PSA screening for prostate cancer

When it comes to men’s health, one could argue that the prostate-specific antigen (PSA) test to screen for prostate cancer is the most critical test a man can have. It is also the most controversial. That’s because many experts believe prostate cancer is the exception to the rule that early detection of cancer saves lives. In fact, PSA screening may actually result in more harm than good.

Most men in the United States over age 50 get the PSA screening test, which was approved by the FDA in 1994. Many men have the test repeatedly. That’s probably no surprise, given that Americans value the early diagnosis of cancer along with the prompt and often aggressive treatments that follow. And for most types of cancer, that plan of attack makes sense.

But the PSA test can’t differentiate between slow-growing, harmless prostate cancers and the less common aggressive, potentially deadly tumors. In fact, the test doesn’t even diagnose cancer. Rather, depending on the numerical score, it may prompt a biopsy, which is the only way to detect cancer. If pathologists see cancer cells in the tissue sample, they try to estimate how aggressive the cancer is based on its appearance under the microscope.

The upside of PSA screening is that early diagnosis of aggressive prostate cancers can improve survival rates — tumors are caught before they significantly threaten survival. But when screening uncovers cancers that would never cause symptoms or harm during a man’s lifetime, it results in the major downsides of PSA screening: overdiagnosis and overtreatment.

That’s because a cancer diagnosis usually leads to treatment, and all prostate cancer treatments carry a substantial risk of side effects, including erectile dysfunction and urinary incontinence. As a result, though diagnosing aggressive cancers can be lifesaving, diagnosing harmless cancers may do more damage than good.

Two eagerly awaited studies — one conducted in the United States and the other in Europe, the results of which were published in The New England Journal of Medicine in March 2009 — were supposed to settle the debate over the value of the PSA test. (See “PSA screening and mortality” below.) Does PSA screening save lives by allowing doctors to treat aggressive cancers early? Or does it harm men who would never die from the disease by subjecting them to the side effects of surgery, radiation, or hormone therapy? Might a small subset of men benefit from screening? How often should screening be done? While each study offers some answers, the only definitive conclusion that seems to have come to the fore is that there is no “right” answer for everyone.

The American study

The Prostate, Lung, Colorectal, and Ovarian (PLCO) Cancer Screening Trial began in 1993. Over the next eight years, 76,693 men between the ages of 55 and 74 with no history of prostate, lung, or colorectal cancers volunteered for the study, which took place at 10 medical centers in the United States. Investigators randomly assigned half of the men to be offered annual PSA testing for six years, along with an annual digital rectal examination (DRE) for four years. The PSA and DRE results of men in the screening group were reported to them and their physician. A PSA level above 4.0 ng/ml was reported as abnormal and usually resulted in a prostate biopsy. Men in the control group continued to receive their usual medical care, which may have involved some prostate cancer screening or none at all. In both groups, men who were diagnosed with prostate cancer were treated by their personal physicians. PLCO researchers found that men in the two groups had similar treatments.

After seven years, the researchers found a relative increase of 22% in the rate of prostate cancer diagnosis among the men offered regular PSA screening compared with those who received their usual medical care. Even though PSA screening increased the diagnosis of prostate cancer, it did not improve the prostate cancer survival rate, and there were no real differences in the numbers of deaths from other causes between the two groups. For about two-thirds of the men, complete follow-up data were available for 10 years, and the results after 10 years were similar to the findings after seven years.

The PLCO trial will continue until all the volunteers have been evaluated for 13 years. Researchers are compiling information on the side effects of treatment and participants’ perceptions of their quality of life. They will also track additional deaths among the study participants.

The European study

Like the American study, the European Randomized Study of Screening for Prostate Cancer (ERSPC) began in the early 1990s. Investigators identified 182,000 men between the ages of 50 and 74 through population registries in seven countries. However, because some countries only enrolled men between ages 55 and 69, the ERSPC focused mainly on this core group, which totaled 162,243 men. Investigators then randomly assigned half of the men to be offered PSA screening once every four years, on average; those with a PSA level of 3 or 4 ng/ml or higher were referred for a prostate biopsy. The other half — the control group — had their usual medical care. Men who were diagnosed with prostate cancer were treated by their own physicians according to local guidelines.

