What Makes a Marker Useful?

There is great interest in the development of new markers that can aid cancer patients and their physicians in the process of clinical decision-making. There are three general categories of such markers:

Diagnostic markers are important aids in assessing the tissue of origin of a malignancy or the specific subtype of a tumor.

Prognostic markers indicate the likelihood of outcome (tumor recurrence, patient survival) regardless of the specific treatment the patient receives. For example, in most solid tumors the spread of cancer cells to lymph nodes indicates an increased likelihood of tumor recurrence independent of the particular form of therapy the patient receives following surgery.

Predictive markers indicate the likelihood of response to a specific therapy. For example, breast cancers that express the estrogen receptor tend to respond to hormonal therapies such as tamoxifen.

Some markers can be used to monitor a patient’s response to therapy or to detect the growth of metastases. Markers of precancerous conditions are also sought, in order to serve as the basis for screening strategies or to identify populations at high risk for cancer who might benefit from preventive measures.

There is increasing interest in defining molecular targets for new therapeutic agents. For example, treatment with the monoclonal antibody Herceptin is offered to patients whose tumors have amplified the Her2/neu gene or over-express the Her2/neu gene product.

Markers are generally considered to be expressed within the neoplastic cells. However, useful markers may be molecules that are associated with non-neoplastic cells that may be components of the stroma of a cancer or the infiltrating host inflammatory cells. Also while some markers are secreted into the blood or other body fluids other markers are restricted to expression within neoplastic tissues. However, as a marker is characterized, it is important that the investigator determine where the marker is most easily measured because it may have a large impact on the design and sensitivity of an assay to measure that marker.

The successful outcome of research in cancer biomarkers is the development of a new assay, procedure or technique that provides information useful to physicians and patients in designing the course of cancer treatment. A new marker is obviously significant if it will have a clear impact on those decisions that clinicians struggle with today. The nature of a “difficult decision” depends on the type of cancer, the stage of the disease and the range of treatment options.

Ultimately a useful marker is one that meets two criteria:

1. It can be measured reproducibly by means of a reliable and widely available assay.
2. It conveys information about the disease that is meaningful to the physician and patient.

It may not be clear at the outset whether a new marker will be most useful for the purpose of cancer detection, diagnosis, prognosis or predicting response to specific therapies. Validation of a marker for cancer screening in asymptomatic people will generally require testing large numbers of subjects, most of whom will never develop cancer, and waiting many years for outcome data to accumulate. Validation of a marker and test for predicting response to a specific therapy may take less time and be incorporated into the design of phase II or phase III clinical trials of novel therapeutic agents.

The assay used to measure or evaluate expression of a marker is as important as the marker itself. Assay techniques that were originally developed for use with cultured cells or animal tissues frequently must be modified for use on the patient tissues or fluids sent to the clinical laboratory. Most clinical laboratory assays, with the exception of some blood and urine tests, are designed to work on tissues that have been preserved in some way. Most diagnoses of cancer are made by microscopic examination of a specimen that is fixed in formalin and embedded in a block of paraffin. Sections of frozen tissue may be used to make a preliminary diagnosis. Assays that require fresh tissue or tissue preserved in fixatives other than formalin are generally more complicated and more expensive to perform.

Clinical laboratory assays require scrupulous attention to control for intra- and inter-laboratory variability in the scoring and reporting of results. Often the investigators scoring laboratory assays are blinded to the clinical outcomes and vice versa, to minimize bias. Quality-control procedures should be designed with active participation by the statistician and should be in place from the very beginning whether a study is to be performed at a single site or at multiple institutions.

The following considerations may help investigators to recognize a potentially useful marker:

1. Is there a clinical need for this marker? The definition of clinical need is pertinent whether the marker is a single analyte, such as a gene or gene product, or a classifier composed of multiple analytes. Preliminary data will be required to support the importance of a marker by demonstrating that it is associated with an outcome such as survival or recurrence of disease; that it improves the classification provided by known clinical parameters such as stage of disease or nodal status; or that it identifies some other subset of patients.
2. Is the marker or classifier evaluated in an assay that has good intra- and inter-assay reproducibility? Is there a well-defined scoring system that is clearly associated with an important clinical outcome? How difficult would it be to reproduce the assay in other laboratories?
3. Has the marker or classifier been examined in normal as well as abnormal or diseased tissue? Has it been examined in different organ sites? This is not so much to establish the exact distribution of the marker but to help determine the setting where the marker might have greatest value. It is important to develop positive and negative controls that may be useful in standardizing the assay.
4. Can patient populations be defined for which this marker might have utility? What is an expected range for the prevalence of this marker in any populations of potential interest? Very rare (<5%) or very prevalent (>95%) markers are likely to be useful in more circumscribed settings.
5. Can the marker be measured in the types of specimens that will generally be available?