Over the last 50 years, there has been a dramatic decrease in mortality from infectious and cardiovascular disease. However, the same cannot be said about cancer survival, as despite the increasingly sophisticated treatment protocols, mortality from non-haematological cancers remains high. The traditional approach towards managing cancer has always been a “one size fits all”, based primarily on observational studies. This has proven to be costly, and often ineffective or inappropriate. As a result, some patients with aggressive disease may have been under-treated, whereas others with slowly progressive disease may have been over-treated. Our increasing understanding of the molecular biology underlying cancer has created an enthusiasm for a more personalised treatment, recognising the heterogeneity of the disease. The aim of modern cancer therapy should be to deliver the right treatment and dose, to the right patient, at the right time. The availability of reliable cancer biomarkers is key to achieving this outcome by:

Stratifying patients according to the aggressiveness of their disease (prognostic markers).
Estimating the likelihood of response or resistance to the treatment (predictive markers).
Identifying the effect of the drug on the patients (pharmacodynamic markers)

1 Introduction and definitions

Cancer is an umbrella term for many diseases that arise from almost any tissue or organ in the body, which differ in as many properties as they share. Although most cancers tend to invade locally and spread distantly, they have different aetiologies, molecular composition, and diagnostic and treatment methods.

Cancer remains one of the leading causes of mortality and morbidity worldwide with a lifetime risk estimated at 40% [1].

In 2008 there were 12.4 million new cancer cases worldwide and 7.6 million cancer-related deaths. This has risen to 14.1 million and 8.2 million respectively in 2012 (according to the International Agency for Research on Cancer’s online database GLOBOCAN). The same database predicts that annually there will be 19.3 million new cases of cancer by 2025 [2]. This is partly due to an ageing population, but mostly due to environmental factors such as exposure to carcinogens, predisposing lifestyles (smoking, obesity, diet, and a sedentary lifestyle). It is also estimated that 30% of all cancers can be prevented or avoided by modifying these risk factors.

There has been a recent shift towards a “personalised” cancer management as it is clear that certain treatments work best in a certain group of patients. This highlights the need to develop novel and reliable cancer biomarker tests with high sensitivity and specificity for early detection, diagnosis, staging and monitoring treatment response.

Unfortunately, despite the recent advances in our understanding of tumour biology, it has proven challenging to translate candidate biomarkers from laboratory to a bedside setting. In addition to this, those biomarkers that are currently in clinical use are far from being ideal.

In this review, I will focus on answering the following questions: what is a cancer biomarker? What types of biomarkers are there? What are their established and potential clinical uses? I will also briefly touch on why many markers do not perform as well as would be expected in clinical practice and what could be done in the future to improve this.

So what is a cancer biomarker?

According to the Biomarkers Definitions Working Group of the National Institute of Health, a biomarker in general is any cellular, biochemical and/or molecular characteristic which can be measured and evaluated objectively as a surrogate of a normal biological, pathological process or a pharmacological response to a particular treatment [3].

In particular, a cancer biomarker is a biological molecule produced either by the tumour cells or surrounding tissues or microenvironment which can be measured as an indicator of an underlying cancerous process [4].

Cancer biomarkers can be sampled in the circulation (blood, serum or plasma), in secretions (stools, urine, nipple discharge) or any other human biological fluid. It can also be derived from biopsy or surgical resection specimens.

For a marker to be ideal, it needs to fulfil certain criteria [5]:

1. The marker should only be produced by the tumour cells.

2. It needs to correlate with the tumour burden.

3. It should allow for sufficient lead time.

4. It should be present in quantities sufficient to be measured.

5. It should not be detectable in healthy individuals nor those with benign tumours.

6. There should be an easy and cheap test to measure it.

To date, the U.S. Food and Drug Administration has approved 19 protein cancer biomarkers (table 1) [5]. These are routinely used in clinical practice, but they are far from being ideal markers.

Biomarker	Specimen	Clinical use	Cancer type	Methodology
α-fetoprotein (AFP)	Serum	Staging	Nonseminomatous testicular	Immunoassay
Human chorionic gonadotropin-β (β-hGC)	Serum	Staging	Testicular	Immunoassay
Carbohydrate antigen 19–9 (CA 19–9)	Serum	Monitoring	Pancreatic	Immunoassay
Carbohydrate antigen 125 (CA 125)	Serum	Monitoring	Ovarian	Immunoassay
Carbohydrate antigen 15.3 (CA 15.3)	Serum	Monitoring	Breast	Immunoassay
Carbohydrate antigen 27.29 (CA 27.29)	Serum	Monitoring	Breast	Immunoassay
Carcinoembryonic antigen (CEA)	Serum	Monitoring	Colorectal	Immunoassay
Fibrin/fibrinogen degradation products (FDP)	Serum	Monitoring	Bladder	Immunoassay
Human epidermidis protein 4 (HE4)	Serum	Monitoring	Ovarian	Immunoassay
Prostate specific antigen (PSA)	Serum	Screening and monitoring	Prostate	Immunoassay
Thyroglobulin (TG)	Serum	Monitoring	Thyroid	Immunoassay
Epidermal growth factor receptor (EGFR)	Tissue	Prediction	Colorectal	Immunohistochemistry
v-kit Hardy-Zuckerman 4 feline sarcoma viral oncogene homolog (KIT)	Tissue	Prediction	Gastrointestinal	Immunohistochemistry
Estrogen receptor (ER)	Tissue	Prognosis and prediction	Breast	Immunohistochemistry
Progesterone receptor (PR)	Tissue	Prognosis and prediction	Breast	Immunohistochemistry
v-erb-b2 erythroblastic leukemia viral oncogene homolog 2 (HER2-neu)	Tissue	Prognosis and prediction	Breast	Immunohistochemistry
Nuclear matrix protein 22 (NMP-22)	Urine	Screening and monitoring	Bladder	Immunoassay
Bladder tumor antigen (BTA)	Urine	Monitoring	Bladder	Immunoassay
High molecular CEA and mucin	Urine	Monitoring	Bladder	Immunofluorescence

Table (1) List of tumour markers approved by the FDA [5].

