Ultrasound criteria for risk stratification of thyroid nodules in the previously iodine deficient area of Austria - a single centre, retrospective analysis

Thyroid nodules are a common finding in any given population, with an estimated prevalence of 20-76% on ultrasound examination [1]. The vast majority of these nodules – regardless of whether they were initially palpable or incidental findings (e.g. upon carotid artery sonography) – are benign [2].

Therefore it is of high importance to discern truly benign thyroid nodules from those at higher risk [3]. Thyroid cancer is a tumor entity with steep increase in incidence, albeit with the majority of new cases belonging to the group of low risk tumor stages [4]. How to select the thyroid nodules for further assessment by fine needle aspiration is currently still a matter of debate [5]. Especially with the widespread use of high resolution neck ultrasound, many thyroid nodules are detected as incidentalomas [1, 2].

There is no typical sonographic pattern of thyroid cancer. Various sonographic criteria have been proposed to estimate the risk of malignancy in thyroid nodules. In 2009, two different proposals for a so-called TIRADS scoring system based on ultrasound nodule patterns were published [6, 7]. The concept was inspired by the widely used BIRADS system for assessing breast lesions, but partly been criticized for its complexity. It is still under debate whether this system can easily be applied to routine clinical use [5], and several modifications have been proposed during the last years [8‐10].

In August 2017, the European Thyroid Association Guidelines for Ultrasound Malignancy Risk Stratification of Thyroid Nodules in Adults have been published [11]. Prior to this, the British Thyroid Association has published ultrasound features indicative of malignant nodule in its Guidelines for the management of thyroid cancer in 2014 [12], and the American Thyroid Association presented sonographic patterns suspicious for thyroid cancer in their 2015 guidelines [13], leading to some modifications of the Korean Thyroid Association guidelines in 2016 [14].

A number of studies evaluated the sensitivity and specificity of various criteria in patient samples from e.g. Korea [8, 15, 16], France [9, 10], or Poland [17]. We selected the following six criteria: mild hypoechogenicity [7‐9, 11, 12], marked hypoechogenicity [7‐9, 11, 12], microlobulated or irregular margins [7, 8, 9, 11, 12,], microcalcifications [7‐9, 11, 12], taller than wide shape [7‐9, 11, 12], and the absence of a thin halo [7, 12].

The presence of two or more of the six ultrasound criteria mild hypoechogenicity, marked hypoechogenicity, microlobulated or irregular margins, microcalcifications, taller than wide shape, and a solid component have been proposed to identify nodules at risk in a large Korean study evaluating more than 1600 patients including follow-up data which was the basis for the TIRADS Kwak score [8]. In the TIRADS French score system the four criteria irregular shape, irregular margins, microcalcifications, and marked hypoechogenicity are classified as highly suspect whereas mild hypoechogenicity in the absence of any of the four high suspicious features is the criterion for intermediate risk [11].

Comparable to France and several other European countries, Austria has been a moderately to even severely iodine depleted area with a high prevalence of endemic goiter, functional autonomies and cretinism, especially in the alp regions, for a long time [18]. In Austria, table salt has been iodized by federal law since 1963. The initial concentration of 10 mg KI per kg salt was increased to 20 mg/kg in 1990 when urine iodine secretion was still found to be in the mildly deficient range, currently it is 15-20 mg/kg. This strategy proved to be successful in greatly reducing the incidence of the above mentioned consequences of iodine deficiency [19].

Yet recent data indicate that iodine intake is still insufficient in at least part of the Austrian population, especially in pregnant women [20]. Due to decreased intake of table salt (as advocated for preventing hypertension) and the widespread use of not iodinated industrial salt, iodine intake might be insufficient even in the general population [20, 21].

The aim of our study was to assess six ultrasound criteria indicating thyroid cancer mainly published in the TIRADS Kwak score [8] and the TIRADS French score [9, 10] in the former iodine deficient Greater Vienna area. The study was conducted as a single centre, retrospective analysis of all nodules with postoperative histologic data available over the time course of 10 years (2004-2014). Five blinded experts rated the sonographic images according to the presence of published criteria. Diagnostic values of these criteria were then determined and compared to the published literature.

