Igor said:I looked on your site j-dumas.home.mindspring.comj-dumas.home.mindspring.comOpen ↗ and must admit that you have done a lot
of work. I'm just wondering, if the "two peak" profile, which you present for novolog is not just
result of incorect fit of the glucose curve?!? I.e. you write, that you use "8th order least
square fit", do you mean that you used a polynom of order 8 (i.e. X^8)?? Just the use of polynom
of such an high order can cause the "two peak effect". Have you tried to fit some lower order
polynom? The result should be very different after the transformation...
BTW, probably with the polynomial fit you should be able to calculate an analytical solution to
the minimal model and by correct estimation (propagation) of errors you shold be able to estimate
the precision of the result...
I just don't belive that an insulin shot can have two peaks...
Hi Igor,
I remember seeing a graph that Dr Anderson of Eli Lilly had on Humulin R pharmacodynamics, (glucose
disposal from an insulin dose, not insulin in the bloodstream from the dose), it had double and
triple peaks. This was umpublished glucose-clamp data that Dr Anderson, (MD now in charge of
worldwide clinical trials for the Humulin insulin product line), would quietly show other MDs at the
American Diabetes Association convention about 10 years ago. You have to remember the liver uses
about 60% of total insulin and it turns on and off as it builds glycogen stores. This modulating
effect causes multiple peaks in glucose uptake in the normal subjects used in these studies (so no
long-term antibody binding delay effects in these subjects). So two or three peaks is not unusual in
the glucose infusion profile for normal subjects. Now add in injection site differences, (the Lilly
data was for abdominal injection sites that are warm/constant temperature so fast, constant
absorption from the subcutaneous tissue happens), where the SC absorption is modulated by
vasodilation/constriction (capillary size changes from temperature) and glucose disposal has more
noise in the data. Finally, add antibody binding effects and multiple peaks are the norm. So if I
see a single peak insulin action profile, I'd say it was theoretical and not real measured data.
This can easily be seen in glucose-clamp dextrose infusion profiles to prove the point.
Next, the order of the polynomial fit is chosen to keep physiological parameters believable. One
parameter used is the glucose uptake by the brain (CNS). Many polynomial fits are "almost perfect"
but show the CNS is sourcing glucose instead of using glucose. The liver and GI tract are the only
organs that can source glucose. So this "near perfect" polynomial fit fails on a physiological basis
and is marked as incorrect in a search algorithm. So you pick the "best" polynomial fit on a
physiological basis.
HTH,
--
Jim Dumas T1 4/86, background retinopathy, rarely hypoglycemic: <1/mo. lispro+R+U+NPH daily,
moderate exercise, typically <6% HbA1c
