Current Oncology Reports

, 16:403

Current Status of Molecular Biomarkers in Endometrial Cancer

Authors

    • Department of Clinical ScienceUniversity of Bergen
    • Womens’ ClinicHaukeland University Hospital
  • H. B. Salvesen
    • Department of Clinical ScienceUniversity of Bergen
    • Womens’ ClinicHaukeland University Hospital
Gynecologic Cancers (NS Reed, Section Editor)

DOI: 10.1007/s11912-014-0403-3

Cite this article as:
Werner, H.M.J. & Salvesen, H.B. Curr Oncol Rep (2014) 16: 403. doi:10.1007/s11912-014-0403-3
Part of the following topical collections:
  1. Topical Collection on Gynecologic Cancers

Abstract

In spite of the high and increasing incidence of endometrial cancer, our current models for prediction of prognosis and even more treatment response are suboptimal, and molecular biomarkers to assist clinical decision making are needed. In this review an overview is given of the currently known as well as promising prognostic and predictive biomarkers in endometrial carcinoma. Key clinical challenges, where use of molecular biomarkers can meet clinical needs, are highlighted. The current status for the presently most promising prognostic and predictive biomarkers in endometrial carcinoma is reviewed. DNA ploidy, p53 status, hormone receptor level, HER2, stathmin, L1 cell adhesion molecule expression and other biomarkers are discussed in relation to the scientific robustness of various essential steps in biomarker development and (current) clinical applicability for individualizing treatment strategies. Tumour heterogeneity and its consequences for biomarker assessment and the importance of developing standardised tests for implementation are discussed. To improve the development and clinical uptake of biomarkers, several strategies are proposed.

Keywords

Endometrial cancerMolecular biomarkersPrognostic biomarkersPredictive biomarkersBiomarker validationClinical utilityGynaecologic cancer

Introduction

Endometrial cancer is the commonest pelvic gynaecological cancer in the westernised parts of the world [1]. In view of increasing life expectancy and BMI, the incidence will more than likely increase further [24]. With localised disease, most patients face good prognosis, but after systemic relapse, survival is abysmal and averages 7–12 months, in spite of treatment [5, 6]. Our abilities to predict which patients are at increased risk of facing relapse of endometrial cancer or, for example, which patients will respond optimally to treatment are suboptimal [7, 8]. Biomarkers, defined as characteristics that are objectively measured and valuated as an indicator of normal biological processes, pathogenic processes, or pharmacological responses to a therapeutic intervention [9], can help us to disentangle the complexity in each individual case and support our decision making capabilities. So far, the widely applied biomarkers in endometrial cancer are not molecularly based; International Federation of Gynecology and Obstetrics (FIGO) surgical stage, tumour subtype and histological grade are used in the clinic to advise on prognosis and the need for adjuvant treatment, but have uncertain survival benefit. Up to 20 % of the presumed low-risk tumours, on the basis of these biomarkers, will eventually recur, and roughly 50 % of the patients with presumed high-risk tumours will be fully cured of their disease [7, 8]. Therefore, not only do we need biomarkers to better identify those patients who are currently left undertreated, but also to avoid overtreatment of patients who would never have had a recurrence, even more so in light of the endometrial cancer population, which is often elderly and co-morbid [4, 10]. Further, biomarkers may guide systemic treatment choice or risk of toxicity.

Importantly, for any biomarker to be of clinical interest, it should provide incremental information beyond what is already known from applying state-of-the-art clinicopathological variables and other known predictors, and as such should be tested in addition rather than independently of these.

Basic studies, especially through the advent of multiple (high-throughput) omics platforms, have given us the opportunity to study the molecular make-up of the tumour at many different ‘levels’, using different cellular molecules, such as DNA, RNA, microRNA and protein, also including epigenetic and metabolomic changes to identify tumour aberrations on a large scale. There has been enormous progress in the discovery of such molecular biomarkers in recent years. Many are studied in tumour tissue, but increasingly blood, urine or ‘in situ’ measurements, including functional imaging, are performed and may have the advantage of being more accessible, in particular, for sequential measurements during therapy. Some of these aberrations may be developed further as molecular biomarkers and may tell us something about the carcinogenic process itself or may be directly linked to, for example, the prognosis or therapy resistance. Both individual markers and so-called signatures or panels of markers in combination can be relevant.

