Wait, so we’re doing PSA again? The ProtecT trial and the rise, fall, and resurrection of prostate cancer screening

Yesterday we talked about PSA, the molecule you hate to need, and certainly one of the most frustrating cancer screening methods in history. We reviewed how trials of population-level screening for asymptomatic patients demonstrated a small mortality benefit, but at the cost of significant overdiagnosis of prostate cancer. And yet the USPSTF is changing their recommendation from a grade D “recommends against” to a grade C “offer to individual patients based on risk profile.” Today, we’re going to figure out what made them change their mind.

We’re going to do this 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. A trial out of the UK, this is the most recent of several studies addressing the question “So you have localized prostate cancer. Now what?” The answer may surprise you! Or not, I guess, if it already matches your pre-formed intuitions. Let’s get started!

If You’re Only Going to Read One Paragraph

At a median follow up of 10 years, there was no difference in prostate cancer-specific death or all-cause mortality between an initial approach of surgery, radiation therapy, or active surveillance in patients with PSA-detected localized prostate cancer. The active surveillance group was more likely to have disease progression or metastatic disease over this time period, but presumably less likely to experience adverse events from radiation or surgery given that 45.2% of patients in this group did not require any such intervention over the 10 year period. This is a demonstration of the safety and efficacy of active surveillance as an approach to localized prostate cancer; the resultant minimization of harms related to surgery and radiation is part of why the USPSTF is changing its PSA screening recommendation from a D to a C.

Patients, Intervention, Comparator, and Outcomes

During the 10-year enrollment period, men between the ages of 50 and 69 who underwent PSA screening in the UK were recruited to the study. 82,429 were evaluated for inclusion, of whom 2664 were diagnosed with localized prostate cancer. Of these, 1643 agreed to undergo randomization (a high dropout rate, but you can’t fault the investigators for it — not hard to imagine why randomness is not desirable for some people in selecting cancer treatment). Diagnosis of localized disease and eligibility for any of the three therapies were the only criteria for inclusion apart from age (and having a prostate, at which most men excel).

Patients were risk stratified by Gleason score (< 7, exactly 7, or 8 to 10) and then randomized to one of three trial arms: radical prostatectomy, radiation therapy, or active monitoring. The monitoring group received serial PSA every 3 months for the first year and every 6 to 12 months thereafter; an increase in serum PSA of at least 50% over the baseline value from 12 months prior triggered reflex to further evaluation. There was a protocolized referral for radical curative-intent intervention if members of the monitoring group were found to have progression of disease. The radiotherapy group received neoadjuvant androgen-deprivation therapy (ADT), radiation, and then serial PSA to monitor for recurrence. The surgery group received radical prostatectomy (without specification of laparoscopic vs open technique, which given the 10-year span of the trial from 1999-2009 is likely to have evolved over the course of the study), followed by serial PSA monitoring. All groups were referred for salvage therapy as needed.

The primary outcome was prostate-cancer mortality at 10 years. All-cause mortality, rate of metastasis, rate of clinical progression, primary treatment failure, and treatment-related complications were planned secondary endpoints. It’s worth noting that the treatment-related effects measured here are not necessarily the ones patients are scared of — the investigators reported death, major bleeding, and cardiovascular events, whereas all of my patients are more afraid that they will suffer from incontinence or erectile dysfunction after treatment, rates of which are substantial. While these outcomes were assessed in this trial (that paper I just linked is a different study by the same investigators with the same population), they are presented as a separate analysis rather than alongside the mortality data. I get focusing on big headline events, but it’s pretty physician-centered to talk about the risks and benefits of prostate cancer treatment without talking the penis problems that keep patients up at night.

A Bit of Context

First of all, if you haven’t read yesterday’s post, I’d page back now. The history of PSA screening is really prerequisite to understanding the trial design the motivation behind the question it asks.

The outcomes from these different treatment methods in localized prostate cancer have been evaluated previously in two major trials. The Scandanavian Prostate Cancer Group Study Number 4 (SPCG-4), which randomized 695 men to surgery versus “watchful waiting,” demonstrated a prostate cancer-specific mortality benefit to surgery at 10 years (relative risk of death 0.50 for surgery, P = 0.02, NNT ~ 23), as well as a lower risk of distant metastases. This held up at a staggering 23 year follow up.

