This work is licensed under a Abstract
The aim of this study was to quantify the cancer-specific death (CSD) benefit of radioactive iodine therapy (RAIT) in older patients with N1b differentiated thyroid carcinoma, a population designated by the ATA-2025 guideline as a ‘consider RAIT’ zone despite unproven mortality benefit, and to translate this recommendation into clinically actionable subgroups using a competing-risks framework. We used data from the Surveillance, Epidemiology, and End Results (SEER) programme (2004–2015) and analysed papillary thyroid carcinoma (PTC) cases aged ≥55 years with N1b, M0. After 1:1 propensity score matching, cumulative incidence function and Fine–Gray models evaluated associations between RAIT and CSD. Stratifications included tumour size, positive lymph node burden (PLN > 5), and ETE. A nomogram was developed for 1-, 3-, and 5-year CSD prediction, and absolute risk reduction (ARR) and number needed to treat (NNT) were estimated. Among 1,142 patients, 648 were matched (n = 324/group; median follow-up = 69 months). Five-year CSD was 14.1% without RAIT versus 5.1% with RAIT (P = 0.001), yielding ARR ≈ 9% and NNT ≈ 11. RAIT was independently protective (SHR 0.33, 95% CI: 0.14–0.76). Benefit concentrated in tumours 2–4 cm, PLN > 5, and ETE-positive strata. The nomogram showed strong discrimination and calibration. In patients aged ≥55 years with N1b PTC, RAIT confers measurable mortality benefit, most evident in 2–4 cm, PLN > 5, or ETE-positive disease. Integrating ARR, NNT, and a validated nomogram, this study converts the conceptual ‘consider RAIT’ into clinically actionable ‘RAIT-favoured’ and ‘RAIT-optional’ pathways.
Keywords: radioactive iodine, papillary thyroid carcinoma, competing risk analysis, elderly patients, nomogram
Introduction
Differentiated thyroid cancer generally carries an excellent prognosis; however, patients with lateral neck nodal metastasis (N1b) who are aged ≥ 55 years face substantially higher risks of recurrence and death, which has led to wide practice variation in postoperative radioactive iodine (RAIT) indications and intensity (1). The American Thyroid Association (ATA) 2025 guideline, within the recurrence/persistence risk framework, subdivides the traditional intermediate-risk category into intermediate-low and intermediate-high – both designated as ‘consider RAIT’ – whereas high-risk or distant metastasis cases are recommended RAIT and low-risk cases are generally not recommended RAIT. Features such as N1b involvement, higher tumour diameter, a higher number of positive lymph nodes (e.g. >5), and extrathyroidal extension (ETE) are explicitly listed as scenarios in which RAIT is more likely to be favoured (2). In parallel, the European consensus clarifies the three purposes of RAIT – remnant ablation, adjuvant therapy, and treatment of the known disease – and advocates selective use in the ‘consider RAIT’ zone (3).
‘Consider’ does not equate to proven benefit on mortality endpoints. Real-world analyses have suggested that age, nodal burden, and invasive features may modify the likelihood of benefit from RAIT, but most studies have prioritized overall survival (OS) or cancer-specific survival (CSS) rather than cancer-specific death (CSD) as the core endpoint; moreover, among older patients, other-cause death (OCD) has seldom been handled within a competing-risks framework, risking over- or underestimation of RAIT’s true oncologic effect (4). For example, analyses from the Surveillance, Epidemiology, and End Results (SEER) programme indicate that among older N1b patients with a higher nodal burden, RAIT benefit is more likely to be observed, yet stratification thresholds and statistical methods have been heterogeneous, limiting bedside implementability (5). Meanwhile, safety signals – including second primary malignancies in younger patients and short- to mid-term haematologic suppression and salivary/lacrimal toxicities after RAIT – have been documented, and although the underlying evidence is heterogeneous, these findings underscore the need to identify net-benefit subgroups within the ‘consider’ population (6, 7, 8, 9).