Although with respect to the development and clinical implementation of molecular biomarkers some other cancer types, such as breast cancer and lung cancer, are clearly ahead of endometrial cancer, there are many exciting developments in endometrial cancer. Also, much can be learnt, although not directly applied, from the experience in other cancer types.

This review provides an overview of the status of molecular biomarkers in endometrial cancer as well as promising approaches to accelerate biomarker development.

Types of Biomarkers

Different biomarkers serve different aims. A frequently used classification divides them into prognostic biomarkers, related to the survival of the patient independent of treatment and often implicated in the oncogenic processes; predictive biomarkers, related to the expected response of a patient to a treatment; and pharmacodynamical biomarkers, related to systemic treatment titration. They are vital in this era of personalised medicine, to help tailor treatment to the individual patient’s needs.

Prognostic Biomarkers

As indicated, for prognostic estimation, no molecular biomarkers have proven stronger than the currently used clinicopathological variables. However, molecular biomarkers have been shown to be of clear additional value in the diagnostic process, for example considering ambiguous tumours, where the identification of certain ‘subtyping’ biomarkers supports their correct classification.

An area of intense investigation is related to preoperative identification of risk factors, such as lymph node metastases, to optimise the surgical stratification. In endometrial cancer, 6–8 % of patients will have lymph node metastases at diagnosis, 4–6 % will have pelvic lymph node metastases (with or without paraaortic) and only 2 % will have para-aortic lymph node metastases [11, 12]. No survival benefit has been shown to be associated with the lymphadenectomy procedure itself in clinical trials [1315]. The procedure is also not well defined, and significant morbidity is known to result from it [12, 16]. Lymphadenectomy, however, is part of a complete surgical staging procedure in endometrial cancer, and as such helps to stratify correctly patients for adjuvant treatments [17]. To counter the trend of more aggressive surgery without clear survival benefit, different strategies are being developed, including sentinel node procedures [1820] and preoperative identification of high-risk patients focusing on lymph node metastasis, cervical stromal infiltration or deep myometrial involvement using molecular biomarkers from preoperative specimens [2124], blood samples [2528] or imaging such as ultrasonography, MRI and PET/CT [11, 2932] for their utility in selecting those patients in which the potential benefits of the procedure outweigh the additional morbidity associated with, in particular, complete pelvic and aortic lymphadenectomy.

Further, in endometrial cancer currently no assays are available to be included in the clinical algorithms that can facilitate monitoring of disease, such as cancer antigen 125 in ovarian cancer. Clinical symptoms and imaging results are currently the main pillars to detect endometrial carcinoma progression or recurrence. Reliable markers in urine or blood would facilitate disease monitoring at likely significantly reduced costs related to imaging and with less burden and fewer potential adverse events for the patient. Molecular biomarkers such as growth differentiation factor 15, cancer antigen 125 and human epididymis protein 4 are being studied at present for this goal, but are not routinely applied clinically [2528].

The best validated prognostic biomarkers in endometrial cancer to date, with some, although not wide, application in the routine clinical setting, are oestrogen receptor (ER) and progesterone receptor (PR) [24, 33, 34], p53 [3537] and DNA ploidy [3841], and which are to a degree histological subtype specific. OR and PR positivity is common in endometrioid tumours, whereas overexpression of p53, aneuploidy and the widespread presence of chromosomal gains and losses much more characterise non-endometrioid tumours. Thus, possible prognostic effects should be analysed in a subtype-specific manner. Loss of OR and PR has been clearly linked to aggressive tumours, also specifically within the endometrioid subtype [42]. Overexpression of p53 is associated with reduced survival, similarly to the situation in multiple other cancer types [35, 37]. Also, DNA ploidy has been shown to be an independent prognostic marker for endometrial cancer, identifying high-risk patients who were presumed to be at low risk on the basis of clinicopathological variables in multiple but not all studies [3941]. A recent review of the subject commented on the absence of definite studies to show its prognostic value in endometrial cancer [43] in spite of the investigation sometimes crossing decennia. This is an aspect that is applicable to most endometrial biomarkers, and there is an urgent need to validate the most promising biomarkers in large prospective studies to obtain a final answer on their prognostic capabilities and hopefully their incorporation in clinical paradigms, where improved prognostication is badly needed.

