Using PSA to determine prognosis

The 186,000 or so men likely to be diagnosed with prostate cancer in the United States this year will face a myriad of choices and ask countless questions. “Am I going to die of this disease?” is likely the most frequent query. Until relatively recently, this question was nearly impossible to answer. After all, it wasn’t until the late 1980s that testing for prostate-specific antigen (PSA) began; before that, most prostate tumors were detected during a digital rectal examination (DRE), when they were advanced enough to be felt through the rectal wall.

The advent of PSA screening meant that prostate cancer could be detected at an early stage, perhaps more than a decade before it would cause symptoms. But then what? What did a particular PSA test result mean for an individual patient? Did the same numerical PSA score in two different men mean the same thing? Was there any significance to a PSA level that rose quickly instead of gradually? Could doctors use PSA to determine which patients would be the best candidates for surgery? For radiation? For a clinical trial? Was there any way to know who might be cured and who was likely to die of the disease?

These are just a few of the many questions that spurred Dr. Anthony D’Amico to study PSA, its rate of change, and its role in a patient’s prognosis. He also wanted to know whether the prognosis would change if he opted for a different treatment method. After studying how more than 1,800 men fared years after treatment, he and his research team released their findings in a 1998 landmark study in The Journal of the American Medical Association.

The team stratified patients into risk groups depending on three factors — one of which was PSA. The findings that defined the risk groups became known in medical circles as the “D’Amico criteria.” And although more than a decade has elapsed since the study was published, doctors continue to rely on these criteria to help patients make treatment decisions.

We talked with D’Amico, chief of genitourinary radiation oncology at Brigham and Women’s Hospital and Dana-Farber Cancer Institute in Boston. He is also a professor at Harvard Medical School and the principal investigator of several federally funded studies. A prolific researcher, D’Amico has written more than 150 peer-reviewed original publications and editorials for medical and scientific journals. Here he shares what inspired his PSA research, how it’s relevant in the clinic, and his own recommendations for patients.

Early PSA research

What inspired your research on PSA and how quickly it changes?

My patients inspired me. A gentleman came in with a PSA of about 4 ng/ml, and he asked me, “Do you think it matters that my PSA was 1 ng/ml last year?” The man I had seen just before him had a PSA of about 4, too, but his PSA the year before had been 3.2 ng/ml. That piqued my curiosity. His question was what inspired me to look at the rate of change in PSA before treatment as a potential prognostic factor.

Patients also inspired me and my colleagues to look at PSA levels after surgery. With some patients, the PSA would be 0.1 after surgery, then 0.5, and then, all of a sudden, 3 ng/ml, and their bone scans would turn positive and their CT scans would light up. But then there were patients whose PSAs would go from 0.1 after surgery, to 0.15, to 0.19, to 0.21 ng/ml over the course of several years. Their PSA never seemed to translate into anything. Looking at the differences in these cases raised questions in my mind.

Change in PSA is probably most important after treatment because the only thing that really affects it then is the disease itself. Before treatment, all sorts of things can affect PSA, making it a less reliable indicator of what’s happening with cancer. After surgery, it should be coming down and stabilizing or staying undetectable. When you see a rapid rise in PSA after treatment, you know that patient is going to have trouble soon. We don’t know whether having more treatment immediately is better for this patient than deferred treatment, but that question has spurred clinical trials. Looking at PSA changes after treatment also helps us stratify patients and determine which patients should enroll in clinical trials of novel agents.

So if a patient’s PSA is doubling every two months after treatment, can you assume that the PSA was doing the same thing before treatment?

We can definitely correlate post-treatment relapses with pretreatment PSA velocity, or how quickly the PSA rises. We did a study showing that a pretreatment PSA that increased by more than 2 ng/ml in a year is the strongest predictor that the PSA will double in less than three months after surgery. That’s important because when the PSA doubling time is less than three months, the patient has a significantly shorter life expectancy.

How did your risk classification system develop?

The risk groups are really based on three parameters: PSA level; Gleason score, which assigns a numeric value to the cancer based on what it looks like under the microscope; and DRE findings.

When I first started practicing medicine, age was the only prognostic factor that was used to determine if a man with prostate cancer would have surgery or radiation. If he was young — less than 70 — he had surgery. If he was older, or not healthy, or both, he had radiation. All factors other than age were thrown out.

