Introduction
Although only 5 per cent of the adult population have clinically palpable thyroid nodules, it is estimated that up to 68 per cent of the population will have at least one thyroid nodule.Reference Guth, Theune, Aberle, Galach and Bamberger1 With the growing use of imaging, incidental thyroid nodules are being increasingly reported,Reference Kant, Davis and Verma2–Reference de Jong, McNamara, Winter, Roskell, Khan and Mihai4 with approximately 16 per cent of CT or MRI studies of the neck identifying such nodules.Reference Yousem, Huang, Loevner and Langlotz5 This trend is reflected in data from the South Korean screening programme, which showed a 15-fold increase in the number of thyroid cancers being identified following its introduction.Reference Ahn, Kim and Welch6 Despite this, the mortality rate from thyroid cancer has remained static.Reference Davies and Welch7 Thyroid cancer is the most common endocrine malignancy but constitutes only 1 per cent of all diagnosed malignancies,Reference Coleman, Babb, Damiecki, Grosclaude, Honjo and Jones8 even though occult thyroid carcinoma is identified in roughly 30 per cent of autopsy studies.Reference Seshadri9
Thyroid nodules can be classified using modern ultrasound scan (USS) techniques, coupled with fine needle aspiration cytology (FNAC) when indicated, as guided by the American Association of Clinical Endocrinologists, American College of Endocrinology and Associazione Medici Endocrinologi (AACE/ACE/AME) guidelines, as well as the British Thyroid Association (BTA) national guidelines.Reference Perros, Boelaert, Colley, Evans, Evans and Gerrard10, Reference Gharib, Papini, Garber, Duick, Harrell and Hegedüs11 Although up to 95 per cent of thyroid nodules are benign, approximately 25 per cent of nodules sent for cytology are classified as indeterminate (Thy3/Bethesda III/IV), with a 5–30 per cent malignancy rate.Reference Kant, Davis and Verma2, Reference Seshadri9 In addition, there is a group of patients for whom FNAC is either not feasible (too small, unsafe location or patient declines) or yields a cytological non-diagnostic result [Thy1/Bethesda1 (1–43 per cent)], with 53 per cent of these remaining non-diagnostic on repeat testing.Reference Seshadri9, Reference Cowen, Dave, Neale, Ward, Repanos and Nasef12
Ultrasound scan findings are key in guiding further management, with the AACE/ACE/AME guidelines suggesting that nodules under 1 cm in diameter can be safely managed conservatively.Reference Gharib, Papini, Garber, Duick, Harrell and Hegedüs11 This includes subcentimetre nodules diagnosed as papillary thyroid cancers (PTC), a subcategory sometimes referred to as microcarcinomas that constitutes 50 per cent of all thyroid cancer diagnoses.Reference Kant, Davis and Verma2, Reference Singh Ospina, Iñiguez-Ariza and Castro13 Molecular testing appears to have a promising role in further evaluating indeterminate thyroid nodules, with reported high negative predictive values (up to 97%Reference Steward, Carty, Sippel, Yang, Sosa and Sipos14). However, this is currently not widely available in the UK and remains an area of ongoing research. PET-CT scanning also has a high negative predictive value (96%)Reference Vriens, de Wilt, van der Wilt, Netea-Maier, Oyen and de Geus-Oei15 but, due to cost and limited accessibility, its use is not currently recommended in this clinical context.Reference Gharib, Papini, Garber, Duick, Harrell and Hegedüs11 Obtaining a second opinion via repeat USS-guided FNAC can improve diagnostic accuracy, as can USS-guided core biopsy, which has been shown to reduce the need for diagnostic surgical procedure by 25%.Reference Alqahtani3
Despite these technological advances, a substantial patient cohort with indeterminate thyroid nodules continues to pose significant challenges to clinical decision-making and contributes to cost pressures across healthcare systems. At present, there remains a paucity of clear long-term guidelines for the management of these patients.Reference Seshadri9 Both the BTA and AACE/ACE/AME guidelines suggest follow-up USS when appropriate for certain indeterminate (U3/TIRADS Class 3) thyroid nodules.Reference Perros, Boelaert, Colley, Evans, Evans and Gerrard10, Reference Gharib, Papini, Garber, Duick, Harrell and Hegedüs11 Although certain studies have made suggestions regarding the recommended interval between scans (every six to 12 weeks, extended to three to five years, with the AACE/ACE/AME suggesting 12 to 24 months, extending to two- or three-yearly intervalsReference Kant, Davis and Verma2, Reference Gharib, Papini, Garber, Duick, Harrell and Hegedüs11), there is limited evidence regarding the appropriate endpoint for such ultrasonographic follow-up. Patel et al. suggested that 2.5 years was adequate based on a cohort of 194 patientsReference Patel, Conybeare, Panesar, Badrol and Sood16 while Gaunt et al. demonstrated no significant disease progression on USS of Thy3f nodules at 27 (16–40) months.Reference Gaunt, Moore, Huvenne, Dhami, Eades and Balasubramanian17 Teliti et al. recently suggested a follow-up duration of 3.3 years for TIRADS 3a (Thy3a) nodules,Reference Teliti, Chytiris, Coperchini, Cerutti, Grillini and Gallo18 whereas Piticchio suggested a 10-year follow-up.Reference Piticchio, Wolde Sellasie, D’Arrigo, Galeano, Barca and Prinzi19
With increasing pressure on healthcare resources, particularly within the National Health Service (NHS)Reference Perros, Boelaert, Colley, Evans, Evans and Gerrard10 and emerging evidence supporting a conservative approach including for patients diagnosed with microcarcinoma,Reference Ito and Miyauchi20 there is a clear need to refine surveillance strategies. The aim of this study was to establish a feasible and safe endpoint for the surveillance of these ‘truly’ indeterminate nodules (BTA U3, Thy3/Thy1).
