During the period when treatments were planned, numerous randomized controlled trials (RCTs) with conflicting results regarding the ablation success of 30 mCi (1.1 GBq) and 100 mCi (3.7 GBq) doses were published [13,14,15,16,17]. Despite the ATA 2015 guidelines recommend a 30 mCi (1.1 GBq) dose for remnant ablation, clinicians quickly adopted a strategy of ‘saving a chance of definitive treatment for a later occasion’ by opting for the full dose of RAI. As a result, the majority of patients received an adjuvant treatment dose of 100 mCi (3.7 GBq).

Our matched patient cohort primarily aims to evaluate the effectiveness of RAI therapy in patients who have undergone total thyroidectomy (TT) in our country. Even though the ATA 2015 guidelines do not recommend routine RAI ablation in low-risk cancer patients, and suggest using low-dose RAI if ablation is performed, certain factors may cause different outcomes in our country [4]. The patient group included in our study comprises individuals from various regions of the country, most of whom underwent TT not performed by endocrine-specific surgeons. Only approximately 6% of the patients in our study underwent central neck dissection, and 2% underwent lateral neck dissection. The mean postoperative Tg levels are 5.30 ng/ml (95% CI: 3.95–6.66) for the RAI group and 3.96 ng/ml (95% CI: 2.95–4.96) for the Non-RAI group, respectively. The recently published prospective phase-3 ESTIMABL2 study reported that 42% of patients underwent lymph node dissection (LND), with up to 25% undergoing lateral neck dissection, which is unusual for low-risk patients. In contrast, the ESTIMABL2 study, while the mean or median postoperative serum Tg levels were not specified, it was reported that out of 555 patients with postoperative serum Tg results, 507 had levels below 1 ng/ml [18]. Previously, the HiLo study, which investigated the difference between low and high RAI doses, included patients who had undergone LND in 75% of cases, with median postoperative serum Tg levels at 2.3 ng/ml [19]. Along with these randomized studies, various studies and meta-analyses have examined the success of RAI therapy based on results from patients who underwent LND in 50% or more of cases [20, 21]. In these studies, the presence of a very small thyroid remnant likely made TT combined with LND sufficient on its own, or a low dose of RAI therapy was adequate. However, this was insufficient for our study group.

Notably, The long-term recurrence rate in our RAI group is around 1%, which is similar to the 2% rate reported in the ESTIMABL1 study. However, the recurrence rate in our Non-RAI group differs from that in the ESTIMABL2 study. In the ESTIMABL2 study, during a 36-month follow-up period, 4.1% of Non-RAI patients experienced recurrence (BE with Tg levels > 2.0 ng/ml). Although the ATA 2015 guidelines have established treatment response categories for patients receiving RAI, there is currently no standardization for the Non-RAI group. In our Non-RAI group, a significant proportion of patients experienced BE, with rates of 5.8% when using a cut-off of Tg > 2.0 ng/ml and 9.4% with a cut-off of Tg > 1.0 ng/ml.

With a median follow-up period of 6.5 years, we observed an increased ablation rate in the RAI group, while a lower proportion of patients in the Non-RAI group achieved NED (99.3% vs. 90.6%, p < 0.002). Although the ESTIMABL2 study has only published 36-month follow-up data, showing that 95.6% of RAI patients achieved NED, it is likely that long-term outcomes will reach around 98%, as seen in the ESTIMABL1 study. Nevertheless, given that our study cohort had a higher postoperative remnant level compared to the surgical group in these studies, achieving statistical significance in comparisons is unlikely.

Our study, similar to these prospective studies, found that postoperative serum Tg levels are a significant predictor of recurrence in both patient groups (p < 0.001). In the non-RAI group, the area under the curve (AUC) was 0.658 (95% CI: 0.478–0.837), with a cut-off of 2.4 ng/ml yielding a sensitivity and specificity of 62%. For the overall patient cohort, the AUC was 0.646 (95% CI: 0.481–0.810), with a cut-off of 3.1 ng/ml, resulting in a sensitivity of 64% and specificity of 65%. These AUC values indicates that while Tg levels are a useful marker, their predictive accuracy is moderate, but statistically, they are not significant. This suggests that while Tg levels are a useful marker, their accuracy in predicting recurrence is limited.

Our study aims to guide clinics in the treatment of low-risk thyroid cancer with a focus on achieving a low LND rate and minimizing incisions. The low recurrence rates achieved with standard RAI treatment highlight this situation clearly, showing a recurrence difference of up to 7–10 times. Studies that perform a high frequency of central and even lateral LNDs on low-risk patients are far from guiding daily practice. Many guidelines, including those from the American Thyroid Association, do not recommend routine central LND and state that lateral LND should be recommended only with proof of metastatic disease through fine-needle aspiration biopsy. Extensive surgery leads to more significant issues compared to RAI-related salivary gland problems, such as vocal cord paralysis, permanent tracheostomy, permanent hypoparathyroidism, and extensive neck scarring [22,23,24,25]. Although the ESTIMABL2 study did not provide data on complications from extensive surgery, it aims to protect against the potential side effects of RAI. However, its results could only be applied to centers with high-volume endocrine spesific surgeons, whereas a wider range of thyroid surgeons, including general surgeons, could benefit from our findings.

Our study has certain limitations, mainly stemming from its retrospective design and the absence of a multicenter approach. The applicability of the findings may not be universally generalizable due to varying genetic and societal background factors, including differences in iodine deficiency status at the time of diagnosis compared to countries like the USA, France, or the United Kingdom, which have adequate iodine levels [26,27,28]. Additionally, the use of Mahalanobis distance matching in our study introduces specific limitations, notably Dimensionality bias. This bias occurs when the number of variables considered in the matching process is high relative to the sample size, potentially leading to inaccurate or suboptimal matches. As a result, the effectiveness of matching might be compromised, affecting the robustness of our findings.