Principles of allostasis: optimal design, predictive regulation,
pathophysiology and rational therapeutics.
This chapter compares two alternative models of physiological regulation. The first
(“stability through constancy”), has dominated physiology and medicine
since Claude Bernard declared, “All the vital mechanisms.
..have only one object – to preserve
constant the conditions of .
.. the internal environment”. His dictum has been interpreted literally
to mean that the purpose of physiological regulation is to clamp each internal parameter at a
“setpoint” by sensing errors and correcting them with negative feedback (Cannon, 1935: Figure
1). Based on this model physicians reason that when a parameter deviates from its setpoint value,
some internal mechanism must be broken. Consequently they design therapies to restore the
“inappropriate” value to “normal”.
The homeostasis model has contributed immeasurably to the theory and practice of
scientific medicine, so to criticize it might almost seem absurd. Yet, all scientific models
eventually encounter new facts that do not fit, and this is now the case for homeostasis. In
physiology, evidence accumulates that parameters are
constant. And their variations, rather
than signifying error, are apparently designed to
error. In medicine, major diseases now
rise in prevalence, such as essential hypertension and type 2 diabetes, whose causes the
homeostasis model cannot explain. For in contrast to the hypertension caused by a constricted
renal artery and the diabetes caused by immune destruction of insulin-secreting cells, these
newer disorders present no obviously defective mechanism. And treating these diseases with
drugs to fix low-level mechanisms that are not broken turns out not to work particularly well.
The chapter will expand upon each of these points.
Figure 1. Alternative models
describes mechanisms that
a controlled variable
by sensing its deviation from a
“setpoint” and feeding back to
correct the error.
describes mechanisms that
the controlled variable
by predicting what level will be
needed and overriding local
feedback to meet anticipated
The second model,
(“stability through change”), takes virtually the opposite
view. It suggests that the goal of regulation is
constancy, but rather, fitness under natural
selection. Fitness constrains regulation to be efficient, which implies preventing errors and
minimizing costs. Both needs are best accomplished by using prior information to predict
demand and then adjusting all parameters to meet it (Figure 1). Thus allostasis considers an
This essay is dedicated to the memory of Howard A. Schneiderman, who recruited me to experimental biology and