QUEBEC CITY--For the first time, a randomized clinical trial of
screening with prostate-specific antigen (PSA) has been performed and
shown to reduce the risk of prostate cancer mortality.
The large Canadian study showed a 69% reduction in mortality in men
who participated in screening, as opposed to those who did not. All
but one prostate cancer death in the screened group arose from a
tumor detected at the initial screening visit.
"This study demonstrates the feasibility and efficacy of early
diagnosis and early treatment of prostate cancer," Fernand
Labrie, MD, said at the plenary session of the ASCO meeting. "If
we take the data as they are, potentially 27,000 of the 39,000 annual
prostate cancer deaths in the United States could be prevented by
early diagnosis and treatment."
Whether physicians or the public "take the data as they are"
remains to be seen. At the plenary session, formal discussion of the
results focused on possible methodologic flaws in the study.
In particular, questions were raised about the potential difference
in risk of death from prostate cancer between those men who accepted
PSA screening and those who did not accept the invitation to be screened.
The study results also raised questions about the use of 4 ng/ml as
the upper limit of normal for PSA. Investigators in the trial used 3
ng/ml, which increased the cancer detection rate at the initial and
follow-up visits, said Dr. Labrie, director of research at the Laval
University Medical Center, Quebec City.
The trial involved more than 46,000 men ages 45 to 80 who were
enrolled and followed between January 1, 1989, and December 31, 1996.
Dr. Labrie noted that the study began at a time when PSA screening
had little support as a means to prevent prostate cancer death.
31,000 Invited to Be Screened
Of the total patient population, almost 31,000 men were contacted and
formally invited to be screened for prostate cancer. Ultimately,
23,800 declined the invitation but were included in follow-up with
another 15,237 men who were not invited to screening.
In the screened group, the initial evaluation included PSA and
digital rectal exam (DRE) in all men. Patients who had abnormal PSA
levels or abnormal DRE were also evaluated by transrectal ultrasound.
PSA screening and DRE were repeated during early follow-up visits,
but DRE was subsequently dropped from later follow-up examinations.
Patients who were not screened had no regularly scheduled clinic
visits as part of the study but were followed by usual medical practice.
By the end of 1996, five patients in the screened population had died
of prostate cancer vs 137 in the control group. Expressed in terms
that account for follow-up years, the screened group had a prostate
cancer mortality of 15 per 100,000 man-years, compared with 48.7 per
100,000 in the control group.
The investigators separated the patients who declined invitations to
screening from those who were never invited, the latter group
representing what Dr. Labrie called "the true control group."
The uninvited cohort had a prostate cancer mortality of 41.6 per
100,000 man-years, compared with 53 per 100,000 among the
invited-declined group. The difference between the two groups was not
Four of the five prostate cancer deaths in the study group were
associated with cancers detected at the initial screening visit. All
four of these cancers were metastatic, a finding that Dr. Labrie
interpreted as confirming the effectiveness of current therapy for
early-stage prostate cancer. Overall, 367 cancers were detected in
the screened group, and none were metastatic when detected after the
"If we start screening men at age 50, there will be very few
cancers that will not be localized; metastatic prostate cancer almost
disappears," he said.
An economic analysis of the findings revealed an estimated cost of
about $3,000 for diagnosis of a single case of prostate cancer at
initial visit. Total cost per each cancer diagnosed at any visit was
about $10,000, a figure that compares favorably with costs associated
with the diagnosis of breast and cervical cancers, Dr. Labrie said.
The results also show that use of 3 ng/ml as the cutoff for normal
PSA has a substantial impact on cancer detection. Dr. Labrie said
that a cutoff of 4 ng/ml as the upper limit of normal would have
missed 13% of cancers at the initial visit and 21% of cancers at
follow-up visits. "It also became clear that only PSA screening
should be done at follow-up visits, since approximately 5,000 DREs
are required to find one case of prostate cancer if the PSA level is
below 3 ng/ml," Dr. Labrie said.
In a formal discussion of the results, Peter Boyle, MD, focused on
the investigators decision to combine the uninvited cohort with
men who were invited to screening but declined. He argued that the
invited-declined group should have been included with men who were screened.
"A randomized clinical screening trial should always be analyzed
on an intention-to-screen basis," said Dr. Boyle, director of
epidemiology, European Institute of Oncology, Milan, Italy.
Dr. Boyle calculated the death rates from prostate cancer on an
intention-to-screen basis, looking at the ratio of death rates in
those men who were invited to be screened compared with those men who
were not invited to be screened.
With this analysis, the relative risk of prostate cancer death was
about 1.1 to 1.16. "This is in tremendous contrast to the 70%
reduction in death from screening (relative risk, 3.25) which you get
when you look at the particular groups as Dr. La-brie has shown,"
However, Dr. Labrie said that the intent-to-screen analysis performed
on actual data, even including prevalent cases at time of study
initiation, show a relative risk of prostate cancer death of 0.94, a
When accounting for the compliance rate and years of follow-up, he
said, this corresponds to a 63% decrease in prostate cancer death in
men actually screened. "This is a figure that confirms the
previous analysis and practically eliminates the possibility of
selection bias," Dr. Labrie said.
Dr. Boyle, however, said that although the "preliminary findings
from this randomized study are of great importance, I think that
there needs to be much further analysis of this dataset before final
conclusions can be drawn about the study results."
Dr. Boyle cited the large compliance problem in the study: Only about
23% of those men invited to be screened actually complied. "So
it is understandable that the authors wish to adjust for this in some
way," he said. "Unfortunately, the method presented to take
this into account is, I think, strongly susceptible to the
possibility of selection bias."
Dr. Boyle also would have liked to have seen information about the
total number of prostate cancers detected. "We would like to see
increases in prostate cancer incidence along with the decline in
mortality," he said.
Finally, Dr. Boyle cited the lack of a "death committee"
for the study to allocate deaths to prostate cancer or other causes,
rather than relying solely on the death registry for Quebec.
New Statistical Methods
Nonetheless, Dr. Boyle believes that many of the issues raised by the
data could be resolved using advanced statistical epidemiology methods.
"There are statistical methodologies available to get a better
understanding of the impact of compliance and contamination and the
extent of the problem of selection bias and ascertainment bias,"
Dr. Boyle congratulated Dr.Labrie and his group on having the
"insight and foresight" to undertake the first randomized
trial of prostate screening with PSA, but he emphasized that "no
issue, and certainly not an issue as complex as prostate cancer
screening, can be resolved by the results from one individual trial."