Clinical outcome and influencing factors of differentiated thyroid cancer with biochemical incomplete response

ABSTRACT

Purpose

The aim of our study was to investigate the clinical outcomes and predictive factors in patients with biochemical incomplete response (BIR) after initial radioiodine therapy (RAI).

Materials & Methods

This retrospective study enrolled 198 patients with BIR from two institutions and stratified them into a training cohort (Qingdao cohort, n = 144) and an external validation cohort (Beijing cohort, n = 54). The patients were classified as having incomplete response (IR) or non-IR disease at the final follow-up. Univariate and multivariate analyses identified IR predictors in the training cohort. A nomogram was developed using the training cohort and was validated in the validation cohort using calibration curves, decision curve analysis (DCA), and receiver operating characteristic analysis.

Results

Age, recurrence risk, pre-stimulated thyroglobulin (ps-Tg), and BRAF^V600E mutation were independent predictors of IR in the training cohort. The nomogram identified ps-Tg as the strongest predictor of IR risk, followed by the BRAF^V600E mutation, age, and recurrence risk. The calibration curves demonstrated excellent agreement between the predicted and observed probabilities. DCA confirmed favorable clinical utility.

Conclusions

This validated nomogram provides a clinically useful tool for quantifying IR risk in BIR patients, supporting personalized management decisions after initial RAI.

Plain Language Summary

In the present study, we followed 198 biochemical incomplete response (BIR) patients from two hospitals to understand which factors affect treatment outcomes. The research revealed four important predictors: age, recurrence risk category, pre-stimulated thyroglobulin (ps-Tg) levels, and the BRAFV600E gene mutation. Using these factors, we created an easy-to-use nomogram model tool that helps doctors to quantify the risk of incomplete response (IR). When tested with BIR patients from both hospitals, this tool proved reliable. It will help doctors better personalize treatment plans for thyroid cancer patients based on their individual test results and characteristics.

1. Introduction

The standard treatment for differentiated thyroid cancer (DTC) involves total thyroidectomy, followed by selective postoperative radioiodine therapy (RAI) and thyroid-stimulating hormone (TSH) suppression therapy [1,2]. Notably, RAI significantly reduces recurrence rates and improves overall survival in patients with DTC [1]. Thyroglobulin (Tg) has long been considered a valuable marker for detecting persistent disease during follow-up in patients with DTC who have received RAI [1,3–5].

Dynamic risk stratification (DRS) for patients with DTC was first introduced by Tuttle et al. [6], who revealed significant changes in the risk categories of patients with DTC. In short, the DRS system is based on post-treatment Tg and anti-Tg antibody (TgAb) levels, as well as imaging studies performed during follow-up. The DRS system was adopted according to the American Thyroid Association (ATA) 2015 guidelines [1], and therapeutic responses are classified into four categories: excellent response (ER), indeterminate response (IDR), biochemical incomplete response (IR; BIR), and structural IR (SIR). The DRS system is notably more effective in predicting long-term prognosis than either the American Joint Committee on Cancer/TNM staging system or the ATA risk stratification criteria [5,6].

The BIR category is defined as persistent abnormally stimulated Tg, suppressed Tg or elevated TgAb levels in the absence of structural disease [6]. BIR is relatively common, occurring in 11–22% of patients with DTC [1]. Up to 17–20% of patients may develop structural disease at the last follow-up [3,4]. Therefore, in this study, we aimed to evaluate the clinical outcomes and predictive factors in patients with BIR after initial RAI.

2. Materials & methods

2.1. Patients

This retrospective study was conducted at the Affiliated Hospital of Qingdao University. The study was approved by the Institutional Review Board of the Affiliated Hospital of Qingdao University, which waived the requirement for written informed consent.

The inclusion criteria were as follows: (i) patients who underwent total thyroidectomy with selective cervical lymph node dissection, (ii) patients with a histopathological diagnosis of DTC, (iii) patients with an intermediate- or high-risk of initial recurrence, (iv) patients who received only one standardized RAI between January 2014 and December 2022, and (v) patients whose therapeutic response was assessed as BIR at 6–12 months after initial RAI therapy [1,6]. Patients with positive anti-TgAb, concurrent malignancies, or receipt of other antitumor therapies were excluded. This exclusion criterion specifically addressed potential confounders from non-RAI interventions (e.g., targeted therapy or external beam radiation) that might independently influence Tg levels or imaging findings, thereby ensuring accurate assessment of initial RAI therapy outcomes.