In light of the widespread use of ultrasonography and frequent findings of thyroid incidentalomas, strategies that systematically and reliably identify those thyroid nodules with a higher risk of malignancy are highly warranted.

In iodine deficient areas like Austria the spectrum of thyroid malignancies differs from iodine sufficient areas with a relatively higher proportion of follicular carcinomas and the occurrence of anaplastic carcinomas in multinodular goiters (which almost disappeared after the introduction of iodization of table salt) [22].

We therefore conducted a retrospective analysis of sonographic images from thyroid nodules with available postoperative histological data from Austrian patients (mainly from Vienna and surroundings). Since the published sonographic criteria relate to papillary carcinomas only, we also calculated a statistical analysis restricted to PTC vs. benign lesions.

At the time the study was designed, several criteria were repeatedly studied and published [7‐9, 12]. According to previously published reports, we analyzed mild hypoechogenicity, marked hypoechogenicity, microlobulated or irregular margins, microcalcifications, and a taller than wide shape. As retrospective assessment of the composition of a nodule (solid, mainly solid, mainly cystic) in two dimensional images seemed problematic to us, we decided not to include this criterion in our analysis. These five criteria are also among the proposals for the TIRADS French [9] and the recently published EU-TIRADS scoring system [11] and are included in the system proposed in the recently published ATA 2015 guidelines [13]. Due to the rather small sample size, however, we decided to restrict our analysis to applying the most widely used criteria without applying a formal scoring system. In addition, we evaluated the absence of a thin perinodular halo. There is no hard evidence from studies, but the halo sign traditionally has been regarded as an important ultrasound sign for the risk stratification of thyroid nodules in central Europe [23, 24] and is still discussed as a suggested feature for thyroid cancer in several diagnostic algorithms [12, 25, 26]. Noteworthy, the halo sign has also recently been discussed also in the European Thyroid Association Guidelines for Ultrasound Malignancy Risk Stratification of Thyroid Nodules in Adults [11] and is one of the nine sonographic features initially identified in the white paper of the ACR TIRADS Committee [26]. Our findings indicate that the absence of a thin halo is also helpful for the diagnosis of thyroid cancer, at least in the former severely iodine depleted population of Austria.

Overall we found a good accordance of the published risk markers with the lesions evaluated in this sample. Our findings support the validity of the TIRADS Kwak criteria [8], the ATA 2015 criteria [13] as well as the TIRADS French [10] and the EU-TIRADS [11] criteria for the former severely iodine deficient area of Austria. One notable exception is the criterion of mild hypoechogenicity. In contrast to marked hypoechogenicity, mild hypoechogenicity could not discriminate between benign and malignant lesions in our sample. This is different to TIRADS Kwak [8]. ATA 2015 does not discriminate between mild and marked hypoechogenicity. In EU-TIRADS, mild hypoechogenicity indicates intermediate risk but no high risk for thyroid cancer [11], indicating that also in the previously iodine deficient population of France, mild hypoechogenicity is not that strict criterion for malignancy as in iodine sufficient regions such as Korea.

There are several possible reasons for this finding. Firstly, all the nodules included in this sample were considered to be reason enough to operate on the patient. Reasons leading to the operation decision were not systematically recorded, but included suspicious results in fine needle aspiration, nodule size, local complaints, patient’s wish, and also suspicion of malignancy in ultrasound. Thus, unambiguously benign lesions were underrepresented or (e.g. simple cysts) even absent from the study sample.Yet another possible reason for this finding is the effect of iodine deficiency that might results in changes in thyroid tissue that are only partially reversible upon iodine supplementation. In a large French sample (France being a formerly iodine deficient region), mild hypoechogenicity conferred only intermediate risk for thyroid cancer [27], in contrast to the iodine sufficient region of Korea.

Of note, 48% of the patients included in the study were born before the year 1963, when iodization of table salt was introduced in Austria and therefore spent at least parts of their early lives in severe iodine deficiency. Only one of the patients was born after 1990 when iodization of table salt in Austria was doubled to 20 mg/kg, because iodine supplementation was still considered insufficient [27]. The different iodine status in Austria and France could also be the explanation mild hypoechogenicity being classified as an intermediate risk factor in the TIRADS French study, but not discriminating in our sample.