Additionally, considerable heterogeneity in molecular characteristics of the subtypes included in the non-endometrioid group underscores that this group needs to be specified further as pertaining to serous carcinoma, clear cell sarcoma or carcinosarcomas for the results of studies to be more clinically useful. Table 1 lists prognostic molecular biomarkers in endometrial cancer that have shown independent prognostic value, in different stages of development or clinical use.
Table 1

Prognostic molecular markers in type 1 and type 2 endometrial carcinoma

 

Pathway involved

Function

Alteration

Type 1 (%)

Type 2 (%)

Evidenceb

Risk of extrauterine disease

Stage

Tumour markers

         

PTEN [44]

PI3K

TSG

Mutation/LOH/deletion/hypermethylation

37-61

0-11

2

   

 Oestrogen receptor/progesterone receptor [24, 33, 42, 45•]

 

TF

Loss of expression

27-30

76-81

6

   

PIK3CA mutations [46]

PI3K

OG

Mutation

26-36

26-36

4

   

PIK3CA amplification [46]

PI3K

OG

Amplification

2-14

46

4

   

PIK3R1 [47, 48]

PI3K

TSG

Mutation

43

12

1

   

 MSI [4951]

DNA repair

OG

Promotor hypermethylation

25-30

0-10

3

   

TP53 [50, 52]

DNA damage

TSG

Mutation

5-20

80-90

2

   

ERBB2 (HER2) [53]

RTK/EGFR signalling

OG

Amplification, overexpression

Rare

30

2

   

 CDH1 [54, 55]

WNT

TSG

LOH or promoter hypermethylation

22

60-90

2

   

FGFR2 mutations [56, 57]

FGF signalling

OG

Mutation

16

NA

4

   

KRAS [50, 58]

RAS–RAF–MAPK

OG

Mutation

10-30

0-10

5

   

CTNNB1 [49, 59]

WNT

OG

Mutation

14-44

0-5

4

   

 ARID1A [6062]

Transcription regulation

TSG

Mutation

29-39

18-26 (clear cell)

5

   

 Aneuploidy [3941]

Chromosomal instability

OG

LOH, aneuploidy

Rare

Frequent

5

   

STMN1 [23, 63]

Microtubule cytoskeleton

OP

Overexpression

15

64

5

   

 POLE [49]

Polymerase epsilon catalytic subunit

DNA repair

(Hot spot) mutations

7

*

3

   

 L1CAM [64•]

Immunoglobulin superfamily

OP

Protein positivity

18 % (stage 1)

*

3

   

 GPER [45•]

Alternative oestrogen receptor

TF

Loss of expression

23

63

5

   

Serum markers

         

 CA125a [27, 65, 66]

Epithelial surface protein

Glycoprotein

Increased serum levels (>30 U/l)

   

Sensitivity 0.31

Specificity 0.86

4

 HE4a [27, 65, 67, 68]

Natural immunity

 

Increased serum levels (>70 nmol/l)

   

Sensitivity 0.83

Specificity 0.53

4

 GDF15 [28]

TGF-β superfamily

Cytokine

Increased serum levels (over upper tertile)

   

AUC 0.65

 

5

ARID1A AT-rich interactive domain containing protein 1A, AUC area under the curve, CA125 cancer antigen 125, EGFR epidermal growth factor receptor, FGF fibroblast growth factor, GDF15 growth differentiation factor 15, GPER G-protein-coupled oestrogen receptor, HE4 human epididymis protein 4, L1CAM L1 cell adhesion molecule, LOH loss of heterozygosity, MAPK mitogen-activated protein kinase, MSI microsatellite instability, NA not available, OG oncogene, TF transcription factor, TGF transforming growth factor, TSG tumour suppresor gene, PTEN phosphatase and tensin homolog, PIK3CA phosphatidylinositol-4,5-bisphosphate 3-kinase, catalytic subunit alpha, PIK3R1 phosphoinositide-3-kinase, regulatory subunit 1 (alpha), STMN1 stathmin 1, CDH1 ecadherin, CTNNB1 beta-catenin, WNT want, FGFR2 fibroblast growth factor receptor 2, POLE polymerase epsilon, catalytic subunit, OP oncoprotein

* Biomarker have either not been found (POLE) or investigated (L1CAM) in type 2 cancer

aSome studies found improved hazard ratios for the combination of HE4 and CA125.

