Key words multikinase inhibitors - advanced/metastatic thyroid cancer - network meta-analysis
Introduction
Although thyroid cancer (TC) constitutes the most common type of endocrine
malignancy, it comprises a diverse range of histopathological entities the most
common being well-differentiated thyroid cancer (WDTC). Generally, with the
exception of anaplastic cancer, TC has been considered as having an overall good
prognosis with most patients being cured by surgery although a subset with medullary
thyroid cancer (MTC) may also exhibit an aggressive course. Traditional treatment
for WDTC besides surgery includes radioactive iodine (RAI) and thyroid-stimulating
hormone (TSH) suppression therapy, whereas surgery is the main treatment for MTC
[1 ]. Nevertheless, some patients are
diagnosed with established locally advanced and/or distant stage disease or
even exhibit progress under the aforementioned standard therapies. The management
of
these patients is a multidisciplinary field with many recent innovations in TC
genetics and molecular pathogenesis with pertinent therapeutic implications
particularly in patients with MTC a subset of who may develop in the context of a
familial syndrome [2 ]. Importantly, systemic
cytotoxic chemotherapy has disappointing response rates in patients with
unresectable and/or metastatic TC and also comes at high toxicity price
[3 ]. However, recent randomized clinical
trials (RCTs) have demonstrated prime anti-tumor activity of novel, small molecule
multikinase inhibitors (MKIs) in TC, in particular RAI-refractory WDTC (RR-WDTC) and
medullary TC (MTC), resulting in the approval of certain MKIs by the Food and Drug
Administration (FDA).
Molecular alterations in the MAPK and the PI3k/Akt pathways have been
recognized in the last few years as playing a pivotal role in gene expression linked
with proliferation, cell migration and apoptosis inhibition of TC cells [2 ]
[4 ]
[5 ]. In WDTC, BRAFV600E constitutes a key
mutation and is associated with aggressive histopathological features and dismal
clinical outcomes [6 ]
[7 ]. Other key mutations are the H-, K-, and
N-RAS mutations found in follicular and poorly-differentiated TC; and PTEN
deletions, also encountered in follicular TC [2 ]
[8 ]. In the MTC counterpart, most cases occur
sporadically, whereas approximately 25% are linked with multiple endocrine
neoplasia 2A (MEN2A) or MEN2B [9 ]. In
particular, the majority of patients with sporadic form of MTC have somatic
mutations identified in the gene encoding the RET protein, whereas patients with
MEN2A or MEN2B exhibit germline RET mutations with close genotype-phenotype relation
[10 ]. In sporadic MTC without RET
mutations (35%), RAS gene mutations are commonly encountered [11 ]
[12 ]. Finally, the vascular endothelial growth
factor (VEGF) and MET pathways are associated with angiogenesis, invasion, and
promote metastasis in MTC [13 ]
[14 ].
To date, a number of MKIs have been used in advanced and/or metastatic TC,
including the MKI of VEGFRs 1–3, RET, RAF (including BRAFV600E), and PDGFR
β, sorafenib [15 ]; the MKI of RET,
VEGFR, and EGFR tyrosine kinases, vandetanib [16 ]
[17 ]; the MKI of VEGFRs 1–3, FGFR
1–4, PDGFR α, RET, and KIT signaling networks, lenvatinib [18 ]; and the MKI of MET, VEGFR, and RET,
cabozantinib [19 ]. However, translation of
these results into clinical practice faces certain challenges, as a therapeutic
reference standard for RR-WDTC and MTC in cases of advanced and/or
metastatic disease is currently lacking and these MTTs have only been compared with
placebo; hence, complicating clinical decision making in selecting one MKI agent
over another.
Additionally, the RCTs in MKIs for RR-WDTC and MTC report treatment-related
toxicities according to standard guidelines, that is, the National Cancer Institute
Common Terminology Criteria for Adverse Events, and therefore constitute a complete
resource of MKI-related toxicities. The present systematic review and network
meta-analysis of MKIs for TC provides a comprehensive summary and a comparison of
all the available randomized evidence on the antitumor activity and toxicities of
novel MKI therapies in advanced and/or metastatic TC.
Materials and Methods
The present study was designed and conducted according to the Cochrane Guidelines
for
Systematic Reviews and Meta-analyses of Interventions and their extension for
network meta-analysis [20 ]
[21 ]
[22 ]
[23 ].