The training cohort was obtained from the Affiliated Hospital of Qingdao University (Qingdao cohort), whereas the external validation cohort was obtained from Peking Union Medical College Hospital (Beijing cohort).

2.2. Collection of main epidemiological characteristics

The RAI therapy preparation protocol for DTC patients followed the 2015 ATA guideline recommendations [1]. All patients underwent standardized preparation, including levothyroxine withdrawal for 3–4 weeks and maintenance of a low-iodine diet, with TSH levels confirmed > 30 mIU/L prior to RAI therapy. Serum Tg was measured under these hypothyroid conditions one day before RAI administration and designated as pre-stimulated Tg (ps-Tg).

The main epidemiological characteristics collected included age, gender, histological type, tumor size, multicentricity, presence of the BRAF^V600E mutation, lymph node metastasis, initial recurrence risk, ps-Tg level, and TgAb status. Each patient’s disease stage was reclassified according to the eight edition of the American Joint Committee on Cancer TNM staging system [7].

2.3. Laboratory analysis

Tg and TgAb levels were quantified using an electrochemiluminescence immunoassay, following institutional protocols [2,8]. The sensitivity of Tg measurement was 0.04 ng/mL, with a detection range of 0.04 to 500 ng/mL. TgAb levels greater than 40 IU/mL were considered positive.

2.4. RAI protocol and follow-up

According to the 2015 ATA guidelines, all patients underwent total thyroidectomy followed by adjuvant RAI [1]. The administered ¹³¹I activity ranged from 100 to 150 mCi, depending on the patient’s disease status [3]. Planar therapeutic ¹³¹I whole-body scans and SPECT/CT were performed 3–7 days after ¹³¹I administration. The therapeutic response was first evaluated at 6–12 months after the initial RAI, in accordance with the 2015 ATA guidelines [1].

2.5. Definitions of clinical outcomes

At the last follow up, therapeutic responses were categorized as ER, IDR, BIR, or SIR based on the DRS classification criteria from the 2015 ATA guidelines [1]. Subsequently, patients with BIR or SIR were classified into the IR group, whereas those with ER or IDR were classified into the non-IR group [2,9,10].

2.6. Statistical analysis

Continuous variables are expressed as medians and ranges, whereas categorical variables are summarized as frequencies and percentages. The Mann-Whitney U test was used to compare ps-Tg levels between groups. Receiver operating characteristic (ROC) curve analysis was used to determine the optimal threshold for ps-Tg levels. Univariate and multivariate analyses of ps-Tg and other factors were performed using binary logistic regression analysis. A nomogram was developed using the training cohort to predict IR rates in patients with BIR after initial RAI, providing a clinical decision-making aid. The predictive performance was validated in the validation cohort.

Harrell’s concordance index (C-index) was used to quantify the discriminatory performance of the nomogram. The predictive ability was further assessed using ROC curve analysis, with performance quantified by the area under the curve (AUC). Calibration curves were used to visualize the agreement between nomogram-predicted probabilities and observed outcomes. Decision curve analysis was used to evaluate the nomogram’s clinical usefulness by illustrating the net benefit across a range of threshold probabilities. Nomogram development, C-index calculation, ROC analysis, calibration curve generation, and decision curve analysis were conducted using R software (version R4.2.1; http://www.r-project.org/). Statistical significance was set at P < 0.05, and all P-values were two-tailed.

3. Results

3.1. Baseline clinical characteristics

During the study period, 198 patients with BIR who met the inclusion criteria were enrolled. The Qingdao cohort (n = 144) constituted the training cohort, and the Peking cohort (n = 54) served as the validation cohort. The baseline clinical characteristics of patients with BIR are presented in Table 1. The baseline clinical data were similarly distributed between the training and validation cohorts.

Table 1.

Clinicopathologic characteristics of 198 DTC patients with BIR after initial RAI.