b1 preclinical studies, 2 retrospective analysis, 3 retrospective analysis of prospectively collected series, 4 retrospective analysis of population-based series, 5 retrospective analysis of prospectively collected population-based, multicentre series, 6 prospective study (ongoing)

Predictive Biomarkers

In view of their working mechanisms and the distribution of molecular aberrations within and between cancer types, targeted treatments are not expected to work for the entire population. Further, as an example, US Food and Drug Administration approval of HER2-targeting therapy occurred roughly 25 years after the discovery of HER2 [69, 70]. Such a time lag is no longer considered acceptable, and the development process from bench to bedside and approval, although significantly improved, should improve further to make new drugs available to the public at an earlier date. Taking into consideration also the substantial costs associated with clinical trials, calculated to be up to one billion dollars [71], alternative set-ups of clinical trials integrating potential predictive biomarkers could be advantageous. Although intuitively linked to targeted therapies, predictive biomarkers are equally applicable to conventional therapies such as chemotherapy and hormonal therapy, where the rates of response, especially to single-drug treatments, are cited as around 25–30 % [72].

Overall, in cancer, about a dozen predictive biomarkers have demonstrated robust clinical utility and have already been incorporated into clinical algorithms. They include well-known examples such as ERBB2 amplification for using HER2-targeted therapies in breast cancer [73, 74] and BCR-ABL gene fusions for using imatinib in chronic myeloid and acute lymphoblastic leukaemia [75]. Also, specific side effects have been shown to serve as a predictor of clinical effect, such as skin rash severity using tyrosine kinase inhibitors in patients with advanced squamous cell lung carcinoma [76].

Endometrial cancer has in this respect lagged behind, and no (molecular) biomarkers are available to guide treatment decisions, and even hormonal treatment is often initiated without knowledge of the hormone receptor status, in particular in the metastatic lesion, which has been demonstrated to be discordant from the primary lesions in 18–45 % of cases [45•, 77]. Research though has shown that patients with positive receptor status show better response and longer median survival, suggesting that hormone receptor assessment of metastatic lesions may have a yet unexplored potential as a predictive marker in endometrial carcinomas [78].

The phosphatidylinositol 3-kinase (PI3K) pathway is one of the most frequently aberrated pathways in endometrial and other cancers, and a lot or research has focused on it. Unfortunately, a recent study including trials focusing on molecular determinants of outcome with single-treatment mammalian target of rapamycin inhibitors did not identify any biomarker (combination) that was able to do so [79•], including amongst others PIK3CA mutations that in a smaller study, including different gynaecological cancers, including endometrial cancer, were positively associated with response to combination treatments, including PI3K pathway inhibitors [80]. Relatively small sample sizes in a heavily pretreated patient population may have influenced the lack of significant predictive biomarkers detected.

In a recent study combining preclinical data with a retrospective analysis of prospectively included, routinely treated patients, stathmin level was associated with response to taxanes, one of the most frequently used chemotherapeutics in endometrial cancer [81]. Importantly, stathmin is expressed in all tissues and plays an important role in the cell cycle as a critical regulator of microtubule dynamics, thus providing some of the necessary biological substantiation. Table 2 summarises potential predictive biomarkers under development in endometrial cancer.
Table 2

Predictive molecular markers in endometrial carcinoma

Tumour marker

Biological rationale

Aberration

Predictive of response to

Evidencea

Hormone receptors [42, 78]

Receptor binding necessary for treatment functioning

Loss of expression

Endocrine treatment

5

Stathmin [81]

Deactivation compulsory for mitotic spindle formation

Overexpression

Taxanes

6 (planned)

PI3K pathway aberrations [79•, 80]

Important pathway in endometrial cancer

Pathway activation

PI3K pathway inhibition

7b

KRAS mutations [82]

Preclinical evidence for resistance to PI3K inhibition

Pathway activation

PI3K pathway inhibition

7

HER2 [83]

Overexpression in breast cancer indicative of response

Amplification, overexpression

HER2 treatment

7c

BRCA-like [84]

DNA repair

Impairment

PARP inhibitors

 

PARP poly(ADP-ribose) polymerase, PI3K phosphatidylinositol 3-kinase,

a5 retrospective analysis of prospectively collected population-based, multicentre series, 6 prospective study (ongoing), 7 clinical trial(s)

bTested in two trials with different results

cTested in one trial with negative results

Pharmacodynamic Biomarkers

Pharmacodynamic biomarkers prove the direct pharmacological effect of a drug, and may be used to determine optimal drug dosage and avoid drug-related toxicity. No such biomarkers are known in endometrial cancer to date.

