Both the prevalence and the course of human diseases are influenced by patient age. In the field of thyroid cancer, age is a well-recognized prognostic risk factor, patients with differentiated thyroid carcinoma (DTC) older than 55 years having a higher risk of dying from DTC [20, 21] and developing structural recurrence [9] than younger age groups. In addition, DTC among the pediatric population has a different biology from that of adults [23]. This is so relevant that age was proposed as the pivotal feature during the initial assessment of the risk of recurrence of DTC [24]. Thus, according to the above concepts, it is reasonable to consider that the performance of TIRADSs may change with age. The present study reports interesting new findings that warrant extensive discussion.

First, despite some slight differences between age groups, the performance of ACR-TIRADS in discriminating benign from malignant TNs did not significantly change across the patient lifespan. In fact, both the distribution by category of TNs and the rate of FNAC indication were not significantly different between young and elderly patients. Moreover, the rate of UN-FNAC and NNP-FNAC also remained unchanged over the patient lifespan. Second, the frequency of cancers observed in the age groups was quite similar. In addition, there were no difference between groups in gender, TN size, and cancer size. On the other hand, a mildly significant difference was found in cancer types between patients older than 65 years versus adult patients of younger age (mostly middle adulthood).

Besides a few studies including either young or elderly patients [10,11,12,13,14], literature on the age effect on TIRADS is scant. A recent study by Walter et al. which evaluated whether cancer prevalence changes according to Bethesda FNAC categories found that the rate of malignant FNAC decreases as the age of the patient increases. [25]. Moreover, they observed that age significantly influences the malignancy rates of ACR-TIRADS. An interesting example of this is that the cancer rate of TR5 of patients aged 20–39 years, 40–59 years, and ≥ 60 years was 64.7%, 45.9%, and 22.6%, respectively. Even if the latter study used FNAC as the reference standard with possible selection bias, the results are of particular interest at this time when the I-TIRADS project is ongoing [18]. The present study is not likely to confirm those data since it is probable that different case selection used (i.e., histological reference from a series of patients undergoing surgery) determines apparent discordance between results. However, when we look at FNAC series, we find populations undergoing FNAC for many reasons (i.e., strict indication according to TIRADS and other). In some cases, FNAC may be performed only as a cautionary measure, even if worrisome US signs are lacking. Young people may be incidentally diagnosed with TN, and physicians are prone to require further investigation, especially when the lesion is suspicious. In a retrospective review of US images of patients undergoing surgery, it was noted that these effects usually disappear. In addition, while FNAC series substantially include only PTC that can be recognized on cytology [26], histological series like the present one also include FTC and MTC that are not or are only partially detectable on FNAC [27, 28]. The most accurate ROC-derived threshold of ACR-TIRADS to distinguish benign from malignant nodule observed between the third tertile and the other tertiles (Fig. 3) merits consideration. These data appear to be in line with the different cancer type distribution according to age (Table 1). In fact, the FTC rate in elderly individuals was higher than in younger patients. Since FTC generally presents as isoechoic without typical features of PTC [27], the best cut-off decreased from > 4 in the first two tertiles to > 3 in the third. This is yet more firm proof that the TIRADSs were basically conceived on the basis of PTC US presentation [26]. Most studies based on FNAC series do not include FTC that is usually cytologically indeterminate, has low/intermediate US risk, and then may be infrequently operated on. The latter data can be reported only in histological series like the present one [26]. From this point of view, a slightly different performance of ACR-TIRADS across ages is present. Accordingly, we should consider that elderly people may harbor different types of cancer compared to young patients, resulting in lower TIRADSs reliability. In any case, the cytological and histological series are useful to further extend our knowledge about TIRADS performance, while the upcoming I-TIRADS will benefit from the additional data. To summarize, the present study shows that when we consider all histological series and cancer types, the performance of ACR-TIRADS does not change according to patient age. The present findings represent a novelty in the TIRADS literature.

Ultrasonography is a safe and low-cost procedure and is widely used in TN patients. However, how to improve the accuracy of thyroid US is currently a hot topic. Researchers have indeed studied the limitations of US and indicated that implementing the use of AI may provide greater insight into thyroid disorders thereby improving their treatment [15,16,17, 29]. At this time, most publications / articles in this field are focused on assessing TN US images related to datasets, despeckling algorithms, segmentation algorithms, and classification algorithms. These data are encouraging and open up new perspectives for clinicians. Furthermore, as investigated in the present study, US and TIRADSs still have non-imaging-related limitations to be resolved. On the whole, TIRADSs are radiological recommendations that do not consider the patient profile (e.g., age, gender, comorbidities, and medications). Our data showed that patient age should not modify the performance of TIRADS, although our findings diverge from those reported by other authors [10,11,12,13,14]. Both our own and previous studies may suffer from selection bias (see above). Furthermore, whether we can currently consider TIRADSs reliable independently of patient age remains to be clarified. In addition, no studies have demonstrated whether TIRADSs can diagnose cancer at an early stage, even considering that ACR-TIRADS indicates FNAC when the TN size is large [30]. From this standpoint, further algorithm-based proposals including patient profile are of particular interest for clinicians. Our advice is that further studies be conducted comparing the performance of humans using TIRADSs versus other imaging-based applications and models. Examples include filtering algorithms [31] or semantic segmentation models based on encoder-decoder architectures or computer-aided diagnosis [32].

The limitations of the present study should be mentioned. First, this is a retrospective analysis of US images of patients undergoing thyroid surgery for any indication. Second, reviewing US images stored in PACS by multiple raters may represent bias, even if discordant assessment was resolved via mutual consultation to achieve a final consensus. Third, pediatric patients were not found in the institutional database during the study period, making analysis of this age group non-feasible. A major strength is that the present study includes only histological diagnosis, thus avoiding the bias of cytological evaluation as a reference standard. Overall, the sample size and the selection strategy should be taken into account as a potential bias of the results. As previously mentioned, we aimed to enroll a series of patients with histological diagnosis to avoid selection bias included in the FNAC series. Even if assuming only histology as a reference standard cannot exclude biases (i.e., a histological series enrolls only patients with indication for surgery), we chose to enroll histological cases to include cancers other than PTC [23,24,25], increase as much as possible the number of TNs of each patient, and avoid the limitation associated with FNAC-based series. However, the cancer rate that we found (i.e., 46 among 114 patients) may represent an overestimation of the generally anticipated cancer frequency in TN series. Nevertheless, all thyroid carcinomas were included in our series with satisfactory relative frequencies (36 PTCs, nine FTCs, and one MTC) and the TIRADS performance could be explored not only in PTC, as usually occurs in FNAC series.

In conclusion, the present study shows that the performance of ACR-TIRADS is not significantly influenced by patient age across his/her lifespan. More studies are needed to verify these data in order to contribute to paving the way for I-TIRADS.