Search strategy and study selection
Our aim was to identify all potentially eligible RCTs comparing systemic MKIs in
advanced and/or metastatic TC. A broad search algorithm using MeSH terms
and text words in the abstract in combination with a therapeutic intervention
and a study design filter was developed. Search strategy and the applied filters
regarding treatment selection and study design are presented in the Supplement
(Supplementary Table 1S ). The PubMed, Embase, SCOPUS, Web of Science and
the Cochrane Central Register of Controlled Trials were searched through until
March 25, 2019. No language or date restrictions were applied. The website of
ClinicalTrials.gov for potentially eligible unpublished trials was also searched
through. Key search terms included thyroid cancer, therapy, and randomized
controlled trial. We included RCTs comparing a MKI with placebo or different
doses of the same agent reporting disease control rate, objective response rate,
progression-free survival and/or side-effect incidence. Two of the
authors (MT and KD) worked in duplicate independently and screened all
potentially eligible titles and abstracts, as well as the full-text manuscripts
of all potentially relevant trials to finalize eligibility. Disagreements were
resolved by consensus between MT and KD or discussed with a third author (GK).
Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA)
guidelines for reporting were followed [21 ]
[22 ]
[23 ]. A study protocol for this
meta-analysis was not published or registered before the study was
undertaken.
Outcomes and data extraction
The primary outcomes with regards to MKIs’ antitumor activity were
disease control rate (DCR), objective response rate (ORR) and progression-free
survival (PFS). Secondary outcomes were MKIs’ safety profile and
included serious side effects (SE). Absolute values of DCR, ORR and SE; hazard
ratios (HR) with 95% Confidence Intervals (CI) for PFS rates were
extracted. Data on tumor type, study size and industry sponsorship were also
extracted. Two of the authors (MT and KD) extracted all data in duplicate and
independently. As in study selection, discordances in data extraction were
resolved by consensus.
Risk of bias and quality of evidence assessment
The risk of bias for all included studies was assessed with the Grading of
Recommendations Assessment, Development and Evaluation (GRADE) using the
Cochrane Risk of Bias Tool [24 ]
[25 ]. Scores were given for the following
standard domains: random sequence generation, allocation concealment, blinding
of participants and personnel, blinding of outcome assessment, completeness of
outcome data, for each domain. Two of the authors (MT and KD) assessed all RCTs
in duplicate and independently. Disagreements on risk of bias assessments were
resolved by consensus (between MT and KD) and/or discussed with a third
reviewer (GK).
Statistical analysis
We conducted a network meta-analysis with a frequentist approach. The
investigated end-points included DCR, ORR, PFS and serious SE (overall and
divided by organ/system) for RR-WDTC and MTC. We applied a continuity
correction for studies with a 0 cell count by adding 0.5 to all cell frequencies
[26 ]. We assessed heterogeneity by the
between study-variance τ, Cochran Q and I2 . We assessed
inconsistency by the global inconsistency test and implemented the inconsistency
model in which direct and indirect estimates were compared and a calculation of
the between-design part of Cochran Q analysis. Having fit the consistency
and inconsistency model, we produced network forest plots to evaluate the effect
size of each RCT and each treatment. However, neither consistency nor
heterogeneity for all the investigated networks could be confidently determined;
hence, consistency and lack for heterogeneity was assumed for all networks in
the study.
Interval forest plots were used with combined effect estimates (i. e., OR
and HRs with 95% CIs and size of boxes proportional to the inverse of
the SEs). We ranked therapies using the surface under the cumulative ranking
(SCURA) command in STATA to identify superiority among the investigated
treatments. The data were presented with ranking plots and clustered ranking
plots (DCR-PFS plot) of competing therapies for RR-WDTC and MTC separately, as
appropriate. We stratified the meta-analysis by subgroups of MKI serious SE
profile across different organ/systems. We used a random-effects model
to present study- specific odds ratios (OR) [27 ]. To explore heterogeneity between the studies the I2
statistics were used [28 ]. When
I2 was >0.50% the statistical heterogeneity was
considered substantial. The level of statistical significance was set at
5% (two-tailed p <0.05). We used the mvmeta application in the
STATA package (version 13.1; StataCorp, College Station, TX, USA) [29 ]
[30 ].
