In April 2010, an international team of researchers from academia and drug company GlaxoSmithKline reported that dutasteride, a drug already approved for the treatment of benign prostatic hyperplasia, reduced the chances that men considered at high risk for prostate cancer would develop the disease. The four-year trial included more than 8,100 men and met the gold standard for clinical trials: it was randomized, double-blind, and placebo-controlled; it studied parallel groups at multiple medical centres; and it assessed outcomes with biopsies at two years and four years. In the end, men who took dutasteride were 23% less likely to have a positive biopsy for cancer than those on the placebo. GSK submitted this data in its application to the US Food and Drug Administration to market the drug for prostate cancer prevention — this January, the FDA said No.

Although it is not unusual for the FDA to reject a drug application supported by apparently positive data, this case illustrates the particular challenges surrounding clinical trials for cancer prevention. When the aim is to decrease the incidence of cancer in large populations, studies on preventive agents require large patient cohorts — sometimes approaching 20,000 participants — and take years or even decades to perform. This combination makes them especially unwieldy compared to tests with therapeutic compounds, which can much more quickly be seen to work, or not, by testing them exclusively in people who have the disease. In cancer prevention drug trials, the usual gold standard barely rates a bronze.

Credit: JAHI CHIKWENDIU/THE WASHINGTON POST VIA GETTY IMAGES

Since preventives are intended for apparently healthy patients, trials require a high confidence that the anticipated anticancer benefit will outweigh any harmful side effects. In the dutasteride trial, statistical analysis showed that the decrease in cancer was driven mainly by a reduction in less serious tumours that might not even require treatment. In addition, men who took the drug were slightly more likely than those on a placebo to develop more aggressive tumours. The FDA's expert advisory panel concluded that the prevention benefits failed to outweigh this risk.

Researchers say two things are needed to decrease the length and size of prevention studies. One is to identify high-risk populations to be the preferred subjects for the trials. The second is surrogate endpoints that can provide evidence of whether a preventive drug is working — and do this in just a few years, rather than decades. The key to both is finding better biomarkers — the genes, proteins, and cellular metabolites that can be measured and associated with the development of cancer.

Patterns of these biomarkers that can be uniquely linked with one type of cancer can make it easier to estimate an individual's cancer risk. Selecting highest-risk patients for studies increases the statistical power of trials with a smaller number of participants. As a second benefit, high-risk cohorts can also shorten trials. If epidemiological studies show, for example, that a known percentage of patients carrying a certain gene will develop cancer within five years, researchers can restrict a prevention trial to those patients and run it for just that duration. Moreover, patients and regulators are likely to be more tolerant of side effects if the targeted users have a high chance of developing cancer without intervention.

The designers of the dutasteride trial did select participants judged to be at higher risk of developing prostate cancer. However, they did so by looking for elevated levels of prostate-specific antigen (PSA) — a protein whose utility as a biomarker for prostate cancer is a matter of debate. If a fully validated biomarker for prostate cancer had existed, GSK might have been able to design a dutasteride trial that required fewer participants and could have yielded a more definitive outcome. In particular, looking at the drug's effect (or lack thereof) on the biomarker may have clarified whether the increase in detected higher-grade cancers was due to the drug or simply an artefact of the tumours becoming more easily detected owing to dutasteride's shrinking of the prostate.

Some biomarkers may even function as the surrogate endpoints needed to shorten prevention trials. If, say, a specific group of proteins reliably increases in the blood of patients during the earliest, precancerous stages of disease, doctors could monitor those proteins rather than relying on biopsies to detect malignancy. Molecular biomarkers of potential toxicity, such as the activity of drug-metabolizing enzymes, could also help researchers monitor subjects' safety and response to drug candidates in clinical trials.

Scott Lippman, an oncologist and cancer prevention researcher at the University of Texas MD Anderson Cancer Center, has proposed fully integrating biomarkers chemoprevention development. After evaluating biomarkers in animal models, researchers would do epidemiologic studies linking the biomarkers to human cancers. They would next model the likelihood that patients with specific biomarkers will develop cancer. Then, in a 'phase 0' step between preclinical and phase I clinical trials, researchers could test sub-therapeutic doses to assess a drug's behaviour in healthy patients without risking harm. Lippman argues that this approach could yield better decisions on whether to undertake a lengthy, and costly phase III trial — and speed the development of preventive agents. Indeed, the fact that GlaxoSmithKline skipped some of these steps might have played a role in the FDA's decision on dutasteride. The drug inhibits the enzyme that converts testosterone to the more potent 5α-dihydrotestosterone. But neither molecule is yet a validated biomarker for prostate cancer.