MicroRNA Expression Profiles in the Management of Papillary Thyroid Cancer

  1. James C. Leea,b,c,d,
  2. Justin S. Gundarac,d,
  3. Anthony Gloverc,d,
  4. Jonathan Serpella,b and
  5. Stan B. Sidhuc,d,e

  1. aDepartment of Surgery, Monash University, Melbourne, Victoria, Australia; bEndocrine Surgery Unit, Alfred Hospital, Melbourne, Victoria, Australia; cKolling Institute of Medical Research, Sydney, New South Wales, Australia; dUniversity of Sydney, Sydney, New South Wales, Australia; eEndocrine Surgical Unit, Royal North Shore Hospital, Sydney, New South Wales, Australia


Papillary thyroid cancer (PTC) is the major contributor to the dramatically increasing incidence of thyroid cancer. Low-risk PTC shows the most rapid rate of increase because of changing trends in neck imaging and the use of fine needle aspiration to investigate thyroid nodules. The need for a paradigm shift in the management of these patients, to provide personalized treatment and surveillance plans, has led to the focus on molecular biomarker research. MicroRNAs (miRNAs) compose a class of molecules with promising applications for every stage of PTC management, including diagnosis, prognosis, treatment, and surveillance. Although most of the miRNA studies are currently preclinical, given the rapid progress of scientific discovery, clinical trials will not be far away. Thyroid clinicians will be expected to have good insights into the current status of PTC-related molecular translational research. This article focuses on the potential roles of miRNA in PTC management in the context of contemporary recommended clinical practice.

Implications for Practice:

The rapid increase in the diagnosis of papillary thyroid cancer has led to a demand for strategies to improve risk stratification, in order to accurately tailor management. MicroRNA biomarkers are a promising tool to aide current clinicopathological features in preoperative and postoperative risk stratification. Furthermore, circulating miRNA biomarkers also have the potential to augment thyroglobulin in long-term, noninvasive surveillance after initial treatment. Other areas of miRNA research in the management of papillary thyroid cancer include improving diagnostic accuracy of cytology samples, and targeted therapy in advanced disease.


The incidence of thyroid cancer has increased approximately threefold over the last three decades in many Western countries [14]. This increase is almost entirely attributable to an increase in the incidence of papillary thyroid cancer (PTC), which is the most common type of thyroid cancer. The American Thyroid Association (ATA) first published treatment guidelines for patients with thyroid nodule and differentiated thyroid cancer (DTC; including PTC) in 1996 and has since updated the guidelines in 2006 and 2009. The updated guidelines reflect the changing evidence from quality studies, and there have been significant advances in the diagnosis and therapy of DTC during this time.

The recommendations in the ATA guidelines were based predominantly on clinical studies [5]. There is limited application of molecular genetics in the current clinical management recommendations. In recent years, researchers have made dramatic advances in understanding the role of miRNAs in normal and malignant biological processes. A wealth of literature supports the notion that miRNAs are involved in the modulation of a myriad of cellular processes, including cell proliferation, apoptosis, invasion, and metastasis. With improved understanding, investigators are working toward harnessing the potential of miRNA in the clinical management of malignant diseases.

This review gives a brief background on miRNA and summarizes the potential of the translational application of miRNA research in the context of current recommended management of PTC.

MicroRNA Background

MicroRNA Discovery

MicroRNAs (miRNAs) belong to a class of noncoding small RNAs that are approximately 22 nucleotides in length. The founding member of this class of RNAs, lin-4, was discovered as early as 1993 in the worm Caenorhabditis elegans (C. elegans). It was noted that lin-4 did not code for a protein but instead had antisense complementarity with multiple sites in the lin-14 mRNA [6, 7]. Further studies subsequently demonstrated that the pairing of lin-4 RNA with the lin-14 mRNA resulted in translational repression of the lin-14 message. This, in turn, effected the transition from first to second larval stage [6, 7].

It was not until 2000 that the next member of this class of RNA was discovered. The let-7 RNA was also a gene in C. elegans, encoding another approximately 22-nucleotide regulatory RNA. The let-7 RNA had a role in promoting the transition from late-larval to adult stage, in a similar manner that lin-4 promoted progression from first to second larval stage [8, 9].

Owing to their roles in the developmental transitions, lin-4 and let-7 RNAs were named small temporal RNAs (stRNAs) [10]. Although 7 years passed between the discovery of the first and second stRNAs, the explosion of discovery of new stRNAs had begun. Within a year of the discovery of let-7, more than 100 additional genes of similar characteristics were reported in humans, flies, and worms. Many of these newly identified stRNAs did not appear to be involved in the temporal developmental transition process; instead, they appeared to be differentially expressed in different cell types, and many had as yet undetermined functions. It was then that the term microRNA was coined to refer to the stRNAs and all the other tiny RNAs with similar features [11]. The growing discovery of new miRNAs is nothing short of an epic phenomenon. Over the past decade, more than 2,000 miRNAs have been sequenced in humans, with more in other species.

MicroRNA Expression and Maturation

The first step in the expression of a miRNA is the transcription of the miRNA gene into the primary miRNA transcript (Pri-miRNA) by an RNA polymerase (Fig. 1). Following transcription, still within the nucleus, the Drosha-DGCR8 (DiGeorge critical region 8) microprocessor complex cleaves both strands of the stem on the Pri-miRNA [12]. The product of the cleaved Pri-miRNA is the shortened precursor miRNA (Pre-miRNA) [13].

Figure 1.

Schematic diagram depicting miRNA production. The Pri-miRNA is transcribed from a miRNA gene within the nucleus. The Pri-miRNA is then cleaved to become the Pre-miRNA, which is exported into the cytoplasm. Dicer processes the Pre-miRNA into the mature miRNA duplex. One of the strands is the guide strand, which is incorporated into the miRNA effector machine, known as the RNA-induced silencing complex. Cleavage or translational repression of the target mRNA occurs depending on the complementarity of the miRNA to the mRNA. Figure copyright Beth Croce, Bioperspective. Reproduced with permission.