After about nine years of observation, 214 men in the PSA screening group and 326 men in the control group had died from prostate cancer. The relative risk of dying from prostate cancer was reduced by 20% in the screening group and by 27% among those actually screened. What does that mean to the typical American man? In the United States, the lifetime risk of dying from prostate cancer is nearly 3 in 100, or 3%. A 27% reduction in the relative risk of dying from prostate cancer means that, with screening, the risk would drop to 2.19 in 100, or 2.19%.

However, that modest benefit came at a steep price: The researchers calculated that 48 men who are not at risk of dying from prostate cancer would have to be treated for screening to prevent one death from the disease over nine years. In other words, 48 men would risk the side effects of treatment to save one life.

As in the PLCO trial, ERSPC investigators will continue to track prostate cancer deaths and monitor treatment side effects and quality of life for the men.

Despite these major new studies, PSA testing remains a personal decision — one that can have dramatic ramifications. To help readers sort out the risks and benefits of PSA screening and learn more about the two studies, we interviewed Dr. Gerald Andriole, chairman of the Department of Urology at the Washington University School of Medicine in St. Louis and principal investigator of the PLCO trial, and Dr. Fritz Schröder, from the Department of Urology at the Erasmus Medical Center in the Netherlands and principal investigator of the ERSPC.

Dr. Andriole talks about the PLCO trial

We weren’t expecting any results from the PLCO trial to be published for a few years. Why did you decide to publish data now?

The PSA test was increasingly used in the late 1980s and early 1990s, and it received FDA approval for prostate cancer screening in 1994. Around that time, a drop in prostate cancer mortality occurred fairly quickly, and there have always been questions about whether that was due to screening or more aggressive and improved treatment. The early results of our study strongly suggested that the drop in prostate cancer mortality was more likely due to treatment than screening. That was the impetus to publish. The data and safety monitoring board didn’t mandate that we publish, but they did indicate that it could really be of benefit to people who are thinking about why we have seen such a reduction in prostate cancer mortality.

Also, in 2008, the U.S. Preventive Services Task Force recommended that doctors not screen men for prostate cancer after the age of 75. It was felt that the results from the PLCO trial supported that recommendation — or at least supported stopping screening in men with a limited life expectancy.

So you’re saying that the drop in the prostate cancer death rate is probably not due to PSA screening, but to treatment.

I think that more men with prostate cancer started getting treated — with surgery, radiation, or hormone therapy — and that was effective in prolonging their survival and increasing their chances of dying of something else. So I think it’s the combination of more treatment — and more aggressive treatment — that explains the drop in prostate cancer mortality in the 1990s.

What were the main operational and organizational similarities and differences between the ERSPC and the PLCO trial? How do you reconcile the differences?

In the PLCO trial, we screened with annual PSA tests for six years, and we reported results to the patient and his doctors. We did not mandate any evaluation after screening, and we did not mandate any treatment, even after a biopsy. In the European study, they had an explicit follow-up procedure after an abnormal PSA screening, though that procedure varied a little bit from country to country. Most of the European centers used a PSA of 3 ng/ml or higher as an indication for a biopsy, but that varied, and some centers prescribed ancillary tests, such as a DRE and transrectal ultrasound, at certain PSA levels.

With the ERSPC, I have a feeling that men who were in the screening group might have gotten slightly more aggressive treatment for prostate cancer than those in the control group because screening tended to happen at large medical centers. That’s where the men would get their biopsies, too, and if they had cancer, where they tended to be treated. But men who were randomly assigned to the control group were probably more likely to stay in their own community or small town for medical care, and it may have taken more time for them to get a biopsy and treatment. If you look at Appendix 7 of the ERSPC paper in The New England Journal of Medicine, you’ll see that higher percentages of men in the screening arm had surgery [33.7%] or radiation therapy [20%] than in the control arm [20.7% and 12.3%, respectively], though for quite a large number of the men, we don’t know what the treatment was [12.5% in the screening group and 26.9% in the control group]. Of course, you have to adjust for the fact that there were differences between the two arms in terms of the cancers’ stage at diagnosis, but it could be that some of the difference we see in mortality between the groups in the European study is due to these treatment differences rather than just screening differences.