Characteristics	Cohort, no. (%)		χ2/U	P value
	Training (n = 144)	Validation (n = 54)
Gender			0.022a	0.881
Female	89 (61.8%)	34 (63.0%)
Male	55 (38.2%)	20 (37.0%)
Age (years)			0.055a	0.815
< 55	107 (74.3%)	41 (75.9%)
≥55	37 (25.7%)	13 (24.1%)
Multicentricity			0.055a	0.815
No	64 (44.4%)	23 (42.6%)
Yes	80 (55.6%)	31 (57.4%)
Histological type			0.351a	0.554
PTC	139 (96.5%)	53 (98.1%)
FTC	5 (3.5%)	1 (1.9%)
BRAFV600E mutations			0.167a	0.683
NO/NA	78 (54.2%)	31 (57.4%)
Yes	66 (45.8%)	23 (42.6%)
Tumor size (cm)	2 (1.2, 3)	2 (1.05, 3)	3803.5b	0.814
Recurrence risk			0.035a	0.852
Intermediate	98 (68.1%)	36 (66.7%)
High	46 (31.9%)	18 (33.3%)
T stage			0.818a	0.845
T1	47 (32.6%)	15 (27.8%)
T2	25 (17.4%)	9 (16.7%)
T3	40 (27.8%)	15 (27.8%)
T4	32 (22.2%)	15 (27.8%)
N stage			0.147a	0.701
N1b	113 (78.5%)	41 (75.9%)
N0-1a	31 (21.5%)	13 (24.1%)
ps-Tg (ng/mL)	39.91 (26.973, 73.478)	34.155 (27.45, 75.005)	3784.5b	0.773

Abbreviation: DTC: differentiated thyroid cancer; BIR: biochemical incomplete response; RAI: radioiodine therapy; PTC: papillary thyroid cancer; FTC: follicular thyroid cancer; NA: Not applicable; T: tumor; N: node; ps-Tg: pre-ablation stimulated thyroglobulin. ^ameans chi-squared test; ^bmeans Mann-Whitney U test.

3.2. Clinical outcomes of patients with BIR in the training and validation cohorts

At the last follow-up (median 49.5 months, range 7.3–103.4 months after initial RAI), the proportions of ER, IDR, BIR, and SIR in the training cohort were 20.8% (n = 30), 25.0% (n = 36), 32.0% (n = 46), and 22.2% (n = 32), respectively. In the validation cohort, ER, IDR, BIR, and SIR were observed in 18.5% (n = 10), 33.3% (n = 18), 24.1% (n = 13), and 24.1% (n = 13), respectively. IR was observed in 54.1% (n = 78) of the training cohort and 48.1% (n = 26) of the validation cohort.

3.3. Predictive value of ps-Tg in patients with BIR and IR in the training cohort

The ps-Tg level in the IR group was 56.5 (35.0, 127.6) ng/mL, which was significantly higher than the ps-Tg level of 29.7 (21.3, 47.2) ng/mL observed in the non-IR group (Figure 1(A)). ROC curve analysis of patients with BIR (Figure 1(B)) identified an optimal ps-Tg threshold of 51.2 ng/mL for predicting IR, with a corresponding sensitivity of 65.4%, specificity of 83.3% and an AUC of 0.774, respectively.

Figure 1.

(A) Comparison of the ps-Tg level for IR and non-IR in patients with BIR after initial RAI. (B) ROC curves of ps-Tg for detecting IR in patients with BIR after initial RAI. Abbreviation: ps-Tg: pre-ablation stimulated thyroglobulin; IR: incomplete response; non-IR: non-incomplete response; BIR: biochemical incomplete response; RAI: radioiodine therapy; ROC: receiver operating characteristic; AUC: area under the ROC curve; ***P＜0.001.

3.4. Risk factors for IR in the training cohort

Upon analyzing the relation between clinical characteristics and IR in patients with BIR after initial RAI, 10 factors were included in the univariate analyses (Table 2). The results indicated that patients aged ≥55 years (P < 0.001), those with multicentricity (P = 0.025), larger tumor size (P = 0.047), T4 stage (P = 0.015), positive BRAF^V600E mutation (P < 0.001), high recurrence risk (P < 0.001), or ps-Tg ≥51.2 ng/mL (P < 0.001) had a high likelihood of developing IR. By contrast, no significant differences were found for gender (P = 0.269), histological type (P = 0.790), or N stage (P = 0.051). Multivariate analysis revealed that age ( < 55 years/≥55 years) (odds ratio [OR]: 4.102, 95% confidence interval [CI]: 1.369–12.296, P = 0.012), recurrence risk (intermediate/high) (OR: 3.572, 95% CI: 1.318–9.677, P = 0.012), ps-Tg ( < 51.2 ng/mL/≥51.2 ng/mL) (OR: 10.813, 95% CI: 4.145–28.213, P < 0.001), and BRAF^V600E mutation (no or not available/yes) (OR: 5.821, 95% CI: 2.336–14.503, P < 0.001) were significantly associated with the risk of IR (Table 3).

