Why the Average Result Was Not the Whole Story

ACHIEVE gives us two findings that must be held together: the hearing intervention did not reduce three-year global cognitive decline in the overall cohort, while a prespecified sensitivity analysis suggested a reduction among participants at increased risk for cognitive decline but not among those at decreased risk. The question here is how to compare those findings without allowing the more hopeful subgroup signal to replace the primary result—or allowing the average result to erase meaningful variation. The answer supports a careful interpretation, not a personalized promise.

Keep the primary result in view

The overall ACHIEVE cohort is the first comparison to read because it is the study's broadest randomized answer under the conditions that were assigned. Across the 977 enrolled adults, the hearing intervention was compared with the successful-aging health-education control, and the primary endpoint was change in a standardized global-cognition factor score over three years. The official study summary says that the intervention did not reduce global cognitive decline in the overall cohort.

That result is the anchor for two reasons. First, it describes the comparison that includes the full randomized population rather than a selected slice of it. Second, it concerns the prespecified primary endpoint, not an outcome chosen after seeing which pattern looked most interesting. A subgroup or sensitivity analysis can reveal that people respond differently, but it should not silently become the new primary answer simply because it points toward benefit.

The overall result does not mean that every participant had the same experience or that no pathway changed for anyone. An average is a summary across people whose baseline risk, hearing circumstances and response may differ. If an intervention has a larger effect in one part of a population and little effect in another, the combined estimate can be close to zero even though the underlying responses are not identical. That statistical possibility explains why subgroup analysis can be valuable. It does not, by itself, show that the subgroup difference is real, large or useful for individual decisions.

The opposite overreading is also possible. A null average can tempt readers to say that hearing aids do not matter for cognition, or even that the entire hearing-care idea failed. That conclusion goes beyond the result. ACHIEVE tested a supported clinic-based hearing intervention against an active health-education programme, in selected older adults, using three-year global cognitive decline as the primary outcome. The result rules out a broad claim of average benefit on that endpoint under those conditions. It does not rule out benefits that the study did not adjudicate, effects that emerge over another period, or differences in other populations and interventions.

The distinction between “no average reduction” and “no one can benefit” is not a semantic escape hatch. It is the difference between evidence of absence for a specified comparison and an absence of evidence about every other comparison.