2 Cancer biomarkers development

In 2001, the National Cancer Institute’s Early Detection Research Network (EDRN) devised five hurdles before a potential marker can be approved for clinical use. The strength of evidence required increases with each phase. One major downside for this process is that time it takes from initial discovery until clinical approval, which can take a decade or more [6].

2.1 Phase 1: preclinical exploratory studies

This initial phase involves selecting potential biological molecules (“candidate markers”), which should discriminate between cancer and healthy individuals. This is done by two approaches [7]:

“Knowledge-based”: this is hypothesis- driven, developed by deductive methods from the ever increasing understanding of the molecular biology of cancer. Only molecules that play a role in carcinogenesis are chosen.
“Discovery-based”: this relies on the new high-throughput -omics technologies, which select molecules that are expressed differently in cancer patients and healthy individuals.

Despite the large number of potential markers suggested, very few actually make it beyond the discovery stage. The main reason is usually the small difference in the concentration when compared to the controls [8].

2.2 Phase 2: Clinical Assay Development

The next step, following a selection of a potential candidate marker, is to develop a test sensitive and specific enough to measure the marker reliably. To progress to the next phase of development, a biomarker test must show [9]:

a. Analytical validity: the test needs to be reliable, reproducible and accurate.

b. Clinical validity: the test needs to differentiate the healthy and cancer groups, with distinct clinical outcomes.

c. Clinical utility: the test needs to make a positive impact on the clinical outcome.

2.3 Phase 3: Retrospective analysis:

Once a reliable assay is established, it is first tried on historic stored samples, to determine whether it can truly detect the cancer, and also to set a cut-off value for clinical use. At this phase, however, the test is unable to stage the cancer [6].

2.4 Phase 4: Prospective analysis:

This analysis aims to determine whether the biomarker test is able to detect the cancer at its early stage of development. This looks at the potential for the test to be used as a screening test. Asymptomatic individuals, who have a positive result, are then followed up to confirm or refute the diagnosis [6].

2.5 Phase 5: Cancer control studies:

Through large scale studies, the aim of this phase is to determine whether use of the biomarker test reduces mortality or morbidity of cancer in the population studied [6].

3 Types of biomarkers:

Broadly speaking, biomarkers can be divided into three categories: prognostic, predictive or pharmacodynamic markers. It is worth noting that one marker could serve more than one purpose and therefore could fit into more than one category.

3.1 Prognostic markers

A prognostic marker is any marker that predicts the outcome in the absence of a therapy, or predicts an outcome that is different in a patient with a similar type of cancer but without the marker [10]. Therefore, prognostic markers can differentiate between good outcome tumours and poor outcome tumours and can be used to decide on how aggressively the tumour should be treated. Thus, prognostic markers are of particular importance at the time of diagnosis of cancers that vary widely in outcome (e.g. breast or prostate cancer) [10].

In reality, however, no prognostic marker can accurately predict the outcome of the disease in an individual patient. Instead, it provides a probability of outcome in a group of patients [11].

In the last few years, hundreds of prognostic biomarkers have been studied, but only a few have made it to clinical use.

The role of prognostic markers will be to minimise overtreatment of those patients who have indolent malignancies, hence reducing the side effects of treatments, and to avoid undertreating those patients who have aggressive malignancies by offering them the appropriate systemic or local therapies [11].

3.2 Predictive markers

Broadly speaking, a predictive marker is any molecule or factor which provides information on how likely a particular patient is to benefit from a specific therapy [12]. Predictive markers therefore allow a certain degree of personalization as they assess the probability that the tumour will respond to the chosen drug or other intervention. By prospectively dividing patients into groups of those who responded and those who did not respond to a particular anticancer therapy, and correlating this with the presence or the absence of the marker, it may be possible to adjust future treatment choices according to the presence or absence of such markers. In addition to this, predictive markers would allow for considerable cost savings as anticancer drugs would only be given to those patients who will most likely benefit from them [13].

3.3 Pharmacodynamic markers

Pharmacodynamic analysis provides information on the effects of the drug on the body. This would include both early effects on the target of the drug (i.e. whether the drug has any effect on inhibiting its target) and its effects on downstream events [14].

Pharmacokinetics on the other hand describes what the body does the drug, i.e. the process by which the drug is absorbed, distributed, metabolised and finally eliminated from the body [14].

Endpoint pharmacodynamic markers include assessments of factors such as protein phosphorylation markers, measures of cellular proliferation, cell cycle regulator and epigenetic changes. In, cancer, Pharmacodynamics markers are particularly useful in choosing chemotherapeutic agents doses below their cytotoxicity level [14].

4 Potential clinical uses

Different biomarkers can be used at multiple stages of the disease progression (i.e. patient’s status). Potential areas of application of biomarkers include:

– Prior to cancer diagnosis:

 How likely is the patient to develop this type of cancer in the future?

