Treatment of Pediatric Graves' Disease Is Associated with Excessive Weight Gain

Abstract

Context:

Little information is available about changes in body weight and body mass index in children before, during, and after treatment for Graves' disease (GD).

Objective:

Our objective was to examine changes in body weight after treatment for GD in children as related to clinical features.

Design:

The medical records of 43 pediatric patients with GD [35 girls and eight boys, aged 4.0–18.5 (mean 10.9) yr] were examined. Patients were included if clinical data were available for 1 yr before and after the diagnosis of GD.

Main Outcome Measures:

Weight, height, body mass index (BMI) z-scores, and thyroid hormone levels were assessed.

Results:

Overall, patients presented with an average BMI z-score of −0.02 ± 1.05 that was not different from the normal population (P = 0.921) or their premorbid values (P = 0.07). However, in the subset of patients who were initially overweight or obese in the premorbid state, the BMI decreased significantly during the development of hyperthyroidism (P < 0.05). After initiation of treatment, patients gained significant amounts of weight over the first 6 months leading to elevated BMI z-scores (P < 0.0001), and elevations in BMI persisted in about 25% of the patients.

Conclusion:

Excessive weight gain within 6 months of treatment is seen in children treated for GD, and the gain in weight can persist.

Graves' disease (GD) is the most common cause of hyperthyroidism in children and adolescents (1–3). In 2008, it was estimated that approximately 8000 children in the United States were treated for this disease with an estimated prevalence of one in 10,000 children (4). As early as 1895, secretions from the thyroid gland were observed to influence the metabolic rate of humans and influence body weight (5). It is now well recognized that thyroid hormones influence growth and body mass with hyperthyroidism often associated with weight loss at presentation, although normal or increased body weight can be seen (6–9).

Studies in adults have reported sustained increases in body weight after initiation of treatment of GD that exceeds prediagnosis weight with most of the weight gain (54–67%) occurring within the first 3 months of therapy (6, 7, 9–14). In adults treated with ¹³¹I for GD, sustained increases in BMI have been observed 1 yr after treatment with an average gain of 2.33 kg/m² (13). In another study of adults treated with ¹³¹I, at 4 yr of follow-up, females had higher weights than GD-pretherapy values (12).

At present, there is little information about body weight in children after treatment for GD. In a study of 57 hyperthyroid children, body weight was observed to increase by 14.3% by the second follow-up visit over an undefined period (15). Factors that influenced weight gain in that study were not identified, and the follow-up duration was limited.

To further examine this issue, we examined changes in weight in children with GD who had follow-up data available for 1–3 years after therapy onset. We now report demographic and clinical features associated with long-term changes in body weight and BMI in boys and girls treated for GD.

Patients and Methods

The medical records of children between the ages of 4–18 yr who were diagnosed with GD and treated at the Yale Pediatric Thyroid Center over the past 5 yr (2006–2011) were reviewed. GD was defined by the presence of biochemical hyperthyroidism [elevated circulating total T₄ (TT₄) and/or free T₄ (fT₄) levels and suppressed TSH levels], elevated thyroid-stimulating Ig levels, increased ¹²³I or ¹³¹I uptake by the thyroid on nuclear medicine scans, goiter, and proptosis. Patients were included for analysis if information was available for height, weight, and thyroid hormone levels after initiation of therapy for 1 yr and for 1 yr before treatment onset (the premorbid period).

Body mass index (BMI) was calculated using the following formula: BMI = weight in kilograms/height in square meters. BMI values were compared with gender- and age-specific reference values using the Centers for Disease Control and Prevention 2000 growth charts choosing the nearest half-month of age resulting in BMI z-scores (16). Per standard criteria, normal weight was defined as BMI-for-age scores between the third and 85th percentiles. Underweight was defined as a BMI-for-age at or below the third percentile (BMI z-score ≤ −1.9). Overweight was defined as BMI values between the 85th and 94th percentiles for age (+1.04 ≤ BMI z-score < +1.67). Obesity was defined as a BMI-for-age values at or above the 95th percentile (BMI z-score ≥ +1.67) (17, 18). Premorbid BMI z-scores were calculated from height and weight measured by the primary care provider and documented in the medical record.

Euthyroidism was defined by a TT₄ normal range of 5.0–10.6 μg/dl and fT₄ values of 1.0–2.2 μg/dl. TSH normative values were adjusted for age according to Lazar et al. (19).

