To understand the impact of physical activity (PA) on health, valid accelerometer count cut points must be applied to measure PA. Because cut points may be population specific, we aimed to establish accelerometer cut points for moderate PA (MPA) and vigorous PA (VPA) (defined as ≥3 and ≥6 metabolic equivalents, resp.) in young-to-middle-aged obese-to-severely obese subjects. Data from 42 subjects (11 men; body mass index 39.8±5.7; age 43.2±9.2 years) who performed a treadmill calibration using the Actigraph GT1M, were analyzed using ordinary linear regression (OLR), linear mixed model regression (MIX), and receiver operating characteristics curves (ROC 1; ROC 2). Cut points obtained from the models were quite different (612 to 1646 counts/min for MPA; 3061 to 7220 counts/min for VPA). We argue that the MIX approach, which resulted in cut points of 612 and 4980 counts/min for MPA and VPA, respectively, is the most appropriate method to establish accelerometer cut points in this setting. We conclude that accelerometer cut points are lower in young-to-middle-aged obese-to-severely obese subjects compared to young normal-weight subjects and that care should be taken when analyzing PA level in groups that vary in age and degree of obesity.