Background Self-administered questionnaires such as the International Physical Activity Questionnaire (IPAQ) have been widely used to assess physical activity (PA): in cohort studies to estimate the impact of PA on long term health outcomes; and in randomised controlled trials (RCTs) of PA interventions. Accelerometry is increasingly being used to measure PA but we are not aware of any data assessing whether accelerometry and questionnaires give similar results in PA trials. The PACE-UP trial provides such data.

Methods The PACE-UP cluster RCT compared a pedometer-based walking intervention, by post (n = 339) or with nurse support (n = 346) to usual care (n = 338). Participants wore an accelerometer at baseline and 12 months, giving total weekly minutes of moderate-to-vigorous physical activity (MVPA) in ≥10 minute bouts (accelerometer MVPA). IPAQ was completed for the same seven days, giving total weekly minutes of moderate+vigorous activity in ≥10 minute bouts (IPAQ MVPA) and total weekly minutes of walking in ≥10 minute bouts (IPAQ walking). Baseline accelerometer MVPA was compared with baseline IPAQ MVPA and IPAQ walking. Treatment effects were estimated as change in PA from baseline to 12 months for each outcome measure.

Results At baseline (N = 807), objective PA from accelerometry was lower than self-reported PA from IPAQ: mean weekly minutes 99 (s.d. 104) accelerometer MVPA, 173 (s.d. 278) IPAQ MVPA and 314 (s.d. 308) IPAQ walking. Agreement between accelerometer and IPAQ measures was low: IPAQ MVPA r² = −0.006 (p = 0.87), IPAQ walking r² = 0.18 (p < 0.001). At 12 months (N = 614), both intervention groups had significantly increased their accelerometry MVPA compared with control: postal group 42 (95% CI 23,62) minutes/week and nurse group 43 (23,63). IPAQ walking also showed increases at 12 months: postal group 68 (17,118) and nurse group 54 (4,105). However, IPAQ MVPA showed non-significant decreases compared with control: postal group −11 (−65,44) and nurse group −49 (−103,6).

Conclusion The trial results depended on the outcome measure used. IPAQ MVPA found non-significant decreases in MVPA in the intervention groups, possibly explained by participants learning to assess and differentiate different PA levels. In contrast, IPAQ walking provided the largest estimate of treatment effects, but with considerably less precision than accelerometry estimates. The PACE-UP intervention was designed to increase walking so it is unsurprising that the intervention groups reported more walking at follow-up and also that they overestimated their increase. We conclude that using accelerometry as an objective measure both reduces bias and increases precision compared to questionnaires, when estimating the benefits of PA interventions.