Phase two, and the conditions for starting it
Understanding who is vulnerable requires observing people before loss of control. That is a second study, and it starts only if the first passes its gates.
Why there is a phase two
The first phase studies people who have already lost control. It can say what need looks like, how stable its measurement is, and whether it anticipates relapse. It cannot say who was vulnerable beforehand, because comparing people afterwards is not observing them before.
That is the question phase two addresses. It is not stated as an opening promise: it is stated as an extension, with the activation criterion written in advance.
The activation criterion
Phase two begins if, and only if, both of the following hold in the clinical population:
- The parameters pass the within-person repeatability gate.
- The parameters add information about outcome beyond what is already available.
If the first gate fails, there is nothing to carry forward: an unstable instrument stays unstable in a different cohort. If the first passes and the second does not, the result is a reliable measure with no prognostic value — publishable, but not extendable.
What changes in the design
| Phase one | Phase two | |
|---|---|---|
| Who | People in treatment | Exposed people, not yet addicted |
| When measured | After loss of control | Before it happens, if it happens |
| Question | Does the dynamics anticipate relapse? | Does the dynamics precede loss of control? |
| Outcome | Relapse at 3 and 6 months | Onset of problematic use over time |
| Comparison | Within person | Between those who will and will not develop |
The instrument stays the same. The population and the time horizon change. This is deliberate: if the protocol changed, one could not claim the two phases measure the same thing.
What it costs
A prospective study on exposed people runs for years, requires a cohort enrolled before anything happens, and a far larger number of participants, since only a minority will develop addiction. It is an order of magnitude beyond the first phase, which is why it cannot be the project's front door.
A middle path, inside phase one
One thing can be done straight away, at contained cost: include in the cross-sectional study a group of exposed people who have not lost control.
It does not settle the vulnerability question — these remain people who have not lost control, and we do not know whether they will — but it supplies the reference range needed to say what is abnormal, and allows a check on whether the parameters distribute differently. It is a lead, not proof, and should be presented as such.