Results
Study selection and risk of bias assessment
We initially screened 1347 titles and abstracts from all databases and additional
548 clinical trials in clinicaltrials.gov and identified 25 potentially eligible
RCT reports (Supplementary Fig. 1S ). Finally, a total of seven RCTs
reported DCR, ORR and/or PFS and were included in the network
meta-analyses. Some of the RCTs were reported in more than one publication,
whereas the results of two RCTs were solely available from clinicaltrials.gov,
that is, not yet published. Only patients with locally advanced and/or
metastatic TC were included. In particular, four RCTs included RR-WDTC and three
included MTC. A total of 1934 unique patients were recruited; four different
MKIs were evaluated. All RCTs in the network meta-analysis were industry
sponsored. RCT characteristics are provided in Supplementary Table
2S .
Among seven included RCTs, high risk for bias in random sequence generation
(selection bias), allocation concealment (selection bias), blinding participants
and personnel (performance bias), blinding the outcome assessment (detection
bias), incomplete outcome data (attrition bias), and selective reporting
(reporting bias) was evident in none of the studies (Supplementary Table
3S ).
Antitumor activity in RR-WDTC
Four RCTs compared DCR and PFS for three different MKIs in RR-WDTC
(Supplementary Fig. 2S a ). The network meta-analysis found that all
MKI monotherapies studied were highly effective compared to placebo both in
terms of DCR and PFS analysis. The corresponding figures ([Fig. 1a, b ]) present the estimated summary
effects for all comparisons of DCR and PFS, respectively. Specifically,
sorafenib exhibited an OR with regards to DCR of 0.11 [95% CI:
0.03–0.40 and a PFS_HR of 1.99 (95% CI: 1.62–2.46)]; for
vandetanib 300 mg the corresponding pooled estimates were DCR_OR: 0.26
(95% CI: 0.06–1.24) and PFS_HR: 0.99 (95% CI:
0.82–1.20); and for lenvatinib, DCR_OR was 0.26 (95% CI:
0.05–1.33) and PFS_HR was 0.99 (95% CI: 0.81–1.22),
respectively ([Fig. 1a, b ]). A
clusterrank plot of PFS and DCR is given in [Fig. 1c ]. The quality of evidence in RR-WDTC was high for all the
included studies.
Fig. 1 a Interval plot of Disease Control Rate (DCR; Odds ratios
with 95% Confidence Intervals); b Interval plot of
Progression-free Survival (PFS; Hazard ratios with 95%
Confidence Intervals); and c respective clusterrank plot of PFS
and DCR in radioiodine refractory well-differentiated thyroid
cancer.
Antitumor activity in MTC
Three RCTs assessed ORR for two different therapies in MTC (Supplementary Fig.
S2 b ). One RCT for MTC in the network meta-analysis did not report DCR,
but only ORR; hence, ORR network analysis was undertaken. The network
meta-analysis demonstrated that cabozantinib with regards to ORR analysis was
highly effective. Cabozantinib resulted in the highest ORR [OR: 85.32
(95% CI: 5.22–1395.15)] followed by vandetanib 300 mg
[ORR_OR: 3.31 (95% CI, 0.68–16.22)] and vandetanib
150 mg [ORR_OR: 0.60 [95% CI, 0.16–2.33)] ([Fig. 2a, b ]).
Fig. 2 a Interval plot of Objective Response Rate (ORR; Odds
ratios with 95% Confidence Intervals) and b Rankogram of
estimated probabilities of each treatment being the best based on ORR in
Medullary Thyroid Cancer.
Two RCTs only assessed PFS for two different therapies (cabozantinib and
vandetanib 300 mg) in MTC. The lowest hazard for progression was found
after treatment with cabozantinib treatment [HR: 0.28 (95% CI,
0.19–0.40)], followed by vandetanib 300 mg [HR: 0.46
(95% CI, 0.31–0.69)]. Both therapies significantly reduced the
hazard for progression compared with placebo. The quality of evidence in MTC was
high for all the included studies.
Serious toxicities profile
Four RCTs compared serious SE for three different MKIs in RR-WDTC ([Fig. 3a ]). Sorafenib exhibited an OR for
serious SE of 0.30 (95% CI: 0.12–0.72); for vandetanib
300 mg SE_OR was 0.49 (95% CI: 0.15–1.68); and for
lenvatinib SE_OR was 0.96 (95% CI: 0.31–3.01), respectively
([Fig. 3a ]). For MTC, the summary
effect estimates were SE_OR: 1.20 (95% CI, 0.45–3.19) for
vandetanib 300 mg, SE_OR: 1.07 (95% CI, 0.23–4.92) for
vandetanib 150 mg, and SE_OR: 0.41 (95% CI: 0.22–0.75)
for cabozantinib, as compared to placebo ([Fig.
