Before deciding to light your first cigarette, imagine knowing your exact chance that this will lead to cancer. Although smoking is the main cause of lung cancer, only 10%–20% of smokers and former smokers actually develop the disease. The reasons for this — and for the changes that lead to many cancers — have eluded researchers for decades.

Lung cancer, as the biggest cause of cancer death worldwide, is a priority of prevention strategies. And to make the biggest impact in cancer prevention, it is vital to target those individuals with the highest risk of developing cancer. This is where biomarkers come in. Individuals identified as being high risk using biomarkers could receive counselling for lifestyle changes (see Breaking the cancer habit, page S16), or they might be eligible for chemoprevention (see First line of defence, page S5). Even before that point, using biomarkers to select people for cancer prevention studies would allow for more powerful trials (see Designing smarter cancer prevention trials, page S20). Indeed, there isn't a clear distinction between the types of biomarkers that help identify risk before cancer has developed and those that indicate that cancer is just starting to grow.

The process of carcinogenesis takes years, if not decades. The search is on to discover and validate the often-subtle, microscopic changes in the constituents of the blood, sputum, urine or tissue samples that herald cancer. Prior to activated oncogenic pathways, there is the possibility of identifying sluggish DNA repair mechanisms, changes in gene expression, or detecting the low-level immune response to the presence of a nascent tumour.

This research activity is not without its controversies. There have been numerous false starts, where promising biomarkers did not stand up to rigorous testing. To be useful, a biomarker needs to have sensitivity — that is, the likelihood that it detects disease — of at least 90%. The other key quality is specificity — the probability that a positive signal is a true sign of disease and not an error. That, too, should be 90% or more for a biomarker to be of clinical value. Although no lung cancer biomarkers yet meet that 90/90 standard, there are several promising candidates.

Changing expression

Avrum Spira, a pulmonologist at Boston University, has been using bronchoscopy to brush cells from the bronchial airways of healthy smokers and non-smokers, followed by gene expression profiling to compare thousands of genes. He has found that the cells lining smokers' airways show signs of genomic changes related to inflammation and cell proliferation, even when they appear normal with standard bronchoscopy. By comparing these results to those of smokers with suspected lung cancer, Spira's group has identified an 80-gene signature that could identify patients with early-stage lung cancer with about 90% sensitivity. This signature, says Spira, “is the proverbial canary in the coalmine”.

However, bronchoscopy is an invasive procedure, so Spira broadened his search to more accessible parts of the airway. He recently found abnormal gene expression in epithelial cells of the nose and mouth that resemble those in the bronchial airway1. Analyzing these cells using a simple swab could serve “as a mass screening tool in population-based studies,” he predicts.

Digging into the pathways underlying these precancerous changes, Spira, together with genomicist Andrea Bild from the University of Utah, found phosphatidylinosital 3–kinase (PI3K ) signalling pathway, already known to be involved in the development of cancer2. By comparing PI3K activity levels in cells from apparently healthy smokers to those of smokers with mild-to-moderate abnormalities, they found that this pathway “might be activated before tumorigenesis,” says Spira, making PI3K a prime biomarker candidate.

The PI3K pathway might also be used for chemoprevention. Early trials have shown that the administration of a compound that decreases PI3K activity causes regression of abnormal lesions in the airways of high-risk smokers3.

DNA damage

As part of daily living, DNA frequently sustains damage. If not repaired, this can lead to mutations that replicate, resulting in abnormal and then cancerous growths. Certain mechanisms usually prevent this from occurring. The enzyme 8-oxoguanine DNA glycosylase (OGG1) repairs DNA by excising damaged bases (see DNA repair duties,S21). Biochemists Zvi Livneh and Tamar Paz-Elizur, at the Weizmann Institute in Rehovot, Israel, discovered that levels of OGG1 can also be used to predict an individual's risk of developing lung cancer.

By measuring OGG1 concentration in blood samples, Livneh and Paz-Elizur found that 40% of people with lung cancer had low levels of the enzyme compared to 4% of healthy individuals. Smokers with low OGG1 activity were 5 to 10 times more likely to develop lung cancer than smokers with normal OGG1; when compared to non-smokers with normal OGG1 activity, the risk skyrocketed to 120 times more likely. The same blood test could be broadened to other cancers. For example, smokers with lower OGG1 activity are 70 times more likely to develop head and neck cancer than non-smokers with normal enzyme activity4.

