Coupled Cardiovascular-Renal State Variables and Feedback Topology

Why the Topology Matters

Blood pressure and fluid regulation are not controlled by one isolated organ or one variable. A change in circulating volume can affect cardiac output and pressure; altered pressure and renal perfusion can, in turn, change renal handling of sodium and water. Those changes feed back on the amount of fluid available to the circulation. A model that omits one of these links may fit a narrow set of observations yet give misleading predictions when an input, disease state, or treatment changes.

The first task is therefore to define the quantities in the model and specify which quantities influence which others. Feedback topology is the structure of those directed influences, including their signs and, where relevant, their delays. It is distinct from the equations that quantify the size and timing of each response.

Define the Core Quantities Before Drawing Arrows

A useful starting set of quantities is blood volume, exchangeable sodium, mean arterial pressure, cardiac output, and renal blood flow. Each needs an explicit meaning and unit. For example, let \(V_b\) denote blood volume in liters, \(N_e\) exchangeable body sodium in millimoles, \(P_a\) mean arterial pressure in millimeters of mercury, \(Q_c\) cardiac output in liters per minute, and \(Q_r\) renal blood flow in liters per minute.

These quantities are related but not interchangeable. Blood volume is not the same as total body water or extracellular-fluid volume. Exchangeable sodium is a body stock, not a sodium concentration and not a direct measurement of blood volume. Pressure describes a hemodynamic condition; it does not by itself specify cardiac output or renal flow. Cardiac output and renal blood flow are rates of blood delivery, not stored amounts.

A model can make any of these quantities a state if it supplies an appropriate dynamic rule. Alternatively, a quantity may be calculated from other states as an algebraic output. For instance, a model focused on slow fluid adaptation might treat pressure as a rapidly adjusting output, whereas a model focused on short-term cardiovascular transients might give pressure its own dynamic equation. The choice should reflect the question and the observations available, rather than a desire to label every measured variable a state.

For a stored amount, the basic bookkeeping rule is inflow minus outflow. If \(I_{\mathrm{Na}}\) is sodium input and \(E_{\mathrm{Na}}\) is total sodium loss, both in millimoles per day, then a simplified sodium balance is

Similarly, a simplified blood-volume balance can be written as fluid entering the circulating compartment minus fluid leaving it. This equation is a bookkeeping structure, not a complete whole-body water model: exchanges with other fluid compartments and the processes determining renal and nonrenal losses may need to be represented separately.

where \(I_b\) and \(O_b\) are the rates of entry to and exit from the blood-volume compartment, expressed in liters per day. The sodium and volume balances must not be collapsed into one equation. Sodium retention can contribute to changes in extracellular-fluid volume over time, but the relationship also depends on water balance and distribution between compartments.

Map the Feedback Links

A directed diagram begins with plausible influences, not with an assumption that every pair of variables has a direct connection. A basic integrated topology can include the following paths:

• Blood volume can influence cardiac filling and thereby influence cardiac output over the modeled operating range.
• Cardiac output can influence arterial pressure; the quantitative pressure–flow relation requires a separate specification.
• Arterial pressure can influence renal perfusion and renal handling of sodium and water. The magnitude and timing of these effects depend on the model.
• Renal sodium excretion removes sodium from the exchangeable body stock, while sodium intake adds to it.
• Changes in sodium and water balance can influence extracellular-fluid volume and, through distribution and filling, blood volume.
• Pressure-sensitive cardiovascular control can alter cardiac or vascular function, while renal regulatory pathways can alter renal excretion. These pathways should be included at the level of detail the model supports.

A positive link means that, locally and with other influences held fixed, an increase in the upstream quantity tends to increase the downstream quantity. A negative link means it tends to decrease it. These are local statements: a link may change strength or even direction in another physiological range. The sign of an arrow does not supply its gain, delay, or exact functional form.

