**4. Discussion**

In this work, we focused our attention on prophylactic CLND and the influence of occult lymph node metastases on the prognosis in patients with DTC.

The first issue to consider is that in our study, there was an intrinsic bias in the selection of patient candidates for pCLND: in fact, this treatment was reserved to cases in which preoperative evaluation including family history, physical examination, US features, FNAC findings, and intraoperative examination suggested a tumor with potentially aggressive behavior.

In our series, the patients who underwent pCLND more frequently had a cytologic diagnosis of PTC and, in addition, were younger than the patients who underwent TT alone. These aspects can be explained by the fact that the surgeon chooses to perform a pCLND in the case of certainty of a malignant nodule: in this case, the operator feels authorized to perform an aggressive intervention to achieve a radical excision of the tumor, mostly in younger patients with a longer life expectancy. On the other hand, when facing patients in whom preoperative diagnosis is uncertain, the surgeon seems inclined to more conservative surgery, preferring a prudent approach to prevent postoperative complications.

However, unlike what could be expected considering the inherent bias of our study in the selection of the surgical approach, at histopathological examination, the two groups appeared comparable: the nodule size and the presence of aggressive features including multicentricity, angioinvasivity, and extrathyroidal extension were similar between the two groups.

At this point, a consideration should be made with regard to the indication for pCLND. The ATA guidelines suggest this approach in cases of advanced primary tumors (T3 or T4). Even if our indications were larger than those purposed by the ATA, our work seems to indicate that preoperative and intraoperative evaluation have low reliability in establishing what tumors have pathologically aggressive features that could benefit from a pCLND.

The incidence of tall cell carcinoma, which has been largely described as an aggressive variant of PTC [29–31], was significantly higher in the pCLND group. This fact could be explained by the higher prevalence of the FNAC diagnostic for malignancy in the pCLND group: in fact, tall cell carcinoma is associated with considerable alterations of the cells that result in a higher incidence of Tir4 and Tir5. In contrast, the higher incidence of follicular variant of PTC in the TT group could be explained by the fact that this subtype of tumor presents less cellular abnormalities, thus is more often associated with inconclusive or negative FNAC, and consequently with a more conservative approach.

As expected, the incidence of lymph node metastases was higher in the pCLND group (25.3%) than in the TT group (4.7%). However, if we consider only the 99 patients in the TT group in which an evaluation of the N status was possible because at least one lymph node was excised, the real incidence of lymph node metastases was 15.2%, thus similar to the other group. Furthermore, this finding explains the fact that the lymph node ratio was significantly higher in the TT group, where the denominator of the fraction was smaller because a smaller number of lymph nodes was excised. These findings suggest that the incidence of lymph node metastases is higher when a larger number of lymph nodes is excised [24,28].

Considering the ATA risk stratification for structural disease recurrence, patients in the pCLND group were more often classified in the intermediate risk group. This could be explained by the fact that pCLND allows for more accurate staging of the tumor, ensuring a better assessment of the N status; in fact, the intermediate class of risk includes tumors in which more than five lymph nodes are involved. Thus, it is likely that pCLND allows to upstage tumors that otherwise would have been classified as low risk tumors, as already reported in the literature [28,32,33]. This fact also explains the higher incidence of patients who underwent RAI therapy after surgery in the pCNLD group.

Overall, our work failed to demonstrate an advantage on prognosis in patients who underwent pCLND. However, if we consider only patients at intermediate and high risk of recurrence, pCLND significantly improved the disease-free survival. We think that this is the key point because tumors at low risk of recurrence have a good prognosis, thus pCLND could be considered an overtreatment that does not modify the course of the disease; on the other hand, the real value of pCLND is expressed in tumors at intermediate and high risk of recurrence, which benefit from an aggressive surgery, with a reduction of recurrence rate.

These findings are in accordance with a recent meta-analysis by Zhao et al., which included 22 studies with over 6000 patients, where pCLND proved to be effective in reducing the risk of loco-regional recurrence [34]; another meta-analysis regarding pCNLD in patients who underwent hemithyroidectomy was consistent with this result [35].

The secondary outcome of our work was to assess whether pCLND was burdened by a higher incidence of postoperative complications. In our series, the occurrence of hypoparathyroidism and RLN injury was higher in the pCLND group, but this difference was not significant, suggesting that pCLND could be a safe procedure with an acceptable incidence of complications. However, this finding should be carefully considered and contextualized: in fact, the same meta-analysis of Zhao et al. that we previously reported, demonstrated a higher incidence of transient and permanent hypoparathyroidism and of transient RLN injury [34].

Furthermore, we must underline that the overall incidence of permanent hypoparathyroidism (9.8%) in our study was higher than the ones usually reported in the literature. As already stated in the Methods section, we defined hypoparathyroidism on the basis of PTH value; probably, this assessment overestimates the incidence of hypoparathyroidism compared to the centers that use only serum calcemia as a criterion, which can be easily influenced from oral calcium supplementation.

Finally, we considered the influence of lymph node metastases on prognosis. When excluding the patients in which the N status was not assessed because no lymph node was excised (pNx), the incidence of recurrent disease was considerably higher in patients with lymph node metastases (pN+), reaching up to 20% than in patients with uninvolved lymph nodes (pN0), with an incidence of 2.8%. It is also interesting to observe that in the group in which the N status was not assessed (pNx), the incidence of recurrences was almost twice (5%) that in patients in the pN0 group, perhaps suggesting that some of these recurrences could have been avoided if a lymphectomy had been performed.

As already mentioned, current guidelines are discordant regarding pCLND. In the ATA guidelines published in 2015, pCLND assumes a marginal role in the treatment of differentiated thyroid carcinoma [1]. The latest NCCN guidelines published in 2019 have eliminated, compared with the previous edition, pCLND in DTC [27]. Considering the negative impact on the prognosis and the high incidence of occult lymph node metastases, and taking into account the difficulties in establishing preoperatively and intraoperatively what tumors could have aggressive behavior, we think that indications for pCLND could be revised in order to achieve more efficacious treatment of aggressive tumors. Such considerations are in accordance with a recent meta-analysis of Zhao et al., which included over 4000 patients, reporting a poor sensitivity of US in detecting metastases of the central compartment (pooled sensitivity of 33%, range 10–57%) with an incidence of lymph node metastases of 48%, suggesting for these reasons that indications to pCLND should be extended to all patients with DTC [36].

This study has some limitations. First, this is a single center, retrospective study. The study was performed in an endemic iodine deficient region, with a high incidence of autoimmune thyroiditis; therefore, the generalization of our results to other populations should be made carefully. Finally, the real incidence of disease recurrence could be underestimated in our study, considering that the mean follow-up is 55.4 months, and that these kind of tumors are generally indolent, and recurrences can appear up to 10 years after surgery.
