Radioactive Iodide Therapy for Pediatric Graves' Disease

Primary Aims. The investigators propose to assess the safety of 131I use in children with hyperthyroidism due to Graves' disease (GD). The investigators will measure whole body radiation exposure following 131II therapy in children treated for GD. The investigators will assess the effects of GD treatment on chromosome structure.

These studies will involve collaborative efforts with Dr. Patrick Zanzonico (Memorial Sloan-Kettering Cancer Center), who is an expert in 131I dosimetry and Dr. James Tucker (Wayne State University), who is an expert in cytogenetic effects of radiation. The studies will involve children treated for GD at University of Florida University and Baylor University. These studies have been designed with the help of the University of Florida Center for Clinical Investigation, Biostatistics Support Unit, which will be involved in data analysis.

Characteristics of study population. The investigators will recruit a total of 150 patients diagnosed with GD younger than 18 years of age. All subjects are to be treated with 131I. In this trial, children will not be randomized to treatment, but will be treated per physician prescribed care. To ensure an equal distribution of age and gender between the two groups of children, the investigators stratify enrollment by gender (male vs. female) and age (5-10 yrs, 10-15 yrs, 15-18 yrs).

Two sites will enroll patients to achieve the desired sample size: Baylor College of Medicine and University of Florida University. These sites have been selected for the following reasons. (1) These are large centers where radioactive iodide has been used for decades. (2) Each site has treated large number of children with radioactive iodide. (3) Each site has computerized patient databases and contact information for children treated with 131I is known. (4) The investigators have working relationships with collaborators at these sites. Based on the relative patient volumes of Baylor and University of Florida, the investigators anticipate that 70% of patients will come from Baylor and 30% from University of Florida. Calculations supporting the sample size are detailed in each of the two aims below.

Patient eligibility. Eligibility criteria include the following:

1. A diagnosis of GD based on initial laboratory studies showing a suppressed Thyroid Stimulating Hormone (TSH) (i.e. \<0.01); high total triiodothyronine (T3), high total thyroxine (T4) and/or free T4 level; an elevated thyroid stimulating immunoglobulin (TSI) titer; increased and diffuse uptake of 123I, 131I, or 99Tc in the thyroid gland.
2. Age \<18 years at the time of GD disease diagnosis.
3. Non-smoking parents. Subject enrollment. Practitioners in the University of Florida Pediatric Thyroid Center and the Baylor Pediatric Endocrinology Division will identify eligible individuals for study participation. Patients will be enrolled after appropriate consent/assent procedures have been satisfied. At the time of collection, the investigators will record age, gender, current treatment, and treatment history for Graves' disease (i.e., antithryoid drugs (ATDs) and/or 131I therapy including dose).

These studies will only be performed on children treated with 131I as part of physician prescribed clinical care. Children will not be treated with 131I for the sole purpose of generating subjects for this study.

After treatment is decided upon by the physician and the patient, then the patient will be offered participation as to provide balanced enrollment for each treatment/age/gender category.

1. Primary aim (1) and secondary aim (i): Perform dosimetry to assess whole body and tissue specific radiation exposure in children treated with 131I and determine potential cancer risk from these data. At present, no data are available to assess whole body and tissue-specific radiation exposure for children treated with 131I for GD. Theoretical modeling has been done, but this has not been based on actual data. Knowing the exposure of specific organs to radioactivity can be used to determine tissue-specific risk for malignancies. The investigators thus propose to perform a cross-sectional dosimetry study on children being treated with 131I to determine tissue specific and whole body radiation exposure. These studies have been designed by Dr. David Cheng (University of Florida University), Dr. Patrick Zanzonico (Memorial Sloan-Kettering Cancer Center; NY), and Dr. James Dziura (University of Florida University).

   Patient's Total-Body Mass and Administered Activity. On the day of administration of the therapeutic administered activity of 131I, the patient will be weighed. Immediately prior to administration, the therapeutic administered activity of 131I will be measured in a dose calibrator on the 131I setting, and the activity and the dates and times of assay and of administration recorded. This activity will be prescribed by the treating physicians at University of Florida or Baylor.

   Gamma Camera Imaging. All 131I gamma camera whole-body scanning will be performed using a 20% photopeak energy window (i.e. 364 keV + 10% = 328 to 400 keV) and a scan speed of 10 cm/min for all scans. The scan length will be set for each patient to include the entire patient and the same scan length will be used for all scans of a given patient. The exact dates and times of each whole-body scan will be recorded. The time post-administration of each whole-body scan will be calculated as the time interval (in hours) from the date and time of 131I administration and the date and time of whole-body scanning.

   Determination of Organ and Total-Body Activities. The determination of organ and total-body activities will use each patient as his or her own calibration standard. The patient will undergo a conjugate-view whole-body scan within 30 to 60 minutes after the 131I administration (i.e. nominally specified as time 0) but before the first post-administration void or bowel movement. In addition, scans will be performed at one day and, four days after administration of the dose. A blood sample (10cc) will also be drawn to measure the amount of radioactive iodide in the blood at these times and for assessment of DNA damage markers. For each patient, the net (i.e. background-subtracted) geometric-mean count rate for the total body for this initial scan thus corresponds to 100% of the administered activity. As noted, this scan will be performed at 30 to 60 minutes post-administration to allow some dispersion of the activity throughout the body, so that the effects of scatter and attenuation are grossly the same for this initial scan as for subsequent scans of the patient. For the time-0 and each subsequent conjugate-view whole-body scan, the posterior (lower-detector) gamma-camera image is "mirrored" to align it with the anterior (upper-detector) image.