Abbreviations: AGO, Argonaute protein; miRNA, microRNA; Pre-miRNA, precusor microRNA; Pri-miRNA, primary microRNA; RISC, RNA-induced silencing complex.

The Pre-miRNA produced by the nuclear processing is then exported into the cytoplasm by exportin-5 [12]. In the cytoplasm, the Pre-miRNA is cleaved into a double-stranded miRNA duplex by Dicer, an RNase enzyme [14]. One of the strands of the duplex is the guide strand, whereas the complementary strand is the passenger strand. The guide strand is incorporated into the miRNA effector machine, known as the RNA-induced silencing complex (RISC), and the passenger strand is degraded.

The RISC is the machinery responsible for gene silencing and consists of a miRNA strand, a member of the Argonaute (AGO) protein family, and other structural and facilitating proteins [12]. The guide miRNA strand recognizes the target mRNA transcript, whereas the AGO protein (with other facilitating proteins) mediates the corresponding action [12, 15]. The nature of the mediated action depends on the degree of complementarity between the miRNA and mRNA. If there is significant yet incomplete pairing of the miRNA and mRNA sequences, mRNA translation is repressed. It is this incomplete nature of miRNA-mRNA matching that allows a single miRNA to target multiple mRNAs and multiple miRNAs to target the same mRNA sequence [16]. In the less commonly occurring event of complete base pairing, the mRNA transcript is degraded [15].

MicroRNA Deregulation

MicroRNAs contribute to oncogenesis either as tumor suppressors or oncogenes. Their deregulation is a result of genomic abnormalities similar to those for protein-coding genes: chromosomal rearrangements, genomic amplifications or deletions, and mutations [17]. In any given cancer, a combination of abnormalities in protein-coding and noncoding genes can be identified [18]. Among the possible mechanisms, the contemporary view is that aberrant gene expression is the main mechanism underlying the functional changes of miRNAs in cancer cells. This is characterized by abnormal levels of expression for mature and/or precursor miRNA sequences compared with the normal tissue of that organ. The abnormal expression level is otherwise known as deregulation [17]. In addition, defects in miRNA processing and maturation also contribute to their deregulation and subsequently cause diseased states [12].

MicroRNA Versus mRNA

Several features of miRNAs make them attractive diagnostic biomarkers. They are regulators upstream from mRNAs, and each miRNA is able to target multiple protein-coding genes within or across pathways [19]. Unlike mRNAs, miRNAs do not need to be translated to proteins to exert their effects; therefore, their measured expression may correlate more closely to the functional status of the gene. Consequently, new miRNA markers can be tested for biological effects by generic sequence-based methods [19]. Furthermore, miRNAs show superior stability and maintain their expression profiles in archival formalin-fixed paraffin-embedded samples, allowing utilization of a vast amount of resources already available in most centers [1922].

MicroRNA in the Management of PTC

With the increased diagnosis of predominantly indolent PTC, biomarkers that can efficiently select the minority of patients with aggressive tumors are urgently needed. This would allow personalized treatment planning to adequately treat the minority with aggressive tumors and avoid overtreatment in the majority with indolent disease. The following discussion focuses on PTC; the potential utility of miRNA in other forms of thyroid cancer is not discussed.

MicroRNA in PTC Diagnosis

Although the clinical diagnosis of PTC is not usually a problem, it is important to appreciate the development of miRNA research in PTC, as a foundation for other clinical applications.

He et al., in their seminal paper published in 2005, were the first to apply a miRNA profiling capability to PTC [23]. One of the significant findings was that in PTC, the key differentially expressed miRNAs appeared to be overexpressed, although global under-expression of miRNA was normally associated with cancers in general. This finding has since been confirmed in other reports [24, 25].

He et al. compared the miRNA expression of 30 PTC tumors and the normal tissue samples of their corresponding thyroid glands, using the Affymetrix HG-U133 plus two array chips [23]. The pairwise comparisons indicated that 23 miRNAs showed differential expression, with 17 of them overexpressed and 6 underexpressed. Of the overexpressed miRNAs, 6 demonstrated a fold change of >1.5-fold. The most overexpressed miRNAs were miR-146b, miR-221, and miR-222, showing increases of 11- to 19-fold. These findings were further confirmed on semiquantitative reverse transcription polymerase chain reaction (RT-PCR) and Northern blotting. All underexpressed miRNAs had a fold-change of less than twofold [23]. Incidentally, in a 2006 publication, Calin and Croce pointed out that a less-than-twofold change in miRNA expression in some instances might also be biologically significant [17]. Using as few as 5 miRNAs, He et al. were able to define cancer status in 12 blind tumor samples with 100% accuracy [23].

He et al. also found that miR-221 expression was increased in nonmalignant adjacent thyroid tissue, albeit to a lesser degree [23]. This suggested that miR-221 overexpression was either an early genetic event in PTC carcinogenesis or that some carcinogenic miRNA changes were global phenomena within the thyroid gland [23].

Significant overexpression of miR-146b, miR-221, and miR-222, reported by He et al. [23], was confirmed in many subsequent studies by other groups [2527]. Using microarray and Northern blotting, for example, Pallante et al. also showed that miR-221 and miR-222 were significantly overexpressed in PTC tissue compared with normal thyroid tissue of the contralateral, unaffected lobe [26]. In their study, the third overexpressed miRNA reported was miR-181b [26]. Further confirmation of these findings was achieved by quantitative RT-PCR comparing the miRNA precursors in an external cohort of 39 PTC and 8 follicular adenoma tissue samples. Means of approximately 13-fold overexpression were seen in all 3 miRNA precursors in the PTC samples; they then extended this technique to fine needle aspiration (FNA) samples [26] (as discussed in MicroRNA in Thyroid FNA).

