Objectives  To determine whether adherence interventions should be administered to all medication takers or targeted to nonadherers.

Data sources and study selection  Systematic search (Medline and Embase, 1966–2009) of randomized controlled trials of interventions to improve adherence to medications for preventing or treating cardiovascular disease or diabetes.

Data extraction  Articles were classified as (1) broad interventions (targeted all medication takers), (2) focused interventions (targeted nonadherers), or (3) dynamic interventions (administered to all medication takers; real-time adherence information targets nonadherers as intervention proceeds). Cohen's d effect sizes were calculated.

Data synthesis  We identified 7,190 articles; 59 met inclusion criteria. Broad interventions were less likely (18%) to show medium or large effects compared with focused (25%) or dynamic (32%) interventions. Of the 33 dynamic interventions, 6 used externally generated adherence data to target nonadherers. Those with externally generated data were less likely to have a medium or large effect (20% vs. 34.8% self-generated data).

Conclusion  Adherence interventions targeting nonadherers are heterogeneous but may have advantages over broad interventions. Dynamic interventions show promise and require further study.