   The regions of interest (ROIs) will be manually drawn around the organs of interest (the thyroid, salivary glands, liver, intestines, stomach, and urinary bladder) and the total body. For each organ, its ROI may be drawn in the scan in which it is best visualized and then copied and pasted onto the other whole-body scans, translating and/or rotating the ROI as needed on these other scans to accurately superimpose it on the organ. Note that, for each scan, a single background (BG) ROI, drawn outside of but close to the body, may be used.

   Statistical Analysis. OLINDA-based Calculation of Organ Absorbed Doses and Effective Dose. The OLINDA dosimetry program will be used to assess absorbed doses (11, 76). In OLINDA, the investigators will select the "Fraction and Half-times" option (in OLINDA's "Kinetics Input Form") and enter the best-fit parameters of the respective time-activity functions (A/100% and Ta and, if, applicable, B/100% and Tb) for the specified source regions - the thyroid, salivary glands, liver, intestinal contents, stomach contents, urinary bladder contents, red marrow, and total body. Note that OLINDA requires the zero-time intercept values in fraction (not %) of the administered activity. Click "hr" radiobutton for the "Half-life Units" and the "Biological" radiobutton for the "Half-lives." Also in OLINDA, the investigators will select iodine-131 (I-131") as the nuclide (in OLINDA's "Nuclide Input Form") and the anatomic model most closely approximating the age or total-body mass of the patient as the model (in OLINDA's "Model Input Form"). Then, select the "Main Input Form" and click the "Doses" button to calculate the organ doses and the effective dose.

   Evaluation of Radiation exposure. Distributions of the primary outcome measures for the total body and organ-specific radiation exposure (described above) will be summarized graphically (boxplots) and numerically (means, standard deviations, medians, interquartile ranges).

   Radiation exposure (e.g., absorbed dose of 131I in the total body and specific organs) will be compared across specific categories of the administered dose of 131I, as well as across age groups, and gender using Analysis of Variance (ANOVA). The investigators will also evaluate whether there were outcome differences by study site. Should data not comply with distributional assumptions required of the ANOVA, alternative non-parametric techniques (i.e. Kruskal-Wallis test) will be considered. The investigators will correlate the administered dose of 131I with the absorbed dose of the radioactive agent, using Spearman's Rank Correlation. In all analyses, alpha of 0.05 will be used.

2. Primary aim (2) and secondary aim (ii): Assess chromosomal translocations in children treated with 131I and evaluate chromosomal translocations as related to patient's age and 131I exposure. Low-level, whole body irradiation is a risk factor for cancer 58 . The prolonged use of certain medications is associated with the risk of cancer in some circumstances as well. Current 131I therapy for Graves' disease in children and adults aims for ablation sufficient amounts of thyroid gland to result in a hypothyroid state. This treatment, though, will also be associated with low-level whole body irradiation11. Studies of adults, who have been treated with 131I, have revealed small increases in rates of stomach and breast cancer. Although it has been suggested that children are more prone to carcinogenic risks of low level irradiation58, there have not been any studies with a sufficient sample size to assess long-term cancer risk in children treated with 131I.

Recent data convincingly show that chromosome translocations are associated with long-term cancer risks. Chromosome translocations are a molecular sign of ionizing radiation exposure. Importantly, translocations persist for decades after radiation exposure22. This persistence makes chromosomal translocations the gold-standard aberration type for performing radiation dosimetry when there is a lag between the time of exposure and assessment. Normative data for chromosomal translocations are available, as related to age and gender20.

The investigators therefore propose to perform an observational cohort study of children treated for Graves' disease to assess chromosomal translocation. These studies will be performed on the children in which dosimetry is performed, as detailed above. The investigators will stratify enrollment by gender and age to ensure a comparable distribution of these characteristics. The chromosome translocation studies, at baseline and at the 12 month-follow-up.

Treatment with 131I. Patients will be treated with 131I as detailed above. Sample Collection. Blood will be obtained for chromosome translocation analysis at baseline and at 12 months after treatment with 131I, or after receiving surgery or ATDs. For blood collection, a heparinized vacutainer will be used to collect 5 ml of blood. Blood will be obtained at the time of routine phlebotomy for assessment of thyroid hormone levels.