Between 2005 and 2013, eight publications reported on the differential miRNA expression profiles of human PTC samples compared with normal or benign thyroid tissue (Table 1); however, these publications were heterogeneous in terms of the included sample populations, tissue studied, and methodology. The heterogeneity arose from different study designs and questions, but the common ground was the use of microarray on at least 10 samples including PTC and control tissue types. Most studies included both classical and follicular variants of PTC, whereas some did not define the subtype of PTC samples included. Some studies also included other thyroid malignancies [28, 29]. The control or comparison groups varied from normal thyroid tissue [27, 30] to multinodular goiter (MNG) [25] or benign follicular lesions [28, 29]. Despite these inconsistencies, some similar trends of miRNA deregulation can be observed in PTC samples. The most consistently overexpressed miRNAs in PTC were miR-221, miR-222, and miR-146b. Other commonly reported overexpressed miRNAs were miR-155 and miR-181b. There was less consistency in the downregulated miRNA set.

Table 1.

Overview of studies using miRNA microarray to compare PTC and control

MicroRNA in Thyroid FNA

Although more histological subtypes of PTC were being profiled, work was also being done on thyroid FNA samples. Pallante et al. were the first to report three highly overexpressed miRNAs (miR-221, miR-222, and miR-181b) in thyroid tumors in the context of FNA cytology samples [26]. All three miRNAs showed higher expression levels in seven of eight PTC FNA samples compared with normal thyroid cells from FNA of non-neoplastic nodules. With this finding, new doors were opened for the utility of miRNA profiling in the diagnosis of thyroid malignancies. The potential role of miRNA profiling in thyroid pathology was transformed from postoperative confirmation or subtyping in tissue samples to preoperative diagnosis or subtyping in FNA samples. This development has major implications for the way PTC patients may be managed, if it enters clinical application.

Currently, approximately 3%–6% of FNA samples are considered indeterminate [32]. It is the management of this small subgroup that can be significantly improved with molecular diagnosis preoperatively. Keutgen et al. reported that by using a panel of 4 miRNAs (miR-222, miR-328, miR-197, miR-21), thyroid malignancies could be diagnosed in FNA samples of indeterminate lesions, with 100% sensitivity and 86% specificity and overall accuracy of 90% [33]. Other studies using a similar study design of indeterminate FNA samples, with single or a panel of miRNAs, failed to achieve such high accuracy [29, 34, 35]. Agretti et al. [35] and Shen et al. [34] were able to distinguish benign thyroid tissue from PTC using miRNA expression of FNA samples; however, the discriminatory power of the technique was poorer in the subgroup with indeterminate cytology, limiting the clinical application [35]. Using 1 miRNA on 125 indeterminate FNA samples, Vriens et al. achieved accuracy of 75% and negative predictive value of 81% [29]. These rates are not yet suitable for clinical use. The consensus is that further work is required before miRNA expression in FNA samples can be of clinical utility [36].

MicroRNA in Staging and Prognostication

One of the challenges in managing PTC patients is identifying the small subgroup with aggressive disease and providing those patients with more extensive treatment and intensive long-term surveillance. As the number patients with indolent PTC increases, there is an even greater need for accurate and efficient identification of those with aggressive disease. Accurate stratification at initial diagnosis and treatment is essential to avoid overtreatment and unnecessary surveillance of the majority and to provide adequately intensive management for the minority with aggressive disease.

Various molecular markers have been investigated in the hope of more accurate risk stratification [3740]. The most studied marker is BRAF (the B-isoform of the Raf gene). The BRAFV600E mutation has been shown to be associated with aggressive clinical behaviors of PTC by some investigators [37]; however, the evidence is still inconclusive, with contradictory reports [38]. Other molecular markers being investigated include cell-cycle regulators p27, p21, and cyclin D1 and immunohistochemical markers such as CEACAM-1 (carcinoembryonic antigen-related cell adhesion molecule 1), OPN (osteopontin), and E-Cadh (E-cadherin) [39, 40]. None of these markers have reached wide acceptance in clinical practice so far, including BRAF.

Following reports on the role that miRNAs play in PTC tumorigenesis, and thus their potential diagnostic utility, investigation also progressed to testing the prognostic ability of miRNA expression profiling. In one of the earliest studies on this subject, Gao et al. demonstrated miRNA differential expression in three subpopulations of a PTC cell line (IHH-4) with increased lymph node metastatic potency compared with the control subpopulations [41]. They found that 11 miRNAs from the array of 509 miRNAs were significantly differentially expressed among 3 pairs of subpopulations with high and low metastatic potential. Five of these miRNAs were upregulated in the metastatic subpopulations, and six were downregulated.

Three miRNAs (miR-146b, miR-221, and miR-222) consistently found to be overexpressed in PTC tissue, compared with normal thyroid tissue, appeared to also confer high-risk features such as extrathyroidal extension, lymph node metastasis, distant metastasis, recurrence, and BRAFV600E mutation. In 100 PTC samples, Chou et al. showed that tumors with the BRAFV600E mutation had higher miR-146b expression than those without the mutation [42]. In a follow-up study, Chou et al. reported that patients whose PTC expressed high levels of miR-146b had poorer overall survival. They further demonstrated that the BCPAP human papillary thyroid cancer cell line showed increased ability with regard to cell migration and invasion when transfected with miR-146b mimics [43]. Yip et al. demonstrated that miR-146b and miR-222 were overexpressed in aggressive PTC, defined as PTC associated with distant metastasis or recurrence. They also demonstrated downregulation of miR-34b and miR-130b in the PTC samples associated with aggressive biology [30]. Lee et al. reported that miR-146b and miR-222 were overexpressed in PTC with recurrence and were more strongly associated with PTC recurrence than BRAF expression. They postulated that the quantifiable nature of miRNA expression is more suited to the role of prognostication than the dichotomous nature of BRAF expression [44]. Zhou et al. found that overexpression of miR-221 was associated with extrathyroidal extension, lymph node metastasis, advanced disease stages III and IV, and the BRAF mutation [45].

