This is really interesting, and something I hadn’t thought about before. Doing a quick literature search, there is also previously existing evidence that high levels of dietary iron may impart a diabetes risk. So the effects seen in the paper don’t seem crazy, but I did come out of this with a couple of questions/​comments.
Are the new lower estimates of anemia rates in India just due to changing cutoffs, or are they also because of existing supplementation/​fortification programs working? If programs switched form mass supplementation to screen-and-treat, would they end up still giving a large fraction of the population supplements?
The confidence intervals reported in the preprint seem a tiny bit suspicious to me, given that the lower bounds are all between 1.001 and 1.01. Sometimes that’s just what comes out of the analysis, but it’s also what you’d expect to see if the authors had been p-hacking.
Thanks for the comment and apologies for the delay in responding!
That’s a great point and you’re right that lower estimates are likely influenced by prevailing mass iron supplements. In their paper estimating the new cutoffs, they exclude individuals meeting the criteria for iron-deficiency using the WHO, 2020 cut-offs (12 μg/​L for 1–4 years, <15 μg/​L for 5–19 years), since they need a healthy sample on which to base the new cut-offs on. It seems quite probable that this sample would look different if you excluded those who were taking supplements already. I posted the table from the paper with the percentage difference of anemia by age group b/​w WHO and new cut-offs and the drop is pretty substantial for most groups. So, I still expect a drop in the supplements but someone’s got to run some models to quantify this drop more accurately.
Interesting, could you elaborate more on how exactly it could be p-hacking? I have a decent understanding of p-hacking in neuroscience and it goes as follows—you collect data from hundreds of voxels, apply different connectivity outcome measures and see which voxels slide under the p<0.05 and then claim that was your outcome measure all along and don’t correct for multiple comparisons. In the case, would the claim be that the authors ran multiple samples on the data and chose the one which showed an association above 1.001?
This is really interesting, and something I hadn’t thought about before. Doing a quick literature search, there is also previously existing evidence that high levels of dietary iron may impart a diabetes risk. So the effects seen in the paper don’t seem crazy, but I did come out of this with a couple of questions/​comments.
Are the new lower estimates of anemia rates in India just due to changing cutoffs, or are they also because of existing supplementation/​fortification programs working? If programs switched form mass supplementation to screen-and-treat, would they end up still giving a large fraction of the population supplements?
The confidence intervals reported in the preprint seem a tiny bit suspicious to me, given that the lower bounds are all between 1.001 and 1.01. Sometimes that’s just what comes out of the analysis, but it’s also what you’d expect to see if the authors had been p-hacking.
Thanks for the comment and apologies for the delay in responding!
That’s a great point and you’re right that lower estimates are likely influenced by prevailing mass iron supplements. In their paper estimating the new cutoffs, they exclude individuals meeting the criteria for iron-deficiency using the WHO, 2020 cut-offs (12 μg/​L for 1–4 years, <15 μg/​L for 5–19 years), since they need a healthy sample on which to base the new cut-offs on. It seems quite probable that this sample would look different if you excluded those who were taking supplements already. I posted the table from the paper with the percentage difference of anemia by age group b/​w WHO and new cut-offs and the drop is pretty substantial for most groups. So, I still expect a drop in the supplements but someone’s got to run some models to quantify this drop more accurately.
Interesting, could you elaborate more on how exactly it could be p-hacking? I have a decent understanding of p-hacking in neuroscience and it goes as follows—you collect data from hundreds of voxels, apply different connectivity outcome measures and see which voxels slide under the p<0.05 and then claim that was your outcome measure all along and don’t correct for multiple comparisons. In the case, would the claim be that the authors ran multiple samples on the data and chose the one which showed an association above 1.001?