Comparison of Initial Risk Stratification Methods in Predicting Treatment Outcome in Differentiated Thyroid Cancer

Hong Kong J Radiol 2021 Mar;24(1):4-14

 

 

Department of Clinical Oncology, Pamela Youde Nethersole Eastern Hospital, Hong Kong

 

  Full Article in PDF

Differentiated thyroid cancer (DTC) is the most common endocrine malignancy and consists of papillary and follicular subtypes.[1] DTC is the fifth most common cancer among women in Hong Kong and statistics have shown that the incidence of DTC was rising in the early 21st century globally, although studies, many in Asia, have indicated that this is due to more screening, and not due to an actual increase in disease rates.[2] [3]

 

Treatments for DTC commonly consist of surgical excision, radioactive iodine (RAI) ablation, and thyroxine suppressive therapy, as well as radiotherapy in some cases of metastatic disease. For some patients with small tumours <1 cm without lymph node involvement or high-risk factors, active surveillance with ultrasonography may also be considered.[4] With adequate treatment, patients with early stages of DTC usually have excellent overall survival (OS) with 5-year OS of approximately ≥95%. Patients presenting in stage III or IV may have 5-year OS of 40%.[5] Therefore, patients with low-risk disease may be given less intensive treatment and follow-up so as to minimise their exposure to radiation and medication. It is therefore important to have good risk stratification methods to assist oncologists in risk stratification of patients to allow formulation of an appropriate treatment and follow-up plan.

 

In the past, several oncology centres or collaborations across the globe devised risk stratification methods for DTC. These include the American Joint Committee on Cancer and the International Union Against Cancer (AJCC/UICC), the Mayo Clinic, European Organisation for Research and Treatment of Cancer (EORTC) Thyroid Cancer Cooperative Group, the Memorial Sloan Kettering (MSK) Cancer Center, the National Thyroid Cancer Treatment Cooperative Study (NTCTCS) and the MD Anderson Cancer Center (MDA). Given the numerous risk stratification methods available and their being mainly drawn from an overseas population, local data and analysis will further aid clinicians to make clinical decisions in local settings.

 

With increasing prevalence of DTC due to increased screening, clinicians are anticipating a higher number of patients with DTC in their clinics. Therefore, apart from the clinical applicability of each risk stratification method, their practicality and reproducibility are also essential factors for consideration of use in day-to-day clinical practice. Some risk stratification methods have included many clinicopathological factors for categorisation while others have complicated scoring systems, making them difficult to be adopted widely in daily clinical practice.

 

In the present study, we aimed to review the treatment outcome of DTC in our hospital and to evaluate the efficacy, practicality, and applicability of different risk stratification methods in a local setting.

 

This was a retrospective review of patients diagnosed with DTC and treated and followed up in the Department of Clinical Oncology of Pamela Youde Nethersole Eastern Hospital from 1 January 1994 to 31 December 2005. Our study protocol was approved by the Hong Kong East Cluster Ethics Committee and conducted in accordance with the Declaration of Helsinki. Patient consent was not required as it was a retrospective study based on historical clinical records.

 

In our centre, a department protocol on management of DTC is available for all doctors as a reference. In general, patients with DTC were considered for adjuvant RAI if they had undergone total thyroidectomy, unless they had very low-risk disease with a favourable disease profile. Otherwise, a RAI dosage of approximately 1100 to 3700 MBq was recommended for patients, depending on their risk profile, such as tumour size and presence of lymph node involvement. After ingestion of RAI, patients would have whole-body scans within 1 week and a repeat scan approximately 6 months later. Patients were started on a thyroid-stimulating hormone–suppressive dose of thyroxine unless contra-indicated, with dosage adjusted later based on treatment response. If patients continued to have RAI-avid residual disease on follow-up scans, RAI was repeated. External beam radiotherapy of up to 70 Gy over 35 fractions was considered for patients with structural residual disease not amendable to further surgery.[4]

 

Patient hospital records with primary diagnosis of ‘thyroid cancer’ during the above stated period were retrieved from the Clinical Data Analysis and Reporting System. Records including consultation notes, operative records, pathology reports, imaging reports including ultrasonography and radioiodine scans, and blood results were reviewed.

 

A MEDLINE search on DTC, staging and risk stratification methods was done and found 17 documented risk stratification methods. They include the AJCC/UICC Tumor-Node-Metastasis system (AJCC/UICC TNM) 7th edition, EORTC classification,[6] the Metastases, Age, Completeness of surgery, Invasion and Size system (MACIS),[7] the Age, Grade, Extent and Size classification (AGES),[8] the Age, Metastases, Extent and Size system (AMES),[9] the MSKCC Grade, Age, Metastases, Extent and Size classification (MSKCC GAMES),[10] the MDA,[11] the Clinical Class system (Clinical Class),[12] the University of Münster system (Münster),[13] the NTCTCS classification,[14] the Ohio State University system (OSU),[15] the Noguchi classification (Noguchi),[16] the University of Murcia system (Murcia),[17] the Cancer Institute Hospital system (CIH),[18] the Ankara Oncology Training and Research Hospital system (Ankara),[19] the Sex, Age and Grade system (SAG),[20] and the DNA, Age, Metastases, Extent and Size system (DAMES).[21]

 

Of the 17 risk stratification systems, only nine could be applied to our patients based on the patient data available. The methods excluded and the various reasons accountable for exclusion are: AGES (lack of tumour grading data), OSU (lack of data on number of intrathyroidal foci), SAG (lack of microscopic description of nuclear atypia), NTCTCS (lack of tumour grading data), CIH (lack of data on size of lymph nodes), Ankara (lack of angioinvasion data), Murcia (lack of histological subtype data), and DAMES (lack of DNA ploidy data).

