Computational Mechanobiology in Cartilage and Bone Tissue Engineering: From Cell Phenotype to Tissue Structure - Computational Modeling in Tissue Engineering - page 342

variability can pose a problem for the corroboration of these theories. Similarly,
model predictions can benefit from covering a range of possible outcomes based
on statistical considerations of the underlying populations and variabilities. Pre-
dicting a range of performance measures for an implanted tissue engineered
construct and its probability of success can aid decisions regarding its design or
even just whether or not to implant a certain construct. The need for nondeter-
ministic simulations and appropriate validation metrics has long been recognised
in traditional engineering, especially in the design of safety relevant systems [ 86 ].
These validation metrics should be based on a statistical evaluation of both the
experiments and the computations. In bioengineering, the variations that occur are
usually larger than those observed in controlled technical systems. Focussing on
scenarios for which mechanoregulatory aspects are of importance, for example for
load bearing implants, variability arises from environmental factors, such as
implant positioning and activity levels, i.e. external loading, as well as genetic
factors such as the level of mechanosensitivity.
The variability of bone geometry and mechanosensitivity observed in a popu-
lation was modelled to simulate the clinical trial of low and high-stiffness intra-
medullary prostheses in 100 patients [ 92 ]. Variable mechanosensitivity was
included by modulating the parameter values that determine the onset of bone
resorption/deposition in the bone remodelling algorithm used. When only one
averaged ideal patient was simulated, i.e. no variability included, the deterministic
simulation predicted that the low stiffness implant would migrate less. The sim-
ulation of the clinical trial with 100 different patients, however, predicted no
statistically significant difference between the two implants. The study concluded,
that stochastic simulations might be more validatable against an actual clinical trial
and may eventually overcome the lack of falsifiability of this type of model [ 92 ].
Modelling certain cellular processes in a stochastic fashion [ 90 ] introduces only
a minor degree of nondeterminism in the simulation outcomes. Khayyeri et al. [ 59 ]
simulated tissue differentiation inside an in vivo bone chamber in which the tissue
could be loaded in compression and investigated the influence of the load mag-
nitude, implant positioning, blood pressure and bone marrow MSC density on the
distribution of phenotypes. The study concluded that these environmental factors
were not sufficient to fully explain the high degree of variability observed in the
experiments, namely the emergence of two distinct experimental groups. In a later
study [ 61 ], the authors thus hypothesised that a variable mechanosensitivity
between individuals could be responsible for the variability observed in the tissue
differentiation experiments. Simulations of the bone chamber experiment
were performed and the effect of stochastically up- or down-regulating the
cellular process rates for proliferation, differentiation and apoptosis in an indi-
vidual-specific manner was investigated. Simulation results were highly sensitive
to MSC activity. The simulations not only predicted the general patterns of tissue
differentiation but also produced a variability akin to that observed in the exper-
iment. Specifically, the simulations predicted the dichotomy, i.e. the emergence of
two distinct differentiation patterns. Results such as this can have important
implications for the corroboration of mechanoregulation theories.