How accurate are your lab tests?

Gretchen Becker Health Guide
  • Imagine that your last hemoglobin A1c result was 5.9. Now it's time for a new test, and the results show that your A1c is 6.1. Does this mean your control has gotten a lot worse?


    Not necessarily.

    We have to remember that all these lab tests we get are subject to all kinds of errors. First, they're performed by human beings, and all human beings occasionally make mistakes.

    When I was first diagnosed, my hemoglobin A1 (the old test that didn't break down into A1c) was 16. This was equivalent to an A1c of about 13. Obviously not good! I was put on metformin and given the standard dietary advice.

    Add This Infographic to Your Website or Blog With This Code:

    I drastically reduced my food intake, especially starches and sugars, went back for another test in several weeks, and then saw my doctor. He said my A1 hadn't changed. I told him that was impossible. "If it had gone from 16 to 14, I'd believe you, but it's impossible that it hasn't changed when I've changed my diet so much (read more about dietary changes for diabetes here). My fastings have come way down." He gave me the patronizing look that doctors can give and obviously didn't believe me.

    I went home and ate a candy bar. Why bother to diet if your A1 doesn't change. Then the telephone rang. It was the doctor. He said he'd been looking at the wrong A1 test sheet. My new A1 was actually 8 (equivalent to an A1c of about 7).

    So it's always possible that, like my doctor, someone working in the lab read the wrong paper or the wrong line on the paper giving the results. [I once got a letter from my college congratulating me on my recent marriage. I wrote back to say I had a policy never to get married before the first date. It turned out someone typing up the lists had skipped a line.]

    Many tests are actually performed by machines. Machines also make errors or get out of calibration. I used to use a OneTouch Profile meter that measured blood glucose (BG) levels in terms of whole blood, not plasma, and the results were always higher than the lab results, ranging from about the expected 12% up to about 25%.

    One time I was in a research study at the Joslin Diabetes Center that involved testing my BG level every 5 minutes. To do this, they took the blood, spun it down in a centrifuge to remove red blood cells, and then measured the BG level of the resulting plasma in a Beckman spectrophotometer that was supposed to be very accurate.

    The results they were getting were almost exactly the same as what I was getting on my Profile meter, which should have been reading higher. I mentioned it to the researchers, and they said perhaps their Beckman spectrophotometer needed recalibration.

    So it's always possible that the machine used to do some test is out of calibration. Or perhaps the chemical solutions used for the test were made up improperly, or had gotten too old. Or perhaps some test tube was contaminated.

    I once decided to test my meter by taking 5 readings at the doctor's office just before they drew blood for a blood test. I did this several times, and every time, both my Ultra and my Profile gave results that were much higher than what the lab reported.

  • The next time I went for a blood draw, I mentioned this to the phlebotomist who drew the blood, and she laughed and explained that they didn't actually send the blood over to the lab for about 5 hours, and during this time some of the glucose in the blood is destroyed.

    I was aghast and said this meant that if a doctor was testing to see if someone had diabetes, they might get a normal result when in fact their BG levels were high. She said in that case she'd do an immediate test with her own meter. But she wouldn't know to do this if it was simply a routine blood test. I stopped getting blood drawn there and went directly to the hospital lab from then on.

    Add This Infographic to Your Website or Blog With This Code:

    So it's always possible that your blood hasn't been properly stored before it's tested.

    I was once in a clinical study at Joslin that required two admissions to Women & Brigham hospital. At the hospital, they'd do complete blood tests, and between hospital admissions, the blood tests were done at Joslin.


    Here are the results of my HDL tests, each one taken about 2 weeks apart:


    Joslin: 68 (normal range 35 to 95)
    W&B: 50 (normal range 35 to 60)
    Joslin: 66
    W&B: 48

    Now, it's possible that my HDL was really bouncing around like that and it was just a coincidence that it was 66 to 68 when I was at Joslin and 48 to 50 when I was at Women's & Brigham, but I don't think that's likely. I think each lab did the test slightly differently and got different results, even though these are both considered top-notch research facilities.