I felt we really needed to consider other factors and put some structure into the system. So we decided to figure out who didn’t do well after surgery or radiation. We learned quite quickly that men with high PSAs did poorly, as did men with high Gleason scores and men at advanced clinical stages of the disease. That became the basis for the risk system: if a patient had one of these high-risk factors — a PSA over 20; or a Gleason score of 8, 9, or 10; or a tumor that you could feel during a digital rectal exam — surgery or radiation alone wasn’t likely to be enough for that patient. We knew he’d need additional treatment.

Later on, we were able to show that these risk factors not only predicted PSA recurrence, or biochemical recurrence, but also death from prostate cancer. That’s when risk classification became clinically relevant. [See “What’s biochemical recurrence?” below, and Figure 1 and Table 1, above.]

But those aren’t the only factors you consider today, are they?

I’ve often said that those three factors — PSA, Gleason score, and physical exam findings — aren’t enough to figure out which group a man falls into. Today, we put icing on the cake. We have PSA velocity, which can move a low-risk patient into a high-risk classification. The presence of perineural invasion on the biopsy can reclassify a low-risk person as being at intermediate risk. If more than half of the cores taken during a biopsy are positive, a patient moves up a risk group. PSA may also be affected by testosterone levels. Men with low testosterone actually have higher PSA levels than what the PSA test shows. For example, a patient with very low testosterone and a PSA of 2 ng/ml might be in the same risk group as someone with a normal testosterone level and a PSA of 10 ng/ml. In some cases, a patient may have a tertiary grade, too. [See “Tertiary grade,” below.]

So a lot of fine-tuning has occurred since the initial risk-based system was developed. But providing the original structure really helped people to further the idea. Now we have nomograms that treat the risk factors as continuous functions and yield a probability of failure from 1% to 100%. These nomograms essentially take the risk groups and make them points on a continuum as opposed to set categories.

Backing up, how did you determine that a change in PSA greater than 2 ng/ml in the year before diagnosis meant a higher risk of prostate cancer death?

We looked at data from more than 1,000 men with a median PSA of 4.3 ng/ml who were screened and diagnosed with prostate cancer and then had a radical prostatectomy. We then looked at the men who had a change in PSA in the year prior to diagnosis and divided them into four groups. The first group had a PSA change of 0.5 ng/ml or less; the second group, 0.51 to 1 ng/ml; the third group, 1.01 to 2 ng/ml; and the fourth group had a change greater than 2 ng/ml in a year. We found that the men in the first, second, or third group did fine. Prostate cancer death rates were very, very small at 10 years — about 1% following surgery. But for those in the fourth group, the death rate from prostate cancer was huge — up to 28% within seven years of radical prostatectomy. We saw that and we were struck. [See “PSA velocity after surgery,” below.]

One thing to keep in mind about PSA velocity that is not in the study report, but which is true, is that when we refer to people in the cohort with “velocity greater than 2 ng/ml,” we’re including men whose PSA velocity was 2.1, 6.5, and so on. Since there’s no upper limit, the relationship of PSA velocity to cancer death is going to be driven by those at the extremes. I don’t want people to think there’s something magical about a change of 2 ng/ml. Of course, that’s what was published, so everybody remembers it, but 6 is worse than 4, which is worse than 2.5 ng/ml. The bigger the change, the worse it is. For a man with a velocity of 2.1 ng/ml in the previous year, his absolute risk of death from prostate cancer is nowhere near the 19-fold increase that was mentioned. His absolute risk is closer to 1.4.

You mentioned perineural invasion. Can you give us your thoughts on that? Not everyone believes that’s an important pathologic finding on a biopsy.

There are a couple of studies that have looked at cancer death in otherwise low-risk patients with and without perineural invasion [PNI], or cancer around a nerve, on the biopsy who have been treated with radiation or surgery. They show that about 5% to 10% more men who were at low risk of disease progression die of prostate cancer within 10 years of treatment if they had PNI. [See “Impact of PNI,” below.]

We know that PNI can mean that there’s higher-grade disease or that there’s extracapsular extension. But I don’t want people to think that every man with low-risk disease and PNI has an increased risk of death. That’s not true. However, there is some subset of men with PNI who truly don’t have Gleason 6 disease. They’ve got Gleason 7 or 8 disease, and PNI is often a harbinger of higher-grade disease, but not always.

Is that generally accepted?