Data were analyzed as related to different posttreatment periods. The time of diagnosis and treatment onset was designated as 0 months [time zero (T0)]. Follow-up periods after treatment onset were designated 3, 6, 9, 12, 15, 18, 24, 30, and 36 months. The 36-month follow-up period was selected as the end point for analysis because the number of patients with follow-up longer than 36 months was limited.

Patients with comorbid conditions that could affect body weight, including celiac disease, malignancy, diabetes mellitus, and inflammatory bowel disease, were excluded.

This study was approved by the Yale University Human Investigation Committee.

Statistical analysis

Data were collected and recorded on spreadsheets. The one-sided t test was used to compare the BMI z-score premorbid values and at T0 with the normal population (BMI z-score = 0). The Mann-Whitney U test was used to analyze the distribution and differences of BMI z-scores over different groups, which did not meet the criteria for normal distribution or when there were fewer than 30 subjects in each group. The Spearman rank test was used to analyze correlations and dependence between continuous variables. Friedman's two-way ANOVA was used to compare matched BMI z-scores over time when the number of subjects was less than 30. The Wilcoxon signed rank test with Bonferroni correction was used to analyze where differences exactly occurred for paired observations. The χ² test was used to compare binomial variables with each other. Descriptive statistics are given as the mean ± sd and as median and range. Statistical analyses were performed with SPSS version 18.0 (SPSS, Chicago, IL). P < 0.05 was considered to be statistically significant.

Results

Characteristics of the study population

Forty-three children met eligibility criteria (Table 1). Two patients were treated with ¹³¹I initially. One patient was treated within 2 months after diagnosis by total thyroidectomy. Most patients were treated with antithyroid drugs (n = 40) initially. Two patients were treated with propylthiouracil (PTU). Thirty-eight patients were treated with methimazole (MMI). Four patients were changed to PTU after developing hypersensitivity reactions to MMI. Seven patients received definitive therapy within 3 months after initiation of MMI therapy, because three of these patients developed hypersensitivity reactions to the medication. In the other four cases, definitive therapy was requested by the patient's family after initial stabilization.

Table 1.

Clinical characteristics of cohort at diagnosis and start of GD therapy (T0)

	n (%)	Mean ± sd	Range
Male (%)	8 (22.9)
Female (%)	35 (77.1)
Age at diagnosis (yr)		10.9 ± 3.8	4.0–18.5
Follow-up (months)		37 ± 23	9–101
Weight (kg)		41.2 ± 18.6	14.9–86.5
Weight z-score		−0.01 ± 1.01	−2.37–2.1
Height (m)		1.45 ± 0.21	1.03–1.83
Height z-score		0.02 ± 0.75	−0.6–0.52
BMI (kg/m2)		18.4 ± 3.8	13.3–29.2
BMI z-score (sd)		−0.02 ± 1.05	−2.57–2.24
BMI percentile (%)		48.8 ± 30.2	0.5–98.8
Free T4 (ng/dl)		5.5 ± 2.8	2.1–14.4
Total T4 (μg/dl)		21.3 ± 10.8	9.4–70.9
TSH (mIU/liter)		0.03 ± 0.04	0.01–0.16

Normal free T₄ range is 1.0–2.2 μg/dl; normal total T₄ range is 5.0–10.6 μg/dl; and normal TSH range is 0.3–7.5 mIU/liter according to Lazar et al. (20).

Height and weight data were available for 3.6 ± 1.6 yr before GD diagnosis and for 3.1 ± 2.1 yr after treatment onset. Mean premorbid BMI z-scores, which were standardized from values 1 yr before diagnosis, were 0.25 ± 1.32 and were similar to BMI z-scores over the preceding 3 yr for all patients. These values were not significantly different from the normal population (BMI z-score = 0; P = 0.224). Twenty-nine patients had premorbid BMI z-scores within the normal range (−0.26 ± 0.85), eight patients were overweight (1.37 ± 0.22), and five patients were obese (2.19 ± 0.46).

At the onset of GD therapy, 30 patients had decreased BMI z-scores, and 13 patients had increased BMI z-scores compared with 1 yr or more before the diagnosis of GD. For children that had premorbid BMI z-scores within normal ranges (third to 85th percentile, n = 29), there were no significant differences between premorbid and T0 BMI z-scores (P = 0.787). For those patients who were overweight (n = 8) or obese (n = 5) more than 1 yr before GD diagnosis, BMI z-scores decreased significantly (P < 0.05) over the period before diagnosis.

