Paperonium

Child nutrition · PLoS Med · 2020

B12’s effect on infant hemoglobin: the paper’s range is nine times wider than its numbers allow

Printed in the summary andTable 32.7 g/dL wideTable 3’s own numbers give0.3 g/dL wide, 9× narrowerFor scale: each group’s riseduring the study1.0 g/dL−1.33 to 1.37−0.13 to 0.17−2 g/dL−1012no difference
Printed in thesummary andTable 32.7 g/dL wideTable 3’s ownnumbers give0.3 g/dL wide,9× narrowerFor scale: eachgroup’s rise duringthe study1.0 g/dL−1.331.37−0.130.17−2−1012no differenceg/dL

Gap in end-of-study hemoglobin between the groups, with each range of plausible gaps

The paper says

A Nepalese trial gave 600 infants daily vitamin B12 or a placebo. Both groups’ hemoglobin (the blood’s oxygen carrier) ended at 11.6 g/dL. The paper’s summary and Table 3 put the gap at 0.02 g/dL, with a range of plausible gaps from −1.33 to 1.37.

Its own data show

Table 3 also prints each group’s size (283 and 290 infants) and how far infants typically were from their group’s average (0.8 and 1.0 g/dL). The textbook formula its footnote names (the Student t test) turns these into −0.13 to 0.17. The same method reproduces the other 22 ranges in the paper’s two results tables.

A range spans the gaps the data can’t rule out; nearly 300 infants per group make it narrow. The printed one allows a B12 effect over 1 g/dL, as big as each group’s rise during the study. The no-effect conclusion stands, but reviews pooling trials count wide-range ones less, so this one would count 1/82 as much as it should.

“The summary” means the paper’s abstract: the short summary printed at its top, which is the part most readers see.

Read the paper →Full technical write-up →

Effects of vitamin B12 supplementation on neurodevelopment and growth in Nepalese Infants: A randomized controlled trial · PLoS Med (2020) · doi:10.1371/journal.pmed.1003430