Prostate-specific antigen and its discontents: Why you don’t know whether to screen your patients

OK, we’re going to do another two-part update, because I want to talk about the evolving evidence around prostate cancer screening, and the only thing worse than trying to write about all of that in one sitting would be trying to read what came out.

So let’s talk about prostate-specific antigen (PSA), the molecule you just don’t know what to do with. PSA is like your friend who really wants to help you move, but keeps dropping your stuff. Despite the best of intentions, nobody’s sure if he’s doing more harm than good. You all remember that Grade D recommendation the USPSTF gave PSA as a population-level screening tool? Well, turns out they’re changing their mind. “Who is on the USPSTF and what the heck do they really do?” is an interesting topic for some mini-post in the next few weeks, but for today we’re going to focus on the data about PSA screening that underpinned that initial grade D recommendation (the D is for Don’t do it), and tomorrow we’ll go over the changes that have moved them instead to recommend an individualized approach.

Tomorrow, we’re going to look at these changing recommendations through the lens of the ProtecT trial, a 10-year trial by Hamdy and colleagues comparing active surveillance, surgery, or radiotherapy for localized prostate cancer. Today, let’s learn what we need to know to appreciate how those results shaped policy.

A Brief History of PSA

There’s a reason I’ve broken the usual context section out into its own post: there’s just so much data that we need to appreciate here. While I’ve tried to keep a high altitude view, this is going to take a second.

Prostate cancer is the second most common cancer in the US, and the most common among men. As such, there is obvious impetus driving research into effective screening techniques. PSA, a glycoprotein produced by the prostate and secreted into semen, is manufactured exclusively by prostate cells. Elevated serum PSA levels can occur in situations of either benign or malignant neoplasia, as well as prostatic inflammation. Use of PSA for prostate cancer screening in asymptomatic patients has been evaluated in two major clinical trials.

Let me say up front that a major limitation of all of these studies is the failure to include significant proportions of black patients (no trial included them as more than 4% of the study population). Black men get prostate cancer at a rate 60% higher than their caucasian counterparts, and have faster rates of growth and metastasis even controlling for baseline levels of disease. Higher risk means more likely to benefit from screening, so even as we have poured incredible amounts of resources into investigating the benefits of screening PSA for white men, the question of whether African American patients should be screened remains largely unanswered.

First, the Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial (PLCO) randomized 76,693 men to annual PSA screening vs usual care, which resulted in a higher rate of cancer detection in the annual screening group than the usual care group, but no difference in mortality at 10 years. However, this trial had a tremendous rate of contamination, which is to say that patients in the control group actually got more cumulative PSA testing than those in the annual screening group. Recent 15 year follow up data are consistent with the results of the original trial; the authors of this follow up appropriately remark that what this really indicates is that there is no difference between organized versus opportunistic screening with PSA. As a result, this trial is largely written off in ongoing discussions of PSA screening.

Then the European Randomised (FANCY BRITISH SPELLING ALERT) Study of Screening for Prostate Cancer (ERSPC) randomised 182,160 men in 8 European countries to PSA screening versus no offer of screening. Did you read the word “randomised” in a British accent? If not, go back and do so. I’ll wait.

Good? Good.

In stark contrast to the PLCO, the ERSPC did demonstrate a 21% relative reduction in mortality, albeit with a number needed to screen (NNS) for 11 years of 1055 in order to prevent one death. At 13 year follow up, the benefit persisted, with a NNS of 781.

Wait, there IS a mortality benefit? Why am I not screening again?

Overdiagnosis of cancer was a major issue in the ERSPC. Incidence of prostate cancer was significantly higher in the screening group than the control group over this period, with 1.57 times the number of cases in screening versus control. This indicates cases found on screening that would not have presented (and thus presumably not affected quality of life in any meaningful way) without screening.

Let’s pause on this concept for a second. The premise is that both groups have an equal rate of screen-detectable cancer. So if a case is detected by screening but is not incident in the non-screened population over the same period, that represents an indolent cancer for which treatment is not likely to result in medical benefit — after all, those same cancers are not causing any symptoms in the control group, let alone mortality. Because discovering these cancers incurs all the cost and adverse effects of diagnosis and treatment without generating clear medical benefit, it is, in short, a bad thing.

