Early prediction of transient versus permanent congenital hypothyroidism: a retrospective cohort study

Article information

Ann Pediatr Endocrinol Metab. 2026;31(1):38-44

Publication date (electronic) : 2026 February 28

doi : https://doi.org/10.6065/apem.2550066.033

Myung Ji Yoo¹, Ji-Eun Lee¹, Eun Young Joo¹, Jisun Park¹, Young Ju Suh², Su Jin Kim^,1

¹Department of Pediatrics, Inha University Hospital, Incheon, Korea

²Department of Biomedical Sciences, Inha University College of Medicine, Incheon, Korea

Address for correspondence: Su Jin Kim Department of Pediatrics, Inha University Hospital, 27 Inhang-ro, Jung-gu, Incheon 22332, Korea Email: kimsjped@inha.ac.kr

Received 2025 February 25; Revised 2025 May 20; Accepted 2025 May 21.

Abstract

Highlights

· Levothyroxine (LT4) dose requirements at 6, 12, and 24 months may help differentiate transient from permanent congenital hypothyroidism in patients with gland in situ.

· Lower LT4 doses at these time points were associated with a higher likelihood of transient disease.

· Incorporating birth weight and thyroid-stimulating hormone (TSH) modestly improved prediction, supporting consideration of earlier treatment re-evaluation.

Introduction

Congenital hypothyroidism (CH) is the most common endocrine disorder in newborns and, if left untreated, can result in irreversible neurodevelopmental delays [1,2]. The global prevalence of CH has increased by approximately 52% (confidence interval, 4%–122%) when comparing 1969–1980 with 2011–2020 [3]. This increase can be attributed to the expansion and improvement of newborn screening programs, including adjustments in screening strategies, such as lower thyroid-stimulating hormone (TSH) cutoff levels, and increased geographical and socioeconomic coverage [3,4]. Additionally, advances in reproductive technology have contributed to an increased number of premature and low-birth-weight infants, which may influence the prevalence of CH [5-7]. Environmental and genetic factors also contribute to CH development. Environmental factors such as iodine imbalance and endocrine-disrupting chemicals can alter thyroid hormone synthesis [8,9]. Mutations in the TSHR, DUOX2, or PAX8 genes may also impair thyroid development and function [10,11].

For patients with CH and confirmed as having a gland in situ (GIS), guidelines recommend re-evaluation of the hypothalamic-pituitary-thyroid axis at 2–3 years of age [2,12]. However, recent studies suggest that early treatment discontinuation should be considered in selected patients, particularly those with GIS, no first-degree family history of CH, and low levothyroxine (LT4) requirements [12-15]. Therefore, this study aimed to evaluate the ability of LT4 doses at 6, 12, and 24 months to differentiate transient CH (TCH) from permanent CH (PCH) and assess whether incorporating additional clinical and biochemical parameters, such as TSH levels and birth weight, improves predictive accuracy. By identifying reliable early markers, we aimed to determine when the discontinuation of treatment can be safely considered.

Materials and methods

1. Subjects

This retrospective cohort study was conducted at a single tertiary care center (Inha University Hospital, Korea) between January 2013 and December 2022. The study protocol was approved by the Institutional Review Board of Inha University Hospital (approval No. 2023-09-032). Initially, 212 pediatric patients diagnosed with CH were identified. The diagnosis of CH was based on abnormal thyroid function test (TFT) results (low free T4 and/or elevated TSH levels >10 mIU/L) and the presence of a normal thyroid gland (eutopic thyroid) confirmed by thyroid ultrasonography and/or scintigraphy. Eligibility criteria included infants with confirmed CH and GIS. Exclusions were patients lost to follow-up (n=16), incomplete medical records (n=18), age <3 years or follow-up <3 years (n=71), and thyroid dysgenesis (n=2). After applying these criteria, 105 patients were eligible and subsequently categorized into TCH (n=70; those maintaining normal thyroid function without LT4 after withdrawal at 3 years) and PCH (n=35; those requiring continuous LT4) (Fig. 1). The final sample size was based on the number of eligible cases during the study period; no formal sample size calculation was undertaken. Clinical data included gestational age, birth weight, age at CH diagnosis, LT4 dose at diagnosis and during follow-up (6, 12, 24, and 36 months), and family history of thyroid disease. Outcome was defined as maintenance of normal thyroid function without LT4 beyond 3 years (TCH) versus the need for continuous LT4 (PCH), while the exposure was the LT4 dose requirement at specified time points. Other variables included baseline TSH and free T4 levels.

Fig. 1.

Flow diagram of study population. CH, congenital hypothyroidism; GIS, gland in situ.