Three miRNAs (miR-146b, miR-221, and miR-222) consistently found to be overexpressed in PTC tissue, compared with normal thyroid tissue, appeared to also confer high-risk features such as extrathyroidal extension, lymph node metastasis, distant metastasis, recurrence, and BRAFV600E mutation.

The findings in these studies are exciting, although further refinement through larger, externally validated studies is necessary to uncover the true potential of prognosticating with miRNA expression profiling [46]. Although few studies have reported miRNA profiling as a prognostic tool in PTC risk stratification, miRNA classifiers with promising prognostic potentials have been reported consistently across a variety of other cancers. In a 2012 systematic review of 43 studies, Nair et al. reported that several miRNA classifiers provided more accurate prognostic information, for diverse cancers, than the traditional clinicopathologically based tools [46]. They also reported consistencies in the deregulation of specific miRNAs across classifiers for different cancers, leading them to postulate that some miRNA-coordinated regulatory pathways are common to many cancers. They cautioned that errors and biases can occur at every step of the appraisal of miRNAs for prognostic purposes and concluded that the clinical application of miRNA expression measurement as a biomarker for prognosis has been limited so far.

MicroRNA in Treatment

The majority of nonaggressive PTC is adequately treated by surgery, radioactive iodine ablation, and thyroid-stimulating hormone-suppressive therapy; however, a subgroup of patients present with aggressive disease that can be locally invasive, widely metastatic, or repeatedly recurrent. Disappointingly, external-beam radiotherapy and chemotherapy have shown limited success. Recent trials of tyrosine kinase inhibitors have shown promising results and indicate the need for alternative treatment options for advanced PTC [47, 48].

The therapeutic potential of miRNAs as agents of targeted therapies is currently being explored. The ability of a single miRNA to target multiple genes is of great therapeutic advantage. Simultaneous targeting of multiple components of the same pathway, leading to synergistic effects, would confer therapeutic advantage. However, achieving target organ specificity while limiting off-target effects is a major challenge in the translation from bench to bedside [49]. Although the prospect of manipulating miRNAs to complement current therapeutic strategies in cancer is appealing, the practicality is complex.

Despite the challenges, in April 2013, for the first time, a miRNA mimic reached phase I clinical study. MRX34 aimed to restore the lost tumor suppressor function of endogenous miR-34 in patients with primary liver cancer or metastatic tumors to the liver. By restoring the tumor suppressor pathway via miR-34 replacement therapy, apoptosis was induced in tumor cells in vitro and in mouse models; however, off-target effects, dosing, and pharmacokinetics were all areas that required further study [50]. Currently, there is limited research on the miRNA-targeted treatment of PTC. In a recent early stage in vitro study, Lin et al. showed that restoration of miR-101 expression in the K1 PTC cell line significantly reduced proliferation [51].

MicroRNA in Follow-Up

In recent years, there has been intense interest in the feasibility of using miRNAs or miRNA panels as circulating biomarkers for the presence of malignant and nonmalignant diseases. Cellular miRNAs can be released into the circulation or surrounding microenvironment membrane free, protein bound, or packaged within microvesicles (0.1–1 μm) or nanovesicles (<100 nm). Free miRNAs are rapidly degraded by RNases, whereas vesicular or protein-bound miRNAs are protected from degradation. These circulating miRNAs released by diseased cells, however they escape degradation, are currently being investigated as potential noninvasive biomarkers of disease for diagnosis or recurrence. Some of these circulating miRNA biomarkers also double as prognostic factors for survival, staging tools, extracellular communicators, and markers of pathological progression in various cancer types.

In PTC, circulating miRNAs are potential alternatives to serum thyroglobulin (Tg) measurements. Serum Tg measurement as a long-term noninvasive surveillance tool is not applicable to up to 25% of PTC patients because of the presence of Tg antibodies, the performance of less than total thyroidectomy, or the lack of postoperative radioactive iodine ablation. Preliminary studies have demonstrated that circulating levels of let-7e, miR-151-5p, miR-146b, miR-221, and miR-222 are higher in PTC patients compared with healthy subjects [44, 52]. Yu et al. measured miRNA expression in the serum of 106 patients with PTC, 95 patients with benign thyroid nodules, and 44 healthy subjects. They found that serum levels of let-7e, miR-151-5p, and miR-222 were significantly overexpressed in PTC patients compared with patients with benign nodules and healthy subjects. No significant difference was found in the serum levels of these miRNAs between patients with benign nodules and healthy subjects [52].

Preliminary studies have demonstrated that circulating levels of let-7e, miR-151-5p, miR-146b, miR-221, and miR-222 are higher in PTC patients compared with healthy subjects.

In a study comparing plasma miRNA expression before and after total thyroidectomy, Lee et al. reported that plasma levels of miR-146b, miR-221, and miR-222 were overexpressed in 42 PTC patients compared with healthy individuals [44]. They further reported that the plasma levels of all three miRNAs in PTC patients fell to levels comparable to those of healthy subjects. However, in addition to the healthy subjects, Lee et al. also included a third group of patients with multinodular goiter. They found that both preoperative and postoperative plasma miR-146b, miR-221, and miR-222 levels in MNG patients were comparable to those of PTC patients. Consequently, they concluded that plasma miRNA expression is not suitable for de novo diagnosis of PTC but may be a useful adjunct to Tg in long-term surveillance [44]. A noninvasive, miRNA alternative to thyroglobulin would simplify the long-term recurrence surveillance for the subgroup of patients who are unable to make use of serum Tg levels.


There has been significant improvement and standardization in the clinical management of PTC in recent years; however, molecular markers to further improve risk stratification are still anticipated for clinicians to be able to provide biomarker-driven, personalized treatment. Not only are microRNAs potential biomarkers for PTC recurrence and metastasis (miR-146b and miR-222), they also have been shown to be potential tools for long-term surveillance (let-7e, miR-151-5p, miR-146b, miR-221, and miR-222). Work is in progress to ascertain prognostic information by profiling miRNA expression in thyroid FNA samples and by using miRNAs as targeted therapy for treatment-resistant PTC tumors.