Against this background, we focused on patients with papillary thyroid carcinoma (PTC) in SEER 2004–2015 who were aged ≥ 55 years with N1b, M0 disease: treatment‐selection bias was mitigated using propensity score matching (PSM); CSD was specified as the primary endpoint and, within a competing-risks framework cumulative incidence function (CIF) with Gray’s test and the Fine–Gray subdistribution hazard model, the association between RAIT and CSD was quantified; stratification was performed according to guideline-aligned features – tumour size, positive lymph node burden (PLN > 5), and ETE – in keeping with ATA-2025 considerations favouring RAIT; and a nomogram based on significant covariates was constructed to enable individualised prediction of 1-, 3-, and 5-year CSD (4). This study addresses a methodological gap by employing a competing-risks framework with CSD as the primary endpoint, reporting absolute effect measures – absolute risk reduction (ARR) and number needed to treat (NNT) – and delivering a deployable nomogram that translates ‘consider RAIT’ into a multidisciplinary team (MDT)-ready two-tier strategy – ‘RAIT-favoured’ and ‘RAIT-optional’ (1, 2).
Methods
Data source and ethics
This retrospective, population-based cohort used the publicly available SEER database (SEER*Stat v8.4.5; diagnosis years 2004–2015). SEER data are de-identified; the study complied with the Declaration of Helsinki and required no additional institutional review board approval or informed consent.
Study population: eligibility and exclusions
Target population
The target population included adults with PTC and N1b nodal status, aged ≥ 55 years, treated with total thyroidectomy, M0 at diagnosis, and with complete follow-up.
Flow
We initially identified 93,203 PTC cases; 32,410 met core eligibility. We then excluded 31,268 for missing or inconsistent key variables: race (n = 1,347); multifocality (n = 23,404); lymph node dissection or node count (n = 6,318); and insufficient follow-up or survival data (n = 199). The final analytic cohort comprised 1,142 patients (Fig. 1).
Figure 1.
Flowchart of patient selection and study design. Abbreviations: RAIT, radioactive iodine therapy; SEER, Surveillance, Epidemiology, and End Results; PTC, papillary thyroid carcinoma; and PSM, propensity score matching.
Terminology
We use ‘N1b (as recorded in SEER)’ throughout. Because the SEER N category derives primarily from pathological or composite staging, we do not further distinguish cN1b from pN1b in the main text to avoid coding inconsistencies.
Variables and definitions
Exposure
RAIT (yes/no). Clinicodemographic and tumour covariates. Sex; race (white vs non-white); marital status (married vs unmarried); tumour size (≤1, >1–≤2, >2–≤4, >4 cm); positive lymph node burden (PLN > 5; yes/no); ETE (yes/no); multifocality (multiple vs solitary).
Endpoints
OS and CSS; in the competing-risks framework: CSD and OCD.
Propensity score matching (PSM)
were matched 1:1 using nearest-neighbour matching (calliper width 0.1 on the propensity score scale), without replacement (groups: RAIT yes/no). Matching covariates: sex, race, marital status, tumour size, PLN > 5, ETE, and multifocality. Balance was assessed using Love plots of standardised mean differences (SMDs) in conjunction with between-group tests; SMD ≤ 0.10 was taken to indicate adequate balance.
Survival and competing-risks analyses
We generated Kaplan–Meier (KM) curves with log-rank tests for OS and CSS in the unmatched and matched cohorts. In the competing-risks framework, we estimated the CIF and compared groups with Gray’s test. In the matched cohort, we fitted Fine–Gray subdistribution-hazard models, including RAIT, sex, tumour size, PLN > 5, ETE, race, marital status, and multifocality, reporting SHR with 95% CI.
Pre-specified subgroups
Analyses were stratified by tumour size (four levels), PLN > 5 (yes/no), and ETE (yes/no). For each stratum, we compared CIFs for CSD and OCD using Gray’s test (p1 for CSD; p2 for OCD).
Nomogram construction and internal validation
The full cohort was randomly split 3:1 into a training set (n = 799) and an internal validation set (n = 343). Based on significant predictors from the training-set Fine–Gray model (sex, PLN > 5, treatment group, and tumour size), we constructed a nomogram for 1-, 3-, and 5-year CSD prediction. Discrimination was assessed using time-dependent AUC (time-dependent ROC) at 1, 3, and 5 years; calibration was evaluated using bootstrap calibration curves (1,000 resamples).
Missing data and robustness
Cases with missing key variables were excluded per the flowchart (Fig. 1; complete case analysis), with potential selection bias discussed in the limitations. The primary endpoint analyses were performed in the matched cohort; the unmatched cohort is presented for context.
Statistical considerations (time scale and censoring)
All tests were two-sided with α = 0.05. Analyses and visualisation were performed in R (v4.0.3) (core packages: MatchIt and optmatch for PSM; cobalt for Love plots; cmprsk for CIF and Fine–Gray; timeROC, pROC, and bootstrap for AUC and calibration).