A third European country with former iodine deficiency is Poland [28]. Recently an evaluation of four TIRADS classification systems in Polish multinodular goiter patients was published suggesting TIRADS Kwak as a suitable and practicable tool for this patient group. [17].

There is no single ultrasound feature which could reliably distinguish benign from malignant lesions. Some markers (e.g. microcalcifications) have a high specificity but insufficient sensitivity and vice versa. Thus, the combination of several features enhances the diagnostic value of sonography. In a recent Korean study [15], performed on a very large sample of thyroid nodules 10-19 mm in size, a head to head comparison of six risk stratification systems proposed in the literature was performed and yielded, as would be expected, different diagnostic values. Application of TIRADS French (using the number of positive criteria with a cut-off of two or more criteria being present for proposing fine needle aspiration) resulted in a sensitivity of 95% and a specificity of 52%, respectively. In the sample presented here, the cut-off value of two or more positive criteria for PTC resulted in a roughly comparable diagnostic performance (with slightly lower sensitivity of 89% and specificity of 45%). Other stratification systems resulted in even higher sensitivity at the cost of lower specificity.

On the other hand, the pooled expert estimation of malignancy risk yielded sensitivity and specificity very similar to the 3 criteria cut-off condition. Thus, 35% of PTC were mislabelled as benign from the expert panel (although the PTC misjudged as benign were all but one postoperatively staged as pT1).

The experts diagnosed cancer with a specificity of 84%, but the sensitivity was only 52%. For the diagnosis of PTC, the expert’s sensitivity of 64% was only slightly higher. These findings indicate that in real life setting, the accurate differential diagnosis of nodules still remains difficult. Our five thyroidologists, who have long-time experience in interpreting thyroid ultrasound, did not predict malignancy accurately. Using a number of at least two positive ultrasound criteria to define the risk of malignancy yields a higher sensitivity but a lower specificity than expert judgement.

Our findings suggest that mild hypoechogenicity should be clearly differentiated from marked hypoechogenicity. The role of mild hypoechogenicity as a malignancy marker has to be clarified in relation to iodine status: In the sample studied here, in opposition to marked hypoechogenicity - which was significantly more frequent in PTC - mild hypoechogenicity didn’t even show a trend towards higher frequency in PTC. On the other hand, the absence of a thin halo added diagnostic value in the sample presented here and might be worthy of consideration when evaluating nodules for possible malignancy.

The 2015 American Thyroid Association Management Guidelines for Adult Patients with Thyroid Nodules and Differentiated Thyroid Cancer [13] have been published after a long debate including a revised concept for ultrasonographic risk stratification of thyroid nodules. These guidelines do not distinguish between mild and marked hypoechogenicity. Hypoechogenic solid nodules without other risk patterns are classified as intermediately suspicious with a malignancy risk of 10-20%. In our study sample, however, the criterion hypoechogenicity was not found more often in cancer than in benign lesions.

The six criteria were not helpful for diagnosing follicular cancer. As presented in Table 6, the mean number of ultrasound criteria of FTC did not differ from BN, but from PTC. In these patients, the “nodule in nodule sign” could be a criterion for future studies [29]. If, as sometimes is suggested, the scintigraphic pattern (with reduced versus isointense activity) would be helpful in risk stratification of those nodules, is currently unclear. The sample did not include any patient with medullary thyroid cancer. Most probably this is because in Vienna nearly all patients with medullary thyroid cancer are followed up in a single tertiary centre after surgery.

Strengths of the study include the assessment by five thyroid experts in a systematic way and the availability of histological data of all the included nodules. Limitations of the study include the retrospective study design, that only B-mode images were used, the lack of information on the actual iodine status of the patients, the rather small sample size and the single center study design.

In conclusion, we report validity of published ultrasound malignancy risk markers of thyroid nodules in the formerly severely iodine deficient area of Austria for the first time. Our findings support the EU-TIRADS scoring system. With the exception of hypoechogenicity, ultrasound criteria as described in the literature were applicable with good sensitivity for risk adjustment of thyroid nodules in this secondary care setting. Additionally, the missing halo sign was a sensitive malignancy marker in this sample, which might be useful in a screening setting.