A recent article in the Archives of Internal Medicine about using the PSA (prostate specific antigen) for screening for prostate cancer has garnered the attention of some of the medical blogosphere, including Kevin, MD and MedPundit. The full article requires subscription, so I have not been able to read the whole thing, but the gist is that PSA screening, with or without DRE (digital rectal exam) of the prostate does not improve survival times–i.e., men whose cancer is detected using PSA screening don’t live any longer than those cancer is diagnosed by other means.
Color me skeptical about this study — although it’s not entirely a surprising conclusion, either.
PSA is a protein released by prostate tissue which is measurable in the bloodstream. It is not a cancer test–it is a prostate test. This is one area of confusion about this test in many people’s minds. Normal prostate tissue secretes PSA, in amounts roughly proportional to the size of the gland. As the prostate enlarges with age, PSA levels tend to rise, typically rather slowly, due to the growth of a benign tissue called adenoma. This is the stuff which tends to give guys less pressure with urination as they age, as the adenoma may cause a degree of obstruction to the urinary channel. PSA is a rather noisy test as a cancer screen, however: it can be elevated for a host of reasons. Infection or inflammation in the prostate can cause a striking increase in blood levels; instrumentation (such as bladder catheters or endoscopy) can give it a good bump; and of course cancer, which tends to produce more PSA than benign tissue relative to its volume.
Cancer screening is one of those things which sounds great on paper but which suffers from the ravages of statistical math. Let’s say, using simple numbers, that you have a disease, say cancer, which occurs at an incidence of 10 people out of 10,000 (just pulling numbers out of the air). Left untreated, all 10 of these people will eventually die from this disease. If diagnosed early enough, all 10 can be treated successfully, and will not die — at least from this cancer (a very optimistic assumption, but let’s use it for the sake of discussion). So it is obviously desirable to find a way to detect this cancer early enough so that it can be treated before it is too late.
Now let’s say you have a diagnostic test which can detect 9 out of these 10 cancers, if everyone at risk for the disease gets tested. In other words, 90% of the people with cancer will have an abnormal test. That’s called sensitivity: how many people who actually have the disease will have an abnormal test — and 90% (9 of 10) is pretty darn good for most screening tests. So far, so good, right? Well, that’s just the beginning. Let’s say the same test is abnormal in 1 out of 10 people without cancer. This is called specificity: 90% of the people without cancer have a normal test, and 10% have what’s called a false positive. This is also pretty darn good as screening tests go: there’s no perfect test, all have both false positives and false negatives.
Now let’s apply these numbers and see what happens. All 10,000 people get tested; 10 of these statistically have cancer. The test finds 9 of the cancers–this is good (except for the one who gets missed — but that’s why God created trial lawyers). The test is abnormal–therefore indicating cancer — in 1000 people (10% of 10,000) who are in actuality cancer-free.
These 1000 people will need more tests, x-rays, biopsies, etc., to prove that they don’t have cancer — and these subsequent tests also have false negative rates. So of these 1000 people, a few may still have cancer (as far as you can tell), and a whole lot more are worriedthey have cancer, even though the subsequent tests don’t show it, and they don’t actually have it. And, wait–there’s more! The additional tests for the 1000 who do not have cancer, but who have an abnormal screening test, also have problems: they are costly, often uncomfortable, have risks of their own, and even they may not diagnose every cancer. So you end up spending a lot of money, adding additional risk and discomfort, with a whole lot of people to find that relative few who have cancer early enough to treat. Suddenly screening tests for cancer aren’t looking quite so rosy anymore.
Now let’s look at the economics. Let’s say the screening test costs $20, and the subsequent diagnostic studies needed when the test is positive cost $1000 (fairly conservative numbers, by the way). So our screening test cost $200,000 ($20 times 10,000 tests). The subsequent diagnostics for those found with cancer cost $9000 (9 times $1000). The diagnostic expenses for those with an abnormal test but no cancer is $1,000,000 ($1000 times 1000). So we’ve spent $1,209,000 to diagnose 9 early cases of potentially curable cancer: over $134,000 per cancer found — before it is ever treated.
Now, society or public policy may judge that such an expenditure is justified to save these 9 lives. But we have made other assumptions here which are not always true: that if we detect the cancer early, all those so discovered will be cured; that even if they are cured, that they will live longer (since the list of other things which can kill you is rather long); and that failure to treat the cancer will shorten their lives. These seem reasonable assumptions — but they are more often wrong than not.
Take prostate cancer, which the PSA test is designed to detect. It typically occurs in men over 50 years of age, increasing in incidence with increasing age. It is often very slow-growing–so slow that many men who are diagnosed with prostate cancer end up dying of other diseases–heart attacks, strokes, other cancers, etc. Yet it remains a significant health risk: over 30,000 men die of prostate cancer annually in the U.S–and it tends to be a slow, rather painful way to die for many.
With prostate cancer, nearly every assumption we made in our neat little example above is uncertain. First, there is the problem of selecting the population to screen. Not every population has equal or predictable risk. Men under 50 can get prostate cancer — and often have a very aggressive, rapidly growing variety–but the disease is much less common than in 70-year-olds. Men over 70, conversely, have a much higher incidence, but are far more likely to be managed without aggressive treatment, as other diseases are much more likely to pose a mortality risk. African-Americans are at higher risk than Caucasians–but get screened less. Men with a strong immediate family history of prostate cancer are also at higher risk–but tend to have fast-growing aggressive tumors at diagnosis, and even early detection may not result in cure.