Some people have found a relationship, but others have not. This gets back to my point that not every patient with PNI is going to have a worse outcome. If a man has had a 12-core biopsy and one core shows PNI and Gleason 6 cancer, he’ll probably be fine. But someone who has had a 12-core biopsy with seven cores that show Gleason 6 and PNI is more likely to have trouble.

How you handle that depends on how you think as a physician. I take the stance that I would rather overtreat 10 people than undertreat one. What’s the price of overtreatment? Hormone therapy, which I would add to radiation therapy. I tend to use short-course hormone therapy, which is six months of hormone therapy, in patients with PNI and otherwise low-risk disease.

When do you use hormone therapy in addition to radiation?

I will use it for anybody whose disease is not strictly low-risk — and not just by the original D’Amico criteria. Let me outline it. In order for a patient to get monotherapy [a single therapy, such as radiation therapy alone], without hormone therapy, he has to have a Gleason score of 6 or lower, with less than half the biopsy cores being positive, a cancer stage of T1c or T2a [see Figure 2 below], a PSA that doesn’t exceed 10 ng/ml, and a PSA velocity that doesn’t exceed 2 ng/ml in a year, and he cannot have PNI. Nor can the patient have low testosterone, because if he does, as I said, his PSA is not valid.

For anyone who doesn’t meet those low-risk criteria, I will add a short course of hormone therapy to radiation therapy. That includes patients with a Gleason score of 6 or higher with more than half of their biopsy cores positive for cancer; patients with a Gleason score of 7 (4 + 3) or higher; patients whose cancer is a stage T2b or higher; and patients with rapid changes in PSA.

I’d think through some cases a little more carefully — for example, take a patient with a Gleason score of 7 (3 + 4) and five of 12 cores positive, no PNI, stage T1c disease, and a PSA score that was 3.5 ng/ml a year ago and has gone up to 4.1 this year. Does he really need combined therapy? I would look at his family history. I would look at his ethnicity. I would see whether he was regularly screened for prostate cancer or if it was picked up in just the second year of screening. I would also look to see if he had had any prior biopsies. I’d factor these other things into my equation to figure out his treatment. If I can’t find anything else that would affect his risk, I’d recommend monotherapy.

What about the percentage of cancer within a core?

That’s more controversial, but we’ve looked at that, and others have too. Given the possibility of sampling error, if a man has two positive cores out of 12 and each core has 100% involvement, it’s hard to know if that really puts him into a higher risk category. You can get lucky and hit the exact spot where a small cancer is and get two cores that are 100% cancerous. But if you find that six of 12 cores are positive, that’s not just luck. There’s definitely something there, regardless of the percentage of cancer within each core. So I just look at the number of positive cores.

Didn’t the European Organization for Research and Treatment of Cancer (EORTC) study the use of hormone therapy in the same types of patients you’re describing?

The patients in that study were a bit different. Some of them had locally advanced prostate cancer, meaning they could have extracapsular extension or seminal vesicle invasion. They could have stage T3 or T4 disease as long as it didn’t involve regional lymph nodes. That’s more advanced disease than what I’ve been describing. The men in that study also received three years of hormone therapy. [To read this study’s findings, see “Radiation, hormone therapy, and the EORTC,” below.] I don’t question giving them hormone therapy and radiation, but did they need hormone therapy for three years? That study started in 1987, and I think we’re a lot smarter now about hormone therapy than we used to be.

So who should receive longer-term hormone therapy?

I’d be inclined to prescribe two years of hormone therapy, or even a little more, to a man who has high-grade cancer and is young. Although it’s not definitive by any means, a retrospective study will be published later this spring in the International Journal of Radiation Oncology, Biology & Physics showing that it may not be how much hormone therapy you give, but how long the effect lasts — that is, how long the testosterone stays at castrate levels predicts a patient’s longevity with regard to cancer death.

We looked at 206 men who received six months of hormone therapy and radiation or radiation alone in a randomized trial. We measured their testosterone levels at the start and then every three months after the end of hormone therapy. We were able to show that the men who lived the longest, or were the least likely to die from prostate cancer, were those whose testosterone remained suppressed to castrate levels for at least two years. And those men tended to be older; their testosterone rebounded more slowly. The younger men tended to rebound more quickly. So, in my mind, whether you need to give long-term hormone therapy to an elderly gentleman is a question. But for a younger man, you’d be foolish not to do it.

Technology and decision-making

When do you use magnetic resonance imaging (MRI) with an endorectal coil* in your decision-making?

