A local and global sensitivity analysis of a mathematical model of coagulation and platelet deposition under flow

The hemostatic response involves blood coagulation and platelet
aggregation to stop blood loss from an injured blood vessel. The
complexity of these processes make it difficult to intuit the overall
hemostatic response without quantitative methods. Mathematical models aim
to address this challenge but are often accompanied by numerous parameters
choices and thus need to be analyzed for sensitivity to such choices. Here
we use local and global sensitivity analyses to study a model of
coagulation and platelet deposition under flow. To relate with clinical
assays, we measured the sensitivity of three specific thrombin metrics:
lag time, maximum relative rate of generation, and final concentration
after 20 minutes. In addition, we varied parameters of three different
classes: plasma protein levels, kinetic rate constants, and platelet
characteristics. In terms of an overall ranking of the model’s
sensitivities, we found that the local and global methods provided similar
information. Our local analysis, in agreement with previous findings,
shows that varying parameters within 50-150% of baseline values, in a
one-at-a-time (OAT) fashion, always leads to significant thrombin
generation in 20 minutes. Our global analysis gave a different and novel
result highlighting groups of parameters, still varying within the normal
50-150%, that produced little or no thrombin in 20 minutes. Variations in
either plasma levels or platelet characteristics, using either OAT or
simultaneous variations, always led to strong thrombin production and
overall, relatively low output variance. Simultaneous variation in
kinetics rate constants or in a subset of all three parameter classes led
to the highest overall output variance, incorporating instances with
little to no thrombin production. The global analysis revealed multiple
parameter interactions in the lag time and final concentration leading to
relatively high variance; high variance was also observed in the thrombin
generation rate, but parameters attributed to that variance acted
independently and additively.