At the start of treatment (T0), the average BMI was 18.4 ± 4.0 kg/m². This value corresponds with an average BMI z-score of −0.02 ± 1.05 and is not statistically significantly different from the normal population (BMI z-score = 0; P = 0.952) (Fig. 1). Most patients were within normal range of BMI percentiles (n = 36) at treatment onset. One patient was underweight (BMI z-score < third percentile), five patients were overweight (BMI z-score ≥85th percentile), and one patient was obese (BMI z-score ≥95th percentile) at treatment onset.

Fig. 1.

BMI z-scores over time. Whisker plots show BMI z-scores. The upper and lower edges of the boxes represent the 25–75th percentiles, respectively. The horizontal line within the box represents the median. The entire range of values is also shown. The number of patients (n) with data available at each time point are shown at the bottom of the figure. Wilcoxon signed rank test: *, P < 0.01 vs. T0 BMI z-scores; **, P < 0.01 vs. premorbid BMI z-scores. Premorbid is defined as 1 yr before GD diagnosis.

At 3 months after treatment onset, the average BMI was 19.8 ± 4.0 kg/m², which corresponds with an average BMI z-score of 0.45 ± 0.93 and indicates a significant increase in BMI z-scores after initiation of therapy (P < 0.0001). The absolute increase in body weight over this period was 2.9 ± 3.3 kg. Thirty-two patients gained weight over this period (range, 0.5–8.0 kg), and six patients lost weight (range, 0.2–1.8 kg) (Fig. 1). At 3 months, BMI z-scores did not exceed premorbid values (P = 0.512).

BMI z-scores increased until 6 months after treatment onset (mean BMI z-score = 0.79 ± 0.81) and then stabilized at significantly higher values than at T0 (χ² = 22.918; P < 0.0001). Compared with the premorbid BMI z-scores, the BMI z-scores were significantly higher at 1 yr after start of therapy (χ² = 13.133; P = 0.01). Post hoc analysis revealed that most of the weight gain that occurred after treatment onset occurred from 0–3 months (P < 0.0001) and 3–6 months (P < 0.001).

Characteristics of BMI gain in subsets

We observed three major patterns in weight change after treatment onset (Fig. 2). Twenty-two patients (51.1%) had follow-up BMI z-scores that remained within normal ranges (group 1). Their follow-up BMI z-scores never exceeded premorbid z-scores (χ² = 5.571; P = 0.473).

Fig. 2.

BMI z-score gain in patient subsets. Whisker plots show BMI z-scores. The upper and lower edges of the boxes represent the 25–75th percentiles, respectively. The horizontal line within the box represents the median. The entire range of values is also shown. The number of patients (n) with data available at each time point is shown at the bottom of the figure. Group 1 includes patients that are within normal BMI range premorbid, and during follow-up (BMI z-score within third to 85th percentile). Group 2 includes patients that are overweight (BMI z-score ≥85th percentile) or obese (BMI z-score ≥95th percentile) premorbid and during follow-up. Group 3 includes patients that became overweight (BMI z-score ≥85th percentile) or obese (BMI z-score ≥95th percentile) after start of GD therapy. Wilcoxon signed rank test with Bonferroni adjustment: *, P < 0.001; **, P < 0.01. Premorbid is defined as 1 yr before GD diagnosis.

Eleven patients (25.6%) had premorbid BMI z-scores that were overweight or obese. These patients regained their weight after initiation of GD therapy at approximately 9 months, and BMI z-scores stabilized and remained at premorbid values (group 2).

Ten patients (23.3%) became overweight or obese after initiation of GD therapy (group 3). These patients had lower BMI z-scores than group 2 premorbid (P < 0.001), but did not have higher BMI z-scores than group 1 premorbid (P = 0.272).

Gender

Premorbid BMI z-scores were not different between males (0.02 ± 2.07) and females (0.29 ± 1.12; P = 0.864). BMI z-scores at 12 months were higher than BMI z-scores at T0 for both males and females (P < 0.05). After this time point, females remained with high BMI z-scores (P < 0.035), whereas males returned to their baseline (T0) BMI z-scores. BMI z-scores for males never exceeded their premorbid BMI z-scores, whereas female BMI z-scores tended to be higher at all times after treatment onset (Fig. 3).