Author Contributions

Conception/Design: James C. Lee, Jonathan Serpell, Stan B. Sidhu

Collection and/or assembly of data: Justin S. Gundara, Anthony Glover

Data analysis and interpretation: Justin S. Gundara, Anthony Glover

Manuscript writing: James C. Lee, Justin S. Gundara, Anthony Glover

Final approval of manuscript: James C. Lee, Jonathan Serpell, Stan B. Sidhu


The authors indicated no financial relationships.


Critical Appraisal of Translational Research Models for Suitability in Performance Assessment of Cancer Centers

  1. Abinaya Rajana,b,
  2. Richard Sullivanc,
  3. Suzanne Bakkera and
  4. Wim H. van Hartena,b

  1. aDepartment of Psychosocial Research and Epidemiology, The Netherlands Cancer Institute-Antoni van Leeuwenhoek Hospital, Amsterdam;

  2. bDepartment of Health Technology and Services Research, University of Twente, The Netherlands;

  3. cInstitute of Cancer Policy, Kings Health Partners Integrated Cancer Center, King's College London, United Kingdom


Background. Translational research is a complex cumulative process that takes time. However, the operating environment for cancer centers engaged in translational research is now financially insecure. Centers are challenged to improve results and reduce time from discovery to practice innovations. Performance assessment can identify improvement areas that will help reduce translational delays. Currently, no standard method exists to identify models for use in performance assessment. This study aimed to critically appraise translational research models for suitability in performance assessment of cancer centers.

Methods. We conducted a systematic review to identify models and developed a set of criteria based on scientometrics, complex adaptive systems, research and development processes, and strategic evaluation. Models were assessed for linkage between research and care components, new knowledge, systems integration, performance assessment, and review of other models.

Results. Twelve models were identified; six described phases/components for translational research in different blocks (T models) and six described the process of translational research (process models). Both models view translational research as an accumulation of new knowledge. However, process models more clearly address systems integration, link research and care components, and were developed for evaluating and improving the performance of translational research. T models are more likely to review other models.

Conclusion. Process models seem to be more suitable for performance assessment of cancer centers than T models. The most suitable process models (the Process Marker Model and Lean and Six Sigma applications) must be thoroughly tested in practice.


  • C/A:
    Consulting/advisory relationship
    Research funding
    Honoraria received
    Ownership interests
    Intellectual property rights/inventor/patent holder
    scientific advisory board


Translational research is a complex, cumulative, and often unpredictable process focused on moving a single or combination of basic research findings into clinical practice. The recent identification of and attention to this field is not just meant to raise awareness, but also to improve performance in terms of efficiency and effectiveness. A particular challenge to translational research in oncology, as in other clinical fields, are perceptions about unnecessary delays in or complete blockage of translation.

In the fiscal year 2004–2005, global spending on cancer research reached approximately €14 billion ($17.64 billion). The U.S. (dominated by the National Cancer Institute) accounted for most of the spending, with per capita spending almost three times greater than Europe. However, in terms of publications and an increasing trend towards more applied clinical outputs, relative research productivity was better in Europe [1]. Apart from effectiveness issues, translation of research into practice still takes a lot of time. There are claims that translation of only 14% of new health-related scientific discoveries to clinical practice [2] takes an average of 17 years [3]. Ioannidis et al. examined 101 promising claims of new discoveries with clear clinical potential that were reported in major basic science journals between 1979 and 1983; only five resulted in interventions with licensed clinical use by 2003 and only one had extensive clinical use [4].

Imatinib is an example of successful translation from oncology. It shows the time it took for an intervention to reach licensed clinical use based on knowledge that emerged slowly over many decades. The drug focuses on disrupting one specific protein that seems to fuel the cancer while sparing other enzymes. The initial knowledge appeared in the 1960s when scientists first noticed chromosomal abnormalities in the blood of patients with chronic myeloid leukemia. However, it was not until the 1980s that genetic mapping helped determine that chromosomal abnormality produces a cancer-causing kinase enzyme. It took 2 years to create and test 400 molecules to find one that would target this enzyme without disrupting any of the hundreds of other similar enzymes in a healthy cell. Another 8 years of safety testing and development was needed before the drug could be tested with patients, finally giving remarkable results. While clinical trials were being expanded, the U.S. Food and Drug Administration put the drug on fast track for approval in 2001 [5].

Apart from effectiveness issues, translation of research into practice still takes a lot of time. There are claims that translation of only 14% of new health-related scientific discoveries to clinical practice takes an average of 17 years. Ioannidis et al. examined 101 promising claims of new discoveries with clear clinical potential that were reported in major basic science journals between 1979 and 1983; only five resulted in interventions with licensed clinical use by 2003 and only one had extensive clinical use.

Translational research is cumulative. To improve its performance and reduce unnecessary delays, acquiring insight into the process and performance assessment can add value. This means assessing performance in cancer centers against a set of predetermined criteria of the economy, efficiency, and effectiveness of that organization in conducting translational research (adapted from the Organisation for Economic Cooperation and Development definition) [6] with the purpose of supporting continuous improvement and transparent accountability at multiple organizational levels. This would help address delays by identifying areas for improvement, including innovation transfer management, organizational administration of research projects, incentive mechanisms to motivate researchers, and communication strategies between researchers and other key stakeholder groups. These areas can promote multidisciplinary collaboration that in turn can speed the rate at which basic research discoveries eventually become clinically viable health technologies.

For performance assessment, it is essential to know what is being translated and how it is being translated. Initially, models need to be systematically identified and critically appraised before they can be tested in practice. To a large extent, the process of translational research seems to be generic, and it is not clear if a specific model should be preferred for oncology. At present, it is unknown how many models exist and which of those are suitable for performance assessment. Most recent references are based on two studies. Trochim et al. reviewed and synthesized four models to illuminate important issues to evaluate translational research [7]. Morris et al. looked at quantification of translational time lags; in that context, they offered a tentative model based on synthesis of a few models [8]. However, the studies do not specify if they conducted a systematic identification of models, nor did they use systematic criteria to appraise the identified models. Moreover, in the study by Morris et al., it is not clear how many models were used to synthesize their model.