Time scale
Follow-up was measured from diagnosis in months (SEER survival months).
Censoring rules
OS: event = death from any cause; alive/lost was right-censored at last contact.
CSS (KM): event = thyroid-cancer death; OCD treated as non-informative censoring.
Competing risks (CIF with Gray’s test and the Fine–Gray model): primary event = CSD; OCD modelled as a competing event (not censored).
Administrative censoring: primary analyses used the full SEER follow-up; results are reported at 1, 3, and 5 years, with plots to 120 months.
Results
Baseline characteristics
We included 1,142 patients with PTC and N1b, aged ≥ 55 years (RAIT n = 818; no RAIT n = 324) (Table 1). Overall, 79.9% were white and 63.7% were married; tumours were mainly >2–≤4 cm (30.6%) or >4 cm (19.4%); PLN > 5 was present in 51.1%, ETE in 57.7%, and multifocality in 61.7%. One-to-one PSM yielded 648 comparable patients (n = 324 per group). Before matching, only marital status differed (married: 65.5 vs 59.0%, P = 0.044). After matching, balance improved and differences were not significant for sex, marital status, tumour size, PLN > 5, ETE, and multifocality (P = 0.133, 0.124, 0.691, 0.208, 1.000, 0.684, respectively), with a small residual imbalance in race (white: 87.0 vs 80.9%, P = 0.042). Median follow-up for the full cohort was 69 months; 72 CSDs and 65 OCDs occurred. The primary endpoint model was built in the matched cohort (multivariable Fine–Gray; Table 2).
Table 1.
Baseline characteristics of patients with N1b PTC aged ≥ 55 years treated with or without RAIT before and after PSM.
| Characteristics | Before PSM | P value | After PSM | P value | ||||
|---|---|---|---|---|---|---|---|---|
| ALL | No-RAIT | RAIT | ALL | No-RAIT | RAIT | |||
| n = 1,142 | n = 324 | n = 818 | n = 648 | n = 324 | n = 324 | |||
| Gender | 0.181 | 0.133 | ||||||
| Female | 584 (51.1%) | 155 (47.8%) | 429 (52.4%) | 290 (44.8%) | 155 (47.8%) | 135 (41.7%) | ||
| Male | 558 (48.9%) | 169 (52.2%) | 389 (47.6%) | 358 (55.2%) | 169 (52.2%) | 189 (58.3%) | ||
| Race | 0.652 | 0.042 | ||||||
| White | 912 (79.9%) | 262 (80.9%) | 650 (79.5%) | 544 (84.0%) | 262 (80.9%) | 282 (87.0%) | ||
| Non-white | 230 (20.1%) | 62 (19.1%) | 168 (20.5%) | 104 (16.0%) | 62 (19.1%) | 42 (13.0%) | ||
| Marital status | 0.044 | 0.124 | ||||||
| Married | 727 (63.7%) | 191 (59.0%) | 536 (65.5%) | 402 (62.0%) | 191 (59.0%) | 211 (65.1%) | ||
| Unmarried | 415 (36.3%) | 133 (41.0%) | 282 (34.5%) | 246 (38.0%) | 133 (41.0%) | 113 (34.9%) | ||
| Tumour size (cm) | 0.947 | 0.691 | ||||||
| <= 1 | 239 (20.9%) | 69 (21.3%) | 170 (20.8%) | 143 (22.1%) | 69 (21.3%) | 74 (22.8%) | ||
| >1–2 | 332 (29.1%) | 90 (27.8%) | 242 (29.6%) | 190 (29.3%) | 90 (27.8%) | 100 (30.9%) | ||
| >2–4 | 349 (30.6%) | 101 (31.2%) | 248 (30.3%) | 191 (29.5%) | 101 (31.2%) | 90 (27.8%) | ||
| >4 | 222 (19.4%) | 64 (19.8%) | 158 (19.3%) | 124 (19.1%) | 64 (19.8%) | 60 (18.5%) | ||
| LN positive | 0.608 | 0.208 | ||||||
| <= 5 | 559 (48.9%) | 163 (50.3%) | 396 (48.4%) | 309 (47.7%) | 163 (50.3%) | 146 (45.1%) | ||
| >5 | 583 (51.1%) | 161 (49.7%) | 422 (51.6%) | 339 (52.3%) | 161 (49.7%) | 178 (54.9%) | ||
| ETE | 0.845 | 1.000 | ||||||
| No | 483 (42.3%) | 139 (42.9%) | 344 (42.1%) | 278 (42.9%) | 139 (42.9%) | 139 (42.9%) | ||
| Yes | 659 (57.7%) | 185 (57.1%) | 474 (57.9%) | 370 (57.1%) | 185 (57.1%) | 185 (57.1%) | ||
| Focality | 0.841 | 0.684 | ||||||
| Multiple | 705 (61.7%) | 202 (62.3%) | 503 (61.5%) | 410 (63.3%) | 202 (62.3%) | 208 (64.2%) | ||
| Single | 437 (38.3%) | 122 (37.7%) | 315 (38.5%) | 238 (36.7%) | 122 (37.7%) | 116 (35.8%) | ||
Abbreviations: RAIT, radioactive iodine therapy; PSM, propensity score matching; LN, lymph node; ETE, extrathyroidal extension; and PTC, papillary thyroid carcinoma.