Fig. 3.

BMI z-score for gender. Whisker plots show BMI z-scores. The upper and lower edges of the boxes represent the 25–75th percentiles, respectively. The horizontal line within the box represents the median. The entire range of values is also shown. The number of patients (n) with data available at each time point is shown at the bottom of the figure. Wilcoxon signed rank test: *, P < 0.01 vs. T0 BMI z-score; **, P < 0.05 vs. premorbid BMI z-score. Premorbid is defined at1 yr before GD diagnosis.

Therapy

Seventeen patients received ¹³¹I, and four patients underwent surgery after treatment with PTU or MMI. Including those patients who developed toxic reactions to medication soon after treatment onset, the time after initial drug therapy to definitive therapy was 11.5 ± 10.7 months. Excluding the patients with early toxic reactions (within 1 month after treatment onset, n = 3), the time from drug treatment onset to definitive therapy was 13.2 ± 11.1 months. BMI z-scores for patients receiving definitive therapy at start of medical therapy were 0.01 ± 1.03 and at start of definitive therapy 0.60 ± 0.80 (P = 0.116). Gains in BMI z-scores 3 months after definitive therapy were not greater than those who received definitive treatment initially (n = 3; P > 0.617).

Age

The mean age at diagnosis and start of therapy was 10.9 ± 3.8 yr. For purposes of analysis, children were divided into two groups according to age at diagnosis: under 11 yr of age (n = 23) and 11.0 yr of age or older (n = 20). There were no differences in BMI z-scores for the different age groups premorbid (P = 0.827) and at T0 (P = 0.205). Over time, the younger patients had significantly higher BMI z-scores than at T0 (P < 0.028), but these values did not exceed premorbid values (P > 0.10). The older patients had BMI z-scores higher than T0 up to 9 months (P < 0.01). Thereafter, the BMI z-scores returned to baseline. None of the recorded BMI z-scores for the older patients were higher than premorbid BMI z-scores (Fig. 4).

Fig. 4.

BMI z-scores by age. Whisker plots show BMI z-scores. The upper and lower edges of the boxes represent the 25–75th percentiles, respectively. The horizontal line within the box represents the median. The entire range of values is also shown. The number of patients (n) with data available at each time point is shown at the bottom of the figure. Wilcoxon signed rank test: *, P < 0.05 vs. T0. Premorbid is defined at 1 yr before GD diagnosis.

Thyroid function

At presentation, all patients were hyperthyroid (Table 1). At 3 months after diagnosis, 17 patients were hyperthyroid (39.5%), 18 patients (41.9%) were euthyroid, and four patients (7.5%) were hypothyroid. At 6 months, the fT₄ and TT₄ levels of all patients normalized, and TSH levels ranged from normal to abnormal (Fig. 5).

Fig. 5.

Thyroid function tests over time. Scatter plot is of all thyroid hormone test results. Bars represent median values. Shaded areas represent the normal range of values (TT₄, 5.0–10.6 μg/dl; fT₄, 1.0–2.2 μg/dl; TSH, 0.3–7.5 mIU/liter) according to Lazar et al. (19).

For all follow-up blocks and times, changes in BMI z-scores (ΔBMI z-score), TSH levels (ΔTSH), and fT₄ and TT₄ levels (ΔfT₄ and ΔTT₄) were calculated. Only over the period of 3–6 months was there a positive correlation between ΔBMI z-score and ΔTSH levels (r = 0.682; P = 0.007), a negative correlation between ΔfT₄ and ΔBMI z-score (r = −0.671; P = 0.034), and a negative correlation for ΔTT₄ and ΔBMI z-score (r = −0.632; P = 0.021). No correlations were found between fT₄ or TT₄ and BMI z-scores at all other time points.

When we examined TSH levels over time, we observed that TSH levels were either suppressed or within the normal range for age in 98% of determinations. In four individuals, TSH elevations greater than 25 mU/liter were observed once or twice 12 months after therapy onset. Each of these individuals had been treated by surgery or radioactive iodine and reported poor compliance over the month or more before laboratory testing. When weight gain was examined over each of the periods associated with TSH elevations, no significant increases in weight gain were observed.