Table 2.

Univariate analyses for IR in patients with BIR after initial RAI in the training cohort.

Characteristics	n	Non-IR group	IR group	χ2/U	P-value
Age (years)				14.529a	< 0.001
< 55	107	59(55.1%)	48(44.9%)
≥55	37	7(18.9%)	30(81.1%)
Gender				1.220a	0.269
Male	55	22(40.0%)	33(60.0%)
Female	89	44(49.4%)	45(50.6%)
Histological type				0.071a	0.790
PTC	139	64(46.0%)	75(54.0%)
FTC	5	2(40.0%)	3(60.0%)
Multicentricity				5.035a	0.025
Yes	80	30(37.5%)	50(62.5%)
No	64	36(56.3%)	28(43.8%)
Tumor size (cm)		1.6(1.2, 2.9)	2.0(1.2, 3.5)	2078.5b	0.047
T stage				10.489a	0.015
T1	47	25(53.2%)	22(46.8%)
T2	25	15(60.0%)	10(40.0%)
T3	40	19(47.5%)	21(52.5%)
T4	32	7(21.9%)	25(78.1%)
N stage				3.802a	0.051
N0-N1a	31	19(61.3%)	12(38.7%)
N1b	113	47(41.6%)	66(58.4%)
Recurrence risk				13.082a	< 0.001
Intermediate	98	55(56.1%)	43(43.9%)
High	46	11(23.9%)	35(76.1%)
BRAFV600E mutation				26.203a	< 0.001
Yes	66	15(22.7%)	51(77.3%)
NO/NA	78	51(65.4%)	27(34.6%)
Ps-Tg (ng/mL)				34.608a	< 0.001
< 51.2	82	55(67.1%)	27(32.9%)
≥51.2	62	11(17.7%)	51(82.3%)

Abbreviation: IR: incomplete response; non-IR: non-incomplete response; BIR: biochemical incomplete response; RAI: radioiodine therapy; PTC: papillary thyroid cancer; FTC: follicular thyroid cancer; T: tumor; N: node; NA: Not applicable; ps-Tg: pre-ablation stimulated thyroglobulin; ^ameans chi-squared test; ^bmeans Mann-Whitney U test.

Table 3.

Multivariate analysis for IR in patients with BIR after initial RAI using logistic regression in the training cohort.

Characteristics	P-value	OR	95%CI
Age ( < 55 years/≥55 years)	0.012	4.102	1.369–12.296
Recurrence risk (Intermediate/High)	0.012	3.572	1.318–9.677
Ps-Tg( < 51.2 ng/mL/≥51.2 ng/mL)	< 0.001	10.813	4.145–28.213
BRAFV600E mutation(NO or NA/Yes)	< 0.001	5.821	2.336–14.503
Constant	< 0.001	0.110

Abbreviation: IR: incomplete response; BIR: biochemical incomplete response; RAI: radioiodine therapy; ps-Tg = pre-ablation stimulated thyroglobulin; NA: Not applicable; OR: odds ratio; CI: confidence interval.

3.5. Prognostic nomogram for IR rates in patients with BIR after initial RAI

A nomogram model was developed to predict the probability of IR in patients with BIR after initial RAI. The model was based on independently associated risk factors identified through multivariate logistic regression analysis. Each variable was assigned a score, and the total score was calculated by summing the values of all selected variables. A vertical line drawn from the total score to the risk axis yielded the estimated probability of IR. Among the predictors, ps-Tg was the most significant contributor to IR risk, followed by BRAF^V600E mutation, age, and recurrence risk (Figure 2).

Figure 2.