 

The calculations or categories used in the nine risk stratification methods analysed in our study are summarised in Table 1. Disease-free survival (DFS) and OS were evaluated. DFS was defined as the date of diagnosis to the date of relapse of DTC or death. OS was defined as the date of diagnosis to the date of death from any cause.

 

Table 1. Calculation and categorisation of risk stratification methods analysed.

 

DFS and OS of each risk classification system were analysed by comparing their Kaplan–Meier plots with log rank test. The Harrell’s C-index was used to evaluate the discriminative ability of different risk stratification methods in identifying low- or high-risk patients. A C-index of 1 implies that the risk stratification method can perfectly select individuals with discordant events, while a C-index of 0.5 shows that the risk stratification method fails to show any discriminative ability. C-index of >0.65 is considered an acceptable model for predicting outcome. Statistical analysis was performed using SPSS (Window version 22.0; IBM Corp, Armonk [NY], United States) and R statistical software version 3.3.3.

 

A total of 321 patients were included in our study. The median follow-up time was 143.3 months (range, 0.3-461.7). The percentage of patients with <1 year of follow-up was 2.49%.

 

Demographic distribution of our study population is tabulated in Table 2. Females accounted for 81.3% of the study population with median age 46 years. A total of 79.4% of patients presented with thyroid nodules and only 4% of patients reported symptoms of thyrotoxicosis. Papillary carcinoma and follicular carcinoma accounted for 91.6% and 8.4% of the study cohort, respectively. About one-third (34%) of our patients had multifocal disease confirmed on histological examination.

 

A total of 94.4% of patients underwent total thyroidectomy with 29.9% patients receiving planned selective neck dissection according to the preoperative lymph node status. Proportions of patients achieving R0, R1 and R2 resections were 76%, 15%, and 2.2%, respectively. Most of the patients (91.6%) received at least one dose of RAI. For the first postoperative RAI, the most commonly used dose was 80 mCi (93.9%) according to our department protocol. On the first post-ablation whole-body scan, 58.5% of patients showed uptake over the thyroid bed only. On the subsequent follow-up whole-body scan 6 months after RAI, 67.3% of patients showed no significant uptake in the entire body. Of the patients receiving postoperative RAI, 27.9% proceeded to receive a second-dose RAI while 8.9% further received a third RAI. Only about 5% received postoperative external beam radiotherapy with a median dose of 60 Gy (range, 50-64).

 

In our study, the 10-year and 15-year OS were 93% and 88.2%, respectively. The 10-year and 15-year DFS were 90.9% and 77%, respectively. The DFS Kaplan–Meier curves for risk stratification methods with C-index of ≥0.65 are shown in Figure 1. The OS Kaplan–Meier curves for risk stratification methods with C-index of ≥0.65 are shown in Figure 2.

 

Figure 1. Disease-free survival curves by the (a) EROTC, (b) MDA, (c) AJCC/UICC TNM, and (d) MACIS risk stratification methods.

 

Figure 2. Overall survival curves of (a) EORTC, (b) MDA, (c) AJCC/UICC TNM, (d) Noguchi, (e) MSKCC GAMES, and (f) MACIS risk stratification methods.

 

Distribution of patients in the different risk stratification methods is tabulated in Table 3. Table 4 shows the C-index of the risk stratification methods and their rankings, in terms of DFS and OS, respectively. Comparing the C-indices, EORTC had the highest discriminative power for both predicting recurrence (C-index = 0.694, 95% confidence interval [CI] = 0.599-0.790) and OS (C-index = 0.825, 95% CI = 0.725-0.924). MDA came in second in predicting recurrence (C-index = 0.669, 95% CI = 0.584-0.753) and OS (C-index = 0.769, 95% CI = 0.690-0.849). The AJCC/UICC TNM staging, which we commonly used in our daily practice, ranked third with a C-index of 0.651 (95% CI = 0.561-0.740) for predicting recurrence and C-index of 0.747 (95% CI = 0.630-0.863) for predicting OS.

 

Table 3. Patient distribution according to risk stratification methods (n = 321).

 

Table 4. Harrell’s C-indices and ranking of various risk stratification methods for disease-free and overall survivals.