Although currently many of the proposed and applied biomarkers in endometrial cancer are single markers, it is expected they will progress into multimarker panels in the future, and this is expected to increase the diagnostic accuracy achieved by applying single markers such as shown in a small study on endometrioid endometrial cancer patients for markers related to PI3K, WNT/β-catenin, p53 and microsatellite instability [85, 86].

The ease of accessibility of a specimen to determine the biomarker status is of importance, especially when considering biomarker determination in recurrent disease or in a frail and vulnerable population as is often the case in endometrial cancer. Although many biomarkers are determined in tumour tissue, where assessment in formalin-fixed paraffin-embedded tissues is still the most applicable method in a routine clinical setting, there is a clear tendency to consider non-invasive techniques, such as determination of biomarkers in blood or urine or using functional imaging, to name a few. This will facilitate introduction in the clinic, especially if repeated determination is required.

Biomarkers and Tumour Heterogeneity

Heterogeneity is arguably one of the bigger challenges in cancer and a potential confounder in many studies, considering subclonality within a primary tumour [8789] or a primary tumour with a concurrent or later metastases [58, 77, 9094]. Biomarker switches, defined as a different biomarker status in metastatic compared to primary tumour, are a frequently reported phenomenon in multiple cancers, and for some biomarkers repeated taking of biopsy specimens has been advised clinically should the result impact treatment decisions [95, 96]. Single tumour biopsies, the standard approach in most cases, are likely to underestimate this heterogeneity, and multiple biopsies, both in space and in time, will help us decipher the tumour subclonality and diversity [87] to treat our patients more adequately. A pooled analysis in breast cancer indicated that repeated taking of biopsy specimens might impact treatment directly in up to one in every seven patients [97]. Although this will undoubtedly pose logistical challenges, patients are in this context shown to be more than willing to accept the burden of repeated taking of biopsy specimens, and up to 90 % of them recommend this to fellow patients [98]. The same article quoted a technical success rate for repeated taking of biopsy specimens of 97 % [98].

How To Enhance Biomarker Development

The last decade has seen great efforts in biomarker research which have resulted in a myriad of potential biomarkers for endometrial cancer, predominantly prognostic (see Table 1). Few have significantly changed clinical routines. Concerning predictive biomarkers, in spite of the previously mentioned clear successes in some cancer types [7375], the take-off of successful implementation of predictive biomarkers to support the use of targeted treatments in the clinic is slow.

One of the reasons underlying the slow uptake of molecular biomarkers in the clinic is the lack of analytical and clinical validation of the biomarkers and (consideration of) their clinical utility. Figure 1 highlights how these critical aspects have been addressed so far for a number of (promising) biomarkers. Some aspects can and should be addressed in preclinical development stages, and when starting clinical validation, even more so in a multicentre setting, analytical robustness should be in place.
https://static-content.springer.com/image/art%3A10.1007%2Fs11912-014-0403-3/MediaObjects/11912_2014_403_Fig1_HTML.gif
Fig. 1

Validation and clinical utility of endometrial cancer biomarkers. The figure reflects the scientific robustness of various essential steps in biomarker development for a number of known and promising biomarkers in endometrial cancer. Red indicates a low level of evidence, such as a limited number of studies or non-prospective studies. In the case of HER2, it has also been used to indicate no suggested predictive value for the biomarker owing to the negative outcome of clinical utility of HER2 treatment as a single drug. Orange indicates a considerable level of evidence, but can also indicate that existing high-level evidence is predominantly obtained in a different cancer type, and thus needs to be explored further in the context of endometrial carcinoma. It may also mean the biomarker has not been implemented widely in the clinic for various reasons. Green indicates the availability of a solid body of scientific evidence and/or (proven) clear clinical utility. As the level of evidence may differ between the prognostic and the predictive abilities of a biomarker, these have been indicated separately with the top box for the prognostic and bottom box for the predictive biomarker respectively. EEC endometrioid endometrial cancer, ER oestrogen receptor, HR hormone receptors, LN lymph node, PR progesterone receptor