– At diagnosis:

 Does this patient have cancer? What is the grade of this cancer?

– After diagnosis:

 Prognosis: what will be the outcome if no therapy is given?

 Predicting treatment response: which treatment is most appropriate?

 Pharmacodynamics: what is the optimal dose for the anticancer drug?

 Monitoring response: is the therapy effective?

– Post treatment:

 Recurrence: has the patient’s cancer recurred?

The spectrum of patients to whom this would apply would range from those who are unaffected, but perhaps at increased risk of developing certain cancers and therefore would need advice on whether they should adopt a preventive or a screening strategy, to those patients who have been diagnosed at an early stage and now need advice on the appropriate treatment, to those who are disease-free and are concerned about recurrence, and finally, to those whose disease has metastasized.

Interestingly some biomarkers can only be used at a specific stage of the disease whereas others can serve more than one purpose.

4.1 Risk assessment

Risk assessment is searching for the earliest indicator that an asymptomatic person will develop a certain cancer. Traditionally, cancer has been thought of as a genetic disease. However, genetics alone cannot explain sporadic cancers, nor the incidence in those individuals with no prior family history. We now know that epigenetic changes also play a key role in in the various stages of neoplastic development. Genetic factors include translocations and point mutations. Epigenetic changes include DNA methylation and non-coding microRNA expression [15].

Epigenetic modifications, in particular, often constitute the hallmark of exposure to certain risk factors, and as such, are candidates to be used as biomarkers for risk assessment before the histopathological changes occur. Here, I discuss their potential usages in some types of cancer.

4.1.2 Bladder cancer

After prostate cancer, bladder cancer is the most common genitourinary neoplasm. Biomarkers that have been linked to its development include:

DNA methylation of RARB2, APC, JUP, DAL1, APAF1 and DAPK1 [16].
Histone alterations in the coding regions of RASSF1A, PTGS2, PTEN and CDH1 [17].
Abnormal expression of miRNAs (miR-141, miR-155, miR1233 and miR32) [18].

The use of these biomarkers could therefore be an alternative to the more expensive and more invasive procedure of cystoscopy and biopsy to diagnose bladder cancer.

4.1.2 Cervical cancer

Candidate markers that were suggested for cervical cancer risk assessment include SOX9 hypermethylation, which can distinguish between the CIN-1/normal and the SCC/CIN2/3 stages of disease [19]. Other genes that were found to be hypermethylated include the DKK3 gene (which also correlated with disease-free survival) [20].

4.1.3 Hepatocellular carcinoma

The role of single nucleotide polymorphisms in miRNA coding regions in the development of liver cancer, has been demonstrated by Zhou et al. [21]. Their study analysed tissues from 186 primary HCC patients, compared to 483 controls. RASSF1A and methylation of genes targeted by the CRC suppressor proteins were involved in the early stages of liver carcinogenesis. Chronic hepatitis B and C carriers, who are high risk of developing HCC, were found to have an abnormal gene expression profile, abnormal methylation and genomic instability [22].

4.1.4 Pancreatic ductal adenocarcinoma

This cancer still carries a poor prognosis, mainly because it is often diagnosed late. Epigenetic markers may play a role in the early detection of this neoplasm. Candidate markers include the KRAS mutation and p16 methylation, both of which are present in pancreatic cancer [23]. A population- based study by Li et al. has shown hypermethylation of mismatched repair genes hMLH1 and hMLH2 [24]. Another study looked at methylation of cell-free DNA in pancreatitis and pancreatic cancer patients; this showed a distinct pattern in each group [25].

4.2 Screening

Many cancers can be cured or managed better if detected early. Screening is therefore an important part of cancer research, with the potential of improving clinical outcomes, survival and quality of life for patients.

Screening is the use of a test to identify those individuals at high risk, from a certain disorder, and selected for further investigations.

For a screening test to be useful, it needs to fulfil all or most of the criterial of an ideal tumour marker, described earlier. The World Health Organisation (WHO) has set a number of criteria that a screening test should have [26]:

The condition screened for should be important.
The condition should be treatable.
There should be a test for the disease and diagnostic facilities should be available.
The disease should have a latent stage.
The population should accept the test.
There should be sufficient knowledge on the natural history of the disease.
The screening programme should be economically viable.

Several biomarkers are currently being evaluated for potential use as a screening tool, including the use of:

Alpha fetoprotein (AFP) in the screening for hepatocellular cancer, in high risk subjects, e.g. chronic hepatitis B and C carriers.
CA-125, in combination with transvaginal ultrasound, in ovarian carcinoma.
Prostate specific antigen (PSA) for prostate cancer.
Faecal occult blood testing (FOBT) for colorectal cancer (CRC).

This table summarises the biomarkers being currently investigated [27]:

Cancer	Biomarker	Reduction in mortality
Prostate	PSA	Conflicting data
Ovarian	CA-125	Unknown
HCC	AFP	Yes, in 1 study
CRC	FOBT	Yes
Neuroblastoma	VMA / HVMA	No
Gastric	Pepsinogen	Unknown
Gestational trophoblastic disease	HCG	yes

Perhaps with the exception of the use of FOBT in screening for CRC, none of these markers has been unequivocally proven to lead to a reduction in mortality.

Below, I look at the use of AFP in screening for HCC.