In the PLCO trial, what happened when a man had a PSA test? How did you capture the information you needed?

We would mail a letter to the man and his physician that merely reported the PSA level or the DRE result with no specific recommendation about follow-up, though we did say that we considered a PSA level over 4 ng/ml to be abnormal and advised diagnostic evaluation. If a man had an elevated PSA, generally the first thing that his doctor did was repeat the test, and in some of those men, the PSA level was actually lower or had normalized. So if a man’s PSA level was slightly over 4 and then went down below 4 ng/ml, he would probably be monitored without treatment. But if a man in the PLCO trial was younger, had a family history of prostate cancer, or had a high PSA density, he was more likely to have had a biopsy.

We did a study analyzing who had a biopsy in the screening arm of the study, and we concluded that the biopsy rate following a positive PSA screening test in the PLCO trial was about the same as the national rate. It’s true that it took about two to three years for most of the men to have a biopsy following an abnormal PSA screening test, but that’s real life. Many of them have other health issues that need more urgent attention.

In addition to obtaining all of the records related to the follow-up of men with abnormal PSA tests, we periodically assess participants’ compliance in getting screening, look for changes in tumor grade and Gleason score over time, and calculate the overall prevalence of cancer. (For more details about follow-up and compliance with the PSA screening regimen, see “PLCO participant follow-up” below.)

So it sounds as if the PLCO trial was designed to mimic what happens in terms of follow-up in a real medical practice.

Exactly. We have a screening site here in St. Louis. We enrolled about 17,000 patients, and quite a number of them ricochet back and forth from their family doctor to our center for a biopsy and treatment, but not all of them. There are patients who came here for screening from two to three hundred miles away. Generally, they saw a urologist in their community to follow up.

Have you done any kind of analysis or comparison of patients who are treated in the community versus those who are treated in large academic medical centers? That would lend credence to your observation about treatment differences in the ERSPC.

We have not done that. I don’t know if doing such an analysis would show anything. But, in the PLCO, we have analyzed who received treatment, the type of treatment, at what age, and what other medical conditions they had, and we could find no difference between the two arms of the study.

Was there any difference in Gleason scores between the screening and control arms of the study?

Yes. As you would expect, the Gleason scores tended to be lower for cancers discovered in the screening arm. But keep in mind that the Gleason scores were not assigned or reviewed at one central facility, so there may have been some variability in determining the scores. We are in the process of trying to analyze Gleason scores centrally, and we’ve collected about 3,000 radical prostatectomy samples to review.

You’re not reviewing biopsy samples?

Not at the moment. We have a big tissue repository associated with the PLCO trial, and a laboratory at UCLA was making slides with tissue from a number of cancers, including prostate cancers treated with radical prostatectomy, and we asked them to make extra slides for us so that we could do this central Gleason score review.

Was there any consistency in the number of tissue samples, or cores, taken during biopsies or in how the biopsies were done?

We do not know, but suspect there is no significant difference between men in the screening and control groups. While some critics have said that we should not leave any of these issues to chance or a given practitioner’s preference, in reality, patients and practitioners are different. As a urologist, my main concern may be a man’s elevated PSA. For his family doctor, the No. 1 priority might be the man’s diabetes and high cholesterol, so the elevated PSA gets put on the back burner. He might watch the PSA to see if it goes up, and he might do a biopsy at a later time, but that might not be his primary concern at the moment. It may not be consistent, but it’s realistic. And according to the analyses we’ve done thus far, the biopsy practices in the PLCO trial seem to reflect good medical judgment.

Statistically speaking, what was the PLCO trial originally designed to show?

A 20% reduction in prostate cancer mortality in the screening arm versus the “usual treatment” arm, the same as with the ERSPC. Our original enrollment goal was 75,000 men, and we accrued a bit more than that.

