Does allometric scaling remove the dependency of flow-mediated dilation on baseline artery diameter without the need for shear rate normalization?
Allometric scaling properly isolates the flow-mediated response from baseline artery diameter, rendering shear rate normalization unnecessary.
A ratio index (FMD%) is used ubiquitously to scale (by simple division) brachial artery flow-mediated dilation (Ddiff) in direct proportion to baseline diameter (Dbase). It is now known that Ddiff is inversely proportional to Dbase rendering FMD% wholly inappropriate. Consequently, FMD% is still substantially dependent on Dbase. Although this problem is grounded in statistics, normalization of FMD% for the change in arterial shear rate (ΔSR) has been proposed to remove this Dbase-dependency. It was hypothesized that, if the flow-mediated response is scaled properly to Dbase in the first place, shear rate normalization would not be needed to remove Dbase-dependency. Dedicated software (Digitizelt) was employed to extract the data from a seminal study on FMD% normalization. The underlying allometric relationship between Dbase and peak diameter (Dpeak) was described. The re-analyses revealed that the absolute change in arterial diameter was strongly inversely proportional to Dbase (r= - 0.7, P < 0.0005). The allometric exponent for the Dbase-Dpeak relationship was 0.82 (95% CI: 0.78-0.86) rather than the value of 1 needed for appropriate use of FMD%. The allometric approach completely eliminated the originally reported dependency on Dbase without any need for ΔSR normalization (r=0.0, P=0.96). The correlation between ΔSR and FMD% reduced from 0.69 to 0.37, when adjusted for Dbase. In conclusion, this new re-analysis of data from an influential study demonstrates that the FMD%-Dbase correlation is caused by the inappropriate size-scaling properties of FMD% itself. Removal of Dbase-dependency via FMD%/ΔSR normalization is not essential at all if allometric scaling is applied to isolate the flow-mediated response in the first place. Consequently, the influence of ΔSR on this properly scaled response can also be isolated and quantified accurately without the confounding influence of Dbase.
Greg Atkinson (Wed,) studied this question.