Risk Stratification in Differentiated Thyroid Cancer

From Detection to Final Follow-Up

R. Michael Tuttle; Ali S. Alzahrani


J Clin Endocrinol Metab. 2019;104(9):4087-4100. 

In This Article

Abstract and Introduction


Context: Modern management of differentiated thyroid cancer requires individualized care plans that tailor the intensity of therapy and follow-up to the estimated risks of recurrence and disease-specific mortality.

Design: This summary is based on the authors' knowledge and extensive clinical experience, supplemented by review of published review articles, thyroid cancer management guidelines, published staging systems, and original articles identified through a PubMed search, which included terms such as risk stratification, staging, clinical outcomes, and differentiated thyroid cancer.

Main Outcome Measures: In the past, risk stratification in differentiated thyroid cancer usually referred to a static estimate of disease-specific mortality that was based on a small set of clinicopathological features available within a few weeks of completing initial therapy (thyroidectomy, with or without radioactive iodine). Today, risk stratification is a dynamic, active process used to predict the appropriateness for minimalistic initial therapy, disease-specific mortality, risk of recurrence, and the most likely response to initial therapy. Rather than being a static prediction available only after initial therapy, modern risk stratification is a dynamic, iterative process that begins as soon as a suspicious nodule is detected and continues through final follow-up.

Conclusions: Dynamic risk assessment should be used to guide all aspects of thyroid cancer management, beginning before a definitive diagnosis is made and continuing through the final follow-up visit.


Risk stratification in differentiated thyroid cancer has traditionally used a relatively small set of clinical and pathological factors to create models that predict disease-specific mortality or overall survival.[1–7] Although clinically useful, these models provided static estimates of risk with information available within the first few months of initial therapy and demonstrated suboptimal, long-term outcome predictions for any individual patient.[1,6]

Over the last decade, additional models have been developed that provide predictive information with regard to other clinically relevant outcomes, such as the risk of having persistent disease after initial therapy, the risk of structural or biochemical disease recurrence, and the likelihood of going into remission following initial therapy in adult patients with thyroid cancer.[6,8–14] Furthermore, rather than using information that is only available at one particular point in time, these new models emphasize the importance of dynamic risk assessment, where the initial risk assessment is modified over time as new data become available. These dynamic risk assessments allow us to integrate response to therapy assessments with the underlying individual tumor biology to provide real-time risk assessments at any point in the course of the patient's disease. Thus, the modern view of risk stratification begins with the identification of a suspicious nodule (peri-diagnostic period) and continues through the phases of diagnosis, treatment, adjuvant therapy, and follow-up (Figure 1). Whereas the general concepts of risk-adapted management and follow-up are applicable to pediatric thyroid cancer,[15] anaplastic thyroid cancer,[16] and medullary thyroid cancer,[17,18] we will focus this review specifically on differentiated thyroid cancer that has been studied more extensively.

Figure 1.

Risk stratification in thyroid cancer is best viewed as a dynamic, iterative, active process that begins in the peri-diagnostic period and extends through final follow-up. AJCC, American Joint Committee on Cancer; ATA, American Thyroid Association.

From a practical standpoint, postoperatively, we use the eighth edition of the American Joint Committee on Cancer/tumor node metastasis (AJCC/TNM) staging system to predict disease-specific mortality and the American Thyroid Association (ATA) risk stratification system to predict the risk of recurrent or persistent disease (Figure 1).[19,20] These initial risk estimates are then modified over time using the descriptions from the ATA guidelines to define the patient's response to therapy at any point during follow-up, as excellent (no evidence of persistent/recurrent disease), biochemically incomplete [abnormal thyroglobulin (Tg) or rising Tg antibodies in the absence of identifiable structural disease], structurally incomplete (structural evidence of persistent/recurrent disease), or indeterminate (nonspecific findings that cannot be confidently classified as benign or malignant).[21] These modified risk estimates are then used to plan ongoing management.