OGG1 is only one of an unknown number of DNA repair enzymes; low levels of any of them could be associated with cancer development. Livneh and Paz-Elizur have expanded their research to include two additional DNA repair enzymes — AAG and APE1 — to cover people with “different risk factors to develop a certain cancer”, says Livneh. A study is underway to access their performance and results are expected in mid-2011.

It is unlikely that any single test, however many markers included, will be sufficient to gauge the risk of cancer development. “We have an additional ongoing study which explores a two-stage protocol for lung cancer prevention,” says Livneh. The first stage involves Livneh and Paz-Elizur's DNA repair biomarkers plus five biomarkers developed by other groups. These biomarkers measure: alteration in gene expression; levels of DAP kinase, an enzyme involved in programed cell death; antibodies to mutant p53, a sign that a cell's tumour suppressor system is damaged; markers of inflammation; and variations in cancer-related genes. “Together these biomarkers are expected to yield a better risk assessment than one type alone,” says Livneh. Individuals identified as high risk will be tested using spiral computed tomography (CT). “For such a high-risk group, spiral CT early detection of lung cancer might be cost-effective and life saving,” adds Livneh.

In the initial stages of cancer, the body is often able to recognize abnormal cell changes and raise a response, producing auto-antibodies. However, this response is limited, and in the later stages of cancer, the immune system becomes compromised and can no longer identify and attack cancer cells. Auto-antibodies are therefore prime candidates for biomarkers of early stage cancer.

By examining auto-antibody formation in presymptomatic individuals who later went on to develop lung cancer, Samir Hanash, at the Fred Hutchinson Cancer Center in Seattle, Washington, has identified three important antigens — annexin-1, 14-3-3 Theta and LAMR-1 — regarded by the immune system as foreign5. So far, specificity of these biomarkers is high but sensitivity lingers around 60%. The challenge for Hanash is to find additional candidate antigens that improve on the performance of this 3-antigen panel.

These figures might be improved by looking for even earlier signs of cancer. Through the Women's Health Initiative and Physician's Health Study, Hanash has access to blood samples that were collected up to eight years before a patient was diagnosed with lung cancer. In addition, he is searching for biomarkers of lung cancer in former smokers and in people who never smoked. “It turns out that most of the blood markers we have identified among smokers are also applicable to non-smokers,” says Hanash.

In spite of major investment in biomarker development over the past 15 years, the field of cancer prevention biomarkers looks woefully thin. One of the main reasons, according to Eleftherios Diamandis, a clinical biochemist at the University of Toronto, is because of poor study design with weak endpoints and little statistical rigor6. Furthermore, most research efforts have focused on biomarkers that monitor treatment. In fact, most biomarkers in clinical use are not suitable for population screening or for early diagnosis, observes Diamandis.

Diamandis claims that previous research into cancer biomarkers were looking in the wrong places. Too often efforts have focused on genetic markers, which in terms of cancer “represent 'digital information' — yes or no. This is not true for metabolomic or proteomic biomarkers, which are associated with quantitative changes”, he says. But such biological markers are delicate. “They can be influenced by sample collection and storage methods, benign diseases, and even diet and drugs,” he explains. A difficulty of identifying quantitative biomarkers that are both highly sensitive and highly specific is that data analytical biases are introduced. “It is not surprising that seemingly spectacular data on new biomarkers are subsequently found to be not reproducible, and therefore unsuitable for use in clinical practice,” Diamandis concludes.

George Poste, head of the Complex Adaptive Systems Initiative at Arizona State University in Tempe, agrees that biomarker research is yet to deliver on its promise for these and other reasons. Part of the problem, he says, is that until recently, most investigator-initiated research has been too small and non-uniform to yield meaningful results. A lack of standardization in sample collection and processing, the use of cell lines instead of patient biopsies for research, and an insufficient number of patient samples are reasons for the dearth of meaningful biomarker development. More-over, the field needs much more funding to encourage collaborative research and a 'big science' approach, says Poste. Government and industry funding must step up to the plate, he adds.

Poste cites the US National Cancer Institute's Cancer Human Biobank and the UK's Biobank as successful examples of big science and what it can do when the community invests in this research. He notes that historically the lion's share of cancer funding has gone to treatment, not prevention. “But the real issue is how can we catch cancer very early on, before it spreads,” says Poste. This is the realm of biomarkers. “If we can find cancer in its earliest stages, it might be possible in the future to prevent it.”