1. Sample size calculation. The investigators will test the hypothesis that translocation frequencies are higher in subjects receiving 131I compared to subjects receiving alternative treatment (ATDs or surgery only) for GD. Since there is a low level of chromosomal breaks in healthy children20, if there is an increase in chromosomal translocation it should be possible to detect significant increases with a relatively small sample size. Our estimates of sample size are based on rates of translocation described by Sigurdson who observed rates of 0.2 translocations per 100 cell equivalents in children under 20 y. Given these baseline rates and using the PASS 2005 module for Poisson regression, the investigators estimated that a sample size of 135 children treated with 131I and 135 treated with ATDs or surgery will provide 80% power at the two-sided 0.05 significance level to detect a doubling of the chromosomal translocation rate between the two groups of patients at 12 months after tre. The investigators will aim for 1/3 of the children being in each of the following age groups: 5-10 yrs, 10-15 yrs, 15-18 yrs. The investigators will enroll 150 subjects in each group to accommodate a potential 10% loss to follow-up.
2. FISH assay for chromosome aberrations. Personnel in Dr. Tucker's laboratory will determine the frequency of chromosome translocations using Fluorescence In Situ Hybridization (FISH) whole chromosome painting probes. All samples will be coded so that the Tucker laboratory will not know the radiation exposure history of the subjects. Cell cultures will be initiated 24-48 hr after phlebotomy in Dr. Tucker's laboratory and processed according to routine cytogenetic methods. Approximately 1,800 metaphase cells will be evaluated per subject, and this will be equivalent to 1,800 x 0.56 = 1,000 metaphase cells (define as cell equivalents; CEs) as if the full genome had been scored. All translocations in cells will be enumerated and the frequency of translocations per 100 CEs will be used as the dependent variable in the statistical analyses.

3. Data Analysis. Data analysis will be conducted in collaboration with the Biostatistics Unit of the University of Florida Center for Clinical Investigation. All analyses will be performed using SAS v9.2 (SAS Institute, Cary, NC) with a two-sided 0.05 type I error used to evaluate statistical significance. Frequency distributions and missingness will be examined for each variable. The investigators will omit from consideration in further analyses variables with homogenous distributions or with a high degree of missingness and collapse categorical variables with underrepresented levels. Associations between independent variables will be examined using Spearman correlation coefficients, principal components and hierarchical clustering (PROC VARCLUS).

   Demographic (age, gender and race of child and primary caretaker), socioeconomic (parental education and income), and clinical variables (e.g., duration of Graves disease, ATD treatment, and 131I dose) will be compared between the two treatment groups at baseline using t-tests for continuous variables and chi-square tests for categorical variables. Meaningful clinical differences will be reported and adjusted for in the multivariate analysis of chromosomal translocation described below.

   (ii) Comparison of Chromosomal Translocation Frequencies. The number of chromosomal translocations at baseline and at 12 months post treatment will be determined using means and confidence intervals. These data will also be compared to our data for healthy children.

   The investigators will compare the number of chromosomal translocations at baseline and at 12 months between groups in a multivariate model, using zero-inflated Poisson mixed model analysis69. The zero-inflated Poisson model accommodates the increased variance that is typical of count data with a large proportion of zeroes. Furthermore, through the inclusion of random effects, the mixed model analysis will allow for the correlation from repeated observations. The mixed model also accommodates individuals with incomplete observations (i.e. lost to follow-up) under the assumption that given the observed data the missing data is not dependent on unobserved values.

   The following fixed effects will be used: treatment group (131I treatment vs. surgery or ATD group), selected covariates (e.g., child age and gender, 131I dose, parental education or income), time (baseline and at 12months), and time by treatment group interaction. A random effect will be used to account for possible correlation between the number of chromosomal translocations at baseline and 12 months for the same subject. The investigators will also explore whether the effect of treatment on chromosomal translocations is modified by the age of child and the received radiation dose, by including three-way interaction terms between treatment, age and time or treatment, radiation dose and time.

   Several strategies will be imposed to accommodate the likelihood that missing data will occur during this study. Prevention is the most obvious and effective manner to control bias and loss of power from missing data. Postcard and telephone visit reminders will be delivered to participants prior to protocol specified collection times. Alternative contacts will be identified on entry into the study to minimize loss-to follow-up. Timely data entry combined with weekly missing data reports will trigger protocols for tracking and obtaining missing data items or outcome assessments. Despite these prevention efforts, it is reasonable to assume missing data will occur. The primary analysis method will use a likelihood-based mixed model which accommodates incomplete observations and operates under the assumption that the missing data is missing at random (MAR)69. Missing data patterns and reasons for dropout will be compared between the treatment groups. T-tests, cross-tabulations and logistic regression will be used to evaluate whether withdrawal is dependent on any observed variables.

4. Statistical Considerations: Describe the statistical analyses that support the study design.

The investigators used Power Analysis and Sample Size software (PASS 2005) to estimate precision around a mean of radiation exposure (e.g., expressed as mean percent of administered activity in an organ/total body or Olinda-based mean organ absorbed dose and mean effective dose). A sample of 150 subjects produces a 95% confidence interval equal to a mean plus or minus 0.16 standard deviations. From a previously published study of children with Graves disease (ages 7-18 years)42, the investigators estimated that our sample of 150 patients can be stratified into patients who will receive 150-200 Gy and 200-300 Gy. Given these proportions, the investigators will be able to estimate stratum-specific 95% confidence intervals around the means with a precision of 0.32 standard deviations for the two smaller strata and 0.16 standard deviations for the larger strata.