Escalation When No Dose Meets the MTD Rule

When Escalation Reaches Its Planned Ceiling

A 3+3 trial can reach the highest dose allowed by its protocol without observing enough dose-limiting toxicities (DLTs) to trigger the rule’s toxicity-based stopping decision. This situation can feel like a conclusion is missing: the trial has escalated as planned, but it has not observed the pattern ordinarily used to identify an MTD. The distinction matters. Reaching a dose ceiling is a study-design fact; meeting the MTD rule is a decision based on observed toxicity counts.

This tutorial uses the standard 3+3 rules and assumes that doses are tested in an ordered sequence, DLT outcomes are classified according to the protocol, and participants counted in a cohort have completed the DLT observation window or otherwise meet the protocol’s evaluability criteria. It also assumes there is no protocol-authorized dose above the stated maximum. If these assumptions do not hold, the trial’s own rules govern the decision.

The rule is based on counts at each dose, not on a claim that the observed proportion is the true toxicity rate. For example, 1 DLT among 6 participants is an observed proportion of \(1/6=16.7\%\); it is not proof that the population DLT probability at that dose is 16.7%. The 3+3 rule is a prespecified decision procedure, not a precise estimate of the dose-toxicity curve.

What the Rules Imply at the Highest Dose

Let the tested doses be \(d_1,\ldots,d_J\), in increasing order, with \(d_J\) the highest planned dose. At each dose, the cohort outcomes determine whether the rule permits escalation, requires further enrollment at that dose, or stops escalation for toxicity. If the trial reaches \(d_J\) and its outcomes remain below the stopping threshold, the protocol’s dose ceiling prevents further escalation even though the toxicity rule itself has not called for stopping.

This can be seen directly from the structure of the conventional MTD decision. Let \(j^*\) denote the first dose index at which the stopping threshold is reached. When \(j^*\) exists and a lower dose was tested, the rule points to \(d_{j^*-1}\) as the MTD candidate. If no tested dose reaches the stopping threshold, there is no \(j^*\), so this rule cannot identify a dose immediately below a toxicity-stopping dose.

The result is therefore usually reported as MTD not reached or MTD not identified within the tested dose range. The highest tested dose may be described as the highest dose evaluated, and its observed DLT count should be reported. It should not automatically be labelled the MTD merely because it was the last dose tested. The trial has not observed the upper toxicity boundary that the conventional rule uses to select a dose below it.

Resolve the Cohort Before Concluding

Being at the maximum planned dose does not remove the within-dose cohort rules. In particular, 1 DLT among the first 3 participants requires expansion to 6 at that same dose. The decision cannot be made from the initial 3 alone. If the additional participants have completed observation and the total remains at 1 DLT among 6, the toxicity rule permits escalation. If there is no higher planned dose, the study reaches its ceiling without a toxicity-based MTD determination.

Similarly, incomplete DLT follow-up can make the apparent count misleading. A participant whose outcome is not yet evaluable cannot simply be treated as having had no DLT unless the protocol explicitly defines that participant as evaluable under its rules. Before interpreting the highest-dose outcome, check the DLT window, evaluability classifications, and whether any outcomes remain pending. These checks do not change the 3+3 thresholds; they determine whether the counts being compared with those thresholds are final and appropriate.

A Practical Decision Sequence

At the upper end of the dose range, separate the toxicity decision from the administrative limit. Apply the cohort rule first; then record whether any further escalation is actually authorized. The following sequence makes that distinction explicit.

Reporting, Alternatives, and Limitations

A clear report gives enough information to distinguish a toxicity decision from a study limit. It should name the maximum planned dose, the number treated there, the number with DLTs, whether the cohort was expanded when required, and why escalation ended. For example: “At the maximum planned dose of 50 mg, 1 of 6 participants had a DLT. The standard 3+3 rule permitted escalation, but no higher dose was authorized; therefore, no toxicity-defined MTD was identified.” This wording reports the count and the decision without converting an observed fraction into a population probability.

There are two different actions a protocol might take after reaching this point. It may end dose escalation at the prespecified ceiling, leaving the MTD not reached. Alternatively, a future protocol or formally approved amendment may authorize testing additional doses, after which the rule can be applied to those doses. The second action is not an automatic consequence of the first: dose increases must follow the governing protocol and safety oversight. Continuing beyond the ceiling without authorization is not a valid way to complete the 3+3 rule.

The main alternative interpretation is to call the highest tested dose the MTD because it was tolerated by the observed cohort. That label is not supported by the conventional rule when no dose has triggered its toxicity boundary. The highest dose may still be of interest for later evaluation, but “highest tested dose” and “MTD identified by the 3+3 rule” are different statements. The rule’s small cohorts also mean that counts below the threshold do not demonstrate negligible risk; they only determine the rule’s next action.

The conclusion is specific to the prespecified 3+3 procedure and tested range. A different dose-finding method can use different criteria and may make a different recommendation from the same data. Such a recommendation would need to be stated as the result of that method, not retrofitted as the conclusion of this 3+3 trial. The conventional rule also cannot resolve questions about doses never tested, long-term toxicity outside the defined DLT window, or whether a selected dose is optimal for efficacy.

Common Mistakes and Checks

• Calling the maximum dose the MTD by default. The dose ceiling is an administrative limit, not evidence that a toxicity threshold was reached. Report the ceiling dose as the highest tested dose unless the protocol’s MTD rule supports a different label.
• Skipping expansion after 1/3. One DLT among the first 3 requires 3 additional participants at that dose under the standard rule. The final count among 6 determines whether escalation is permitted.
• Confusing permission with completion. A count of 0 or 1 DLT among 6 permits escalation; it does not mean escalation occurred. At the maximum planned dose, the rule can permit a step that the protocol does not make available.
• Turning an observed proportion into a population rate. A count of 1 among 6 is an observed proportion of 16.7%, not proof that the true DLT probability is 16.7%.
• Ignoring unresolved follow-up. If a participant’s DLT outcome is not yet final under the protocol, do not present the cohort count as complete without explaining how that participant is handled.

A useful final audit is to ask: Did any dose actually meet the toxicity stopping threshold? Was every required cohort expansion completed? Was the highest dose the protocol maximum? If the answers are “no,” “yes,” and “yes,” respectively, then the trial reached its ceiling without a toxicity-defined MTD under the standard rule.