Another important thing to emphasize is that the men in the study who were screened for prostate cancer were also screened for lung and colorectal cancers. We didn’t have three separate cohorts. We probably could have enrolled fewer patients to assess screening for lung and colorectal cancers, but we needed 75,000 men to have a study that would be strong enough to find a difference in prostate cancer mortality with screening, if one existed.

What about cause of death? Was that centrally reviewed or determined by individual physicians?

There was a central committee that reviewed clinical and autopsy data. Over all, about 15% of the men in the study have died, and if they had prostate, lung, colorectal, or any metastatic cancer, the committee reviewed the data to confirm the cause of death. Committee members are “blinded,” so they don’t know who had a radical prostatectomy, for example, or who was in the screening arm or the control arm. After seven years of follow-up, only 94 deaths have been attributed to prostate cancer, with 50 deaths in the screening group and 44 deaths in the control group.

But with only 94 deaths, if a few of them were reclassified, it could change the study’s results.

That’s unlikely. There were only small differences between what the death certificate said and the death committee’s review. We’re going to continue to follow men who have not died, look at the age of the men who are dying and any other medical conditions they have. There could be subsets of men who will benefit from screening. Remember, in the European study, only men in specific age groups benefited from screening.

Let’s talk about the contamination rate, which is the percentage of participants in the control group who shouldn’t have had a PSA test or a DRE but did.

Critics could say that the level of screening in the control group could be substantial enough to water down any modest effects of screening in the screening group. In our control group, the rate of PSA testing was 40% in the first year, and it increased to 52% by the sixth year. But even with that magnitude of contamination, the trial is still adequately powered to find a 20% or greater drop in prostate cancer mortality due to screening, if one exists.

What about screening before the study started?

About the same number of men in each group had had a PSA test in the three years before the trial began — 34.6% in the screening group and 34.3% in the control group.

Did any of them have a biopsy before the study started?

About 4% to 5% of the men had actually had a biopsy prior to entering the PLCO trial and being randomly assigned to the screening or control group. That’s not a huge number, but in retrospect, we probably should have excluded them from the study. When you add that to the amount of screening prior to the start of the trial, it could take longer for this trial to show a difference in mortality with screening.

Are the participants still being screened?

No. But we contact each participant yearly and do an annual health survey. We want to follow each person for a minimum of 13 years.

What did you think of the media coverage of the studies?

It’s been a confusing story. You have one positive study and another one that’s not positive, so what do you tell patients? I would say that these are early mortality results. The PLCO trial does not yet tell me whether younger patients will or won’t benefit from screening. We need to wait longer to find out. But in my mind, the converse is true: if you are elderly or you have other medical conditions and your life expectancy is only about 10 years, PSA screening probably isn’t necessary, regardless of your actual age.

I really liked the comment in the ERSPC study paper that 48 men would need to be treated to prevent one death from prostate cancer. That was eye-opening, and it’s why we need to be smarter about what we’re doing and who gets treated. Once a cancer is diagnosed, there’s a real tendency among patients and doctors to treat it.

We need to find ways to determine which cancers are potentially lethal and need treatment and which ones can be left alone.

That’s right. Looking at cancer under the microscope and predicting how aggressively it will behave seems like 100-year-old technology. We’ve got to do better. But keep in mind that the study was designed in the early 1990s, and we didn’t know as much then as we know now about the implications of various PSA levels, PSA velocity, and other factors.

You’ve spent a substantial amount of your career on this study. It must be disheartening to hear people criticize it and say it was poorly planned.

I think some people have criticized it because it doesn’t show what they want it to show. But it’s not over yet. In time, it may well show that screening saves some lives. And even when a trial has a negative result, we can learn from it. With the PLCO trial, we want to look at when it might make sense to pursue PSA screening — when it should be started and when it should be stopped.

Have you changed the way you practice as a result of either of these studies?

I feel much more confident in stopping screening of men who probably won’t live another 10 years. I’m also much more inclined to suggest active surveillance, to watch men with prostate cancer and do another biopsy a year or two later to see whether they have an aggressive cancer.

