26

Yesterday I donated blood. When the nurse performed a finger prick test of iron using my left hand, she indicated that the level was too low. However, she then asked to perform the test again on my right hand. Surprisingly, the iron level was adequate.

She said that this is a well known phenomenon, specifically that the dominant hand often has higher iron levels. (For references, I am right handed.)

How can this be?

I would have thought that, given blood circulates throughout the entire body, it would be uniform everywhere.

Ian Campbell
  • 3,257
  • 2
  • 7
  • 29
Peter Flom
  • 363
  • 2
  • 7

1 Answers1

37

I'm guessing the "iron" test was actually a hemoglobin test (though I'm certainly no expert in blood donation screens...).

I suspect your nurse is reporting a folk tale rather than any true difference. Unfortunately, medical professionals are not always an excellent source of scientific knowledge.

I looked for papers that have actually compared measurements in the two hands. Here's one:

Patel, A. J., Wesley, R., Leitman, S. F., & Bryant, B. J. (2013). Capillary versus venous haemoglobin determination in the assessment of healthy blood donors. Vox sanguinis, 104(4), 317–323. https://doi.org/10.1111/vox.12006

Capillary fingerstick samples were assayed by HemoCue in 150 donors. Fingerstick samples from two sites, one on each hand, were obtained from a subset of 50 subjects. Concurrent venous samples were tested using both HemoCue and Cell-Dyn devices.

They're taking finger pricks and comparing them to venous samples; the finger pricks are a convenient real-world method, the venous samples are considered to be a "gold standard" reference.

Their result that is most relevant to this question is:

Substantial variability in repeated fingerstick HemoCue results was seen (mean hemoglobin 13.72 vs. 13.70 g/dL, absolute mean difference between paired samples 0.76 g/dL). Hand dominance was not a factor.

So, there's a lot of variability in the measurement, but no actual difference between the hands.

Just speculating: the procedure in the clinic is to test the non-dominant hand because people prefer to have the discomfort in the non-dominant hand. If the test comes back adequate, that's the end of it: blood donation starts. If the test comes back inadequate, they repeat the test in the other hand.

A fair number of these retests are occurring in people who have a "true" hemoglobin level above the threshold, but whose initial test was below the threshold simply due to the variability in the test. So, we have a classic "regression to the mean" situation, and the second test often exceeds the threshold.

Because of the selection bias in which people get retested in the dominant hand, the nurse develops a theory (or it gets spread around among other professionals) that the dominant hand produces a higher reading.

Bryan Krause
  • 14,047
  • 4
  • 34
  • 54
  • Thanks! That makes sense. – Peter Flom Oct 04 '23 at 17:08
  • 2
    Strictly speaking, you should not repeat measurements that are near the threshold and keep the best of 2 (or N) values. Because in this case, for values near the threshold, you are artificially selecting values that are inflated due to random fluctuation. In practice, you achieve the same result as if you were lowering the threshold. – pygri Oct 05 '23 at 12:49
  • 3
    @pygri It somewhat depends on the type of error as well as what the original safety benchmark was based on. If your threshold is based on taking the maximum of two tests, then only doing one test is artificially increasing the threshold instead. Or, if the mechanism of the test is such that errors are always underestimates and therefore the result sets a lower bound on the real value, then multiple testing is just helping you to get a more accurate lower bound. – Bryan Krause Oct 05 '23 at 14:51
  • Excellent explanation. An additional (reasonable) assumption here is that too-low hemoglobin is more common than too-high hemoglobin, as retests based on too-high values would regress to the mean in the opposite direction, resulting in the opposite myth that the dominant hand generally returns a lower value. – Nuclear Hoagie Oct 05 '23 at 19:08
  • 6
    @NuclearHoagie My guess, without any experience doing this, I'm not a nurse and don't have any clinical role, is that low readings are common for procedural/uninteresting reasons, so that it's worth doing a second test. Maybe not all of the sample gets in the machine, maybe the sample includes lymph fluid in addition to blood and is slightly diluted, not sure what the exact failure modes are. Whereas, a high reading is just a high reading, and if it's high enough to be concerned for the donor there's no sense doing a second reading. – Bryan Krause Oct 05 '23 at 21:10
  • 3
    I find that a fascinating insight into why the nurse might end up thinking it's the dominant hand that tends to give the higher reading. It's extremely convincing that people would on balance tend to have at least a slight preference for having the *first* test on the non-dominant hand. And the *second* test doesn't even exist unless the first one gave a low reading, so it stands to reason there will be a good few cases where the second test "succeeds", which to a non-statistician could easily give the impression it makes a difference which hand you test (first! :) – FumbleFingers Oct 06 '23 at 18:02
  • 5
    @FumbleFingers It's a good lesson for anyone analyzing data to remember that it's really important to understand how data are generated. A statistician looking at a spreadsheet of data would easily reach the same conclusion as the nurse if they didn't bother to look at the process that generated the data. And it's a good idea to have multiple people looking it over, because it's easy to miss a subtle influence. Also something to keep in mind when reading some claim about data reported on social media by non-experts (or experts, for that matter)... – Bryan Krause Oct 06 '23 at 18:07
  • 2
    @BryanKrause: This answer and comment is turning into a real goldmine for me, thanks! How data are (is?! :) generated can obviously be crucial, but until now I'd have tended to think that implied being careful to avoid "cherry-picked" data. Other "misleading" influences will obviously come into play with other datasets, and they may be more obvious or less obvious. But this one strikes me as sufficiently subtle you'd have to be on the ball to be sure of noticing it! So I don't think our hypothetical nurse here is "unthinking" - she's just not great at *lateral thinking.* – FumbleFingers Oct 06 '23 at 18:34
  • @FumbleFingers Plus, this was probably a phlebotomist, which are not a bachelor's-trained position. – Azor Ahai -him- Oct 26 '23 at 02:53