 

Precise clinicopathological staging is crucial in both patient management and communication in doctor-patient and doctor-doctor settings. A good staging system provides reliable estimation of risk of recurrence and disease-specific mortality for individual patients, hence allowing clinicians to make evidence-based decisions on the aggressiveness of adjuvant treatment, intensity of follow-up, and for patient education and counselling.[22] Moreover, a widely adopted staging and risk stratification system can also allow clinicians around the globe to communicate effectively, providing a common language for medical discussion and research studies. An effective staging system should adequately offer predictability, practicality, and reproducibility in order to serve the multiple purposes mentioned above.

 

Many systems have been proposed and studied for risk stratification of DTC. Each system employs a slightly different set of clinical and pathological factors. Age and presence of metastases are well-recognised prognostic factors in DTC and hence included in most risk stratification methods.[23] [24] [25] [26] In our study, the EORTC, MDA, and AJCC/UICC TNM methods came in the first three positions in terms of discriminative ability to differentiate low-risk from high-risk population, which is in concordance with previous studies of this topic.[27] [28] This may be accountable by the heavy weighting given to age or the presence of distant metastases in these systems. For instance, in the EORTC calculation, distant metastases give the highest contributing score of 15 to 30 depending on the number of metastases. Meanwhile, for AJCC/UICC TNM staging method, all patients aged <45 years belong to stage I or II regardless of tumour size, local extent, or nodal involvement. For the MDA method, age and presence of distant metastases are the only consideration factors, eliminating other less influential clinicopathological factors included in other systems.

 

Direct comparison of the different risk stratification methods can be difficult as each method includes different histologies of DTC. All histologies of thyroid cancer were included in the AJCC/UICC TNM and EORTC systems, while only papillary thyroid cancer is included in the MACIS, Clinical Class, and Noguchi methods. Both papillary and follicular thyroid cancers are included in the MDA, AMES, MSKCC GAMES, and Münster methods. As papillary, follicular, medullary and anaplastic thyroid cancer each has a distinct disease behaviour and prognostic curve, the survival data across two or more histology types are more difficult to be interpreted and compared directly with data from other studies. We recognise that medullary and anaplastic thyroid cancers are rare disease entities, and it is challenging to recruit adequate number of cases for reliable analysis.

 

Practicality and reproducibility are essential consideration factors in everyday clinical practice. Some risk stratification systems involve a more complicated calculation, allowing different clinicopathological factors to have different weighting on the risk stratification outcome. For instance, EORTC assigns a score to each clinicopathological factor (eg, 12 for male sex, 15 for one distant metastasis) and requires the sum of the scores. Another example is in MACIS, where clinicians need to perform multiplications and summation of various clinicopathological factors (eg, multiplying age by 0.08 for patients aged ≥40 years, multiplying tumour size in cm by 0.3). These may render certain risk stratification methods less convenient to be applied in day-to-day practice. Therefore, in the aspect of practicality, the AJCC/UICC TNM and MDA systems appear more intuitive and clinician-friendly for interpretation and patient counselling. The pros and cons of each risk stratification method with a good discriminative power are listed in Table 5.

 

Table 5. Pros and cons of various risk stratification methods.

 

Several risk stratification methods have not been included in our study as certain required variables are not routinely available in our pathology reports. Some of these pathological characteristics include size of lymph nodes, nuclear atypias, and DNA ploidy. According to the thyroid cancer structured reporting protocol by the International Collaboration on Cancer Reporting, certain parameters, such as exact number of tumour foci, size of lymph nodes, and tumour grading are not required to be included in the pathology reports.[29] These parameters have only been reported in limited studies to be correlated with prognostication. A prospective study in collaboration with pathologists will be needed if these parameters are to be investigated in the future.

 

Limitations in our study include the inherent bias in its retrospective design and the lack of complete pathological data required for all risk stratification methods. The long follow-up time in this study implies that there may have been changes in practice, especially with the emerging utility of thyroglobulin to guide clinical management. In addition, further analysis with the newer version of the AJCC/UICC TNM staging system would elucidate its applicability in our local population.

 

Building on the basis of initial risk stratification, the role of dynamic risk stratification is also coming to light in the current era.[30] [31] We recognise that initial staging is only the beginning of the risk stratification process. A more tailor-made risk stratification specific to each individual patient will require further clinical information on follow-up, such as postoperative serum thyroglobulin and post-RAI scan findings.[31] [32] [33] [34] With the increasing availability of molecular analysis in the modern era, certain molecular markers such as BRAF or RET may also be factored in for more accurate risk stratification.[35] [36] Therefore, although both AJCC/UICC TNM and MDA can offer a reliable and convenient means for initial risk stratification, we recommend that it should be coupled with the American Thyroid Association risk stratification system and further dynamic risk stratification (based on presence of biochemical or structural evidence of relapse over time) to best estimate an individual patient’s survival and risk of recurrence.

 

AJCC/UICC TNM, EORTC, and MDA risk stratification methods are all reliable tools for initial risk stratification in DTC. We recommend the use of AJCC/UICC TNM staging in daily practice for its practicality and global recognition, in combination with the American Thyroid Association and dynamic risk stratification to best predict individualised risk of recurrence and survival in patients with DTC.