Wednesday, June 16, 2010

Useless Medical Screening Tests



Let me begin by defining what I mean by useless: the results of the test are of no use to the patient and cannot be used to improve the patient's health. There are four classes of screening tests, and I will discuss them each in turn, as well as discussing how to interpret tests and their basic limitations. Just bear in mind (except for a few tests such a red cell blood count and oxygen saturation) that a negative test proves absolutely nothing, that any abnormal test should be repeated before acting on it or worrying about it, and that NO DOCTOR EVER GOT SUED FOR ORDERING A TEST, but only for not ordering one.

The first consideration is the ability and skill of the interpreter (and in the case of ultrasounds such as cardiac echoes, the skill of the technician) and the skill of the laboratory tech.. There are two different terms to describe the error due to the observer/reader/interpreter: inter-observer variation and intra-observer variation. Recall that at the bottom of all mammogram reports is a statement that they can miss cancer 10% of the time. The inter-observer variation, for X-rays, is the percentage of times that a another radiologist differs from the interpretation of the first radiologist. The intra-observer variation is the percentage of times the first reader will disagree with him/herself when he/she re-reads the X-ray one year from now. Neither percentage is ever reported with the test result (which is usually 5 to 15 %).

Next, we have the standard error of the test. The "normal" range given for a blood test is usually the United States average (?men ?women? ?ethnic group ?age range) for the test in question plus or minus two standard deviations. Therefore at least 5% of patients are "abnormal" in any given test without having an illness. Racial variation is never mentioned, but I know from experience that females of a certain racial group average 3.5K in their white blood cell count, and not 4.5K. For that matter, all Inca Indians are blood type O, which will affect some of their blood tests. However, the standard error of the test itself is never given (hematocrit from the hospital lab = 45% +/- what %?). No blood test is 100% accurate, and the error bars are vital for proper interpretation of the test. I will not even start to discuss the diurnal variation (cortisol is higher in the AM, serum iron is higher in the PM), or the fact whether a male gives an AM urine sample standing or lying down can affect the % of protein in the urine, and that many joggers will test positive for blood in their stool (and possibly their urine) on the days that they jog. So when you see a numerical lab test posted, you are told the "normal" range, but never the probable per cent error in the measurement itself, which may be larger than the gap between the patient's value and the "normal" one. And what if a patient's test value has an annual variation? We know that gastric and duodenal ulcer bleedings used to peak in the spring and the fall. What else can vary with time? For that matter, has any one seen a table of the variation in the basic lab tests throughout a female's menstrual cycle? And please remember that the absolute systematic lab error and the interpretation variability of the observer, lab tech or radiologist should be added because these errors are independent of each other.

A screening test is ordered for one of four reasons: (a) you are having surgery, and the anesthesiologist or surgeon insists on it, (b) the doctor thinks it will be of benefit in managing your health, (c) the patient asks for it, either for himself or because a spouse or a friend, or an article suggested it, and (d) a van pulled up to his nursing home or city hall and offered the screening test for a "special" price (usually an ultrasound of your carotid and femoral arteries, and your abdominal aorta.).

The fourth reason seems to me to be morally and ethically wrong on the part of the tester. He/she performs the test, and then takes absolutely no responsibility for applying the results to you, but instead essentially abandons you (after taking your money) and tells you to take the results to your doctor. This is a violation in spirit of the Hippocratic Oath.

When the patient or spouse or friend asks for a test, I used to try to explain why that test might not be indicated (e.g. a cardiac stress test in a young patient who has had chest pain from an anxiety attack). But, since even "healthy" young adults can drop dead suddenly, the better part of valor is to order not one, but two stress tests (echo and thallium) and then refer the patient to a cardiologist as well, thus minimizing the chance of a malpractice suit. As far as Chest CT scans for coronary artery calcification is concerned, why not cut to the chase and do a stress-thallium test, to see dynamically how well the coronary arteries nourish the muscles of the heart?

Almost all screening blood tests for cancer are suspect, with the exception of alpha-fetal protein or beta-HCG in patients with chronic hepatitis or cirrhosis who you are screening for hepatocellular cancer. The other tests: CA-125, CEA, CA-19-, CA-15-, etc, are not useful to test for the presence of the disease, nor is a vaginal ultrasound to look for ovarian cancer. The blood tests are, however, useful after the cancer has been found and operated on, to screen for recurrence of cancer.

I have deliberately omitted any mention of PSA, because there is not yet any conclusive evidence that treating prostate cancer saves lives.

The jury on whether CRP has causative value or is just a marker similar to the elevator arrow in the lobby is still out. The latest study was a meta-analysis; for a general criticism of this type of study I refer you to an earlier blog of mine "Analysis of Meta-Analysis".

