrual cycle was not recorded at the time of serum collection, we cannot clearly interpret any differences between cases and controls in terms of premenopausal serum ferritin levels, although there was a significant risk of breast cancer for premenopausal ferritin (Table 2).

image

Figure 2.  Distribution of serum ferritin (natural log transformed) in post-menopausal breast cancer cases and nested (risk-set matched) controls according to menopausal status at the time of serum collection. (a) Post-menopausal sera from cases; (b) premenopausal sera from cases; (c) post-menopausal sera from controls; (d) premenopausal sera from controls.

Download figure to PowerPoint

Table 3 examines in greater detail the joint effects of ferritin and radiation on breast cancer risk. Owing to the paucity of premenopausal breast cancer cases, detailed analyses of joint effects are limited to post-menopausal cases. Because analyses of joint effects can be sensitive to the form of mathematical model used for the main effects of individual factors,(33) stratification based on tertiles of log serum ferritin was used rather than modeling the risk based on continuous values. We also note for the purposes of assessing joint effects, the use of tertiles does not impose a constraint on the form of ferritin risk. There is the suggestion of a non-linear effect of ferritin on risk, but the departure from linearity is not significant.

The tertiles of log ferritin are separated by approximately 1 unit and the risk for the middle and high tertiles is approximately double compared with the lowest tertile; thus, it is not surprising that the overall linear increase is in the order of between 1 and 2 (Table 2). For radiation, a linear ERR model has been well established elsewhere,(25,34) and so is used in the present analysis. The difference in deviance for additive and multiplicative models of post-menopausal breast cancer risk for post-menopausal serum ferritin and radiation was 0.797, corresponding to an evidence ratio of 1.48,(31) which means that the probability is <60% that the multiplicative model fits better (a difference in deviance of at least 5.9 would be required to judge one model better with 95% confidence). The best estimate of mixture parameter was >1.0, suggesting that the joint effect may be multiplicative or supermultiplicative (synergistic). With premenopausal serum ferritin, the additive model was slightly favored (difference in deviance 1.70; evidence ratio 2.34; probability approximately 75%) and the best estimate of mixture was less than zero (subadditive). However, the results are purely qualitative and should be interpreted with caution because there is insufficient power to state with confidence on which scale the joint effects are better fit to the data.

Discussion

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Acknowledgments
  7. Disclosure Statement
  8. References

Cancer of the breast is one of the most radiogenic malignancies in atomic bomb survivors(26) and the excess lifetime risk is greatest in women exposed at a young age.(34) The purpose of the present nested case-control study was to test the hypothesis that body iron stores, as measured by serum ferritin levels, are associated with the risk of breast cancer in female atomic bomb survivors and, further, to determine whether the impact of radiation exposure is greater in women with higher body iron. Excess body iron may increase risk directly through oxidative stress mechanisms, and may also act to sensitize cells and tissues to the damaging effects of radiation exposure by depleting cells of their ability to scavenge radical.(2)

As predicted, we found a significant association between serum ferritin and risk. Other measures, such as transferrin saturation and total iron binding capacity, have been used in some previous studies of body iron and cancer,(4,35) although, among otherwise disease-free people, serum ferritin is the best indicator.(32,36) Some chronic diseases and inflammation can yield elevated serum ferritin that does not reflect body iron status,(36) but all of our cases of breast cancer had blood drawn at least 2 years prior to diagnosis. It should also be noted that serum ferritin levels are not a perfect indicator of body iron status and thus there is a degree of exposure misclassification in the present study.

Human iron absorption from dietary sources is tightly regulated. However, particularly for heme iron, this regulation is not perfect(36) and body iron increases with age, especially after menopause for women. Uptake depends on the heme and non-heme iron content of the diet, as well as on a variety of genetic polymorphisms that alter the regulation of iron uptake.(37) The deleterious effects of genetic conditions that result in iron-overload, such as hereditary hemochromatosis (HFE), appear to result exclusively from the iron overload and not other effects of these genetic variants themselves.(32) Therefore, our interest has been on excess body iron in the general population.

The serum samples used in the present study were taken from women surviving at least 24 years after the bombing and up to 26 years before cancer onset. If elevated body iron increases susceptibility to radiation exposure, then this would presumably only be relevant at the time of radiation exposure. Therefore, the question is, to what extent are an individual’s body iron stores stable over time? This would depend on genetic make up relevant to iron absorption and processing(37) and the stability, over time, of the typical diet. There is some evidence for long-term correlation from a family study of the heritability of body iron levels in both HFE mutation carriers and non-carriers.(38) In that study, among 276 HFE wild-type participants in the hemochromatosis and iron overload screening (HEIRS) study who did have transferrin saturation and serum ferritin above a screening threshold, the residual heritability (after adjustment for demographic covariables) of serum ferritin was 0.64, a very high number. The largest racial group among the 276 was “Asian”, with 98 participants. That implies, at least among people with higher than average body iron levels, that serum ferritin levels are stable over time.

The results of an analysis for an interaction between ferritin levels and radiation were ambiguous (Table 3), a common problem with observational studies.(39,40) Although there was some indication of an interaction, it did not reach statistical significance. Alterations of radiation sensitivity by body iron stores are important possibilities that require further investigation.

The potential limitations of the present study include confounding by factors such as alcohol consumption or age at menarche and menopause. We did take account of menopausal status in our analyses, but we could not adjust for age at menarche or alcohol intake. Most studies do show an association of age at menarche and breast cancer risk; however, in the bomb survivor population, Land et al.(41) did not find such an association, whereas Goodman et al.(42) did. Alcohol consumption among Japanese women of this age group is probably quite low. In particular, Hori et al.(8) report ferritin levels according to alcohol intake among a sample of 215 Japanese women in 2006 (mostly under 50 years of age). Of these women, 93% consumed two or fewer drinks per day (0–19.9 g/day) and, of the 7% consuming more than two drinks (>20 g/day), serum ferritin was 32.7 ng/mL (compared with 29.3 ng/mL among those consuming no drinks per day). In addition, the impact of alcohol consumption on risk is very modest at low levels of consumption, with RRs of 1.13 for 15–24 g/day and 1.21 for 25–34 g/day.(43) For a confounder to entirely account for our results, it would have to be a strong risk factor for breast cancer and strongly predict ferritin level. Therefore, we believe unspecified confounding does not account for our results.

Acknowledgments

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Acknowledgments
  7. Disclosure Statement
  8. References

The Radiation Effects Research Foundation (RERF), Hiroshima and Nagasaki, Japan, is a private, non-profit foundation funded by the Japanese Ministry of Health, Labor and Welfare (MHLW) and the US Department of Energy (DOE), the latter in part through the National Academy of Sciences. This publication was supported by RERF Research Protocol RP #06–02 and by US National Cancer Institute contract NO1-CP-31012 and the Japanese Ministry of Education, Culture, Sports, Science and Technology grants 14031227 and 15026220.

References

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Acknowledgments
  7. Disclosure Statement
  8. References