Results

1. Clinical characteristics of the study population

This study included 105 patients, with 70 diagnosed with TCH and 35 with PCH (Table 1). The male-to-female ratio did not differ significantly between groups (TCH, 41:29 vs. PCH, 20:15; P=1.000). Low birth weight (<2,500 g) was significantly more frequent in the TCH group (54.2% [38 of 70]) than in the PCH group (17.1% [6 of 35], P<0.001). Similarly, preterm birth (<37 weeks) was significantly more common in the TCH group (58.6% [41 of 70] vs. 28.6% [10 of 35], P=0.004). A family history of thyroid disease was present in 21.0% (22 of 105) of all patients, with no significant intergroup difference (TCH, 18.6%; PCH, 25.7%; P=0.396). The mean age at LT4 treatment initiation was 27.3±19.0 days, with no significant intergroup difference (TCH, 29.6±20.3 days vs. PCH, 22.0±14.5 days, P=0.068).

Table 1.

Patient characteristics by study group

3. ROC curve and multivariate logistic regression analysis

ROC analysis evaluated the ability of the LT4 dose per body weight at 6, 12, and 24 months to distinguish between the TCH and PCH groups (Fig. 2, Table 2). At 6 months, the area under the curve (AUC) was 0.649 (95% confidence interval [CI], 0.536–0.762), with an optimal cutoff value of 3.25 μg/kg, sensitivity of 67.7%, and specificity of 57.8%. At 12 months, the AUC was 0.746 (95% CI, 0.643–0.849), with an optimal cutoff value of 3.30 μg/kg, sensitivity of 51.7%, and specificity of 92.9%. At 24 months, the AUC was 0.794 (95% CI, 0.699–0.889), with an optimal cutoff value of 3.02 μg/kg, sensitivity of 53.6%, and specificity of 94.3%. Fig. 2 presents the combined ROC curves for the LT4 dose at each time point. A logistic regression model incorporating birth weight, TSH level, and LT4 dose per body weight was used to generate adjusted ROC curves (Fig. 3). In this model, the AUC at 6 months was 0.740 (95% CI, 0.637–0.843), with a sensitivity of 95.2% and a specificity of 60.4%. At 12 months, the AUC was 0.782 (95% CI, 0.685–0.879), with a sensitivity of 95.2% and specificity of 64.6%. At 24 months, the AUC was 0.833 (95% CI, 0.745–0.921), with a sensitivity of 90.5% and specificity of 66.7%. DeLong test compared the AUC values between time points (Table 3). We found no significant differences in any of the comparisons (6 months vs. 12 months: P=0.127; 6 months vs. 24 months: P=0.052; 12 months vs. 24 months: P=0.177).

Fig. 2.

Receiver operating characteristic (ROC) curves of various thresholds of levothyroxine dose per body weight (μg/kg) at 6, 12, and 24 months of age for distinguishing between transient congenital hypothyroidism and permanent congenital hypothyroidism. The area under the curve values were 0.649 (95% confidence interval [CI], 0.536–0.762) at 6 months, 0.746 (95% CI, 0.643–0.849) at 12 months, and 0.794 (95% CI, 0.699–0.889) at 24 months.

Table 2.

Summary of receiver operating characteristic curve analysis of levothyroxine doses at 6, 12, and 24 months of age

Fig. 3.

Receiver operating characteristic (ROC) curves of various thresholds of thyroid-stimulating hormone, birth weight, and levothyroxine dose per body weight for predicting transient congenital hypothyroidism at 6 months (area under the curve [AUC], 0.740; 95% confidence interval [CI], 0.637–0.843), 12 months (AUC, 0.782; 95% CI, 0.685–0.879), and 24 months (AUC, 0.833; 95% CI, 0.745–0.921).

Table 3.

DeLong test results for area under the curve comparisons at different time points

Multivariate logistic regression models demonstrated that LT4 dose was a consistent and significant predictor of TCH at 6, 12, and 24 months (odds ratio [OR]: 0.44, 0.44, and 0.17; P=0.018, 0.011, and 0.001, respectively). In contrast, TSH levels were not significantly associated with TCH at any time point. Birth weight showed a statistically significant association at 6 and 24 months, and borderline significance at 12 months (Table 4). Multicollinearity was not observed in any of the multivariate models (all VIFs <1.4). To evaluate potential effect modification by TSH, interaction terms between LT4 dose and TSH were tested in logistic regression models at each time point. The interaction term was statistically significant at 12 months (OR, 0.973; 95% CI, 0.948–1.000; P=0.046), while no significant interaction was found at 6 or 24 months (Table 5).

Table 4.

Multivariate logistic regression results

Table 5.