Any parting comments?

What’s sobering is that even if PSA screening reduces prostate cancer mortality, the cost of screening is extraordinarily high in financial and human terms. We could save money by stopping screening at a certain age or when life expectancy hits a certain point. We could reduce the economic and human toll if we screen more thoughtfully and, when we find cancers, treat them a little less aggressively, if at all.

Dr. Schröder talks about the ERSPC

When we interviewed you several months ago,* you didn’t think you’d have results until 2010 or 2011. What happened that allowed publication to move forward so quickly?

Right after we spoke, our data monitoring committee said that we had a statistically significant difference in prostate cancer mortality between the control and screening arms, so we had to stop recruiting volunteers but continue follow-up. And according to our original plan, that meant we were to publish the findings to date.

You reported study findings on men who were between the ages of 55 and 69 when they entered the study. What about men in the study who were outside that age range?

That age range was set in our protocol when we forged the European collaboration including all of the centers in 1994. Some of the centers also gathered data on men ages 50 to 54 and 70 to 74; data on these subsets are included in our report but were not tested for statistical significance.

Did you have any discrepancies about cause of death in your study?

There is total unanimity about the causes of death among study participants because all of the centers used the same algorithm, which was published, to determine cause of death throughout the study. [See “Determining cause of death” above.] We looked only at prostate cancer information and whether or not the man was alive; committees looking at cause of death were blinded with respect to the arm of the study the person was in.

What did you think of the press coverage of your study and the PLCO trial?

I thought it was a pity that the two studies were looked at in the same way. There’s one poorer study and one better study. We weren’t scheduled to have sufficient follow-up to find a difference until we had analyzed all of the data through December 2008, but we found it two years early. So, The New England Journal of Medicine put a lot of pressure on us to explain exactly what the power calculations of our study were. We were very surprised to see that statistical power isn’t even mentioned in the PLCO trial. Estimates of the power of that study are so low that we’re unlikely to see any significant differences in future analyses.* The main reason for that is that the difference in the proportion of men undergoing screening in the two arms of the PLCO trial is only 33%: 85% of the men in the screening group were actually screened, and 52% of those in the control group were screened. The PLCO trial should’ve excluded previously screened men from the study to improve the power of the study.

The study noted that you’d have to screen 1,410 men and treat 48 to save one life in nine years, prompting many men to ask whether screening is worthwhile. What do you think? Should medical practice be altered, given these findings?

Absolutely, yes. Until now, we only knew the disadvantages of screening; we didn’t know the potential advantages. We now know that PSA screening reduces the risk of death from prostate cancer by 20% — and by 27% in men who are actually tested.* But that message has to be balanced with the enormous amount of overdiagnosis and overtreatment — diagnosing and treating prostate cancers that aren’t likely to cause health problems during a patient’s lifetime. People who are overdiagnosed and overtreated suffer from the side effects and the costs and efforts involved in getting treated without getting any of the benefits.

One of the major advances of our study is that we can offer patients some information about how many men need to be treated to save a life. With breast cancer, the number of patients that need to be treated to save one life is around 10. For prostate cancer, due to the enormous amount of overdiagnosis, it is 48 over nine years. That’s a number that’s difficult to accept.

But our report did not specifically outline the factors that determine the number needed to treat. So, for example, if we didn’t do biopsies on the men who had a PSA between 3 and 4 ng/ml, which was about 8% of the men, we wouldn’t have detected as many cancers. That would bring down the number needed to treat considerably — and probably make it more acceptable.

What do you tell a patient who has had radiation or a radical prostatectomy and asks you, “Do you think I really needed to be treated?”

I think patients are justified in asking that question. From now on, once the biopsy is done, I would use our risk indicator or another nomogram to help decide whether a prostate cancer needs treatment; we did a study that showed 49% of men diagnosed with prostate cancer through screening have indolent disease. [See “Indolent prostate cancer” above and “How aggressive is my cancer?” below.] Identifying indolent cancers won’t decrease the number of cancers that are diagnosed, but it will decrease the number requiring active treatment. That’s a good beginning.