Table 2.
Multivariate Fine–Gray subdistribution hazard analysis for CSD in patients with N1b PTC aged ≥ 55 years.
| Characteristics | Subdistribution proportion hazards model | ||
|---|---|---|---|
| HR | 95% CI | P value | |
| Gender | |||
| Female | Reference | ||
| Male | 1.34 | 1.12–3.38 | 0.030 |
| Race | |||
| White | Reference | ||
| Non-white | 0.57 | 0.19–1.68 | 0.310 |
| Marital status | |||
| Married | Reference | ||
| Unmarried | 1.01 | 0.41–2.49 | 0.980 |
| Tumour size (cm) | |||
| <= 1 | Reference | ||
| >1–2 | 1.57 | 0.17–14.28 | 0.690 |
| >2–4 | 3.24 | 0.36–29.6 | 0.300 |
| >4 | 12.88 | 1.71–97.16 | 0.013 |
| RAIT treatment | |||
| No RAIT | Reference | ||
| RAIT | 0.33 | 0.14–0.76 | 0.009 |
| LN positive | |||
| <= 5 | Reference | ||
| >5 | 2.85 | 1.14–7.13 | 0.025 |
| ETE | |||
| No | Reference | ||
| Yes | 0.79 | 0.3–2.09 | 0.640 |
| Focality | |||
| Multiple | Reference | ||
| Single | 1.00 | 0.4–2.46 | 0.990 |
Abbreviations: RAIT, radioactive iodine therapy; PSM, propensity score matching; CSD, cancer-specific death; OCD, other-cause death; PTC, papillary thyroid carcinoma; CI, confidence interval; and HR, hazard ratio.
Survival analysis
In both the unmatched cohort (OS: Supplementary Fig. S1A; CSS: Supplementary Fig. S1B (see section on Supplementary materials given at the end of the article)) and the matched cohort (OS: Supplementary Fig. S1C; CSS: Supplementary Fig. S1D), survival curves favoured RAIT, all log-rank tests: P < 0.001 (full figures available in the Supplementary materials).
Competing risks
Treating OCD as a competing event, RAIT was associated with consistently lower CSD (Fig. 2). Unmatched cohort: 1-, 3-, and 5-year CSD CIFs (no RAIT vs RAIT) were 7.5, 11.0, and 14.1 vs 0.3, 2.7, and 5.4%, respectively (Gray’s test: P < 0.001); the corresponding OCD CIFs were 1.2, 6.6, and 10.7 vs 0.5, 2.7, and 5.1%, respectively (P = 0.004). Matched cohort: 1-, 3-, and 5-year CSD CIFs were 7.5, 11.0, and 14.1% (no RAIT) vs 0.3, 3.5, and 5.1% (RAIT), respectively (P = 0.001); OCD CIFs were 1.2, 6.7, and 10.7 vs 0.3, 2.2, and 4.6%, respectively (P = 0.0226). The 5-year ARR in CSD after matching was ≈ 8.97%, corresponding to a NNT of ≈ 11–12 (see Supplementary Table S1 for detailed estimates).
Figure 2.
CIF curves comparing RAIT versus no RAIT for cancer-specific death (CSD) and other-cause death (OCD) in the matched cohort. (A) Overall matched cohort; (B, C, D, E) tumour-size subgroups: (B) ≤1 cm, (C) 1–≤2 cm, (D) 2–≤4 cm, (E) >4 cm; (F and G) positive lymph node (PLN) subgroups: (F) PLN ≤ 5 and (G) PLN > 5; (H and I) extrathyroidal extension (ETE) subgroups: (H) ETE positive and (I) ETE negative. Abbreviations: RAIT, radioactive iodine therapy; CIF, cumulative incidence function; CSD, cancer-specific death; OCD, other-cause death; PLN, positive lymph node; and ETE, extrathyroidal extension. A full colour version of this figure is available at https://doi.org/10.1530/ERC-25-0420.
