C-Peptide Kinetics for Reconstructing Insulin Secretion

Why C-Peptide Helps Reconstruct Secretion

Insulin and C-peptide are released together when proinsulin is cleaved in pancreatic beta cells. Their concentrations do not, however, provide interchangeable measures of secretion. Insulin undergoes substantial first-pass extraction by the liver and is cleared relatively rapidly, whereas C-peptide largely avoids hepatic first-pass extraction and has different, generally slower, clearance. Consequently, a measured insulin concentration reflects both secretion and insulin-specific disposition. C-peptide concentrations, interpreted through a kinetic model, can instead be used to estimate the insulin entering the portal circulation before hepatic extraction: prehepatic insulin secretion.

The inference is a model-based reconstruction, not a direct reading of the measured concentration. C-peptide moves between a central and a peripheral compartment and is eliminated from the central compartment. The amount in the peripheral compartment influences how quickly C-peptide returns to the central compartment. A single concentration, therefore, does not generally determine the instantaneous secretion rate; the kinetic parameters and the recent concentration history matter.

The Two-Compartment Mass Balance

Use first-order transfer between compartments and first-order elimination from the central compartment. The rate constants \(k_e\), \(k_{12}\), and \(k_{21}\) have units min\(^{-1}\): \(k_e\) describes central elimination, \(k_{12}\) transfer from central to peripheral, and \(k_{21}\) return from peripheral to central. With \(A_1=V_1C\), the mass balances are

These equations assume linear, time-invariant kinetics over the observation period; fixed compartment volumes and rate constants; C-peptide assay values that represent central plasma concentration; and no unmodelled source of C-peptide. The kinetic parameters must be appropriate to the individual or population being analyzed. In particular, renal dysfunction or a change in renal function can alter C-peptide clearance and invalidate parameters calibrated under different conditions.

Rearranging the first mass balance gives the instantaneous secretion rate when the central concentration's derivative and the peripheral amount are known:

This expression makes the dynamic correction explicit. A rising concentration does not translate directly into secretion: some C-peptide is being transferred or eliminated, and some may be returning from the peripheral compartment. Estimating secretion from measured data therefore requires reconstructing the unobserved compartment state as well as smoothing or modeling the time course.

Why the Concentration Response Is Biexponential

The two-compartment model implies that a secretion input produces a response with two kinetic time scales. This can be shown by taking Laplace transforms of the mass-balance equations, assuming zero initial amounts for the input-response calculation. Eliminating \(A_2\) yields

where \(\lambda_f\) and \(\lambda_s\) are the positive fast and slow decay rates, respectively. They are the roots determined by

Partial fractions show why the response is biexponential. If \(\lambda_f\ne\lambda_s\), then

Consequently, the impulse-response kernel for a unit amount of secretion is

For an arbitrary secretion history, the concentration is the convolution \(C(t)=\int_0^t h(t-\tau)S(\tau)\,d\tau\), plus the model's response to any nonzero initial compartment amounts. This is the basis of deconvolution: infer the input \(S(t)\) that, after passing through the modeled kinetics, could have generated the measured concentration curve. The derivation also explains why simply multiplying a changing concentration by a clearance is not a valid general reconstruction: that shortcut omits distribution dynamics and the contribution of prior secretion.

From Sampled Concentrations to Secretion

Measured concentrations arrive at discrete times and contain assay and sampling error. Direct numerical differentiation of adjacent observations can amplify that error: a small difference between noisy concentrations is divided by the time interval, producing a highly variable slope. In practice, a model-based reconstruction uses the entire time series and the kinetic equations. For example, with the state vector \(x(t)=(A_1(t),A_2(t))^\mathsf{T}\), define

If secretion is assumed constant at \(S_j\) during a sampling interval of length \(\Delta\), the exact state update is \(x_{j+1}=\Phi x_j+\Gamma S_j\), where \(\Phi=e^{M\Delta}\) and \(\Gamma=\int_0^\Delta e^{M\tau}b\,d\tau\). The observed concentration is \(A_1/V_1\), not the full state. Reconstruction estimates the interval secretion rates and unobserved states so that the predicted concentrations agree with observations, while respecting the model and plausible nonnegative secretion. The assumption of constant secretion within an interval is an approximation; shorter intervals can represent faster changes but may be harder to estimate from noisy data.

Assumptions, Diagnostics, and Alternatives

A secretion reconstruction is credible only to the extent that its kinetic model and input data are credible. Before interpreting an estimate, check that the kinetic parameters and units match the concentration assay, that the sampling schedule resolves the time scale of interest, and that the estimated states and secretion rates are physiologically plausible. Compare observed concentrations with model-predicted concentrations over the full time course. Systematic residual patterns—such as consistent underprediction during a rapid rise or overprediction during a decline—can signal an inadequate kinetic model, misspecified parameters, or an input representation that is too coarse.

Other warning signs include strongly negative reconstructed secretion, large oscillations between adjacent time points, or estimates that change greatly when one observation is removed. Negative values and rapid oscillations are often signs of an ill-conditioned inverse problem or excessive reliance on noisy derivatives, not evidence of biologically negative secretion. Smoothing, a coarser time grid, or a constrained model can improve stability, but each adds assumptions and may obscure genuine rapid changes. If clearance is uncertain, assess how plausible alternative parameter values change the reconstructed secretion rather than treating one parameter set as exact.

The main alternatives answer different questions. A steady-state calculation, \(S=k_eV_1C\), is simple and useful when concentrations and secretion are stable and the kinetic model is appropriate; it does not reconstruct a changing secretion profile. Direct insulin measurement is useful for describing circulating insulin exposure, but because insulin has distinct clearance and hepatic first-pass extraction, it is not by itself a direct estimate of prehepatic secretion. Two-compartment C-peptide deconvolution explicitly accounts for distribution and clearance, but requires kinetic parameters, sufficiently informative sampling, and a stable model. It cannot eliminate uncertainty caused by assay error, parameter mismatch, or unmodeled changes in physiology.