Materials and methods
A retrospective audit was carried out at a high-volume tertiary referral centre over a five-year period (1 January 2017 to 31 December 2021) using the Computer Picture Archiving and Communications System (PACS). A search for ‘U3’ in all neck, thyroid, parathyroid and associated neck USS was undertaken. The search produced 1652 scans in 1179 patients. Electronic Health Records (EHRs) were subsequently accessed to interrogate USS, cytology and histopathology reports, operation notes, clinic letters and (when appropriate) GP letters. All patients with a first USS report of a U3 thyroid nodule were included in the study, even if this was prior to the start of the study period. Nodules that started as U4 or U5 and then downgraded to U3 were excluded. Nodules that started as U2 were only included when the USS report was upgraded to U3 (indeterminate). The number of USS and whether the nodule was upgraded or downgraded from the initial U3 report, was then noted, along with their dates, cytology of FNACs over time and final histology, when appropriate. In addition, age and comorbidities (at first USS), sex, operation date, date of death or discharge (when appropriate) were also noted. Data were anonymised and stored on a password-encrypted computer in a Microsoft Excel database (Microsoft Corp., Redmond, Washington, USA). The dataset was analysed using Python 3.8.2 (Python Software Foundation, Wilmington, Delaware, USA).21 A p-value <0.05 was considered statistically significant, based on two-tailed hypotheses. KaplanMeierFitter (KMF) and CoxPHFitter (CPH) models were implemented with the use of the KMF and CPH Fitter functions of the Python package lifelines.Reference Davidson-Pilon, Kalderstam, Jacobson, sean-reed and Zivich22
Results and analysis
Of the 1179 patients identified, 66 had ‘no U3 or above’ nodules, 46 were initially U4 or U5 and then downgraded, and 12 were not thyroid nodules. One further USS was for a second opinion only and was therefore not followed up at all. These were therefore all excluded, resulting in 1054 patients being included in the study. The age–sex distribution is depicted in Figure 1, showing a predominance of women (846 women to 208 men).
Age and sex distribution of included patients, which is in keeping with other published data, confirming a reliable sample with no obvious skew.
Of the 1054 patients, 563 (53 per cent) were followed until discharge or death, for an average of 667 (11–4372) days; at the time of writing 234 (22 per cent) are still being followed up, with a further 257 (24 per cent) follow-up plans unclear. Furthermore, 151 of these were truly lost to follow-up (nine of whom recurrently did not attend or refused to be treated, with a further 93 never being referred to ENT/endocrinology. Finally, eight moved out of the area or were treated elsewhere and five were treated in the private sector.
In total, 70 patients (6.64 per cent) were found to have died during the follow-up period. Of those, only two died of thyroid cancer (anaplastic in both cases) with the majority of patients dying of other (unrelated) cancers (28) or cardiovascular disease (14). As this was conducted over the Covid-19 pandemic era, there was also a significant Covid-19 mortality rate (seven) (Figure 2). Lung (six), breast (four), renal (four) cancer and lymphoma (four) sitings were the most common causes of cancer-related mortality. Of the two patients with anaplastic thyroid cancers, one patient presented as a clinical neck lump and was found to have a contralateral U3 nodule. The other patient presented after 11 years of follow-up for her stable U3 nodule, with a clinically new neck lump, five months after her last stable USS. Therefore, neither of these two cancers was picked up on USS surveillance, and neither was the result of dedifferentiation and malignant transformation of the original U3 nodule, for which they had been kept under surveillance.