The current study aims to identify models of translational research using a systematic literature review and critically appraise them by using common criteria that were specifically developed for this purpose. The rationale is to identify the models that are most suitable for assessing the performance of cancer centers in translational research.


A systematic literature review was carried out to identify translational research models using a combination of search terms in four databases: PubMed, Embase, Trip Database, and Scopus (supplemental online data). The first search included scientific terms and common expressions for translational research and terms associated with models and performance assessment, whereas the second search included scientific terms and common expressions for translational research and different phases of translation (Fig. 1). In addition, we tracked the references and citations for a few papers that were identified through the previous search method, which either proposed a model and/or identified other models. We did not limit our search to models specific to oncology nor to the year of their publication.

Figure 1.

Search terms used to identify models of translational research.

Criteria Development to Appraise Models

At present, there is no standard methodology to assess the suitability of translational research models for performance assessment purposes. We developed a set of criteria (CR; Table 1). The models were awarded a yes or no answer for each question, in which yes meant that the model seemed suitable for performance assessment. Model appraisal focused on how translational research was presented in terms of its main purpose, component(s) that can be evaluated, strategies to evaluate the identified components, and testing of the chosen strategies in practical settings. To validate our focus, we referred to a range of literature from both medical and nonmedical disciplines, such as organizational management.

Table 1.

Critical appraisal of translational research models

With reference to the scientometric analysis conducted by Jones et al., we deduced that translational research emerged to link the research and care components (CR1) [29]. Cancer research is a complex adaptive system in which the components must be regularly assessed to improve their performance (CR2 and CR3) [30]. Fifth-generation research and development suggests that performance assessment strategies should integrate organizational systems to link the process of translation that occurs through cross-boundary learning and knowledge flow (CR4) [31]. Using the theory of the evaluation of strategic options by Johnson and Scholes, we framed criteria for evaluating the strategies of the models for suitability and feasibility (CR5 and CR6) [32]. A seventh criterion based on acceptability (CR7) was meant to check if models have been tested or applied in practice. This last criterion is not been presented in Table 1 because we were able to assess only one model.


Identified Translational Research Models

A total of 2,397 studies were identified after removing the duplicates (Fig. 2). Title screening showed that the majority of studies were related to specific biomedical discoveries focusing on basic and translational issues. Many studies referred to animal models, not conceptual models. Only 385 papers contained a description of translational research. Abstract screening led to 182 papers that contained bench-to-bedside issues; 89 studies used descriptive statements to define translational research. Only 12 studies that contained and described a model were included in the resulting appraisal. Of these, 6 studies described the main phases/components for translational research within different translational blocks (T models) [2, 9, 1214, 19]. The remaining 6 papers mapped the steps/processes for translational research (process models) [7, 2226]. Both type of models start at basic discovery; the following phases extend to clinical trials or even beyond to widespread diffusion or population impact (Fig. 3).

Figure 2.

A systematic review to identify models of translational research.

Figure 3.

An overview of translational models and process models of translational research.

Overview of T Models

The terminologies and position of the types of translations are inconsistent in all T models. Overall, the T blocks identify the specific translational areas that are also barriers for translation, but steps to overcome these barriers and improve performance are not clearly addressed.

Type 1 Translation

In the six models, descriptions of type 1 translation (T1) have similar starting points but are phrased differently. T1 encompassed “basic research to patient based research” [9], “basic science research to human clinical research” [2], “basic science research (phase 0) to early human trials (phase 1) and early clinical trials (phase 2)” [14], “basic biomedical science to clinical efficacy knowledge” [13], “basic biomedical to clinical science knowledge” [12], and “gene discovery to health applications” [19]. Because of these variations, it is hard to establish where T1 ends.

Type 2 Translation

The description of type 2 translation (T2) is also inconsistent over all models. T2 encompassed “patient oriented to population oriented research” [9], “human clinical research to practice based research” [2], “early clinical trials (phase 2) to late clinical trials (phase 3)” [14], “clinical efficacy knowledge to clinical effectiveness knowledge” [13], “clinical science knowledge to improved health” [12], and “health applications to evidence-based guidelines” [19].

Type 3 Translation

The location and extent of type 3 translation (T3) also varies in all models. T3 encompassed “population-based research to basic research” [9], “practice-based research to clinical practice” [2], “late clinical trials (phase 3) to implementation phase (phase 4)” [14], and “clinical effectiveness knowledge to improved population health” [13]. In Sung et al.'s model, there was no T3 [12]. In Khoury et al.'s model, T3 was the translation of guidelines to health practice [19].

Type 4 Translation

Only Khoury et al.'s model contained type 4 translation (T4), which was the translation of practice to population health impact [19].

Overview of Process Models

Three process models used T terminologies. The early translational pathways by Ernest et al. [23] used the T1-T2 model, but the pathways were mapped only for T1. They were developed to aid the transformation of scientific discoveries into new clinical modalities for oncology—specifically risk assessment modalities (biospecimen-based risk assessment devices and image-based risk assessment) and interventive modalities (agents, immune response modifiers, interventive devices, lifestyle alterations).

The biomedical research continuum by Drolet and Lorenzi [22] consisted of a zone of translation with three translational chasms (T1-T3): T1 was laboratory to clinical research between basic science discovery to proposed human application; T2 was safety and efficacy research between proposed human application and proven clinical application; T3 was implementation and adoption research between proven clinical applications and clinical practice. A pathway, inquiry, and action for each chasm were given.

The Lean and Six Sigma applications to clinical and translational research by Schweikhart and Dembe [24] used the T1-T4 phases by Khoury et al. [19] to improve the efficiency of translational research. Each phase consisted of business management strategies for process assessment. The Process Marker Model by Trochim et al. identified key steps of translational research, which were not represented by Ts but described as three integrated systems: basic research system, clinical trials system, and practice-based system. The model aims to evaluate the process of translational research in order to reduce the time lag [7].