Multivariable Fine–Gray model
In the matched cohort, RAIT was independently associated with lower CSD (SHR 0.33, 95% CI: 0.14–0.76, P = 0.009) (Table 2). Relative to ≤1 cm, >4 cm tumours conferred higher risk (SHR 12.88, 95% CI: 1.71–97.16, P = 0.013), whereas 1–≤2 cm and 2–≤4 cm did not reach significance (SHR 1.57, P = 0.690; SHR 3.24, P = 0.300). PLN > 5 was an independent risk factor (SHR 2.85, 95% CI: 1.14–7.13, P = 0.025). Male sex was associated with higher risk (SHR 1.34, 95% CI: 1.12–3.38, P = 0.030). Race, marital status, ETE, and multifocality were not statistically significant (all P > 0.050).
Pre-specified subgroups
Gray’s tests on stratified CIFs showed patterns directionally concordant with the matched Fine–Gray model. Tumour size: ≤1 cm (Fig. 2B) and 1–≤2 cm (Fig. 2C) were not significant or trend-level (p1 = 0.891 and 0.063 for CSD); 2–≤4 cm (Fig. 2D) showed a significant reduction in CSD (p1 < 0.001); >4 cm (Fig. 2E) was not significant (p1 = 0.062). The corresponding p2 values for OCD were 0.622, 0.193, 0.011, and 0.510, respectively. Nodal burden: PLN ≤ 5 (Fig. 2F) was not significant (p1 = 0.063), whereas PLN > 5 (Fig. 2G) showed a significant reduction (p1 < 0.001); the corresponding OCD p2 = 0.198 and 0.056. ETE: ETE positivity (Fig. 2H) showed a significant reduction (p1 < 0.001), whereas ETE negativity (Fig. 2I) was not significant (p1 = 0.152); the corresponding OCD p2 = 0.106 and 0.105 (here, p1 = Gray’s test for CSD; p2 = Gray’s test for OCD).
Nomogram and internal validation
The cohort was split 3:1 into training (n = 799) and validation (n = 343) sets. A nomogram predicting 1-, 3-, and 5-year CSD was constructed from significant training-set predictors (sex, PLN > 5, treatment group, tumour size; Fig. 3). Discrimination was good: AUCs in the training set were 83.6% (1 year, 95% CI: 73.9–93.2), 80.0% (3 years, 73.1–86.8), and 75.3% (5 years, 69.1–81.5); in the validation set, AUCs were 82.9% (1 year, 68.8–96.9), 78.0% (3 years, 68.1–88.0), and 71.2% (5 years, 60.4–81.9) (Fig. 4A and B). Calibration plots showed good agreement between predicted and observed risks (Fig. 4C and D).
Figure 3.
Nomogram predicting 1-, 3-, and 5-year cancer-specific death (CSD) probabilities in the matched cohort based on the Fine–Gray subdistribution-hazard model. Predictors included sex, positive lymph node burden (PLN > 5), RAIT treatment group (group: yes = received RAIT; no = did not receive RAIT), and tumour size. The total points correspond to individual 1-, 3-, and 5-year CSD probabilities. Abbreviations: RAIT, radioactive iodine therapy; CSD, cancer-specific death; and PLN, positive lymph node. A full colour version of this figure is available at https://doi.org/10.1530/ERC-25-0420.
Figure 4.
Discrimination and calibration of the Fine–Gray nomogram for predicting cancer-specific death (CSD). (A and B) Time-dependent receiver operating characteristic (ROC) curves and corresponding area under the curve (AUC) values for 1-, 3-, and 5-year CSD in the training (A) and validation (B) sets. (C and D) Calibration plots showing agreement between predicted and observed probabilities in the training (C) and validation (D) sets. Abbreviations: CSD, cancer-specific death; ROC, receiver operating characteristic; AUC, area under the curve. A full colour version of this figure is available at https://doi.org/10.1530/ERC-25-0420.
