Why Dose Recommendations Need an Error Check
A dose calculation can be arithmetically correct and still be wrong for the patient. A common reason is that a quantity is used on a different scale from the one assumed by the model: indexed kidney function is treated as an absolute clearance, or a model’s maturation adjustment is applied twice. Other errors arise when body size is scaled without a specified limit or when a model is used outside the population that supports it.
The purpose of an error check is not to substitute an improvised correction for a validated regimen. It is to establish what the recommendation actually assumes, whether the patient and measurements meet those assumptions, and whether the resulting dose follows the stated model or protocol. The checks below focus on four recurrent failure modes in special-population dosing.
Check the Scale of Kidney Function
Estimated GFR is often reported indexed to a standard body surface area (BSA) of \(1.73\ \mathrm{m^2}\), with units of \(\mathrm{mL/min/1.73\,m^2}\). Some dose models instead require absolute GFR in \(\mathrm{mL/min}\). These values are not interchangeable, particularly for a child whose BSA differs substantially from \(1.73\ \mathrm{m^2}\).
The conversion follows from the definition of an indexed value. If absolute GFR is multiplied by \(1.73/BSA\) to produce the indexed estimate, rearranging gives the conversion back to an absolute value:
This conversion is appropriate only when the reported estimate is in fact indexed to \(1.73\ \mathrm{m^2}\), and when the dosing method calls for absolute GFR. It does not make different estimating equations interchangeable. The clinician or analyst must also confirm that the equation, units, and kidney-function measure match the model or protocol.
Worked Example: Converting an Indexed Estimate
Hypothetical data. A child has an indexed eGFR of \(60\ \mathrm{mL/min/1.73\,m^2}\) and a BSA of \(0.80\ \mathrm{m^2}\). Suppose, solely for this example, that the dose rule defines its renal categories using absolute GFR in \(\mathrm{mL/min}\). Convert the value before applying that rule.
The required conversion is:
The units of \(\mathrm{m^2}\) cancel against the indexing factor, leaving \(\mathrm{mL/min}\). If the hypothetical rule has a category boundary at \(30\ \mathrm{mL/min}\), the converted value is below that boundary. Using the reported number \(60\) as though it were absolute GFR would place the child on the wrong side of the boundary. This calculation assumes the eGFR is indexed as stated, the BSA is appropriate for the same patient and time point, and the rule truly requires absolute GFR.
Do Not Apply Maturation Twice
Pediatric clearance models may account for body size, maturation, or both. A maturation function represents age-related change in a process such as drug clearance; it is distinct from a body-size scaling function. The key audit question is whether the proposed formula already contains each adjustment. Adding a second adjustment merely because the patient is young can lower a predicted clearance—and raise a derived dose adjustment—without support from the model.
For example, consider a model defined as adult reference clearance multiplied by an allometric body-size term and a maturation function. In that model, the maturation function is already part of the prediction. A separate “immaturity correction” is not justified unless the model or dosing rule explicitly defines an additional, nonduplicative adjustment. The calculation also assumes that the model’s age and weight definitions, reference values, and applicable population are used as specified.
Worked Example: Finding a Duplicated Maturation Adjustment
Hypothetical model. For a child weighing \(10\ \mathrm{kg}\), suppose a model defines clearance as \(CL=CL_{\mathrm{adult}}(WT/70)^{0.75}M(age)\), where \(CL_{\mathrm{adult}}=5.00\ \mathrm{L/h}\) and the model’s maturation function at the child’s age is \(M(age)=0.60\). Assume the model is validated for this child’s age and weight, and use the specified body-weight units and reference weight.
First calculate the body-size term:
Then apply the two factors specified by the model:
The model prediction is approximately \(0.697\ \mathrm{L/h}\). Multiplying by \(0.60\) again would give \(0.4183\ \mathrm{L/h}\), but that result is not the prediction of the stated model: it counts the same maturation function twice. This example assumes the model has exactly the displayed structure; a different model may define its covariates differently, so its equation must be checked rather than inferred from a label such as “pediatric clearance.”
Check Weight Scaling and Dose Limits
Weight-based dosing can be linear, allometric, capped, or defined by another regimen-specific rule. These approaches are not interchangeable. Blindly extending a linear milligram-per-kilogram rule to a very large body weight can produce a dose above a specified maximum or beyond the weights used to support the rule. Conversely, applying an arbitrary cap where none is specified is also an error.
A cap is part of a regimen only when it is stated by the relevant label, protocol, or validated dosing method. It should not be invented from the calculation. Likewise, a model-based alternative should not be substituted just because it produces a more plausible-looking number: that alternative needs its own evidence and defined use conditions.
Worked Example: Applying a Stated Maximum Dose
Hypothetical regimen. A fictional regimen specifies \(5.00\ \mathrm{mg/kg}\) for an adolescent and a maximum dose of \(400\ \mathrm{mg}\). An adolescent weighs \(90.0\ \mathrm{kg}\). Assume the regimen applies to this patient and that no other adjustment is required for this illustration.
The uncapped weight calculation is:
Because the stated maximum is \(400\ \mathrm{mg}\), the regimen’s rule gives:
Reporting \(450\ \mathrm{mg}\) would ignore an explicit condition of this hypothetical regimen. The conclusion depends on the cap being part of the rule; it is not a general claim that all weight-based doses should be capped at \(400\ \mathrm{mg}\). If the patient’s characteristics fall outside the regimen’s evidence base, the cap alone does not establish that the regimen is suitable.
