To the Editor: We read with great interest the article by Lotan et al1 recently published in this Journal on the subject of TMPRSS2-ERG gene fusions in prostatic ductal adenocarcinoma. In this article, Lotan et al noted the poorly understood role of TMPRSS2-ERG rearrangements in the initiation and progression of prostate carcinoma.

At present, there are large and important differences in the results of multiple articles studying the relationship between ETS fusion status and clinical outcome. Inconsistent results may be due to several factors, including study population, sample size, method of cancer detection, source of tissue, method of gene fusion detection, type of gene fusion sought after, primers used, follow-up, and different clinical end points. Often, this fundamental data are not well delineated in the methods sections, making comparisons even more difficult.

We would like to highlight one major factor that is inadequately addressed in the paper by Lotan et al1: Cohort selection. Many studies have evaluated patients with PSA-screening-detected prostate cancer, who underwent radical prostatectomy.2, 3, 4 Other studies have focused on patients detected incidentally and placed on expectant therapy or watchful waiting.5, 6 In the paper by Lotan et al,1 the population was selected on the basis of a rare subtype of prostate cancer (ie, ductal adenocarcinoma). We and others have observed large frequency differences for ERG-rearranged prostate cancers depending on how they were selected.2, 3, 4, 5, 6, 7 The selection of cases may help us understand these differences in ERG rearrangement frequencies, but to date we cannot offer a clear explanation. It is intriguing to speculate that the low frequency of the TMPRSS2-ERG fusion in incidentally identified prostate cancers by TURP is similar to ductal adenocarcinoma that may have been more often identified for symptomatic reasons. They both share a central localization within the prostate and are not detected by PSA screening.

Another issue related to cohort selection that needs to be better delineated in prostate cancer biomarker studies is clinical end point. The variability of end points in different studies makes results difficult to interpret. Some groups have reported a significant relationship between the presence of ERG rearrangement and worse clinicopathological indicators, whereas others have not found this relationship. Still other groups describe an association with a better outcome. Most studies have used PSA biochemical failure as the study end point. In fact, it affects how a patient is treated and likely his perception of the disease. However, it is important to point out that biochemical recurrence is not a reliable predictor of disease progression or cancer-specific death.8, 9, 10, 11

The confusion caused by mixing study cohorts unfortunately surfaces in this paper. The study by Demichelis et al6 is grouped into the category of PSA biochemical studies showing association with clinical outcome. In fact, this study explored for associations with the development of metastatic disease or cancer-specific death. It represents a population-based cohort from a distinct catchment area of Örebro, Sweden. All cancers were detected incidentally without PSA screening and followed on a Watchful Waiting protocol. Therefore, the study assessed the natural history of fusion prostate cancer. Clearly, the Örebro cohort is distinct from a set of men selected through PSA elevation, at specific medical centers, and treated with radical prostatectomy. The Demichelis study was included among a group of studies that reported a relationship between the presence of the gene fusion and shorter interval to biochemical recurrence. However, in this study a statistically significant association between the gene fusion and prostate cancer death was observed.

The study by Lotan et al,1 also failed to reference another population-based study from the United Kingdom by Attard et al5 that showed consistent finding with that of Demichelis et al.6

It is worth mentioning that in this article the cases were evaluated by FISH for ERG gene rearrangements using break-apart probes for 5′ERG and 3′ERG. This is an indirect test and cannot directly confirm the fusion of TMPRSS2-ERG. Recently, there is an increasing evidence of other 5′ partners.12, 13, 14 Therefore, Lotan et al describe the occurrence of ERG rearrangements in prostatic ductal adenocarcinomas rather than the TMPRSS2-ERG fusion gene itself.

In conclusion, we like to point out that taking into account population selection, treatment, and follow-up is key for appropriate comparison among studies. Comparisons should be made between similar cohorts and end points.