Analytical validation refers to the ability of an assay to accurately and reliably measure the analyte of interest in the laboratory and in specimens that are representative of the population of interest and therefore is suited for its intended purpose [99101]. Clinical validation requires the correlation between the biomarker and an associated clinical variable of interest such as survival, lymph node metastases or response in the population of interest. A clear biological rationale would hereby strongly advocate the case of individual biomarkers. Clinical utility implies that the results of the biomarker assay lead to a clinical decision that, with a high level of evidence, has been shown to improve patient outcome [99, 100]. More focus on the clinical utility and the future clinical context, starting at an early phase in biomarker development, will increase the likelihood of successful implementation and applicability in personalised treatment approaches. Inclusion of biomarkers in clinical trials may speed up the process for both and lead to the simultaneous approval of the drug and its companion diagnostic biomarker [102].

Effect estimates in ‘highly cited studies’ of a high percentage (86 %) of biomarkers, as recently demonstrated for 35 top-cited biomarkers, appear (much) inflated when tested in independent datasets or meta-analyses. This assumes that the initial publications on biomarkers are often too optimistic or claim exaggerated associations [103]. Importantly therefore, biomarkers should be validated in independent datasets or, for example, on different platforms and preferentially in population-based settings. The Enhancing the Quality and Transparency of Health Research (EQUATOR) network (http://www.equator-network.org/) is an umbrella organisation focused on research article quality improvement through the development of guidelines on accurate and complete reporting of key aspects. The reporting recommendations for tumour marker prognostic studies are abbreviated REMARK [104], augmenting evaluation of the generalisability of study results.

Systems biology is a research approach where computational modelling is combined with extensive molecular and patient outcome data and reiterative testing to obtain a more integrated analysis than is possible with more traditional research approaches. This systems approach has been shown to be able to, through unbiased analysis, further elucidate the complexity of the molecular aberrations and wider cellular networks in cancer, and as such has the potential to play an important role in the earlier stages of novel biomarker and/or treatment target identification [105].

Large, prospectively collected biobanks with multiple data samples per patient, including tumour samples, but also blood, urine or imaging, together with high-quality patient annotation are a valuable asset to test or screen interesting biomarkers generated from basic research in a timely fashion for their clinical potential and prior to potential inclusion in clinical trials. Recently, this approach demonstrated that low tumour blood flow assessed by preoperative functional MRI was associated with poor prognosis and disorganised angiogenesis with coexisting microvascular proliferation in primary endometrial carcinomas from parallel hysterectomies. This suggests that hypoxia and pathological angiogenesis are associated with progression and metastatic spread and that the biological process may be measured and monitored by advanced imaging techniques [106•].

It is also becoming clear that different cancer types share similar molecular aberrations, such as endometrioid endometrial cancer and colorectal cancer and, for example, serous endometrial cancer with basal breast and high-grade serous ovarian cancer [49]. Promising results of biomarker studies in, for example, basal breast cancer, which is much more intensively studied than endometrial cancer, may therefore function as an interesting starting point for the less frequent, but prognostically unfavourable, serous endometrial cancers. Basket studies, where cancers from different sites of origin are combined on the basis of similar molecular aberrations, are another way to exploit tumour similarities [107]. This may also prevent uncontrolled biomarker or targeted treatment use, in response to increasing patient demand for targeted therapies, in cancer types where they have not undergone rigorous testing in clinical trials and where their assumed effect may or may not be similar to those in more studied cancer types. An example is the use of HER2-targeted therapies, which, when studied in HER2-overexpressed endometrial cancer, did not repeat the breast cancer results and therefore did not meet the expectations [83]. Patients should therefore be included as much as possible in clinical trials, rather than being treated off-label in individual cases, omitting formal testing of the drug for different indications.

Costs are closely related to successful implementation of biomarkers. Long-term cost reductions through biomarker implementation, leading to better patient selection or avoidance of unnecessary treatments and potential serious side effects, would increase the likelihood of benefit necessary to allow cost-effective clinical implementation. An important area where prognostic biomarkers will have clear clinical utility and possibly result in cost reduction in endometrial cancer is stratification for surgical evaluation of potential lymph node metastasis and adjuvant treatment, where the current routinely applied methods are poorly defined, with low reproducibility and with low documented predictive values.