Recognize Extrapolation Beyond the Evidence
A prediction is interpolation when the patient’s relevant characteristics lie within the data range and population supporting the model; it is extrapolation when the calculation extends beyond that support. A formula will usually return a number outside its development range, but the existence of a number does not establish that the prediction is reliable there. Extrapolation can be especially consequential at extremes of age, body size, or kidney function, where the relationship may change or the data may be sparse.
The audit should compare the patient’s age, weight, renal function, and other model covariates with the population and ranges used to develop and evaluate the method. It should also distinguish a documented dosing rule for a subgroup from a mathematical extension of a model fitted elsewhere. If the patient is outside the supported range, the correct response is to identify the limitation and seek an applicable evidence-based method or clinical review—not to present the unvalidated output as established.
Worked Example: A Renal Prediction Outside the Studied Range
Hypothetical model. A dose-fraction model was evaluated only for absolute GFR from \(20\) to \(80\ \mathrm{mL/min}\), and within that range it defines the fraction as \(GFR/60\), with a maximum fraction of \(1\). A patient’s absolute GFR is \(12\ \mathrm{mL/min}\). For arithmetic comparison, directly extending the formula would give:
The units cancel, so \(0.200\) is dimensionless. However, \(12\ \mathrm{mL/min}\) is below the model’s evaluated range of \(20\) to \(80\ \mathrm{mL/min}\). The computed fraction is therefore an extrapolated value, not a validated recommendation. The calculation is useful for detecting the boundary problem, but it does not justify using \(0.200\) as a dose fraction. A method supported for this degree of renal impairment is needed.
A Practical Audit and Its Failure Modes
A consistent audit makes errors easier to locate before a dose is acted on. It starts with the definition of every input, not with the final arithmetic. In particular, check whether weight is total, ideal, or another defined measure; whether GFR is indexed or absolute; and whether age-dependent maturation is already incorporated. Then check the model’s range and any explicit dose limit.
Record the exact formula or protocol, including units, covariates, reference values, and any specified maximum or category boundary.
Check that weight, age, and kidney-function values are measured or estimated as required by the rule and refer to the same patient and relevant time point.
Mark where body size, maturation, and renal function enter. If a factor appears twice, determine whether the model explicitly requires both applications.
Compare the patient with the model’s supported population and range. Verify that the units in each operation cancel to the expected units for clearance or dose.
Distinguish a calculation supported by the rule from an extrapolated value or an unresolved input mismatch; do not imply that arithmetic resolves those problems.
This audit is not a replacement for model diagnostics or clinical review. Subgroup performance can differ from overall performance, and a formula may be poorly calibrated even within its nominal range. Warning signs include abrupt dose jumps at a category boundary, implausibly large changes from small input changes, inconsistent units, or predictions that change materially when an already-included factor is removed. Such signs prompt investigation; they do not by themselves identify the correct replacement dose.
| Potential error | What to check | Why the check matters |
|---|---|---|
| Indexed GFR used as absolute GFR | Reported units, BSA, and the scale required by the rule | The patient may be assigned to the wrong renal category |
| Maturation applied twice | Whether the clearance equation already includes a maturation function | Clearance may be underpredicted and derived dosing altered without support |
| Weight scaling extended without a stated limit | Whether the regimen specifies a maximum and whether the patient is in range | The arithmetic may exceed the rule or its evidence base |
| Use outside model range | Age, weight, GFR, and population used for development and evaluation | A numerical output may be an unsupported extrapolation |
Why These Checks Matter
The main alternatives are not competing universal formulas; they are methods whose suitability depends on the dosing question and evidence. A specified fixed maximum may govern a weight-based regimen, while another regimen may use a body-size model without that cap. Renal adjustment may require an indexed estimate, an absolute estimate, or a particular estimating equation. A maturation-aware model may differ from one using a simpler age or weight rule. The right comparison is therefore between the patient’s situation and the method’s stated assumptions—not between whichever outputs look most reasonable.
A common mistake is to say, “The formula produced a dose, so the dose is supported.” The correction is to separate calculation from justification: a valid calculation requires correct inputs and arithmetic, while a valid recommendation also requires a method appropriate to the patient and within its supported use. When those conditions are unclear or unmet, report the uncertainty or limitation rather than silently changing the formula.
Check Your Understanding
Answer each question, then compare with the solutions below.
- An indexed eGFR is \(45\ \mathrm{mL/min/1.73\,m^2}\) and BSA is \(1.20\ \mathrm{m^2}\). What is the corresponding absolute GFR, and what must be checked before using it?
- A clearance model includes a maturation function. A separate calculation applies the same maturation factor again because the patient is young. What is the error?
- A hypothetical regimen specifies \(6.00\ \mathrm{mg/kg}\) with a maximum of \(300\ \mathrm{mg}\). What dose follows for a \(60.0\ \mathrm{kg}\) patient, assuming the regimen applies?
- A model was evaluated for absolute GFR from \(25\) to \(90\ \mathrm{mL/min}\). Is its formula automatically validated for a patient at \(18\ \mathrm{mL/min}\)? Explain.
Solutions. (1) \(45\times 1.20/1.73=31.21\ \mathrm{mL/min}\). Before use, verify that the estimate is indexed as stated and that the dosing rule requires absolute GFR and accepts the estimating method. (2) The maturation effect is counted twice; use the model as specified unless it explicitly defines another independent adjustment. (3) The uncapped calculation is \(6.00\ \mathrm{mg/kg}\times60.0\ \mathrm{kg}=360\ \mathrm{mg}\); applying the stated maximum gives \(\min(360,300)=300\ \mathrm{mg}\). (4) No. \(18\ \mathrm{mL/min}\) is below the evaluated range, so a direct formula output is extrapolation and requires an applicable evidence-based method rather than being treated as validated.