27 May 2009
Measurement of blood prostate specific antigen (PSA) is currently the most widespread screening method for prostate cancer. However, the benefits of PSA screening still remain unclear despite the publication of two large-scale prostate cancer screening trials (see previous news), as the benefits of testing may be outweighed by the harms of potential misdiagnosis and unnecessary intervention.
Recent linkage and genome-wide association studies have helped to elucidate the genetic basis of this common complex disease, with numerous genetic variants identified as being associated with the condition, fuelling the hope that genetic testing may provide better risk-prediction for prostate cancer. Despite modest risk estimates from individual SNPs, Zheng and colleagues have previously suggested that combining five SNPs with family history may be useful for predicting prostate cancer (Zheng et al. (2008) NEJM 358(9):910 and see previous news). But how does this genetic test for prostate cancer perform and does it have any clinical utility?
Recent work by Li-Wan-Po and colleagues [Li-Wan-Po et al. (2009) Public Health Genomics DOI:10.1159/00218710] attempts to answer this question. Using the cumulative association of five genetic risk factors (SNPs) with family history, Li-Wan-Po et al. calculate the positive and negative predictive values in order to determine how good a predictor this test would be. Despite the positive predictive value increasing with the number of factors included, the test performs poorly even when five or more factors were included, with only 44% of positive tests accurately predicting prostate cancer. They also highlight that although the odds ratios increase as the number of SNPs included increases, the test performance actually decreases because the prevalence of the combined genotypes is low. The highest risk class – with nearly 4.5-fold higher risk of developing prostate cancer relative to the population average – represents less than 1% of the population, making it inappropriate for screening.
Zheng et al. have also previously shown that genotypes at the five genetic variants do not correlate with disease state, Gleason score, or PSA level at diagnosis, suggesting that this genetic test cannot identify those men at a higher risk of a more clinically aggressive disease outcome. When combined with the fact that the genetic test does not even predict prostate cancer as well a single PSA test between the age of 44 and 50 (which is not currently recommended by the UK National Screening Committee), the performance of the genetic test is clearly woefully inadequate for either clinical or public health purposes. Furthermore, combining PSA testing with the genetic test only leads to a marginal, though statistically significant, improvement in test performance [Sun et al. (2008) Prostate 68: 1257-1262].
Comment: Li-Wan-Po et al. are correct in pointing out that “the benefit of screening for prostate cancer is still uncertain with PSA measurement” and “that the proposed 5-genetic-variant test for risk estimation of prostate cancer is inadequate”. Even with the addition of further genetic variants to the PSA test, only a small improvement in test performance is gained. Analyses currently suggest that adding these SNPs to existing measured clinical factors (such as age, Gleason score and serum PSA level) provides no additional benefit in identifying men that are at a higher risk of a more clinically aggressive form of prostate cancer. The technical performance of this test is inadequate and does not help to address any of the harm issues currently muddying the waters with PSA screening. Although results are promising, and susceptibility tests for prostate cancer based on these genetic variants are already available direct-to-consumer, there is still a long way to go before the test can be considered “ready for prime-time”. This study provides a useful case-example of the some of the issues that need to be considered in the development of genetic screening tests and their suitability for wider clinical use.