Contrasting with the rather indiscriminative approach of clinical trials in the past, improved definition of likely responsive populations much earlier in the process would potentially be helpful. This approach includes integrating predictive biomarkers in clinical trials from phase 1 either as an integral marker, i.e. required for the trial to proceed to select patients for treatment assignment, or as an integrated marker, i.e. performed on all patients but not needed for clinical decision making [101, 108]. The requirements for analytical and clinical validation of the biomarkers would be dependent on their role within the trial, where the requirements in the case of integral biomarkers will be much higher than, for example, for exploratory or integrated ones [101]. For integrated biomarkers, analytical validation should have been confirmed not only in the research setting, but since it is used for clinical decision making, also for a clinical laboratory setting, where the requirements are naturally more stringent; in the USA, for example, such clinical laboratories have to be Clinical Laboratory Improvement Amendments (CLIA) compliant (http://wwwn.cdc.gov/clia/default.aspx) [109]. Clear guidelines should be available for assay performance. If a likely target group is known, this should have been identified at the start of the study. Equally important, the goal of biomarker incorporation should be clearly defined and weighed against the risks inflicted on the patient though biomarker determination [101]. Window-of-opportunity studies, where standard treatment is briefly delayed to test novel treatments, with access to tumour tissue in the primary operation are another possibility to study predictive biomarkers and novel treatments at an early, not heavily pretreated stage.

In short, biomarkers need to be developed that are much more closely related to clinical needs and (targeted) drugs, including the related quality standards, and the notion that biomarkers can substantially improve clinical trials, already at an early stage, as predictive or ‘companion’ biomarkers, needs to be emphasised more strongly. To do so, much more intense communication and collaboration among molecular biologists, biostatisticians, physicians, and pharmacists is needed, applying a clear translational research approach [102].

We and others have shown that biomarker status can change significantly during tumour development, and repeated assessment in metastatic disease when most of the targeted therapies are applied will help to understand the true potential of a biomarker and its likely efficacy in the clinic [45•, 58, 77, 81, 96]. The accessibility of the biomarker and the burden for the patient from the performance of the assay will constantly need to be balanced against the clinical added value of the assay. Therefore, a strong emphasis on the development of biomarkers or biomarker assays that can be performed in blood or urine or using (dynamic) imaging in parallel with biomarker assessment in biopsied tumour specimens, which are so far most commonly used in the clinic, will increase the likelihood of future clinical implementation.

The run time of the assay is, especially for metastatic cancer patients awaiting the start of treatment, a very relevant issue, where awaiting test results for 2 months would be unacceptable. Tests and test methods therefore should be designed with the clinical reality in mind.

Conclusion

Currently, even the best validated (prognostic) biomarkers in endometrial cancer, p53, OR/PR and DNA ploidy, are not widely and routinely implemented in clinical use. To optimise and individualise treatment in endometrial cancer, there is an urgent need to implement the currently promising biomarkers in the clinic and further develop prospective, predictive and pharmacodynamic biomarkers. Many of the principles in developing clinically useful biomarkers are shared with other cancer types, such as the necessity to ensure rigorous analytical and clinical validation as well as the joint development of drugs and predictive biomarkers and early implementation of biomarkers in clinical trials. A much more intense multidisciplinary collaboration between molecular and cell biologists, drug development teams and clinicians will be needed to develop those biomarkers that meet key clinical needs. Finally, as an approach to handle tumour heterogeneity and repeated biomarker assessments, possibly both in time and in space, may be needed. Development of serum-based assays, advanced imaging biomarkers and, to a degree, biomarker panels may circumvent the problem of repeated taking of biopsy specimens.

Acknowledgments

The European Network for Individualized Treatment in Endometrial Carcinoma (ENITEC) and the European Society of Gynaecological Oncology (ESGO) are thanked for their support.

Compliance with Ethics Guidelines

Conflict of Interest

The authors report the following conflicts:

The University of Bergen/Haukeland University Hospital/Dana Farber Cancer Institute/Harvard University through Bergen Teknologioverføring AS (BTO) have an interest in some aspects relating to prognostic markers for endometrial cancer through (pending) intellectual property rights (US 12/962,946 and US 13/991,947).

Human and Animal Rights and Informed Consent

This article does not contain any studies with human or animal subjects performed by any of the authors.

Copyright information

© Springer Science+Business Media New York 2014