Introduction
Thyroid cancer currently accounts for 3.1% of all new cancer cases in the US but is projected to be the fourth most common cancer in the nation by 2030 [
1,
2]. The annual incidence has been steadily rising over the last four decades. There was an approximately 4% annual increase during 2005–2014 which has been attributed in part to improved detection of subclinical, indolent cancers [
2‐
4]. However, this short term rise is also consistent with the long term annual increase of 3% observed during 1974–2013 [
5].
Thyroid cancers are classified by three main histological types, differentiated thyroid cancer (DTC)—which includes papillary, follicular, Hürthle cell and poorly differentiated carcinomas—medullary thyroid cancer (MTC) and anaplastic thyroid cancer. Among patients diagnosed with thyroid carcinoma during 2010–2014, DTC was by far the most common, accounting for 96% of all diagnoses [
6]. Approximately 93% of DTC cases were papillary, 5% were follicular and 2% were Hürthle cell; poorly differentiated carcinomas were the least common but were not well quantified [
6]. Medullary and anaplastic carcinomas represented 1.7% and 0.8% of all thyroid cancer diagnoses, respectively [
6]. Anaplastic carcinoma is the most aggressive thyroid cancer with mortality approaching 100% [
7]. Although most patients with DTC and MTC can be cured by thyroidectomy, particularly when diagnosed at early stages, some DTC cases may also be managed with post-surgical radioiodine ablation (RAI) therapy [
2,
8,
9]. Prognosis is altered dramatically with the development of metastatic disease [
6]. The 10-year survival for DTC and MTC cases with distant metastases is 40–42% [
10,
11]. Among DTC patients who are radioiodine refractory, 10-year survival declines to 10% with a median survival of 3–4 years [
10,
12]. Until recently, treatment options for advanced DTC and MTC have been limited since conventional chemotherapy is largely ineffective in controlling disease progression [
9].
Advances in the understanding of the molecular pathogenesis of thyroid cancer have led to the development of targeted agents aimed at combating advanced disease. Small molecule kinase inhibitors (SMKIs) modulate cell signaling pathways responsible for tumor cell growth and proliferation [
13,
14]. The National Comprehensive Cancer Network (NCCN) and American Thyroid Association favor the use of SMKIs for management of radioiodine refractory metastatic DTC and progressive or symptomatic MTC [
9,
15]. The US Food and Drug Administration (FDA) approved sorafenib (in 2013) [
16] and lenvatinib (in 2015) [
17] for treatment of locally recurrent or metastatic, progressive DTC refractory to RAI. The NCCN recommends consideration of lenvatinib (preferred) or sorafenib for progressive and/or symptomatic disease with selection individualized for each patient according to the likelihood of response and comorbid conditions [
9]. Vandetanib (in 2011) [
18] and cabozantinib (in 2012) [
19] are FDA-approved for progressive, metastatic or unresectable MTC. Both are NCCN-recommended in the setting of symptomatic disease or progression [
9]. Both are indicated as first-line therapy and there are no data indicating which SMKI should be considered first, hence leaving the decision to the discretion of the provider. Often MTC patients will receive both agents during the course of their disease. Other commercially available SMKIs may be considered for DTC and MTC if clinical trials are not an option for a patient [
9] and are used off-label in clinical practice [
20].
The current literature on SMKI treatment patterns shows extensive use of these agents in the first-line setting among patients with advanced DTC and MTC [
20‐
23]. Further, studies that examined progression-free survival among patients treated with SMKIs in both the first- and second-line setting found evidence of clinical benefit in the second-line setting, suggesting that SMKI should be considered after first-line failure [
20,
22‐
24]. While previous research has provided important insight into the use and efficacy of SMKIs beyond the clinical trial setting, studies to date were conducted prior to the commercial availability of lenvatinib in 2015 and most were single site investigations [
20‐
23]. The overall aim of this study is to provide a more contemporary assessment of SMKI treatment patterns in the US with a broader representation of patients. To this end, a national administrative claims database was used to examine real-world prescribing patterns among patients who initiated any NCCN-recommended SMKI therapy for advanced thyroid cancer during 2010–2016. In this analysis, we examined SMKI regimens used in the first-, second- and third-line of therapy and duration of each line of therapy.
Discussion
Multiple SMKI therapies are currently available for patients with advanced thyroid cancer [
9]. Our results suggest that SMKI treatments patterns have changed since the introduction of lenvatinib in 2015. Over the entire study period (2010-2016), sorafenib was the predominant regimen in both the first-line (36.9% of all patients) and second-line setting (24.7%). Among patients starting SMKI therapy in 2015 and later, lenvatinib became the most common first-line regimen: 43.4% of patients in 2015 and 66.7% of patients in 2016 received lenvatinib in the first line of therapy. Lenvatinib was also the most common second-line regimen among patients initiating therapy in 2015.