Discussion
In patients aged ≥ 55 years with N1b, M0 PTC, RAIT was independently associated with lower CSD after propensity score matching (Fine–Gray SHR 0.33, 95% CI: 0.14–0.76). Within a competing-risks framework, RAIT consistently reduced the cumulative incidence of CSD across 1-, 3-, and 5-year follow-ups, corresponding to an absolute 5-year risk reduction of about 9% and a number needed to treat of ≈ 11–12. A clear concentration pattern emerged: tumours 2–≤4 cm, PLN > 5, and ETE-positive strata showed significant Gray’s tests, whereas >4 cm did not (P = 0.062). This distribution suggests that, in older N1b patients, nodal burden and an invasive phenotype – rather than size escalation alone – better delineate the net-benefit zone for RAIT (1, 10).
Alignment with ATA-2025 is notable. The guideline refines the intermediate-risk category into intermediate-low and intermediate-high (both ‘consider RAI’) and highlights N1b, increasing tumour size, higher numbers of positive nodes (e.g. >5), and ETE as features favouring RAIT (2, 3). Our findings are directionally concordant yet extend this framework by i) demonstrating – on a mortality endpoint using competing-risks methods – that benefit is concentrated in ≥2 cm, PLN >5, and ETE-positive subgroups, with no clear CSD gain in lower-burden strata, and ii) operationalising ‘consider’ via quantified ARR/NNT and a deployable nomogram, enabling MDT-ready stratification into ‘RAIT-favoured’ versus ‘RAIT-optional’ (4). Comparisons with prior real-world evidence support these inferences. SEER-based cohorts of N1b patients suggest that advanced age and heavy nodal burden underpin observable benefit from RAIT; some studies have proposed higher thresholds (e.g. PLN ≥ 9) to reveal such effects (5). Using a PLN > 5 versus ≤ 5 categorisation, we reproduced the same directional signal for CSD in adults aged ≥ 55 years, echoing ATA-2025’s emphasis on ‘>5’ as a burden cue favouring RAIT (2). While larger tumours are more often treated with RAIT in observational data (11), our finding of a clearer benefit at 2–≤4 cm but not at >4 cm suggests diminishing marginal explanatory power of size alone in older patients – potentially owing to higher OCD and limited events reducing statistical power, alongside unmeasured confounding (extent of surgery, RAIT activity and preparation, thyroglobulin, and imaging) not captured in SEER (12). By contrast, ETE and nodal burden are pathobiological surrogates for microscopic residuals or subclinical dissemination – the actionable targets of adjuvant RAIT (13, 14).
Beyond mortality endpoints, relapse scenarios are also relevant when interpreting RAIT in older N1b patients. Locoregional recurrence is more common in those with heavy nodal burden or ETE, features that reflect microscopic residual disease along lymphatic pathways and at the thyroid bed – the principal biological targets of adjuvant RAIT. Distant recurrence, although infrequent in differentiated PTC, increases with age and nodal burden and may further contribute to cancer-related mortality in this population. Although the SEER programme does not capture locoregional or distant relapse, the concentration of benefit we observed in patients with PLN > 5 and ETE-positive disease is biologically consistent with these relapse patterns and supports a precision approach to RAIT selection within the ATA-2025 ‘consider RAIT’ zone (15, 16).
That not all high-risk patients reached significance likely reflects the interplay between methodological and biological factors. In older adults, OCD genuinely competes with CSD in competing-risks analyses (rather than being treated as non-informative censoring); as comorbidity and treatment-related toxicity rise, the apparent oncological signal of RAIT may be partially offset (17). Although dose–toxicity data are heterogeneous and generally of low certainty, available evidence suggests that haematological suppression and salivary/lacrimal adverse effects occur within 1–12 months post-RAI and tend to increase with administered activity, supporting the use of the most tolerable (lowest effective) administered activity in those most likely to benefit (7, 8, 9). Safety signals for second primary malignancies in younger populations further underscore focussing RAIT on patients with a high probability of net benefit (6, 18).
Clinically, within the ATA-2025 ‘consider RAIT’ zone for patients aged ≥ 55 years with N1b, tumours ≥ 2 cm, PLN > 5, and ETE-positive disease constitute RAIT-favoured scenarios, in which we observed significantly lower CSDs; by contrast, tumour diameter ≤1 cm or 1–≤2 cm, PLN ≤ 5, and ETE-negative strata are RAIT-optional, warranting shared decision-making that integrates thyroglobulin, imaging, comorbidities, and patient preferences (10). With strong discrimination and calibration, the nomogram supports MDT communication of individualised 1-, 3-, and 5-year CSD risks and absolute differences with versus without RAIT (16).
