Since the prostate-specific antigen (PSA) screening test was introduced into clinical practice in the early 1990s, detection of prostate cancer has greatly increased and patient management has improved. However, despite the sensitivity of this test, PSA lacks specificity as a serum marker for prostate cancer. Consequently, the need for discovery of more specific biomarkers to improve the early detection of the disease is paramount. Recent technological advances have made this possible, and, in the 1 July issue of Cancer Research, George L. Wright and co-workers describe how they have used a serum-protein fingerprinting technique to help accurately distinguish between prostate cancer, benign prostate hyperplasia (BPH) and healthy tissue.

The authors used a protein-biochip surface-enhanced laser-desorption/ionization mass spectrometry (SELDI) approach to detect proteins that were affinity-bound to a protein-chip array. They then used an artificial intelligence learning algorithm to reduce the number of proteins found down to the number that are required to differentiate prostate cancer from noncancer cohorts. Serum samples were taken from 167 patients with prostate cancer, 77 patients with BPH and 82 unaffected healthy men. SELDI detected 779 protein peaks after clustering and peak alignment, which were narrowed down to nine protein masses necessary for classification.

So, how specific is this classification method? The overall sensitivity was 83%, the specificity was 97% and the positive predictive value was 96% for differentiating prostate cancer from BPH and unaffected healthy men. This compares very favourably with the PSA test, in which sensitivity is >90% but specificity is only 25%. It follows that use of the serum-fingerprinting method will lead to a substantial reduction in unnecessary biopsies — something which causes considerable anxiety in men who, if classified accurately, would not have needed a biopsy.

Other advantages of this screening technique include earlier detection of prostate cancer — the authors' experience suggests that prostate cancer might be suspected 5 or more years earlier with this fingerprinting approach than with PSA screening. Because of tumour microheterogeneity, it is perhaps not surprising that the use of multiple biomarkers is likely to be more effective than use of a single marker. The next step is to identify other biomarkers that can differentiate aggressive cancers from nonaggressive cancers, to make this classification system for early detection as effective as possible.