4.2.1 Use of AFP in screening for hepatocellular cancer

HCC is relatively rare in the West, where is associated with chronic infection to hepatitis C. World-wide, HCC is the 3rd most frequent cause of cancer-related deaths. It is particularly prevalent in China, South East Asia and Sub-Saharan Africa where it is linked to chronic hepatitis B infection [28]. The presence of a distinct group at risk of HCC lead to the development of two screening tests targeted at this group, with the aim of detecting HCC: measurement of serum AFP and liver ultrasound assessment.

In order to assess the reliability of these tests, two large RCTs were carried out in China:

In the first study, 19,000 subjects with chronic hepatitis B, aged 35-55 were randomised to either receive AFP+USS or the usual standard of care. An AFP cut off of 20µg/L was used, the sensitivity for HCC at this level being 69% and the specificity 95%. The positive predictive value (PPV) for AFP measurement alone was 3.3%. When AFP was combined with ultrasound, the sensitivity increased to 92%, but the specificity dropped to 92.5%. Overall, the outcome analysis showed a 37% reduction in mortality from HCC [29].

The second trial showed that screening with AFP alone lead to an earlier detection of HCC, but did not demonstrate a reduction in overall mortality [30].

Singal et al. performed a meta-analysis of 13 other RCTs; they found the ultrasound sensitivity to be 63% in detecting early stage HCC. The addition of AFP to ultrasound only improved this marginally to 69% [31].

Although hard evidence is not available, several guidelines by expert panels from around the world advocate that high-risk individuals should undergo a form of screening. These guidelines, however, differ on what test should be used:

The National Academy of Clinical Biochemistry (NACB) advises measuring AFP, together with performing an ultrasound, every 6 months. It recommends a cut-off of 20 µg/L. If AFP is found to be rising, the patient should be referred for further investigations, even if the ultrasound is negative [27]. Similar guidelines are given by the National Comprehensive Cancer Network (NCCN). NCCN also recommends the addition of CT or MRI where AFP is rising [32].

In contrast, the American Association for the Study of Liver Disease (AASLD) discourages the use of AFP in screening, except where ultrasound is not available [33].

In Europe, the joint guidelines by the European Association for the Study of Liver (ESL) and the European Organisation for Research and Treatment of Cancer (EORTC) oppose the use of AFP due to concerns over a high rate of false-positive results [34].

4.3 Diagnosis

The lack of specificity and sensitivity of tumour markers in the detection of early disease prevents their use as primary diagnostic markers. They are instead used to aid in the diagnosis when there is uncertainty about other differentials. An example includes the marker CA-125 which is used to differentiate between a benign or malignant pelvic mass [35]. The tumour marker AFP could also be used to elucidate liver lesions. The American Association for the Study of Liver Disease recommends, that where the lesion is > 2 cm, then AFP can be used to confirm the likelihood of this being HCC if AFP > 200 µg/L. This removes the need to perform an invasive biopsy. The same cannot be said if the mass lesion is smaller than 2 cm [36].

Here, I discuss an example of the use of CA-125 in diagnosing ovarian adenocarcinoma.

4.3.1 The use of CA-125 in ovarian cancer diagnosis

CA-125, also known as MUC16, is a membrane glycoprotein, encoded by the MUC16 gene, and belongs to membrane-associated mucins. Mucins are produced by epithelial cells in the GI, respiratory and GU tract, where they play a role in lubrication and protection of the epithelial surfaces. Altered forms of mucins are seen in various pathological processes such as asthma or COPD. More importantly, it is known that the expression and glycosylation of these mucins become abnormal in many adenocarcinomas, including ovarian cancer. This may confer a survival advantage to the tumour cells.

By comparing the levels of CA-125 between cancer patients and healthy controls, a cut-off value of 35U/ml was established (i.e., 99% of healthy individuals would have a level below that threshold) [37].

CA-125 has a poor specificity and sensitivity for ovarian cancer, because its level is also raised in various inflammatory conditions, such as cirrhosis, hepatitis and pelvic inflammatory disease. It is also raised in other malignancies, such as pancreatic, lung, liver and gastric cancers. Because of this, a single measurement of CA-125 cannot be reliably interpreted, and cannot be used as a screening nor diagnostic test in ovarian cancer. In order to make this test more useful, it needs to be combined with other diagnostic modalities, such as CT or ultrasound, particularly in those who fall into the risk group [38].

4.3.2 The use hCG in germ cell tumours:

The human chorionic gonadotropin (hCG) hormone has a variety of physiological functions, but is also used as a tumour marker in many cancers, including the gestational trophoblastic disease and germ cell tumours.

Germ cell tumours, despite being rare, constitute an important diagnosis not to be missed in young males, as it is highly treatable. They represent 90% of all testicular malignancies.

Testicular cancers fall into two types, according to the marker expressed: seminomas and non-seminomatous germ cell tumours (NSGCT). NSGCTs carry a poorer prognosis, as they are more aggressive, and usually require a more complex treatment regimen. Therefore, it is essential to distinguish between these two types at time of diagnosis, to ensure the correct management plan is followed. hCG is detected in both the syncytiotrophoblast cells of NSGCT and in 20% of seminoma patients; however, high serum concentrations of hCG (>300-1000 IU/l) are almost always indicative of NSGCTs [39].

In females, hCG can be used to differentiate between the most common germinal cell tumour; i.e., dysgerminoma, which usually secretes hCG, from yolk sac tumours, which only secrete α-fetoprotein and only rarely hCG [40].