The Need to Knowledge model by Lane and Flagg [25] identified unmet needs that lead to the generation of knowledge through the outputs of three activities: research discovery, prototype intervention, and product innovation. It recognized that knowledge implementation and beneficial societal impacts involve effective communication of each successive knowledge state to the relevant stakeholders. Finally, Ogilvie et al.'s model is a framework to advance translational research that identifies a pivotal role for evidence synthesis that translates knowledge of nonlinear and intersectoral interfaces to the public realm [26].

Oncology-Specific Models

It was difficult to confirm which of the appraised models are currently being used to inform translational research in cancer centers in Europe and/or the U.S. However, only one model was specifically developed for oncology: the early-stage translational pathways by Ernest et al. [23]. They used the T1-T2 model proposed by the President's Cancer Panel [18]. This was one of the first models in translational research to emerge and is also known as bench-to-bedside-to-practice. The pathways were developed in T1 phase to facilitate the process of basic discoveries in cancer to be developed into clinical modalities, but they have not been adopted in practice.

Evidence From Appraisal of T Models and Process Models

The process models were more favorable when appraised against our criteria than T models (Table 1), suggesting that they may be better suited for performance assessment in cancer centers. There is only one similarity between the two types of models: they view translational research as accumulation of new knowledge. The differences are that process models more clearly address systems integration, link research and care components, and were developed for evaluating and improving the performance of translational research. In contrast, T models tend to review other models; their purpose is to present the phases of translational research but not to assess and improve its performance.

Three process models (Lean and Six Sigma applications, the Process Maker Model, and the Need to Knowledge Model) seem to have been developed to evaluate translational research. In particular, the first two models scored highest in the appraisal (Figs. 4 and 5). They track the time between various steps of the different translational phases in order to improve translational process efficiency. Lean and Six Sigma is the only model that clearly gave evidence that it had been tested in practice in a process improvement project focused on redesign of the scheduling system at the clinical trials unit of Ohio State University [24].

Figure 4.

Examples of process marker models at three levels of scale. Reprinted from [7] with permission from Wiley.

Abbreviations: FDA, U.S. Food and Drug Administration; IRB, institutional review board.

Figure 5.

Example of a process improvement project at a clinical trial unit using Lean techniques. Reprinted from [24] with permission from Wolters Kluwer Health.

Possible Implementation of Lean and Six Sigma Techniques in Performance Assessment of Translational Research

A research process improvement project involving redesign of the scheduling system in the clinical trials unit of the Ohio State University (Fig. 5) used a five-stage intervention. The aim was to improve the efficiency of the patient scheduling process by replacing paper-based calendar system with a more coherent data-driven computerized scheduling system. It is a practical example of the applicability of Lean and Six Sigma techniques in assessing and improving the performance of translational research.

In stage 1, an environmental scan was undertaken by a research team to determine stakeholder needs, as well as to sufficiently identify and understand various steps that are involved in the patient accrual and scheduling process, including protocol requirements, the total number of trials being conducted, software requirements, inpatient bed capacity, number of available nurses and other staff per shift, examination and treatment room availability, number of expected visits and specific visit number in the sequence of protocol. The improvement strategy was to develop acuity measures to gauge resource intensity in each step.

Next, in stage 2, the team identified and mapped each process step and relationship between those steps using value stream maps or process flow maps. As an improvement strategy, they developed different scheduling algorithms based on acuity measures and other factors. In stage 3, the team identified obstacles for and inefficiencies between patient scheduling and planning of the available resources. The improvement strategy led to the development of standardized scheduling instructions for physicians and patients to improve resource utilization.

In stage 4, the team performed repeated field testing of various scheduling algorithms. As an improvement strategy, an acuity table with estimates for each activity was calculated. For example, the activity of “simple specimen collection” was given an acuity score of 5. A scheduling algorithm matched the scores with key internal and external factors (e.g., availability of a specific number of research staff per shift, room availability, protocol-related requirements) to optimize patient and staff scheduling on a given day.

Finally, in stage 5, an assessment of organizational structure and culture was done in the research unit to evaluate readiness for change. The improvement strategy led to cross-disciplinary training of research staff to make them understand and use the new patient scheduling system. The concerns and suggestions by staff regarding the practical use of the system were addressed during the training. The above stages led to the adoption of the system in daily practice [24].

Drawing on this example in more generic terms, a five-stage intervention for applying performance assessment models in translational research in cancer centers should address the following:

  1. Environmental scanning to understand key activities in translational research

  2. Elaborating different algorithms in which the identified key activities will be efficiently performed

  3. Evaluation of these algorithms by performing continuous improvement cycles to check which algorithm is most suitable

  4. Using estimates (metrics such as frequency/duration) to map the key activities identified and correlating that to key internal and external factors that may affect those estimates

  5. Training of research staff on the new system and ensuring that its implementation within the cancer center is acceptable to key stakeholders

Based on these stages, qualitative and quantitative indicators can be derived.


This study aimed to identify models of translational research and appraise their suitability for performance assessment of cancer centers. We managed to identify 12 models of translational research: six T models and six process models.

T models contribute to our understanding of translational research by mapping its key components. However, these components vary from model to model, confirming the statement of Australia's chief scientist, Professor Ian Chubb: “If you were to ask ten people what translational research means, you're likely to get ten different answers” [33]. It is not clear whether the variations in T models reflect actual variations in practice or are related to specific objectives or circumstances of various stakeholders. These variations may also reflect models being developed for specific research and/or clinical domains. In contrast, process models identify methods to facilitate, track, and assess knowledge flows and interfaces along the continuum, including multiple starting points for innovation, pathway mapping, process markers, using strategies and tools from business management, and inclusive evidence synthesis.

