Why a Dose Ladder Must Be Designed Before Escalation
A rule-based escalation trial needs a finite, ordered set of dose levels before participants are assigned to dose cohorts. The ladder determines which doses are available and how far apart they are; a separate decision rule determines whether the trial proceeds to another level. Conflating these two tasks can create ambiguity precisely when safety decisions must be clear.
The ladder should describe the regimen consistently: dose amount, route, schedule, and any relevant duration or loading component. A sequence expressed only as “Level 1, Level 2, Level 3” is not sufficient unless each level maps unambiguously to a complete regimen. The dose-limiting toxicity (DLT) definition and observation window must also be fixed for interpreting participant outcomes, as established in the earlier material on defining DLTs and observation windows.
The levels must be ordered by the intended amount of study treatment, with units and regimen components held constant unless the protocol explicitly defines a change. A ladder in milligrams per day cannot be compared directly with one in milligrams per administration if the schedule differs. Nor does a larger administered dose necessarily imply proportionally larger exposure or biological effect. Those relationships depend on pharmacology and participants; the ladder is a design specification, not an exposure model.
Choosing the Spacing
Spacing is a compromise. If levels are too close, the study may need many cohorts to cover the intended range. If they are too far apart, a small number of steps can span the range, but a transition may be too large for the protocol’s safety constraints or may leave clinically useful doses untested. The appropriate spacing depends on the drug, route, available formulation, prior evidence, and protocol requirements. There is no universally correct ratio.
Two common mathematical constructions are equal additive spacing and equal multiplicative spacing. With equal additive spacing, the difference between adjacent doses is constant. With equal multiplicative spacing, the ratio between adjacent doses is constant. Multiplicative spacing is often useful when dose-related changes are considered on a relative rather than absolute scale, but this does not establish that toxicity or exposure changes linearly with dose.
The exponent is \(1/(K-1)\), because \(K\) levels contain \(K-1\) intervals. This is an important counting detail: using \(K\) intervals would fail to land on the planned final dose. A protocol may instead specify a maximum adjacent ratio or a maximum absolute increase. Those constraints must be checked after choosing candidate doses, rather than assumed to follow automatically from the spacing formula.
Worked Example: Equal-Ratio Levels From 5 to 80 mg
Suppose a hypothetical oral regimen is to have eight levels from 5 mg to 80 mg per administration, with the same schedule at each level. Find equally spaced theoretical levels on a multiplicative scale.
There are \(8-1=7\) intervals, so $$ r=\left(\frac{80}{5}\right)^{1/7}=16^{1/7}\approx1.486. $$ Thus each unrounded level is 1.486 times the preceding level, using the exact ratio \(16^{1/7}\) in the calculations.
The theoretical levels are \(5r^0,5r^1,\ldots,5r^7\), which are approximately 5, 7.43, 11.0, 16.4, 24.4, 36.2, 53.8, and 80 mg. For example, the fifth level is $$ D_5=5(16^{1/7})^4\approx24.38\text{ mg}. $$ The final level is \(5(16^{1/7})^7=80\) mg before rounding, as required.
These are mathematical candidates, not necessarily deliverable doses. If the formulation allows only particular tablet strengths or the protocol requires whole-milligram doses, the candidate list must be converted into feasible doses and then rechecked for order, range, and permitted increments. The equal-ratio calculation alone does not show that this ladder is safe or appropriate.
Apply Practical Constraints Before Finalizing
A theoretical ladder becomes a protocol ladder only after the proposed levels have been checked against practical constraints. Relevant constraints can include available dosage forms, preparation accuracy, a minimum administrable dose, a maximum dose supported by prior evidence, and explicit limits on how much a participant’s dose may increase between levels. A protocol may constrain increments as a percentage, an absolute amount, or both.