4.4 Prognosis and Treatment Selection

Prognostic markers are measured, independent of treatment, to provide information to oncologists about the likely clinical outcome. They can also be used at the time of diagnosis to predict the likelihood that the cancer has spread to lymph nodes or whether it has metastasised. This could be used to make a decision on adjuvant therapy in patients who undergo curative surgery for instance [41].

On the other hand, predictive markers are used during treatment selection, to decide if a specific patient group is more likely to benefit from a particular treatment, i.e., the concept of “personalised medicine”. Somatic markers, such as epidermal growth factor receptors (EGFR), have traditionally been used as predictive markers. More recently, there has been a focus on the methylation status of genes or the expression of miRNA to guide treatment [42].

Here, I discuss the use of such markers in various types of cancer:

4.4.1 Breast cancer

Predictive marker

Predictive markers in breast cancer include hormone receptors such as the oestrogen receptor (ER) and the progesterone receptor (PR), which are used as targets for hormone therapy. Current guidelines, therefore, recommend hormone therapy for receptor- positive patients both as adjuvant and in metastatic disease [43].

Another predictive marker used is HER-2, which is the target of inhibitors such as trastuzumab and lapatinib. Patients who were HER-2 negative did not respond to these monoclonal antibodies [44].

Prognostic marker

The hormone receptor HER-2 is used as a prognostic marker too; HER-2 positive patients usually carry a poor prognosis. A similar prognosis in seen in HER-2 positive gastric cancer patients [45].

4.4.2 Colorectal cancer (CRC)

Predictive marker

The monoclonal antibodies cetuximab and panitumumab have been used in the treatment of metastatic CRC, which target the EGFR receptor. However, studies have not shown a correlation between the expression of EGFR and response to treatment [46]. Further studies on the genes in the EGFR signalling pathway have shown that anti-EGFR antibodies are only effective in metastatic CRC exhibiting the wild type of KRAS, NRAS, BRAF and PIK3CA [47]. Recent guidelines, therefore, only recommend the use of anti-EGFR antibodies for patients with the wild type KRAS and MRAS [48].

Prognostic marker

In familial adenomatous polyposis (FAP), a mutation of the APC tumour suppressor gene predisposes to the development of adenoma and adenocarcinoma. Screening of individual carriers is therefore recommended. DNA mismatch repair deficiency (dMMR) has been shown to lead to genetic instability, especially in the region of highly repetitive DNA known as microsatellites. An individual with a high frequency of instability at microsatellites is particularly vulnerable to developing CRC [49].

4.4.3 Lung cancer

Predictive marker

Patients with non-small cell lung cancer (NSCLC) get routinely tested for the EGFR kinase mutations in exons 19 or 21. Those who are positive are then offered the tyrosine kinase inhibitors gefitinib and erlotinib [50].

Another predictive biomarker is the anaplastic lymphoma kinase gene (ALK), as mutations in this gene lead to overexpression and activation of the ALK fusion protein. Patients with this mutation are offered the ALK inhibitor crizotinib [51].

Thus, testing for the EGFR mutation and ALK gene rearrangement is now included in the workup for NSCLC.

Prognostic marker

Various biomarkers were studied for their potential to be used as prognostic markers, e.g., the KRAS mutation in NSCLC. However, the results were disappointing as it was found to be a poor prognostic marker [52].

4.5 Monitoring response

One of the main uses of cancer markers is to monitor the effect of systemic therapy. In general, the same markers used in diagnosis or screening are used to monitor response. Broadly speaking, if the level of the marker starts decreasing, then this would indicate a good response to treatment, and vice versa. The only exception to this is the initial spike seen in people with advanced malignancies on starting treatment, which perhaps could be explained by transient tumour cell necrosis.

Below, I discuss the use of various markers in monitoring response:

4.5.1 The use of CA 19-9 in monitoring response to chemotherapy in pancreatic adenocarcinoma

Several studies have looked at the reliability of CA 19-9 in treatment monitoring. Willett et al. [53] took serial measurements in 42 patients before and following treatment with 5-FU. They compared the results with restaging imaging CTs. The results showed a correlation between CA 19-9 changes and disease progression (P = 0.009). Another study by Halm et al. [54] analysed CA 19-9 levels in 36 patients receiving gemcitabine and showed that decreasing levels correlated with the median survival (P = 0.001). Despite these results, and other concurring studies, the guidelines from the NACB Panel do not recommend using CA 19-9 on its own; instead, it advises to combine this with imaging to decide on the success of the treatment.

4.5.2 The use of PSA in monitoring response:

The PSA test was approved in 1986 by the FDA to monitor response in prostate cancer. Pound et al. [55] followed patients for 5 years after radical prostatectomy (RP); they found that no patient had a recurrence without a corresponding rise in their PSA. The level of PSA should be undetectable after RP. The American Urological Association and the European Association of Urology define a biochemical recurrence as a PSA of > 0.2 ng/mL [56].

PSA is also used to monitor response after radiotherapy (RT). This is slightly more complex, as the PSA will be detectable even after treatment, due to the residual prostate tissue. The Phoenix criteria were set in 2005, which define biochemical recurrence after RT as a rise of > 2 ng/mL above the post-radiation nadir [57].

4.5.3 Circulating tumour cells (CTCs) for prognosis and monitoring of breast cancer:

CTCs are cells that are shed by the tumour (primary or metastatic), and therefore can be sampled in serum (“liquid biopsy”).