The large majority of SMKI regimens were single agents which is consistent with previous retrospective analyses [
22,
23]. Our results also show frequent use of agents that are NCCN-recommended but not approved by FDA for advanced DTC and MTC; sunitinib was often used in all lines of therapy and pazopanib was a common first-line regimen. Patients were observed on average for 16.6 months after initiation of the first line of therapy but duration of follow-up varied considerably between SMKI regimens. Patients receiving lenvatinib in first-line therapy had the shortest mean duration of follow-up time (8.3 months) while patients treated with vandetanib in the first-line setting averaged the longest duration of follow-up (22.9 months). This pattern likely coincides with the timing of FDA approval of agents for advanced thyroid cancer relative to the end date for inclusion in the study (May 2016): lenvatinib was the last drug to approved by FDA (2015, for DTC) while vandetanib received the earliest approval (2011, for MTC). The remaining single agent regimens either received FDA approval for advanced thyroid cancer by 2013 (cabozantinib for MTC and sorafenib for DTC) or were available for other indications before the study start date (pazopanib, sorafenib and sunitinib).
We examined time to treatment discontinuation of each SMKI regimen in first-, second- and third-line therapy as a proxy for clinical benefit as cancer treatments are typically discontinued when they are no longer effective or tolerable. Treatment duration is a composite measure that does not distinguish between efficacy, tolerability and other causes of therapy discontinuation. However, it is a commonly reported metric in clinical trials and often a key outcome in observational studies of oncology treatments [
30‐
33]. Kaplan–Meier estimates of median time to discontinuation differed significantly by regimen in the first line of therapy: sorafenib had the shortest estimate (2.8 months) and vandetanib had the longest estimate (16.2 months). No significant differences between regimens were observed in subsequent lines of therapy. Overall, estimates of treatment duration in the first and later lines of therapy were lower than observed in clinical trials [
24,
34,
35]. This likely reflects differences in the profile of patients and patient management between clinical practice vs. the controlled environment of clinical trials. Moreover, large inter-patient variability in SMKI treatment duration has been observed outside the clinical trial setting [
21]. The side effect profile of SMKI regimens may also have influenced treatment duration [
36]. Sorafenib use has been associated with occasional lack of tolerability, necessitating dose reductions/re-escalation, drug interruptions/re-start and withdrawal (particularly in early treatment cycles) [
20,
21,
37] while vandetanib is generally better tolerated than sorafenib [
36]. In addition, comorbid conditions may have impacted treatment duration or warranted dose reductions. It is also possible that the pharmacy claims-based algorithm used to define lines of therapy and duration of therapy in this study [
28,
29] underestimated duration of therapy because dose reductions or drug interruptions were not observable in this analysis. The definition of each line of therapy considered medication fills and timing of the refills. If a 60-day or greater gap in medication supply occurred prior to the initiation of a new agent, the gap signaled the end of a line of therapy. However, a 50% dose reduction in a 60-day prescription could cause a 60-day delay in the next prescription refill and result in misclassifying a patient as ending that line of therapy.
Patients receiving sorafenib, sunitinib, lenvatinib or vandetanib as first-line therapy commonly received the same regimen in the second line of therapy. A similar pattern was evident in the third-line setting but to a lesser degree. This finding was unexpected given the availability of several SMKI for use in this population [
9]. Sequential use of the same therapy may also reflect a drug holiday to allow patients to recover from side effects followed by a re-challenge, or, as noted above, a dose reduction. Alternatively, patients may have been re-challenged with the same drug for reasons other than side effects. For example, a patient may have started therapy and subsequently stopped if there was no evidence of disease progression but later re-started the same therapy after disease progression occurred. Use of second and later lines of therapy was generally modest. Among the subset of patients with ample follow-up time (which produced the largest estimates), only 53.7% of patients received a second or later line of therapy. This is consistent with the rate of 49% observed in a previous study examining multiple first-line SMKIs in advanced DTC and MTC [
23]. The results of recent retrospective studies support clinical benefit of SMKI in the second-line setting [
20,
22,
23]. In a French study of patients treated with sunitinib or sorafenib, median progression-free survival was similar in both the first-line (7.0 months) and second-line setting (6.7 months) [
22]. Owonikoko et al. also observed clinical benefit of second-line therapy in a US study examining several first- and second-line SMKI therapies; however efficacy diminished in the second-line vs. first-line setting (median progression-free survival 4.6 months vs. 16.2 months) [
23]. In addition, the results of a phase 3 clinical trial of lenvatinib showed relatively similar median progression-free survival among patients who had received no prior tyrosine kinase inhibitor therapy (18.7 months) and among patients treated with one previous tyrosine kinase inhibitor regimen (15.1 months) [
24]. Our results also suggest benefit of SMKI in the second-line setting: among all patients, median times to treatment discontinuation were 5.0 months in first-line vs. 5.1 months in second-line treatment. However, it is important to note that our estimates of treatment duration for the second-line setting represented 35.4% of the patients with first-line therapy and should not be interpreted to suggest similar efficacy of SMKI in first- and second-line setting. Further, there were insufficient data to identify specific treatment sequencing to substantiate benefit in the second-line setting.