Cause of death of all patients who have died within the study. This confirms that cardiovascular disease and other cancers are the biggest cause of death, with thyroid cancers making up only 3% (2 patients) and neither of these were directly related to the U3 nodule under surveillance.
Out of the total cohort, 107 patients proceeded to develop a thyroid malignancy. The histogram (Figure 3) shows the distribution of new cancer diagnoses. In total, 84 patients (79 per cent) were diagnosed within the first 12 months and very few (13 per cent) were diagnosed after two years, with only nine (8%) after three years.
Histogram showing time from first ultrasound scan to thyroid cancer diagnosis, when that was made. This confirms that the majority of diagnoses were made within the first year and very few new diagnoses after 3 years.
Survival analysis was undertaken using the Kaplan–Meier estimator to show the time from initial USS to a diagnosis of thyroid cancer for the first five years, summarised in Figure 4 with the green line (confidence intervals are shaded). The largest proportion of cancers (i.e. the steepest part of the curve) was diagnosed within the first 12 months. This result can be compared with the progression to Thy4 (orange) and Thy5 (blue), which is largely static.
Kaplan-Meier chart showing time from first ultrasound scan to confirmed Thy4, Thy5 or thyroid cancer diagnosis, whereby the nodule is no longer indeterminate, but now a likely cancer with a definitive treatment pathway. The widening of the lines with time takes loss of follow-up into account. Despite this, most endpoints are reached within the first 12 months.
Of the 107 patients (10 per cent) who eventually received a diagnosis of cancer, 18 (17 per cent) had an initial FNAC of Thy1 (inadequate), seven (6.6 per cent) of Thy2 (benign), nine (8.4 per cent) of Thy3a (indeterminate with atypical features), 37 (35 per cent) of Thy3f (follicular), 14 (13 per cent) of Thy4 (suspicious) and 17 (16 per cent) of Thy5 (malignant).
On examination, 222 (21 per cent) of the nodules were less than 10 mm in size. Most nodules were more than 10 mm, with the 10–20 mm group being the largest category (30 per cent). The nodule size did not correlate with the Thy status, except that the smaller nodules were less likely to be sampled for FNAC (Figure 5). In total, 559 (53 per cent) nodules remained undiagnosed cytologically, with 427 (41 per cent) unsampled and 132 (13 per cent) classified as Thy1 status. Of the 384 (36 per cent) nodules that were larger than 10 mm, 270 (26 per cent) were unsampled and 114 (11 per cent) were classified as Thy1. The reason for the lack of sampling in this cohort ranged from patient wishes and accessibility of the nodule, to referrer status.
Graph showing the distribution of nodule size. As expected, it shows a normal distribution.
Overall, 162 (15 per cent) nodules had a Thy3f cytology, with a roughly even split with overall histopathology results [51 (31 per cent) malignant, 65 (40 per cent) benign and 46 (28 per cent) had no formal histopathology], which is in keeping with the published literature of Thy3f malignancy rates.
Subgroup analysis
Table 1 demonstrates a subgroup analysis of the rate of malignant diagnosis by Thy status; 1.8 per cent of those not sampled and 4 per cent of Thy1 nodules did progress to a malignant diagnosis, but all within the three-year surveillance period. None of the patients with a U3 nodule and cytology result no higher than Thy1 developed a malignancy after three years from first USS. There were no deaths in these groups attributable to the presenting nodule; 1.8 per cent of the Thy2 progressed to a malignant diagnosis, one of which was after three years, but presenting with an increase in size.
Subgroup analysis of the rate of malignant diagnosis by Thy status
Table 1 Long description
Rates are listed by Thy status for Total, Confirmed malignancies as a count with percent, Diagnosis after three years as a count with percent, and Mortality as a percent. Totals range from 16 in Thy5 to 429 in Not sampled. Confirmed malignancies are 8 of 429 at 1.8 percent for Not sampled, 5 of 124 at 4 percent for Thy1, 4 of 225 at 1.8 percent for Thy2, 7 of 57 at 12.3 percent for Thy3a, 45 of 164 at 27.4 percent for Thy3f, 21 of 39 at 53.8 percent for Thy4, and 14 of 16 at 87.5 percent for Thy5. Diagnosis after three years is 0 for Not sampled and Thy1, 1 of 225 at 0.4 percent for Thy2, 2 of 57 at 3.5 percent for Thy3a, 3 of 164 at 1.8 percent for Thy3f, 3 of 39 at 7.7 percent for Thy4, and 0 for Thy5. Mortality is 0 for all rows except Thy3f, which is 1 of 164 at 0.6 percent and is marked with an asterisk.