For a proportional cap \(c\), where \(c\) is the largest permitted ratio between consecutive levels, the condition is \(D_{j+1}/D_j\leq c\). For an absolute cap \(a\), the condition is \(D_{j+1}-D_j\leq a\) in the specified dose units. If both apply, every adjacent pair must satisfy both conditions. A ladder that misses a constraint is not repaired by saying that the levels are “approximately” spaced; it must be revised and checked again.
Worked Example: Rounding Under a Maximum-Increment Rule
Consider a hypothetical regimen with a 10 mg starting dose, a planned upper dose near 75 mg, and a protocol limit that no adjacent levels may differ by more than a factor of 1.5. Suppose doses must be whole milligrams. One feasible candidate ladder is 10, 15, 22, 33, 49, and 73 mg.
Check each adjacent ratio using the rounded doses: $$ \frac{15}{10}=1.500,\quad \frac{22}{15}\approx1.467,\quad \frac{33}{22}=1.500,\quad \frac{49}{33}\approx1.485,\quad \frac{73}{49}\approx1.490. $$ Every ratio is at most 1.5, and each dose is greater than the preceding one. The absolute increments are 5, 7, 11, 16, and 24 mg; these would also need checking if the protocol imposed an absolute cap.
For comparison, rounding an ideal sequence without checking could produce a pair that exceeds the permitted ratio. Therefore the correct procedure is to select practical candidate doses, calculate the actual adjacent ratios (and absolute differences if relevant), and revise any pair that violates a limit. This ladder reaches 73 mg rather than exactly 75 mg; a protocol should state whether this is the intended upper level or whether an additional feasible level is needed.
Rounding is not a cosmetic final step. It can change the spacing, create duplicate doses, or make an increment exceed a prespecified cap. The protocol should list the doses that will actually be administered, not only an equation or unrounded target values. If dose preparation requires a body-surface-area calculation, concentration-based volume, or another conversion, the units and rounding convention for that conversion must also be specified consistently.
Control Movement, Skips, and Inserted Levels
The default ladder is sequential: levels have fixed labels and each has a fixed regimen. A separate rule-based algorithm determines what evidence permits a transition. This tutorial does not define a specific escalation algorithm. At ladder-design stage, the important task is to make the available transitions explicit and ensure that no participant is assigned a dose through an informal relabeling or undocumented exception.
Skipping an untested level is not simply a shorter way to use the ladder. It changes the size of the transition and removes information about the skipped dose. If skipping is permitted at all, the protocol should state who can authorize it, what evidence and safety review are required, whether the skipped level remains available later, and what maximum increment still applies. A prohibition on skipping is also a valid rule and can be easier to apply consistently. These are protocol choices, not universal statistical requirements.
An inserted level can be useful when the original ladder has a gap that is judged too wide or when practical information makes a new dose necessary. The protocol should specify how the new dose is selected, who approves it, how its position and identifier are recorded, and whether assignment can proceed to it before the amendment is effective. Inserting a dose during an ongoing trial changes the set of available options; it must not be handled by silently changing the meaning of existing level labels.
Worked Example: Inserting a Level Without Changing the Existing Ladder
Suppose a hypothetical protocol lists 10, 20, 40, and 80 mg per administration. Before enrollment at 40 mg, the design team identifies the 20-to-40 mg gap as too large for the intended protocol. They propose a new level halfway between those doses on a multiplicative scale.
The geometric midpoint \(D\) satisfies \(D/20=40/D\), so \(D^2=20\times40\), giving $$ D=\sqrt{20\times40}\approx28.28\text{ mg}. $$ If the formulation permits a 28 mg dose, the revised local sequence is 20, 28, and 40 mg. Its adjacent ratios are \(28/20=1.400\) and \(40/28\approx1.429\), both smaller than the original ratio of 2 between 20 and 40 mg.
The protocol must still decide whether 28 mg is feasible and whether the revised ladder satisfies all increment limits. If approved, the amendment should identify the new level, its complete regimen, the effective date, and the assignment rules. The existing 40 mg level remains 40 mg; the new level is added rather than created by renaming the old levels.
