Graves’ Hyperthyroidism (GH), caused by Graves’ disease (GD), is a common endocrine disorder characterized by excessive thyroid hormone production (1–3). The treatment options for GH include antithyroid drugs, surgery, and Radioactive iodine (RAI) therapy. Antithyroid drugs (ATD), particularly methimazole, are currently considered the preferred treatment option (4–6). However, in certain circumstances such as liver disease, congestive heart failure, advanced age with comorbidities, or periodic paralysis, RAI may be preferred (4, 5).

For both patients and clinicians, the non-remission rate is a crucial factor in determining the choice of treatment. The relapse rates are 52-53% following ATD, 8-15% following RAI, and 0-10% following thyroidectomy (7, 8). Although, RAI therapy is a widely used treatment due to its effectiveness, convenience, and low price, various factors can impact the non-remission rate of GH. Therefore, it is important to identify the risk factors for non-remission and to develop strategies to minimize the probability of non-remission. However, to our knowledge, there is lack of study to develop a dynamic prediction model for non-remission in GH treated with RAI.

Nomogram is a valuable statistical tool that can transform predictive models into visual graphs, providing a numerical probability of a specific clinical event (9). Therefore, this study aimed to develop an online dynamic nomogram to assist physicians in providing a personalized therapy for GH.

Over the years, RAI therapy has been one of the most important methods for treating GH (4). Many physicians prefer to use a large dose of the isotope to achieve early hypothyroidism and avoid the need for another dose of iodine-131. However, there are still 8-15% of patients who do not achieve complete remission after the initial treatment and require a second round of RAI therapy 3-6 months later. Numerous studies have been conducted to determine the ideal dosage of iodine-131 that would increase the likelihood of achieving euthyroid in patients. However, these studies often overlooked other predictive factors that could influence the treatment outcome (13–15). In our study, we proposed four predictive factors: RAIU_24h, Teff, TID, and IDPG, and used these four factors to construct a nomogram model for predicting non-remission after the first RAI therapy.

RAIU_24h plays a crucial role in determining the iodine-131 dose and is included in various iodine-131 dose calculation formulas. Numerous studies have been conducted to investigate the relationship between RAIU_24h and the efficacy of RAI therapy, both domestically and internationally. In our study, we observed patients with high RAIU_24h had better outcomes with RAI therapy, which aligns with the conclusions of studies such as NordykeRA and NordykeRA, where higher RAIU_24h was associated with higher cure rates (16–18). Contrary to our findings, some studies have shown a negative correlation between the cure rate of RAI therapy for GH and RAIU_24h, indicating that higher RAIU_24h is associated with lower cure rates (15, 19–21). There are several possible explanations for this phenomenon. Firstly, there may be an inverse relationship between the sensitivity of thyroid tissue to radiation and RAIU_24h, as demonstrated in the study by Di Martino F et al (22). Secondly, the iodine-131 dose for treatment decreases as RAIU_24h increases, and low therapeutic dose of iodine-131 may reduce the uptake of iodine-131 in thyroid tissue (23, 24). Despite the inconsistency in results, RAIU_24h is still considered an important factor influencing the success of RAI therapy.

Teff refers to the time required for the radioactive nuclide iodine-131 to decrease to half of its original amount in the body, due to the combined effects of physical decay and biological activity. It reflects the duration of iodine-131 residence in thyroid tissue, which directly affects the actual absorbed dose of beta radiation in thyroid tissue, thereby influencing the extent of damage to the thyroid tissue by iodine-131. Yu et al. (11) found that a longer Teff increases the likelihood of developing hypothyroidism, while other study (25) reported no significant difference in Teff between the remission and non-remission groups after RAI therapy. In this study, both univariate and multivariate logistic regression analyses found that shorter Teff is associated with non-remission, which is consistent with the current general view. Shorter Teff reflects a high metabolic state of thyroid cells, resulting in a shorter residence time and lower absorbed dose of thyroid tissue (26).

The IDPG and TID can impact the actual absorbed dose of the thyroid tissue and therapeutic effect of RAI. Previous studies have shown that lower do IDPG result in higher failure rates (27). In both the univariate and multivariate logistic regression analyses in this study, the TID and IDPG identified as positive risk factors affecting treatment outcomes. This is consistent with the general consensus (28). We speculate that when higher TID, especially higher IDPG, are administered, it ensures sufficient beta radiation within the thyroid tissue to exert a destruction, resulting in satisfactory treatment outcomes. Currently, the typical IDPG in clinical practice is 70-150 µCi/g, but evidence suggests that in order to achieve a hypothyroid state in patients with GH, the IDPG needs to be greater than 150 µCi/g (29, 30). The ATA guideline recommends an IDPG of 50-200µCi in clinical practice, while the Chinese guideline suggests 70-150µCi for IDPG (4, 10). However, there is evidence suggesting that in order to achieve hypothyroid state, the IDPG needs to be greater than 15 MBq (405 µCi) or dosimetry aiming at delivering at least 300 Gy to the thyroid tissue for functional ablation (31, 32). Peters’ research has shown that RAI efficacy is demonstrated for glands up to 80 ml, and it is important to note that the patients presented in our study fit within previously shown gland sizes for RAI efficacy (<80 gm). In this study, glands were in the modestly enlarged range, and glands >80 gm were generally excluded. These findings are in agreement with those reported by Peters (31, 32). It is worth noting that in Table 2 , the TID range in the validation group was 9.1 ± 5.0 mCi, which is notably lower than the commonly used fixed dose of 555 MBq (15 mCi). This preference for lower treatment doses is influenced by patient-centered or cultural considerations, aiming to restore euthyroidism rather than induce hypothyroidism in the initial management of GH using single RAI therapy. The choice of treatment target may differ from practices in other regions, such as the United States, where the standard approach is to use RAI doses that induce hypothyroidism. It should be emphasized that using lower doses aimed at restoring euthyroidism may result in lower long-term remission rates and could potentially limit the applicability of the model.

Through univariate and multivariate analyses, these four factors were ultimately identified as independent factors affecting the non-remission rate after initial RAI therapy in GH patients. However, in order to more accurately predict the probability of non-remission, we did not choose to use a single independent factor for prediction, but instead combined these four factors to construct a more predictive nomogram model. As shown in ROC curve analysis, it can be clearly seen that the nomogram model performed better in terms of predictive ability compared to single independent factor, both in the training set and validation set. Additionally, the nomogram model also demonstrated excellent consistency and clinical net benefit in calibration curve and DCA curve analysis. By applying this model, we can guide the adjustment of dosage before RAI administration to achieve a lower rate of non-remission and reduce the chances of a second RAI therapy.

However, it is important to acknowledge the limitations of our study. Firstly, this was a retrospective study with some inevitable bias. Secondly, the sample size of two center was small. The novel nomogram model was only validated internally. Therefore, a multicenter clinical study with a larger sample size could be conducted for external validation to verify the clinical benefits in the subsequent research. Thirdly, the patients included in this study had moderately enlarged thyroid glands, not markedly enlarged glands. This may limit the applicability of our model in treating GH patients with very large goiters. Nevertheless, this novel nomogram model does have the potential to improve clinical practice of personalized therapy.

This study was the first to develop and validate online nomogram based on the independent risk factors to dynamically predict non-remission probability in individuals with GH after initial RAI therapy. This novel model demonstrated superior performance and discriminative power, providing vital information for nuclear medicine physicians to design customized clinical treatments for their patients.