In the case of breast cancer, CTCs were evaluated as prognostic markers, and their role in monitoring response to treatment. The DETECT study [58] evaluated 221 patients with metastatic breast cancer. They used the CellSearch® Assay for CTC enumeration. The median survival was 18.1 months for CTC- positive patients, compared to 27 months for CTC- negative patients (P < 0.001).

In the area of disease monitoring, Liu et al. [59] followed up 68 patients with metastatic breast cancer for a median period of 13.3 months. Their results demonstrated a statistically significant correlation between the CTC levels and disease progression seen on imaging.

For this reason, a phase III trial is currently underway in the USA (SWOG S0500) to determine whether CTC monitoring can be used to check treatment response early, and change to a different regimen if the initial one does not work [60].

4.6 Monitoring recurrence

Many of the tumour markers used in screening and diagnosis, are also used for surveillance post-operatively, or following remission. Table 2 summarises some examples of these. Biomarkers can show early signs of recurrence, before clinical radiological changes show. This provides a lead time which can be advantageous, as it is believed that the sooner the treatment is initiated, the better is the outcome [61].

Cancer Marker(s)	Marker(s)
Colorectal	CEA
Hepatocellular	AFP
Pancreatic	CA 19-9
Ovarian	CA 125
Breast	CA 15-3
Prostate	PSA
Germ cell	AFP, HCG
Lung (non-small cell)	CYFRA 21-1, SCC
Lung (small cell)	NSE, proGRP
Melanoma	S100
Trophoblastic	HCG
Thyroid (differentiated)	thyroglobulin

Table 2 Examples of tumour markers used in surveillance [62]

Below, I discuss the usefulness of such markers in various cancer types.

4.6.1 CA-125 in patients with ovarian cancer:

Serial CA-125 measurements after therapy can detect recurrence with a median lead time of 4 to 5 months [63]. However, the clinical value of this is still not clear. A recent prospective randomised trial found no survival benefit, when treatment was restarted following detection of a rising CA-125 [64]. The negative finding could be partly attributed to the absence of any effective second line therapies in ovarian cancer. Current guidelines regarding serial measurements therefore differ; where the NACB panel (National Academy of Clinical Biochemistry) in USA encourages this [65] and the EGTM (European Group on Tumour Markers) panel opposes it [66]. Because of the absence of solid evidence, and contradicting guidelines, it is probably best to follow the patient’s wishes when deciding on this.

4.6.2 CA 15-3 in breast cancer

There is no clear evidence to support the use of CA 15-3 in breast cancer recurrence surveillance, despite this being the most widely used worldwide [67]. As a consequence, guidelines vary on the use of CA 15-3 in asymptomatic breast cancer patients. The NACB does not recommend routinely offering this test for early detection of recurrence [65]. The EGTM, however, advises using this test in this setting. Ultimately, the wishes of the patient should be respected.

4.6.3 Carcinoembryonic Antigen in colorectal cancer

Various studies have shown a better outcome in those patients undergoing surveillance with CEA following curative surgery [68]. European and American guidelines therefore recommend its use in clinical decision making. The EGTM recommends that patients with Dukes’ B and C CRC should be offered CEA measurement 2 to 3 monthly, for at least 3 years following diagnosis [69].

There is, however, no consensus on the cut-off point that should be used. The EGTM defines a “significant” increase if the level of CEA 30% or more than the previous level. It also recommends repeating this test a month later, to confirm this increase, before instigating further investigations [69].

5 Cancer biomarkers in clinical practice: reality and limitations (case study: health economics of prostate specific antigen)

Prostate cancer is the most commonly diagnosed malignancy in men worldwide. The prevalence has steadily increased recently due to an ageing population and the increased usage of the PSA screening test.

The aim of screening asymptomatic men for prostate cancer, as with any screen program, is to detect incidence early and to afford the person the best possible clinical outcome and quality of life. However, when considering mass population screening one must weigh the benefits and harms of such a program, and whether it is financially viable.

The PSA test is used to diagnose prostate cancer, which is suspected when the level is increased. This is usually combined with a digital rectal examination. Confirmation of the diagnosis would require a biopsy and histology.

In its latest policy review in 2010, the UK National Screening Committee concluded that there was no benefit from introducing a national screening programme for prostate cancer. However, a Prostate Cancer Risk Management programme is available to help men make an informed choice about undergoing screening or not.

5.1 The evidence available

There have been five major randomised controlled trials recently that looked at the effectiveness of PSA- based screening: the European Randomised Study of Screening for Prostate Cancer (ESRPC) [70], the US Prostate, Lung, Colorectal and Ovarian (PLCO) [71] cancer screening trial, the Norrkoping and Stockholm studies in Sweden and the Quebec study in Canada. The Cochrane Collaboration reviewed these RCTs and conducted a meta-analysis to evaluate the effect of screening on cancer- specific and all-cause mortality. In total, 341,342 participants took part. Among these RCTs, the ESRPC and PLCO where the largest and were considered to have a low probability of bias [72]. Table 3 summarises these two studies:

The ESRPC found that the PSA screening reduced prostate cancer-specific mortality significantly (rate ratio = 0.84; 95 % confidence interval: 0.73-0.95) [70]. On the other hand, the PLCO found no significant difference between the screening and the control groups (RR=1.15; 95% CI: 0.86-1.54) [70]. When the ESRPC was limited to those aged 55 to 59, the reduction in mortality was 21%. Overall, the meta-analysis concluded that PSA screening did not reduce prostate cancer- specific mortality (RR=1.00, 95% CI: 0.86-1.17 nor all-cause mortality (RR=1.00, 95% CI: 0.96-1.03) [72].