A skip exception can be described just as concretely. For example, if a supply interruption makes one listed dose unavailable, a protocol might require enrollment to pause rather than allow an automatic jump to the next available dose. If an exception is allowed, its eligibility conditions and approval path must be written in advance. The team should not infer permission to skip from the fact that another dose is physically available.
Justification, Diagnostics, and Limitations
The equal-ratio formula follows directly from multiplying the same ratio across all intervals. If there are \(K-1\) intervals and each multiplies dose by \(r\), then the final dose is \(D_1r^{K-1}\). Setting this equal to \(D_K\) and solving for \(r\) gives the formula: $$ D_K=D_1r^{K-1} \quad\Longrightarrow\quad r^{K-1}=\frac{D_K}{D_1} \quad\Longrightarrow\quad r=\left(\frac{D_K}{D_1}\right)^{1/(K-1)}. $$ This derivation assumes positive starting and final doses and \(K\geq2\). It is an arithmetic construction, not a theorem about the biological dose-response relationship.
Before the ladder is approved, check it systematically:
- Confirm that every level specifies the same regimen components and consistent units, or clearly documents any intended change.
- Verify that levels are strictly increasing and distinct after rounding.
- Calculate every adjacent ratio and difference; do not check only the first and final levels.
- Confirm that each dose can be prepared and administered using the available formulation and procedures.
- Check that labels, tables, pharmacy instructions, and decision documentation all refer to the same dose values.
- Make the rules for escalation, exceptional skips, and amendments unambiguous and prespecified.
Warning signs include duplicate rounded levels, an adjacent increase above a stated cap, inconsistent dose units, a “maximum dose” that differs across protocol sections, or a transition rule that relies on an unstated exception. Another diagnostic concern is a ladder justified solely by a convenient mathematical pattern. Equal ratios do not guarantee equal increments in exposure, DLT probability, or clinical effect. Those questions require evidence and methods beyond ladder construction.
An arithmetic ladder is simpler to calculate and may fit a setting where equal absolute increments are meaningful. A geometric ladder gives equal proportional increments and can cover a wide range with fewer levels. A custom ladder may better reflect formulation, prior evidence, or operational constraints, but it requires transparent justification and careful checking. Model-based dose-selection methods can use outcome and exposure information differently; a fixed rule-based ladder should not be described as if its spacing alone performed that modeling.
The ladder also has limits. It restricts choices to the specified levels unless a formal amendment adds another. A coarse ladder may leave relevant doses untested, while a dense ladder may increase operational burden without ensuring more informative decisions. No spacing scheme can remove uncertainty about participant response, and no dose sequence should be interpreted as proving that toxicity increases monotonically. The ladder is a controlled structure for a trial, not a substitute for safety review or a complete dose-finding analysis.
Check Your Understanding
Answer each question, then compare your reasoning with the solutions below.
- A ladder has six levels from 4 mg to 64 mg. What is the equal ratio, and how many intervals are there?
- A protocol permits no adjacent dose ratio above 1.4. Does the sequence 10, 14, 20 mg satisfy this condition? Show the checks.
- Why is a dose ladder not itself an escalation algorithm?
- A proposed rounding step makes two theoretical levels equal. What should the design team do, and why is simply retaining both labels inadequate?
- What information should be documented if a new level is inserted during a trial?
Solutions. (1) Six levels have five intervals. The ratio is \(r=(64/4)^{1/5}=16^{1/5}\approx1.741\). (2) The ratios are \(14/10=1.400\) and \(20/14\approx1.429\). The second exceeds 1.4, so the sequence does not satisfy the rule and must be revised. (3) The ladder lists available doses, whereas an escalation algorithm specifies the evidence and conditions under which a transition is allowed. (4) The team should choose distinct feasible doses and recheck all adjacent increments. Duplicate doses are not distinct treatment levels and could make assignments and records ambiguous. (5) Document the new dose and complete regimen, its identifier and position, the rationale and approval, the amendment’s effective date, and the rules governing use of the revised ladder.