OBJECTIVE: To determine the correlations between signal-averaged electrocardiography (SAECG) indices and various anthropometric parameters in obese women. SUBJECTS: Twenty-seven healthy obese women (mean age 29.7 AE 9.6 y, mean body mass index 35.2 AE 6.0 kg/m 2 ). MEASUREMENTS: SAECG was performed using a Marquette, MAC 12/15 system. Bidirectional ®ltering, low-pass ®ltering 250 Hz and 25, 40 and 80 Hz high-pass ®lter settings were employed on the same averaged beats. At least 200 beats were averaged to achieve a noise level`1 mV. Body mass index, waist to hip circumference ratio, neck circumference, arm circumference and conicity index were determined for all subjects. Skinfold thicknesses were measured in triceps, biceps, subscapular, supra-iliac and para-umbilical regions. RESULTS: Signi®cant correlations were observed between the root mean square (RMS) voltage of the terminal 40 ms of the ®ltered QRS complex at 25 and 40 Hz and skinfold thicknesses measured at different sites. Signi®cant correlations were obtained between the noise level at 25 and 40 Hz and skinfold thicknesses measured at subscapular, biceps and triceps regions also. The only SAECG parameter correlated signi®cantly with the noise level was the RMS voltage at 25 and 40 Hz high-pass ®lter settings. CONCLUSIONS: In obese women SAECG must be performed with 80 Hz high-pass ®lter setting which eliminates the random noise originating from the subcutaneous adipose tissue. In addition RMS voltage which shows positive linear correlations with the noise level of 25 and 40 Hz high-pass ®lter settings might be inappropriate for the de®nition of abnormal SAECG criteria in obese women.