5.2 The UK CAP Study

Despite the promising results of the ESRPC study, the use of PSA was still controversial because of issues of over-diagnosis and overtreatment of clinically insignificant tumours. To overcome this, The UK Cluster randomised trials of PSA testing for prostate cancer (CAP) was setup to evaluate whether testing men aged 50 to 69 would be cost-effective and reduce mortality [73]. The study randomised patients into either those who get screened (the intervention arm) and those who would receive the standard clinical care (comparison arm). The study recruited 573 GP practices in England, Scotland and Wales. The first results were published in 2016.

The study compared active monitoring, radical prostatectomy, and external-beam radiotherapy for the treatment of prostate cancer. A total of 82,429 men aged 50 to 69 received a PSA test; 2664 were found to have localized prostate cancer, and 1643 underwent randomization to active monitoring (545), surgery (553), or radiotherapy (545).

Overall, there were 17 deaths: 8 in the active-monitoring group (1.5 deaths per 1000 person-years; 95% CI 0.7 to 3.0), 5 in the surgery group (0.9 per 1000 person-years; 95% CI, 0.4 to 2.2), and 4 in the radiotherapy group (0.7 per 1000 person-years; 95% CI, 0.3 to 2.0). The study concluded that at 10 years, prostate cancer–specific mortality was low irrespective of the group assigned to, with no major difference among treatments. Surgery and radiotherapy groups showed a lower incidence of disease progression and metastases compared to the active monitoring group [74].

5.3 Cost effectiveness of screening in the UK

Based on the results of the ERSPC trial, the ScHARR (University of Sheffield School of Health and Related Research) published a report in 2013, which devised a model to compare four screening policies to not screening [75]:

Policy 1: a single screen at age 50

Policy 2: screening every four years from age 50 to 74

Policy 3: screen every two years from age 50 to 74

Policy 4: screening every year from age 50 to 74

The screening impact model predicts that:

A one-off screen at age 50 is the same has no screen
Annual screening had a marginal impact on age- specific incidence compared to longer interval screening
One- off screening was associated with the least over-diagnosis rate.
The average life years gained increased with shorter intervals between screening, ranging from 2-4 extra days of life gained for a one- off screen at age 50, to 27-67 extra days for yearly screening.
However, every life gained would require 17-32 years of prostate cancer management. Consequently, policy 2-4 are associated with loss of discounted QALYs ranging from 0.016 to 0.023 man invited to screening.
The cost of policy 1 is estimated to be £58 million, increasing to £1 billion for policy 4.

5.4 Implications for policy

Despite the promising result of the ERSPC trial of a 21% reduction in prostate cancer- specific mortality, the evidence is still not sufficient to justify a universal PSA screening programme. There are still unresolved issues surrounding over-diagnosis and over-treatment of clinically insignificant cancers. The current available evidence tips the balance more towards harm, than benefit.

PSA screening is therefore not seen as cost-effective, and the UK National Screening Committee does not recommend a PSA-based universal screening programme. This, however, does not preclude men who want to have the PSA test. Indeed, a Risk Management Programme is available to provide information on the pros and cons, and allow an informed choice.

6 Other (non-serum) biomarkers

As discussed earlier, there is a shift in trend towards “personalised” treatments for cancer patients, as our understanding of the basic molecular biology advances. This applies to the traditional treatments, such as surgery and radiotherapy, as well as targeted systemic therapies. A good example of this is hormone therapy in breast cancer. The advances in functional and molecular imaging has to lead to their use as cancer biomarkers. There are three main areas in which imaging can optimise cancer treatment:

– Prognosis, i.e., to determine how aggressive the tumour is.

– Prediction, i.e., to determine the presence of molecular targets to therapy.

– Response: to determine the earliest signs of response or no response.

Here, I discuss the role of somatostatin receptors imaging in the management of neuroendocrine tumours (NETs).

Neuroendocrine tumours (NETs) are a distinct group of tumours, which can be found anywhere in the body, arising from neuroendocrine cells [76]. They all the share the ability to secrete biogenic amines and hormones. They are characterised by their low prevalence and slow proliferation, which makes their diagnosis difficult [77]. Common sites include the gastrointestinal and bronchopulmonary tracts. As all NETs express the somatostatin receptor (SSTR), this has been targeted using somatostatins (SST).

SST and SSTR

SST is found in the cerebral cortex, the hypothalamus, brainstem, pancreas and gastrointestinal tract, and is produced by immune, inflammatory and neuroendocrine cells [78]. There are five G-protein coupled receptors, which facilitate the function of SST (SSTR1 to 5) [78]. For example, pancreatic endocrine and carcinoid tumours primarily express SSTR2 [79]. The overexpression of SSTR by NETs was exploited in imaging using radiolabelled SST analogues [80]. However, the clinical use of SST is limited due to its short half-life in the plasma [81]. Imaging with SST analogues, such as octreotide or octerotate, is widely used instead as the first line imaging for NETs [82].

Clinical use of radiolabelled SSTR in NETs

1) Single Photon Emission Computed Tomography (SPECT) Imaging:

¹¹¹In-pentetreotide is the first choice for detecting receptors for SST analogues. It can, for instance, differentiate scar tissue from a recurring tumour after pituitary surgery [83]. Another agent, ¹¹¹In-DTPAOC, is also used and especially in NET scintigraphy. Overall, Somatostatin Receptor Imaging (SRI) has been proven to have a high sensitivity for localising NETs [84].