That’s evolved over time. We published the results of a study suggesting that the best yield for this technology in terms of sensitivity and specificity is in the patients at intermediate risk. In low-risk patients, it has very little utility because it’s almost always wrong. The only time it’s right is when it shows seminal vesicle invasion, and that happens in only 2% of low-risk patients. If you want to cross all of your T’s and dot all of your I’s, then you could do this test in all low-risk patients, but only two times out of 100 will it give you something that will change patient management. I’ve also learned that you don’t need to do it in patients at high risk because you already know their cancer is extensive.

I use endorectal MRI most commonly in patients who have more than one determinant of intermediate risk — that is, if they have a nodule and a Gleason score of 7, or a nodule and a PSA level over 10 ng/ml. If patients have only one determinant of intermediate risk, such as stage T2b, MRI isn’t likely to be very helpful. But if they’ve got multiple determinants, the yield is much higher. That’s when it becomes more cost-effective and clinically useful.

What about the anatomic location of the nodule?

Every study I’ve read, including one from the University of California–San Francisco published in Radiology last year, concludes that MRI in the right patient population will clearly show extracapsular extension about 80% of the time. But because the images can be so hard to interpret, what we think is the actual location of a nodule turns out to be right only about half the time. Therefore, I don’t use it to determine whether to spare a nerve on one side or the other, because I think the images can fool you.

But you have used MRI to help guide the placement of brachytherapy* seeds, haven’t you?

Yes. I’m really proud of that work. I think it allowed many people to maintain their quality of life because they suffered fewer side effects. In a very select group of low-risk patients, we differentially dosed the brachytherapy seeds throughout the prostate. The peripheral zone, where most cancers arise, got the full dose. The transition zone, where the cancer is less likely, got about 80% of the full dose. And the anterior base, where these select patients often have no cancer, got half the full dose. With that lower dose at the anterior base, the urinary retention rate, the rate of acute, long-term urethral stricture [narrowing of the urethra], and the incidence of rectal damage were all markedly reduced.

One of your editorial board members, Dr. James Talcott, studied quality of life in men treated with brachytherapy and proved that patients had better quality of life with regard to gastrointestinal and genitourinary function in the months after the procedure when the MRI approach was used compared with the traditional use of ultrasound alone to place the seeds. [See “Seed placement and quality of life,” below.]

So you’re not worried about tumors in the anterior base of the prostate?

In this very select group of people, we rarely see anterior base tumors. Period. When we do, they are usually cancers that have grown from the peripheral zone anteriorly. But in a very low-risk cohort, that shouldn’t be the case.

We’ve treated almost 800 men with this approach. Five years later, the PSA control rate is 95%. That’s not only because of the technique, though. It’s also because of the patients we select. When the 5% who experienced biochemical recurrence had saturation biopsies, the cancers that were found were not found in the anterior zone. They were actually found in the transition zone. So I think the anterior base can be spared in this select group.

Couldn’t you use ultrasound to place the seeds in the same way?

In a way, we do. It’s called MRI-ultrasound fusion. We do an MRI and an ultrasound and fuse the images together. That gives us more information. This is the technique we currently use to place brachytherapy seeds.

Can patients who live in rural areas have MRI-guided brachytherapy?

It’s unlikely. They can get ultrasound. MRI-ultrasound fusion is a software package that hospitals can buy, but they need to have a physicist who is savvy enough to know how to use it. It’s available in institutions in or near large cities, but if you’re far from any major institution, you’re not going to find it.

Biochemical failure: What’s next?

At the beginning of our discussion, we talked about biochemical failure, or biochemical recurrence. How is that defined?

After surgery, biochemical failure occurs when a patient has a PSA level greater than 0.1 ng/ml on two consecutive PSA tests. After radiation, it’s not as straightforward; the definition of biochemical failure keeps changing. The most recent one is the Phoenix definition, which is the PSA nadir, or low point, plus 2 ng/ml. I understand the reason it was changed — it alerts you to the most significant failures, the ones that lead to distant metastases and cancer death. But it’s really understating all the PSA failures.

You might say that’s a good thing because patients who have very protracted PSA failures may have disease that never becomes clinically significant. I think the reason the new definition was adopted is that older men typically have radiation therapy, so a protracted PSA failure for them might never be a problem. But when younger men have radiation therapy, particularly brachytherapy, I think it’s more relevant. Indolent PSA failure in a man who is in his 50s could become cancer death.