Nomogram predicting IR rates for patients with BIR after initial RAI. Interpretation Guide: for an individual patient: 1) find their value on each variable axis (age, recurrence risk, ps-Tg, and BRAF^V600E mutation). 2) for each variable, draw a vertical line up to the “points” axis at the top to read the points contributed by that variable. 3) sum the points from all variables to get “total points.” 4) locate this “total points” value on the bottom “total points” axis. 5) draw a vertical line straight down from “total points” to the “risk” axis to read the predicted probability. Example: a patient with age = 45 years (0 points), intermediate recurrence risk (0 points), BRAF^V600E mutation positive (74.4 points) and ps-Tg = 200 ng/mL (100 points) has total Points = 174.4. Drawing down from 174.4 on the “total points” axis gives a predicted IR risk of approximately 87.4%. Abbreviation: IR: incomplete response; BIR: biochemical incomplete response; RAI: radioiodine therapy; ps-Tg = pre-ablation stimulated thyroglobulin; NA: not applicable.

3.6. Validation of the nomogram

The nomogram demonstrated robust predictive accuracy for IR, achieving a C-index of 0.878 (95% CI: 0.823–0.934) in the training cohort (Figure 3(A)). Calibration curves showed excellent agreement between nomogram-predicted and observed probabilities of IR after initial RAI in both cohorts (Figure 4(A,B)). ROC analysis yielded an AUC of 0.878 for IR prediction in the training cohort (Figure 3(A)) and an AUC of 0.917 in the validation cohort (Figure 3(B)). Furthermore, decision curve analysis demonstrated the favorable clinical utility of the nomogram in predicting IR in patients with BIR after initial RAI (Figure 4(C,D)).

Figure 3.

ROC curve in a IR risk nomogram model for patients with BIR after initial RAI. (A) training cohort, and (B) validation cohort. Abbreviation: IR: incomplete response; BIR: biochemical incomplete response; RAI: radioiodine therapy; ROC: receiver operating characteristic; AUC: area under the ROC curve.

Figure 4.

Performance evaluation of the nomogram for predicting IR rates in training and validation cohorts. Calibration curves of the nomogram for IR rates prediction of the training (A) and validation (B) cohorts. DCA of the nomogram for IR rates prediction of the training (C) and validation (D) cohorts. Abbreviation: IR: incomplete response; DCA: decision curve analysis.

4. Discussion

BIR does not necessarily indicate that a patient has persistent DTC. In some cases, incomplete thyroidectomy may lead to persistently low Tg levels because of nonmalignant thyrocytes [4]. To accurately identify patients with true BIR, clinicians must integrate initial static assessments with prognostic factors during follow-up, as discussed in this study.

Tg, a key biochemical marker in DTC, accurately reflects residual or recurrent tumor burden in patients who have undergone total thyroidectomy [1,11]. However, TgAb can interfere with Tg measurements, compromising the reliability of disease monitoring [1]. To address this limitation, both Tg and TgAb levels were systematically measured in our study, and those TgAb-positive patients were excluded.

In the training cohort, we analyzed 144 patients with BIR after initial RAI, over a median follow-up of 49.5 months. During the follow-up period, 45.8% of patients reverted to non-IR, whereas 54.2% progressed to IR (remaining BIR or SIR). Our primary findings demonstrate that ps-Tg can predict the efficacy of RAI, with an optimal cutoff threshold of 51.2 ng/mL. In this context, ps-Tg serves as a key predictive marker for IR. Additionally, our multivariate analysis revealed that a ps-Tg level ≥51.2 ng/mL, age ≥55 years, high initial recurrence risk, and the presence of the BRAF^V600E mutation were independent predictors of IR, highlighting their potential utility in tailoring treatment strategies and refining prognostic evaluations for individual patients.

The 2015 ATA guidelines emphasize Tg’s role in risk stratification, recommending suppressed Tg < 0.2 ng/mL and ps-Tg < 1 ng/mL as benchmarks for remission after total thyroidectomy and RAI [1]. Rosario et al. [12] reported only 3.3% persistent/recurrent disease among 181 intermediate-risk DTC patients with ps-Tg < 2 ng/mL during 5-year follow-up after RAI. Similarly, Tian et al. [13] observed just 2.9% recurrence in 69 high-risk DTC patients with ps-Tg < 1 ng/mL after total thyroidectomy and RAI, with median follow-up of 5.6 years. Thus, DTC patients with undetectable or low postoperative ps-Tg levels typically exhibit favorable outcomes regardless of age, initial staging, or risk stratification. In our study, ps-Tg emerged as the strongest predictor of IR, consistent with prior research showing that higher ps-Tg levels (e.g., > 10 ng/mL) correlate with persistent/recurrent disease [4,14]. These data collectively underscore Tg’s central importance in guiding personalized management of DTC patients.