2) Positron Emission Tomography (PET)

⁶⁸Ga-labelled somatostatin analogues are widely used in PET imaging (the main compounds used are ⁶⁸Ga-DOTATATE and ⁶⁸Ga-DOTATOC) [85]. One particular study, involving 38 patients, demonstrated a detection sensitivity of 82% for ⁶⁸Ga-DOTATATE PET/CT [86]. Where the lesions are small and with low tracer uptake, 68Ga-DOTATOC was found to be superior to ⁶⁸Ga-DOTATATE [87]. When comparing PET/CT to SPECT and conventional diagnostic CT, ⁶⁸Ga-DOTATOC was found to be significantly superior [88]. A study comparing ⁶⁸Ga-DOTATATE and 68Ga-DOTATOC PET/CT reported a similar accuracy [85] (figure 1).

Fig 1: higher uptake BY LESIONS in 68Ga-DOTATOC than 68Ga-DOTATATE imaging. [85].

3) Somatostatin Receptor Targeted Radionuclide Therapy

The first choice treatment for NETs is surgery. The vast majority will then require further treatment with somatostatin analogues or peptide receptor radionuclide therapy (PRRT) [89]. The expression of SST2 is an essential criterion to qualify for radiolabelled octapeptide SST analogues [79] or ⁹⁰Y- and ¹⁷⁷Lu-DOTATATE/DOTATOC [90]. SSTR PET imaging is used to select candidates for PRRT. Patients who show a high 68Ga-DOTATOC uptake were treated with ⁹⁰Y-DOTATOC [91]. On the other hand, those who show increased uptake of ¹¹¹In-octreotide in scintigraphy are treated with ¹¹¹In/⁹⁰Y-octreotide [92].

Molecular imaging is therefore a promising non-invasive method which can be used in the diagnosis, treatment choice and monitoring response. PET/CT, in particular, has been shown to play a key role in predicting treatment outcome, and helps in decision making on therapy.

7 Conclusion and future prospects

Despite the massive research output in the field of cancer biomarkers, their translation into clinical practice has been disappointing. Very few, if any, markers have made into the bedside in the last 30 years. So why has it proven difficult to implement new tumour markers? One explanation may rely on how we define success and failure. Whilst a new therapy, which would extend survival by a few months, may be hailed as a breakthrough, the same cannot be said about a new biomarker with a modest improvement in reliability. Ioannidi [93] classified biomarker failures into four categories:

a) Type A failure (clinical reversal): this happens when a clinically established marker is later proven to be not as effective.

b) Type B failure (validation failure): this is where a marker is promising in early studies, but not in large studies.

c) Type C failure (non-optimised clinical translation): this is where a promising marker does not show any clinical reliability.

d) Type D failure (promotion despite non clinical evidence): this occurs when a marker is promoted despite lacking any evidence to its usefulness.

However, despite these setbacks, the recent advances in the fields of genomics (e.g., next-generation sequencing), have allowed us to identify a new breed of markers, such as circulating tumour cells (CTCs), cell-free tumour DNA (cfDNA) and microRNA, which are collectively known as “liquid biopsy”. These would allow us repeat sampling and could be used at various stages of the diagnostic or therapeutic journey of the patient.

CTCs are cells of tumour origin, which can be sampled non-invasively to monitor tumour progression. CTCs generally occur at very low concentrations, which can present a challenge for their detection. Currently, the only assay approved by the FDA is the CellSearch® System [94]. These cells have been detected in various types of cancer, including those of the breast, lung and colon. Their clinical utility, however, is still under investigation.

Circulating cell-free DNA is secreted by normal and tumour cells, as a result of cell injury or necrosis [95]. The ability to detect a significant surge can be used to differentiate those patients with malignant tumours, from those have benign masses (or those who are normal) [96]. Currently, however, this promising area of investigation is not felt to be ready yet to be used clinically [97].

Another promising marker is microRNA (miRNA), which are fragments of non-coding RNA, responsible for regulating various genes of cellular apoptosis and division [98]. It has been observed that the expression of these miRNAs can be unique to a particular tissue, and is different between normal and cancer patients. Further research is needed to investigate their potential use in clinical practice.

To conclude, it is clear that we all carry enormous amount of information in our blood, which can be harnessed by various marker assays, to optimise cancer screening, diagnosis and treatment.

What Are Cancer Biomarkers and How Can They Be Used in Oncology Practice?

What Are Cancer Biomarkers and How Can They Be Used in Oncology Practice?

1 Introduction and definitions

2 Cancer biomarkers development

2.1 Phase 1: preclinical exploratory studies

2.2 Phase 2: Clinical Assay Development

2.3 Phase 3: Retrospective analysis:

2.4 Phase 4: Prospective analysis:

2.5 Phase 5: Cancer control studies:

3 Types of biomarkers:

3.1 Prognostic markers

3.2 Predictive markers

3.3 Pharmacodynamic markers

4 Potential clinical uses

4.1 Risk assessment

4.2 Screening

4.3 Diagnosis

4.4 Prognosis and Treatment Selection

4.5 Monitoring response

4.6 Monitoring recurrence

5 Cancer biomarkers in clinical practice: reality and limitations (case study: health economics of prostate specific antigen)

5.1 The evidence available

5.2 The UK CAP Study

5.3 Cost effectiveness of screening in the UK

5.4 Implications for policy

6 Other (non-serum) biomarkers

7 Conclusion and future prospects

Professor