Limitations are inherent to observational, registry-based analyses: residual confounding persists; SEER dataset lacks key variables (RAIT administered activity and preparation, margin status and extent of dissection, post-operative thyroglobulin, and locoregional recurrence), and differences in staging versions or recoding may introduce noise (11). The PLN > 5 versus ≤5 dichotomy precluded evaluation of higher cut-offs (e.g. ≥9); the non-significant >4 cm result may reflect sparse events together with competing mortality; and the nomogram, although internally validated, warrants external and prospective validation (19, 20, 21, 22, 23).
In summary, within the ATA-2025 ‘consider RAIT’ framework, we provide competing-risks evidence on a mortality endpoint for older patients with N1b PTC, quantify absolute effect measures – ARR and NNT, and supply a deployable nomogram that translates ‘consider’ into an actionable RAIT-favoured and RAIT-optional pathway. Future work should validate this stratification externally and prospectively, integrating RAIT activity selection and toxicity surveillance to maximise net benefit while minimising unnecessary exposure (24, 25, 26).
Conclusion
Within the ATA-2025 ‘consider RAIT’ framework, our competing-risks analysis demonstrated that in patients aged ≥ 55 years with N1b, M0 PTC, the mortality benefit of RAIT was most evident in those with tumours 2–4 cm, PLN > 5, or ETE positivity. By integrating ARR, NNT, and a validated nomogram, this study translates the conceptual ‘consider’ into actionable ‘RAIT-favoured’ and ‘RAIT-optional’ pathways, bridging quantitative evidence and multidisciplinary decision-making. External and prospective validation is warranted to refine this precision-guided use of adjuvant RAIT.
Supplementary materials
Declaration of interest
The authors declare that there is no conflict of interest that could be perceived as prejudicing the impartiality of the work reported.
Funding
This work was supported by the Hangzhou Health Science and Technology Project (Grant No. B20253339) and the Basic Public Welfare Research Program of Wenzhou Science and Technology Bureau (Grant No. Y20240475).
Datasets analysed
The datasets analysed during the current study are available in the SEER repository (https://seer.cancer.gov/data/).
References
- 1.Kang IK, Kim K, Park J, et al. Central lymph node ratio predicts recurrence in patients with N1b papillary thyroid carcinoma. Cancers 2022. 14 3677. ( 10.3390/cancers14153677) [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Ringel MD, Sosa JA, Baloch Z, et al. 2025 American Thyroid Association Management Guidelines for adult patients with differentiated thyroid cancer. Thyroid 2025. 35 841–985. ( 10.1177/10507256251363120) [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Pacini F, Fuhrer D, Elisei R, et al. 2022 ETA consensus statement: what are the indications for post-surgical radioiodine therapy in differentiated thyroid cancer? Eur Thyroid J 2022. 11 e210046. ( 10.1530/ETJ-21-0046) [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Alshwayyat S, Al-Akhras A, Ghazou A, et al. Assessing radioiodine therapy long-term outcomes in differentiated thyroid cancer using nomograms. Sci Rep 2024. 14 23349. ( 10.1038/s41598-024-72002-0) [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Jian J, Wei M, Li X, et al. Impact of positive lymph nodes and RAI therapy on survival in N1b papillary thyroid carcinoma. Front Endocrinol 2025. 16 1551075. ( 10.3389/fendo.2025.1551075) [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Pasqual E, Schonfeld S, Morton LM, et al. Association between radioactive iodine treatment for pediatric and young adulthood differentiated thyroid cancer and risk of second primary malignancies. J Clin Oncol 2022. 40 1439–1449. ( 10.1200/JCO.21.01841) [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Berta DM, Teketelew BB, Cherie N, et al. Effect of radioactive iodine therapy on hematological parameters in patients with thyroid cancer: systematic review and meta-analysis. Front Endocrinol 2025. 16 1562851. ( 10.3389/fendo.2025.1562851) [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Musso L, Maltese C, Beretta G, et al. Low-dose versus standard-dose radioiodine therapy in differentiated thyroid cancer: focus on tolerability in a retrospective evaluation. Pharmaceuticals 2025. 18 443. ( 10.3390/ph18040443) [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Reinecke MJ, Ahlers G, Burchert A, et al. Second primary malignancies induced by radioactive iodine treatment of differentiated thyroid carcinoma – a critical review and evaluation of the existing evidence. Eur J Nucl Med Mol Imaging 2022. 