Based on our appraisal, two process models seem to be most suitable for performance assessment of cancer centers: the Process Marker Model and Lean and Six Sigma applications. Process markers can help cancer centers assess the performance of translational research by tracking the time taken between markers, such as prepiloting of studies, submission of research proposals, funding of studies, the start and end of data collection for studies, and inclusion of the study in research synthesis (e.g., publications or mainstreaming of research activities) that leads to subsequent stages of translational research. Process markers can include both process steps as well as reflect the transfer process per step (known as subprocess markers). Process markers can be defined for phases in clinical trials, proposal submission, Institutional Review Board approval, funding of proposal, accrual of first subject, closed to accrual, and presentation and publishing of results etc [7]. Process markers might help to identify and possibly reduce the time between different phases of clinical trials in cancer.

Lean and Six Sigma applications are complementary to the Process Marker Model and might help cancer centers define markers more clearly. For example, in basic research, process makers could include turnaround time of toxicology results, transfer of samples in laboratory, and response to regulatory requests. In clinical trials, cancer centers could track the unnecessary time and/or added value per process step for biostatistical consultations, minimizing protocol amendments, checking if placebos are needed, patient recruitment campaigns, patient monitoring process, and elimination of early-phase design errors [24].

The fear regarding [performance] assessments among some stakeholder groups is that it might jeopardize serendipity that is characteristic for many research processes, fail to capture research excellence that might exist partially or completely outside the scope of assessment criteria, and enable bureaucrats to take control of fields they do not really comprehend. A cautious and stepwise approach is therefore advisable if cancer centers intend to use these models for performance assessment.

However, the models still have some limitations. Lean and Six Sigma applications are derived from a nonmedical field. Although their pilot results are positive, they need to be tested in other phases of translational research to fully validate their use along the continuum. The Process Marker Model lacks precisely stated operational definitions of markers and an inferential statistical analysis framework [7]. In addition, although markers primarily measure time lags, qualitative value related criteria are still lacking.

The five-stage intervention for the possible implementation of Lean and Six Sigma techniques can be adapted to different phases of the translational research continuum. It can aid performance improvement from basic science along the continuum to population impact. However, defining activities or markers for the earlier phases is relatively easier than for later phases, such as population impact. These later phases tend to be beyond the primary scope of some comprehensive cancer centers. Hence, inclusive evidence synthesis is needed to understand the later phases from a broader public health perspective [26] before performance assessment models can be implemented.

Translational research is not a simple linear process. Some may argue that its complex and unpredictable nature prohibits the use of models for performance assessment. The fear regarding such assessments among some stakeholder groups is that it might jeopardize serendipity that is characteristic for many research processes, fail to capture research excellence that might exist partially or completely outside the scope of assessment criteria, and enable bureaucrats to take control of fields they do not really comprehend. A cautious and stepwise approach is therefore advisable if cancer centers intend to use these models for performance assessment. As a first step, acquiring structured insight into the various aspects of the translational process and comparing these results between cancer centers might help centers identify improvement opportunities. For that purpose, more precise operating definitions are needed at three levels: performance dimensions, performance indicators, and sufficiently detailed metrics [34]. It is hard to say whether the cancer field has specific needs, but all stakeholders, including clinicians, should be open to the idea that models from other medical and/or nonmedical fields can also be used to assess cancer centers. These models should be thoroughly tested in practice to know their potential for actual performance assessment.

The strengths of this study are that, to our knowledge, this is the first time that a systematic review has been undertaken to identify models of translational research that were appraised using a set of criteria. These criteria were based on a range of issues for translational research identified from relevant literature. Undoubtedly, the criteria that we used can be critiqued. However, it is necessary for cancer centers to carefully select models for performance assessment and our framework provides a basis for that. The criteria can be refined with views from key stakeholder groups (e.g., basic researchers, clinical researchers, clinicians, funding agencies, senior executives, and patients).

There are two possible limitations to our study. First, we could not check if all the models had been tested and implemented in practice. One could argue that the elements of these models are supported by “findings” or evidence from academic or experiential literature. The second limitation is that, because of a lack of consensus on terminologies in translational research, it was hard to identify models. Therefore, there could be models that we did not consider in this appraisal. To increase the possibility of identifying models in future, the title, abstract, and keywords of studies should clearly use a common term and/or commonly associated terms of translational research. Substitutions such as “bench-to-bedside,” “implementation science,” and “biomedical research” should be restricted to the main content of the papers, with clear explanation of these terms that can help the reader understand the model. Addition of a specific MeSH term for models in databases (e.g., conceptual models of translational research) may be useful to ensure that models are easily listed and identified.


Performance assessment can help improve the process of translational research by identifying areas for improvement in its management, knowledge exchange, and engagement of multidisciplinary teams to deliver efficient and effective translational research, which would help reduce unnecessary time lag. Two models of translational research appear to be more suitable for performance assessment: the Process Marker model and the Lean and Six Sigma applications to clinical and translational science. It will be necessary to thoroughly test them in practice. Finally, cancer centers need to come to a consensus on terminologies in translational research, which will help to identify and select models for performance assessment that can improve the performance of translational research for the benefit of patients.

Author Contributions

Conception/Design: Abinaya Rajan, Richard Sullivan, Wim van Harten

Collection and/or assembly of data: Abinaya Rajan, Suzanne Bakker

Data analysis and interpretation: Abinaya Rajan, Wim van Harten

Manuscript writing: Abinaya Rajan, Wim van Harten

Final approval of manuscript: Abinaya Rajan, Richard Sullivan, Suzanne Bakker, Wim van Harten


The study was supported by the FP7 program of the European Commission as part of a Eurocan platform project (An Excellence Designation System for Comprehensive Cancer Centers in Europe). The project consists of 28 cancer institutions across Europe that currently collaborate on different work packages to help advance translational cancer research.


  • C/A:
    Consulting/advisory relationship
    Research funding
    Honoraria received
    Ownership interests
    Intellectual property rights/inventor/patent holder
    scientific advisory board


Copyright © 2014 - Material published on this website is protected by copyright. You are expressly prohibited from electronically storing any material found here or republishing it by any means.