There were 3.44 excess diagnosis cases of prostate cancer per 1000 person years in the screening group, which, compared to the 0.11 per 1000 person-years mortality reduction, is not a small number. What’s the harm of this overdiagnosis? According to best available estimates, among patients with normal baseline function more than 90% undergoing prostatectomy will have some degree of sexual dysfunction at 36 months after surgery, as will > 75% of radiotherapy patients. Numbers for urinary incontinence are not as bad, though 43% of prostatectomy and 18% of radiotherapy patients who have normal function at baseline will have at least some degree of incontinence post-treatment. So there is a clear risk to overdiagnosis and over-treatment of cancers that may never progress over a decade of follow up, in exchange for a mortality benefit for one in every 781 patients over 13 years.

To figure out whether there is a net benefit here relies on complex models including quality-adjusted years of life (attempting to answer the question “how many years living with urinary incontinence are worth one year in perfect health?”). The authors of the ERSPC state that their model shows a net benefit of screening — the USPSTF was more skeptical. Given the rates of overdiagnosis at more than 30 times the rate of mortality benefit, they recommended against screening at a population level. This did NOT mean that we should not screen patients with symptoms, a strong family history, or other significant risk factors, but it did mean that EPIC started telling us to perform “PSA counseling” with men between the ages of 50-69 rather than “PSA screening.” In short, if you’re going to screen patients, you had better be selecting high-risk patients so that the significant burden of adverse events from overdiagnosis is not likely to outweigh the benefits of screening. (Now you’re starting to understand why I drew these graphs the other day — every time I end up talking about population-based interventions I draw one of these in my head.)

But what if you could reduce the harms of overdiagnosis? What if instead of subjecting screen-detected patients with localized disease a radical intervention and all of the undesirable outcomes that go along with this we offered them a chance to prove their disease is indolent? Tomorrow: The ProtecT trial, a comparison of surgery versus radiation versus active surveillance for screening PSA-detected localized prostate cancer.

Blog Stuff

A quick announcement: Following part two of this piece tomorrow, we’re moving from 5 days a week to Monday/Wednesday/Friday updates, with Tuesday/Thursday reserved for guest posts as available. This is for several reasons. Mostly, while I love writing this thing, it is a crap ton of work to do so, and my sanity demands a little more time to be a person. Moreover, the readers we do have are awesome, but this is a resource I want to get in front of everyone who can use it, so I’m going to be spending more time publicizing it, as well as expanding to an associated podcast. Third and perhaps most importantly, you need more perspectives than just mine. Risa’s post last week was awesome, and shows how great IM HEAT can be when we get more people involve. My friend Aalap Shah, an EM doc, has a piece coming up on Wednesday that I am super jazzed for. I would love to get enough guest pieces to fill in Tuesday and Thursday updates and actually continue 5 posts per week that way, but I’m not going to be able to coordinate that unless I drop some of this workload first.

So I’m going to take the time saved and use it to coordinate a small army of writers (which you should be a part of — e-mail us at with ideas!), to get this blog in front of as many sets of eyes as humanly possible, and maybe sometimes to have a long, leisurely dinner with my wife. Tuesday/Thursday updates will be available as they come in. If you want to keep up, make sure you follow us!

Twitter: @imheatpod
Facebook: @imheatpod
Instagram: @imheatpod
E-mail: In your Inbox

You’re all the best. Never change. Unless the change is getting smarter by keeping up with awesome new evidence in medicine, in which case I’m cool with it.

3 thoughts on “Prostate-specific antigen and its discontents: Why you don’t know whether to screen your patients

  1. I think you have to delineate between overall mortality and prostate cancer related mortality, which are two different things. The study states that they only used overall mortality to gauge similarity of the arms as prostate cancer related deaths are such a small portion of the overal deaths.
    But why should I as a patient care what kills me? Those who aren’t screened may be more likely to die of prostate cancer than a heart attack, but they still live to the same age. In addition, we’ve established that quality of life is decidedly lower among those who are screened.


    1. Certainly you could make the counter-argument that it’s not the job of the screening to reduce overall mortality, merely mortality from that specific disease. If it moved the needle on that endpoint but not on mortality overall, it might still be worth the trouble. However, certainly you’re right that if the intervention is not producing an overall mortality benefit it is less likely to be worth the quality-of-life issues that come along with screening.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s