Kim et al. [15] identified a high frequency of gene mutations in patients aged ≥55 years at diagnosis, which is linked to more aggressive disease pathways than those observed in younger patients. In addition, Shah et al. [16] found that high-risk patients diagnosed before age 55 years had twice the likelihood of achieving ER as those diagnosed later. Our results demonstrated that patients aged ≥55 years exhibited a higher risk of IR than that of their younger counterparts (OR: 4.102, 95% CI: 1.369–12.296, P = 0.012). This finding aligns with existing literature indicating that advanced age is a poor prognostic factor for DTC [17,18]. Age-related changes in thyroid cancer biology, reduced immune responses, and potential alterations in iodine uptake efficiency may have contributed to this observation [1]. These results underscore the need for more aggressive follow-up and individualized therapeutic approaches in older patients.

DTC is generally characterized by favorable survival outcomes; however, tumor recurrence remains a significant challenge in clinical management, particularly for high-risk patients [19,20]. Recurrence in patients with DTC is associated with higher mortality rates, more aggressive disease behavior, and resistance to standard treatments, including RAI [21]. In this study, we demonstrated that initial recurrence risk stratification is an independent risk factor for IR in patients with BIR. High-risk recurrence categories are strongly associated with increased risk of persistent disease and treatment resistance [1]. These findings align with previous studies that emphasize the importance of early identification of high-risk patients and their appropriate stratification for management [19,21,22]. Importantly, while patients with low recurrence risk generally have a favorable prognosis and near-normal life expectancy, those in high-risk groups are more likely to require adjuvant therapies and more intensive follow-up because of the greater likelihood of recurrence and disease progression. This supports the role of recurrence risk stratification as an essential tool in clinical decision-making.

In the present study, the BRAF^V600E mutation was significantly associated with an elevated risk of IR (OR: 5.821; 95% CI: 2.336–14.503; P < 0.001). The BRAF^V600E mutation is the most common genetic alteration, present in approximately 45% of patients with DTC [23]. This mutation has been implicated in a diminished response to RAI because of its effects on thyroid cancer biology [24]. Azouzi et al. [25] demonstrated that BRAF^V600E is associated with reduced sodium/iodide symporter expression, leading to impaired iodine uptake in thyroid cancer cells. This finding underscores the critical role of iodine avidity in determining RAI success. Additionally, Brose et al. [26] reported that patients harboring the BRAF^V600E mutation exhibited poorer clinical outcomes, including high rates of persistent or recurrent disease after RAI.

Patients with DTC who undergo initial treatment without imaging evidence of disease, yet fail to achieve the desired Tg levels often experience a spontaneous reduction in Tg, even in the absence of therapies beyond thyroid-stimulating hormone suppression [27,28]. Elevated ps-Tg levels in patients with DTC are generally associated with worse prognosis. Our recent study identified several factors contributing to elevated Tg levels after surgery, with 60% of cases attributed to biochemical disease [29]. Managing patients within this diagnostic “gray area” poses a significant challenge. The most striking finding of our study was the association between elevated a ps-Tg level ≥51.2 ng/mL and markedly increased risk of IR (OR: 10.813; 95% CI: 4.145–28.213; P < 0.001). These results strongly support the use of ps-Tg as a reliable biomarker for predicting response to RAI.

While prior studies have characterized outcomes in BIR patients, none have developed a validated, multivariable tool for individualized risk stratification [3,4]. Our study addresses this gap by: integrating molecular, clinical, and biochemical predictors (BRAF^V600E, ps-Tg, recurrence risk, age) into a prognostic nomogram. Establishing a clinically actionable ps-Tg threshold (51.2 ng/mL) specifically for BIR patients, unlike prior studies that focused on heterogeneous cohorts [4,29]. Externally validating the model across geographically distinct populations, enhancing generalizability beyond single-center studies.

Our nomogram translates complex risk stratification into an intuitive visual tool, enabling clinicians to identify poor outcomes patients (e.g., ps-Tg ≥51.2 ng/mL, BRAF^V600E positive, age ≥55 years, high-risk). This tool may improve survival through timely intervention while reducing overtreatment in low-risk subgroups. It directly addresses the ATA’s call for “personalized management” in DTC and offers a practical solution to a common clinical dilemma.