Recently, the move toward deferred intervention (active surveillance) of very low-risk thyroid cancers and a more minimalistic approach to thyroid surgery has expanded the risk-stratification horizon to include not only the intraoperative and postoperative time periods but also the peri-diagnostic time frame that begins with the detection of a suspicious thyroid nodule (Figure 1).[21–25] In this peri-diagnostic period, it is important to identify low-risk thyroid cancers that may be eligible for either an active surveillance management approach (with or without cytological confirmation) or for a minimalistic surgical intervention, such as thyroid lobectomy without neck dissection.[23,25,26] Conversely, it is equally important to identify, in the peri-diagnostic period, those patients who would be most likely to benefit from more aggressive initial interventions that could include total thyroidectomy, with or without prophylactic or therapeutic neck dissection, radioactive iodine treatment, external beam radiation, or upfront systemic therapy.

It is also important to recognize that highly sensitive disease-detection tools can often detect small foci of papillary thyroid cancer that may not require immediate diagnosis and therapy. The 2015 ATA guidelines provided several specific examples where an observational management approach, often without cytologic confirmation of disease, is recommended as the preferred or alternative management approach to small-volume disease.[21–25] For example, an active surveillance observational management approach is allowed for carefully selected patients with either highly suspicious subcentimeter asymptomatic thyroid nodules without the need for cytologic confirmation or biopsy-proven, very low-risk thyroid cancers, such as intrathyroidal papillary microcarcinomas, in locations not adjacent to trachea or neurovascular structures without evidence of lymph node metastasis.[21] Furthermore, an observational management approach is also allowed for patients with persistent/recurrent small abnormal cervical lymph nodes, asymptomatic stable or slowly growing distant metastasis, and stable or declining abnormal Tg or Tg antibodies.[21]

As it is clear that not all detectable findings require immediate diagnostic or therapeutic intervention, it is imperative that we develop a risk-stratification decision-making framework to differentiate actionable findings from non-actionable findings (Figure 2). Whether we are considering a highly suspicious subcentimeter thyroid nodule without cytologic confirmation of disease, a biopsy-proven thyroid nodule with low-risk thyroid cancer, or persistent/recurrent disease in the neck or elsewhere, we find it useful to consider five key factors that when taken together, allow us to predict the likelihood that a specific tumor focus represents clinically important disease that may require additional evaluations, ongoing observation, or therapeutic intervention (Figure 2). Both tumor size and tumor location are the major factors that determine whether a tumor focus is likely to cause clinically substantial invasion into local structures, such as the recurrent laryngeal nerve, airway, gastrointestinal tract, major vessels, or other important structures.[27–29] A third important factor is the tumor growth rate (measured as tumor volume doubling time), with an observational management approach being much more appropriate for tumors either anticipated to have a slow tumor growth rate or with actual documented slow growth rates over time.[29–33] Obviously, tumors that are either symptomatic or likely to have symptomatic progression would be considered actionable. Finally, patient preference plays a key role when deciding whether a particular lesion is actionable or non-actionable, as it is important to integrate the patient's understanding of the risks and benefits of intervention vs observation with their value system and goals.[34,35] In addition to providing initial guidance as to whether the detectable lesion is actionable at the time of detection, ongoing re-evaluation of these same factors, using the basic concepts of dynamic-risk stratification, can also assist the clinician in the determination of when it is time to transition from an observational management approach to active therapeutic intervention.[22]

Figure 2.

Highly sensitive detection tools often detect small-volume disease that may or may not require action. Key factors that differentiate actionable from non-actionable findings include tumor volume, location, growth rate, symptoms, and patient preference.

Thus, risk stratification has moved from a single postoperative static assessment of the risk of disease-specific mortality to an all-encompassing evaluation of the patient that is continually modified over time, beginning from the first detection of a suspicious thyroid nodule and continuing throughout the life of the patient.