|
|
|
|
Occupational and leisure-time physical activity is typically assessed by questionnaires with relatively low validity, prone to random error and bias Rather crude categories for physical activity created for analyses may have led to misclassification
|
Consider various characteristics of physical activity, including nature, frequency, intensity, duration, accumulation, breaks, postures, cardiovascular and biomechanical loading Prevent physical activity misclassification by not using crude categories and by not using arbitrary sample distributions for cut-points (eg, tertiles or quartiles). Existing cohort studies could be re-analysed using such methods
|
|
The possibility of residual confounding, with differences in the way confounding factors were considered and how these factors were measured Some evidence suggested effect modification, for example, by gender or cardiovascular disease
|
Adjust for relevant confounding factors or conduct stratified analyses, using individual participant data, or re-analysing existing cohort data Males and females should be analysed separately to explore gender differences
|
|
|
Controlled experimental studies are the preferred design option, but they may be difficult to conduct. Alternative research designs could therefore be considered.10 For example, negative control studies could be used. Alternatively, a natural experiment of an occupational group making a transition from highly physically demanding work to robotised physically inactive work could be analysed
|