Personally, I think the definition of PSA failure should factor in the health and age of the patient. For a younger man undergoing radiation, I will use the same definition we use in surgery, and that is anything above the nadir.

You see a lot of patients with biochemical failure after primary therapy with surgery or radiation. Obviously one plan won’t work for all of them, but how do you counsel them?

I look at all the factors that are relevant. I look at their age and their health. I look at how quickly their PSA is rising, their Gleason score, and the time interval to biochemical failure. I then consider two questions: First, “Does something need to be done and if so, when?” Second, “What’s the anxiety level of this particular patient?”

Let’s start with the patient who is not anxious, which is rare. If he is not anxious, then I look at his age and health. If he’s older, has other health problems, and his PSA is rising slowly, meaning a doubling time longer than a year, I will not treat him. But if he’s young and healthy, I will recommend that he see a medical oncologist to discuss early treatment, because I think those are the people where we don’t know, given that they’re going to live long enough, that early treatment may not be helpful.

Now, if a man’s PSA doubles rapidly, meaning in less than six months, I have a tendency to treat, even if he is old and frail. But if he’s had radiation already, should he get hormone therapy? If so, should he get continuous or intermittent hormone therapy? [Intermittent hormone therapy involves alternating cycles of treatment and no treatment.] These are unanswered questions. For young, healthy people, I favor continuous therapy. For people who aren’t as healthy, I favor intermittent therapy. I also watch their testosterone levels because I’m not sure that older men or men with other health problems recover their testosterone as quickly as younger, healthy men.

What do you do if the patient is anxious?

If he’s anxious, all of this changes. You can’t say, “I’m not going to treat you,” and send him out the door. Instead, I talk about intermittent therapies or clinical trials. That way, the patient feels like he can do something, and that helps with the anxiety. The anxiety level in this population is huge.

What do you do for patients who experience biochemical failure following brachytherapy? Do you re-treat them if there’s a localized recurrence?

Let’s consider a man who’s in his early 60s and has a while to live, has no other illnesses, had low-risk disease to begin with, and his PSA is rising very slowly, consistent with a local recurrence. He has three choices. The first option is salvage prostatectomy, but not many doctors do this procedure. And the patient who is considering it needs to have his eyes open because the risk of serious side effects, such as permanent incontinence, is great.

Salvage cryotherapy, which involves freezing tissue, is another option. But I don’t favor it, mainly because you have to put a warmer in the urethra to protect it. At the apex, which is where a lot of the recurrences are, the warmer will heat the prostate tissue and the cryotherapy will not be effective. I’d stay away from it.

The third option is intensity-modulated radiation therapy (IMRT), CyberKnife, or another form of radiation therapy. The patient would have to have a saturation biopsy of 40 or more cores so you could tell exactly where the tumor is located before you start the second therapy. But there will still be side effects. Even with MRI-guided salvage brachytherapy, we have shown that 10% to 15% of patients will have complications that require surgery, such as a colostomy or urostomy.

What about high-intensity focused ultrasound (HIFU)?

HIFU is investigational, so if a man wants to go that route, he needs to know that it’s not a routine procedure. It needs to be done, in my opinion, as part of a clinical trial.

So you categorize patients as low-risk if their PSA doubling time is greater than a year and high-risk if it is less than six months. What about patients in the middle?

A paper in The Journal of the American Medical Association looked at this question. The researchers looked at PSA doubling times of less than three months, three to nine months, nine to 15 months, and more than 15 months. The patients in the first two groups did poorly. The patients in the nine-to-15-month group did fairly well. [See “PSA doubling time,” below.] So as the PSA doubling time gets closer to a year, you’ve got about 10 years to play with. In a younger man, that’s not very long, so I would start treatment, probably radiation therapy combined with hormones. But an older gentleman could be fine; you could start hormone therapy at a later time, maybe when his PSA doubling time drops or when he becomes symptomatic.

With so many different types of patients and so many possible treatment scenarios, as well as the fact that so many patients are being diagnosed with prostate cancer much earlier, does your original risk classification system still help stratify patients and inform treatment decisions?

The risk classification system will continue to be clinically useful until genetic-based molecular markers have been found and shown to be effective — and until testing for them becomes available. In the future, we will be able to use gene signatures to help determine a patient’s best course of treatment and prognosis.

Originally published June 2009;  last reviewed March 21, 2011.