    If I'd had an HDL measured at 48 at one lab, took a drug and found it was 68 at another lab, it would look as if the drug were effective when in fact it was simply a difference in lab measurements.

    Because labs don't always come up with the same results, when Joslin does research studies that depend a lot on changes in A1c values, they usually send the blood to more than one lab. One time, from blood taken from the same sample, one lab said I had an A1c of 6.4 (normal range 3.5 to 5.7) and another lab said I had an A1c of 6.0 (normal range 4.4 to 5.8). A third lab said I had an A1 (the older test) of 7.3 (normal range 4.5 to 6.9). So all the labs agreed that my A1c/A1 levels were slightly above the normal range, but the exact numbers were slightly different.

    So it's always possible that your doctor or your hospital has changed the lab that did your lab tests, or perhaps the lab has changed its methods, or its personnel, and this has caused a slight change.

    Of course, the incidents I mentioned above are all annecdotal, not controlled trials of the accuracy of various tests. But some test might be shown to be 98% accurate and your test could be among the 2% that weren't very accurate.

    Some tests, including the A1c, can be performed in many different ways, and in some cases there's no standardization of the results to ensure that they all mean the same thing. Today, thanks in part to the efforts of the National Glycohemoglobin Standardization Program, most A1c methods can be standardized so that as long as you don't have an unusual hemoglobin type, they give the same results as other methods.

  • C-peptide methods have not yet been standardized as well as the A1c. A recent paper has discussed the problem of standardizing C-peptide measurements. A C-Peptide Standardization Committee is supporting this type of research.

    Previous research had shown a great deal of variation between labs, especially at higher C-peptide levels. This study showed the same thing, as well as variation between samples tested on different days and even between duplicate samples tested on the same day. The researchers were able to improve the precision of the results by calibrating the results with a precise reference method. But there was still some difference between results measured on identical samples.

    Add This Infographic to Your Website or Blog With This Code:

    So what does all this mean? Should you not pay attention to lab tests at all because they might be erroneous?

    Not at all.

    What it means is that you should view your lab results as fairly accurate indications of your health, but not absolute numbers. If your A1c goes up a little or down a little, it's suggestive, especially if it's the type of change you expected. For instance, if you were on holiday and ate more carbs than usual and your A1c is up slightly, that makes sense. If you decided to take control of your life and kept more strictly to your diet and your A1c is down, that makes sense.

    But if you spent the past month pigging out on pies and cakes and your A1c is down, something is out of whack. If you lost blood by donating to the blood bank, or if you severed an artery in an accident or have some kind of internal bleeding, that could explain it. But it's always possible that the results were because of lab error.

    It might also be because you went to a new doctor who used a different lab. Or even if you went to the same doctor, the lab that practice used might have changed, or the same lab might have started using a new technique. It never hurts to call the lab and ask if your doctor doesn't know.

    Whenever possible, it's best to have all your lab tests done at the same place. That will reduce variation caused by different labs.

    But even if you have your lab tests done at the same place, with the same methods, making some major change in your diabetes treatment on the basis of just one lab test isn't always wise. When you have a urine dipstick test that suggests that you're spilling protein into your urine (a sign of kidney damage), for example, standard procedure is not to treat until you've had several positive tests, as several things can cause false-positives.

    So too, if some other lab test is wildly out of line and your doctor wants to make a major change in your treatment, it might make sense to have the test repeated just in case the results were not accurate.

    What is important in many of these tests is trends. If your A1c bounces around between 5.8 and 6.2, you can assume your true A1c is probably around 6. But if you have levels of 5.6, then 5.8, then 5.9, and then 6.2, it's unlikely that lab errors would show such a definite trend, and your control is most likely worsening.

  • So don't sweat the small changes in lab results. Look for trends. Then try to figure out how you can make the trends go in the direction you want them to go. With the A1c that means going down.

    Add This Infographic to Your Website or Blog With This Code:


Published On: July 28, 2008