3b ]). We conducted a subgroup meta-analysis per organ/system
in serious SE. Our findings showed a varying MKI SE profile across both RR-WDTC
and MTC diagnoses, more commonly involving metabolic/nutritional
disorders (OR: 2.07, 95% CI: 0.82–5.18) and gastrointestinal SE
(OR: 1.63, 95% CI:1.00–2.66) ([Fig. 4 ]).
Fig. 3 a Multikinase Inhibitor Serious Toxicities risk assessment
(odds ratio with 95% Confidence Intervals) in Radioiodine
Refractory Well-Differentiated Thyroid Cancer and b Mulitikinase
Inhibitor Serious Toxicities risk assessment (odds ratio with
95% Confidence Intervals) in Medullary Thyroid Cancer.
Fig. 4 Multikinase Inhibitors Serious Toxicities profile per
organ/system in patients with Advanced and/or Metastatic Thyroid Cancer
(Odds Ratios with 95% Confidence Intervals).
Representation in international guidelines
Among the five published RCTs included in the present meta-analysis, four RCTs
were included in the latest American Thyroid Association (ATA) consensus
guidelines for WDTC and MTC and two RCTs in the European Thyroid Association
(ETA) guidelines for MTC [31 ]
[32 ]
[33 ]. However, the ETA guidelines for WDTC
are from 2008; hence, MKIs for RR-WDTC are not considered in these guidelines
and an ETA recommendation is still awaited.
Discussion
Herein, we present a systematic review and network meta-analysis of available RCTs
evaluating the antitumor activity and safety profile of MKI therapies for advanced
and/or metastatic TC. We identified seven RCTs that randomized 1934 patients
with advanced and/or metastatic TC to four different MKI therapies. Our
results suggest a range of MKI monotherapies that are superior to placebo, including
sorafenib, vandetanib (300 mg), and lenvatinib in RR-WDTC, and vandetanib
and cabozantinib in MTC. Our findings point towards a higher efficacy among the
investigated MKIs with regards to antitumor activity of lenvatinib in RR-WDTC. In
addition, MKI exhibit a broad range of risk for serious SE with regards to each
drug’s safety profile with varying treatment-related SE across different
organs/systems; hence, favoring a more patient-tailored approach with the
anticipated toxicities guiding clinicians’ therapeutic decisions. In
particular, serious SE across both RR-WDTC and MTC diagnoses, showed that there is
evidence of a profile more commonly involving metabolic/nutritional
disorders and gastrointestinal SE. Our results also highlight the need for further
research in assessing serious toxicities and effects on quality of life for
different MKI therapies.
In MTC, the ZETA trial on vandetanib (300 mg) reported an ORR as high as 45
vs. 8% in the placebo group and also a PFS benefit with a median of 30.5 vs.
19.3 months in the placebo group (HR 0.46, 95% CI 0.31–0.69, p
<0.001). On the contrary, in the EXAM trial on cabozantinib, participants
needed radiological evidence of disease progression to be eligible. The study
reported an ORR of 28 vs. 0% in the placebo group and demonstrated a PFS
prolongation of 11.2 vs. 4.0 months in the placebo group (HR: 0.28, 95% CI
0.19–0.40; p <0.001. With regards to each agent’s safety
profile, unlike vandetanib, cabozantinib was not linked with QTc prolongation.
However, even if the PFS for cabozantinib appears to be considerably shorter than
that of vandetanib, the EXAM and ZETA populations may not be directly comparable,
due to between study differences with respect to eligibility criteria, i. e.
a MTC population with progressive disease in the EXAM trial. Hence, safe conclusions
cannot be derived regarding which agent is superior and the results of our ORR
network meta-analysis in the setting of MTC have to be interpreted with caution.
Therapeutic decisions regarding MKI selection in MTC should probably be based on
patient general status and comorbidities with particular focus on the anticipated
toxicity profiles across different organ/systems.