For example, the sequence \(V_b \rightarrow Q_c \rightarrow P_a \rightarrow E_{\mathrm{Na}} \rightarrow N_e \rightarrow V_b\) can form a feedback loop. If the first three links are locally positive, increased sodium excretion decreases the sodium stock, and a larger sodium stock supports greater extracellular volume, the product of link signs around this loop is negative. This describes a negative feedback direction under the stated assumptions. It does not establish that the loop is stable or fast: stability also depends on gains, delays, nonlinearities, and other loops.

State Choice, Inputs, and Timescale

External inputs are quantities imposed on the modeled system rather than generated by its internal feedback. They may be constant, time-varying, or event-based. Examples include sodium intake, water intake, an intravenous fluid infusion, hemorrhage, a change in posture, or administration of a drug that alters cardiac, vascular, or renal function. An input should be represented at the point where it acts. For example, an infusion may enter a fluid balance, while a drug may modify a response or a pathway parameter; treating both as a generic “volume effect” can hide their distinct mechanisms.

Timescale determines which quantities need explicit dynamics. A model of minute-to-minute responses may need dynamic pressure and cardiac-output states. A model of long-term fluid adaptation may instead represent some cardiovascular responses as rapid mappings and focus its dynamics on slower changes in body stocks. A model may also use a mixed structure: some variables have differential equations, while others are calculated algebraically. Each choice should be declared so that the model's interpretation is clear.

Justifying a Feedback Sign

The sign of a loop can be checked by following a small perturbation around it. Under a local, fixed-operating-point approximation, assign each link a signed gain: positive if an increase in one quantity increases the next, negative if it decreases it. The sign of the product around the loop is the loop sign.

where \(g_i\) is the local gain for link \(i\). An even number of negative links gives a positive loop sign; an odd number gives a negative loop sign. This is a sign-accounting result, not a stability test.

For the volume–output–pressure–sodium loop described above, suppose \(V_b\) to \(Q_c\), \(Q_c\) to \(P_a\), and \(P_a\) to \(E_{\mathrm{Na}}\) are positive; \(E_{\mathrm{Na}}\) to \(N_e\) is negative; and \(N_e\) to \(V_b\) is positive. The loop contains one negative link, so its overall sign is negative. The reasoning depends on the local link assumptions and on treating sodium stock as supporting volume over the relevant slower timescale.

Diagnostics, Alternatives, and Limitations

A topology diagram is useful only if it is internally consistent and sufficiently explicit. Check the following before estimating or simulating a model:

• Check units and stocks. A stored amount has units such as mmol or L; its balance has units per time. A pressure or flow cannot be added directly to a sodium amount.
• Check arrow direction and sign. Distinguish “pressure affects renal excretion” from “renal excretion affects pressure.” Both may occur through different paths, but they are not the same link.
• Check for missing routes. A model with sodium intake but no sodium loss, or fluid input but no fluid outflow, may accumulate a stock indefinitely by construction.
• Check for duplicate effects. If a drug's effect on renal excretion is already represented in a pathway, adding a second direct drug-to-volume effect may count the same mechanism twice.
• Check the assumed range and delay. A local sign may not hold across all conditions, and a delayed response cannot be treated as instantaneous without justification.
• Check identifiability against measurements. If only pressure is observed, multiple combinations of volume, cardiac output, and vascular responses may produce similar pressure patterns.

A fully coupled model is appropriate when the question concerns interactions or perturbations that can propagate through several organs. A one-way model can be useful when feedback is negligible for the question or when the upstream quantity is deliberately treated as fixed. A quasi-steady or algebraic representation is simpler when fast dynamics are not of interest and the relevant mapping is supported by data. These approaches are not interchangeable: simplifying the topology may improve tractability, but it can prevent the model from predicting responses that depend on the omitted feedback.

A signed diagram also has limits. It does not specify response magnitudes, nonlinear saturation, delays, or patient variability. It cannot establish stability from loop sign alone, and a plausible physiological arrow is not proof that the corresponding effect is identifiable from a given data set. These limitations motivate writing explicit equations and testing model behavior after the topology has been chosen.