OBJECTIVE: To test whether BMI cut-off points for obesity, reflect adequately the actual obesity status, in a sample of perimenopausal women. For study's purposes, a new bioelectrical impedance analysis (BIA) equation was estimated. METHODS: 115 Greek, middle-aged women were tested. Body composition was estimated by dual X-ray absorptiometry and BIA method. Waist (WC) and hip circumference (HC) and skinfolds were also measured. RESULTS: The BIA equation predicted fat free mass (FFM) from height, weight, age and resistance (R(2)=0.88, S.E.E.=1.89 kg). The bias was not significant and the limits of agreement +/-3.6 kg. BMI, FFM, body fat percentage (BF%), waist-to-hip ratio (WHR), WC and HC did not differ between pre (N=37) and postmenopausal (N=48) women (at p=0.05). Both BF% and BMI correlated with WHR and WC (r(BF%-WHR)=0.287, p=0.009; r(BMI-WHR)=0.355, p=0.001 and r(BF%-WC)=0.72, p<0.0001; r(BMI-WC)=0.81, p<0.0001). The mean values for BF%, WC and WHR for women with normal BMI were 36.15% (+/-4.19), 72.53 cm (+/-3.64) and 0.749 (+/-0.05), whereas values for overweight women were 41.42% (+/-3.16), 83.06 cm (+/-7.55) and 0.787 (+/-0.05) and for obese these were 47.40% (+/-3.67), 95.10 cm (+/-8.52) and 0.814 (+/-0.05), respectively. CONCLUSION: The body composition analysis of the study sample revealed undesirably high BF%, even in subjects with BMI below 25, in whom BMI did not reflect the body fat content. However, BF was mainly distributed at the lower part of the body. Whether BMI or BF% is a more sensitive index for obesity related diseases, in perimenopausal period, remains to be defined.