|F. Labrie, MD, PhD, Laval University, Quebec City, Canada, described a dramatic decrease in prostate cancer deaths by screening 8,137 men. First visit included prostate-specific antigen (PSA) test and digital rectal examination (DRE). An abnormal PSA or positive DRE signalled use of transurethral echography (TRUS) and if needed, prostate biopsy for diagnosis of the clinically localized cancer. Compared with 38,056 nonscreened men, the reported prostate cancer deaths decreased from 48.7 per 100,000 men-years (control) to 15.0 per 100,000 (screened and treated). Remarks by the invited discussant point to a need for other statistical analyses.|
|Screening methodology|| Labrie and coworkers began the screening program in 1988, well before the
potential was apparent. The original sample consisted of male voters from
metropolitan Quebec City who registered in 1985. This list yielded 46,193
men between 45 and 80 years and available for screening.
Screening invitations were mailed by random selection to 30,956 (67%) men. An additional 992 men invited themselves and participated in the screening program. The remaining men served as nonscreened controls. Quebec Province records provided death certificate information for prostate cancer deaths covering Jan 1, 1989 through Dec 31, 1996.
A synopsis on subject assignment and compliance follows:
Prostate cancer screening consisted of:
|Results of screening|| This 8-year study showed only 5 prostate cancer deaths in the
screened/treated group and 137 in the control group (p&60;0.01). Death
rates expressed per 100,000 men-years were:
* significant (p=0.02)
** highly significant (p<0.01)
|Presenter recommendations||Labrie advocates screening men with a strong family history of prostate cancer and African Americans starting at age 40 and others at 50. Use of 4 ng/mL PSA as the upper normal limit would have led to missing 13% of prostate cancers at first visit and 21% at follow-up visits. Therefore, the acceptable PSA level needs to be set at 3 ng/mL.|
|Some remarks of invited discussant||
P. Boyle, MD, European Institute of Oncology, Milan, Italy, saluted
the investigators for undertaking the first randomized, prospective trial
to examine PSA in prostate cancer screening. Recent developments in
statistical methodology as related to disease screening merit review
The discussant recommended using the statistical model developed by Cuzick et al (Stat Med 1997;16:1017) to adjust for noncompliance and contamination in randomized trials. Because the present study encountered a 23% compliance rate in the invited group, it was susceptible to selection bias. In Boyle's opinion, the data need analysis. (Labrie et al used the Cuzick model and found a 6% decrease in prostate cancer mortality after comparing invited and uninvited men on an intent-to-screen basis. Cancer specific deaths decreased between 54% and 100% after adjusting for noncompliance and contamination, respectively. This alternative statistical approach confirmed the originally reported results on screening and early treatment).
Another short-coming relates to the death certificate. Current research teams include a death committee for assigning causality to such primary outcome data. Boyle finds the report of Nystrom et al (J Med Screen 1996;3:85) applicable for controlling possible bias in examining death rate data. (Labrie considers the Quebec Province records to be a reliable cancer death registry for the invited and uninvited men followed in this prospective trial).
Presented at the American Society of Clinical Oncology (ASCO) Meeting,
May 16-19, 1998
Copyright © 1998 Conrad Group Inc. All Rights Reserved
Eugene A. Conrad, PhD, MPH / ISSN 1078-2230 / June 1998
Send Comments to: ConradNote@aol.com