Then there’ s the problem of the sensitivity and specificity of the PSA test itself. Normal values for PSA increase with age, because the prostate generally gets larger with age. But size of the prostate at any age varies widely with each individual, is not necessarily age-dependent — and therefore what is normal for your age group may not be normal for you. There are several different assays (laboratory techniques) which can give differing values on the same patient, making comparison of sequential values tricky. Random variations in single values are also common, often for inexplicable reasons–and can lead to unnecessary evaluation and biopsy in some cases. And lastly, you can have prostate cancer with a normal PSA. A recent large screening study — where biopsies were done on men with normal digital exams and PSA values — found prostate cancer in a surprisingly large percentage of men in whom the PSA and physical exam were entirely normal.
When a patient with an elevated PSA needs further evaluation, this generally involves evaluating the prostate with ultrasound, and often involves a biopsy of the prostate. But ultrasound is only fair at identifying abnormal or suspicious areas in the prostate, and biopsy has a false negative rate (i.e., the patient has cancer but the biopsy comes back benign) estimated at 3-5%.
Now let’s assume we’ve found prostate cancer after a biopsy done for an elevated screening PSA. Our neat little example above implies that treatment of the cancer is performed, leading to cure and a longer life. But it’s not nearly so cut and dry in the real world. Many prostate cancer patients–especially younger ones with more aggressive forms of the disease — may be relatively advanced even when discovered by screening PSA, and treatment may not cure them or prolong their life. Conversely, some men with earlier, less aggressive cancer, may well have never had a problem with the cancer if it had not been diagnosed at all. Others with clinically aggressive disease, which might otherwise kill them in time, die from unrelated causes before the cancer gets to this stage. And although there are some indicators of aggressiveness and prognosis which have some predictive value (extent of disease at diagnosis, initial PSA value, pathologic appearance or Gleason score), crystal balls are in short supply, both for predicting future behavior of prostate cancer, and for the likelihood of demise from unrelated causes.
Is it any wonder that studies of the value of screening with PSA are going to have a tough time sorting this mess out?
Oh, and by the way: this is the discussion you’re supposed to have with every patient who requests a screening PSA, according to the AMA and the NCI. And you wonder why most physicians laugh at such recommendations? Even when you have such a discussion, patients will always opt for the screening PSA. Always. And if they decline to have it done, and later find out they have prostate cancer–you’ll get sued for failure to diagnose it. So much for the big-shot’s recommendations.
So what to do about PSA screening? And why do I suspect this recent study disproving its value is likely wrong?
The answer lies in the big picture, and in experience.
I’ve been around this business for quite a while (surprising but true), and took care of a lot of men with prostate cancer before PSA was available, or widely used. Back in the 70’s, about 70% of men who were diagnosed with prostate cancer were at an advanced–and incurable–stage of the disease. The only tool we had for screening was the digital exam, which can only detect cancer when it is large enough to be felt–in other words, more advanced. In the 1980’s, when PSA began to see widespread use, about 43,000 men a year died of prostate cancer.
Today, the large majority of men diagnosed with prostate cancer are detected at a much earlier stage. In 2005, about 30,000 men died of prostate cancer–a reduction of over one-third in 20 years. What has changed? Not the therapy, really. Both surgery and radiation techniques have improved–but most of these improvements lie in reduction of treatment complications, rather than better cure rates. Chemotherapy has never been effective as a curative treatment, and is not so today–although it is improving at prolonging survival from widespread cancer. Men are better educated about prostate cancer and the importance of getting regular checkups for it–although most men are happy to skip their annual rectal exams when given half a chance.
What has changed is the widespread use of PSA for early detection.
The bottom line is, there are only three ways to suspect you may have prostate cancer: an abnormal prostate exam, an elevated PSA, or the findings of widespread cancer on other diagnostic tests. The last means the cat’s out of the bag; the first is relatively insensitive until the local tumor is large. But the PSA–often misleading by being abnormal when no cancer is present–is nevertheless how the large majority of men come to find out they have the disease. And this is conclusion is not merely anecdotal medicine–there is good evidence that regular screening reduces the risk of death by prostate cancer.
So what’s a fella to do, given all this conflicting information?
Here’s some recommendations I would make:
- If you’re under 40, you don’t need a PSA.
- If you’re over 75, and don’t already have prostate cancer, you don’t need a PSA
- If you’re under 75, but have other high-risk medical problems (e.g., severe heart disease), you don’t need it.
- If you have a strong immediate family (father, brother) history of prostate cancer (not just cancer in general), especially if they were diagnosed at less than 60 years of age, get a baseline PSA at 40, another at 45, and annual PSA and DRE after that.
- If no family history, begin with DRE and PSA annually at age 50.
- A PSA when you have a prostate infection or urinary tract infection will often be high. Ask your doctor to wait at least 3 months after treatment before checking one.
- If your PSA comes back high, when previous ones were normal, be sure to have it rechecked before further evaluation. If a repeat comes back normal, have a third one checked 3-6 months later. The trend in PSA values is much more important than any single value.
- If your doctor feels a lump on your prostate, but your PSA is normal, don’t be too reassured: get in to see a urologist.
Ignore the swirling controversies about PSA played out in paper-thin detail by the media–they have absolutely no clue what they’re talking about. Be a good boy, and get your prostate checked and a PSA. And tell them Dr. Bob sent you.