Discussion

It has long been recognized that thyroid hormones influence the metabolic state with hyperthyroidism increasing basal metabolic rates (5). It is recognized that hyperthyroid patients may experience weight loss as the condition is developing (20). Some individuals, however, may not present with weight loss, presumably due to increased caloric intake (9). Should caloric intake remain constant as the hypermetabolic state ebbs, it is predictable that body weight will increase. Studying a cohort of children, we now show that treatment of GD is associated with increases in body weight, and in about 25% of children with GD the excessive weight persists for at least 3 yr.

In the children studied, it was interesting that BMI z-scores at the time GD was diagnosed were comparable to those over the preceding years when the children were presumably not affected with GD (9, 15, 21). Increased caloric intake during hyperthyroid states has been reported in humans (8, 22) and in animals (23), with a more than 50–100% increase in caloric consumption observed. We thus presume that increased caloric consumption while the children were hyperthyroid was sufficient to prevent weight loss.

Over the first 3 months after treatment onset, considerable weight gain was seen as the hyperthyroid state improved but while the majority of children were biochemically hyperthyroid. Weight gain continued until 6 months, after which BMI z-scores stabilized. These data show that the vulnerable period for weight gain spans the period when the hyperthyroid state is undergoing biochemical resolution shortly after treatment onset. In adults who have been treated for GD, it has been observed that early weight gain is greatest in subjects in whom T₄ values normalize quickly (6). This observation is reflected in our findings, because there was a strong correlation between thyroid hormone level changes and changes in body weight and BMI z-scores over the 3–6 months after initiation of treatment.

When we assessed demographic factors associated with weight gain over the first 6 months of treatment, we found that children who were less than 11 yr of age at diagnosis experienced the greatest weight gain. Young girls were more likely than young boys to experience weight gain as well. Because some patients were treated with different forms of therapy over the first 6 months, we assessed the potential contributing roles of different forms of therapy. Comparable patterns of weight gain were seen irrespective of therapy, suggesting that correction of the hyperthyroid state is the key to excessive weight gain rather than the treatment modality. It is also important to note that during periods of transient TSH elevations 1 yr after treatment onset, excessive weight gain was not triggered. Such elevations were seen in the children on replacement levothyroxine therapy after definitive treatment and not in the children on antithyroid drugs.

In the lone other pediatric study of this issue, increases in weight were seen by the second office visit after the onset of GD therapy. Although the time period was not defined, and there were no long-term follow-up data, weight gain was seen in this other study (15). Our data expand upon these observations showing that weight gain is persistent in about 25% of the children.

Due to the retrospective character of this report, we cannot assess whether the weight gain is due to increased caloric intake or decreasing metabolic expenditure as the hyperthyroidism improves. Studies of hyperthyroidism show that high basal metabolic rate improves with the induction of euthyroidism. In adults, basal metabolic rate decreases from 2087 kcal/24 h at diagnosis to 1601 kcal/24 h at 12 months of euthyroidism have been observed (24). We are unaware of comparable studies in children.

Increasing data show that increased body weight in children is associated with an increased risk of cardiovascular disease and diabetes in adulthood (25–27). Thus, because 24% of our patients became overweight after treatment, if the increases in BMI persist, these findings are of potential long-term importance. Adults with GD are recognized to be at an increased risk of cardiovascular disease (28). At present, it is not known whether excess body mass potentially contributes to this problem and whether post-GD-treatment weight gain is a contributory factor.

We recognize several limitations of our study. We did not have comparable follow-up data on all patients, because although we had follow-up data on 27 and 24 patients at 1 and 2 yr, respectively, we had follow-up data at 3 yr in only 14 individuals. Thus, it will be important for other groups to evaluate this issue. In addition, because we did not have data on pubertal status before diagnosis, it was not possible for us to examine outcomes as related to pubertal development. Because of the retrospective nature of our study, we also did not have data about body composition to assess whether changes in weight represent increased fat mass. In prospective future studies, it will be interesting to examine this issue.

Considering our observations, we suggest that weight be carefully followed after GD treatment onset. Dietary counseling should be considered in the hopes of mitigating the rise in weight after GD treatment onset. Thus, in addition to managing the hyperthyroid state and monitoring for adverse effects of therapy, we must now add weight gain management to the duties of the treating practitioner.