If someone asked me that question retrospectively, I would go back to his clinical findings, log on to the risk indicator on my laptop, and fill in the five key numbers. If I found that the patient’s chances of having an indolent tumor were 68%, knowing what we know now, I probably would have advised him not to be treated.

What stance should the American Urological Association,* the American Cancer Society, and other organizations take with respect to routine PSA screening for men at average risk of prostate cancer?

We know now that screening saves lives. In the United States, about 30,000 men die of prostate cancer a year. Among those who consent to it, screening reduces the risk of death by 27%, which is more than 8,000 lives per year. That’s a large number. But we need to be clear that if prostate cancer is diagnosed, there’s a big chance that the tumors are insignificant or not life-threatening. To put that into perspective, men need to talk to their health care providers.

Many men may look at the relative risk of dying from prostate cancer, which is low, along with the negative findings from the PLCO trial, and wonder why they would have a PSA test in the first place.

Again, you shouldn’t mix the PLCO information with our data. We were surprised that these two studies were published together, with the implication being that they employed the same level of scientific rigor. The PLCO trial is very unlikely to ever show a significant difference between the screening and control groups. It’s easy to point out why. For one thing, 44% of the men in the study had had at least one PSA test before they were randomly assigned to a study group. That takes away the potential for the study to detect cancer in a large number of the participants. And the contamination rate in the PLCO trial was 52%.

What was the contamination rate in your study?

We estimate that our contamination rate was approximately 20%, but that varies a bit among the countries based on local attitudes and health care policies. The rate of effective contamination — the number of asymptomatic men in the control arm who had their PSA tested outside of the trial, subsequently had a biopsy, and were then diagnosed with cancer — was only 3%. [See “PSA contamination” below.]

Are you still enrolling patients in the ERSPC?

No. We locked the database on Dec. 31, 2008, but we will continue to evaluate the participants and publish our data, probably every two years. Of course, we’ll have to take into account increasing contamination of the control group over time, and we’ll make a better effort to quantify contamination. But any contamination that occurs in the control group now will have very little influence on mortality. That’s because any cancer cases in the control group that are detected now by screening are very unlikely to decrease the mortality rate within the next two to four years.

In light of the most recent findings, what do you tell patients about screening?

If a man comes to me and says, “I’m thinking about having a PSA test,” I tell him that screening will reduce his chance of dying of prostate cancer by 27% over nine years. Then I explain the downside, telling him that there’s a big chance that if he’s diagnosed with cancer, he will have indolent disease. I would then advise him not to undergo aggressive treatment. Even in cases of aggressive disease, his chances of being overtreated are still considerable.

In the United States, PSA screening has typically been done on an annual basis. Did you come to any conclusions about how often men should be screened?

We are very worried about the effect of our paper on screening. If people conduct screening more often or decrease the amount of time between tests, we’ll have more biopsies. That is something that we’d regret because it would further increase the rate of overdiagnosis.

We believe that future screening intervals should be determined based on a PSA value at a certain age. For example, a 55-year-old man with a PSA of 1.5 ng/ml could be screened less often than another man of the same age with a PSA of 3.5. The volume of the prostate should also be taken into account; a larger gland produces more PSA. Annual screening may be useful for people at high risk of prostate cancer. But the average man with a low PSA (below 1 ng/ml) and a normal DRE could be screened every four years or so.

Will you publish papers about differences in quality of life between participants who were treated and those who weren’t?

The papers are being written. They may show that screening doesn’t make sense when you factor in changes in quality of life due to treatment. But at 11 years, we are likely to see an increased difference in mortality with screening, and perhaps when we factor in quality of life, then the picture will change.

Do you think more men will be screened as a result of this study?

It will depend on how physicians in different countries respond to this new information. In my view, it would make a lot of sense if more people decided to be screened because they could reduce their risk of dying of prostate cancer by 20%. Personally, I would decide to be screened on the basis of these data. When I started this study in 1991, I wasn’t sure whether screening would be helpful in any way, but the uncertainty I had is now gone.

Originally published September 2009; last reviewed March 17, 2011.