Our study had several limitations. First, the retrospective design and relatively small sample size may limit the statistical power of subgroup analyses and generalizability of the findings, although we mitigated this limitation through external validation using geographically distinct cohorts (Qingdao vs. Beijing). Second, while emerging evidence implicates immune markers, tumor microenvironment alterations, and additional genetic variants in RAI response, our model intentionally focused on widely accessible clinical and molecular predictors (including BRAF^V600E and ps-Tg) to ensure immediate clinical applicability. Thus, we are designing a prospective multicenter trial to comprehensively integrate immune profiling, tumor microenvironment biomarkers, and expanded genomic markers (e.g., TERT, RAS) for validating the combinatorial predictive value of these factors in IR among BIR patients.

5. Conclusion

High ps-Tg levels, positive BRAF^V600E mutation status, advanced age, and high recurrence risk may predict an increased likelihood of IR in patients with BIR after initial RAI. A novel nomogram integrating these predictors was developed and demonstrated robust discrimination and calibration for individualized IR risk estimation. The model visually and accurately predicted the probability of IR in patients with BIR, offering a high net clinical benefit.

Funding Statement

The present study was supported by grants from the Shandong Provincial Natural Science Foundation [grant no. ZR2022QH127].

Article highlights

• Focus on a critical clinical dilemma: This study specifically targets patients with differentiated thyroid cancer (DTC) exhibiting biochemical incomplete response (BIR) after initial radioiodine therapy (RAI), a group with significant risk for persistent disease but unclear individual prognosis.

• Multi-center retrospective design: The study leverages data from two independent institutions (Qingdao training cohort, n = 144; Beijing external validation cohort, n = 54), enhancing the robustness and generalizability of the findings.

• Identification of key independent predictors: Through rigorous univariate and multivariate analyses, age ≥55 years, high recurrence risk, pre-stimulated thyroglobulin (ps-Tg) ≥51.2 ng/mL, and presence of BRAF^V600E mutation were identified as significant independent predictors of developing incomplete response (IR) in BIR patients.

• ps-Tg as the dominant risk factor: The developed predictive model establishes ps-Tg level as the single strongest predictor of IR risk in patients with BIR, followed by BRAF^V600E mutation status, age, and recurrence risk.

• Development of a clinically practical nomogram: A user-friendly visual nomogram was constructed incorporating the four independent predictors (age, recurrence risk, ps-Tg, BRAF^V600E) to quantify the individual risk of IR for patients presenting with BIR after initial RAI.

• High predictive accuracy demonstrated: The nomogram exhibited excellent predictive performance, achieving a high concordance index (C-index) of 0.878 in the training cohort and 0.917 in the external validation cohort.

• Robust validation across metrics: Validation included calibration curves showing excellent agreement between predicted and observed IR probabilities in both cohorts, ROC analysis confirming high discriminatory power (AUCs 0.878 & 0.917), and decision curve analysis (DCA) demonstrating favorable clinical utility.

• External validation strengthens reliability: The successful application and high performance of the nomogram in an independent external validation cohort significantly bolster its credibility and potential for broader clinical application.

• Tool for personalized management: This validated nomogram provides clinicians with a practical, evidence-based tool to stratify IR risk in BIR patients, facilitating more informed decisions regarding surveillance intensity and potential additional interventions.

Author contributions

Congcong Wang, Xufu Wang and Xinfeng Liu contributed to the conception and design of the study. Shuhui Wang, Yutian Li, Yaqi Lu, Xia Shen, Wei Sun, Jiao Li, Na Han, Chenghui Lu, Guoqiang Wang, Yingying Zhang, Zenghua Wang, Zengmei Si, and Yansong Lin assisted with data acquisition. Congcong Wang, and Guohua Qin conducted the statistical analyses and drafted the manuscript. Xufu Wang, Xinfeng Liu, Yansong Lin, and Fengqi Li revised the manuscript. All authors contributed to the article and approved the submitted version.

Disclosure statement

The authors have no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.

Reviewer disclosures

Peer reviewers on this manuscript have no relevant financial or other relationships to disclose.

Ethical declaration

The studies involving human participants were reviewed and approved by Ethical Committee of the Affiliated Hospital of Qingdao University (Ethics approval number: QYFY WZLL 29,555). The institutional review board that approved this study waived the requirement for written informed consent due to the retrospective nature of the research involving anonymized patient data.

Data availability statement

The datasets used during the present study are available from the corresponding author upon reasonable request.