By contrast, the more recently-initiated PIVOT trial (1994 vs 1989 for SPCG-4) found no difference in either prostate cancer-specific mortality (HR 0.63, P = 0.09) or all cause mortality (HR 0.88, P = 0.22) between immediate radical prostatectomy or active surveillance. Certainly you could look at the cancer-specific numbers and their P value of 0.09 and wonder if there’s a trend towards a benefit, one that might make itself more obvious with a larger sample size. Still, the SPCG-4 found that benefit with an even smaller sample, so why the difference?

There are a few possible reasons. The most obvious one (and the one submitted by the authors of ProtecT) suggest that the difference may be due to the fact that only 10% of SPCG-4 patients had disease detected via PSA screening — indeed, the trial was initiated before PSA screening was ever implemented on a meaningful scale. Cancers detected by DRE or clinical symptoms may represent more advanced disease at time of diagnosis as compared with screening-detected disease. In both PIVOT and ProtecT, all cancers were detected via PSA screening, which could raise our number needed to treat (this goes back to the risk of overdiagnosis that we discussed yesterday). So we could reasonably interpret this discrepancy to mean that patients presenting with clinical symptoms of cancer may do better with immediate intervention, but there is as yet no data to suggest this is the case for those whose cancer was detected by screening alone.

Another possibility is simply that we have gotten so much better at treating advanced disease over the intervening years, with new androgen deprivation therapy and ever-better radiation techniques, that the benefits of avoiding metastatic disease that used to translate into a mortality benefit now just translate into needing fancy new therapy. If that were the case, you would likely expect to see decreased mortality in all groups, as both salvage therapy post-intervention and primary therapy for mets in the surveillance group give better results.

So — is it lower-risk patients or better treatment of advanced disease? As always in medicine, the answer is “It’s probably both.” Look at the results of ProtecT and you’ll see what I mean.


Median patient age was 62, median PSA was 4.6, and median Gleason score was 6, with 77% of patients scoring a 6 on pathology. 76% of patients had stage T1c disease. All of these were similar across the three groups. 99% of patients were white, a continued limitation of trials related to a disease where demographics can be prognostic. At 9 months after randomization, 88% of the active monitoring group, 71% of surgery group, and 74% of the radiotherapy group had received their assigned treatment, rates similar to prior studies.

There was no difference in prostate cancer mortality at 10 years between any of the groups (P=0.48). There was also no difference in all cause mortality (P=0.87). The active monitoring group did have significantly higher rates of disease progression and new metastases (22.9 and 6.3 events per 1000 person-years, respectively) compared to both surgery (8.9 and 2.4) and radiation (9.0 and 3.0) groups. There was no difference between the surgery and radiation groups for these outcomes. With regard to number needed to treat (NNT), 27 men would need to be treated with surgery or 33 men treated with radiation to prevent one patient from having metastatic disease. The NNT to prevent one case of clinical progression is 9 for either intervention.

Because I complained above about their not being presented side-by-side, let me say here that in the separate sexual and urinary function outcomes study, self-reported outcomes in surgical patients were generally worse at all time points. Radiation therapy had an initial worsening of function in both sexual and urinary domains (especially nocturia), but after 24 months from initiation of therapy function in the radiation group was no longer significantly different than the active monitoring group.

One way to read this is as a win for radiation therapy: with clinical outcomes equivalentmto surgery and better patient-reported sexual and urinary symptom outcomes, it appears to be the curative intervention of choice. Anthony D’Amico, a radiation oncologist, argues in an accompanying editorial that the correct interpretation of the data is that active monitoring should only be considered if the patient is unlikely to live long enough to suffer from the increased rates of metastases that come along with forgoing immediate curative intervention. Certainly he is right that these interventions can prevent the physical and psychological symptoms of metastatic cancer, as well as sparing the patient from the side effects of androgen deprivation therapy (and you thought the erectile dysfunction was bad with prostatectomy!). However, there’s one number I just can’t get over.