To our knowledge, this is the first study to examine real-world SMKI treatment patterns since the approval of lenvatinib, in 2015. The primary strength of this study is that the patient sample, drawn from a large, national commercial insurance database, was considerably larger than previous studies and had broader geographic representation [
20‐
23]. The demographics of our study sample (mean age: 61.2 years; 51.6% female) are consistent with patients undergoing SMKI treatment in recent clinical trials (median age 63–64 years; 48–50% female [
24,
38]). Our results suggest that, at most, only a slight majority of patients receive a second or later line of SMKI therapy in real-world clinical practice. Future research should examine the impact of treatment sequencing on clinical outcomes. While our study lends support to previous research suggesting clinical benefit in the second-line setting [
20,
22,
23], future research is needed to establish consensus for the optimal choice of SMKI in second and later lines of therapy. Assessment of patient characteristics that are predictive of clinical outcomes in second and later lines of therapy is also needed to identify patients who would benefit from SMKI therapy after progression on first-line therapy. In addition, the availability of SMKI regimens for use in routine clinical practice is still relatively recent, particularly for lenvatinib. Thus, some patients may already have experienced significant disease progression prior to SMKI treatment. Future studies are needed to build on our findings as the use of SKMI regimens becomes further established in the general population.
The results of our study are limited by its retrospective nature. In particular, our results are based on data from an administrative claims database and have limitations typical not only of claims-based research in general but also the specific disease state examined. First, claims data are collected for reimbursement, not research, and may be subject to miscoding and errors of omission. Any errors or inconsistencies in the documentation of codes for diagnoses or medications may lead to misclassification of patients into regimen cohorts. However, the rate of these errors is expected to be low because of the comprehensive documentation of codes required for reimbursement. Furthermore, we would not expect any systematic differences in the occurrence of errors between the cohorts of interest. Secondly, claims data lack certain clinical information (e.g., tumor staging, histology, disease progression, radioiodine refractory status, prescribing patterns) that may be relevant to the treatment patterns observed in this study. For example, ICD-9-CM and ICD-10-CM diagnosis codes were used to identify patients with thyroid cancer; however, these codes do not distinguish DTC from MTC or other forms of primary thyroid cancer which may have influenced SMKI treatment regimen. Thirdly, treatment patterns were identified based on pharmacy claims for SMKI. However, pharmacy claims do not indicate if the patient took the medication at all or as prescribed. Fourth, we used an established prescription claims-based algorithm to identify distinct lines of therapy [
28,
29]. This included treatment gaps of 60 days or more (based on days’ supply of therapy) to demark the end of a line of therapy. However, the initiation of a new line of therapy due to disease progression after a 60-day or greater break in therapy is indistinguishable from restarting the same regimen after a drug interruption or a lengthy dose reduction. Fifth, we used time to treatment discontinuation as a proxy for clinical benefit in each line of therapy; however, our database lacked clinically-based measures of efficacy (e.g., survival, Response Evaluation Criteria In Solid Tumors [RECIST] response rates, side effects, rationale for therapy discontinuation). Further, the observation window for lenvatinib was brief (2015–2016), resulting in an insufficient number of patients to estimate treatment duration for any line of therapy in this cohort. Sixth, claims databases may not capture complete information for patients who participated in clinical trials during the course of the study. Nevertheless, our objective was to examine real-world use of SMKI and the literature suggests that less than 5% of oncology patients enroll in clinical trials [
39,
40]. Seventh, although the overall study sample was considerably larger than previous studies of SMKI treatment patterns [
20‐
23], the sample sizes in the subgroup analysis of SMKI regimen by LOT were relatively small which is associated with low statistical power. Finally, the results are based on a sample of patients with commercial or Medicare Advantage insurance and may not be generalizable to patients with other forms of insurance or the uninsured.