* New anaplastic cancer after an 11 years’ follow-up.
Here, 12.3 per cent of the Thy3a nodules were malignant and 27.4 per cent of the Thy3f nodules were malignant. Two of the Thy3a nodules had a malignancy picked up after three years, with no explanation given. Three of the Thy3f nodules were diagnosed with a malignancy after three years. One was initially diagnosed as Thy2 and the other two were later upgraded to U4, one of which was delayed by extensive cardiac investigations. One of the Thy3f patients refused surgical procedure and underwent surveillance for 11 years. An anaplastic nodule was then picked up within six months of the final unchanged USS, presenting as a new neck lump, resulting in the mortality figure.
Table 2 describes the rate of malignant diagnosis according to the size of the nodule on the first USS; 6.8 per cent of nodules smaller than 10 mm, 8.4 per cent between 10 and 25 mm and 14.2 per cent over 25 mm were found to be malignant. Of those for which no size was given, 8.5 per cent were malignant.
Rate of malignant diagnosis according to the size of the nodule on the first USS
Table 2 Long description
Rates are listed by nodule size group, with columns for Total, Confirmed malignancies percent, Diagnosis after three years percent, and Mortality percent. For nodules under 10 mm, total 221, confirmed malignancies 15 at 6.8 percent, diagnosis after three years 4 at 1.8 percent, mortality 0. For nodules 10 to 25 mm, total 403, confirmed malignancies 34 at 8.4 percent, diagnosis after three years 2 at 0.5 percent, mortality 1 at 0.3 percent with an asterisk marker. For nodules over 25 mm, total 359, confirmed malignancies 51 at 14.2 percent, diagnosis after three years 3 at 0.8 percent, mortality 0. For the row labeled No size, total 71, confirmed malignancies 6 at 8.5 percent, diagnosis after three years 1 at 1.4 percent, mortality 0. Across all rows, diagnosis after three years ranges from 0.5 to 1.8 percent, and mortality is zero except in the 10 to 25 mm group.
* New anaplastic after an 11 years’ follow-up.
Discussion
The data show that only two out of 1054 patients (0.2 per cent) died of thyroid cancer and, even in those two patients, the thyroid cancer was not the result of malignant transformation of the original U3 thyroid nodule for which they had been kept on imaging surveillance.
Most patients [84 (79 per cent)] who were eventually diagnosed with thyroid cancer were diagnosed within the first 12 months, with very few being diagnosed after 36 months. For the population of patients who still had an uncertain diagnosis at 36 months (Thy1 or not sampled), no patients had a malignancy diagnosis after a three-year surveillance period.
This study indicates that a three-year follow-up is sufficient for patients with a U3 thyroid nodule and no other sinister features in their history, clinical examination, USS grading or Thy grading. For those with significant comorbidities deemed high risk for general anaesthesia and surgical procedure, it could be argued that follow-up of a U3 thyroid nodule after the initial USS and FNAC would not change their outcome, adding instead to the patient’s and hospital’s burden. This should be a case-by-case multidisciplinary decision made jointly with the patient, as there may be other reasons that warrant a longer follow-up.
Our subgroup analysis by size of nodule (Table 2) demonstrates a trend of increasing chance of malignancy with increasing size. However, the number of nodules over 10 mm (our current guidelines) with a diagnosis after 36 months was too small (0.7 per cent) to recommend any change in guidelines in relation to size.
The incidence of new thyroid nodules and thyroid cancer diagnoses is rising (mostly due to incidental findings on imaging) but mortality has remained static.Reference Davies and Welch7 The impact of this overdiagnosis on patients’ Quality of Life (QoL) and healthcare systems is costly on a global scale. Overdiagnosis commonly results in lifelong follow-up, which, aside from the substantial healthcare cost-related implications, inevitably increases anxiety for patients and their families, as well as disrupting their social and professional lives.Reference Cowen, Dave, Neale, Ward, Repanos and Nasef12, Reference Gaunt, Moore, Huvenne, Dhami, Eades and Balasubramanian17, Reference Garas, Gupta, Mallick, Athanasiou, Patel and Darzi23 These QoL concerns are the same, irrespective of whether a surgical procedure is carried out or not.Reference Garas, Jarral, Tolley, Palazzo, Athanasiou and Zacharakis24 Another adverse effect on QoL as a result of this thyroid cancer ‘epidemic’ is that the increase in thyroid cancer diagnoses, combined with their relatively good prognoses, often results in patients feeling that their cancer is trivialised by both clinicians and family/friends alike.Reference Garas, Jarral, Tolley, Palazzo, Athanasiou and Zacharakis24 For our cohort of patients who are not likely to have thyroid cancer, lifelong follow-up is likely to only adversely impact their QoL rather than serve as a reassurance to them.
