Progression from Insulin Resistance to Hyperglycemia

From Compensation to Progressive Hyperglycemia

Insulin resistance does not imply that glucose must immediately rise. If beta cells increase insulin secretion enough to compensate, glucose can remain near its regulated range even as insulin sensitivity falls. Hyperglycemia emerges when the combined ability of sensitivity and beta-cell capacity to support glucose control becomes inadequate. A further feedback can then arise: sustained high glucose may contribute to glucose toxicity, which in turn impairs sensitivity and beta-cell capacity.

The model here represents that progression with four state variables: glucose concentration \(G\), insulin sensitivity \(S\), beta-cell capacity \(b\), and a dimensionless glucose-toxicity burden \(T\). It is a reduced-order model, not a detailed account of every tissue flux or secretion pathway. Its purpose is to make the joint feedback explicit and to show how the model’s assumptions determine its predictions.

A Reduced-Order Progression Model

Assume glucose adjusts more rapidly than sensitivity, capacity, and toxicity. Represent its adjustment toward a quasi-steady level \(G_{\mathrm{eq}}\) by a first-order relaxation equation. Let sensitivity and capacity recover toward reference values in the absence of toxicity, while toxicity increases their loss. Let glucose above a specified threshold \(G_{\mathrm{tox}}\) produce toxicity, and let toxicity also decay.

Here \(G_{\mathrm{base}}\) is a glucose baseline in mmol/L, \(A\) is a positive glucose increment in mmol/L, and \(\tau_G,\tau_S,\tau_b,\tau_T\) are positive time constants in days. \(S_0\) and \(b_0\) are positive reference values. The coefficients \(k_{ST}\) and \(k_{bT}\) have units of inverse days per unit of \(T\); \(k_T\) has units of \(T\) per day per mmol/L. The maximum function means toxicity production is zero at or below \(G_{\mathrm{tox}}\), and proportional to the glucose excess above it. The term \(-T/\tau_T\) describes decay at every glucose level.

The inverse dependence of \(G_{\mathrm{eq}}\) on \(Sb\) is a compact phenomenological assumption: less sensitivity or less capacity raises the glucose level required by the model’s balance. It is not a universal physiological law or a substitute for fitting a more detailed flux model. The reference values, parameters, and thresholds must be chosen for the purpose and data at hand. In examples below, \(7.0\) mmol/L is an illustrative hyperglycemia cutoff, not a claim that this model establishes a clinical diagnosis.

What the Equations Imply

If \(S\) and \(b\) are approximately constant, the glucose equation approaches \(G_{\mathrm{eq}}\). Solving the equilibrium relation for the product \(Sb\) gives a direct measure of how much combined compensation is required to keep glucose below a chosen level. For an illustrative cutoff \(G_H>G_{\mathrm{base}}\), the model predicts \(G_{\mathrm{eq}}\geq G_H\) when \(Sb\leq A/(G_H-G_{\mathrm{base}})\).

This result shows why considering sensitivity or beta-cell capacity alone can be misleading. A decline in \(S\) can be offset by a rise in \(b\), but only while the product remains sufficient. The model’s toxicity feedback changes that product over time, so a static threshold calculation is a starting point rather than a full progression analysis.

For a fixed glucose concentration above \(G_{\mathrm{tox}}\), toxicity production is positive, but the net rate of change depends on the current toxicity burden. Setting \(dT/dt=0\) gives the toxicity equilibrium at that fixed glucose. Below this equilibrium, \(T\) increases; above it, \(T\) decreases. At or below \(G_{\mathrm{tox}}\), production is zero and positive \(T\) decays.

For fixed \(G\), the equation is linear in \(T\), so its solution is \(T(t)=T^*+[T(0)-T^*]e^{-t/\tau_T}\). This justifies the direction-of-change interpretation: the difference \(T(t)-T^*\) retains its initial sign while shrinking exponentially.

Feedback Equilibrium and Progression

At a steady state with fixed parameters, the sensitivity and capacity equations can also be solved explicitly as functions of \(T\). Setting each derivative to zero yields \(S=S_0/(1+\tau_S k_{ST}T)\) and \(b=b_0/(1+\tau_b k_{bT}T)\). These expressions make the feedback clear: at a positive toxicity burden, both states are below their reference values. Together with the glucose equation and \(T^*\), they define a coupled equilibrium, not three independent calculations.

The derivation follows by rearranging, for example, \(0=(S_0-S)/\tau_S-k_{ST}TS\): multiplying by \(\tau_S\) gives \(S_0=S(1+\tau_S k_{ST}T)\), and division gives the stated expression. A positive-toxicity equilibrium must also satisfy \(T^*=\tau_T k_T(G^*-G_{\mathrm{tox}})\) and \(G^*>G_{\mathrm{tox}}\). If the resulting candidate violates that inequality, it is not an equilibrium on the positive-production branch; the zero-production branch must instead be checked.

Assumptions, Diagnostics, and Alternatives

The model’s interpretation depends on several assumptions: glucose relaxes toward the stated equilibrium on the timescale \(\tau_G\); \(S\), \(b\), and \(T\) are adequate summaries of slower progression; glucose toxicity is driven by excess above a fixed threshold; and the loss terms are proportional to both \(T\) and the state being impaired. The model also assumes the parameters remain constant during the modeled period. These assumptions are simplifications, not consequences of the equations.

When fitting or applying this structure, check whether it can reproduce both transient and steady-state behavior. Useful warning signs include systematic residual patterns in glucose over time, implausible fitted values of \(S\), \(b\), or \(T\), and parameter estimates that change greatly when the observation window or initial conditions change. Because glucose depends on the product \(Sb\), glucose observations alone may not distinguish low sensitivity from low capacity; independent measurements or informative experimental challenges may be needed to constrain them. A threshold-driven toxicity term can also make predictions sensitive to small changes near \(G_{\mathrm{tox}}\). Examine that sensitivity rather than treating the threshold as precisely known.

A static calculation using \(Sb\) can be useful for locating a compensation boundary, but it cannot describe delayed feedback or recovery. A more mechanistic flux model can represent secretion, insulin action, and glucose production separately, but requires additional states, parameters, and data to identify them. This reduced model is useful for organizing a progression hypothesis and testing its qualitative implications; it cannot, by itself, identify a biological cause, establish an individual diagnosis, or replace validation against relevant data.