A factorial quitting study tests components: map both assignments

Find both allocation questions

In a two-by-two design, one factor has two levels and the other factor also has two levels, producing four combinations. Read what each level actually contains, including support that every group receives. The design is about how components are allocated, not merely how many names appear in a participant diagram.

Cochrane’s trial-variants chapter explains factorial designs and the importance of interaction. This article uses a fictional nonclinical support example to help a reader map a report. It does not claim that the example has been tested or that combining two techniques will improve quitting.

Cochrane: Chapter 23: Including variants on randomized trials

Draw the four combinations in words

For an invented example, factor A is an optional planning worksheet versus no worksheet, and factor B is a scheduled peer conversation versus no scheduled conversation. The four combinations are neither component, worksheet only, conversation only, and both. Shared background support would still need to be listed separately.

The example shows why a study with only worksheet, conversation, and both groups is not automatically the same two-by-two design. Check the actual allocation method and comparison structure. A promotional headline that calls any multi-arm study factorial may conceal which questions the design can answer.

Look for interaction rather than assuming addition

The effect of one component may depend on whether the other is present. That is an interaction question. Read whether the researchers planned to assess it, how it was analyzed, and what uncertainty remains. A statement that the combined group looked best does not itself establish a reliable interaction or prove that the benefits add together.

A marginal component comparison can involve participants across the other factor’s levels. Do not casually count those as unrelated trials or assume that a separate report describes new participants. The structure of the design and the effect being estimated determine how its results should be interpreted; an introductory guide cannot settle all statistical choices.

Match the headline to the component actually tested

Write the factors, combinations, comparator content, outcome, follow-up, and unresolved interaction issue in a small reading note. If a marketed app contains only one superficially similar component, ask whether that same intervention was tested. Presence of a planning screen is not proof of equivalence to a research worksheet or a full support package.

When reporting the finding, retain whether it concerns a component across conditions or a particular combination. Keep treatment decisions with a qualified professional when clinical interventions are involved. The useful outcome here is a clear design map that prevents a combined result from becoming an unsupported claim about one component or a different product.