It has been shown that patients with low-risk thyroid cancer (under 45 years of age at diagnosis, with stage I disease) can be safely discharged at five years.Reference Garas, Jarral, Tolley, Palazzo, Athanasiou and Zacharakis24 Extrapolating this to our current cohort of over 1000 patients with indeterminate thyroid nodules, not only are they not high risk, but most individuals will ultimately turn out to be benign. Based on this, following up all patients with indeterminate thyroid nodules for a longer time than those with histologically confirmed low-risk differentiated thyroid cancer seems unwarranted.
As in all studies, there are certain limitations that need to be considered. The main limitation relates to the fact that this is a retrospective study from a single unit and, as such, the current findings may not reflect global practice. However, the study is large, with a cohort exceeding 1000 patients, with heterogenous follow-up expected by multiple surgeons. Of equal importance, the follow-up period was far-reaching and in certain cases exceeded five years, which is the standard cutoff for the follow-up of most solid organ cancers. Moreover, the results are in keeping with other published studies.Reference Patel, Conybeare, Panesar, Badrol and Sood16, Reference Gaunt, Moore, Huvenne, Dhami, Eades and Balasubramanian17
Another limitation relates to the fact that most initial USS examinations were undertaken by sonographers as opposed to radiologists (14.9 per cent). Most sonographers do not perform FNACs even if a suspicious nodule is identified and may have limited abilities regarding interpretation of such nodules (particularly if the nodule is incidental); therefore, an additional USS undertaken with a radiologist with FNAC was required before an initial diagnosis could be made. This practice is routine in the NHS today due to the exceedingly high demand for USS and is largely reflective of real-world practices, both in the UK and internationally.Reference Issa, Mueller, Hussein, Albuck, Shama and Toraih25 The number of patients lost to follow-up was low [151 (14.3 per cent)] and, to an extent, inevitable as a result of the healthcare service restrictions due to the Covid-19 pandemic. As a result, no comments can be made regarding follow-up guidelines regarding timing and number of USS, all of which were out the scope of this study and have been reviewed elsewhere.Reference Arambewela, Wijesinghe, Randhawa, Bull, Wadsley and Balasubramaian26
At the same time, there are important strengths characterising this study, making the reported findings of great clinical value with direct application in the real-world setting. Those strengths relate to the large study cohort of 1054 patients from a high-volume tertiary referral centre, one of the largest at national level. Moreover, the study was conducted over a five-year period, with many patients having even longer follow-up times. Furthermore, the dataset is robust and complete, including the time and cause of death for all patients. Finally, because this study was conducted over a timeframe that included the Covid-19 pandemic, many more patients who would have been under surveillance would alternatively have received diagnostic surgical procedures and, therefore, perhaps this is a better cohort for answering the key question regarding the optimal length of follow-up in the real world.
SummaryWhat is already known on the subject
• Thyroid nodules are being increasingly reported with the increased use of imaging modalities.
• Approximately 10 per cent of thyroid nodules will be classified as indeterminate, despite advances in diagnostic techniques.
• Repeat ultrasound scanning and surveillance are currently advocated for these nodules but with no widely accepted set endpoint.
• Health provision is a finite precious resource.
What this paper adds to our understanding
• Our data suggest that two years would be a sensible follow-up time period for truly indeterminate nodules.
• Our data suggest that patients with significant comorbidities could be discharged earlier.
Conclusion
This study has highlighted that most thyroid cancers are diagnosed within 12 months of their first USS, with very few patients dying due to thyroid cancer. Crucially, this study has also shown that patients lacking a definitive benign diagnosis after three years of surveillance with serial USS (and FNAC when indicated), and with no other clinical suspicion (plus those with significant comorbidities at high anaesthetic and surgical risk), may be safely discharged at the clinician’s discretion at that timepoint.
Acknowledgments
We would like to thank the IT department of University Hospitals Birmingham NHS Foundation Trust for their assistance in data collection.
Funding
None. This study did not receive any grants from funding agencies in the public, commercial, or not-for-profit sectors.
Competing interests
No commercial interests to declare.
Ethical standards
Ethical approval was not required as per the principles of the Declaration of Helsinki as this was an anonymous, retrospective analysis of case notes and investigations.