49 3247–3256. ( 10.1007/s00259-022-05762-4) [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Newman SK, Patrizio A & Boucai L. Decision variables for the use of radioactive iodine in patients with thyroid cancer at intermediate risk of recurrence. Cancers 2024. 16 3096. ( 10.3390/cancers16173096) [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Weis H, Weindler J, Schmidt K, et al. Impact of radioactive iodine treatment on long-term relative survival in patients with papillary and follicular thyroid cancer: a SEER-based study covering histologic subtypes and recurrence risk categories. J Nucl Med 2025. 66 525–530. ( 10.2967/jnumed.124.269091) [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Mihailović J. Evolving paradigm in radioactive iodine therapy for differentiated thyroid cancer: historical perspectives, current practices and future directions. Diagnostics 2025. 15 1438. ( 10.3390/diagnostics15111438) [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Zhao H & Gong Y. Radioactive iodine in low- to intermediate-risk papillary thyroid cancer. Front Endocrinol 2022. 13 960682. ( 10.3389/fendo.2022.960682) [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Guo H, Zhang N, Hu Y, et al. Radioactive iodine therapy strategies for distinct types of differentiated thyroid cancer: a propensity score–matched analysis. Front Endocrinol 2023. 14 1158581. ( 10.3389/fendo.2023.1158581) [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Lee YM, Sung TY, Kim WB, et al. Risk factors for recurrence in patients with papillary thyroid carcinoma undergoing modified radical neck dissection. Br J Surg 2016. 103 1020–1025. ( 10.1002/bjs.10144) [DOI] [PubMed] [Google Scholar]
- 16.Medas F, Canu GL, Boi F, et al. Predictive factors of recurrence in patients with differentiated thyroid carcinoma: a retrospective analysis on 579 patients. Cancers 2019. 11 1230. ( 10.3390/cancers11091230) [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Lauretta R, Bianchini M, Mormando M, et al. Focus on thyroid cancer in elderly patients. Endocrines 2023. 4 757–772. ( 10.3390/endocrines4040055) [DOI] [Google Scholar]
- 18.Persons EM, Hussein MH, Herrera M, et al. The impact of radioactive iodine on outcomes among pediatric and adolescent thyroid cancer patients: a SEER database analysis. Cancers 2025. 17 107. ( 10.3390/cancers17010107) [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Margulis I, Gordon Y, Stein N, et al. Predictive factors for failure in patients with N1b papillary thyroid carcinoma. World J Surg Onc 2025. 23 307. ( 10.1186/s12957-025-03694-5) [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Schott M, Schott-Ohly P, Krieg S, et al. The prognostic impact of radioiodine therapy in patients with papillary thyroid cancer: a propensity score matched case-control study. Horm Metab Res 2024. 56 770–778. ( 10.1055/a-2423-4849) [DOI] [PubMed] [Google Scholar]
- 21.Zhao M, Shi X, Zhang J, et al. Radioactive iodine ablation can reduce the structural recurrence rate of intermediate-risk papillary thyroid microcarcinoma: a meta-analysis. Comput Math Methods Med 2022. 2022 1–8. ( 10.1155/2022/8028846) [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Tang T, Zhi J, Zhang W, et al. Surgery and radioactive iodine therapeutic strategy for patients greater than 60 years of age with differentiated thyroid cancer. J Healthc Eng 2022. 2022 1–9. ( 10.1155/2022/4348396) [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Dai P, Zhao W, Zheng X, et al. Effect of radioactive iodine therapy on cancer-specific survival of papillary thyroid cancer tall cell variant. J Clin Endocrinol Metab 2024. 109 e1260–e1266. ( 10.1210/clinem/dgad580) [DOI] [PubMed] [Google Scholar]
- 24.Indra B, Qodir N, Pramudhito D, et al. Effectiveness of radioiodine therapy on preventing recurrence in differentiated thyroid carcinoma: a systematic review. J Egypt Natl Canc Inst 2025. 37 39. ( 10.1186/s43046-025-00293-z) [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Stegenga MT, Visser WE, Peeters RP, et al. Radioactive iodine in differentiated thyroid cancer: effect on detection of distant metastases comparing 4 guidelines. J Endocr Soc 2025. 9 bvaf051. ( 10.1210/jendso/bvaf051) [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Jaiswal G, Grimes M, Bononi P, et al. Usefulness of preradioactive iodine scans in thyroid cancer. J Endocr Soc 2025. 9 bvaf128. ( 10.1210/jendso/bvaf128) [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.