TSH, or Thyroid Stimulating Hormone is another useless screening test. The body runs on free T4 (actually free T3, but you need to measure that only if you suspect T3 toxicosis). If the TSH is off, then doctors measure the free T4. But I just start with the free T4---why stick the patient twice? I of course repeat the measurement of free T4 along with TSH before starting treatment. Anyway, historically the TSH was easier and cheaper to measure because free T4 is measured by radioimmunoassay (and Rosalind Yarrow got the Nobel Prize for developing the technique of measuring hormones by radioimmunoassay, but because she was a woman and worked at a VA hospital, she couldn't get her early work published in the Journal of Clinical Investigation). But now both T4 and TSH are same day measurements, so if your doctor screens with only the TSH, he is doing you a disservice. Some doctors will measure the free T4 and the TSH at the same time, which makes perfect clinical sense. There is also no evidence that if the TSH is elevated and the free T4 is normal that you need thyroid supplement. In fact, excess thyroid hormone can lead to osteoporosis.

Periodic chest-Xrays and sputum for cytology every 3 months have not been found to be useful to discover lung cancer at an early, curable stage, according to two studies, one at the Mayo Clinic, and the other at Johns Hopkins. The question of the usefulness of spiral chest CT to detect early lung cancer is being studied right now.

Question: how often should a pap smear, a mammogram, and a stool for blood, all of which are useful in detecting cancer at an early and curable stage, be offered to the patient? Once a year, once every 6 months, once every three months, once every month? No one has ever tried to determine the optimum testing interval for any of these tests.

If an anesthesiologist or surgeon requires a test, then you have no choice, unfortunately, because the surgery will not be done without them. An EKG in the previous 6 months is a reasonable request, because the repeat heart attack rate is elevated during surgery for 6 months after a heart attack. However numerous studies (see the Cochrane reports) have shown that pulmonary problems and bleeding problems can generally be detected by a proper history and interview, and a Chest Xray, and the clotting tests PT and PTT are not routinely needed. On the other hand, if there is unexpected bleeding or pneumonia after surgery, someone might get sued if the tests were not done, even if the USPHS guidelines do not recommend them.

Finally we have the question of how to handle a patient who refuses a test. It would seem logical to just make a note in the chart. However, several years ago, a female who refused pap smears for 5 years in a row developed cervical cancer that was ultimately fatal. Despite the documentation of the patient's refusal, with the note in the chart signed by the patient, the plaintiff lawyer for the estate claimed that if the doctor had really explained the risks of not doing the test, and the patient had really understood them, she never would have refused the test. The California jury found for the plaintiff! Consequently, many doctors will "fire" patients who refuse to do certain tests.

I know of no medical use from the results of analyzing the heavy metals in your hair.

FINGERPRINTS: I know this doesn't quite fit the topic, but this should be checked out. It is absolutely true that fingerprints are unique to a given individual, and even identical twins have different fingerprints. However, that does not translate into 100% accuracy in identifying a "latent" fingerprint on an object as coming from a particular person. Recall how a California lawyer's fingerprint was identified as being on some wrapping paper they found at the deliberate explosion in the Madrid subway station but later on they found the real culprits in Europe? Two fingerprints apparently "agree" if a fingerprint "expert" says they agree, much as a psychiatrist testifies that you are sane or insane. However, if you look further into this problem of identification, there is no agreed upon standard as to what constitutes a match. How many "points" have to agree? 6,7,8? And what if some points disagree? And no fingerprint expert will ever admit that there is no "gold standard", but only a judgment call.

Lie detectors are absolutely forbidden to be used to establish guilt or innocence in US federal or state courts. They are too unreliable and non-reproducible, as well as operator/interpreter dependent. There was an article about this in JAMA 3 to 5 years ago. But the CIA and FBI still use them, and DA's often request them.

Urine screening drug tests. The best book about this is probably still Abbie Hoffman's "Steal This Urine Test". The urine tests usually only test for a chemical fragment of the forbidden chemical, and are therefore not very reliable. Also cutoff values are needed for a test to be positive. If the cutoff is set too low, there will be many false positives. For instance, co-nicotine is a metabolite of ordinary vegetables (especially tomatoes) as well as nicotine. So if your urine tests positive for co-nicotine, are you a cigarette smoker or a vegetarian? And remember when Elaine on the "Seinfeld" show flunked her urine test for opium because she had eaten a poppy seed bagel? That has happened in real life as well.

Finally given the fact that 5% of patients statistically are expected to be abnormal on any given blood test, the odds that you will have one abnormal result from a panel of 20 tests when you are healthy is approximately 50%!

And ask your internist if he/she personally reviews ALL Xrays and ultrasounds taken of his/her patients with the radiologist, whether or not the report is normal. I always do, and the Xray report gets modified 10% of the time. Just as in my intern days, the most common overread is cardiomegaly on a chest Xray because the patient did not take a deep enough breath.

1 comment:

  1. The week of October 25 The National Cancer Institute published a study in the Journal of the American Medical Association showing once again that the screening of current or past smokers with a chest X-ray does NOT decrease the death rate from lung cancer.

    ReplyDelete