The GRADE system was applied to assess the risk of bias of the included RCTs and the
confidence in effect estimates for all comparisons. Importantly, the end points
assessed in this network meta-analysis were DCR, ORR and/or PFS, instead of
overall survival (OS), which is the most relevant clinical end point. However to
date, no clinical trial has demonstrated an OS benefit from the use of any MKIs in
advanced and/or metastatic TC, although this is likely to be due to the high
rates of crossover in the placebo groups of the included RCTs and the data
immaturity. Additionally, MKIs have been linked with a wide range of toxicities that
have an impact on the patients’ quality of life. Determining the right time
and choice of agent to initiate the MKI treatment represents one of the most
important future tasks, as for example the radioiodine refractory feature per se is
not sufficient to determine if a WDTC patient is a good candidate for MKI therapy
and the currently available biomarkers, that is, thyroglobulin and calcitonin lack
a
predictive value with regards to treatment selection and monitoring response to
MKIs. In addition, the clinical effects of the investigated MKIs in WDTC have not
been clearly linked with mutation status, e. g. BRAF or RAS mutations for
patients treated with lenvatinib. On the other hand for the MTC counterpart,
exploratory assessment of ORR and PFS in the EXAM trial exhibited a larger treatment
effect of cabozantinib in patients with RET M918T mutation–positive tumors
[19 ]. Finally, it remains to be determined
the exact sequencing of lines of treatments upon disease progression since there are
many treatments not yet tested in RCTs, including radioiodine resensitization,
immunotherapy, novel inhibitors of specific molecular targets as tyrosine or MEK
kinases as well as checkpoint factors [34 ].
Patients who are candidates for MKIs should be thoroughly counseled on the potential
risks and benefits of this particular therapy, as these agents are associated with
many SE including fatigue, hypertension, hepatotoxicity, skin changes and numerous
gastroenterological disorders. These potential SE have a certain probability of
negatively impacting quality of life and necessitating dosage reductions or
treatment discontinuation. Nevertheless, MKIs are linked with more severe risks
including thrombosis, bleeding, heart failure, hepatotoxicity, gastrointestinal
tract fistula formation, and intestinal perforation [35 ]. In the present meta-analysis, serious SEs network analysis for
RR-WDTC exhibited a higher risk in patients treated with lenvatinib, whereas in MTC,
vandetanib at a dose of 300 mg was associated with a higher risk for serious
SE. Further analysis per organ/system demonstrated a varying MKI SE profile
for advanced and/or metastatic TC, commonly involving
metabolic/nutritional disorders and gastrointestinal SE; hence, discouraging
MKI therapy in patients with certain comorbidities, e. g. active or recent
intestinal disease, including recent GI bleeding, diverticulitis, inflammatory bowel
disease, recent bowel resection as per 2015 ATA guidelines [31 ].
This study has some limitations. Most RCTs had an unclear risk of bias due to lack
of
reporting details on random sequence generation and allocation concealment. Due to
the very low number of RCTs included in the analysis, the assessment of
inconsistency, as well as heterogeneity and publication bias was limited.
Nevertheless, the low number of RCTs may have introduced imprecision to the network
comparisons, namely wide 95% CIs may indeed include statistically
significant, yet clinically irrelevant effects. Finally, our study is subject to
biases or confounders encountered in the original RCTs; hence, the findings are
generalizable only to patient groups eligible for these trials. However, the
strengths of our study were that we applied a comprehensive search strategy with a
sensitive search algorithm, obtaining data also from unpublished RCTs targeting all
available randomized evidence. We included studies reporting DCR, ORR and/or
PFS in patients with RR-WDTC and MTC separately; thus ensuring directness.
Additionally, these well-defined TC populations and patient outcomes resulted in a
network with rather high transitivity. Nevertheless, a comprehensive analysis of
serious treatment-related SE reported in RCTs on MKIs is critical, as it may
constitute a reference for clinicians treating TC.
Our systematic review and network meta-analysis have implications for clinicians,
researchers and guideline committees. It provides a comprehensive overview of the
randomized evidence on MKI therapies for advanced and/or metastatic TC as
well as the best possible comparison of therapies that have not been directly
compared in RCTs with the aims to assist clinical decision making and guide further
research in the field. MKI antitumor activity results, and serious toxicities
profile, as presented in our network analysis, but also across different
organ/systems may aid in therapeutic decisions and also facilitate the
implementation of surveillance strategies for MKI-related toxicities. In the era of
personalized medicine, validated predictive biomarkers including histopathological
and molecular parameters with the potential to guide MKI treatment selection at the
individual level are warranted. Toxicities encountered in MKI treatment prompt
further investigation with particular focus on quality of life aspects in TC
patients to achieve a balance between antitumor activity and toxicities.