45.2%. That is the number of men in the active monitoring group who required no radical intervention over the 10 year follow up. Just under half of all screen-detected prostate cancer appear to do exactly nothing for 10 years after it is diagnosed. And now that we know the patients that do progress over that time don’t tend to die from having delayed treatment, even if they do have associated morbidity, I really have a hard time recommending early intervention. However, reasonable people could disagree, and if a patient felt that avoiding progression of disease was more important than avoiding potentially unnecessary treatment, I could in good conscious recommend intervention.

So how does this all play into USPSTF grading? Well, when you can reduce the total number of interventions through selective use of active monitoring, you increase the number needed to harm of screening (i.e. do less harm per person screened). This goes back to those graphs about risk I can’t seem to stop referencing, in which we decrease the adverse effects of the intervention so as to increase net benefit. The change from grade D to C is effectively an acknowledgement that, as management of screen-detected disease evolves to be more effective with fewer adverse events, there now exists a defined population whose risk of cancer-related morbidity is high enough to have a convincing net benefit. The fact that many patients will do fine with active monitoring is part of this equation.

Perhaps the most interesting number in the ProtecT trial isn’t a between-group comparison at all, but rather the overall mortality of about 1% at 10 years in all three groups of this trial. Prostate cancer treatment has been and continues to be revolutionized, so perhaps it shouldn’t surprise us to see that both prostate cancer and all-cause mortality were significantly lower than in prior trials. Ultimately, that’s the takeaway for me: we have gotten really, really good at treating advanced prostate cancer. So if you have localized disease and want to roll the dice that you’ll be like one of the 45.2% of people in the active monitoring arm of this trial who went 10 years without disease progression, this may buy you a metastasis that could have been avoided with early radical intervention. But we’re just so good at treating those that it’s unlikely to shorten your life. Still, if you care more about the risk of mets, which are certain to impact the quality of your life even if they don’t shorten it, it is not unreasonable to elect a curative intervention at time of diagnosis. If you do so, it is difficult to see why prostatectomy would be preferred over radiation therapy.

Prostate cancer and PSA will always be a scary topic. But the demonstration by the ProtecT trial that active surveillance can spare over half of monitored patients a radical intervention without any increase in mortality at 10 years gives us a way to offer patients screening without necessarily exposing them to the risks associated with overdiagnosis. Still, longer follow up will be necessary to see if this approach remains noninferior over a prolonged period.

Tomorrow on IM HEAT

An emergency room doc talks about pulmonary embolism in a way that won’t make you vomit. No really! Get ready for a really interesting piece by my friend and colleague Dr. Aalap Shah.

Don’t forget to follow along! Sign up for daily e-mails or click those fancy logo buttons at the top right of the page to slot us into your favorite feed.

2 thoughts on “Wait, so we’re doing PSA again? The ProtecT trial and the rise, fall, and resurrection of prostate cancer screening

  1. Great take, Bill! This trial was a monumental effort and all oncologists were surprised it got done because it’s incredibly hard to accrue patients for any surgery vs radiation trial–both referring physicians and patients usually come in with strong preferences. Interestingly, the investigators ran a qualitative study before opening this trial examining what barriers existed to accrual and found that a big one is physicians (urologist or rad onc) explain treatments to the best of their ability but with a bias towards what their hammer is. So Protect had highly trained research nurses explain surgery, radiation, and AS…and accrual was way easier in this trial than all the many surgery vs RT trials that ad been attempted (and terminated) before. Other radiation vs surgery trials are opening for lung cancer and following a similar recruitment strategy. Protect has opened up a lot of opportunities.


    1. Very cool Jahan! That’s great for research methods, but has disturbing implications for practice. For example, seems to imply that patients I send to urology are less likely to be offered radiation, and more interestingly that the patients whom I manage primarily are probably more likely to receive surveillance than either intervention. Perhaps we should have coordination nurses to help navigate any treatment decision like this; hope that’s a study for another day.


Leave a Reply

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

WordPress.com Logo

You are commenting using your WordPress.com 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 )

Google+ photo

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

Connecting to %s