Multivariate logistic regression with LT4 × TSH interaction term

Discussion

This study aimed to identify predictive factors that differentiate TCH from PCH in infants using GIS. Several studies investigated LT4 dose thresholds as predictors of TCH [13,14,16-18]. Nagasaki et al. [15] proposed that early re-evaluation and treatment discontinuation might be feasible at 12 months for patients requiring <1.7 μg/kg/day at 12 months or <1.45 μg/kg/day at 24 months. Similarly, Özer et al. identified a newborn TSH<45 mIU/L and an LT4 dose <2.0 μg/kg/day at discontinuation as predictors of TCH [17]. Although these findings support the use of LT4 dose thresholds as useful indicators, their applicability to different populations remains uncertain. Our findings suggest that LT4 dose requirements over time are important indicators for distinguishing between TCH and PCH. However, their diagnostic accuracy remains moderate, necessitating a multifactorial approach that incorporates biochemical and clinical parameters. An LT4 dose <3.25 μg/kg/day at 6 months could serve as an indicator of TCH with moderate specificity and sensitivity consistent with previous findings [14,16]. The ENDO-European Reference Network guidelines [12] also support the use of biochemical markers, including LT4 dose thresholds, to guide early treatment cessation while ensuring patient safety. However, our results indicate that relying solely on LT4 dose thresholds may be insufficient, as moderate sensitivity and specificity necessitate complementary diagnostic tools, such as TSH levels, thyroid ultrasound, and growth patterns [15,17]. To further evaluate the diagnostic accuracy of the LT4 dose, we performed a logistic regression analysis of birth weight, TSH, and LT4 dose per body weight. The ROC curve analysis showed an increase in AUC values compared to LT4 dose alone (6 months: 0.649→0.740; 12 months: 0.746→0.782; 24 months: 0.794→0.833) with improved sensitivity and specificity. However, DeLong test revealed no statistically significant differences between time points (6 months vs. 12 months: P=0.127; 6 months vs. 24 months: P=0.052; 12 months vs. 24 months: P=0.177), suggesting that while the adjusted model provides better overall classification performance, the improvement over time was not statistically significant. Multivariate logistic regression models confirmed that LT4 dose was a significant predictor of TCH at 6, 12, and 24 months, while TSH was not significantly associated with the outcome. Birth weight also demonstrated predictive value at 6 and 24 months, but showed borderline significance at 12 months. This may reflect differences in the relative influence of birth weight across time points, as the LT4 dose becomes a stronger predictor around 12 months. Notably, an interaction between LT4 and TSH was statistically significant at 12 months, suggesting that TSH may modify the predictive value of LT4 at that time point. No significant interactions were observed at 6 or 24 months. This highlights the need for a stable and early predictive model rather than relying on LT4 dose adjustments over time.

Compared with international studies, domestic research suggests slightly higher LT4 dose thresholds at the time of treatment discontinuation [13,16,18]. Studies published in the Annals of Pediatric Endocrinology and Metabolism [16,18] reported LT4 dose thresholds for treatment discontinuation. Our institution follows a conservative approach and rarely escalates the LT4 doses unless TFT abnormalities emerge. This may explain the relatively lower LT4 dose requirements observed at 12 and 24 months compared with other domestic studies [16,18]. Although the LT4 dose alone provided moderate diagnostic accuracy, the lack of statistical significance in AUC improvement over time suggests that no single parameter is sufficient for determining the early discontinuation of treatment. Instead, a multifactorial approach that incorporates biochemical and clinical markers may be a more reliable strategy for assessing the feasibility of ending treatment earlier. While LT4 dose thresholds can serve as a useful initial screening tool, additional factors such as baseline TSH levels, thyroid imaging findings, and genetic factors should be considered when considering an earlier discontinuation. Our findings support the clinical utility of LT4 dose thresholds for guiding early treatment discontinuation in infants with TCH. Patients without genetic abnormalities, a family history of thyroid disease, or high LT4 requirements may benefit from early cessation, thus reducing the unnecessary burden of medication. However, given the moderate diagnostic performance of LT4 thresholds, caution is warranted in borderline cases, particularly in those with elevated initial TSH levels [15,16,19].

Despite these promising findings, this study has several limitations. As a retrospective analysis, it was constrained by the availability of medical records, particularly regarding family history, which showed no significant intergroup difference despite being recognized as a risk factor for PCH in international guidelines [12,15]. Additionally, we did not assess the long-term outcomes following the early discontinuation of treatment, which remains an important area for future research. Future studies should incorporate prospective designs with genetic and imaging data, because thyroid gland morphology and genetic mutations (e.g., DUOX2, TG, TSHR mutations) may provide additional insight into long-term thyroid function outcomes [15,20,21]. Establishing standardized LT4 dose adjustment protocols tailored to different clinical settings will help refine treatment strategies and ensure patient safety.

Article information Continued